Behavioral Ecology Vol. 15 No. 1: 71-82
© 2004 International Society for Behavioral Ecology
Sex ratio variation in the bumblebee Bombus terrestris
a Department of Behavioural Biology, Utrecht University, P.O. Box 80086, 3508 TB Utrecht, The Netherlands b Zoological Institute, Department of Population Ecology, University of Copenhagen, Universitetsparken 15, 2100 Copenhagen, Denmark
Address correspondence to J. J. Boomsma at the University of Copenhagen. E-mail: JJBoomsma{at}zi.ku.dk.
Received 15 January 2002; revised 20 December 2002; accepted 6 February 2003.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Patterns of sex allocation in bumblebees have been enigmatic and difficult to interpret in either a Fisherian context or in a kin-selection perspective. We gathered data on several hundred laboratory-reared colonies of the bumblebee Bombus terrestris and analyzed sex allocation as a function of diapause duration and a series of variables describing colony development. Our analyses addressed both sex allocation patterns across different cohorts of laboratory colonies reared at different times and sex allocation patterns across individual colonies within these cohorts. We used path analysis to test a hypothetical model linking a sequence of colony-development variables to the crucial reproductive parameters at the end of the colony life cycle. We show that (1) population-wide patterns of sex allocation show equal investment in the sexes and are thus consistent with queen control, but not with worker control. (2) A significant part of the colony-level and cohort-specific variation in sex allocation is related to the hibernation conditions of founding queens: Queens with longer than average winter diapause produce larger cohorts of first and second brood workers, switch to haploid eggs early, and produce colonies that raise mostly males and few new queens and vice versa. (3) Colony-level sex allocation is significantly related to the time span between the switch point (date of first haploid egg laid by the queen) and the competition point (date of first haploid egg laid by one of the workers): the longer this period, the more male biased the sex ratio. (4) The breeding constraints of an annual life cycle, the short reproductive season, and the presumably high premium on early produced males imply that bumblebee workers have no realistic options to capitalize on their relatedness asymmetry toward the different kinds of reproductive brood by biasing the sex ratio.
Key words: Bombus terrestris, bumblebees, colony development, queen control, reproductive strategies, sex allocation.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Since the seminal contributions by Hamilton (1964
, 1972) and Trivers and Hare (1976), the study of sex ratio conflict between queens and workers of eusocial Hymenoptera (ants, social bees, social wasps) has been a prime test case of adaptive evolution through kin selection. After being highly controversial initially (see, e.g., Alexander and Sherman, 1977), Trivers and Hare's hypothesis of worker control over sex allocation has been repeatedly and convincingly confirmed in recent years, most particularly for ants (Evans, 1995; Sundström, 1994; Sundström et al. 1996), but also for primitively eusocial bees (Mueller, 1991) and wasps (Queller et al., 1993). The central prediction tested by Trivers and Hare (1976) was that the population-wide sex allocation ratio should be female biased, reflecting the asymmetric average relatedness of workers to sisters and brothers. However, much of the recent support for Trivers and Hare's hypothesis has been obtained by testing split sex ratio predictions based on relative relatedness asymmetry in individual colonies (Boomsma and Grafen, 1990, 1991). This approach allows statistically more powerful tests of the worker control hypothesis than the analysis of population-wide sex allocation ratios (e.g., Trivers and Hare, 1976; Nonacs, 1986) and can potentially provide an explanation for the often bimodal distribution of colony-level sex allocation ratios.
Manipulation of the queen-produced primary sex ratio by workers to achieve a significant female bias can be efficient (Aron et al. 1994, 1995; Sundström et al., 1996) and sophisticated (Chapuisat et al., 1997) in ants, but it is far from clear whether such responses can be equally rewarding in more primitively eusocial bees and wasps with annual colonies. In an early study, Bulmer (1981) pointed out that queens in annual societies are in a much better position to win the sex ratio conflict with their daughter workers than ant queens heading perennial societies. In addition, small annual societies are likely to have more intense and direct conflict over sex allocation than are perennial societies (Bourke, 1999), and workers in such societies may also be more constrained in their manipulative strategies to oppose the reproductive interests of their mother queen (Bulmer, 1981). This implies that annual eusocial species are highly attractive model systems to study the potential significance of social constraints on adaptive sex-ratio responses. The most suitable of these are the social wasps and bumblebees that have single-queen colonies (i.e., potentially high worker–queen conflict over sex allocation), moderately sized colonies, and progressive provisioning of offspring, similar to ants.
In spite of this suitability, the available data sets on sex allocation and worker–queen conflict in wasps and bumblebees are few (for recent reviews, see Bourke and Franks, 1995; Crozier and Pamilo, 1996). Bourke (1997) summarized the comparative data on sex allocation in bumblebees and showed that observed patterns within and across nonparasitic species are generally too male biased to fit conceptual explanations based on kinship theory and worker control. Following earlier authors (Duchateau and Velthuis, 1988; Müller et al., 1992; Plowright and Plowright, 1990), Bourke hypothesized that queen control, protandry (i.e., the relative timing of male and queen production) and perhaps local resource competition are decisive factors that have shaped sex allocation strategies in bumblebee colonies. The protandry hypothesis was originally developed by Bulmer (1983) and was shown to be a likely outcome of selection in annual social insects. In a more recent study on Bombus terrestris, Beekman and Van Stratum (1998) confirmed this male bias and showed that colonies (n = 41) generally specialize in either male production or queen production, as expected from an earlier study by Duchateau and Velthuis (1988). The review by Bourke (1997) demonstrates that our current lack of detailed understanding of sex allocation in bumblebees is at least partly due to the very limited comparative data available, as the total data set that Bourke could analyze comprised only 154–174 colonies from 11 nonparasitic species. These low (14–16 colonies per species on average) sample size per species and the fact that the separate case studies included both field-collected colonies and cohorts of laboratory colonies raised under varying conditions, imply that estimation errors are substantial compared to, for example, the currently available comparative sex ratio data in ants (for a review, see Crozier and Pamilo, 1996).
We analyzed sex ratio data from 152–197 colonies of a single population of the European bumblebee B. terrestris, reared in 11–13 cohorts in the same laboratory. We addressed questions regarding population-wide sex allocation equilibria and facultative colony-level sex ratio biasing and related central aspects of sex allocation to key parameters of hibernation-diapause and colony development, partly including data from an additional 18 cohorts comprising another 254 colonies. Our objectives were to (1) estimate what part of the cohort-specific variation in sex allocation has a direct proximate link to hibernation conditions; (2) test the hypothesis of Owen and Plowright (1982), Duchateau and Velthuis (1988), Müller et al. (1992), Duchateau (1996), and Bourke (1997) that queens of B. terrestris control sex allocation; (3) analyze the relationship between sex allocation and a series of colony-development variables to elucidate the mechanisms of effective queen control that have recently been proposed by Bourke and Ratnieks (2001; see also Kukuk, 1992); (4) examine the constraints that prevent bumblebee workers from significantly influencing sex allocation and male production in light of the Bourke and Ratnieks (2001) hypothesis that workers delay laying eggs until they have information showing that doing so is in their kin-selected interest.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Origin of queens and rearing procedures
Basic information on the cohorts of laboratory colonies used in our study is given in Table 1. Cohorts of colonies were reared in constant darkness (but with red light of a wavelength invisible for bumblebees) at 28°C and 60% relative humidity, in total confinement and with ad libitum pollen and sugar water. Details on nest-boxes, daily feeding regime, and maintenance routines are given in Duchateau and Velthuis (1988)
. On four occasions (1989, 1993, 1994, 1995), we collected a random sample of queens in spring in the Gimborn Arboretum, 15 km from the Uithof campus of Utrecht University. Offspring of the colonies reared from these wild queens were mated in flight cages. After overwintering in boxes with peat dust for 2–8 months at a constant temperature of 5°C, we used these queens to start new cohorts of laboratory colonies.
|
As there was considerable variation among cohorts and individual colonies in reproductive output and sex ratio (see Results), it could not be avoided that the genetic diversity across colonies in the laboratory-reared cohorts was lower than in the cohorts started from queens collected in the field. We have approximated the genetic diversity per cohort by deriving the number of genetically independent haplotypes (i.e., alleles at a hypothetical marker locus with an infinite number of alleles) from pedigree information available for the total sequence of laboratory cohorts, assuming exclusively single queen-mating (Schmid-Hempel and Schmid-Hempel, 2000
; Table 1). Inbreeding could not always be avoided in the laboratory cohorts, and we have approximated this by quantifying the proportion of colonies that produced diploid males (Table 1). Diploid males are morphologically indistinguishable from normal haploid males but can be unambiguously assessed because they are produced in the first and second brood, which normally only contain female (worker) offspring (Duchateau et al., 1994). Colonies producing diploid males always occurred in cohorts of relatively low genetic diversity (Table 1) and were excluded from analyses, but colonies from the same cohort that produced no diploid males were kept in the analysis.
Key variables of colony development and their interactions
The short life cycle of bumblebees makes it possible to study and quantify the key variables of colony development in ways that would be impossible for perennial social insects. In our study, we checked colonies every day and quantified the following variables: (1) number of days required for the founding queen to produce her first egg after having been set up in a nest-box (egg1; range 2–39 days); (2) number of days from the first egg until the first worker emerged (worker1; range 17–32); (3) number of workers in the first brood (n-workers1; range 2–18); (4) number of workers in the second brood (n-workers2; range 6–66); (5) number of days between the emergence of the first worker and the production of the first haploid egg by the queen (switch point; range -2–33); (6) number of days between the emergence of the first worker and the first reproductive activity by workers (competition point; range 7–40); (7) number of workers in the third brood (n-workers3; range 0–294); (8) number of new queens produced (n-queens; range 0–203); (9) number of males produced (n-males; range 0–337). As far as cohort-level means were used, they were arithmetic means, weighing the contributions of colonies equally. However, cohort-specific means of queen and male production were weighted means, adjusted for differences in colony productivity, as these "cumulative" means provide the best estimate of population-wide sex allocation (Boomsma, 1989; Bourke and Franks, 1995).
The oviposition of the first haploid (male) egg produced by the queen was not observed directly, following Duchateau and Velthuis (1988), but was inferred by subtracting 25 days from the date of eclosion of the first adult male. In a series of direct observations, the average developmental time for males in B. terrestris was estimated to be 25.1 days (SD = 0.88). The small variation around this mean of 25 days makes it possible to accurately estimate the start of haploid egg production by mother queens (switch point; Duchateau and Velthuis, 1988). This event is followed by a rather sharp transition, resulting in almost exclusive haploid egg production only a few days after the first haploid egg is laid (Figure 1).
|
A similarly sharp transition as described for the switch point to haploid eggs takes place in the rearing of queens from the last diploid eggs (Duchateau and Velthuis, 1988
). Development of queen larvae takes 30 days on average (SD = 1.07), and only 3 days after the first young larvae are channeled into queen development, 95% of all diploid larvae become queens instead of workers (Duchateau and Velthuis, 1988). In time, the rearing of queens is highly correlated with the start of worker reproductive activities (r =.83, p <.01; see also Bloch, 1999; Duchateau, 1991; Duchateau and Velthuis, 1988), so that the first eggs that develop into queens are produced about 12 days before the start of these worker reproductive activities (12 was the average across the cohorts analyzed here; the study by Duchateau and Velthuis [1988] used an average of 10). The start of reproductive activities by the workers (competition point) was recorded directly. This transition is clearly visible in each colony and is characterized by the following syndrome: (1) The sudden appearance of multiple open eggcups, (2) active oviposition by workers, (3) mutual egg-eating by both queen and workers, and (4) direct aggressive encounters between queen and workers (Duchateau, 1989; Duchateau and Velthuis 1988). In the laboratory colonies analyzed here, the start of worker reproductive activities was mostly derived from the presence of multiple open eggcups. In case of doubt between one day and the next, the additional characteristics 2–4 (which require longer observation) were used.
The significance of worker oviposition and the apparent competition between queen and workers in the final stage of colony development is still poorly understood because it seems that normally very few worker-laid eggs develop as a consequence of high rates of egg removal by queens and nest-mate workers (Duchateau, 1989, 1996; Duchateau and Velthuis, 1988; van Doorn and Heringa, 1986; but see Bourke and Ratnieks, 2001). Because of the low success of late eggs, we terminated all colonies 25 days after the onset of worker reproductive activities (competition point), so that we were sure to have collected only offspring of the queen. This procedure implies that we have missed a number of late males that might have emerged from some of the colonies (either as queen-sons or as worker-sons) if these colonies would have remained healthy (which is unlikely at this late developmental stage, particularly for colonies in the field). As a consequence, our sex ratios are less male biased than they otherwise would have been, and this represents a difference in comparison to other laboratory studies on bumblebee sex allocation (Beekman and Van Stratum, 1998; Müller and Schmid-Hempel 1992a,b; Müller et al. 1992; Owen and Plowright, 1982; Owen et al., 1980). However, as the late males have been inferred to have low reproductive success because of limited mating opportunities late in the season (Beekman and Van Stratum, 1998; Bourke, 1997; Owen and Plowright, 1982; Richards, 1977; Duchateau MJ, Velthuis HHW, and Marien J, unpublished data), we believe that our recording of the reproductive investments of colonies is a rather accurate reflection of the normal field situation in which colonies seldom survive beyond the stage at which we terminated them and in which early "protandrous" males are believed to obtain most of the matings (Bourke, 1997; Bourke and Ratnieks, 2001; Duchateau MJ, Velthuis HHW, and Marien J, unpublished data). Field data directly assessing the emergence of bumblebee males are not available, but a higher male bias in the sex ratio under favorable late-summer conditions has been observed for Polistes wasps (Strassmann, 1984).
Statistical analysis
We analyzed data at different levels. First, we made an overall analysis of the total sex ratio variation among the 197 colonies from the 13 cohorts for which data on both queen and male production were available (see Table 1). Second, we looked at trends in reproductive allocation and sex ratio across the means per cohort. The main objective of the latter analyses was to clarify the significance of the variable duration of hibernation diapause of mother queens, which was constant for all colonies within a cohort and known (range 2–8 months) for all cohorts of laboratory-reared queens (27 out of 31; Table 1). As deviations from unimodal normal distributions were absent or relatively minor, these analyses were done with parametric correlation and regression techniques. Third, we extended the analysis again to the colony level, pooling the data for the cohorts for which data on both queen and male production were available. This implied that our data set was reduced to 152 colonies, because the duration of hibernation of 2 of the 13 cohorts was unknown (queens collected after natural hibernation in the field) and because missing values for one or several of the remaining key variables excluded some of the remaining colonies.
Analyzing colony-level sex allocation data in social insects has various complications (Boomsma and Nachman, 2002). One of them is that sex-ratio decision-making is a sequential process, where feedback connections between a series of variables may affect the final outcome. The analysis of sex allocation in large perennial societies of ants may be reasonably approximated by including a limited number of covariates such as relatedness asymmetry, colony size, colony productivity, and body size in the analysis (Boomsma et al., 1982; Deslippe and Savolainen, 1995; Nonacs, 1986; Rosenheim et al., 1996; Sundström, 1994), so that logistic regression is a good choice (Boomsma and Nachman, 2002). However, this approach is less realistic in annual species, where it is less clear whether covariates are causes or consequences of sex allocation decisions. In such cases, path analysis (Backus, 1995; Herbers, 1990) is a superior technique, provided that the number of females and males per colony are large enough so that sex ratio sampling error per colony is relatively small.
We used the path-analysis LISREL 8 program (Jöreskog and Sörbom, 1993), which implies that the actual numbers of queens and males produced were entered as separate dependent variables. The short colony life cycle of B. terrestris and the detailed data on colony development available for our study colonies allowed us to specify a single hypothetical path-model with a single independent variable (hibernation) and nine sequential dependent variables, representing the natural order in which colonies develop (Figure 2). Our model ranks these key variables of colony development and reproduction on a time axis (from left to right in the figure) and assumes that variables occurring later in time can only be influenced by variables happening earlier during colony development. The number of possible links was somewhat restricted (to prevent the analysis from requiring memory beyond the capacity of our computer) by assuming that variables did not exert influence more that three or four time-steps ahead and by assuming that variables before the second worker brood were only affecting later events through the size of the second brood. Three additional links, however, were added to the path diagram after they appeared to be significant in additional runs, after the basic model had been tested (see Results).
|
Because of the inevitable reduction of genetic diversity in a number of our laboratory cohorts (Table 1), the effective sample size may have been less than suggested by the total number of colonies involved. To remain on the safe side, we have therefore based our conclusions of the colony-level analyses only on relationships that were significant at the p =.01 level. Cohort-level analyses of means were assumed not to have been affected by reduced genetic diversity or inbreeding, so that normal p =.05 significance levels were used.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Patterns of sex ratio variation in the total data set
The total variation in sex allocation among colonies was clearly bimodal, with relatively many colonies producing either mostly males or mostly females (Figure 3a). Most colonies produced between 50 and 200 queen equivalents (
100–400 male equivalents, using a female-to-male cost ratio of about 2; see also below). There was a significant negative correlation between the number of females and males produced per colony (r = -.666; p <.001; n = 197; Figure 3a), which implies that the controlled ad libitum feeding regime of colonies makes the trade-off between queen and male production clearly visible. However, relatively many colonies produced mixed sex ratios, and there was no indication that these colonies were less or more productive overall. If there had been such productivity differences, the negative correlation in Figure 3a would have been convex or concave. Neither log-transforming the number of queens and males nor analyzing their squared values improved the linear correlation coefficient (r = -.666), so significant deviations from linearity seem unlikely. A total of 10,648 queens and 22,021 males were reared in the 197 laboratory colonies (13 cohorts), which gives an overall numerical proportion of queens of 0.326 and a proportional investment in queens of 0.450–0.505, depending on whether one uses the average dry weight ratio of an individual queen and male (2.11; see Duchateau and Velthuis, 1988) or its approximate 0.7 power energetic equivalent (Boomsma, 1989; Bourke, 1997; Helms, 1994). Alternative estimations by taking sex ratio means across all colonies (weighing them equally and ignoring productivity differences) gave similar results: a numerical proportion of queens of 0.424 and an average female:male sex investment ratio (proportion of investment in females) of 0.477–0.500. The pattern of mean queen and male production across cohorts showed a similar negative correlation (Figure 3b) as the sex ratio variation among colonies (Figure 3a).
|
A significant proportion (28%; F12,182 = 6.87; p <.001) of the total variance in sex ratio among colonies appeared to be due to variation among means per cohort (Figure 3b). Assuming that laboratory cohorts are independent samples of the same population and that the cohort-specific differences in sex allocation reflect natural variation that may also occur in the wild, we considered it worthwhile to analyze the variation across cohorts in more detail. The mean cumulative female:male sex ratio across cohorts was 0.313, with an SD of 0.192 and an SE of 0.053. Applying the conversion factors as above, the corresponding proportional weight and energy investments in queens were 0.414 and 0.460, respectively (SD = 0.204; SE = 0.057). Both these estimates were significantly lower than 0.75, the sex allocation ratio expected under complete worker control (p <.001), but not significantly different from 0.50, the sex allocation ratio expected under complete queen control (p >.1). However, the distribution of the cohort means also indicated that worker–queen conflict theory over sex allocation cannot be tested when only a single random cohort of bumblebees is reared without controlling for the effect of overwintering conditions (see below). With a cohort sample size of 1, the SE equals the SD, implying that our observed overall means are neither significantly lower than 0.75 (.1 < p <.2, one-tailed test), nor significantly different from 0.50 (p >.5, two-tailed tests). The variation across B. terrestris laboratory cohorts (Figure 3b; range of cohort-specific means of proportional energetic investment in queens: 0.161–0.780) is similar to the variation of single cohorts across nonparasitic bumblebee species reported by Bourke (1997
; range: 0.075–0.638). These previously evaluated means did not allow species (cohort)-specific conclusions on the significance of worker control and queen control.
Effect of queen hibernation on sex allocation of B. terrestris cohorts
Across the 11 cohorts for which both queen and male production were measured, hibernation duration was significantly negatively correlated (r = -.75; p =.01) with the number of queens produced per cohort, whereas the correlation between hibernation duration and the number of males produced was positive (r =.526), but not significant (p =.10). Figure 4 presents this contrast for a larger sample, including also the cohorts for which only queen production was measured (Table 1). In this larger data set, the correlation coefficient for queens was of similar significance (-0.411; p <.05), and a highly significant difference was also found when both relationships were tested directly against each other for differences in the regression slopes as plotted in Figure 4 (ANCOVA, F1,34 = 9.975; p <.01). The result remained essentially the same when only the fully overlapping hibernation durations of 2–5 months were included in the analysis (F1,28 = 8.151; n = 32; p <.01) or when the analysis was restricted to the 11 cohorts with artificially hibernated queens for which both queen production and male production (see Table 1) were measured (F1,18 = 8.649; p <.01). When using the regression lines of Figure 4 to infer the expected population-wide sex allocation ratio (using the 1.69 cost-ratio conversion) as a function of hibernation duration at 5°C, we would thus expect a 75% proportional investment in queens at a winter diapause duration of 1 month, monotonously decreasing to 45% at 3 months and 20% at 6 months.
|
Colony-level sex ratio variation and path analysis
The result of the path analysis comprising the entire data set of 152 colonies for which all variables were measured is given in Figure 5. As indicated by the model statistics in the figure legend, the overall fit of the path analysis was excellent, with 16 of the 19 plotted links being significant at the p <.001 level. Most of the links hypothesized in Figure 2 were confirmed as statistically significant. Figure 5 quantifies these hypothesized links, omits the links that were hypothesized as possible but that did not come out significant, and adds a few links that were excluded from the predicted path model (because they extended over many time steps), but which turned out to be significant when added to the analysis one by one after testing the basic path model of Figure 2. Hibernation had a direct significant effect on many of the key variables that determine early colony development: Founding queens that had been in hibernation diapause longer than average had larger first and second worker broods, earlier switch points and earlier competition points. The strength of these effects increased with time (n-workers1 < n-workers2 < switch point), but the direct effect on the competition point is weaker. Hibernation did not seem to have a direct effect on the rate of colony initiation (egg1 and worker1), although the appearance of the first worker (worker1) is obviously earlier when the first egg was laid earlier and later when the first brood (n-workers1) is large. An unexpected effect of egg1 on the size of the third worker brood (n-workers3) was also found. These and other remaining links in Figure 5 represent indirect effects (i.e., links expressed after the overall effect of hibernation has been taken into account).
|
of 0.01 for each hypothesized link. Variables are described in text. The t values for specific causal relationships are given as figures (black and gray background for positive and negative effects, respectively) on the arrows linking the boxes in the plot. All values >3.36 are significant at p <.001. General LISREL 8 goodness-of-fit test statistics: 
2 = 46.51 (df = 25, p =.0056); root mean square error of approximation = 0.076; adjusted goodness-of-fit index = 0.88The significant path links later in colony development show that, first, larger first and second worker broods promote early switch points and early competition points, independent of the effect of hibernation. Colonies following this trajectory produce predominantly males. The switch point thus induces a direct trade-off between male production and combined (third worker brood and queens) female production. However, the number of workers in the second brood has no direct effect on the production of queens and third brood workers. Second, colonies with an early competition point, sometimes even preceding the switch of the queen to haploid eggs (this happened in 7 of the 152 colonies = 4.6%), produce predominantly queens but few males and few third brood workers. This implies that the competition point induces a trade-off between queen and male production, but one that is mostly opposite to the trade-off induced by the switch point. Accordingly, the sex ratio is correlated both with the absolute timing of the switch point and the competition point (both are in fact positively correlated; Figure 5) and with the time span between these two crucial switches in the life cycle of a B. terrestris colony. In other words, colonies with an early switch point also tend to have an early competition point and a male-biased sex ratio (Figure 5), but it is the extent to which colonies deviate from this simple positive correlation that determines most of the sex ratio (Figure 6). When a queen has been laying haploid eggs long before the competition point (i.e., the higher the positive value for the time span between the competition point and the switch point), the colony will produce fewer new queens and many males. On the other hand, when the difference between switch point and competition point is small or negative (when the competition point occurs before the switch point), few or no males but relatively many queens are produced (Figure 6a,b). The colony sex ratio thus becomes increasingly male biased with larger (positive) differences between these two crucial switches in the colony cycle (Figure 6c).
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Observed patterns of bumblebee sex allocation have been highly puzzling from an evolutionary point of view. In a review of available comparative data, Bourke (1997)
showed that population-wide sex allocation ratios of bumblebees were not sufficiently female biased (and often too male biased) to fit Trivers and Hare's (1976) predictions of worker-controlled sex allocation. A later study by Beekman and Van Stratum (1998) and our present results confirm this pattern and document that sex ratios of B. terrestris are highly split. Although colony-level resource differences (Beekman and Van Stratum, 1998; Nonacs, 1986) can theoretically induce split sex ratios in the absence of relatedness asymmetry (Boomsma and Grafen, 1990, 1991), we find this resource-based explanation unconvincing for our present results because all colonies were reared under controlled conditions and with ad libitum food. We will therefore discuss our results primarily in the light of a recent conceptual framework offered by Bourke and Ratnieks (2001), which hypothesizes that bumblebee queens achieve their optimal sex ratio by preventing queen production in half of the colonies, an idea originally developed by Pamilo (1982) and recently applied in Pheidole ants by Helms (1999).
In the present study we analyzed more colonies of a single bumblebee species B. terrestris than were available for 11 species in the review by Bourke (1997). Our sample size was also considerably larger than the 41 laboratory colonies in another recent study of B. terrestris by Beekman and Van Stratum (1998). However, the main strength of our data set is that we measured and analyzed most of the key variables of queen diapause and colony development that potentially affect the production of new queens and males. This allowed us to document, first, that variation in mean sex allocation across cohorts reared in the same laboratory is substantial, that there is a highly significant negative correlation between the number of queens and males produced in each colony (Figure 3), and that a significant part of this variation is related to the duration of hibernation-diapause (Figure 4). The study of Beekman and Van Stratum (2000) pointed also in this direction, but it did not firmly establish this trend: colonies (n = 21) tended to produce more males and fewer queens after 4 months of queen diapause compared with a similar set of colonies (n = 27) whose queens had diapaused for 2 months.
Second, we documented that average sex allocation ratios in B. terrestris are less female biased than expected (assuming single mating of queens) under worker control (both before and after partialing out the effects of hibernation), but closely match a 1:1 allocation expected under queen control (Figure 3). This implies that the present data provide stronger evidence for queen control over sex allocation than any previous studies in bumblebees (see Bourke, 1997). We realize that this conclusion is somewhat dependent on the cost-ratio conversion factors used and that one study (Beekman and Van Stratum, 1998) used an energy content cost ratio that yields higher investments in queens. However, we believe that this may easily overestimate female bias, as energetic costs due to respiration (which are normally not measured) tend to shift cost ratios in the opposite direction (Boomsma, 1989).
Third, our studies showed that the timing of the switch of queens to haploid egg production is of overwhelming importance for sex allocation and late colony development (third worker brood and competition point; t values > 10 in the path-links to both the number of males and the number of queens produced; see Figure 5). This reinforces the conclusion of Duchateau and Velthuis (1988) and Duchateau (1991) based on a smaller sample size. A novel finding of this study is that the switch point is severely affected by hibernation duration, both directly and indirectly via the number of workers in the first and second brood. Hibernation even directly affects the (usually later occurring) competition point, but this effect is less strong.
Finally, the switch point, competition point, and the size of the third worker brood are highly positively correlated (Figure 5), but it is the absolute number of days that pass between the two crucial switches to haploid egg-laying that ultimately determines most of the variation in the sex allocation ratio (Figure 6). This variable time window for the production of young queens was known from the earlier work by Duchateau (1991), but the present study confirms its generality across different cohorts of laboratory-reared colonies. The opposing effects of switch point and competition point on the sex ratio explain why these two transitions may be only weakly correlated in some studies (Bloch, 1999; Duchateau and Velthuis, 1988; note, however, that they are highly correlated in the present study; see Figure 5).
Hibernation diapause and life expectancy of queens
For B. terrestris bumblebees, in which, under natural conditions, all queens have diapause, it has been difficult to find direct relationships between queen condition (expressed in weight and/or size) and the duration of her diapause. Under controlled conditions and using a similar range of diapause durations as in our study, Beekman et al. (1998) showed that queens below a certain mass at the onset of diapause will normally not survive hibernation. They also found that diapause duration was significantly negatively correlated with queen survival, a result that was also obtained in another sample of our study population (r = -.615, n = 25, Duchateau, unpublished data). However, Beekman et al. (1998) did not find an effect of diapause temperature (range -5 to 15°C) on queen survival, suggesting that metabolism during diapause is independent of temperature. Finally, Beekman and Van Stratum (2000) found that 10 colonies whose queens had not diapaused at all developed faster and produced relatively many new queens, suggesting an effect of diapause duration on colony development and reproduction. Overall, these previous results thus agree with our present findings that duration of diapause under controlled conditions is consistently correlated with many variables of colony development, colony productivity, and sex ratio.
We hypothesize that hibernation duration is a direct (although probably noisy) predictor of life expectancy for individual queens and thus for entire colonies. In particular queens that have survived long hibernations have suffered higher metabolic losses of body reserves and are likely to have incurred some reductions in their average potential life span, selecting for a correlated response that avoids a type of colony organization whose reproductive benefits take more time to be realized. Queens may thus use their hibernation experience to choose between alternative developmental pathways for their colonies: a more rapid colony development with an early switch to haploid eggs, so that mostly males are produced, or a more prolonged colony development in which more new queens are produced and in which male production starts relatively late if at all. We realize that spring queens are able to replenish their body reserves by active foraging in spring. However, given the large demands during the colony founding stage, it seems likely that this will at best partly compensate for losses incurred during hibernation because there are almost certainly trade-offs between foraging and other essential allocation efforts in this periods.
Beekman et al. (1998) also showed that larger queens (with large radial wing cells) survived diapause better than smaller queens, but our study indicates that queen body size is not important as a further predictor of colony development and colony productivity. Data on the diameter of radial cells of queen wings were available for a subset of our colonies, but this parameter was never significantly correlated with any of the key variables we used. Because including radial cell size of queens in our formal analyses would have given a much higher number of entire-colony deletions due to missing values, we chose to ignore this variable. We note, however, that this does not exclude that variables connected to queen weight or queen fat reserves could have a significant effect on development and reproduction of colonies. In particular, we expect that the ratio between queen size and weight (or fat reserves) directly after diapause could be a useful proxy for life expectancy, so that relatively light queens would have early switch points and male-biased colonies, and relatively heavy queens would have late switch points and female-biased colonies. Life expectancy of queens is thus likely to be affected by three factors: feeding and rearing conditions in the maternal colony (Ribeiro et al., 1996), relative shelter of the overwintering burrow of each queen, and overall harshness of the particular winter diapause. The first two factors cause variation in queen life expectancy within years and laboratory cohorts, whereas the third factor induces variation across years and laboratory cohorts. When variation in the third factor is small relative to the variation in the first two factors, it seems likely that split sex ratios with averages around 1:1 would also be found in field populations.
The duration of queen hibernation thus seems to have an important indirect and cumulative effect on colony development and sex allocation. Lack of standardization for this and related factors in laboratory cultures may therefore explain at least part of the large variation reported in sex allocation ratios across and within bumblebee species (see Bourke, 1997). Clearly, if our laboratory cohorts had all been reared after a hibernation period of 4 months, the cohort-specific differences in mean sex allocation would have been less pronounced than in Figure 3b. When using the smaller SDs across cohorts after partialing out the effects of diapause duration on the production of queens and males (Figure 4), the overall proportional investment in queens is significantly lower (one-tailed test) than 0.75 when using the 1.69 conversion (t = 2.127, p =.028) and also when using the 2.11 conversion (t = 1.790, p =.049), but both estimates remain indistinguishable from the queen sex allocation optimum of 0.5. This implies that with controlled overwintering conditions, the distribution of cohort-specific sex allocation means may be just narrow enough to reject the hypothesis of full worker control on the basis of data from a single cohort of colonies only.
Protandry, queen-controlled split sex ratios, and the major transitions in the bumblebee colony cycle
In a recent conceptual study, Bourke and Ratnieks (2001) concluded that by switching early (i.e., at the start of the third worker brood), bumblebee queens effectively prevent workers from producing gynes. In their view, workers comply to this male-biased reproductive agenda of their mother because early males have a higher than average mating success, and they can only detect their mother's switch to haploid eggs with some delay, so that replacing brothers by worker sons would incur both a significant efficiency cost and lower fitness returns because of reduced mating success of the later replacement males. Bourke and Ratnieks (2001) suggest that the second factor (constrained information) is the most important one, but our present results indicate that both factors may be of similar importance. The argument by Bourke and Ratnieks is based on Duchateau and Velthuis's (1988) study of 21 colonies of B. terrestris, indicating that early- and late-switching colonies initially develop at similar rates. However, our present much larger data set shows that early-switching colonies have significantly larger first and second worker broods than late-switching colonies, so that workers should have at least partial information about the likely future development of their colony, independent of their ability to recognize haploid and diploid brood.
The balance in favor of raising queen-sons in spite of their low kin value apparently stays favorable for quite some time, because workers in early-switching colonies seem to actively postpone interfering with their mother's haploid egg production. This is remarkable because workers in colonies heading for a male-biased sex ratio tend to develop their ovaries (Duchateau and Velthuis, 1989; Van Doorn and Heringa, 1986; Van Honk et al., 1981) and seem thus fully able to replace queen-laid haploid eggs by worker-laid haploid eggs. The crucial factor to understand this paradox is to realize that the kin value of nephews (relatedness = 0.375) is only 1.5 times as high as the kin value of brothers (relatedness = 0.25). This means that replacing queen eggs by worker eggs will only be worthwhile if productivity multiplied by mating success will be at least two-thirds of the value obtained when rearing exclusively brothers. From what is known or inferred about loss of colony coherence and productivity after the onset of worker reproductive activities (Duchateau and Velthuis, 1988; Van Doorn and Heringa, 1986) and about the higher mating success of early males (Bourke, 1997; M. J. Duchateau, H. H. W. Velthuis, and J. Marien, unpublished data), this two-thirds threshold may well be impossible to overcome until quite late in colony development. In other words, workers in male-biased colonies would have even lower inclusive fitness if they rebelled against their fate too early, an example of high potential conflict between queen and workers, but low actual conflict because the cost of manipulation outweighs the benefits (Ratnieks and Reeve, 1992).
We have summarized these interpretations of the key processes affecting sex allocation in B. terrestris in a heuristic graphical model (Figure 7). The concave downward curves in the lower panel describe the likely reproductive value (mating success multiplied by relatedness) of males as a function of time in the season, relative to the constant reproductive value of queens (assumed to be 1). The upper curve in this panel gives the value of worker-produced males to other workers (MW to W; assuming that life-for-life relatedness is 0.375), which is equivalent to the value of queen-produced males to queens (MQ to Q) as long as the population sex-investment ratio is 1:1 and the alternative is investing in full sisters or daughters, respectively. The similar lower curve gives the value of queen-produced males to workers (MQ to W; life-for-life relatedness = 0.25). The time spans t1, t2, and t3 indicate the increasing maximal delays that are allowed to raise a worker-son of the same reproductive value as the queen-son that it replaces. In early-switching colonies the competition point is predicted to be delayed until a time threshold, tn, from which it pays to replace queen-sons by worker-sons. Because of the information constraints discussed by Bourke and Ratnieks (2001), it may take quite some time to reach tn, which explains that the most male-biased colonies have a relatively early switch point, a rather late competition point, and thus a large difference between these two switches (see Figures 5 and 6; see also Figure 1 in Bourke and Ratnieks, 2001). The reproductive values in the lower panel of Figure 7 are determined by reproductive decisions about 35 days earlier: the queen-imposed switch point to haploid eggs (S) and the worker-imposed competition point (C). In the upper panel of Figure 7, we have approximated the cumulative probabilities of S and C across a typical sample of colonies that correspond to the reproductive values in the lower panel 35 days later.
|
In female-biased, queen-producing colonies, workers gain higher inclusive fitness than in male-biased colonies. In our interpretation, these are colonies with relatively healthy queens that have emerged from overwintering with an above-average life expectancy and thus relatively good prospects for productivity during the additional 3 weeks or so that are typically required for this pathway of colony development. The reduced production of nonvolatile pheromones by queens (Cnaani et al., 1997
; Duchateau and Velthuis, 1988; Röseler, 1991) in the critical time window before the competition point could well be an honest signal to the workers that their mother is aiming for a long colony cycle and a female-biased sex ratio. This interpretation would thus be an extended version of the idea by Bourke and Ratnieks (2001) that the cessation of pheromone production is a signal to the female larvae to start development as queens and that workers eavesdrop on this signal. It is important to realize that protandry implies that at this point in time the expected mating success of any male yet to be produced has already dropped substantially, so that both queen and workers maximize their fitness by producing new queens. Queens and workers thus agree about the best reproductive option (the hatched area in Figure 7), but they disagree about the second best option, males. This argument remains rather implicit in the model by Bourke and Ratnieks, but it seems to be the best ultimate explanation for the enforcement of the competition point by workers rather predictably 12 days after the first eggs that develop into queens are laid. At this point in time, the new queen production is secure, and the queen may have switched (or is about to switch) to producing haploid eggs.
When the point has been reached that worker egg-laying no longer jeopardizes the efficient rearing of queen brood (i.e., when all queens that a colony can afford are either close to pupation or at least recognizable because of their enhanced growth rate at about 8 days of age; Ribeiro et al., 1999; see also Bortolloti et al., 2001), the optimal worker strategy is to produce many male eggs and to get engaged in a mutual egg-eating arms race with the queen because this will ensure the highest share in the last cohort of males to be raised by the colony. This is an optimal strategy regardless of the success of male production at the end of the colony cycle. Even if the number of late males produced in field colonies and their mating success are both likely to be low, any mechanism that allows queens and workers to increase their share in this limited reproductive option will be selected for when alternative options are no longer available. The expression of worker–queen conflict over male production at this point thus has essentially a very low cost, an inference that also follows from the Bourke and Ratnieks (2001) study.
As argued by Bourke and Ratnieks (2001), information constraints in the detection of haploid queen-eggs by workers are less likely to apply in late-switching colonies because all parties (queen, workers, and even the diploid larvae) agree that raising queens is the best strategy. The number of queen larvae will probably be fine tuned by the quality of the available resources and the number of workers available to forage. This means that there also is a large degree of agreement on the termination of this queen-production project, so that the workers should be able to predict the switch point of their mother much better than in early-switching colonies. The relative constancy of the period between the onset of queen production and the competition point and the inferred steady decline of a queen signaling pheromone during this period (Cnaani et al, 2000) are therefore not surprising. The optimal strategy of late-switching colonies is also illustrated in Figure 7. Queens and workers agree on raising new queens when the highest male curve in the lower panel of Figure 7 has crossed the queen reproductive value line of 1 (the hatched area). From then on, queens should postpone their switch to haploid eggs until the colony has realized the maximum production of new queens for which it has resources.
Following Bulmer (1981, 1983) and Bourke and Ratnieks (2001), we conclude that the early male advantage that has been inferred to characterize the B. terrestris mating system is likely to be the ultimate cause of the queen-controlled split sex ratios observed. Queens in about half of the colonies prevent the production of new queens because they no longer lay diploid eggs at the crucial point in time when potential queen larvae could be raised (Bourke and Ratnieks, 2001). The early-switching colonies thus function as a balancing sex ratio class (Boomsma and Grafen, 1991), imposing the overall queen optimum of a 1:1 sex allocation ratio in the population. Our model in Figure 7 encompasses the conceptual framework of Bourke and Ratnieks (2001) while adding considerable empirical detail based on a large data set. Our study primarily addressed sex allocation and explored the consequences of reduced mating success of late males, whereas the primary emphasis of the Bourke and Ratnieks study was on information constraints and worker reproduction (although they considered sex allocation, costs, and the timing of male production). Combining the information from these two studies, we have now reached a relatively firm understanding of the paradox that bumblebee reproduction is on one hand fully consistent with kin selected worker behavior, but on the other hand it is constrained by semelparity and protandry, so that workers hardly have any power to significantly affect sex allocation or male production.
|
ACKNOWLEDGEMENTS |
|---|
We thank Janine Mariën for her help in the rearing and data collection of many cohorts of colonies and Marcia Ribeiro for rearing cohorts mv90. Herman Wijnne and Paul Westers gave invaluable help with the path analysis, David Nash helped with the final versions of figures, and Vicky Backus, Madeleine Beekman, Andrew Bourke, Paul Schmid-Hempel, and two anonymous reviewers made valuable suggestions for improvement of previous versions of this article. M.J.D. has been supported by several grants from the Netherlands Technology Foundation of the Netherlands Organization for Scientific Research. The stay of J.J.B. in Utrecht has been supported by a visiting fellowship of the Netherlands Organization for Scientific Research.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
Alexander RD, Sherman PW, 1977. Local mate competition and parental investment in social insects. Science 196:494-500.
Aron S, Passera L, Keller L, 1994. Queen-worker conflict over sex ratio. A comparison of primary and secondary sex ratios in the Argentine ant, Iridomyrmex humilis. J Evol Biol 7:403-418.[CrossRef][ISI]
Aron S, Vargo EL, Passera L, 1995. Primary and secondary sex ratios in monogyne colonies of the fire ant. Anim Behav 49:749-757.[CrossRef]
Backus V, 1995. Rules for allocation in a temperate forest ant: demography, natural selection, and queen worker conflict. Am Nat 145:775-796.[CrossRef]
Beekman M, van Stratum P, 1998. Bumblebee sex ratios: why do bumblebees produce so many males? Proc R Soc Lond B 265:1535-1543.[CrossRef]
Beekman M, van Stratum P, 2000. Does diapause experience of bumble bee queens Bombus terrestris affect colony characteristics? Ecol Entomol 25:1-6.
Beekman M, van Stratum P, Lingeman P, 1998. Diapuase survival and post-diapause performance in bumblebee queens (Bombus terrestris). Entomol Exp Appl 89:207-214.[CrossRef]
Bloch G, 1999. Regulation of queen-worker conflicts in bumble-bee (Bombus terrestris) colonies. Proc R Soc Lond B 266:2465-2469.[Medline]
Boomsma JJ, 1989. Sex-investment ratios in ants: has female bias been systematically overestimated? Am Nat 133:517-532.[CrossRef][ISI]
Boomsma JJ, Grafen A, 1990. Intraspecific variation in ant sex ratios and the Trivers-Hare hypothesis. Evolution 44:1026-1034.[CrossRef][ISI]
Boomsma JJ, Grafen A, 1991. Colony-level sex ratio selection in the eusocial Hymenoptera. J Evol Biol 4:383-407.
Boomsma JJ, Lee GA van der, Have TM van der, 1982. On the production ecology of Lasius niger (Hymenoptera: Formicidae) in successive coastal dune valleys. J Anim Ecol 51:975-991.[CrossRef]
Boomsma JJ, Nachman, , 2002. Analysis of sex ratios in social insects. In: Sex ratios: concepts and research methods (Hardy ICW, ed). Cambridge: Cambridge University Press; 93–111.
Bortolloti L, Duchateau MJ, Sbrenna G, 2001. Effect of juvenile hormone on caste determination and colony processes in the bumblebee Bombus terrestris. Entomol Exp Appl: 1–17.
Bourke AFG, 1997. Sex ratios in bumble bees. Phil Trans R Soc Lond B 352:1921-1933.[CrossRef]
Bourke AFG, 1999. Colony size, social complexity and reproductive conflict in social insects. J Evol Biol 12:245-257.[CrossRef][ISI]
Bourke AFG, Franks NR, 1995. Social evolution in ants. Princeton,. New Jersey: Princeton University Press.
Bourke AFG, Ratnieks FLW, 2001. Kin-selected conflict in the bumble-bee Bombus terrestris (Hymenoptera: Apidae). Proc R Soc Lond B 268:347-355.[Medline]
Bulmer MG, 1981. Worker-queen conflict in annual social Hymenoptera. J Theor Biol 93:239-251.[CrossRef]
Bulmer MG, 1983. The significance of protandry in social Hymenoptera. Am Nat 121:540-551.[CrossRef]
Chapuisat M, Sundström L, Keller L, 1997. Sex ratio regulation: the economics of fratricide in ants. Proc R Soc Lond B 264:1255-1260.[CrossRef]
Cnaani J, Borst DW, Huang ZY, Robinson GE, Hefetz A, 1997. Caste determination in Bombus terrestris: differences in development and rates of JH biosynthesis between queen and worker larvae. J Insect Physiol 43:373-381.[CrossRef][ISI][Medline]
Crozier RH, Pamilo P, 1996. Evolution of social insect colonies: sex allocation and kin selection. Oxford: Oxford University Press.
Deslippe RJ, Savolainen R, 1995. Sex investment in a social insect: the proximate role of food. Ecology 76:375-382.[CrossRef]
Duchateau MJ, 1989. Agonistic behaviours in colonies of the bumblebee Bombus terrestris. J Ethol 7:141-151.
Duchateau MJ, 1991. Regulation of colony development in bumblebees. Acta Hort 288:139-143.
Duchateau MJ, 1996. Is kin conflict expressed in the colony cycle of the bumble bee Bombus terrestris? In: XX International Congress of Entomology, Firenze, Italy, August 1996, Firenze; 404.