A Novel Behavioral Assay to Investigate Gustatory Responses of Individual, Freely-moving Bumble Bees (Bombus terrestris)

Generalist pollinators like the buff-tailed bumble bee, Bombus terrestris, encounter both nutrients and toxins in the floral nectar they collect from flowering plants. Only a few studies have described the gustatory responses of bees toward toxins in food, and these experiments have mainly used the proboscis extension response on restrained honey bees. Here, a new behavioral assay is presented for measuring the feeding responses of freely-moving, individual worker bumble bees to nutrients and toxins. This assay measures the amount of solution ingested by each bumble bee and identifies how tastants in food influence the microstructure of the feeding behavior. The solutions are presented in a microcapillary tube to individual bumble bees that have been previously starved for 2-4 hr. The behavior is captured on digital video. The fine structure of the feeding behavior is analyzed by continuously scoring the position of the proboscis (mouthparts) from video recordings using event logging software. The position of the proboscis is defined by three different behavioral categories: (1) proboscis is extended and in contact with the solution, (2) proboscis is extended but not in contact with the solution and (3) proboscis is stowed under the head. Furthermore the speed of the proboscis retracting away from the solution is also estimated. In the present assay the volume of solution consumed, the number of feeding bouts, the duration of the feeding bouts and the speed of the proboscis retraction after the first contact is used to evaluate the phagostimulatory or the deterrent activity of the compounds tested. This new taste assay will allow researchers to measure how compounds found in nectar influence the feeding behavior of bees and will also be useful to pollination biologists, toxicologists and neuroethologists studying the bumble bee's taste system.


Introduction
Plant-pollinator interactions are complex. Pollinators visit flowers to obtain nectar and pollen as food; in turn, pollinators facilitate sexual reproduction in plants. While this relationship is mostly mutualistic, floral nectar and pollen sometimes contain toxins or other plant compounds [1][2][3][4][5] which can harm pollinators. The ecological rationale for the presence of such compounds in nectar and pollen is not clear in all settings. One outstanding question in this field is how pollinators such as bees can detect and avoid flowers with nectar containing toxins.
The bumble bee species, Bombus terrestris (Linnaeus, 1758), is a generalist pollinator that visits the flowers of many plant species including those producing nectar containing toxins 6 . Bumble bees have been shown to avoid consuming solutions containing high concentrations of toxins in a 24 hr two-choice assay 7 . This assay of food consumption described by Tiedeken et al. 7 revealed that bees can detect bitter compounds in solutions. However, this assay was unable to distinguish taste from post-ingestive processes such as malaise that could also affect feeding behavior over this time interval [8][9][10] .
Bees can be restrained in individual harnesses and then stimulated to produce the feeding reflex as an assay of their ability to taste compounds 18,19 . Others have modified the PER assay to study the sensitivity of the antennae or mouthparts to toxins 9,20 . However, bees are subjected to stress during harnessing. This could affect how they respond to compounds 21 .
Here, a new assay is described to assess the behavioral taste response of freely-moving bumble bees to sucrose and quinine, an alkaloid that has previously been reported to be deterrent 9 and toxic 10 to honey bees (Apis mellifera) and bumble bees (Bombus terrestris) 7,22 . Although quinine has not been found in plant nectar, this alkaloid is often used as an aversive stimulus in behavioral and physiological studies in bees

Representative Results
The novel assay is used to test the feeding responses to 1 M sucrose, 1 M sucrose solution plus 1 mM quinine and deionized water alone. The immediate feeding responses to each treatment are determined by quantifying the duration of proboscis contacts with the test solution, the frequency of the feeding bouts and the speed of the proboscis retracting away from the test solution after the first contact during the 2 min test phase. The volume of solution consumed is also measured after the test phase. In this study, we have chosen a bout criterion interval of 5 sec (Figure 1, see Supplemental File) based on previous work by French et al. 25 who used a 5 sec threshold to characterize the proboscis retraction behavior by Drosophila in response to deterrent compounds 25 . Thus, we defined a feeding bout as a contact between the extended proboscis and the solution not interrupted by an absence of contacts of 5 sec or more.
In comparison to sucrose and deionized water alone, adding quinine to sucrose solution evidently deters feeding by bumble bees as they will rapidly move away if they detect an aversive substance (Video Figure 1).
In this experiment, the treatments have a significant effect on the cumulative duration of proboscis contacts during the test phase (ANOVA on the log-transformed data, F 2,31 = 41, p <0.001). The cumulative duration of contact time with sucrose containing quinine is significantly reduced in comparison to sucrose alone (p <0.001) but not to deionized water alone (p = 0.219) (Figure 2). Similarly, the treatments have a significant effect on the cumulative duration of feeding bouts (ANOVA on the log-transformed data, F 2,31 = 27.95, p <0.001, Figure 3A). The cumulative duration of feeding bouts with sucrose containing quinine is significantly reduced in comparison to sucrose alone (p <0.001) but not compared to deionized water alone (p = 0.41). The treatments have also a significant effect on the frequency of feeding bouts (Poisson GLM with a log link function, change in deviance compared to the c 2 distribution: p <0.050), whereby the number of bouts with sucrose containing quinine is significantly higher in comparison to sucrose (p <0.01) but marginally significantly different to the deionized water treatment (p = 0.055, due to one bumblebee displaying seven feeding bouts on water, Figure 3B). Likewise, the speed of proboscis retraction differs significantly between treatments (ANOVA on the log-transformed data, F 2,31 = 5.12, p <0.050). Bumble bees retract the proboscis away from the test solution significantly faster after the first contact with sucrose containing quinine than with sucrose or deionized water alone (p <0.050, Figure 3C). These results suggest that quinine triggers an active avoidance behavior in bumble bees. The treatments also have a significant effect on the total volume of solution consumed (ANOVA on the log-transformed data, F 2,32 = 62.5, p <0.001), whereby the consumption of sucrose containing quinine is reduced in comparison to sucrose (p <0.001) but not to deionized water (p = 0.457) ( Figure 4B). The volume of solution evaporated from the capillary during the test period is negligible. At laboratory conditions (25 ±2°C and 28 ±2% RH), the evaporation varies between 0.033 to 0.883 µl with an average of 0.276 µl and 0.171 µl for deionized water and 1 M sucrose respectively.
In this assay contacts between the antenna and the test solution cannot be prevented. Nonetheless, the percentage of bumble bees using their antennae to taste the feeding solution during the test phase (sucrose: 46.1%, sucrose plus quinine: 60.0% and deionized water: 33.3%) is not significantly different between the treatments (binomial GLM, change in deviance compared to the c 2 distribution: p = 0.450). No effect of the treatments is found on the latency between the first antennal contacts and the test solution and the first contacts of the proboscis (median: 2.67 sec for sucrose; 1.10 sec for sucrose plus quinine; 0.80 sec for deionized water, ANOVA on the log-transformed data, F 2,13 = 0.620, p = 0.550).
In addition, the percentage of bumble bees extending the proboscis to taste the test solution remains constant across the treatments (sucrose: 66.7%; sucrose plus quinine: 50.0%; deionized water: 52.2%; binomial GLM, change in deviance compared to the c 2 distribution: p = 0.840).
Together these results suggest that the antennae play a minor role in the detection of the toxins in this assay. In summary, a 2 min test period is sufficient to assess the effect of the solution on the total amount of food consumed by bumblebees and the deterrent effects of toxic or repellent substances in this assay. Thus, by measuring food consumption and assaying feeding behavior, it is possible to correlate total food consumption to the fine structure of feeding during the assay.

Discussion
With this novel behavioral assay, quinine is shown to deter feeding of the buff-tailed bumble bee. The reduced proboscis contact time and feeding bout frequency with water or the sucrose solution laced with quinine is interpreted here as a refusal to initiate further feeding on nonnutritive or potentially toxic solutions. When quinine is added to 1 M sucrose solution, bumble bees not only reduce the volume of solution they consume, they also retract the proboscis faster, thus reducing contact time between the mouthparts and the solution containing a toxin. Together, these results suggest that quinine is perceived by the gustatory receptor cells on the mouthparts of the bumble bee, as already previously identified in the honey bee 9 . Quinine is a toxin for insects that induces malaise-like behavior in the honey bee 10 and knockdown in the malaria mosquito (Anopheles gambiae) 23 . This assay could well lead to the identification of some deterrent and potentially toxic compounds that are perceived by the taste receptor cells on the mouthparts in the bumble bees.
It is crucial for the microcapillary tube to be filled with a sufficient volume of test solution to last throughout the test phase. It is recommended that at least around three quarters of the microcapillary tube (e.g. 70-80 µl) is filled. However, care should be taken to not completely fill the microcapillary tube to reduce the risk of spillage during the process of scanning and attaching the microcapillary tube to the experimental apparatus. Care should also be taken when presenting the 500 mM sucrose droplet to the bumble bee, so that the experimenter avoids leaking the droplet into the holding tube. The 4 mm hole at the tip of the holding tube is large enough for an adult worker bumble bee to naturally extend its proboscis toward the test solution. However it is possible that bumble bees can taste the solution with their antennae before extending their proboscises. This might affect the probability of proboscis extension as PER could be elicited in bumble bees by stimulating their antennae with a sugar solution 15 . In fact the antennae of Hymenoptera like the parasitoid wasp (Trissolcus brochymenae) 24 or the honey bee 13 are equipped with taste sensilla, allowing them to taste sugars and toxins like quinine. Consequently, initial antennal contacts with solutions containing highly deterrent compounds like quinine could also reduce the motivation of a bumble bee to extend its proboscis and therefore affect the experimental success rate. Although antennal contact with the test solution cannot be controlled, in the present study we did not find any significant effect of antennal contact on proboscis extension toward the test solution. In this assay, immediately setting up the microcapillary tube after the pre-test phase when the bumble bees' antennae are still within the holding tube can reduce the opportunity for the bumble bees to taste the test solution with their antennae.
The main limitation of this assay arises when tracking the proboscis retraction away from the test solution after the first proboscis contact using the motion tracking video software. The video footage only displays 2D movement of the proboscis, so the given output of the speed measurement can be under or over estimated. However with some modifications, this aspect of the assay could be improved.
This assay can be used to observe natural feeding responses toward solutions containing different compounds including natural-occurring plant secondary metabolites. Observing the immediate feeding responses with this assay gives detailed information about how bumble bees detect these compounds. This is advantageous over existing 'go-no go' methods like PER 18,19 and over two-choice assays 7 because this method produces several behavioral response measures including food consumption during a discrete feeding bout.
Measuring several parameters simultaneously allows a better evaluation of the palatability of a compound. For example in our assay, bumble bees avoid consuming water or the sucrose solution laced with quinine. Retraction of the proboscis could be caused by a change in the responses of the sugar receptor cells 12,13 . Our assay shows that bumble bees retract the proboscis faster after contacting the sucrose plus quinine solution than water alone; this could suggest that quinine affects a distinct set of neurons in addition to inhibiting sugar sensing neurons 9,12,13,25 .
Our assay permits the analysis of the temporal pattern of behavioral responses during feeding. A similar protocol where the consumption time and the number of bouts is measured has already been implemented to evaluate the feeding response of Drosophila to nutritive and non-nutritive sugars 26 . We predict that bees will exhibit a more reliable response to feeding stimulants in our assay than in other methods such as PER because the bees are free to move in the holding tube 21 . This technique will permit an exhaustive analysis of the taste thresholds for nutrients and toxins to illuminate the mechanisms of feeding in bumble bees and potentially other bee species.

Disclosures
The authors declare no conflict of interest.