Research articleThe dynamic association between ovariole loss and sterility in adult honeybee workers
Abstract
In the social insects, ovary state (the presence or absence of mature oocytes) and ovary size (the number of ovarioles) are often used as proxies for the reproductive capacity of an individual worker. Ovary size is assumed to be fixed post-eclosion whereas ovary state is demonstrably plastic post-eclosion. Here, we show that in fact ovary size declines as honeybee workers age. This finding is robust across two honeybee species: Apis mellifera and A. cerana. The ovariole loss is likely to be due to the regression of particular ovarioles via programmed cell death. We also provide further support for the observation that honeybee workers with activated ovaries (mature oocytes present) most commonly have five ovarioles rather than a greater or smaller number. This result suggests that workers with more than five ovarioles are unable to physiologically support more than five activated ovarioles and that workers with fewer than five ovarioles are below a threshold necessary for ovary activation. As a worker's ovariole number declines with age, studies on worker ovariole number need to take this plasticity into account.
1. Introduction
Social insect colonies are characterized by reproductive division of labour between the two female castes (the highly fecund queen and the non-reproductive workers). The differential reproductive capacity of these two castes is generally measured by two key traits: ovary state and ovary size. The state of the ovary refers to the degree of oocyte maturation, while ovary size refers to the number of ovarioles (the filaments of the insect ovary in which oocytes develop) [1]. In honeybee (Apis) colonies, the queen and workers differ strongly in both the state and size of their ovaries. The queen has over a hundred ovarioles per ovary whereas workers typically have fewer than 10 per ovary [2,3]. In addition, the queen can produce over 1500 eggs per day whereas the workers have deactivated ovaries and almost never produce eggs [2,4].
In solitary insects, there is a straightforward relationship between female fecundity and ovary size. The number of ovarioles determines the maximum number of eggs that can be produced per unit time, and is positively correlated with the total reproductive output (reviewed in [5,6]). However, for social insect workers there is an additional complication: an ovary with a higher-than-average number of ovarioles can be deactivated so no eggs are produced and the worker has zero fecundity. Nevertheless, ovary size is often used as a proxy for fecundity in social insect workers (for example see [7]). While a honeybee worker's ovary size is only indirectly related to her fecundity, the number of ovarioles she has influences her behaviour. In particular, workers with a higher-than-average number of ovarioles are less likely to perform a retinue response towards the queen [8,9], but more likely to forage at a younger age and preferentially forage for pollen rather than nectar [7,10–13]. Furthermore, ovary transplantation experiments indicate that a causal relationship exists between a worker's number of ovarioles and her behaviour [8,10].
In most insects, ovary size is fixed during the larval or pupal stages, whereas ovary state can be plastic during adulthood [5]. For example, when adult honeybee workers with activated ovaries are transferred from a queenless colony to a queenright colony, their ovaries regress [14]. By contrast, the order-of-magnitude difference in ovariole number between honeybee queens and workers is established pre-eclosion [15,16]. In worker-destined larvae, the majority of ovarioles degenerate via programmed cell death [3,15]. The extent of ovariole loss in the worker-destined larvae is affected by their genotype [17–19], nutrition [20–24] and the presence of the queen [25–27]. However, it is generally assumed that in honeybee workers the number of ovarioles is fixed before eclosion and therefore remains unchanged during adulthood.
Here, we investigate how the ovary state and ovary size of adult honeybee workers is affected by the age of the worker and by the presence or absence of queen pheromone. Also, we examine the previously documented relationship (electronic supplementary material, table S1) between ovariole number and the likelihood of ovary activation, a more direct measure of fecundity. We utilize three diverse honeybee taxa including two species where the number of ovarioles varies widely: the commercial western hive bee (Apis mellifera), the highly reproductive subspecies of South Africa (A. m. capensis) [28,29] and the highly reproductive Asian hive bee (A. cerana) [30]. If ovariole number is correlated with age or exposure to queen pheromone, this would demonstrate that ovariole number is not fixed in the adult worker.
2. Material and methods
(a) Biological material
(i) Apis mellifera in colonies
This experiment was conducted during November 2013 in Sydney, Australia. One queenright colony of A. mellifera of Australian commercial stock was used as the source colony. Age-matched workers were obtained by incubating brood frames overnight at 34.5°C. A random subset of the emerged workers was collected (n = 100) as the 0-day-old time point. All other emerged workers were colour marked on the thorax (Posca Posta Pens, Japan) and placed into each of three queenright (mated queen and brood present) or three queenless (mated queen removed and brood present) host colonies. To increase sample sizes, we marked emerging workers over three consecutive days, using a unique colour for each day and treatment (queenless or queenright host colony). We collected marked workers from each host colony at 7, 14 and 21 days of age and froze them on dry ice (sample sizes in electronic supplementary material, table S2).
(ii) Apis mellifera capensis in colonies
This experiment was conducted during November and December 2014 in Stellenbosch, South Africa. One queenright colony of A. m. capensis was used as the source colony. A random subset of the emerged workers were collected (n = 20) as the 0-day-old time point. The other emerged workers were marked and placed into each of three queenright (mated queen and brood present) or three queenless (mated queen removed and brood present) nucleus host colonies. Workers were collected from each host colony at 7, 14 and 21 days of age (sample sizes in electronic supplementary material, table S2).
(iii) Apis cerana in colonies
This experiment was conducted during April and May 2013 in Kunming, China. Three queenright colonies of A. cerana were used as the source colonies. Emerged workers were collected (n = 20 per source colony) as the 0-day-old time point. The other emerged workers from that source colony were marked with a source-colony-specific colour and placed into each of three queenright (mated queen and brood present) or three queenless (mated queen removed and brood present) nucleus host colonies. Workers were collected from each host colony at 7, 14 and 21 days of age (sample sizes in electronic supplementary material, table S2).
(iv) Apis mellifera in cages
A possible bias in the colony experiments is that a worker's ovariole number may affect her behaviour or longevity. For example, if workers with greater number of ovarioles are more likely to forage earlier in life [11] then they have a greater chance of dying at a young age. We therefore controlled for worker deaths by conducting a cage experiment. This experiment was conducted during December 2013 in Sydney, Australia. One queenright colony of A. mellifera was used as the source colony. A random subset of the emerged workers were collected (n = 100) as the 0-day-old time point. The other emerged workers were placed into each of four replicate laboratory cages [31] (n = 140 per cage). Two of the cages were fitted with a strip of queen mandibular pheromone (Phero Tech Inc.) (i.e. QMP+) and the other two cages had no queen mandibular pheromone strip (i.e. QMP−). The caged workers had a small section of honeycomb and were provided ad libitum with honey, ground pollen and water. Workers were collected from each cage at 7 and 14 days of age (sample sizes in electronic supplementary material, table S2). Very few workers died during this experiment (electronic supplementary material, table S3).
(b) Determination of worker activation state and number of ovarioles
All workers were dissected under dissecting microscopes. The dissections could not be performed blind due to the obvious morphological differences between the treatment groups. The activation state of the ovary was scored as deactivated (transparent, thin ovaries as oogenesis has stopped at the germarium stage), semi-activated (swollen ovaries, as oogenesis has proceeded to the vitellarium stage) or activated (white, large ovaries as oogenesis has almost finished and mature oocytes have developed). For the A. cerana samples, the ovarioles were counted in situ. For the A. mellifera samples, the left ovary was then removed and placed on a microscope slide. The ovary was covered with a drop of water and then with a coverslip to spread the ovarioles. The slide was examined under a microscope and the number of ovarioles counted from the basal end (closest to the oviduct) to obtain an accurate count. Note that for some workers no ovarioles were present and in some other workers the ovariole was present but not attached to the oviduct, so would not be functional.
(c) Statistics
The number of ovarioles was not normally distributed so the data were analysed with non-parametric statistics. We obtained the mean number of ovarioles ± standard error of the mean for the newly emerged workers: A. mellifera (cages) 4.34 ± 0.38, A. mellifera (colonies) 4.54 ± 0.29, A. mellifera capensis 15.90 ± 1.16 and A. cerana 10.82 ± 0.21. To determine whether aged workers had the same number or fewer ovarioles than newly emerged workers we used a Wilcoxon signed-rank test to determine whether the number of ovarioles had declined since emergence, as indicated when the difference between aged workers and the mean of newly emerged workers was less than zero. In addition, we used a Mann–Whitney U-test to determine whether the decline in ovariole number was affected by QMP/queen presence.
In the absence of queen pheromone, adult workers have a greater potential to activate their ovaries than in the presence of queen pheromone. We explored the relationship between ovary activation state and ovary size, as measured by the number of ovarioles. We calculated the mean ovariole number in workers with deactivated (score 1), semi-activated (score 2) and activated ovaries (score 3) from workers that were 7, 14 or 21 days old from the QMP− (cages) or queenless (colonies) treatments. We then determined whether there was an association between ovary size and ovary state using Spearman rank correlations. However, ovariole size may not be linearly related to ovary state. In particular, workers with an average number of ovarioles might be more likely to activate their ovaries [32,33]. To explore this possibility, we determined whether the variance in number of ovarioles differed significantly between ovary state groups (deactivated, semi-activated and activated ovaries). We first performed two-way ANOVAs of ovariole number (predictor variables: cage/colony and age) for each of the three ovary state scores. The residual mean squares (σ2) from these ANOVAs estimate the variability in ovariole number among workers after accounting for any cage/colony and age effects for each ovary state class. We then determined whether there was significantly greater variance in ovariole number in one ovary state group relative to another using Levene's test: 
. L is distributed as F with the degrees of freedom of the respective error mean squares [34] and is unbiased with respect to the differing sample sizes. Note that in A. mellifera queenless colony 3 a 7-day-old worker had activated ovaries and 21 ovarioles (electronic supplementary material, figure S1), 10 more ovarioles than any other worker in that group and an ovariole number more than 7 s.d. from the mean of that group. This individual was a worker–queen intermediate and was removed from the analyses.
(d) Caspase assay
We examined 0- and 7-day-old workers from queenright colonies in the A. mellifera (colonies) experiment, described above to determine the amount of caspase activity, a proxy for the extent of programmed cell death activity in their ovaries. In order to standardize the number of ovarioles, we identified workers with close to five ovarioles in the left ovary (n = 7 per age group). Each sample consisted of the right, deactivated ovary from these individual workers. Caspase activity was then assayed as described in [35]. In brief, caspase activity was detected by using the Caspase-Glo 3/7 assay (Promega), in a 28 µl reaction volume. Triplicate reactions per sample were performed in a LumiTrac 200 96-well plate (Greiner Bio-One). A PHERAstar FS (BMG Labtech) recorded the luminescent signal. The luminescent signal intensity was normalized to its protein concentration to give the caspase activity in the sample. We used a non-parametric Mann–Whitney U-test to determine the effect of age on the caspase activity.
3. Results
(a) Ovariole number declines with age
There was a significant decline in ovariole number as workers aged for QMP+ and QMP− Apis mellifera (cages), queenright A. mellifera (colonies), queenless A. mellifera (colonies) and queenless A. cerana (figure 1; electronic supplementary material, table S4). For QMP+ A. mellifera the number of ovarioles declined 28%, for QMP− A. mellifera 19%, for queenright A. mellifera 23%, for queenless A. mellifera 6% and for queenless A. cerana 25%. The decline in ovariole number appears to be caused by loss of particular ovarioles rather than an overall degradation of all ovarioles (figure 2).
Figure 1. The decline in ovariole number of adult workers following emergence. (a) There was a significant decline in 14-day-old Apis mellifera workers from cages with queen pheromone (QMP+) and without queen pheromone (QMP−). (b) There was a significant decline in 21-day-old A. mellifera workers from queenright colonies, A. mellifera workers from queenless colonies and A. cerana workers from queenless colonies. Error bars are standard error of the mean. Asterisks indicates a statistically significant difference between the mean of the sample's number of ovarioles and newly emerged workers' number of ovarioles (*p < 0.05, **p < 0.01). Figure 2. Ovarioles that are in the process of regression (arrows). (a) Ovary from a 14-day-old queenright Apis mellifera (colonies) worker: four ovarioles and regression of the first ovariole. (b) Ovary from a 0-day-old A. mellifera (colonies) worker: five ovarioles and regression of the second and fourth ovarioles. (Online version in colour.)
In the majority of the experiments, there was no significant difference in the decline in ovariole number when comparing QMP+/queenright and QMP−/queenless workers. The exception was A. cerana where there was a significantly greater decline in ovariole number in queenless workers compared with queenright workers (Mann–Whitney U = 890.5, p < 0.001).
(b) Five might be the optimal ovariole number for activated ovaries
For A. mellifera (cages), A. mellifera (colonies) and A. cerana workers with activated ovaries the average number of ovarioles was approximately five per ovary (figure 3; electronic supplementary material, figure S2). In addition, for A. mellifera workers from QMP− cages there was a positive correlation between ovariole number and ovary activation state: workers with activated ovaries had a higher number of ovarioles (rs = 0.323, p < 0.001, n = 176; figure 3; electronic supplementary material, figure S2). For A. mellifera workers from queenless colonies, there was no significant correlation between ovariole number and activation state (rs = 0.092, p = 0.139, n = 261). For A. m. capensis workers from queenless colonies, there was no significant correlation between ovariole number and activation state (rs = −0.153, p = 0.058, n = 155; figure 3; electronic supplementary material, figure S2). For A. cerana workers from queenless colonies, there was a negative correlation between ovariole number and activation state; workers with activated ovaries had a lower number of ovarioles (rs = −0.836, p = < 0.001, n = 180; figure 3; electronic supplementary material, figure S2).
Figure 3. The distribution of activation state (deactivated, semi-activated and activated) in regards to ovariole number. (a) Apis mellifera workers from cages without queen pheromone (QMP−). (b) Apis mellifera from queenless colonies. (c) Apis m. capensis from queenless colonies. (d) Apis cerana from queenless colonies.
For A. mellifera workers from QMP− cages the variance in ovariole number was significantly greater in workers with deactivated ovaries compared with workers with activated ovaries (F104,15 = 3.037, p = 0.009; electronic supplementary material, figure S3a). Also, workers with semi-activated ovaries had greater variance in ovariole number compared with workers with activated ovaries (F46,15 = 2.782, p = 0.017; electronic supplementary material, figure S3a). For A. cerana workers from queenless colonies, the variance in ovariole number was significantly greater in workers with deactivated ovaries compared with workers with semi-activated ovaries (F74,63 = 2.536, p < 0.001; electronic supplementary material, figure S3d) and in workers with deactivated ovaries compared with workers with activated ovaries (F74,19 = 2.719, p = 0.008; electronic supplementary material, figure S3d).
(c) A high level of programmed cell death occurs in the ovaries of newly emerged workers
Caspase activity was significantly higher in the ovaries of 0-day-old workers, compared with 7-day-old A. mellifera workers (7685 RLU µg−1 and 3482 RLU µg−1, respectively; Mann–Whitney U = 6, p = 0.018).
4. Discussion
Our study shows that adult honeybee workers can lose ovarioles. Therefore, it might be problematic to compare the ovariole number of workers within and across studies, especially if the workers are of different ages. In particular, loss of ovarioles during adulthood is a phenomenon that needs to be taken into account when conducting studies on the association between worker ovariole number and behaviour.
Ovary size in conjunction with ovary state determines a worker's reproductive capacity and therefore influences her direct fitness. It has been previously noted that workers with an average number of ovarioles are more likely to have activated ovaries than workers with a smaller or larger number of ovarioles [32,33]. Though not explicitly noted, data from other studies also support this generalization (electronic supplementary material, table S1). Our study also supports this observation; workers with deactivated ovaries had higher variance in the number of ovarioles compared with workers with activated ovaries. In addition, workers with activated ovaries have approximately five ovarioles. Against this trend, A. m. capensis workers with activated ovaries had 15 ovarioles, in line with their greater ovariole number than workers of other subspecies/species.
Workers with five ovarioles are more likely to have activated ovaries than other workers; however, the direction of causality is unclear. In A. cerana, there may be a physiological trade-off between ovary size and ovary state. Apis cerana workers emerge with a high number of ovarioles and are therefore unlikely to have the nutritional resources available to activate all their ovarioles. We found that queenless A. cerana workers with activated ovaries had far fewer ovarioles than workers with non-activated ovaries (see also [36]). In A. mellifera, workers with five ovarioles are more likely to activate their ovaries than those with fewer ovarioles. This positive association could be due to a threshold of ovary size necessary for activation. An ovary with a higher number of ovarioles is likely to release more endocrine factors [12,37], which may be necessary for ovary activation. In support of this view, we note that the ancestral number of ovarioles for eusocial bees is four per ovary [38].
Worker sterility operates through ‘reproductive control points’ that are triggered by environmental cues (social or nutritional) [1]. Our results suggest that inadequate nutrition, not queen presence, causes ovariole loss in A. mellifera and A. cerana. Interestingly, Drosophila sterility studies also indicate that adult flies can lose ovarioles during adulthood [39]. Therefore ovariole loss during adulthood might be an evolutionarily conserved reproductive control point in insects.
We suggest that the decline in ovariole number in honeybee workers is due to programmed cell death. During both larval [3,40–42] and adult stages [4,43,44], the germ cells inside the ovarioles undergo programmed cell death. Our caspase assay revealed a high level of programmed cell death in the ovaries of newly emerged workers (see also [35]). Death of the germ cells within the ovariole causes the somatic cells of the ovariole sheath to degenerate [3,40–42]. In larvae, the degeneration of an ovariole results in only the basal end of the ovariole sheath remaining [3,42,43,45]. We found that the ovariole of the adult worker degenerates in the same way as the larval ovaries (figure 2). Furthermore, we found some adult honeybee workers lose all of their ovarioles and are therefore completely sterile.
In conclusion, our study reveals that adult honeybee workers not only exhibit reproductive plasticity in regard to ovary state but also in ovary size. Worker sterility, a central characteristic of eusociality, is therefore an environmentally sensitive process that continues throughout the worker's life.
Data accessibility
The datasets supporting this paper are available as part of the electronic supplementary material.
Authors' contributions
I.R. designed all the research, performed the Apis mellifera experiments, performed the caspase experiment, analysed all the data and wrote the paper. B.P.O. designed all the research, performed the Apis mellifera experiments, analysed the data and wrote the paper. M.H.A. designed and performed the Apis mellifera capensis experiment. K.T. designed the Apis cerana experiment. S.D. and X.L. performed the Apis cerana experiment. V.V. designed and performed the caspase experiment. All authors gave final approval for publication.
Competing interests
We have no competing interests.
Funding
The work was funded by Australian Research Council Grants DP150101985 and DP150100151.
Acknowledgements
We thank Melanie Hoang and Rebecca Reid for ovary dissection help. We also thank Prof. Mats Olsson for the use of the PHERAstar.
