Support for the reproductive ground plan hypothesis in a solitary bee: links between sucrose response and reproductive status

Abstract

In social bees, foraging behaviour is correlated with reproductive status and sucrose sensitivity via endocrine pathways. This association led to the hypothesis that division of labour in social insect societies is derived from an ancestral ground plan that functions to synchronize dietary preferences with reproductive needs in solitary insects. However, the relationship between these traits is unknown for solitary bees, which represent the ancestral state of social bees. We used the proboscis extension response assay to measure sucrose response among reproductive females of the solitary alkali bee (Nomia melanderi) as a function of acute juvenile hormone (JH) treatments and reproductive physiology. We also tested long-term effects of JH on reproductive development in newly emerged females. JH did not have short-term effects on reproductive physiology or sucrose response, but did have significant long-term effects on ovary and Dufour's gland development. Dufour's gland size, not ovary development, was a significant predictor of sucrose response. This provides support for the reproductive ground plan hypothesis, because the Dufour's gland has conserved reproductive functions in bees. Differing results from this study and honeybees suggest independent origins of division of labour may have evolved via co-option of different components of a conserved ground plan.

1. Introduction

Bees require floral resources to complete reproductive maturation, to sustain their flight and nest building activities and to provision their offspring. The dietary needs and mechanisms that regulate variation in nectar and pollen preferences have been well documented for honeybees (Apis mellifera) [1–6] and bumblebees (Bombus sp.) [7–9], but almost nothing is known of the nutritional preferences of solitary bees or the mechanisms underlying these preferences. This is important, because one of the major causes of global bee decline is reduced availability of floral resources [10–12]. The vast majority of the roughly 20 000 species of bees do not live in colonies like those of honeybees and bumblebees, and are thus likely to have different nutritional requirements [13].

In honeybees, dietary preference is influenced by a complex set of factors. Among foragers, those who preferentially collect pollen have lower response thresholds to sucrose than do those that preferentially collect nectar [14,15]. These preferences are relatively consistent throughout the lifespan of an individual bee [16–18], and are based, in part, on genetic variation [5,14,19–21]. Detailed experiments with wild-type honeybees, as well as artificial selection of colonies based on pollen stores, have revealed a suite of traits correlated with foraging behaviour that includes reproductive physiology, gene regulatory networks and response to sucrose. Honeybee workers with larger ovaries and higher levels of vitellogenin (a yolk precursor protein) are also more responsive to sucrose and are more likely to be pollen foragers [22–28]. This relationship involves endocrine systems such as the juvenile hormone (JH) and ecdysone pathways [6,29,30], and is further influenced by genes in nutrition-associated signalling pathways [31,32].

The link between this diverse set of traits gave rise to the hypothesis that division of labour in highly social insects like honeybees is derived from ancestrally conserved reproductive regulatory networks. The cornerstone of this ‘reproductive ground plan hypothesis’ is the premise that sensory responses, feeding behaviour and female reproductive cycles are co-regulated via endocrine pathways in the solitary ancestors that gave rise to division of labour [26]. The proposed function of such an integrated signalling network is to bias female dietary preference to those nutrients most essential during specific phases of the reproductive cycle (e.g. a preference for protein during oogenesis). However, whether such a relationship exists in solitary insects that closely resemble the ancestors of social insects is completely unknown.

We aimed to test the reproductive ground plan hypothesis in the solitary alkali bee, Nomia melanderi Cockerell (Hymenoptera: Halictidae). Alkali bees are native to arid regions of the western USA, where they are semi-managed for pollination of alfalfa seed, though they are generalist foragers [33]. Alkali bees are solitary, but nest gregariously in dense aggregations [33]. Alkali bees belong to the subfamily Nomiinae, which is basal to the Halictinae, in which eusociality has evolved at least twice [34,35]. Thus, they are representative of the solitary ancestral state from which division of labour evolved, and are well-suited for testing the hypothesized role of reproductive signalling in sucrose responsiveness.

The relationship between reproductive status and sucrose response can be tested with the proboscis extension response (PER) [36]. In this experimental procedure, immobilized bees are presented with increasing concentrations of sucrose solutions, and response to each concentration is determined based on whether the proboscis is extended in response to the stimulus. While PER has been used effectively to describe the appetitive behaviour of several social bee species, it has been suggested that solitary bees do not elicit a PER to sucrose [37]. These earlier attempts to elicit PER from solitary bees relied exclusively on stimulation of the antennae, but bees also have gustatory and olfactory sensilla on the mouthparts and the tarsi of the forelegs [38–41]. We targeted these alternative gustatory sensilla to measure sucrose response in alkali bees.

We also aimed to describe the short-term and long-term effects of JH on solitary bee reproductive physiology. JH is a highly conserved gonadotropin responsive to nutrient signalling pathways [42,43], and also plays a key role in the worker division of labour that the reproductive ground plan hypothesis aims to explain [6]. However, the effects of this hormone on either sucrose response or acute changes in reproductive physiology have not been tested in a solitary bee. Long-term effects of JH on reproductive development are only known for one solitary bee species [44]. We sought to fill these gaps.

Similar studies with honeybees have assessed reproductive status by measuring ovary size, but reproductive status may also be reflected in the size of the Dufour's gland. The Dufour's gland is a highly conserved exocrine gland with a diverse set of functions, including secretion of chemicals used in oogenesis and egg laying, nest building, larval food and pheromones [45]. We therefore included measurements of this gland in our assessment of reproductive physiology.

2. Material and methods

(a). Study location and animals

This study was carried out between 18–26 June 2015 (experiment 1) and 27 May–8 June 2016 (experiment 2) in Touchet, WA, USA. For experiment 1, we obtained adult female alkali bees from a single, large (approx. 100 × 50 m), well-established bee bed. Alkali bees emerge from winter diapause in early summer, and trapping efforts on this bed between 5 and 14 June 2015 indicated that emergences had finished by mid-June. Thus, all of the females emerging from holes on this bed were probably reproductive and actively nesting. Reproductive status was confirmed during dissection by observation of follicular relics of laid or resorbed oocytes called yellow bodies [46]. For experiment 2, we obtained newly emerged bees from three large, well-established bee beds in 2016 by placing emergence traps over undisturbed surfaces, which indicated that emergence had not yet begun in that area. We checked these traps at least three times a day to collect bees as they emerged from the ground after over-wintering.

(b). Experimental procedure

(i). Experiment 1

We placed emergence traps over dense clusters of nesting holes early in the morning, before the bees had emerged to begin foraging. Thus, our experiments presented bees with the first opportunity to feed for the day, after a natural overnight fasting period. Bees were gently coaxed into 15 ml conical tubes upon emergence, and the tube was placed in a cooler with an ice pack for transport to the laboratory.

Upon returning to the laboratory, we chilled each bee at 4°C for 5 min, and then harnessed it by securing a thin filament of cotton twine around the thread-waist, and attaching it to a plastic drinking straw, following the previously used ‘belt method’ [37]. The plastic straws had been cut in half lengthwise, such that harnessed bees were facing forward in a vertical position (electronic supplementary material, figure S1).

We randomly assigned each bee to one of three treatments: JH, solvent control, sham. JH-III (product J2000, Sigma-Aldrich, St Louis, MO, USA) was dissolved in dimethylformamide (DMF) (Fisher Scientific, Fair Lawn, NJ, USA) at a concentration of 100 µg µl⁻¹. This dose was chosen because it has significant effects on ovary size in bumblebees (Bombus terrestris), which are slightly larger than N. melanderi [47]. We applied the treatments with a pipette tip to the thorax of each bee while they were secured in their harnesses. Bees in the JH and solvent groups received 1 µl of JH-III dissolved in DMF or DMF only, respectively. Bees in the sham treatment group were touched lightly with a clean pipette tip. The bees were then left to rest at 23–26°C for 4 h. This time period was chosen to give bees ample time to recover from handling stress, and is typical of PER assays performed with honeybees [36,48]. The mean ± 1 s.d. time between capture and treatment was 2.04 ± 0.02 h. Alkali bees cease foraging activity by 19.00 [49], and do not store food in their nests [50]. All bees had thus experienced a natural starvation period of at least 19 h at the time of testing.

The sucrose response of each bee was tested following established protocols [2,14], with the exception that we targeted our stimulation to the gustatory sensilla on the mouthparts (i.e. the galea on the maxillae and the labial palps) while they were folded against the head in the resting position (see fig. 3 in [38] for detailed description of the mouthparts) and the foretarsi. The antennae are a more common target of stimulation in honeybees, but stimulation of tarsi and mouthparts have also been previously used [48]. Bees must fully extend the glossa from inside the rest of the labium to ingest food [38]. Thus, stimulating the mouthparts does not allow the bees to ingest any sucrose solution without a full extension of the proboscis. Bees were presented with a toothpick saturated with each of seven concentrations of sucrose solution (0.1, 0.3, 1, 3, 10, 30 and 50% w/v) for a 60 s period, and their response was recorded. Bees that responded in this way were able to lick the toothpick briefly. Bees were presented with water (0%) for 60 s between each concentration of sucrose solution to control the potential for sensitization or habituation that could occur with repeated exposure to sucrose. Each concentration was presented to each bee before moving on to the presentation of water. The mean amount of time it took to complete a round of testing was 12 min. Thus, an average of 12 min passed between each presentation of sucrose solution and the following presentation of water, and an average of 24 min passed between each presentation of sucrose solution and the next presentation of sucrose. A response was recognized as a full extension of the proboscis, including the glossa. Partial proboscis extensions were noted, but were not counted as a response. (We re-ran our analyses including partial extensions as responses to determine what effect this decision had—see results.) The investigators administering the PER test were blind to the JH treatment group of each bee. Upon completion of the assay, bees were placed in individually labelled tubes and flash-frozen in liquid nitrogen. Samples were stored in liquid nitrogen until return to Utah State University, where they were transferred to a −80°C freezer.

A gustatory response score (GRS) was calculated for each bee as the total number of responses to sucrose recorded for each bee. A high GRS (7) indicates high responsivity to sucrose, and a low GRS (0) indicates low responsivity to sucrose. The lowest response-eliciting concentration (LREC) was defined as the lowest sucrose concentration to which each bee elicited a PER. An LREC of 100 was assigned to bees that did not respond to any of the sucrose solutions.

(ii). Experiment 2

Newly emerged females were collected from emergence traps that had been placed over undisturbed surfaces by gently coaxing them into 15 ml conical tubes, and the tube was placed in a cooler with an ice pack for transport to the laboratory. Upon returning to the laboratory, bees were treated with JH-III (product E589400, Toronto Research Chemicals, Inc., Toronto, Ontario, Canada), DMF or sham following the same methods as described for experiment 1. The treatment procedure was repeated 5 days later.

After treatment, bees were placed in perforated plastic deli containers (72 mm h × 90 mm lower diameter × 113 mm upper diameter). They were fed a mixture of sterilized sucrose and pollen: 30 ml of 35% (w/v) sucrose mixed with 2.5 g of finely ground honeybee pollen (Betterbee, Greenwich, NY, USA), and food was replaced daily. Cages were kept at 22–28°C, 40–85% RH and 13 L : 11 D dark with full spectrum lighting. When bees reached 10 days of age, they were chilled at 4°C for 5 min, placed in individually labelled tubes and flash-frozen in liquid nitrogen. Samples were stored in liquid nitrogen until return to Utah State University, where they were transferred to a −80°C freezer.

(c). Dissections

All dissections were performed under a Leica M80 stereomicroscope fitted with an IC80HD camera (Leica Microsystems, Buffalo Grove, IL, USA). Observers were blind to the treatment group or GRS/LREC of each bee at the time of dissection, and measurements were agreed upon by both authors. Measurements were made directly from the calibrated images using software in the Leica Application Suite (v. 4.5). We measured Dufour's gland length and the longest terminal oocyte in the left and right ovaries. The maximum longest terminal oocyte was identified by comparing the left and right terminal oocytes in each bee. Oocytes showing evidence of being resorbed (e.g. opaque white and malformed) were not included in experiment 1. We also measured intertegular width as an indicator of body size and we measured wing wear as an indicator of age for bees in experiment 1. Wing wear was scored separately for each wing according to the following scale: 1, no wear; 2, minimum amount of nicks on edge of wing (nicks less than 0.4 mm): edge still pretty much intact; 3, many nicks on edge of wing (nicks ranging from 0.1 –0.7 mm): edge has ‘feathered’ appearance; 4, maximum amount of nicks such that some parts of edge are no longer there (but less than half wing edge gone); and 5, wing edge gone and large cuts (more than half wing edge gone). Left and right wing wear scores were averaged for each bee. Yellow bodies were noted on the ovaries in all but four bees in experiment 1, indicating most were reproductively active. For bees in experiment 2, we examined the contents of the spermatheca under a compound microscope to identify females that mated in the emergence traps prior to the collection. Females were excluded from analysis if sperm was detected (n = 2) or if the spermatheca was damaged during dissection (n = 1).

(d). Statistical analyses

All statistical analyses were performed in Stata v. 9.2 (StataCorp LP, College Station, TX, USA).

(i). Experiment 1

Ladder of powers analyses revealed no significant deviation from normality in either maximum longest terminal oocyte or Dufour's gland length (maximum terminal oocyte: χ² = 3.95, p = 0.14; Dufour's gland: χ² = 2.49, p = 0.29) so we used parametric statistics on untransformed data. We measured the joint responses of Dufour's gland length and maximum oocyte length to the independent variables treatment (coded as a categorical variable), intertegular width, and wing wear (coded as a categorical variable) with MANOVA, which takes into account the inter-correlation between the two response variables. Intertegular width and wing wear were included in the model to account for variation in body size and age, respectively, that may contribute to variation in reproductive physiology. We used logistic regression to determine whether there were significant differences in the probability of response between treatment groups for each sucrose concentration, and used a Bonferroni correction for multiple testing.

GRS is a count of the number of responses in the sucrose response test, thus we ran several diagnostic tests (fitstat in Stata) to determine whether a Poisson or negative binomial regression was most appropriate for our data. Results revealed that the two models were similar fit, but the BIC and AIC were smaller for Poisson (AIC: Poisson, 4.09; neg. binomial, 4.12; BIC: Poisson, 3.98; neg. binomial, 0.40). The difference of 4.38 in the BIC between the two models conveys positive evidence of the Poisson as a better fit to the model [51]. We, therefore, used a Poisson regression to test predictor variables of GRS, including JH treatment group (as a categorical variable), Dufour's gland length, maximum oocyte length, intertegular width and PER observer (as a categorical variable). We also tested the effects of these independent variables on LREC in an ordered logistic regression.

(ii). Experiment 2

Ladder of powers analyses revealed no significant deviation from normality in either maximum longest terminal oocyte or Dufour's gland length (maximum terminal oocyte: χ² = 4.14, p = 0.13; Dufour's gland: χ² = 2.16, p = 0.34) so we used parametric statistics on untransformed data. We measured the joint responses of Dufour's gland length and maximum oocyte length to the independent variables treatment (coded as a categorical variable) and intertegular width with MANOVA, which takes into account the inter-correlation between the two response variables. Intertegular width was included in the model to account for variation in body size that may contribute to variation in reproductive physiology. MANOVA tests were followed by individual regressions to test the effects of treatment only (coded as a categorical variable) on oocyte and Dufour's gland development independently.

3. Results

(a). Short-term effects of juvenile hormone on reproductive females

Bees treated with JH or DMF prior to the PER assay (experiment 1) had higher mortality than those in the sham group, though this difference was not statistically significant (JH: 32%, DMF: 40%, sham: 18%; χ² = 4.89, p = 0.09, n = 41 JH, 43 DMF, 44 sham). JH does not have a short-term effect on reproductive physiology in reproductive alkali bee females. The maximum longest terminal oocyte and Dufour's gland size were not statistically different between treatment groups (MANOVA: F_20,136 = 1.15, p = 0.31, n = 80). These results were unchanged when we excluded the four females for which yellow bodies were not visible (MANOVA: F_20,128 = 1.11, p = 0.35, n = 76).

(b). Sucrose response in alkali bees

Across all treatment groups, a higher percentage of bees responded to higher concentrations of sucrose (electronic supplementary material, figure S2). There were no significant differences between treatment groups in the probability of a PER at any of the sucrose concentrations (logistic regression at each concentration: p > 0.10 before Bonferroni correction, n = 23–31 per group; electronic supplementary material, figure S2). Across all females (n = 81), the mean (±1 s.d.) GRS was 3.62 ± 2.13. Nearly 10% (n = 8) of bees did not respond to any concentration of sucrose, and more than 13% (n = 11) of bees responded to all sucrose concentrations. This reflects a large amount of variation in appetitive response.

(c). Reproductive status and sucrose response

Reproductive status was correlated with sucrose response in unexpected ways. Maximum oocyte length was not a significant predictor of GRS in alkali bees (p = 0.43; electronic supplementary material, table S1; figure 1a). JH treatments also did not have a significant effect on GRS (p = 0.72; electronic supplementary material, table S1). However, a different aspect of reproductive physiology, Dufour's gland size, was a significant predictor of GRS (p = 0.001; electronic supplementary material, table S1; figure 1b). Females with the longest Dufour's glands were the most responsive to sucrose. Identity of the investigator administering the PER test was also a significant predictor of GRS (p < 0.001; electronic supplementary material, table S1). These results were unchanged when the four females without notable yellow bodies were excluded from the analysis (electronic supplementary material, table S2). These results were also unchanged when we included partial extensions of the proboscis in the GRS (electronic supplementary material, table S3).

Figure 1.

Relationship between sucrose response (GRS) and reproductive physiology in reproductive females: (a) longest terminal oocyte length (p = 0.43, n = 80; electronic supplementary material, table S1); and (b) Dufour's gland length (p = 0.001, n = 81; electronic supplementary material, table S1).

Similarly, LREC was not significantly different between JH treatment groups (p = 0.26; electronic supplementary material, table S4), and did not vary with maximum oocyte length (p = 0.28; electronic supplementary material, table S4), but was significantly predicted by Dufour's gland size (p = 0.002; electronic supplementary material, table S4) and PER observer (p < 0.001; electronic supplementary material, table S4). Females with longer Dufour's glands had lower LRECs. This result was unchanged when we excluded the four females without visible yellow bodies (electronic supplementary material, table S5).

(d). Long-term effects of juvenile hormone on newly emerged, non-reproductive females

Newly emerged females treated with JH or DMF and reared in the laboratory for 10 days (experiment 2) had higher mortality than those in the sham group, though this difference was not statistically significant (JH: 53%, DMF: 57%, sham: 34%; χ² = 3.62, p = 0.16, n = 30 JH, 30 DMF, 32 sham). JH does have a long-term effect on reproductive physiology in young, non-reproductive female alkali bees. The maximum longest terminal oocyte and Dufour's gland size, measured jointly, were significantly larger in JH-treated females than in females in either control group (MANOVA: F_6,80 = 9.42, p < 0.001, n = 45). Follow-up regression analysis of each reproductive measurement revealed that the JH treatments led to a marginally significant increase in maximum oocyte length (overall model: F_2,42 = 2.88, p = 0.07; DMF versus sham: p = 0.56, sham versus JH: p = 0.06, DMF versus JH: p = 0.03, n = 45; figure 2a) and a highly significant increase in Dufour's gland length (overall model: F_2,42 = 30.87, p < 0.001; DMF versus sham: p = 0.37, sham versus JH: p < 0.001, DMF versus JH: p < 0.001, n = 45; figure 2b).

Figure 2.

Long-term effects of JH on reproductive development in newly emerged females: (a) longest terminal oocyte length; and (b) Dufour's gland length. Error bars represent 1 standard error of the mean; letters represent statistically significant differences between groups in a regression analysis (p < 0.05); n = 45 (sham, 20; DMF, 12; JH, 13).

4. Discussion

Gustatory perception has been largely unstudied in solitary bees, despite the critical insight that understanding variation in dietary needs and preferences can provide. To our knowledge, this is the first documentation of sucrose responsivity in a solitary bee. Our results show that while alkali bees generally prefer higher concentrations of sucrose, response to sucrose is variable among reproductive females. This is consistent with demonstrated preferences and variation in social bees [23,52].

Our results have important implications for agriculture, because alkali bees are key alfalfa pollinators [53]. Bees are facing global declines owing to myriad factors, including reduced availability of high-quality forage [10–12]. Furthermore, modern agricultural practices, such as irrigation and pesticide use, have been demonstrated to alter the nutritional composition of floral nectar and pollen, and subsequently influence bee preferences and visitation rates [54–57]. Understanding the dietary preferences of solitary bees, as well as the factors that influence these preferences, is thus a critical first step in any conservation or management plan for wild bees.

Previous reports have suggested that solitary bees may not demonstrate a PER to sucrose [37], but our results provide evidence to the contrary; more than 90% of the bees in our study responded to at least one concentration of sucrose solution. Future students of appetitive behaviour in solitary bees may consider stimulation of alternative gustatory sensilla, such as those on the mouthparts and tarsi. Indeed, some of the earliest work on sucrose response was performed by eliciting feeding responses from stimulation of structures other than the antennae, such as the tarsi [58,59] or the proboscis [60], and early work by von Frisch [61] demonstrated that the gustatory receptors on the mouthparts are responsible for discrimination between different types of sugars in honeybees [38]. In our experience, proboscis extension can be elicited by stimulation of the foreleg tarsi or mouthparts also in other solitary bees (e.g. Megachile rotundata, Melissodes sp., Eufriesea mexicana, K. M. Kapheim 2012, unpublished data).

By using the ‘belt’ method to restrain the bees in our study [37], bees were able to actively touch the sugar water stimulus with both their mouth parts and foretarsi (electronic supplementary material, figure S1). We were thus not able to ensure that each sensilla was equally stimulated. Some bees may have received more stimulation in one region of gustatory receptors, depending on the specific presentation technique of the PER administrator. This may explain the significant differences in GRS and LREC between PER observers if certain observers more consistently stimulated more or less sensitive gustatory sensilla. Importantly, observers were blind to the reproductive status and treatment group of the bees they were testing. The correlation between Dufour's gland size and sucrose response was detectable as highly significant, despite variation introduced by observer effects.

Our results provide partial support for a critical, but heretofore untested, premise of the reproductive ground plan hypothesis by demonstrating a link between some aspects of reproductive physiology and gustatory perception in a solitary bee species that closely resembles the presumed ancestor of social bees exhibiting division of labour. Specifically, we find that female alkali bees with enlarged Dufour's glands are significantly more responsive to sucrose than are females with small Dufour's glands. In basal groups of Aculeate insects, the Dufour's gland functions as an accessory reproductive gland, and its secretions probably lubricate the ovipositor for egg laying or form a protective casing over eggs [62–64]. In ground-nesting, mass-provisioning species, including alkali bees, the secretions of the Dufour's gland are used to line brood cells, to strengthen the nest entrance, or as an anti-microbial coating on larval provisions [65–71]. As such, the size and chemical profile of the Dufour's gland changes during the nesting season, after mating, and with foraging activity in several bee species, including other halictids [72–77]. Additionally, the composition of the Dufour's gland secretions is correlated with fat body size and ovarian development in other halictid bees [78]. This strongly suggests that changes in Dufour's gland size are a physiological reflection of reproductive cycling in alkali bees.

The relationship between sucrose response and reproductive signalling in honeybees has focused entirely on ovary size, and a link to Dufour's gland size has not been investigated. Functional differences of the Dufour's gland in honeybees and alkali bees may explain this knowledge gap. Honeybees nest in cavities, build their brood cells from wax, and mitigate microbial infestations behaviourally or with propolis [79]. Accordingly, the chemical profiles of Dufour's gland secretions of halictid bees and honeybees are very different [45,66,67,70,71,80], suggesting that the Dufour's gland may not play as important a role in reproduction in honeybees as in alkali bees.

Contrary to the original support for the reproductive ground plan hypothesis in honeybees, we did not observe a significant relationship between ovarian cycle and sucrose response in alkali bees. This may reflect variation in the ontogeny of ovary development in honeybees and alkali bees. In honeybees, the number of ovarioles in each ovary varies among individuals, but is fixed upon completion of development, and thus reflects a genetic difference in potential fecundity, rather than a real-time indicator of reproductive status. Individual variation in the number of ovarioles is the metric by which ovarian size is typically measured in relationship to the reproductive ground plan hypothesis in honeybees, because workers do not develop oocytes when in the presence of the queen [24,26]. Alkali bees, like most solitary bee species, always have three ovarioles in each ovary, with no individual variation. Measuring the longest terminal oocyte thus provides a real-time measure of reproductive status in these bees. Alkali bees can lay an egg each day [50], and more than one oocyte develops at the same time (i.e. one in each ovary; K. M. Kapheim 2015, personal observation). Thus, ovarian cycling in solitary bees is irregular and may be too short to influence short-term changes in sucrose response. Alternatively, changes in the size of the Dufour's gland associated with reproductive status may occur over longer periods of time, whereby correlated dietary changes may have a more pronounced impact on reproduction.

Our results reveal that topical application of JH does not have acute effects on reproductive physiology or sucrose response in reproductive alkali bees, but JH does have long-term effects on reproductive physiology when administered to females that have just completed development. In honeybees, the association between JH and gustatory perception is well established, but most previous studies have manipulated JH in newly emerged bees, and measured long-term effects [6,17,81]. Similar to our results, a single application of the JH analogue methoprene to 6-day-old honeybee foragers did not influence food preference [82]. It is possible that there is a critical period at adult emergence upon which JH can influence ovary activation. This suggests that we did not observe an effect of JH on reproductive physiology in reproductive alkali bees, because a single dose of JH or its analogues is not enough to affect short-term behaviour or physiology in reproductive adult bees.

JH is involved in many aspects of reproductive and behavioural division of labour in eusocial bees and ants [83,84], and thus plays a key role in the major hypotheses for how these features of social life evolve [26,85,86]. Despite this, the relationship between JH, behaviour and reproductive development is virtually unknown in solitary bees (but see [44]), and inferences about how these functions have changed over the course of social evolution have necessarily been based on comparisons to distantly related insect species with life histories that do not closely resemble those of the solitary insects that gave rise to eusociality [26,83]. Our finding that JH influences the development of the Dufour's gland and, to a lesser extent, the ovaries in young, non-reproductive solitary bees helps to fill this gap. The role of the Dufour's gland in social organization and behaviour is also becoming increasingly understood [77,87,88], and a recent study in bumblebees indicates that JH may influence the chemical composition of secretions produced by the Dufour's gland in social bees [89]. Our study supports the hypothesis that a JH-mediated role for the Dufour's gland in social communication could stem from an ancestral ground plan.

5. Conclusion

The reproductive ground plan hypothesis was proposed to explain the evolutionary origin of division of labour in social insects. This is based on observations that foraging behaviour, sensitivity to sucrose and reproductive physiology are co-regulated with endocrine pathways in honeybees. Our results demonstrate that PER can be used to test this hypothesis in solitary bees, and that the sucrose perception of alkali bees is highly variable, though higher concentrations of sucrose tend to elicit a greater response. Our experiment provides evidence in support of the reproductive ground plan hypothesis by showing that Dufour's gland size is a significant predictor of sucrose response, though we also find that this relationship cannot be modulated with acute treatments of JH. Also consistent with the reproductive ground plan hypothesis, we find that reproductive development is influenced by JH, when applied to non-reproductive females that have just completed development. Division of labour and colonial living evolved independently in honeybees and halictid relatives of alkali bees. It is therefore not surprising that a presumed ancestral link between sucrose response and reproductive status would be co-opted in different ways over independent evolutionary trajectories, with potential involvement of the Dufour's gland in halictid bees and ovaries in the lineage leading to honeybees. Additional research designed to explore the link between division of labour, Dufour's gland development and sucrose response in social halictids closely related to alkali bees will be needed to understand the mechanisms by which this relationship may have been co-opted for division of labour. Overall, these results provide additional support for the assertion that independent origins of social behaviour evolve via convergent processes, but lineage-specific pathways [90].

Data accessibility

The data used in this manuscript are provided as supplementary tables (electronic supplementary material, tables S6 and S7).

Authors' contributions

K.M.K. designed the study, conducted the experiments, analysed the data and wrote the manuscript. M.M.J. contributed to data acquisition, analysis and edited the manuscript. Both authors approved of the submission of the final manuscript.

Competing interests

We have no competing interests.

Funding

This work was supported by funding from the Utah Agricultural Experiment Station (Project 1297), United States Department of Agriculture Agricultural Research Service Alfalfa Pollinator Research Initiative and Utah State University.