1 Introduction

Dispersal is defined broadly as movement of individuals with potential consequences for gene flow (Ronce 2007), which is a key process sha** the structure and characteristics of populations or groups of individuals and influencing the likelihood of inbreeding and intensity of mate competition with relatives (Pusey and Wolf 1996; Bowler and Benton 2005; Ronce 2007). Dispersal is a multi-causal process that is affected by various ecological constraints, such as habitat quality, population density, and kin interactions, as well as individual phenotypes, such as morphology and endocrine function (Dufty et al. 2002; Bowler and Benton 2005; Clobert et al. 2009). Body size is a fundamental life-history trait that affects an individual’s movement and dispersal capacity (Greenleaf et al. 2007). Body size is also correlated with many other facets, such as metabolism, fecundity, superiority of resources, and mate competition (Thornhill and Alcock 1983; Honěk 1993; Blanckenhorn 2000). Thus, the multiple facets of dispersal-related traits may indirectly affect the structure and characteristics of populations or groups of individuals via dispersal, as proposed in the context of dispersal syndrome; Dispersal syndrome refers to the covariation of dispersal and traits, i.e., a suite of traits correlated with dispersal (Clobert et al. 2009; Stevens et al. 2014).

Social Hymenoptera insects are susceptible to inbreeding and genetic drift because of their small effective population size resulting from limited reproductive individuals (queens). In addition, haplodiploidy causes a decrease in the gene copies (allele) of males in the population’s gene pool, resulting in a smaller effective population size, which is three-fourths of diploid under an equal sex ratio (Hedrick and Parker 1997). Furthermore, in the case of complementary sex determination systems (where sex determines at a single locus), the increased frequency of homozygous genotypes with inbreeding causes diploid males to emerge from fertilised eggs that usually develop into females, i.e. queens and workers (Cook 1993). This causes indirect death of females and not only limits colony growth but also reduces effective population size due to decreasing the number of females (Hedrick and Parker 1997; Zayed 2004, 2009). These characteristics imply the need for effective outbreeding.

Eusocial Hymenoptera insects, such as ants and bees, form temporal male aggregations for mating (Ayasse et al. 2001; Boomsma et al. 2005). The congregation consists of males from colonies nesting in an area (Baudry et al. 1998; Mueller et al. 2012; Beaurepaire et al. 2014; dos Santos et al. 2016), and has a potentially important effect on decreasing the probability of inbreeding (Paxton 2005). Thus, understanding the factors and processes that affect the structure and characteristics of male congregations is crucial. In the mating systems, most eusocial hymenopteran males, including honeybees, cannot independently survive away from colonies (Winston 1987; Hakala et al. 2019), so the underlying process for forming mating congregations is the dispersal of males from natal colonies.

Honeybees (Apis mellifera L.) aggregate and mate at specific aerial mating locations called drone congregation areas (DCAs) (Zmarlicki and Morse 1963; Ruttner 1966). DCAs are found at several locations within the range where males (drones) are capable of flight, and have unique landscape characteristics that orient drones toward these places (Galindo-Cardona et al. 2012; Koeniger et al. 2014; Hayashi and Satoh 2021). Meanwhile, observation using X-band radar on a uniform landscape has shown that drones form flyways that contain many individuals and that the flyways run parallel to prominent landscape features such as the tree line. In this case, many drones aggregate at the intersection of flyways, where interaction between individuals is likely to occur, but they can also aggregate within the flyways (Loper et al. 1992). Thus, although it is controversial (since drones’ congregation does not necessarily occur only in distinct, defined places), the formation of aggregations in their mating is universal.

Drones from approximately 10–100 unrelated colonies aggregate and fly within DCAs to find queens with which to mate (Baudry et al. 1998; Jaffé et al. 2010; Moritz et al. 2013). The genetic structure and size of the congregations vary spatially and temporally (Taylor and Rowell 1988; Kraus et al. 2005; Jaffé et al. 2009; Utaipanon et al. 2019). Variations in drones’ flight abilities, which is affected by their body size, would be a predominant proximate cause limiting dispersal distance and pattern (Gençer and Firatli 2005; Hrassnigg et al. 2005). Furthermore, drones with better flight abilities have a selective advantage for mating with queens because they must stay until queens visit DCAs and can quickly reach the queens before their competitors (Coelho 1996; Koeniger and Koeniger 2000; Koeniger et al. 2005a). Therefore, the body size of drones may not only limit their dispersal to DCAs, but may also influence competitive traits that affect male-male competition in congregations via dispersal.

This study evaluated the effect of body size on dispersal by trap** drones at DCAs at different distances from the colonies, and also tested for a correlation between body size and flight abilities to determine if body size affects dispersal distances and the structure and characteristics of mating congregations. A previous study has shown that the homing range of drones varies with age, suggesting that their flight range likewise varies because their survival strongly depends on homing to the colony when they fail to mate with queens (Hayashi et al. 2022). Thus, this study also evaluated the effect of age on congregation via dispersal.

2 Material and methods

2.1 Drone rearing

Honeybee (Apis mellifera L.) colonies were reared on a farm at the Tokyo University of Agriculture and Technology (TUAT). To obtain the experimental drones, empty drone combs were introduced into the queenright colonies. The honeybee queen laid unfertilised eggs that developed into adult drones in the cells of the drone combs. The combs were removed from the colonies 10 days after the cells were sealed and placed in an incubator at 33 °C in the dark. To obtain samples of known age and colony, the thorax and abdomen of the emerged drones were paint-marked within 24 h after emergence. In experiment 1, drones derived from queens of seven different colonies were used. In experiment 2, drones derived from queens of two different colonies were used. The marked drones were introduced into queenright colonies (> 32,000 workers).

2.2 Experiment 1: Observation of drone size and age in DCAs

The study period was 40 days between late June and the end of July. Experimental drones were introduced into three natal or foster colonies for 51 days, starting from 23 days before the experiment. On average, the number of drones introduced into the colonies per day was 320 ± 45 (mean ± standard error) and in total, 17,958 drones were introduced. Drones were captured at three DCA locations: approximately 200, 500, and 1000 m away from the colonies in the south and southeast direction (Figure 1). Locations of DCAs were previously surveyed by presenting queen-sized black wood pieces applied with 9-oxo-2-decenoic acid (9-ODA) to a height of 10 m; 9-ODA is a queen pheromone that attracts drones (Gary 1962). Identified DCA locations within the study area were relatively open-landed, low elevation and high solar irradiance (Hayashi and Satoh 2021). The places where drones aggregate in the study area were limited, and the distances between the DCAs used in this study and other identified DCAs were over 250 m away.

Figure 1.
figure 1

Map of study site locations. The three DCA locations where drones were trapped are shown by filled circle. Location of colonies where experimental drones were introduced is shown by star symbols. Tall plants, including trees and crops, and structures are shown by shaded and open boxes, respectively. This map is based on data from the Geospatial Information Authority of Japan (GIAJ; Basic Geospatial Information; Perimeter line of structures) and Japan Aerospace Exploration Agency (JAXA; High Resolution Land Use Land Cover diagram).

Full size image

Drone trap** in DCAs was conducted for 1–1.5 h per day, a total of 14 days, with 4–5 days at each location. Drones were trapped by raising a glass fibre pole with a Taylor trap 9 m above the ground (Taylor 1984). Several queen-sized black wood pieces were hung inside the trap, and 500 µg of 9-ODA was applied to one each of the top and bottom pieces. The drones were captured between 13:30 and 15:45. The ages of the trapped drones were recorded based on the marked colour, and their body sizes were measured in the laboratory. The distance between the wing bases of drones, which is referred to as intertegular (IT) span, was measured as an index of their body size (Cane 1987; Greenleaf et al. 2007). The measurements were obtained using a light microscope (cellSens Standard Olympus software version 1.6, Olympus Optical Co., Tokyo, Japan).

2.3 Experiment 2: Evaluation of the relationship between body size or age and flight ability

Experimental drones were introduced into foster colonies of two-story hives with a horizontal queen excluder separating the upper and lower hives (Hayashi et al. 2022). Drones were collected from the upper and lower groups 9–12 days after emergence and used in the flight experiment (n = 74). Differences between collected groups were not considered because no effects on flight parameters were found (flight time, distance, and velocity; GLMM: P > 0.52).

A flight mill was used to evaluate the flight abilities of the drones. A 19-mm-long pin was attached to the dorsal part of the drones’ thorax using instant glue; they were tethered to the tip of the arm of 19.5 cm radius in a roundabout. First, their energy reserves were depleted by stimulating flight until wing movement weakened. Flight was then stimulated by removing a ball of paper that had been used to restrict them (Gmeinbauer and Crailsheim 1993). Second, the drones were weighed, fed 10 µL of 2 M glucose solution using a Hamilton syringe (701 RN, Hamilton Co.), then weighed again to check the feeding amount. After feeding, the drones’ movements were restricted again with the paper ball, and they were allowed to rest. The drones were stimulated to fly again 5 min after feeding; the flight time and number of rotations in the roundabout were measured. The flight velocity was calculated based on the flight time and rotation number. Flight observations were conducted at 25 °C under a 60-W light bulb. The roundabout was surrounded by a strip of black and white at an interval of 2.5 cm to provide the drones with a homogenous the visual environment during flight. The IT span of the drones was measured after flight observation.

2.4 Statistical analysis

2.4.1 Experiment 1

All analyses were performed using R, version 3.5.0 (R Core Team 2018). The relationships of the size and age of drones with the distance of DCAs from the colonies were analysed using generalised linear mixed models (GLMM, R package ‘lme4’) with Gaussian and Poisson error structures, respectively. The age and IT span of the trapped drones were incorporated into each model as the response variables. The distances of DCAs from the colonies were scaled to mean = 0 and standard deviation (SD) = 1 to avoid convergence problems in the models and were incorporated into models as explanatory variables. The IDs of the natal colonies of trapped drones were incorporated into all models as random terms to consider the variance in the colonies of the trapped drones. The statistical significance of the variables in the models was determined using the likelihood ratio test (α = 0.05). Furthermore, to detect a trend in the age of drones in DCAs that the variance of the observed age might mask, age was categorised into old and young groups based on the mean age of drones in the DCAs. The categorical variable was treated as binary data with 1 for young and 0 for old, and a similar analysis was conducted using GLMMs with binomial error structures. To better understand the age structure of drones in DCAs, the difference in age variance between DCAs was evaluated using the F-test, and P values were corrected using the Bonferroni method. To obtain a comparable variance index, the coefficient of variation (CV) for the age of the drones was calculated by dividing the SD by the mean for each DCA.

2.4.2 Experiment 2

The effects of body size and age on the three flight parameters were analysed using GLMMs with a Gaussian error structure. Prior to the analysis, three flight parameters and IT span variables were log-transformed to enhance normality and stabilise the error structures. Flight parameters were incorporated as response variables in each statistical model. The IT span and age of the drones were incorporated into the models as explanatory variables; the IDs of the natal colonies were incorporated into the models as random terms.

3 Results

3.1 Experiment 1: Observation of drone sizes and ages in DCAs

A total of 107, 79, and 29 drones were trapped and marked at DCAs 200, 500, and 1000 m away from the colonies, respectively. A summary of the IT span and the age of drones in each DCA is shown in Table I, and details of the age and natal colony structures are summarised in Supplementary Tables 1 and 2. The IT span of the drones in DCAs was 4.270 ± 0.166 mm (mean ± SD), and significantly increased with the distance of the DCAs from the colonies (GLMM: P < 0.001, Figure 2, Table II). The age of the drones in DCAs was 8.9 ± 3.3 days (mean ± SD), and ranged from 5 to 38 days. No significant correlation was found between the age of the drones and the distance of the DCAs (GLMM: P = 0.125, Figure 3, Table II). Furthermore, the proportion of young drones (< 10 days) decreased with the distance of the DCAs from the colonies (GLMM: P < 0.001, Figure 4, Table II). The CV of age on DCAs 200, 500, and 1000 m away from the colonies were 0.324, 0.457, and 0.262, respectively. The age variance of DCA at 500 m was significantly larger than that at 200 and 1000 m (200 vs. 500 m: F = 0.420, P < 0.001; 500 vs 1000 m: F = 3.071, P = 0.004‬), and the variance between DCAs at 200 and 1000 m was not significantly different (F = 1.288, P > 0.05).

Table I Summary for intertegular (IT) span and the age of trapped drones in DCAs; n indicates the number of trapped drones at three DCAs
Full size table
Figure 2.
figure 2

Box-plot representing intertegular (IT) spans of trapped drones at three DCAs with different distances from the colonies. [Sample size in each DCA: n = 107, n = 79, n = 29 for 200, 500, and 1000 m, respectively; see Table I for details]. Drones’ IT span was significantly correlated with the distance of DCAs from the colonies (GLMM: P < 0.001).

Full size image
Table II Effect of the distance of DCAs from the colonies on drones’ intertegular (IT) span and age and the proportion of young drones (< 10 days old)
Full size table
Figure 3.
figure 3

Box-plot representing ages (days) of trapped drones at three DCAs with different distances from the colonies. [Sample size in each DCA: n = 107, n = 79, n = 29 for 200, 500 and 1000 m, respectively; see Table I for details]. Drones’ age was not correlated with the distance of DCAs from the colonies (GLMM: P = 0.125).

Full size image
Figure 4.
figure 4

Bar-plot representing the proportion of young drones (< 10 days old) at each DCA, categorised based on the mean age of drones on three DCAs. [Number of young and old drones in each DCA: (200 m: n = 93, n = 14 for young and old drones, respectively), (500 m: n = 57, n = 22), (1000 m: n = 13, n = 16)]. The proportion of young drones was correlated with the distance of DCAs from the colonies (GLMM: P < 0.001).

Full size image

3.2 Experiment 2: Evaluation of the relationship between body size or age and flight ability

The drone flight time using the flight mill was 612 ± 150 s (mean ± SD). The flight time was negatively correlated with IT span (GLMM: P < 0.05, Figure 5, Table III). The flight velocity was 1.35 ± 0.37 ms−1 (mean ± SD), and positively correlated with IT-span (GLMM: P < 0.05, Table III). Flight distances were 807 ± 240 m (mean ± SD) and not significantly correlated with IT span (GLMM: P = 0.866, Table III). None of the flight parameters was significantly correlated with the age of the drones (GLMM; P > 0.05, Table III).

Figure 5.
figure 5

Relationship between log-transformed flight parameters and the intertegular (IT) span of drones. The panel shows flight time a, flight distance b, and flight velocity c. Regression lines on the panels are based on estimation obtained from simple variables models derived using the IT span of drones [(GLMM: (flight time; y = 8.661 − 1.591x), (flight distance; y = 6.469 + 0.114x), (flight velocity; y =  − 2.110 + 1.650x)]. Significant (α = 0.05) and non-significant lines for correlation are shown by solid and dashed lines, respectively.

Full size image
Table III Effect of drones’ intertegular (IT) span and age on flight parameters
Full size table

4 Discussion

The effects of body size (IT span) and age on the structure of honeybee drones’ congregation for mating via dispersal were evaluated. Most studies have reported that larger drones have higher mating and reproductive success, suggesting that this is caused by differences in competitive traits based on size, such as mobility and the number of spermatozoa (Coelho 1996; Berg et al. 1997; Schlüns et al. 2003; Koeniger et al. 2005a; Couvillon et al. 2010; Gençer and Kahya 2020). This study found that trapped drones’ body size increased with the distance of DCAs from the colonies and was also correlated with flight abilities (time and velocity) that are linked to dispersal capacity and male-male competition within DCAs (Figures 2 and 5). Thus, multiple facets of body size can influence the structure and characteristics of the congregation via dispersal. Furthermore, the age structure of drones in DCAs varied based on the distance of DCAs from the colonies (Figure 4), but it was not explained by differences in their flight abilities within the observed age (Table II). Thus, the age of drones is likely to influence the congregation through a process that is different from that of size. This study provides insight into factors and processes limiting the components and structure of drones’ congregation in mating, which affects the likelihood of inbreeding and competition between male relatives.

Higher energy is needed for long-distance flights, and the absolute energy demands are predicted to be higher for larger individuals (Grula et al. 2021). A previous study showed that small drones raised in worker cells can take longer flight times and travel greater distances than drones developed in drone cells (Hrassnigg et al. 2005). This study showed that larger drones have shorter flight times; however, flight distances were not correlated with body size (Figure 5). The difference in correlation between flight distances and body size might be caused by the narrow range of body sizes of the drones used here because they were restricted to individuals developed in drone cells. Thus, effect of size on flight distances was ambiguous, but the observed negative correlation of flight time with size indicates that larger drones generally consume more energy per unit time for flight. This implies a similar effect of body size on flight, as observed when the difference in the size distribution of the drones used for comparison is clear (Hrassnigg et al. 2005), and can limit flight distances and times. In this context, drone dispersal distances may decrease with body size, contrasting with the observed relationship between dispersal and body size (Figure 2). Meanwhile, the larger drones showed higher velocity (Figure 5), indicating that they have greater propulsion during flight, which is likely to explain a function of the increased dispersal distances and body size. Flight activities are often affected by external factors, such as wind (Loper et al. 1992); therefore, drones must have stable and powerful flights to reach DCAs that are distant from a colony. Furthermore, body size may involve an increase in the capacity of energy loading for flight, and a higher flight force allows for holding more energy content in crop by reducing the cost per mass. This would have a significant influence, even if the increase in the loading amount is small because drones load highly concentrated nectar as fuel (Hayashi et al. 2016). Thus, the differences based on size are most likely a proximate mechanism for increased dispersal distances.

In the pre-copulation process, a honeybee drone needs the ability to reach and locate mating places (DCAs) and detect queens, suggesting the presence of selection when drones aggregate at the mating place (Jaffé and Moritz 2010). Empirical studies have shown that lower wing size asymmetry, which is known to be associated with higher flight performance in fruit flies (Markow and Ricker 1992), is found in drones caught from DCAs compared with those from colonies (Jaffé and Moritz 2010). The number of visiting drones in DCAs also decreases with the distance of the source colony (Taylor and Rowell 1988; Koeniger et al. 2014). This finding was also observed in the present study (Table I). One hypothesis for the increased body size with the distance of DCAs is that it results in a selection during drone aggregation to DCAs (based on the distance from the natal colony). This affects the congregation’s structure and is likely to have an important function in avoiding inbreeding, competition between relatives, and deleterious factors by dispersal. For example, inbred and Varroa-infected drones with lower flight abilities are likely to be selected during distant flights (Moritz 1988; Duay et al. 2002; Yañez et al. 2012), and more related individuals would be included in DCAs that are near to the colonies (Koeniger et al. 2005b). Particularly, inbreeding in honeybees causes the production of diploid drones due to the increasing frequency of homozygotes at the sex determination locus, and also negatively affects wing symmetry, homeostasis, and foraging in worker bees (Brückner 1978). In addition, honeybees are extremely polyandrous, and queens mate with multiple drones, resulting in high performance and productivity of colonies via increased intra-colonial genetic diversity (Mattila and Seeley 2007; Mattila et al. 2012), but inbreeding causes loss of diversity. Thus, selection related to dispersal and aggregation toward DCAs may indirectly affect mate selection (Jaffé and Moritz 2010), and the potentially high requirement for dispersal may explain the evolution of honeybee drones’ large body size that allows greater and stable dispersal. Our results are based on observations using colonies located at similar sites, but feral Apis colonies have been reported to have clumped distribution in habitats (Oldroyd et al. 1995, 2000; Wattanachaiyingcharoen et al. 2008). Thus, selection via dispersal and aggregation could be one of the underlying principles determining the structure of drone congregations and their dispersal patterns, although it might be limited when colonies are clumped.

A study using harmonic radar reported that drones take long-distance flights for mating and begin as early as the second flight (Woodgate et al. 2021). They can visit several DCAs within a single flight, which suggests that drones require fewer systematic searching processes to locate DCAs (Woodgate et al. 2021). This is supported by our results, in which the mean age of drones in DCAs coincided with the timing for their initiation of flight (Witherelli 1971; Reyes et al. 2019). Meanwhile, DCAs that drones are capable of visiting or locating may vary with age. A study on homing distance toward the hive, which is often used as a proxy for their flight range, showed that the homing distance increases with age (Hayashi et al. 2022). In this study, no significant correlation between age and distance of DCAs from the colonies was found, but the proportion of young drones (< 10 days old) decreased with the distance of DCAs (Figure 4). The lack of a direct correlation between age and distance would be because of the large variance in age. Specifically, approximately 70–80% of the marked drones near DCAs (200 and 500 m) consisted of young drones, and the remaining were a small number of considerably old drones (max. 38 days old). This might reflect the relative ease of accessing DCAs near the colonies for the older drones with lower abilities that are unsuccessful in mating until that time, and also a decreased performance with ageing. Ageing accompanies familiarity with sites via flight experience and various physiological changes in the nervous system and motor activity (Fahrbach et al. 1997; Page and Peng 2001; Hrassnigg and Crailsheim 2005; Degen et al. 2015); thus, factors causing the difference cannot be elucidated here. Nevertheless, an interesting possibility is that the dispersal distance of young drones is limited and can affect the structure of congregations. The drone components for ejaculation and viability of spermatozoa change with age (Locke and Peng 1993; Baer et al. 2012; Stürup et al. 2013; Metz and Tarpy 2019), and these differences may affect the fecundity of queens and the intensity of post-copulatory male-male sperm competition until stored in the spermathecae of queens (den Boer et al. 2010; Gençer and Kahya 2020) via mating with a group of drones that is different in age structure.

Besides dispersal, the observed size and age structure in DCAs is likely affected by various aspects such as the observation methodology, season, and mortality. For example, the trapped drones were collected for measurement of their size at any time, although many drones were introduced into colonies over the study period, so older drones may be present at a low probability. In addition, our study was conducted from late spring to summer, and the longevity of drones is shorter in summer than in spring and autumn (Fukuda and Ohtani 1977; Reyes et al. 2019). Furthermore, honeybee drones undergo age-independent deaths such as predation during flight (Rueppell et al. 2005). Particularly, the number of the trapped drones compared with introduced ones into the colonies was considerably small, which would need to be noted because the observed pattern in DCAs might also be affected by the selection of individuals via potential factors causing mortality. Therefore, further studies are needed to understand the effect of age and body size on the dispersal process and structure of congregations.

5 Conclusion

The results of this study suggest that the body size and age of drones are potential factors limiting the dispersal distance. Given that body size and age are related to competitive traits during mating, these dispersal-related factors could influence the structure and characteristics of drones in DCAs via dispersal.