Rodent odour bait: A new bumble bee conservation tool to enhance nest box occupancy

Bumble bee conservation focuses on supplementing floral resources. Yet, nesting site availability is linked to bumble bee abundance. As a supplement to natural nest sites, nest boxes could be deployed and baited with synthetic lures. As queen bumble bees reportedly establish colonies in abandoned rodent burrows, we hypothesized (1) that queen bumble bees sense, and behaviourally respond to, rodent odour, and (2) that lures of synthetic rodent odour can guide spring queens to nest boxes. We collected headspace odorants from bedding soiled with urine and faeces of house mice, Mus musculus, and identified the 10 odorants that elicited responses from queen antennae. To field‐test attraction of queens to mouse excreta odorants, we tree‐mounted paired nest boxes in florally rich locations, and assigned clean and soiled bedding, respectively, to one box in each pair. Queens established colonies in 17 mouse‐scented boxes and in six unscented boxes. This 43% occupancy rate of mouse‐scented boxes represents a significant improvement over the 10% occupancy rate common for unscented boxes. In a further field experiment, we baited one box in each pair with a synthetic mouse odour lure and found that queens established colonies in 13 baited boxes and in six unbaited control boxes. Specifically, Bombus mixtus established seven colonies in baited boxes and only one colony in an unbaited box. With this proof‐of‐concept that synthetic lures can guide queens to nest boxes, we anticipate that bumble bee conservation programs will soon be able to offer both expanded floral resources and baited nest boxes readily detectable by queens.


INTRODUCTION
Populations of bumble bees in western Europe and North America have declined in recent decades (Cameron et al., 2011;Goulson et al., 2008), with metadata analyses indicating that population trends are highly species-specific (Jackson et al., 2022). Current bumble bee conservation efforts focus on supplementing floral resources but often neglect providing nest sites (Lye et al., 2011) which are key ecological requisites for bumble bees (Kearus & Thompson, 2001;Tscharntke et al., 1998). As bumble bee abundance was found to be positively correlated with the number of rodent holes present in a habitat (McFrederick & LeBuhn, 2006), and as abandoned rodent burrows serve as nest sites for a variety of bumble bee species , offering artificial nest sites (domiciles), such as nest boxes, is an obvious conservation goal. Occupancy rates of domiciles range from 0% to 100% (Johnson et al., 2019; and literature cited therein), with many factors modulating occupancy rates, including geographic range (Europe, New Zealand, North America), habitat quality, prevalence of natural nest sites, time of study (Sladen, 1912 to current), and the design and placement of domiciles. Compared to historical domicile occupancy rates, more recently reported rates seem low (e.g., Barron et al., 2000;Lye et al., 2011) but the reasons are difficult to ascertain. Both the design of nest boxes and the provisioning of insulation material address the bumble bees' need for nesting in a dry and well-insulated cavity (Donovan & Wier, 1978) but the cues queen bumble bees exploit to locate such nest sites are poorly understood (Barron et al., 2000). We predicted that occupancy rates might increase, if nest boxes were baited with lures that attract nest siteseeking queens.
Although it is widely assumed that bumble bees nest in abandoned rodent burrows Fussell & Corbet, 1992;Svensson & Lundberg, 1977), it remains unknown or inconclusive whether queens selectively seek rodent burrows as nest sites (Barron et al., 2000;Hobbs et al., 1960;Sladen, 1912). In an observational study, bumble bees favoured nest sites with grasses from 'field mice' (taxonomic name not provided) (Frison, 1918), and in a controlled experiment bumble bees selected burrows with induced previous occupancy by mice (Fye & Medler, 1954). However, the results obtained by Fye and Medler (1954) could not be substantiated in a repetition of this experiment (Hobbs et al., 1960).
Bumble bees have an exquisite sense of smell (Sprayberry, 2018), and semiochemicals (message-bearing chemicals) play an important role in their life history, mediating-among others-nest entrance marking (Foster & Gamboa, 2010), nestmate recognition (Gamboa et al., 1987), and floral foraging (Leonard et al., 2010). It is entirely conceivable that rodent-derived semiochemicals, such as urine, faeces and fur odorants emanating from burrows, guide Spring queens to rodent burrows. Nest site-searching queens exhibit a characteristic zig-zag flight with an overall forward motion (Kells & Goulson, 2003), reminiscent of odour-tracking behaviour in many animal taxa (Svensson et al., 2014). This flight pattern suggests that queens exploit rodent semiochemicals to locate the often visually obscured entrance holes of rodent burrows.
But even if we were to experimentally demonstrate that queen bumble bees exploit natural rodent semiochemicals for locating nest sites, such a finding would contribute little to bumble bee conservation because natural rodent semiochemicals are not readily available for deployment by conservationists. To optimise the likelihood that queens detect, and adopt, artificial nest boxes for starting their colony, these nest boxes would need to be baited with synthetic lures of rodent semiochemicals. Yet, research in chemical ecology for conservation purposes is lagging. While synthetic semiochemicals are widely used to control pest insects, very few semiochemicals have been developed for biodiversity and conservation studies such as the deployment of pheromones to detect the presence of endangered species and to delineate their distribution range (Larsson, 2016). As our study offered a unique opportunity to apply chemical ecology research for the purpose of bumble bee conservation, we decided to not only test whether queen bumble bees exploit rodent semiochemicals for nest site location but-if so shown-to also identify these semiochemicals for nest box lure development.
Here, we tested two hypotheses: (H1) Queen bumble bees searching for nest sites in spring antennally sense, and behaviourally respond to, natural rodent odour. (H2) Lures of synthetic rodent odour can be used to guide queen bumble bees to nest boxes.
As the source of rodent odour, we selected house mice, Mus musculus, because (i) house mice are prevalent in urban environments (Pocock et al., 2005) where bumble bees thrive (Samuelson et al., 2018), and (ii) many house mouse odorants (sex-attractant pheromone components) have already been identified and field-tested for captures of mice (Musso et al., 2017;Novotny et al., 1985;Takács et al., 2017;Varner et al., 2019Varner et al., , 2020Varner et al., , 2022, thus expediting the development of a synthetic lure as a bumble bee guide to nest boxes.

Experimental house mice as odour sources
As odour sources, CD-1 ® female and male house mice (60-to 90-dayold) were purchased from Charles River Laboratories International Inc. Acquisition of headspace odorants emanating from soiled mouse bedding The procedure was previously detailed (Varner et al., 2019) and is only outlined here. Briefly, corncob bedding (100 g per mouse) soiled over the course of 3 days with urine, faeces, shed fur and skin cells of singly-housed males, or of five group-housed females, was collected and the material from several cages combined and mixed. Onehundred gram aliquots of male-or of female-soiled bedding was then placed into separate Pyrex glass chambers (30 cm Â 15 cm), each connected to a Pyrex glass tube (15 cm Â 15 mm OD) filled with Porapak Q (200 mg) to trap headspace odorants. After drawing charcoalfiltered air through each chamber and the Porapak Q odorant trap at 1 L min À1 for 24 h, Porapak Q-trapped odorants were desorbed in sequence with pentane and ether (2 mL each), dodecyl acetate was added as an internal standard to extracts, and each extract was concentrated to 250 μL. Clean bedding (100 g; control) was subjected to the same procedure.

Gas chromatographic-electroantennographic detection analyses of mouse odorants
To test whether queen bumble bees can sense mouse odorants, we captured wild queens of Bombus vonesenskii and B. mixtus, and we obtained a queen of B. impatiens from a commercially supplied colony.
Because every field-captured queen used in electrophysiological recordings will sensorily be impaired and can no longer establish a colony, we strived to use as few queens as possible to show proof of concept. We also captured and tested wild bumble bee workers of B. vosnesenskii, B. mixtus and B. flavifrons. We tested worker bees in gas chromatographic-electroantennographic detection (GC-EAD) recordings to determine whether the ability to sense mouse pheromone components is conserved in the queens' daughters. Analysing Porapak Q extracts (described above) by GC-EAD and GC-mass spectrometry (MS) allowed us to rapidly locate and identify those rodent odorants that are sensed by queen antennae and that therefore may guide queens in spring to (abandoned) rodent burrows. As general and specific procedures and instruments have previously been detailed (Arn et al., 1975;Gries et al., 2002), they will only be outlined here.
Briefly, the GC-EAD set-up employed a Hewlett-Packard 5890 gas chromatograph (GC) fitted with a DB-5 GC column (30 m Â 0.32 mm I.D., film thickness 0.25 μm). Helium served as the carrier gas (35 cm s À1 ) with the following temperature program: 40 C for 1 min, increasing 10 C min À1 to 280 C. The injector port and flame ionisation detector (FID) were set at 260 C. For each GC-EAD recording (n = 9), we carefully dislodged an antenna from the head of a queen or worker bumble bee and suspended it between two glass capillary electrodes (1.0 Â 0.58 Â 100 mm) adapted to accommodate a bumble bee antenna and filled with a saline solution. The sample size (n = 9) for electrophysiological recordings was kept moderate because capturing not more queens than absolutely necessary was part of an agreement that allowed us access to park and public lands.

Analysis of soiled bedding headspace odorants by GC-MS
Odorants in the headspace of soiled mouse bedding that elicited antennal responses from three queen and worker bumble bees were analysed on a Varian Saturn 2000 Ion Trap GC-MS operated in fullscan electron ionization mode and fitted with a DB-5 MS column (30 m Â 0.25 mm I.D., film thickness 0.25 μm), with helium as carrier gas (35 cm s À1 ). The injector port and ion trap were set at 250 C and 200 C, respectively, and the column oven program was as follows: 40 C for 5 min, then 10 C min À1 to 280 C, held for 10 min. To identify antennally active odorants (see Results), we compared their retention indices (relative to aliphatic alkanes; Van den Dool & Kratz, 1963) and mass spectra with those of authentic standards.

Purchase and syntheses of soiled bedding headspace odorants
Authentic standards were obtained from various sources: 2-heptanone, 4-heptanone, 1-hexanol, 1-octen-3-ol and acetophe- Each week, nest box pairs were monitored for 5 min to determine the bumble bee species occupying boxes. Nine weeks after initiating experimental replicates, one worker from each occupied box was collected while exiting the box and identified to species in the laboratory.
At the end of the season, during 8-16 July, all boxes were dismounted and opened for final recordings of occupancy, colony size, number of brood cells, and evidence for parasitism.
Hypothesis 2. Lures of synthetic rodent odour can be used to guide queen bumble bees to nest sites.

Design of nest box occupancy field experiment 2
The design of field experiment 2 largely followed the design of experiment 1 (see above), with some modifications. During 22-26 March 2021, paired nest boxes (Figure 1a) were mounted on trees in eight F I G U R E 1 Photograph and drawings illustrating (a) the paired nest box design used in two field experiments, and (b, c) a single nest box and its contents (image modified from Wildlife Preservation Canada, 2018) used in experiment 1 (b) and in experiment 2 (c). Boxes were mounted to trees via packaging straps, with 0.4-m spacing within pairs and >2 m between pairs. Each box (20 cm Â 18 cm Â 14 cm; Hobbs et al., 1960) featured a 2-cm hole in the front panel to serve as an entrance (1), and a plastic sheet (2) stapled to the roof to provide protection from rain. All boxes were fitted with unbleached cotton (30 g, 3) as nesting material. Boxes in experiment 1 received-enclosed in a cheesecloth bag (4)-200 g of corncob bedding (5) which was kept clean (control box) or was previously soiled by mice (treatment box). Soiled bedding contained the metabolic waste produced by one laboratory-strain male and female house mouse over the course of 3 days. Boxes in experiment 2 had a 2-cm hole in the bottom panel to accommodate an inverted plastic vial with a flip top lid (89 mm Â 42.9 mm) (6) to facilitate lure replacement and a perforated bottom to enable odour release into the box. The vial contained separate lures for known sex-attractant pheromone components of male mice (7), candidate sex-attractant pheromone components of male mice (8), sex-attractant pheromone components of female mice (9), nonpheromonal urine/faeces odorants of mice invariably eliciting antennal responses from bumble bee queens and workers (10)

Statistical analyses
Predicting that nest box baits enhance nest box occupancy rates, data were analysed with a one-tailed cumulative binominal test, using R (R Core Team 2020, version 4.0.3.). Nest box pairs with either double occupancy (where the sequence of occupancy could not be determined) or with no occupancy (where no queen was present or made a choice) were excluded from analyses.

RESULTS
Hypothesis 1. Queen bumble bees antennally sense, and behaviourally respond to, natural rodent odour.
All these odorants were absent in the extract of clean bedding, except for acetophenone which was present at a 7-fold lower amount. Antennae of queen B. mixtus also responded to a synthetic blend of 1, 7, 9 and 10 ( Figure 3).

Nest box occupancy field experiment 1
Four out of the 46 tree-mounted nest box pairs were tampered with by curious garden visitors and were excluded from data analyses. In  Figure 4). The 8-fold preference of B. mixtus for mouse-scented boxes was statistically significant (p = 0.02), but sample sizes for the remaining species were too small to warrant statistical analyses. With the composition of bumble bee communities in field sites not known neither in experiment 1 nor in experiment 2 (below), it could not be determined whether queens of additional species were absent, were present at low abundance, or were well represented in the community but did not respond to nest box baits.
Hypothesis 2. Lures of synthetic rodent odour can be used to guide queen bumble bees to nest sites.

Nest box occupancy field experiment 2
Of 97 nest box pairs, 16 were tampered with by park and garden visitors and thus needed to be excluded from analyses. Of the remaining 81 nest box pairs, 19 had one of its two boxes colonised. In these pairs, bumble bees established colonies in 13 baited boxes and in six unbaited control boxes (p = 0.084, Figure 5). Among the four species colonising boxes, B. mixtus favoured baited to unbaited boxes (7:1; p = 0.035, Figure 5), whereas the remaining three species did not show a preference. Bombus sitkensis established colonies in each of three baited and three unbaited boxes, B. melanopygus established colonies in each of two baited and two unbaited boxes, and B. flavifrons established one colony in a baited box. The sample size of these three species was too small to warrant statistical data analyses.
As it became evident in early June, all occupied nest boxes (n = 19) were parasitized by larvae of the bumble bee wax moth, Aphomia sociella, much earlier in the season than anticipated. Because of this parasitism, we could neither count the number of brood cells nor measure colony size.
F I G U R E 4 Number of colonies established by Bombus mixtus, B. sitkensis and B. melanopygus in paired wooden nest boxes (see Figure 1a), that were baited (mouse-scented treatment box), or not (unscented control box), with bedding soiled by laboratory-kept male and female house mice, Mus musculus. An asterisk (*) denotes that significantly more colonies were established in mouse-scented boxes (one-tailed cumulative binominal test, p < 0.05); total occupancy refers to colonies from all identified species plus additional colonies that could not be identified to species; occupancy data for B. sitkensis and B. melanopygus were too low to warrant statistical analyses. Bumble bee patterns from Williams et al. (2014) with artwork by Evans et al. (2022).
Abandoned rodent burrows seem to meet essential nest site requirements because many queens establish their colonies in burrows of rodents or shrews such as the wood mouse, Apodemus sylvaticus, common shrew, Sorex araneus, bank vole, Clethrionomys glareolus, and field vole, Microtus arvalis (Fussell & Corbet, 1992;Lye et al., 2012;Sladen, 1912). How queens locate these burrows is not well understood but there are at least two complementary observations suggesting that rodent odorants emanating from burrows guide queens.
Nest site-searching queens exhibit a characteristic zig-zag flight pattern with an overall forward motion (Kells & Goulson, 2003) which is reminiscent of odour-tracking behaviour in many animal taxa (Svensson et al., 2014). This flight pattern suggests that queens exploit rodent odorants to locate the often visually obscured entrance holes of rodent burrows. Direct evidence that rodent odorants guide Spring queens to rodent burrows was presented in an experimental study showing that burrows had higher bumble bee occupancy when they were recently occupied by mice (Fye & Medler, 1954) but these results could not be repeated (Hobbs et al., 1960).
To test the hypothesis that Spring queens indeed antennally sense, and behaviourally respond to rodent odorants, we conducted both laboratory and field studies. As odorants from wild mice are difficult to acquire for testing in electrophysiological and behavioural experiments, we relied on laboratory-strain mice on the assumption that their odorants closely resemble those of their wild counterparts.
For an odorant source, we used bedding material soiled by laboratorykept house mice, drawing on our experience that soiled bedding emanates a plethora of mouse urine and faeces odorants, including sex pheromone components (Varner et al., 2019;unpublished). Analysing headspace odorants of soiled bedding by GC-EAD revealed as many as 10 odorants that elicited responses from bumble bee queen and worker antennae ( Figure 2). Remarkably, four of these odorants (1-hexanol, 3,4-dehydro-exo-brevicomin, 2-sec-butyl-4,5-dihydrothiazole, 2,3,5-trithiahexane) are sex pheromone components of male mice (Novotny et al., 1985;Varner et al., unpublished) and one odorant (4-heptanone) is a sex pheromone component of female mice (Varner et al., 2019). That bumble bee antennae accommodate olfactory receptors tuned to rather unique rodent pheromone components implies a functional role of these odorants in the context of locating rodent burrows as nest sites. After all, sensory receptors are costly to maintain (Niven & Laughlin, 2008) and the physical space taken up by rodent pheromone receptors could otherwise be occupied by receptors facilitating floral foraging (Leonard et al., 2010) or social communication between nest mates (Gamboa et al., 1987).
Experimental evidence that the queens' ability to sense rodent odorants improves their success in finding rodent burrows could be demonstrated conclusively only in field experiments. In the first experiment, we substituted rodent burrows with nest boxes, and presented natural rodent scent in the form of bedding soiled by laboratory-kept mice. In each of 46 replicates, we offered Spring queens a choice between a 'mouse-scented' box and an unscented control box, monitoring nest site selection and colony establishment throughout the breeding season. The 3-fold higher occupancy rate of 'mouse-scented' nest boxes (Figure 4) provides convincing evidence F I G U R E 5 Number of colonies established by Bombus mixtus, B. sitkensis and B. melanopygus in paired wooden nest boxes (see Figure 1a). In each pair, one box was baited with a synthetic lure of house mouse, Mus musculus, odour ( Figure 1c) and the other left unbaited. An asterisk (*) denotes that significantly more colonies were established in mouse-scented boxes (one-tailed cumulative binominal test, p < 0.05); occupancy data for B. sitkensis, B. melanopygus and B. flavifrons were too low to warrant statistical analyses. that rodent odorants serve as a 'road map' to rodent burrow-seeking queens. Preferential selection of mouse-scented boxes by queen B. mixtus, B. melanopygus and B. sitkensis (Figure 4) further indicates that rodent odorants are a universal nest site location cue. All but one of the species present in our field sites more readily located, or preferentially chose, the mouse-scented boxes.
Our findings that mouse-scented nest boxes had significantly higher bumble bee occupancy (43%) than unscented control boxes (17%) offer new opportunities for bumble bee conservation. Our experimental data show that rodent odorants inform Spring queens about potential nest sites and that these odorants enhance occupancy rates of nest boxes. Soiled house mouse bedding, however, is an impractical solution as a bait for bumble bee nest boxes and must be replaced with a synthetic lure for large-scale bumble bee conservation. To this end, we prepared a synthetic lure that-conservativelyincluded all currently known or putative sex-attractant pheromone components of house mice, all non-pheromonal constituents in soiled house mouse bedding that bumble bee antennae sensed (Figure 2), and even ammonia as a universal gas emanating from mammalian urine that blood-seeking horse flies and mosquitoes exploit during host foraging (Kristensen & Sommer, 2000;Venkatesh & Sen, 2017).
We took this 'all-inclusive' approach for lure preparation to maximise the likelihood of lure attractiveness.
Our data provide proof of concept that synthetic lures can be developed for guiding nest site-seeking queen bumble bees to nest boxes. That the synthetic lure was less effective than natural rodent odour may be due to either components still missing from the lure or to suboptimal release dynamics of lure constituents. For example, major urinary proteins in urine deposits of mice bind to sex-attractant pheromone components and facilitate their slow release (Robertson et al., 1993), thus prolonging the effectiveness of pheromonal urine markings (Armstrong et al., 2005). To help ensure that a synthetic nest box lure is adopted widely for bumble bee conservation, the lure composition must be simplified and effective systems for disseminating lure constituents must be designed. Determining the key lure constituents is the first step towards lure development. The sex-attractant pheromone components of mice may be these key constituents because sex pheromones, in general, are 'designed' to stand out, and persist, in chemically noisy settings, thereby enhancing the likelihood of detection by intended receivers (Peake, 2005). This conspicuousness, however, makes pheromones also susceptible to interception, or eavesdropping, by other community members (Peake, 2005), such as queen bumble bees seeking nest sites. Alternatively, generic urine cues such as ammonia may drive queen attraction. In any case, a rigorous systematic approach, such as testing the effect of partial (incomplete) lures with specific lure constituents omitted, is needed to determine the lure constituents that attract queen bumble bees. Once these constituents have been determined, their formulation and sustained release can be studied towards lure development.
In conclusion, we have demonstrated that soiled rodent bedding disseminating mouse pheromone components can guide queen bumble bees in spring to prospective nest sites and that bumble bee antennae respond to mouse pheromone components. That the synthetic rodent pheromone components are not yet as effective as the natural components could be due to missing pheromone components in the synthetic lure, suboptimal pheromone formulation, or both. Our study provides strong impetus for continued bumble bee chemical ecology research to enhance bumble bee occupancy of artificial nest boxes and ultimately to support bumble conservation efforts.