Reproductive ground plan may mediate colony-level selection effects on individual foraging behavior in honey bees

Abstract

The colony-level phenotype of an insect society emerges from interactions between large numbers of individuals that may differ considerably in their morphology, physiology, and behavior. The proximate and ultimate mechanisms that allow this complex integrated system to form are not fully known, and understanding the evolution of social life strategies is a major topic in systems biology. In solitary insects, behavior, sensory tuning, and reproductive physiology are linked. These associations are controlled in part by pleiotropic networks that organize the sequential expression of phases in the reproductive cycle. Here we explore whether similar associations give rise to different behavioral phenotypes in a eusocial worker caste. We document that the pleiotropic genetic network that controls foraging behavior in functionally sterile honey bee workers (Apis mellifera) has a reproductive component. Associations between behavior, physiology, and sensory tuning in workers with different foraging strategies indicate that the underlying genetic architectures were designed to control a reproductive cycle. Genetic circuits that make up the regulatory “ground plan” of a reproductive strategy may provide powerful building blocks for social life. We suggest that exploitation of this ground plan plays a fundamental role in the evolution of social insect societies.

Eusocial insects are characterized by reproductive and task-associated division of labor, cooperative brood care, and the presence of a sterile worker caste. During development and adult life, individuals follow bifurcating strings of regulatory events that yield phenotypes with different morphology, physiology, and behavior (e.g., refs. 1 and 2). Genetic architectures underlying stimulus–response relationships of behavior may have been exploited during social evolution (3). Coupled with the behavioral plasticity of the individual insect, this relationship facilitates self-organization of division of labor within a group (4–6). The resulting social structure provides a foundation from which distinct morphs can be derived through phenotypic compartmentalization (7–9).

In terms of behavior, the best-studied eusocial insect is the honey bee (Apis mellifera). It has been suggested that an adaptive response to colony-level selection can emerge in honey bees through genetic modulation of stimulus–response relationships that affect the task performance of the individual insect (3). Selection on the amount of stored pollen in honey bee colonies (pollen hoarding) changes the probability that worker bees will forage for pollen (10). Pollen-foraging behavior is correlated with the responsiveness to sucrose (11–13). It is unlikely that sucrose responsiveness is a direct cause of foraging behavior (3, 14); however, the associations may have a common causal structure that tunes the sensory system of the bee (14). These findings elucidate how natural selection may change the phenotypes of insect societies through modulation of the lower levels of biological organization (3), and at the same time they question the extent of regulatory decoupling in social evolution. This is because colony-level selection on pollen hoarding affects a whole suite of physiological and behavioral traits including age at onset of foraging (10, 15, 16), locomotor activity at emergence (R.E.P., unpublished data), associated learning performance (13, 17), and sensory perception (18–20). It is likely that the coordinated changes result from pleiotropic gene action (14, 21), but it is not easy to explain why these associations exist.

It is believed that pleiotropy plays an important role in physiological and behavioral integration (9, 22, 23). Pleiotropy facilitates the coordinated expression of complex temporal phenotypes over the life cycle of solitary insects through regulatory switches capable of controlling large number of genes (24, 25). At least partly because of this mechanism, the different stages in the female reproductive cycle, i.e., the previtellogenic phase, vitellogenesis, oviposition, and brood care (if applicable), constitute a coordinated string of associated behavioral and physiological events (24–26). In some social hymenopterans, temporal division of labor seems to be superimposed on a rudimentary ovary-development cycle in which young individuals dedicated to brood rearing have slightly developed ovaries or lay trophic eggs, and poorly developed ovaries are associated with foraging [reviewed by West-Eberhard (8)]. This may indicate that the regulatory circuits underlying temporal division of labor in eusocial insects are derived from a reproductive ground plan [i.e., the ovarian ground plan suggested by West-Eberhard (8)].

In the presence of a queen, honey bee workers have undeveloped ovaries independent of behavioral state (27). However, the yolk precursor vitellogenin, the major yolk protein in oviparous animals, constitutes the main part of the hemolymph protein fraction in bees that do not forage (28, 29). The vitellogenin titer increases from emergence to when the bee is 7–10 days old (29). Workers that loose their queen within a week after emergence develop their ovaries to a larger extent than older bees (30). Additionally, the Cape honey bee (A. mellifera capensis) shows rising vitellogenin levels during the first week of adult life, and workers produce well formed eggs at the age of 7–10 days (31). The physiological state of a young honey bee may therefore be similar to a vitellogenic phase in a solitary female (30, 32). As the bee gets older the vitellogenin titer declines, and at the onset of foraging the production is switched off (33). In solitary bees, oogenesis overlaps with foraging (1). However, female life stages characterized by intense flight are frequently associated with reproductive diapause in insects [the “oogenesis-flight syndrome” (e.g., ref. 34)]. The foraging phase of a worker may constitute a similar regulatory phenomenon at the physiological level (35), although the behavioral phenotype is that of a female provisioning a brood cell or foraging for individual consumption before a (new) period of vitellogenesis. Thus, the framework of the reproductive ground plan may be a valid approach to the evolution of temporal castes in honey bees (as originally suggested by West-Eberhard in ref. 8). However, the question remains as to whether it has explanatory power to address pleiotropic associations between physiological and behavioral traits in individual bees.

In this study we explore the pleiotropic genetic network that responds to selection on pollen hoarding in honey bees and its link to regulatory circuits involved in reproduction. To this end we measured the hemolymph titer and transcription level of vitellogenin in young bees from strains selected for high and low pollen hoarding. We show that both titer and transcription level are higher in the high pollen-hoarding strain during the first 10 days of adult life. The results also indicate that vitellogenin expression is under tight regulatory control. The titer of vitellogenin is causally linked to the reproductive potential of honey bee workers (30, 32), and the suite of behavioral and physiological traits that respond to selection on pollen hoarding thus seems to have a reproductive component. We argue that the main features of the suite can be explained in the framework of the reproductive ground plan and suggest that exploitation of genetic circuits that control the reproductive tuning of insects plays a fundamental role in social evolution. We predict that physiological and behavioral traits governed by the ground plan will be linked in a broad range of social insect species and use data from honey bees to derive specific predictions about these associations.

Materials and Methods

Bees. The experiments were performed at the University of California, Davis. Newly emerged workers were obtained from the high and low pollen-hoarding strains selectively bred by Page and Fondrk (36) as a mix from three colonies per strain. The amount of stored pollen was 1,754, 922, and 1,335 cm² for the high-strain colonies and 52, 26, and 26 cm² for the low-strain colonies (see ref. 36 for additional details on the breeding scheme). The bees were collected for analyses or marked on the thorax with a spot of paint (Testors Enamel) to indicate age and strain. The marked workers were added to three unrelated “host” colonies and collected when they were 5 and 10 days old.

Sample Collection. Bees were anesthetized on ice, and hemolymph samples were extracted with Drummond Scientific (Broomall, PA) micropipettes (1 ± 0.001 μl) after puncturing the abdomen between the third and the fourth tergite with a sterile needle. Care was taken to avoid contaminating the samples with tissue fragments and foregut content from the bees. The hemolymph (1 μl) was dissolved in 10 μl of Tris buffer [20 mM Tris/150 mM NaCl/5 mM EDTA, pH 7.5/1 mM phenylmethylsulfonyl fluoride/5 mM benzamidin/0.7 μM pepstatin/8 μM chymostatin/10 μM leupeptin/0.8 μM aprotinin (Sigma)]. The abdomen was separated from the thorax with forceps, and the intestine was removed. The abdominal segments with adhering fat-body tissue [responsible for vitellogenin synthesis in honey bees (33)] were rinsed in nuclease-free water and frozen in 1 ml of TRIzol reagent (Invitrogen).

Quantification of Vitellogenin Titer. Individual samples of 0.75 μlof hemolymph protein were subjected to one-dimensional SDS electrophoresis by using 7% polyacrylamide gels. Vitellogenin was identified as a single band of 180 kDa (37). The vitellogenin titers were determined by the method of Lin et al. (30) using a β-galactosidase standard (Sigma).

Quantification of Vitellogenin mRNA. Total RNA was extracted from the abdomen by using the first steps of the TRIzol reagent protocol (Invitrogen). After the chloroform extraction, the water phase was mixed with an equal volume of 70% ethanol. From this mixture, RNA was purified by using the RNeasy minikit (Qiagen, Valencia, CA), which includes a DNA-digestion step with RNase-free DNase (Qiagen).

Primers for vitellogenin (5′-GTTGGAGAGCAACATGCAGA-3′ and 5′-TCGATCCATTCCTTGATGGT-3′) and actin (5′-TGCCAACACTGTCCTTTCTG-3′ and 5′-AGAATTGACCCACCAATCCA-3′) were designed by using oligo primer analysis software (Molecular Biology Insights, Cascade, CO). The amplicons were 150 and 149 bp, respectively. For each individual bee, 10 ng of total RNA was analyzed by real-time RT-PCR using the QuantiTect SYBR green RT-PCR kit (Qiagen) according to manufacturer instructions. The reaction was run in triplicate for each sample. The assay was carried out in an ABI Prism 7000 with ABI Prism 7000 sds 1.0 software (Applied Biosystems). To verify that the SYBR green dye detected only the intended PCR product, all reactions were subject to the heat-dissociation protocol following the final cycle of the PCR. After calculating the mean cycle threshold values from the triplicate readings, individual vitellogenin mRNA levels were quantified by the comparative cycle threshold method (38) as a relative quantity. Actin was used as active reference. The target and reference systems had equal efficiencies.

Data Analysis. Two-factor ANOVA was used to detect potential effects of genotype and age on the titer and transcription level of vitellogenin. Fisher's least significant difference test was used to test the significance of post hoc effects. Putative differences in the correlative relationships between the titer and the mRNA level of vitellogenin were analyzed by two-tailed t tests on the Pearson correlation coefficients. Polynomial regression followed by a likelihood ratio test was used to analyze higher-order associations. Calculations were performed with statistica 6.0 (StatSoft, Tulsa, OK).

Results

Vitellogenin Titer. The vitellogenin titers of high- and low-strain bees are significantly different [ANOVA: F_{(1, 90)} = 26.08, P < 0.005 (Fig. 1A); F_{(1, 89)} = 7.97, P = 0.006 (Fig. 1B); and F_{(1, 91)} = 34.89, P < 0.005 (Fig. 1C)]. There is also a significant effect of age on the vitellogenin level [ANOVA: F_{(2, 90)} = 23.46, P < 0.005 (Fig. 1 A); F_{(2, 89)} = 23.08, P < 0.005 (Fig. 1B); and F_{(2, 91)} = 15.41, P < 0.005 (Fig. 1C)]. No interaction is detected between strain and age in two colonies [ANOVA: F_{(2, 90)} = 0.03, P = 0.98 (Fig. 1 A); and F_{(2, 89)} = 0.11, P = 0.90 (Fig. 1B)], but a significant interaction is apparent in the third [ANOVA: F_{(2, 91)} = 3.83, P = 0.025 (Fig. 1C)]. Here, the mean vitellogenin titers of high-strain bees show a continuous increase, whereas they drop in low-strain workers between 5 and 10 days of age (Fig. 1C).

Fig. 1.

Mean vitellogenin protein levels in the hemolymph of 0-, 5-, and 10-day-old workers from colonies selected for high and low pollen hoarding. The high- and low-strain bees are represented by gray and white bars, respectively. The titers are calculated relative to a β-galactosidase standard. Shown are the results of separate experimental setups in which worker bees from the two strains were introduced into three wild-type “host” colonies. Lines indicating standard errors are represented on top of each bar. Significant differences within age cohorts are indicated with asterisks. The cut-off P value is 0.05 by a Fisher's post hoc test.

The difference between the two strains is obvious at emergence. For newly emerged bees, the high-strain workers have more hemolymph vitellogenin than low-strain bees in two of three samples (Fig. 1 A and C). In the third colony setup, the difference is significant according to a simple Student t test (P < 0.005, df = 18), but the strains do not differ in the factorial ANOVA post hoc test (Fig. 1B). At 5 days of age, the high-strain bees have more hemolymph vitellogenin than the low-strain bees in all the experimental colonies (Fig. 1). For 10-day-old bees, the difference between the strains is not significant in one colony (Fig. 1B; P = 0.40 and df = 32 in a simple Student t test). The data set has a particularly high variance in this case (Fig. 1B).

Vitellogenin mRNA. Vitellogenin transcription levels were not determined separately for the experimental colonies described above. The effect of genotype was qualitatively similar in the three hosts (Fig. 1), and reducing the data set while conserving between-host variation seemed reasonable. Therefore, RNA samples from all three colonies were picked at random to form a single data set for each strain.

The vitellogenin mRNA levels of the high- and low-strain bees are significantly different [ANOVA: F_{(1, 63)} = 25.83, P < 0.005]. The data show an effect of age [ANOVA: F_{(2, 63)} = 42.08, P < 0.005], but no interaction effect between strain and age is detected for vitellogenin mRNA [ANOVA: F_{(2, 63)} = 2.43, P < 0.10].

The difference between the two strains is observable in newly emerged bees (Fig. 2). A simple Student t test confirms a significant deviation between the strains at this age (P < 0.005, df = 14), but the difference is not established by the factorial ANOVA post hoc test (P = 0.28, df = 63). At 5 and 10 days of age, the high-strain workers have higher vitellogenin mRNA levels compared with those of the low-strain bees (Fig. 2).

Fig. 2.

Mean vitellogenin mRNA levels in the abdomen of 0-, 5-, and 10-day-old high- and low-strain bees. The data from the two strains are represented by gray and white bars, respectively. The mRNA levels are measured as relative quantities (RQ) (see Materials and Methods for information). Lines indicating standard errors are represented on top of each bar. Significant differences within age cohorts are indicated with asterisks. The cut-off P value is 0.05 by a Fisher's post hoc test.

Hemolymph Titer Versus Transcription Level. To learn more about the relationship between the titer and transcription level of vitellogenin, we compared individual measurements of both variables (Fig. 3). Extreme values may contain important information about associations in data (39), and we therefore included a subset of samples that was chosen deliberately to increase the range of values for the two variables. An analysis of the reduced data set gave results similar to those presented below, with the main difference being a lower but significant P value for the correlative comparison of 5-day-old bees (t₁₆ = 2.01, P < 0.04).

Fig. 3.

Vitellogenin titer plotted against vitellogenin mRNA level. High-strain workers are represented by gray circles, and low-strain workers are represented by white circles. The titers are calculated relative to a β-galactosidase standard, and the mRNA levels are measured as relative quantities (RQ). Each data point is a sample obtained from an individual bee. Shown are data distributions from 0-, 5-, and 10-day-old bees, respectively. The first-order regressions of the mRNA levels on the titers are depicted by dashed and full lines for the high and low pollen-hoarding strains, respectively. In C, the two high-strain samples with titers of >17 μg/μl are not included in the estimation of the regression line.

In newly emerged bees, there is a proportional relationship between the titer and transcription rate of vitellogenin (Fig. 3A). A high vitellogenin titer is associated with a high transcription level in both strains (r_H = 0.82, df = 7; r_L = 0.78, df = 7), and the correlative relationship is not different between the two genotypes (t₁₄ = 0.20, P = 0.86). In 5-day-old bees, however, the dynamics are different (Fig. 3B). The association between titer and transcription level is positive in the high strain (r_H = 0.83, df = 9) and negative in the low strain (r_L = –0.37, df = 10). Additionally, the correlative relationship is significantly different (t₁₉ = 5.63, P < 0.005), which is also the case in 10-day-old bees when the samples with the most hemolymph vitellogenin (>17 μg/μl; Fig. 3C) are excluded from the analysis of the high strain (r_H = 0.81, df = 11; r_L = –0.54, df = 13; t₂₄ = 29.16, P < 0.005). If the samples are included, the titer and transcription rate of vitellogenin are not correlated in the high-strain bees (r_H = 0.05, t₁₂ = 0.18, P = 0.86), and the correlative relationship is not different between the strains (t₂₆ = 1.15, P = 0.14). Clearly, the full data set does not fit a linear model. The two samples in question may be noninformative outliers. Alternatively, they suggest that the relationship between the two variables is nonlinear. When the data from the three age cohorts are pooled within genotype, a significant second-order association between titer and transcription rate is apparent in both the high (χ² = 17.64, P < 0.005, n = 32) and low (χ² = 4.96, P < 0.03, n = 33) strain. This result suggests that the association between the titer and transcription rate of vitellogenin changes from positive to negative when the titer exceeds a certain level or when the bee reaches a certain age.

Discussion

Vitellogenin Dynamics in Bees from Selected Pollen-Hoarding Strains. Our results demonstrate that the hemolymph titer and the transcription level of the yolk protein vitellogenin are significantly different in bees selected for high and low pollen hoarding. The general development of vitellogenin synthesis, with titers increasing the first 7–10 days of adult life (29), is confirmed in our findings. Throughout this period, the high-strain bees have higher vitellogenin titers compared with workers from the low strain. The difference between the strains is not significant in one sample (Fig. 1B). However, the distribution of the vitellogenin titer at a given age is strongly influenced by the age at onset of foraging (28), the amount of brood (29), the availability of pollen (40), and the presence of parasites (41). The difference between the high and low pollen-hoarding strains is unambiguous, although no measures were taken to control host-colony variation in these factors. This finding suggests that the effect of genotype is robust.

The vitellogenin transcription levels in the two strains show the same general trend as the titers of vitellogenin (Fig. 1 versus Fig. 2), which indicates that the higher hemolymph titers in the high-strain bees result from higher rates of vitellogenin synthesis. However, the complex associations between the titer and the transcription level suggest that the causal relationship is not derived from a simple genetic limitation on the maximum transcription rate of vitellogenin. It seems clear that a regulatory mechanism controls the association, and it can reverse the relationship between the titer and the transcription level within days. An inverse relationship between titer and transcription level suggests that the transcription rate is reduced when the concentration of vitellogenin reaches a certain level. A theoretical outline of this self-governing regulatory feedback loop was developed by Amdam and Omholt (42) to explain the vitellogenin dynamics of wild-type honey bee workers. Within their framework, the dynamic differences between the two strains would result from variation in the transcriptional response threshold to the hemolymph vitellogenin concentration. As a result, the transcription of vitellogenin in low-strain bees is down-regulated at a lower vitellogenin titer compared with that of high-strain workers. Alternatively, the observed shift in the relationship between the titer and the transcription level is governed by an age- or time-correlated phenomenon. A longer half-life of vitellogenin in the low-strain bees or differential tuning of a machinery that influences mRNA stability (43) might be possible mechanisms. However, our results cannot distinguish between these alternative hypotheses.

Interpretation Within the Suite of Traits That Respond to Selection on Pollen Hoarding. The fact that the strains differ in their titer and transcription level of vitellogenin at emergence parallels previous findings related to physiological and behavioral characteristics such as the hemolymph titer of juvenile hormone (JH) (D. J. Schulz, G. E. Robinson, and R.E.P., unpublished data), the level of protein kinases A and C (44), plus the transcription level of tyramine receptor (AmTyr1) in the central brain (R.E.P., unpublished data), locomotor activity (R.E.P., unpublished data), and response threshold to sucrose (11, 12, 17). Our findings therefore add to the accumulating evidence that differences between the two pollen-hoarding strains are apparent early in life and correlate with foraging behavior 2–3 weeks later. Our results suggest that the transcription of vitellogenin either starts earlier or increases faster in the high-strain pupae because the production of vitellogenin is initiated during the late pupal stages (45). Both scenarios imply that the strains diverge physiologically at least 60 h before emergence, which is when vitellogenin first can be detected in pupal hemolymph (45, 46). Given the association between early-life physiology and foraging behavior outlined above, this explanation further suggests that the foraging behavior of a honey bee is established before it emerges as an adult.

The onset of vitellogenin synthesis in female honey bee pupae is governed by a slight increase in the endogenous JH level as the ecdysteroid titer declines before emergence (46). In queen pupae, the JH titers increase earlier, decrease later, and reach higher levels compared with workers (46). The vitellogenin transcription rates are subsequently higher in adult queens, but the differential hormonal signatures in female pupae have not been causally linked to variation in adult transcription levels of vitellogenin. However, the fact that the JH titer is higher in newly emerged high-strain bees (D. J. Schulz, G. E. Robinson, and R.E.P., unpublished data) may imply that the hormonal dynamics responsible for the initiation of vitellogenin synthesis differ in the two strains.

JH and ecdysteroids are the key regulatory hormones that control vitellogenesis in a broad range of insect species (47–51). In addition to their gonotrophic role, these hormones govern shifts between behavioral states in the reproductive cycle, e.g., by controlling the shift from nectar to blood-host foraging in female Culex nigripalpus mosquitoes (52), the shift from feeding and sexual behavior to fasting and parental activity in the earwig Labidura riparia (53), the initiation of oviposition in crickets (54), the initiation of sexual behavior of male Agrotis ipsilon moths (55), and shifts between periods of intense flight activity and reproductive behavior in several taxa (34). JH and ecdysteroids affect adult behavior by regulating the growth and central processing of sensory and motor neurons (26, 56–59), and their coordinated regulatory modes have been shown to result in synchronized changes in sensory perception, locomotor activity, and reproductive physiology (26, 60).

In honey bees, both JH and ecdysone appear to have lost their gonotrophic roles in the adult queens and workers (61). JH is not needed for the behavioral shift to foraging (62) or for the growth of the mushroom body neuropil, a brain region involved in learning and memory that increases in size as the bee gets older (63). JH is responsible, however, for a dramatic down-regulation of vitellogenin synthesis in adult workers (33). This negative action on vitellogenin production seems contrary to the norm in solitary insects with a prolonged adult reproductive period (35), yet the gonotrophic role of JH in honey bee pupae may imply that the positive regulatory action seen in solitary adults has been shifted in time to accommodate adaptive production rates of vitellogenin early in life. The initiation of vitellogenin synthesis before emergence may have been instrumental for enhancing the reproductive output of young queens. At the same time, it would provide young bees with a protein source for the production of brood food (64).

If the gonotrophic action of JH has shifted to the pupal stage, then pleiotropic effects on sensory and motor tuning may have shifted as well. In this case, variation in the JH action cascade could explain why the two pollen-hoarding strains show differential sensory tuning and locomotor activity at emergence. The levels of protein kinases A and C play roles in sensory signaling and learning and may be part of such a pleiotropic network. Additionally, the observed differences in AmTyr1 mRNA levels (R.E.P., unpublished data) make sense, because the tyramine signaling pathway seems to be involved in the reproductive tuning of queenless honey bee workers (65). These putative associations, however, do not explain why gonotrophic circuits would respond to selection on pollen hoarding and mediate differential foraging behavior in adult bees.

Interpretation Within the Framework of the Reproductive Ground Plan. In honey bee larvae, a nutrition-dependent JH pulse initiates separate developmental programs for queens and workers (reviewed by Pinto et al. in ref. 66). The mature queen shows an exaggerated reproductive phenotype, whereas workers display a temporal division of labor (reviewed by Winston in ref. 67). Here, young bees (or “hive bees”) perform tasks within the brood nest, and older bees forage for nectar and pollen. The physiological and behavioral machinery required for oviposition is underdeveloped or inactive in workers, but they maintain some reproductive potential while they synthesize yolk protein as hive bees (30, 32). The confinement of this potential to the hive bee stage and the subsequent down-regulation of vitellogenin in foragers may imply that task-associated division of labor in honey bees has evolved through a temporal compartmentalization of cyclic female functions. In this case, the temporal worker castes would derive from states in an inherent reproductive ground plan (7, 8).

The early endocrine switch that separates the developmental trajectories of queens and workers has probably been instrumental for minimizing correlations between the adult morphs (9). Temporal expression of reproductive phases, on the other hand, is likely to be characterized by a much higher degree of interdependency. Explicitly, if the hive bee and forager phases are derived from a common reproductive ground plan, the tuning of the underlying regulatory circuits would be expected to affect both phenotypes. In mosquitoes and black flies (68), reproductive tuning and foraging behavior (nectar or blood meal) are linked (52, 69), because morphogenetic hormones control foraging decisions by modulating sensory perception in association with the reproductive physiology of the insects. With the data currently available, a tentative relationship between reproductive tuning and foraging behavior in bees (i.e., solitary and social) is suggested but not directly supported. In the absence of contrary evidence, however, we suggest that the relationship is similar to the associations found in mosquitoes and flies. In this case, gonoinactive bees would preferentially feed on nectar, whereas individuals tuned for reproduction would hoard ample pollen to provide for brood.

If we partition and extend these associations to social life, the selection response and coupling of traits in bees from high- and low-pollen strains make sense: the regulatory circuits of the reproductive ground plan respond to selection on pollen hoarding through a pleiotropic network involving sensory perception, reproductive potential, and foraging behavior. In this case, nectar foragers would be gonoinactive individuals neurologically tuned for self-maintenance and nectar feeding, whereas pollen would be preferentially hoarded by former gonoactive individuals tuned for reproduction.

Implications, Predictions, and Future Work. By offering a set of regulatory circuits that can serve as building blocks for social life, the reproductive ground plan may be a fundamental director of social evolution in insects (7, 8). Reproductive tuning influences behavioral phenotypes over individual lifetimes, whereas new phenotypes emerge in evolutionary time through developmental or temporal compartmentalization of reproductive states. In this way, complex eusocial societies can emerge from a single genomic template present in solitary ancestors.

The explanatory framework of the reproductive ground plan is a theoretical model derived from indirect evidence. However, the same relationship between sensory perception, reproductive potential, and foraging behavior emerge between honey bee subspecies (70), which suggests that the associations outlined in this article are sound and that the selected strains can be used to elucidate the explanatory potential of the theoretical framework. It is reasonable to predict that the high-strain bees with the more reproductively tuned phenotype should develop their ovaries faster and produce more eggs in the absence of the queen. The association between reproductive tuning and foraging behavior can also be tested at the individual level in wild-type as well as selected bees.

If the ground plan is a major director of social evolution, the correlations of sensory perception, reproductive tuning, and foraging behavior observed in honey bees are linked in a broad range of social insects. Intriguingly, the predicted relationship between sensory perception and reproductive tuning seems to appear in the queenless ant Streblognathus peetersi (71, 72). We believe that future research on these associations will be of fundamental importance to our understanding of social evolution in insects.