1 Introduction

Eusocial insects are well known for their highly efficient foraging and resource acquisition behaviours, facilitated by their complex social organisation (Queller and Strassmann 1998). Less understood are some of their more unusual behaviours, especially the extreme conflicts between some insect colonies that can result in incredibly high mortality. Stealing and robbing the resource stores of another colony (cleptobiosis) is quite common in social insects (Breed et al. 2012), with some species being specialised in such behaviors (Sakagami et al. 1993). Other species (including honeybees) are more opportunistic robbers, and exhibit this behaviour only occasionally (Free 1955). In stingless bees (Meliponini), inter-colony attacks are most often associated with stealing, with some genera being obligate resource robbers (Sakagami & Laroca 1963; Grüter et al. 2016). More extreme inter-colony attacks are often associated with attempts by one colony to take over the nest of another. This has been observed in Africanised honey bees taking over the nests of European honey bees (Schneider et al. 2004), and in several stingless bees species (Grüter et al. 2016).

In one of the most striking examples of this behaviour, two eastern Australian stingless bee species, Tetragonula carbonaria and Tetragonula hockingsi, engage in massive inter-colony fights, often resulting in the death of thousands of workers (Cunningham et al. 2014; Gloag et al. 2008; Heard 2016; Wagner and Dollin 1982). Fights involving these two species may be interspecific or intraspecific (Cunningham et al. 2014; Gloag et al. 2008). T. carbonaria and T. hockingsi are closely related species and morphologically very similar to one another, with T. hockingsi having a slightly larger body. These two species can be clearly distinguished from one another, however, by their different nest structures or by using microsatellite markers (Franck et al. 2004).

The two species now interact more frequently in southern Queensland as the range of T. hockingsi has expanded southwards towards the New South Wales border, possibly as a result of climate change or human-assisted movement of hives (Cunningham et al. 2014). The causes of these fights are not yet fully understood, but a well-supported hypothesis is that the attacks are attempts by colonies to take over other established colonies and thus establish resource-rich daughter colonies. This may partly explain why such an attack is worth the inevitable great loss of individuals (Cunningham et al. 2014).

In the swarms that characterise fights within and between T. carbonaria and T. hockingsi, members of a defending colony fly from their nest to engage attackers in the air close to the nest entrance (Gloag et al. 2008; Wagner and Dollin 1982). In the air, individual workers bite or grasp a single opponent, fall to the ground and fight there until both are killed, resulting in numerous fatalities (Cunningham et al. 2014). A specialised guard caste has recently been discovered in T. carbonaria; these guards are 7% larger than foragers and have more antennal sensilla (Wittwer & Elgar 2018). Both guards and foragers demonstrate aggressive behaviours towards non-nestmates only in the presence of nest odours, highlighting the link between aggression and hive defence (Wittwer & Elgar 2018). It is likely that guards make up only a small proportion of the overall workforce in T. carbonaria, as in the South American stingless bee Tetragonisca angustula (~ 1–2%, Grüter et al. 2012), and that during large inter-colony fights, the entire workforce is recruited into hive defence.

On occasion, two nests may engage repeatedly in fighting swarms over weeks or even months. Multiple repeated attacks by a single T. hockingsi colony on a single defending T. carbonaria colony have been documented over a 5-month period, when eventually, the attackers took over the colony (Cunningham et al. 2014). During this attack, recently emerged adult workers that had not fully melanised (callows) were removed from the nest, but this behaviour was reported to last only a few days. The attackers therefore seem to kill all resident workers in the colony they expropriate. The fate of the developing brood however is unknown and it is possible that they emerge as usual and become workers in the new invader colony (Grüter et al. 2016; Cunningham et al. 2014).

We investigated whether T. carbonaria callows from a different nest are accepted following transfer to a new colony and combined observational data with microsatellite analysis of a series of fights involving Tetragonula stingless bees at a focal nest. We aimed to test further the hypothesis that Tetragonula stingless bees retain the brood of the original colony after nest takeovers, and that a mixed-species cohort of workers can persist for some time following these events.

2 Materials and methods

2.1 The study nests

The experimental introduction of callows and the study of fighting activities were focused on two hive boxes installed at residential properties in a typical sub-tropical residential suburb (Annerley, Brisbane, Queensland, Australia). The hive within which the fighting activities were observed is referred to as the focal nest. A second hive, 1 km away from the focal nest, was used for the experimental introduction of callow workers and is referred to as the callow nest. A third nest, located at the same property as the callow nest (approx. 8 m away and separated by a building), was used to obtain callow bees and is called the donor nest. The focal and callow nests were typical medium-sized T. carbonaria colonies with roughly 3000–5000 workers established in a double-layer standard industrial hive box (Heard, 2016). The internal volume of the boxes was 6 L, and, at the beginning of the observation period, they were about half to two-thirds full of nest structures, including brood, food storage pots and other supporting structures. Both the callow and focal hive boxes were fitted with a transparent plastic viewing window under a removable lid to allow detailed observations.

2.2 Acceptance of newly emerged workers (callows) from unrelated colonies

Callows (newly emerged workers identified by their pale colour) were obtained twice from the donor nest to be introduced as conspecific, non-nestmate workers into both the focal nest and the callow nest. All the callows were marked on the scutellum with PX-20 Paint Markers (Mitsubishi Pencil Company, Hanoi, Vietnam) immediately following collection. At the focal nest, two hundred of these marked conspecific, non-nestmate callows were introduced on the 5th of April 2016, together with 200 marked callows from within the nest (which had been collected for reintroduction on the same day). Another 200 marked conspecific, non-nestmate callows from the donor nest were introduced into the callow nest on the 2nd of May 2016. In both nests, marked callows were initially observed post-release to monitor their acceptance by each host colony. Observations were subsequently conducted at each colony over the succeeding weeks to track the behaviour of the marked workers. This involved a 10-min observation period, three times a day, on 3 days each week, until no marked workers were seen in the nest for 3 consecutive observation days. These observations lasted 81 days.

2.3 Observation of fighting activities at the focal nest

A fight, as indicated by an engaged fighting swarm outside the focal nest, was first noticed on the 16th April 2016, 11 days after the marked callows had been introduced. Daily visual observations of the fights were conducted from this date, both inside and outside the focal nest. In addition, bee specimens were regularly sampled from the focal nest and its surrounds for genetic analysis. We thus determined the sequence of fighting behaviours, the genetic identity of the workers involved, and whether a mixed colony was formed. Specimens were sampled within a radius of 3 m from the focal nest after each day’s fighting and preserved in absolute ethanol. The specimens that were collected outside the focal nest were mainly fighting pairs and these were classified as follows: worker fighting pairs (FP, where two workers had become locked in mortal combat), callow pairs (CP, a newly emerged callow being dragged out from the nest by an attacking bee) and marked pairs (MP, where one worker had been marked (earlier, as a callow) and so was known to be a defending bee in a fighting pair). One of the callow pairs consisted of a dead non-physogastric queen locked with a callow and this is referred to as the non-physogastric queen-callow pair (QP). The numbers of dead callows and other dead bees outside the focal nest were recorded each day. After fighting had ceased completely, ten specimens were sampled from within the nest (hive bees, HB) once a week to develop a genetic profile of the resident colony, through time, after the fights.

2.4 Genetic analysis — species identification and number of colonies involved in fights

The bees that had been collected were identified to species and assigned to a colony using molecular methods. To do this, we genotyped each bee at seven previously identified microsatellite loci: Tc7.13, Tc3.56, Tc4.214, Tc4.287, Tc1.20 and Tc3.302 (Green et al. 2001), and Tc3.155a (redesigned by Cunningham et al. 2014). To avoid contamination, each bee was analysed individually with care taken to exclude all body parts of its opponent (if applicable).

To obtain the genotype of each individual bee at the seven microsatellite loci, DNA was extracted using a DIY spin column protocol (Ridley et al. 2016). Each PCR product was labelled with one of the four fluorescent dyes: FAM, NED, PET or VIC using M13 tails (Schuelke 2000), and the PCR was performed in 12 μl reactions (2 μl sample DNA and 10 μl reaction buffer containing dye labelled M13 primer, forward primer, reverse primer, and buffer). Cycling conditions consisted of 3 min at 94 °C, followed by 35 cycles of 30 s at 94 °C, then 30 s at 55 °C/58 °C, depending on the primer (see Cunningham et al. (2014) for details), and a final elongation step of 10 min at 72 °C. After amplification, PCR products tagged with different dyes were pooled, and the pooled samples were separated on an ABI 3730 DNA Analyzer (Macrogen, Korea) using 500 LIZ as the size standard.

The microsatellite peaks were analysed and confirmed using the microsatellite plugin in Geneious 9.1.5 (Kearse et al. 2012). The Bayesian clustering algorithm implemented in STRUCTURE (Pritchard et al. 2000) was used to infer the species to which each bee belonged. To do this, the microsatellite dataset was combined with previous data obtained by Cunningham et al. (2014). These data consisted of microsatellite alleles from three colonies each of T. carbonaria and T. hockingsi that had had their species identity confirmed by visual inspection of the brood and by COI sequencing. The species identity of each individual bee was then determined based on the posterior probability of assignment to each of two clusters (K = 2) using the ‘no admixture’ model of allele correlation.

Once of each individual bee had been identified to species; it was then assigned to the colony from which it had originated, as follows. Multiple structure runs were performed from K = 1 to K = 10, each consisting of 500,000 iterations following a burn-in of 50,000 and 10 replicates per K value. The most likely value of K was inferred using the delta K method (Evanno et al. 2005) implemented in STRUCTURE HARVESTER (Earl and vonHoldt 2012). Then the 10 runs of the most likely K value were permuted using CLUMPP and plotted using Distruct, both implemented in CLUMPAK (Kopelman et al. 2015).

The STRUCTURE analysis assigned individuals to five clusters (see ‘Sect. 3’) but did not fully resolve the colonies, and the final assignments were made by manually checking the genotypes and parentage within each of the inferred clusters, based on the assumptions of haplodiploidy and single mating in T. carbonaria and T. hockingsi (Green and Oldroyd 2002; Palmer et al. 2002; Vollet-Neto et al. 2018; Smith 2020). Under these assumptions, no more than four genotypes (two heterozygous and two homozygous) and three alleles should be present at each of the seven loci in a single colony (Green and Oldroyd 2002). The assumption of single mating has been demonstrated to be valid in both T. carbonaria (Green and Oldroyd 2002) and T. hockingsi (Palmer et al. 2002) by genotyping pupae, callows and workers from 5 and 20 different colonies respectively.

3 Results

3.1 Acceptance of conspecific non-nestmate callows by T. carbonaria

No aggressive behaviours were observed when marked conspecific non-nestmate callows were introduced into the two colonies. In the callow nest, the introduced workers performed a series of age-related behaviours in a similar sequence to reports in other stingless bee species (Heard 2016; Grüter 2020). These included taking care of brood, construction of nest structure, maintenance of food pots and eventually foraging. None of the marked individuals was observed guarding. The frequency of each of these behaviours with the transition in time is indicated in Figure 1.

Figure 1.
figure 1

Summary of the age-based behavioural development of the externally marked Tetragonula carbonaria workers that were introduced, as callows, into a conspecific nest. The filled areas with different colours indicate the predicted percentage of workers participating in the indicated tasks (see key) based on the log Gaussian curves fitted to the frequency data observed. The coloured dots are the actual data points.

Full size image

In the callow nest, all the marked workers were found performing tasks within the hive until 27 days following the day of release, when the first marked workers were seen foraging. Most of the marked workers were subsequently seen foraging and the last three were observed 78 days following their introduction. In the focal nest, the genotyping confirmed that the introduced conspecific non-nestmate workers (‘Introduced, marked T. carbonaria’, Figure 2) and the marked nestmate workers helped to defend their new nest against the attackers. The marked workers had all been killed by the end of the second week of the fight.

Figure 2.
figure 2

Fighting activities at a single Tetragonula carbonaria nest through time (x-axis). Fighting was first observed on 16th April 2016 (day zero) and the last such behaviour was recorded on 17th June 2016 (63 days), with fighting taking place on most days. a Fights and successful invasions as determined by visual observation of activities within the nest and outside of it. Major fights are marked on the timeline, as determined by observation of renewed defensive behaviours on the surface of the hive box and at the entrance. ‘Successful invasion’ indicates a sudden increase in numbers of bees inside the nest cavity. b Species and the colony to which bee specimens sampled at different times during the fights were assigned, based on genetic data. Category of specimen is given to the left of each set of bars, where non-physogastric queen-callow pair (QP) is a pair with one callow and a queen bee, worker fighting pairs (FP) are pairs of workers that were found locked together within a radius of 3 m from the nest, callow pairs (CP) are fighting pairs with at least one callow, marked pairs (MP) are fighting pairs with at least one of them marked, and hive bees (HB) are workers found in the nest cavity. Each bar represents an individual worker; the first bar of each set of individuals found on the same day is aligned to the time scale. See key for colour code to species of bees and their assignment to colonies; unassigned individuals are grouped into a single category. c Total number of dead bees and dead callows collected each day during the fighting period.

Full size image

3.2 Overview of multiple fights at a single hive box

Varying levels of fighting, in swarms, were observed over 63 days at the focal nest, from the 16th April to 17th June 2016. Fighting swarms were formed almost daily during that period, except for days with heavy rain or storms. No foraging or resource raiding from the focal nest was observed. The ultimate result of this series of fights was that the colony collapsed; the focal nest was empty of living bees roughly 2 weeks after the last fight (i.e. on 1st July 2016). In total, 19,805 specimens were collected outside the focal nest, including 427 ejected callows. Aspects of these fights, including visual observations, the number of dead bees collected daily and the composition of the species and colonies involved (as determined by genetic analysis (see below)), are summarised in Figure 2. Day zero on the x-axis refers to the first day when fighting swarms were observed at the focal nest, which was 11 days after the introduction of marked callows. The details of these fights are expanded in the sections that follow.

3.3 Different colonies (and species) involved in the fighting swarms

Based on the results from genetic screening, five groups of bees were initially identified from the STRUCTURE analysis, with all individuals being assigned with high posterior probability to each cluster in the K = 5 plots. The delta K method (Evanno et al. 2005) indicated that K = 3 was the most likely number of clusters from the data, but K = 4 and K = 5 also had good support. After manually checking the genotypes based on the inference of parentage, at least six different colonies were inferred to be involved in the fights. This included the original T. carbonaria colony established in the hive box (‘Defending T. carbonaria colony’, Figure 2b), the T. carbonaria colony from which marked conspecific callows had been introduced (‘Introduced, marked T. carbonaria’, Figure 2b), one foreign T. carbonaria colony (‘T. carbonaria attacking colony 1’, Figure 2b), and three closely related foreign T. hockingsi colonies (‘T. hockingsi attacking colony 1’, ‘T. hockingsi attacking colony 2’, ‘T. hockingsi attacking colony 3’, Figure 2b). Seven individuals, including six workers and one non-physogastric queen, could not be assigned to any of the identified colonies, nor grouped together into a new colony, and these were referred to as ‘Unassigned individuals’ (Figure 2b). The three attacking T. hockingsi colonies were similar to one another genetically and likely closely related, sharing at least one allele at all seven loci. However, one locus had a high frequency of three different alleles (four alleles in total) and based on the assumption of a single haplodiploid mating per colony in T. hockingsi (Palmer et al. 2002) and T. carbonaria (Green & Oldroyd 2002; Smith 2020), we inferred that they represent separate colonies from one another.

3.4 The four major attacks and the colonies involved

During the fighting period, two sudden increases in worker numbers were observed within the nest cavity (Figures 2 and 3a, b), with both being designated as successful invasions (Online resource 1). The first was observed on day 7 and the second on day 41 (Figure 2a). The fighting activities could be divided into four distinct major fights according to the extent of the defending behaviours observed on the surface of the hive box and at its entrance. A new fight typically started with a noticeable increase in defensive behaviours, and these eased after a few days. The four major fights were initiated on days 1, 24, 46 and 53 (Figure 2b). Peaks in the number of dead bees were evident after the first day of each major fight, when ejected callows were also found. All ejected callows (C) tested belonged to the original colony, regardless of time of collection (‘Defending T. carbonaria’, Figure 2b). The four peaks in the numbers of dead individuals per day were 1665, 1740, 573 and 1417 respectively (Figure 2c), and each of the major fights generally involved only two colonies. Around 38% of the fighting pairs genotyped (n = 52 pairs), including those with a marked worker, were identified as a pair of workers from the same colony (Figure 2b, see Sect. 4).

Figure 3.
figure 3

Photographs of behaviours related to inter-colony fighting of T. carbonaria and T. hockingsi. a Nest before invasion of attacking workers. b Nest after invasion by the attacking workers. c Newly emerged callows (circled in white) present in the brood area following invasion by the attacking workers on Day 14. d Roosting male aggregation outside of the nest. e Males attempting to mate with a non-physogastric queen on the ground.

Full size image

In the first major fight, individuals from the ‘T. hockingsi attacking colony 1’ were mostly responsible for ejecting callows from the original colony (nine out of 11 callows (Figure 2b)). Two others were dragged out by individuals from the defending colony. Workers from the same attacking colony were also the opponents of most marked workers during this fight (69%, n = 13), with the remainder being individuals from the defending colony. The number of deaths was consistently above 400 individuals per day from the beginning of the fighting period until day 9, 2 days after the successful invasion (Figure 2c). The dead bees included a small number of ejected callows each day until day eight. Following the successful invasion, callows were observed working in the brood area together with the workers of the attacking colony (Figure 3c). The number of deaths then remained reasonably low until the second major fight, which started on day 24 (Figure 2c).

In the second major fight, ‘T. hockingsi attacking colony 1’ continued to be involved in the external fighting swarm, but members of a new T. hockingsi colony, ‘T. hockingsi attacking colony 2’ were more numerous. Members of the latter were found in all 16 worker fighting pairs (FP) analysed from this fight, but members from another colony, ‘T. hockingsi attacking colony 3’, and several unassigned individuals, were also found during this fight. This particular fight was the most complex of the four in terms of the number of colonies involved (Figure 2b). Most of the ejected callows (C) collected during the entire fighting period were found during this fight (59.9%), with most found on day 25, and the others on days 26, 41 and 42. A dead, non-physogastric queen that was locked with a callow (QP), and found outside the nest on day 25, was identified genetically as T. carbonaria but did not belong to any of the genetically identified colonies across the study. During this fight, a second peak in the number of deaths was recorded on day 33 before entering a period of less intensive fighting (Figure 2c).

In the third and fourth major fights, two colonies (‘T. carbonaria attacking colony 1’ and ‘T. hockingsi attacking colony 3’) contributed most individuals in both fighting periods (Figure 2b). All the worker fighting pairs (FP) analysed from these two fights included members of these colonies, except for one ‘T. hockingsi attacking colony 2’ individual collected on day 53. These two fights were less than 10 days apart. Despite an increase in defensive behaviour observed on day 53, the fourth major fight can be considered an extension of the third major fight, because it was separated by only a few days, with these mainly being rainy and windy.

Finally, when all fighting had ceased, members of ‘T. hockingsi attacking colony 3’ occupied the hive box, but this lasted for only 2 weeks, when the colony collapsed completely and no more living Tetragonula bees were present (Hive bee, Figure 2b). The dissection of the internal nest structures revealed no food storage pots or brood cells. This was more than 2 months from the first observations of fighting at this nest.

3.5 Reproductive activities in association with the nest takeovers

The appearance of male aggregations on vegetation outside the focal nest (Figure 3d) indicated reproductive activities and perhaps the presence of a virgin queen in the nest. Bees captured in these congregations were identified morphologically as males by inspection with a microscope. These male aggregations were roughly proportional to the size of the fighting swarms. Male aggregations were observed on each day of observation, from day 3 until the end of the fighting activities, except on days with heavy rain. We also found two non-physogastric queens, one each on days 21 and 25. A rare observation of attempted copulation was recorded on day 21, when multiple males attempted to copulate with a non-physogastric queen on the ground (Figures. 2a and 3e; Online Resource 2). This process continued for several minutes until the queen, which appeared to be injured, crawled into nearby vegetation.

4 Discussion

The results of this study show that the extreme fighting behaviours in Tetragonula stingless bees can be intricate, involving multiple attacks from different colonies in waves over extended periods. We also provide further evidence that the primary aim of these massive fatal inter-colony attacks in Tetragonula bees is the takeover of a nest rather than raiding resources. Finally, we find that mixed-species worker cohorts can persist following the successful invasion of a nest, although this is undoubtedly transient.

Fighting swarms in Tetragonula bees were originally proposed to be a response to attempted nest takeovers or resource raiding (Wagner and Dollin 1982). Supporting evidence came from microsatellite genotyping of eight naturally occurring fighting swarms as well as all the known nests within a 10 m radius of the defending nest (Gloag et al. 2008). These results showed that each fight predominantly involved only two colonies and supported the view that nest takeovers were the primary aim of these fatal attacks (Gloag et al. 2008). Another detailed examination of a series of fights between two colonies, with a single T. hockingsi colony attacking a single defending T. carbonaria nest repeatedly over several months, showed that the attacking T. hockingsi colony took over the nest and installed a new queen (Cunningham et al. 2014). The observations we report above, from within the hive, revealed no evidence of robbing or raiding during the invasions, consistent with the hypothesis that fights are primarily aimed at taking over a nest. By combining observations with genetic analysis, we also found that multiple colonies may attack the same colony over an extended period of months. Ultimately, this led to the death of about twenty thousand bees but without the successful establishment of a new colony.

Seven unassigned individuals were found in the collection of genotyped bees, which may suggest the presence of bees from yet other colonies becoming involved incidentally. Members from nearby colonies might join large fights incidentally in response to alarm pheromones released by the workers defending their nest (Gloag et al. 2008; Leonhardt 2017). However, the consistent presence of large numbers of bees from the same colonies in each of the four fighting periods suggests that these attacks are strongly functional rather than mostly involving workers incidentally from other colonies. No other Tetragonula bee colonies were present within at least a 10–20-m radius of the defending nest, suggesting bees from more distant nests may have been involved. Tetragonula carbonaria has a homing range of up to 712 m (Smith et al. 2017), and it is possible that attacking colonies could be located up to this distance away from the focal nest.

Most eusocial insects achieve nestmate recognition by unique cuticular hydrocarbon profiles that are species and colony–specific (Singer 1998; van Zweden and d'Ettorre 2010). In many species, individuals acquire their chemical profile mostly from wax or from other workers in their nest and are thought to do so shortly after emergence (Breed 1983; Breed et al. 1998; Breed and Stiller 1992). Once acquired, the chemical profile can continue to develop as individuals age (Greene and Gordon 2003; van Zweden and d'Ettorre 2010; Vernier et al. 2019). In honeybees, nestmate recognition has been shown to be almost completely determined by cues from the nest wax rather than by the genotype of individual workers, allowing newly emerged (callow) honeybees to be accepted by another colony (Breed et al. 2004; Downs & Ratnieks 2000). Newly emerged workers being accepted by foreign colonies are also reported in several stingless bee species (Grüter 2020; Nunes et al. 2011), consistent with the results of our experiment.

Tetragonula bees from the same colony fighting and killing one another during inter-colony fighting is not unusual (Cunningham et al. 2014; Gloag et al. 2008), but we detected a relatively high frequency of nestmates killing bees from their own colony (e.g. day 6, Figure 2b). Stingless bees tend to use mostly exogenous nest volatiles for nestmate recognition (Buchwald and Breed 2005; Leonhardt et al. 2009). The cuticular profile of stingless bees is also often similar to that of their nest (Leonhardt et al. 2011a). Individuals from different T. carbonaria colonies are more aggressive to each other when they are close to their own nest, and in the presence of nest odours (Wittwer and Elgar 2018). Therefore, this process may be blurred when the chemical profile of two nests is derived from materials collected from the same local environment, leading to non-nestmates being more frequently accepted by the guards (Leonhardt et al. 2011b). Indeed, non-nestmate foragers are regularly accepted into neighbouring nests as ‘guests’ in T. carbonaria when two nests are kept close to each other (Stephens et al. 2017). It is possible that nestmate recognition is more difficult in cases like the one reported here, where multiple colonies successively attacked and invaded the focal colony, because it may take some time for the cuticular profile of the new invaders to reflect the new nest volatiles.

Our observations showed that T. carbonaria callows of the focal colony were ejected during fights, but this process lasted for only a short period. Several days after a successful nest invasion by T. hockingsi, newly emerged T. carbonaria callows were then found working with the invaders in the focal nest (Figure 3c). This implies that brood is retained in the nest following takeovers, and that emerging callows from the previous colony’s brood are accepted. In other words, mixed-species cohorts of workers evidently persist for a while. Monitoring the species changes in commercial stingless beehives over a 5-year period revealed that the change between T. carbonaria and T. hockingsi occurs in both directions (Cunningham et al. 2014). Our results suggest that brood retention would likely also occur when a T. hockingsi nest is taken over by a T. carbonaria colony.

Although mixed species colonies of stingless bees have been reported, they are rare (Grüter 2020). In one case, Nannotrigona testaceicornis invaded a recently split colony of the larger and more aggressive Scaptotrigona depilis from a nest box maintained 1.5 m away, and the two species co-existed for 58 days (Menezes et al. 2009). Also, a naturally occurring mixed colony of Melipona flavipennis and M. fasciata was discovered in a tree cavity (Roubik 1981). In this unusual case, laying queens of both species had co-existed within the nest and shared a brood chamber (Roubik 1981). In contrast to these rare instances, the mixed species colony we describe in this study is a result of the takeover of the hive by the invading colonies.

The association between swarming behaviour and male aggregations, and therefore reproductive activities, was also noted by Cunningham et al. (2014). The dead, non-physogastric queen we found outside the nest was unrelated to any of the colonies identified in our samples, suggesting that queens from colonies that are not involved in the inter-colony fights may also be attracted to fighting swarms. Chemical signals released by male aggregations of stingless bees are known to attract virgin queens (Leonhardt 2017). Stingless bees generally produce extra virgin queens, and these excess queens are often expelled from their own colony by the workers (Jarau et al. 2009; Sommeijer et al., 2003). In the South American stingless bee M. scutellaris, recently mated queens sometimes infiltrate a queenless colony, in a process known as queen parasitism (Van Oystaeyen et al. 2013; Wenseleers et al. 2011). It is not clear what the fate of the two queens that we found in this study might have been, but it is possible that a queen from a different hive to the attacking one could be installed during these inter-colony fights. The link between reproductive activity and fighting in these species does, however, add evidence that extreme inter-colony fighting and hive takeovers are primarily related to the establishment of new colonies in Tetragonula bees.