Juvenile hormone, but not nutrition or social cues, affects reproductive maturation in solitary alkali bees (Nomia melanderi)

Complex social organization, such as that observed among honey bees (Apis mellifera), ants (Formicidae) and vespid wasps (Vespidae), is marked by a high degree of variance in reproductive activity among individuals within a colony. This variation is demarcated among reproductive castes, whereby workers do not reproduce, despite engaging in maternal behaviors (e.g. brood feeding or nest defense), and queens reproduce while largely refraining from brood care (Michener, 1974). Workers of many social insect species have similar reproductive anatomy to queens (e.g. ovaries, spermathecae, glands, ovipositor), yet remain functionally sterile. This suggests the factors that influence the function of these structures differ between queens and workers, and that understanding this variation may provide insights into the physiological basis for the origin of social insect castes.

The factors that differentially influence reproductive activation among social insect castes include nutritional, endocrine and social cues (Kapheim, 2017). For example, both queens and workers acting as nurses within a honey bee colony consume a protein-rich diet, but this protein contributes to egg production only in the queens (Winston, 1987). Similarly, treatment with the juvenile hormone (JH) analog methoprene leads to accelerated ovarian development in queen paper wasps (Polistes canadensis), but increases foraging activity in workers (Giray et al., 2005). Finally, social cues, such as aggression from the queen, can repress endocrine pathways, and thus ovary maturation, in worker bumble bees (Bombus impatiens) and social halictid bees (Megalopta genalis), but aggression directed from workers toward queens does not have the same effect (Kapheim et al., 2016; Padilla et al., 2016; Smith et al., 2009). Understanding how these differences in sensitivity to physiological and environmental cues arise among females and contribute to variation in reproductive activity is thus key to understanding the origins of social insect castes.

One approach toward this goal is to investigate how these factors contribute to variation in reproductive development in solitary species representative of the ancestors that gave rise to social castes. We conducted two experiments to determine how variation in nutrition, JH and the social environment influence reproductive development in a solitary bee that shares similarities with the ancestors of social bees. Alkali bees, Nomia melanderi (Cockerell 1906), are semi-managed, native pollinators of alfalfa seed crops that range throughout the western USA (Cane, 2008). This species belongs to a basal subfamily of Halictidae (Nomiinae) in which eusociality has never evolved, but they are closely related to the Halictinae, in which social behavior is highly variable (Michener, 1974, 2007) and eusociality evolved at least twice (Brady et al., 2006; Danforth et al., 2008; Gibbs et al., 2012). Bees in the family Halictidae shared a common ancestor with the Apidae (e.g. bumble bees, honey bees) approximately 115 mya, and may thus have important differences in reproductive physiology (Cardinal and Danforth, 2013). Alkali bees are considered solitary because each female has her own nest, but they often nest in close proximity to other females in large aggregations (Cane, 2008; Wcislo and Engel, 1996). Importantly, these bees exhibit extended maternal behavior, characteristic of that which was a necessary pre-adaptation to sociality (Batra, 1970; Batra and Bohart, 1969; Schwarz et al., 2007). As such, understanding the factors that influence reproductive physiology in this species can shed light on the physiological basis for social evolution.

Little is known about the factors that influence alkali bee reproductive development. However, a recent study demonstrated that JH accelerates reproductive maturation (Kapheim and Johnson, 2017). It was also recently documented that adult female alkali bees consume pollen on a daily basis (Cane et al., 2016). Pollen is the primary source of dietary protein and lipids for bees (Roulston and Cane, 2000), but whether pollen consumption is necessary for reproduction has not been experimentally tested. We investigated two aspects of reproductive physiology: oocyte growth, which requires proteins for egg yolk (Badisco et al., 2013), and maturation of the Dufour's gland, which secretes lipids used for nest cell construction (Cane, 1981). Our results reveal that vitellogenesis can occur rapidly among newly emerged females, even without mating or nesting opportunity. The initiation of oogenesis and Dufour's gland maturation does not require dietary protein, and females treated with JH were more likely to reach reproductive maturity. This response to JH was not affected by variation in the social environment (i.e. co-housing with an older, reproductive female). This provides important insight into the physiological foundation from which social insect castes evolved, as well as the reproductive physiology of an important pollinator.

This study took place in Touchet, WA, USA, where alkali bees nest in large soil beds near alfalfa seed fields (Cane, 2008). Alkali bees overwinter as prepupae in below-ground nests, and emerge as adults upon completion of development the following summer. We trapped newly eclosed adult females from 27 May to 8 June 2016 by placing emergence traps on three bee beds prior to the start of emergence. Traps were checked at least 3 times a day, and new bees were transferred to the laboratory in individually labeled 15 ml tubes placed inside a cooler with a single ice pack placed under a layer of cardboard.

Upon arrival at the laboratory, bees were chilled at 4°C for 5 min and randomly assigned to a treatment group: sugar water only (sterile 35% sucrose solution), sugar water with pollen (2.5 g sterile, finely ground, honeybee pollen in 30 ml of sterile 35% sucrose solution), and sugar water with pollen plus four sprigs of fresh, untripped alfalfa flowers. (Bees were observed manipulating these flowers to release pollen on a regular basis throughout the experiment.) Bees were placed in perforated plastic cylindrical containers (72 mm×90 mm lower diameter×113 mm upper diameter) and reared in the lab for 10 days. Sugar water/pollen/alfalfa mixes were changed daily. Pollen–sugar mixture was shaken vigorously before each feeding to achieve homogeneity, and then pipetted into feeding troughs made from 1.5 ml microcentrifuge tubes with the tapered tip removed. The cages were kept at 22–28°C, 40–85% relative humidity and full spectrum lighting 13 h light:11 h dark, as has been previously described (Kapheim and Johnson, 2017). At the end of the 10 day rearing period, bees were chilled for 3 min at 4°C, placed in individually labeled tubes, and flash-frozen in liquid nitrogen.

We also collected newly emerged females and reproductive females of unknown age for comparison with lab-reared females. Newly emerged females were collected from emergence traps as described above, but were flash-frozen immediately upon return to the laboratory. Reproductive females were identified as those returning to a nest hole with pollen on their hind legs. They were captured by net, and flash-frozen immediately upon return to the laboratory.

Newly emerged females were collected as in experiment 1. Each bee was randomly assigned to a treatment group: sham control, solvent control or JH. For JH treatments, JH-III (product E589400, Toronto Research Chemicals, Inc., Toronto, ON, Canada) was dissolved in dimethylformamide (DMF) at a concentration of 50 μg μl⁻¹. Bees in the solvent control group received 1 μl of DMF applied to the thorax with a pipette tip. Bees in the JH group received 50 μg JH in 1 μl of DMF applied to the thorax with a pipette tip. Bees in the sham group were touched lightly on the thorax with a clean pipette tip. Hormone treatments were repeated when bees were 5 days old. Bees in each treatment group were randomly assigned to be caged alone or with an older, reproductive female, defined as above. All bees were paint-marked on the dorsal abdomen with a uniquely colored Decocolor^© paint pen (Uchida of America Co., Torrance, CA, USA). All bees were reared in cages, and received 35% sugar water mixed with pollen and fresh alfalfa flowers for 10 days, as described for experiment 1.

Upon collection, all bees from both experiments were stored in liquid nitrogen until return to Utah State University, where they were transferred to a −80°C freezer.

Dissections followed previously reported methods (Kapheim and Johnson, 2017). Briefly, bees were dissected under a Leica M80 stereomicroscope fitted with an IC80HD camera (Leica Microsystems, Buffalo Grove, IL, USA). We measured Dufour's gland and terminal oocyte lengths from images, using software in the Leica Application Suite (v.4.5). The observer was blind to the treatment group of each bee during dissections, and the two authors concurred on measurements. Mating status was determined by examination of the spermatheca under a compound microscope. We excluded newly emerged females with sperm and reproductive females without sperm from further analyses.

Images were further analyzed to classify ovaries into stages of oocyte maturation. Like most halictid bees, alkali bees have three ovarioles in each of two ovaries. The following categories were modified from Duchateau and Velthuis (1989) and Oliveira et al. (2017). Stage I – the oocyte and associated trophocytes occupying a single egg chamber can be distinguished from one another, but the oocyte is much smaller than the trophocytes and is spherical in shape. Stage II – the oocyte is smaller than the trophocytes, and is cylindrical in shape rather than round. Stage III – the trophocytes and oocyte are similar in size, the oocyte occupying 45–55% of the length of the egg chamber. The oocyte is elongated. The trophocytes appear less opaque, and the oocyte appears more solid and full. It is during this stage that vitellogenesis is initiated. Stage IV – the oocyte is much longer than the trophocytes, and occupies more than 55% of the total egg chamber. The trophocytes appear translucent and small in the anterior portion of the egg chamber. Stage IVr (reabsorbing) – the oocyte has all the characteristics of stage IV, but is misshapen and yellow in color, indicating reabsorption. Stage V – the oocyte is large, robust and opaque with no associated trophocytes. Vitellogenesis is complete at this stage. Stage Vr (reabsorbing) – the oocyte has no trophocytes present, but has the characteristics of being reabsorbed. When possible, we measured the length of the maximum terminal oocyte and its associated trophocytes of both resorbing (if present) and viable oocytes, and used these measurements to classify the stage of ovary maturation. We identified the maximum terminal oocyte and maximum stage of oocyte maturation as the longer/more mature from the two ovaries. We used these maxima for statistical analyses.

To determine whether our diet treatments were effective, we quantified the amount of pollen consumed by lab-reared females receiving pollen, relative to reproductive females, by estimating the number of pollen grains in the hindgut. We followed previously described methods (Cane et al., 2016) to estimate pollen grains in six hindguts from each group. Individual hindguts were placed in 0.5 ml microcentrifuge tubes with 50 μl of 70% ethanol and torn apart with forceps. The mix was then vortexed for 5 s, using Vortex-Genie 2 on the highest setting of 10. The shredded gut was then removed using forceps, dabbing the tissue on the sides of the tube to remove excess ethanol and pollen. Each sample was vortexed on the highest setting for 10 s immediately prior to loading 10 μl of the solution into one chamber of a hemocytometer for pollen counting. Pollen grains were counted across the entire chamber under a compound microscope at 20× magnification. Three different 10 μl aliquots were counted for each sample, using the entire chamber each time. For each sample, the average of these three counts was divided by 0.0009 ml, the volume of each hemocytometer chamber, and then multiplied by the volume of ethanol used per sample (0.05 ml).

All statistical analyses were performed in R version 3.2.5 (https://www.R-project.org/). Visual inspection of a qq-plot (R package ‘car’; Fox and Weisberg, 2011) and an Anderson–Darling normality test (R package ‘nortest’, version 1.0-4, https://CRAN.R-project.org/package=nortest) revealed significant departures from normality in the distribution of maximum terminal oocyte and Dufour's gland lengths for experiment 1, but not for experiment 2. We therefore applied a Box–Cox transformation to the data for experiment 1 before running the final model (Venables and Ripley, 2002). For both experiment 1 and 2, we modeled the maximum viable terminal oocyte and Dufour's gland lengths with separate linear mixed effects regressions that initially included intertegular width and treatment (coded as factors: diet for experiment 1, JH×social for experiment 2) as fixed effects, with bee bed of origin as a random effect (Bates et al., 2015). In each case, the variance and standard deviation for the intercept of bee bed were zero, so a linear model without random effects was used for subsequent analyses. Intertegular width was removed from the final models in cases where it was non-significant (P>0.05) – all except Dufour's gland length in experiment 2. We used Tukey post hoc tests to investigate significant differences between treatment groups (Hothorn et al., 2008).

For experiment 2, we repeated the analyses after removing cases where the older, reproductive partner had a smaller intertegular width, maximum terminal oocyte or Dufour's gland than the newly emerged cage-mate to determine whether relative size or reproductive development influenced the outcome of the social treatment.

We compared stages of oocyte maturation across treatment groups in experiment 1 with ordinal logistic regression (R package ‘ordinal’, https://cran.r-project.org/package=ordinal). However, this model was not appropriate for the data in experiment 2 because of the low representation of some values. We compared the proportion of oocyte maturation stage among JH treatment groups in experiment 2 using a chi-square test (R package ‘stats’, https://www.R-project.org/). We also used a chi-square test to compare the proportion of samples with and without resorbed oocytes across groups in experiments 1 and 2.

Final estimates of pollen counts in the hindgut were compared between groups (reproductive, nesting females, sugar+pollen mix, sugar+pollen+alfalfa flowers) with a linear model function (lm), after applying a Box–Cox transformation of the data (Venables and Ripley, 2002).

We used the reproductive and newly emerged females to describe patterns of oocyte maturation in alkali bees. Newly emerged bees activate their oocytes rapidly, even without mating. The modal stage of maturation for newly emerged females was one viable stage I oocyte, with a viable stage I or II oocyte in the other ovary. However, three newly emerged females had viable stage IV or V oocytes in both ovaries. It is possible that these females had spent 1 day or more in their nests upon eclosion, and were therefore slightly older than the others. Newly emerged females did not show any evidence of resorbing oocytes.

The modal stage of maturation for reproductive females was one viable stage IV oocyte and one viable stage IV or V oocyte in the other ovary. The maximum viable terminal oocytes in both ovaries were vitellogenic (stage III, IV or V) for all reproductive females, except one with a pre-vitellogenic (stage II) oocyte in one ovary. Most (70%) reproductive females had at least one resorbing oocyte, and 35% had a resorbing oocyte in both ovaries. Resorbing oocytes were all in either stage IVr or Vr, with 58% in stage Vr. The maximum viable terminal oocytes in ovaries with a resorbing oocyte were in stage III or IV, indicating that oocyte maturation is sequential across ovarioles within an ovary.

Mortality was not significantly different among lab-reared females on different diet treatments (mortality: sugar 40%, sugar+pollen 32%, sugar+pollen+flowers 46%; χ²=1.29, P=0.52, n=92). There were significant differences in both maximum viable terminal oocyte and Dufour's gland length among treatment groups (oocytes: F_4,89=30.68, r²=0.58, P<4.79×10⁻¹⁶, Table S1; Dufour's gland: F_4,101=45.80, r²=0.64, P<2.20×10⁻¹⁶, Table S2). Among these groups, lab-reared, 10 day old females had significantly longer viable maximum terminal oocytes and Dufour's glands than newly emerged females (Fig. 1). However, actively nesting reproductive females had significantly more developed reproductive anatomy than either newly emerged or lab-reared females (Fig. 1). We did not observe significant differences in maximum viable terminal oocyte or Dufour's gland length among females reared in the lab for 10 days on different diets (Fig. 1).

The maximum viable terminal oocytes of newly emerged females were at significantly earlier stages of development than those of lab-reared 10 day old females or reproductive females, but there were no significant differences between females in the latter two groups, all but one of which had vitellogenic oocytes (stage III or higher) (ordinal logistic regression: Z=−4.61, P=3.99×10⁻⁶; Fig. 2). Three lab-reared 10 day old females that received the sugar+pollen diet developed viable mature oocytes (stage V). Approximately half of the lab-reared females had at least one resorbing oocyte (sugar only 50%; sugar+pollen 50%; sugar+pollen+flowers 57%), and all except one (stage Vr) were in stage IVr. The proportion of females with resorbing oocytes among lab-reared females was not significantly different from that of reproductive females (χ²=2.10, P=0.55, n=70 bees). However, most (58%) of the resorbing oocytes in reproductive females were in a more advanced stage (Vr).

The estimated number of pollen grains detected in the hindguts was not significantly different among reproductive females and the two groups of lab-reared females that received pollen in their diet (F_2,15=3.32, r²=0.31, P=0.06, Fig. 3).

Mortality was not significantly different among treatment groups in experiment 2 (mortality: sham/solitary 19%, sham/social 19%, DMF/solitary 20%, DMF/social 47%, JH/solitary 35%, JH/social 44%; χ²=7.04, P=0.22, n=101). There were significant differences in maximum viable terminal oocyte and Dufour's gland length among treatment groups (oocytes: F_5,41=6.68, r²=0.45, P=1.23×10⁻⁴, Table S3; Dufour's gland: F_6,57=8.77, r²=0.48, P=8.97×10⁻⁷, Table S4). Females treated with JH had significantly longer viable terminal oocytes and Dufour's glands than females in control groups, but variation in the social environment did not have a significant effect on these measures of reproductive physiology (Fig. 4).

Reproductive females and newly emerged females paired in the social treatments were similar in size to one another (mean±s.d. ratio of intertegular width 1.03±0.11). On average, the reproductive females had longer viable oocytes and Dufour's glands than their newly emerged cage-mates (mean ratio of maximum terminal oocyte length 1.72±0.62, mean ratio of Dufour's glands 1.13±0.17). However, there were 12 cases in which the reproductive female was smaller and/or had smaller ovaries or Dufour's glands than their newly emerged female cage-mates. Elimination of these 12 cases from the dataset did not change the results (Table S5, S6).

There were significant differences in oocyte maturation stage among JH treatment groups (χ²=23.20, P=7.31×10⁻⁴, n=40 bees). Only females treated with JH had viable mature oocytes (stage V), and sham-treated bees were the only group with pre-vitellogenic (stage II) oocytes (Fig. 5). Most (78%) of the JH-treated females had at least one resorbing oocyte, while 55–58% of the females in the control treatments had resorbing oocytes. However, this difference was not statistically significant (χ²=2.56, P=0.28, n=64). Resorbed oocytes were in stage IVr and Vr in each group.

Variation in reproductive physiology is a hallmark of the social insect societies, in which just one or a few individuals out of thousands within a colony are reproductively active, despite shared genetic influences. Investigating the factors that contribute to reproductive variation in solitary relatives of social insects can provide clues as to how reproductive castes evolve (Kapheim, 2017). Our results demonstrate that solitary alkali bees do not require dietary protein during the initial stages of reproductive maturation, but that JH enhances this process. We also found that, unlike for social bees, interactions between conspecifics do not influence reproductive physiology. This provides important information about the physiological foundation from which reproductive castes emerged.

Access to dietary protein did not limit reproductive activation among newly emerged alkali bees, but it was rarely sufficient for reproductive maturation. Most of the lab-reared females in our study did not develop mature (stage V or Vr) oocytes or Dufour's glands during the 10 day study period, despite having similar amounts of pollen in their hindguts to actively nesting, reproductive females, indicating they had consumed ecologically relevant amounts of pollen during the experiment. Alkali bees commonly begin laying eggs within a few days of eclosion, and some of the newly emerged bees had mature oocytes, indicating that our study period provided ample time for reproductive maturation (Bohart, 1955). It is possible that the completion of reproductive maturation is hastened by ecological cues, such as nesting substrate or access to mates. Seminal fluid is known to trigger oogenesis in several insect species (Avila et al., 2011), and mating limitation is known to influence reproductive activity in other halictid bees (Yanega, 1989, 1992). However, lab-reared females of another halictid bee, M. genalis, reached reproductive maturity when reared, unmated, in the lab for 10 days (Kapheim et al., 2012). If mating is a reproductive limitation in alkali bees, it can apparently be overridden by JH treatment, as JH-treated females in our study were more likely to reach reproductive maturity, even in the absence of mating or ecological cues. Regardless of the role of dietary protein in the initiation of reproductive activation, the fact that alkali bees and other closely related halictid bees consume pollen on a daily basis suggests that protein is likely necessary for sustained reproductive activity throughout the breeding season (Cane et al., 2016; Wuellner, 1999).

Our results are in contrast to results of similar studies of solitary bees in the family Megachilidae, Osmia californica and Megachile rotundata, which showed access to dietary protein is essential for reproductive maturation among newly emerged females (Cane, 2016; Richards, 1994). Unlike alkali bees, Osmia overwinter as adults, begin oogenesis prior to eclosion, and thus eclose with depleted protein stores (Wasielewski et al., 2011). Protein stores have not been measured in newly emerged alkali bees, but remaining reproductively quiescent until eclosion is likely to be less energetically expensive and may be associated with increased availability of nutrient stores for post-eclosion maturation (Hahn and Denlinger, 2007). Alkali bees may therefore be better poised to initiate oogenesis without a dietary protein source. Conversely, pollen is necessary to stimulate vitellogenesis in M. rotundata, which also remains reproductively quiescent until ecolosion. The apparent difference in nutritional requirements for oogenesis among megachilid and halictid bees indicates that the physiological basis for reproductive activity is highly variable among solitary bees. This suggests that assumptions about reproductive physiology among the ancestors of social bees should be made with caution. Additional research on solitary bees from additional families in which eusociality has evolved (e.g. Apidae) are necessary to fully understand variation in nutritional requirements for reproduction.

Our results suggest that JH is a limiting factor in reproductive maturation among alkali bees. Females treated with JH were significantly more likely to have developed viable mature (stage V) oocytes and Dufour's glands, though other groups had mature resorbing (stage Vr) oocytes. This is consistent with earlier results (Kapheim and Johnson, 2017), and provides evidence of a conserved gonadotropic role for JH in alkali bees. In most insects, including bumble bees, JH stimulates the synthesis of vitellogenin, an egg-yolk precursor protein necessary for oocyte maturation (Amsalem et al., 2015; Badisco et al., 2013). This suggests the gonadotropic response to JH may depend on a dietary source of protein. All of the females receiving JH treatments in our study also received dietary protein from pollen, and thus had the nutritional resources necessary to complete vitellogenesis. Future studies are needed to determine how nutrition and JH pathways interact in alkali bee oogenesis.

The path by which JH stimulates Dufour's gland maturation is less clear, as Dufour's gland evolved in the ancestor of Hymenoptera, and secretes chemicals with a wide range of functions within this group (Mitra, 2013). Dufour's gland is likely derived from the colleterial accessory gland in other insects (Mitra, 2013), and reproductive maturity of this gland is induced by JH in cockroaches (Byrsotria fumigate, Periplaneta americana) (Bell and Barth, 1970; Willis and Brunet, 1966). Moreover, JH influences the chemical composition of Dufour's gland secretions in bumble bee (B. terrestris) workers (Shpigler et al., 2014). This, along with our results, suggests that endocrine regulation of Dufour's gland is deeply conserved among insects. Additional research is needed to determine the molecular mechanisms by which JH affects Dufour's gland function.

Unlike for social bees and some solitary bees, variation in the social environment does not influence reproductive physiology among solitary alkali bees. Most of the 20,000 species of bees are solitary, mass-provisioning and annual (Michener, 2007). This means that after mating, females build a nest, provision each brood cell with pollen and nectar, lay an egg on top of those provisions, seal the cell, and die before any of her offspring complete development. For many of these species, a single mating event is the only interaction they have with conspecifics in their entire lives. Conversely, highly eusocial insects have evolved sophisticated forms of social communication. Within colonies of honey bees, bumble bees and sweat bees, behavior and reproductive physiology are dynamically regulated by the behavior and/or pheromones of nest-mates (Alaux et al., 2009; Amsalem and Hefetz, 2010; Arneson and Wcislo, 2003; Grozinger et al., 2003; Huang et al., 1998; Kapheim et al., 2016; Le Conte et al., 2001; Li-Byarlay et al., 2014; Padilla et al., 2016; Smith et al., 2009). Research with a facultatively eusocial halictid bee, M. genalis, suggests that aggression from older, reproductive females can limit reproductive development via JH suppression in newly emerged females (Kapheim et al., 2016; Smith et al., 2009, 2013). However, the degree to which solitary bees respond to social cues varies among species. For example, experimental co-housing of otherwise solitary Ceratina bees (Family Apidae) results in division of labor and reproductive suppression in species that occasionally share nests in nature (Sakagami and Maeta, 1977, 1984, 1989, 1995), but apparently not in species that never share nests, like alkali bees. Although we did not directly measure behavioral interactions among pairs as part of our study, we routinely observed aggressive exchanges among pairs of females in cages. Our results thus suggest that sensitivity to cues from the social environment observed in social halictid bees is not conserved in their solitary relatives. Alkali bees nest in extremely dense aggregations, with up to 713 nests per square meter in a study area (Cane, 2008). At high density, these ground-nesting females are likely to encounter each other regularly as they dig tunnels and build cells, and there is thus likely to be strong selection against physiological sensitivity to social interactions in these populations. Similar studies with additional solitary bees are necessary to identify the circumstances under which sensitivity to the social environment influences reproductive physiology.

This study is the first experimental investigation of dietary, endocrine and social effects on reproductive maturation in a solitary bee closely related to lineages in which sociality evolved, but which is highly divergent (>100 my) from the more commonly studied eusocial honey bees and bumble bees (Cardinal and Danforth, 2013). Our results reveal that the factors contributing to the initiation of reproductive activation and completion of reproductive maturity may be different. Specifically, dietary protein was not essential for the initiation of reproductive activation, but was rarely sufficient for reproductive maturation. JH, however, may be a limiting factor in maturation of both oocytes and Dufour's gland. This provides important insight into how sensitivity to these cues evolved with the origin of reproductive castes in social insects. For example, the effects of JH on ovary and Dufour's gland maturation are apparently conserved between solitary alkali bees and bumble bee workers. However, these JH effects are responsive to social status in bumble bees but not alkali bees (Amsalem et al., 2014; Shpigler et al., 2014). This suggests that different components of the endocrine networks influencing reproductive physiology were independently modified during social evolution. Also, nutrition and cues from the social environment are some of the most important factors in reproductive suppression of workers among social bees (Amsalem et al., 2013; Kapheim et al., 2016; Lawson et al., 2016; Padilla et al., 2016), but these factors did not have a significant influence on variation in reproductive activation in solitary alkali bees. This suggests that changes in how nutrient-sensing and environment-sensing pathways regulate reproductive physiology were especially important in the evolutionary origins of reproductive castes. Further comparisons of the molecular networks underlying the physiological response to nutritional, endocrine and social cues across species are likely to provide key insight into how reproductive division of labor evolves.

Thanks to A. Tripodi for help with imaging spermathecae, E. Klinger for help with pollen counting, and J. Cane for helpful discussion. Thanks to A. Tripodi, D. Cox-Foster and two anonymous reviewers for helpful comments on a previous version of this manuscript. We are grateful to John Dodd and Forage Genetics International for providing lab space and logistical support in Touchet, WA, USA. We thank Mike Ingham, Mark Wagoner and Mike Buckley for access to their bee beds and bees. M. Jolley and F. Dowsett provided valuable assistance in the field.