Trialling techniques for rearing long-tongued bumblebees under laboratory conditions

Bumblebees are important pollinating insects, but many species have suffered declines over the last century. Long-tongued bumblebees have been identified as particularly at risk, partly due to their more selective diet. Attempts to study these species in captivity have been impeded by stress-induced behaviours which cause queens to kill or abandon their brood. Here, we attempt to further develop techniques, using queen pairing and Bombus terrestris cocoons, to successfully rear two common long-tongued bumblebee species (B. pascuorum and B. hortorum) in captivity. Approximately half of queens laid eggs and 29% produced workers. Although challenges remain, there is a great deal to be gained from optimising the captive rearing of these species.


INTRODUCTION
Bumblebees (Bombus spp. ) are ecologically and economically important pollinating insects, but many species have suffered severe declines in recent decades across Europe and North America (Williams and Osborne 2009;Goulson et al. 2015). Nutritional stress, pathogen infection and exposure to pesticides are key drivers in their decline, but these pressures do not act independently of one another and so their effects on bees are not straightforward (Vanbergen and Initiative 2013). Techniques to rear bumblebees in captivity have been developed over the last century (Van den Eijnde et al. 1990;Pouvreau 2004;Velthuis and Van Doorn 2006). However, these experiments have been almost exclusively conducted on short-tongued species (such as B. terrestris and B. impatiens ). In the wild, these species are generally common, characterised by long colony life-cycles and the utilisation of resources from multiple plant groups and habitats (Fussell and Corbet 1992;Goulson et al. 2005Goulson et al. , 2006. While these species are suitable as models for studying some aspects of social insect biology and behaviour, they are not representative of all bumblebee species and differ significantly in their ecological sensitivity and response to stressors (Goulson et al. 2005).
A small number of studies have succeeded in rearing bumblebee species other than B. terrestris or B. impatiens in captivity. Lhomme et al. (2013) provided the first precise protocol for rearing two cuckoo species, B. vestalis and B. sylvestris and Moerman et al. (2016) showed differential responses to pollen diet between B. terrestris , Electronic supplementary material The online version of this article (https://doi.org/10.1007/s13592-019-00707-7) contains supplementary material, which is available to authorized users. B. pratorum and B. hypnorum micro-colonies reared in captivity. Like B. terrestris and B. impatiens , these other species also have short-medium tongue lengths (Falk 2015). However, the bumblebees that have declined most severely tend to be the long-tongued species that rely heavily on flowers from the Fabaceae plant family (Goulson et al. 2005;Biesmeijer et al. 2006). Unlike short-tongued species, many of which have thrived despite land use intensification, long-tongued bumblebees have a restricted range of foodplants and are from smaller colonies with a smaller foraging range (Goulson et al. 2008). As a result, they are more sensitive to local and global pressures (Goulson 2003;Goulson et al. 2005). Unfortunately, these species have remained relatively understudied in laboratory conditions due to difficulties of rearing the colonies in captivity. During laboratory rearing trials, queens generally perform stress-induced behavi o u r s ( P o m e r o y a n d P l o w r i g h t 1 9 7 9 ; Weidenmüller et al. 2002), including failure to settle in nest boxes provided, failure to utilise pollen, neglecting eggs and larvae and direct ovicide or infanticide, resulting in early colony failure. Rearing conditions, such as nest box design and pollen type, have a substantial effect on the health and reproductive success of bumblebees reared in captivity. There is some evidence to show that giving young, long-tongued colonies access to flowers allows natural foraging and improves colony development (Lhomme et al. 2013;Ptáček et al. 2015;Moerman et al. 2016). Many of the long-tonged bumblebees can be characterised as pocket makers, feeding their larvae a solid rather than liquid diet (Den Boer and Duchateau 2006). How this might affect the physiological and behavioural responses of queens, workers and larvae to environmental stressors remains unclear.
The first attempts at long-tongued bee rearing were carried out by Lindhard (1912). At the same time, Sladen (1912) was testing stimuli to encourage oviposition in short-tongued queens. These methods have been combined and developed since although their effects can be highly variable. Stimuli for egg-laying include various interspecies pairings at different life stages, including the provision of cocoons and callow workers.
Providing brood has a stimulatory effect on queens of several bumblebee species (Yoneda 2008;Bučánková and Ptáček 2010), and cocoons of B. terrestris have commonly been used in rearing trials of long-tongued species (Kwon et al. 2003;Bučánková and Ptáček 2012;Ptáček et al. 2015). Allowing cocoons to hatch and callows to remain in the colony, or adding B. terrestris or Apis mellifera callows, can also have a stimulatory effect on queen egg-laying and B. pascuorum and B. ruderarius are at least somewhat encouraged to oviposition by honey bee workers (Ptacek 1983(Ptacek , 1985Ptáček et al. 2015). Callows have also been observed caring for the queens own brood; Ptáček et al. (2015) observed B. terrestris workers feeding B. pascuorum and B. sylvarum larvae. However, in another trial, B. hortorum queens were aggressive towards multiple stimuli and have been observed destroying B. terrestris cocoons and killing honey bee workers (Bučánková and Ptáček 2012). The pairing of queens, first tested by Sladen (1912), can also encourage oviposition and brood care later on (Plowright and Jay 1966;Alford 1975;Duchateau 1985;Ptácek et al. 2000Ptácek et al. , 2015. This sometimes involves an excluder, which keeps queens physically separated in a shared nest box. Once they are placed together, one female usually becomes dominant and begins egg-laying (Ptáček et al. 2015). Ptácek et al. (2000) and Ptáček et al. (2015) found queen pairing effective for B. pascuorum but not for B. humilis or B. ruderarius , who displayed 'resentful' and aggressive behaviours. Of the three B. humilis females they trialled, one broke through the excluder to reach and kill her neighbour. In another study, queens of B. hortorum were more likely to establish an egg cell when kept alone (100%, n = 9), compared with those who were paired (66%, n = 6) or who had been given a honey bee worker (71%, n = 7) (Bučánková and Ptáček 2012). These various interspecific combinations have therefore yielded contrasting results, from cohabitation to killing, but many of the observations are based on a very small number of individuals, making species-level assessments and progress difficult.
To support the conservation of these longtongued bumblebee species, a more detailed understanding of their biological and behavioural response to natural threats is needed. Developing methods to rear them in captivity would be of great benefit to research in this area. In this study, we attempt to further develop techniques in longtongued bumblebee rearing with two common long-tongued species, B. pascuorum and B. hortorum , under laboratory conditions.

MATERIALS AND METHODS
Rearing trials took place at the University of Sussex (UK) in 2017 ( Figure 1). Queens of B. pascuorum and B. hortorum were collected from surrounding chalk grasslands in the spring (March to mid-April). Pollen foraging occurs post-nest establishment (Evans et al. 2007), and so only queens without pollen in their pollen baskets were collected. Bees were stored in ventilated 5 ml Eppendorf tubes cooled beside ice packs for transport back to the lab. Queens were then placed in a dark room (30 ± 1°C, 20% RH) in ventilated 15 × 15 × 15-cm plastic boxes. We observed previously that queens failed to thermoregulate brood and so we used the highest temperature described in similar long-tongued rearing trials (Ptáček et al. 2015). We did not observe any nest fanning behaviour by queens or workers during the experiment that would have indicated overheating. Queens were provided with a piece of cotton wool to simulate nesting material and 50% (v/v) sugar water (10% fructose, 90% sucrose). For the first feeding, sugar water was placed onto the floor of the box or pipetted directly to queens if they were lethargic. Thereafter, sugar water was provided in an external feeder ad libitum (Figure 1c). Queens were given a fresh 1 g pollen ball every 2 days. Pollen balls were made by grinding dry pollen granules to a powder and combining with 50% (v/v) sugar water (10% fructose, 90% sucrose) until a sticky dough was formed. We found that this wet pollen with a high sugar water content was consumed readily by both species. All bees started on mixed Erica , Salix and Prunus pollen (Pollenergie, France) and from April a pre-made polyfloral wildflower mix (mille fleurs, Pollenergie, France) was added, which became 100% of their food by May to simulate natural conditions for workers. All pollen had been freeze-stored prior to use. Pollen balls were generally placed in the box for bees to collect but attempts were made to feed the first larvae by scattering or gently pressing ground pollen onto the brood casing.
Foll ow ing P tá č ek et a l.
(2 0 1 5 ), all B. pascuorum queens were paired to prompt one to become dominant and begin egg-laying ( Figure 1c). We identified dominant individuals as those which spent the most amount of time in the centre of the box and standing on the pollen provided, since this is naturally where they would lay their eggs (Den Boer and Duchateau 2006). Some of these queens also exhibited a lengthening of the abdomen, also observed by Ptáček et al. (2015). Submissive queens (which tended to be the smaller of the two) were more active and tended to roam the edges of the box. Once identified, submissive females were removed and paired together, prompting another to establish dominance. Paired queens that appeared unsettled or not engaged in egg-laying or brood care were also separated and either re-paired or given their own nest box. The action taken for each queen were based on each queen's dominant/submissive characteristics and the availability of other single queens available for pairing and nest boxes (see Supplementary Tables I, II and III for all pairings and actions). When testing the response of two B. hortorum queens being paired, they were highly aggressive, so were separated and no further B. hortorum queens were paired together.
For the first 33 days, we recorded the progress of queens without additional stimuli. From day 34, to further stimulate egg-laying and encourage brood care behaviours, 2-5-day-old laboratoryreared B. terrestris pupae were introduced to the B. pascuorum and B. hortorum queens. Three times a week throughout the experiment, we recorded (i) the day of first egg-laying, (ii) observable ovicide, larvicide or neglect, (iii) brood pupation, (iv) worker emergence, (v) gyne and male production, (vi) aggression between paired queens, or queens and B. terrestris cocoons/ workers and (vii) use of cotton wool as nesting material. All efforts were made to minimise disturbance throughout the experiment.
To check for differences in behaviour and reproductive success between species, we used  Supplementary Tables I, II and III) and the resulting action taken, which was based on each queen's dominant/submissive characteristics and the availability of single queens and nest boxes. c Top left to right: Nesting boxes used to rear single or paired long-tongued bumblebee queens; a B. pascuorum worker and her queen. Bottom left and right: two colonies of B. hortorum utilising cotton wool as nesting and food storage material.
Trialling techniques for rearing long-tongued bumblebees under laboratory conditions Kruskal-Wallis tests for queen survival, the day they first produced eggs or workers, and the number of workers produced. Any pairs of queens that produced workers together were counted as a single queen. Chi-square tests were used to determine if there was a difference between species in whether or not queens produced eggs or workers, and ejected or neglected eggs or larvae.

RESULTS
A total of sixty-eight queens (11 B. hortorum and 57 B. pascuorum ) were used in the experiment, of which forty queens laid eggs (59%) (Figure 1). Six additional pairs of queens showed close co-operative brood care behaviour with their partner, and as a result, it was not possible tell which queen, if not both, were egg-laying (Table I). There were significant differences between species in whether or not they ejected eggs or larvae, and whether or not they produced workers (χ 2 = 10.6, df = 1, p = 0.001 and χ 2 = 9.9, df = 1, p = 0.0002, respectively). B. pascuorum queens were more likely to eject their eggs or larvae (79% of the 34 queens that laid eggs) compared with B. hortorum (16% of the 6 queens that laid eggs), and B. hortorum queens were nearly three and a half times as likely to produce workers (Table I). There were no significant differences between species in their survival, whether or not they laid eggs, the number of days it took to produce their first egg or worker, the number of workers they produced and whether or not they neglected their eggs or brood (respectively: χ 2 = 0.04, df = 1, p = 0.85; χ 2 = 0.09, p = 0.75; χ 2 = 0.21, df = 1, p = 0.64; χ 2 = 0.71, df = 1, p = 0.40; χ 2 = 3.10, df = 1, p = 0.08; χ 2 = 0.95, df = 1, p = 0.32).
Survival of queens ranged from 2 to 178 days (Figure 2), with a 12% mortality rate observed in the first 7 days. Of the 40 queens that laid eggs, two paired B. pascuorum queens (BpW and BpM) and one individual B. hortorum queen (BhF) did so before the addition of any B. terrestris cocoons (Table II). Across both species, 70% of queens ejected eggs and larvae and 33% neglected them. These behaviours were repeatedly observed throughout the experiment. Queens would lay eggs and then either remove them or the developing larvae, or would fail to feed the larvae so they died in the brood case. This cycle continued even after the first workers were produced. Scattering and pressing additional pollen over the larvae was unsuccessful; the larvae were not observed to feed and the pollen dried around them. Nine single queens (B. hortorum queens BhJ, BhK, BhE, BhD and BhF, and B. pascuorum queens BpB and BpAQ) and three pairs of queens (B. pascuorum queens BpAL-BpAM, BpBC-BpBK and BpAD-BpAZ) produced pupae, and all pupae successfully eclosed and emerged as workers.

Rearing of B. pascuorum queens
In the first 10 days, 91% of B. pascuorum queens performed nest-associated behaviours, including carding with cotton wool, making nectar cups and sitting on pollen (Figure 1c). All 57 queens were paired and a total of 79 pairings (up to three per queen, x̄± s.e. 1.4 ± 0.08) were made over the course of the experiment (Tables SI, SII and SIII). Only in one case did a B. pascuorum queen exhibit aggression, where a previously submissive queen killed her new partner. The duration of the other 78 pairings varied between 1 and 129 days (32 ± 4.7). The behaviour of paired queens could be categorised as aggressive (n = 1), unsettled (i.e. avoidance, no egg-laying or interruption of egg-laying) (n = 21), not engaged with brood care (n = 29), settled (i.e. not aggressive or avoidant) (n = 12) or co-operative (jointly caring for brood) (n = 16) (Table SI, SII and III). Of the 34 queens that were observed to lay eggs, 32 (94%) produced their first eggs before the addition of cocoons, and two (6%) 2-5 days after.
At the point B. terrestris cocoons were given to queens (which was determined by the availability in the source colony), 16 of the B. pascuorum queens were paired and 20 were single. The cocoons were always ignored, and workers emerged 1-4 days later. Thirty-one queens actively avoided their B. terrestris worker until the worker was removed. In eight nests, the B. terrestris worker took over and laid its own eggs. When this happened, the B. pascuorum queens ceased their cyclic egglaying and egg or larval ejection and, even after the B. terrestris worker was removed, did not resume egg-laying again for some t i m e . Tw o q u e e n s k i l l e d t h e i r d o n o r B. terrestris after emergence, including one that had successfully produced a worker before the addition of the cocoon (BpAQ; 25 days post-capture; Table II). Only one pair of queens accepted their worker and exhibited positive physical contact (BpAD-BpAZ). In the remaining nests, the B. terrestris workers stayed around the perimeter of the boxes and we observed very few interactions between them and the B. pascuorum queens. We saw no evidence that B. terrestris workers assisted the B. pascuorum queens in brood care.
For five queens (pairs BpAZ-BpAD and BpAL-BpAM and single queen BpB), the addition of a B. terrestris cocoon was followed by the successful production of B. pascuorum workers, 11-15 days later. One additional pairing produced their first worker 41 days later (BpK-BpBC).
Successful queens (i.e. those which produced w o r k e r s ) p r o d u c e d u p t o t h r e e s m a l l B. pascuorum workers each (Figure 1c). No males or gynes were produced. One queen (BpAQ), which had produced a worker before the addition of cocoons, produced another two after the B. terrestris worker was removed. One B. pascuorum worker laid eggs after the death of the queen. These were neglected and died at theL 2 larval stage.

Rearing of B. hortorum queens
Due to their aggression, B. hortorum queens were not paired. All but one of the 11 queens carried out nest building using the cotton wool (Figure 1c). Fifty-four per cent of queens laid eggs, and of these, 83% went on to produce 4 ± 1.4 workers; three of these also produced males (1, 4 and 38 individuals respectively; Table II).
Three queens survived for less than 17 days in captivity and did not lay any eggs in that time (BhG, BhB and BhI; Table II). Of the three queens which survived between 25 and 36 days (BhC, BhA and BhH), only one laid eggs. These were reared up to the L3 larval stage before she ceased brood care and remained relatively still in a corner for 10 days Table I. A summary of colony development, behaviour and longevity of B. hortorum and B. pascuorum bumblebee queens kept under laboratory conditions. B. hortorum queens were kept individuals (n = 11). B. pascuorum queens were kept in pairs or singly after initial pairing (n = 57 queens). Queens were provided with food and sugar water ad libitum and stimuli to encourage oviposition and brood-rearing were provided at various points in colony development (Table II) B. hortorum queens (n = 11) B. pascuorum queens (n = 57) These queens were first observed with eggs or L1 stage larvae on days 8-23, but did not produce their first workers until, at the earliest, day 58 and at the latest, day 109. The donor cocoons used to prompt oviposition and brood care in B. hortorum were given to all queens that survived to this period (from day 41 post-capture) and produced mixed behavioural responses. Queens were not observed interacting with the cocoons and generally continued their own egg-laying and larval feeding. BhF was given a cocoon on the day her first daughter emerged. Within a week, BhF had (for the first time) ejected her remaining larvae. She did not kill or show any aggression to the B. terrestris worker but laid new eggs 4 days later, which later emerged as males. Three other queens responded negatively to the B. terrestris worker when it emerged. BhD killed it within a day of emergence and BhK and BhE actively avoided theirs, ceasing all nest-associated behaviours until the workers were removed. BhJ was the only queen whose interactions with her B. terrestris worker appeared consistently natural (settled physical contact). No evidence of B. terrestris workers engaging in brood care behaviour when paired with B. hortorum queens was observed.

DISCUSSION
Queens of both species readily laid eggs in the study, with more than three-quarters laying eggs if they survived past day 10. Even B. hortorum queens, who were given no initial stimulus beyond nesting material, had generally laid their first eggs approximately 2 weeks after capture. Despite not being a natural Figure 2 Survival of wild-caught a B. hortorum (n = 11) and b B. pascuorum (n = 57) bumblebee queens reared in captivity to produce colonies. Fourteen per cent of queens died in the first 10 days and 76% of queens that survived this period went on to lay eggs.
Table II. Survival and reproductive success (day of first egg, day of first worker and number of offspring produced), of B. hortorum and B. pascuorum queens kept in laboratory conditions. Egg-laying shown in italics denotes cooperation between queens, where it was not possible to tell if one or both queens contributed to egg-laying. B. pascuorum queens were all initially paired and based on their behaviour, reproductive success and survival were either paired (P) or single (S) at different stages of colony production (see Supplementary Tables I, II and III for all pairings and justifications for keeping some queens paired and giving some their own nest box). If a B. pascuorum queen was paired, the number of pairings she had is shown (P1-3). All queens were initially given cotton wool, pollen and sugar water ad libitum at the start of the experiment, then later, a B. terrestris cocoon to further encourage oviposition and care Trialling techniques for rearing long-tongued bumblebees under laboratory conditions pollen choice for long-tongued bumblebees, the diet was sufficiently good for the queens to produce workers (n = 7 single queens and 3 pairs) and reproductives (n = 3). Very few workers were needed for the colony to produce males, demonstrated by a B. hortorum colony Response of queens to the resulting B. terrestris workers: Bullet represents B. terrestris worker accepted by queen (i.e. not killed and not avoided); square represents queen actively avoided B. terrestris worker until it was removed; diamond represents queen killed her B. terrestris worker or showed aggression (queen BhD) which only produced six workers before producing 38 males. Problems arose for both species during larval development, when queens neglected or ejected their young. Although we cannot exclude eggs or larvae being abandoned due to poor health, we suggest this behaviour was more likely a response to stress or a perceived lack of resources (Smith 1985;Parmigiani and Vom Saal 1994). However, we did not observe any queen consuming their abandoned larvae, or failure of the larvae to feed pollen provided by the queen. Queens of both species were found to commit ovicide and larvicide, and this behaviour did delay worker production noticeably in both species, which has also been observed in trials with B. lapidarius (Bučánková and Ptáček 2012).
The cyclic egg-laying-ovicide behaviour presents multiple opportunities to encourage larval feeding and care. It may also suggest that, while queens may be ready to lay eggs and start a colony, they may also be particularly sensitive to external conditions, such as temperature, humidity and disturbance, which int e r f e r e w i t h r e a r i n g . B e h a v i o u r a l l y, B. pascuorum seemed much less aggravated by the artificial conditions and stimuli provided, while B. hortorum queens responded more sensitively and more aggressively if disturbed. This included pollen feeding, during which the lid of the nest box had to be removed. However, given the frequency of ovicide and larvicide in both species, reducing stress post-egglaying should be a continuing priority in rearing trials. The proportion of B. hortorum queens used in the experiment that went on to produce workers (45%) is very promising. A much l o w e r s u c c e s s r a t e w a s a c h i e v e d f o r B. pascuorum , with only eight queens (14%) being involved in worker production. This may be improved by giving younger cocoons to queens as soon as they have acclimatised to their nest boxes. This could not be done in our experiment due to availability of cocoons, but other studies have demonstrated that younger cocoons and earlier exposure can prompt queens to settle quicker in their boxes-and u l t i m a t e l y p r o d u c e w o r k e r s s o o n e r (Bučánková and Ptáček 2012;Ptáček et al. 2015). Previous rearing trials have generally assumed that the presence of cocoons prompts queens to engage in natural brood care activities. This could be as a result of the cocoon's scent (Heinrich 1974;Gamboa et al. 1987), or slightly higher temperature (Barrow and Pickard 1985). While it is possible the queen might assume the brood are her own, nest intrusions occur naturally in the wild (Lopez-Vaamonde et al. 2004;Goulson et al. 2018), and so the cocoons may instead be perceived as evidence of a female competitor being present. This would explain why queens have also been documented destroying cocoons and may stimulate females to assert dominance over the apparent rival via their own egg-laying, essentially stressed into developing a colony. Assuming that this is the case, there must be factors which discourage queens from performing natural rearing behaviours before the addition of cocoons, but which can be at least somewhat overridden when cocoons are added. Given that queens naturally would selfselect nesting sites, it is possible that they do not recognise their captive surroundings and that the addition of cocoons functions at the very least as an indicator that the surrounding environment is a suitable nest.
As a result of intermittent feeding by the queen, the long-tongued workers (mainly B. pascuorum ) that did emerge were small and may have made little contribution to colony development, suggesting queens may need continued support even after their first workers emerge. We did not screen queens for pathogens and it is possible that parasite infections could have affected their propensity to lay eggs, and the size and health of their larvae. Providing B. terrestris cocoons did encourage B. pascuorum queens to engage in more larval care (Table II) but had less effect of B. hortorum queens. After emergence, most B. terrestris workers made little to no physical contact with the queen or her brood, which is in contrast to previous studies in which B. terrestris workers actively engage in brood care and larval feeding for B. pascuorum and B. sylvarum queens (Ptáček et al. 2015).
We did observe co-operative brood care between paired B. pascuorum queens and up to 17 queens appeared to contribute equally to e g g -l a y i n g a n d b r o o d c a r e . F o r B. pascuorum , the value of intra-and interspecific pairings clearly needs further investigation.
B. hortorum and B. pascuorum are amongst the most generalist of the long-tongued bumblebees. They are common in the UK and share the same association with Fabaceae as those species most in decline, making them suitable models for rearing trials and laboratory experiments. In our study, we found that queens of B. pascuorum and B. hortorum could rear small colonies through to the reproductive stage even when fed pollen they may not naturally collect in the wild. We found that queens lay eggs readily under artificial conditions even without the use of cocoons as stimuli; this might be further improved using CO 2 exposure (Röseler 1985;Tasei 1994). It is also clear that species responded differently to captive conditions, as previously shown in other bumblebee species (Bučánková and Ptáček 2012;Ptáček et al. 2015;Moerman et al. 2016). Repeated egg/larvae abandonment remains a problem. Future trials should test methods to manage queen stress from the initial collection through to colony initiation. Trialling various nest box types and pollen diets will further clarify their nesting preferences (Lhomme et al. 2013;Moerman et al. 2016). Since queens may respond positively to stress-inducing stimuli due to a perceived competitive pressure, techniques to restrict stressful conditions to the first egglaying phase might elucidate this. Future research should examine how pollen type and preparation (e.g. water content) affects larval development. Pollen prepared with a higher liquid content than we used here may be more suitable for manual larval feeding. Although rearing long-tongued bees may require more sensitive, species-specific maintenance in captivity, our results show that it is possible. Given the ecological implications and conservation status of many long-tongued species, further work to refine the protocols is clearly worthwhile.