Animals

All bees were collected from hives housed at the University of Otago, New Zealand, from colonies typically containing several tens of thousands of worker bees, brood and a naturally-mated queen. All experiments were performed on Varroa-free bees, as the parasite had not yet infested this region at the time of the experiments. Bees performing tasks characteristic of different age groups (cell cleaning, guarding or returning from foraging) were collected directly from hives. Cell cleaning and guarding are typically performed by young and middle-aged bees respectively, whereas foraging is normally performed by the oldest bees in the colony [10], [17]–[19]. Very young bees can be identified by their lighter color, their relatively slow movements and their poorly-developed flying skills. Bees showing these characteristics that had their heads in empty cells were classified as cell cleaners. Bees that were present at the hive entrance prior to any disturbance and reacted to forceps with biting, raised forelegs and alarm pheromone emission were classified as guards. Bees returning to the hive with full pollen baskets were identified as pollen foragers.

To collect bees of known age, brood frames from the hives were placed under constant darkness in a humidified incubator at 33°C. Each sampling day, during a fixed 4-hour time window (12:00 h–16:00 h) adult bees were collected as they emerged from brood cells. Newly-emerged bees were marked with paint on the thorax so that their age could be determined at a later date, and placed in cages (11×9.5×7.5 cm) containing approximately 50 bees of mixed age. To avoid bias due to seasonal variation, bees were raised to different ages in parallel rather than sequentially. Half of the cages contained a strip impregnated with QMP (BeeBoost, PheroTech 2.5 queen equivalent per cage, changed weekly [20]). Cages with and without QMP strips were maintained in separate incubators (31°C, 70% humidity). Each cage contained a sheet of wax and bees were provided also with food (pollen, honey and sucrose mixture) and water ad libitum.

Stress protocol

Bees were prepared for behavioral experiments immediately after being collected either from the hive (cell cleaners, guards and foragers), or from cages when they were 2-, 4-, 6-, 8- or 16-days old. Bees were handled as described in [9]. Briefly after being cooled down on ice for a few minutes, they were individually harnessed in holders that allowed free movements of the mouthparts and antennae. The bees were starved for 4 to 5 hours to enhance motivation, then exposed for 30 minutes either to IPA or vehicle alone (oil) before undergoing appetitive olfactory conditioning. For IPA exposure, the same protocol was used as described elsewhere [9]; harnessed bees were placed individually in a 35 ml glass vial containing a small piece of filter paper soaked with 25 µl of IPA (24% in paraffin oil, Sigma-Aldrich). Controls were exposed in the same way to vehicle alone (paraffin oil). The closed vials containing the bees being treated with IPA were placed in an air exhaust system throughout the 30 min of exposure period to avoid contamination between groups. Following IPA (or oil) exposure, bees were allowed to recover for 30 min before conditioning.

Sucrose responsiveness

Age, QMP exposure and IPA can all affect sucrose responsiveness [7], [9], [28]–[30], [32]. For this reason, only bees that responded to sucrose stimulation consistently throughout each experiment were included in the analysis. While sucrose response thresholds were not measured directly in our experiments, the percentage of bees in each group that failed to respond consistently to 50% sucrose during conditioning, and/or in the memory tests, was recorded. In the group exposed to neither pheromone, 11% of the bees were discarded. In bees exposed to IPA, 21% of the bees were discarded. In the group reared with QMP, 18% failed to respond, whereas 22% of bees exposed to both pheromones were discarded.

Appetitive olfactory conditioning

The proboscis conditioning paradigm developed by Bitterman and colleagues [21] was used in this study. Each bee received 3 paired presentations of an odor (1-nonanol, 5 µl, pure, Sigma-Aldrich) and sucrose (50%, w/w in water) with a 10 minute interval between each conditioning trial. Prior to conditioning, the conditioned stimulus (CS, 1-nonanol), elicited no response, whereas 50% sucrose (the unconditional stimulus US), applied to the antennae elicited a reflexive proboscis extension response. Any bee that failed to respond reflexively to the US was discarded. Odors were delivered by means of a 20 ml syringe containing a piece of filter paper soaked with 5 µl of pure odorant. To avoid odor contamination, conditioning trials were performed in front of an air exhaust system. Bees were placed in the setup 15 seconds before the CS was presented, for 4 seconds. Sucrose was presented 3 seconds after odor onset and was delivered first to the antennae to elicit proboscis extension and then to the proboscis. Bees were allowed to lick the sucrose solution for 3 seconds. After conditioning, bees that show learning extend their proboscis in response to the odor before the reward is delivered (conditioned response CR). The percentage of bees showing the conditioned response is used as a measure of learning. This assay was used to measure the effect of IPA-induced stress on appetitive learning performances.

Memory test

One hour after the end of the last conditioning trial, bees were placed again in the learning area and presented with the CS alone, without reinforcement. The percentage of bees displaying a conditioned response (CR) was recorded. At the conclusion of each experiment, the proboscis extension reflex was tested again. Bees that failed to respond to sucrose stimulation of the antennae were discarded from the experiment.

Statistics

The data were analysed with generalised linear mixed effects modelling (GLMM) using the R package lme4 [22]. To improve the convergence of the acquisition model, only the second and third trials were analysed (as all groups showed the CR 0% of the time on the first trial). The predictor “Trial” was rearranged such that the third trial was set to zero (as Trial 3), thus becoming the intercept of the model. In this way, model predictors provided information about the probability of responses on the last trial. Slopes for the response of each individual bee (Trial|Subject) across the two trials were included as random effects in this GLMM.

In the first experiment examining learning performance of bees belonging to different behavioral castes, GLMM was used to investigate the fixed effects of conditioning trial (Trial), IPA exposure and Caste, as well as interaction terms for pheromone treatment and Caste. The interaction term was included, as it markedly improved the convergence of the model.

GLMM was also used to investigate the fixed effects of Trial, Age, IPA exposure and QMP-rearing in laboratory-raised bees of different age, as well as all interaction terms for pheromone treatment and age (IPA exposure × QMP-rearing × Age). The interaction term was included, as it markedly improved the convergence of models. Thus, the effects of these variables over the third trial of the conditioning period could be assessed. With these estimates of the final effects of these variables, comparisons could be made both within and between treatment groups.

The probability of bees displaying a conditioned response one hour after the conclusion of conditioning, during the retention test, was examined for both experiments also using a GLM in which all possible terms for pheromone exposure and age (or caste) were included, as in the GLMM.

We also used GLM to analyse the effect of pheromone treatments on two broad age groups; young bees (≤6 d) and bees 8-days and older. The third conditioning trial only was examined for this purpose. Data obtained from bees 6 days of age or less were pooled and compared to pooled data from bees 8 days of age or more. The interaction of three factors (IPA exposure × QMP-rearing × Age) was included.

All statistics were generated in R [23], and results are presented as coefficients ± standard error.

Multiple factors affect learning performance

Consistent with age-related changes in learning behavior described in earlier reports [25]–[29], learning performances were found to be stronger overall in guards and foragers than in the young bees tested in this study, including cell cleaners collected from a queen-right colony. Nonetheless, differences in the learning performances of 16-day old bees and guards were unexpected because these bees were assumed to be similar in age [10], [17]–[19]. However, learning performances can be influenced also by factors such as sucrose responsiveness [30], hormone titers [31] and motivation [32], and it is possible that the learning levels observed in this study were influenced by factors such as these. For instance hormone titers vary greatly during the first days after adult emergence [17], [33]–[34] and may account for the differences in learning performances between bees of different ages observed in the present study. Artificial rearing might also influence learning levels. For example, cage-reared bees were fed ad libitum which may have reduced their motivation to learn [35]. The absence of brood pheromone in the cages might also modify hormone titers in these artificially raised bees [36] and indirectly also, their learning behavior. Nonetheless, despite the artificial environment experienced by laboratory-reared bees, all of the age- and behavioral groups examined in this study showed significant learning and within each group, performances in the 1-hour memory test were highly consistent. Effects of IPA on learning behavior, described previously in foragers [9], were found to be remarkably robust also in this study.

Young bees raised without QMP show strong reactivity to IPA

As young bees performing tasks within the colony generally ignore the “call to arms” signaled by the release of alarm pheromones [11], [24], [37], we were intrigued to discover that learning in young bees could be strongly affected by IPA. As IPA has been shown in previous studies to reduce sucrose sensitivity in bees [7], [9] differences in responsiveness to sucrose (the unconditioned stimulus) can influence learning performance [30], [32], [38], motivation for sucrose was carefully controlled in this study. Only bees that responded consistently to sucrose stimulations were used to examine learning performance. We felt justified in using this approach because IPA-treated bees that respond well to sucrose have been shown to have the same sensitivity to sucrose as controls [9]. Furthermore, there was no clear correlation between the number of bees responding to sucrose and the learning performance in each group. Moreover, in contrast to shifts in sucrose sensitivity [32], IPA altered learning, but not 1-hour memory. This suggests to us that IPA-induced changes in sucrose sensitivity do not account for IPA's ability to decrease learning in worker bees.

IPA affects short-term but not medium-term appetitive memory

The discovery that IPA suppresses responses during the acquisition of short-term memory (STM), but not 1-hour memory retrieval is intriguing. This is especially so, as IPA's effects on appetitive learning have been shown to persist for at least 24 hours [9]. How can IPA's selective effects on learning performance be explained? We propose that physiological stress responses induced by IPA interfere with cellular and molecular processes that underlie STM retrieval, leading at the behavioral level to an apparent reduction in acquisition rate. Our results indicate that medium-term memory (MTM), on the other hand, is resistant to physiological changes induced by IPA.

In bees, as in many vertebrate and invertebrate species, short-, medium- and long-term memories differ not only in time course, but also in their molecular underpinnings [39]–[42]. In a series of elegant studies, Müller and colleagues showed that sucrose stimulation of the antenna causes a transient rise in the activity of the cAMP-dependent protein kinase A (PKA) in the antennal lobes of the bee brain [43]. They showed moreover, that a single pairing of an odor with sucrose extends the duration of PKA activation [44], a process that may potentially be slowed by IPA. The short-term memory that forms as a result of a single conditioning trial is unstable and highly susceptible to disruption [45]–[47]. However, multiple conditioning trials lead to the repeated updating of STM and the formation of stable, longer-lasting memories [42], [45], [48], [49]. Prolongation of PKA activity in the ALs has been shown to be critical for long-term memory formation in bees [44], whereas consolidation of MTM from STM requires ongoing neural activity and the formation of PKM, a constitutively activated form of protein kinase C [50]. Our results suggest the latter is resistant to modulation by IPA because irrespective of whether or not bees were exposed to IPA and/or QMP, bees of the same age exhibited strikingly similar response levels in the 1-hour memory test. It will be interesting in future studies to determine whether IPA-induced decrease of STM affects LTM performance and whether, as predicted, learning-related events in the antennal lobes of the brain are modulated by IPA. Mushroom bodies are intimately involved also in the formation and retrieval of associative olfactory memories [51], [52]. It is entirely possible that IPA-induced stress responses have an impact also on the functioning of learning-related circuits in these so called ‘higher’ centers of the brain.

Stress reactivity in young bees is reduced by QMP

Our results show that young worker bees exposed to QMP respond less to the key component of alarm pheromone, IPA. The calming effect of the presence of the queen (and in particular her pheromonal bouquet) is well known to beekeepers: a hive that lost its queen will have more aggressive workers [53]. Her influence might help explain the low behavioral and olfactory responsiveness to alarm pheromone observed in young bees from queen-right colonies [11], as well as their lower levels of aggression [37]. Furthermore, QMP impairs aversive learning in young bees [15]. Taken together, these results suggest to us that QMP reduces stress reactivity in young worker bees.

QMP had no significant effect on appetitive learning performance in bees up to 6 days after their emergence as adults. However, around this time, bees appeared to go through a transition during which time they were more responsive to QMP and less responsive to IPA. Whether this is an artifact of rearing bees under laboratory conditions, or true also of bees in a normal queen right colony remains to be determined. What is clear from this study is that in young bees, inhibitory effects of IPA on learning performance are suppressed by QMP. Learning in cell cleaners, for example, which we predict to be between 1 and 5 days of age [10], [17]–[19], was not affected by IPA. Our results with bees of known age suggest that stress reactivity in cell cleaners is likely to have been inhibited by exposure to QMP within the colony.

Interactions between IPA and QMP change with age

Responses of 8- and 16-day old bees to QMP were clearly different from those of bees less than 1 week old. In 8-day olds, QMP improved the learning performance of control bees, while IPA exposure failed to have any effect. This age group differed both from the young laboratory-raised bees and the naturally-reared pollen foragers and guards, and our results suggest that an important transition occurs at around 8-days of age (see also [15]). In the older age groups tested, the effects of IPA and QMP, as well as the interplay between these two pheromones, were markedly different to those observed in younger bees. Only older bees reared in presence of queen cues were affected by the alarm pheromone component, IPA. Importantly, pollen foragers and guards, whose learning performance was strongly affected by IPA exposure, came from a queen-right colony. We assume therefore that these bees would have been exposed to queen pheromone. Whether QMP exposure promotes reactivity to IPA at an older age remains to be determined. Age-related shifts in responses to QMP are well documented [10], [15], [54], [55], and correlate strongly with changes both in hormone titers [17], [31] and levels of gene expression in the brain [56]. Interestingly, QMP has been found to inhibit the synthesis of juvenile hormone [12], [13], a hormone that plays a significant role in honey bee behavioral development [17],[18]. While this may provide a clue as to how IPA-induced stress responses are blocked by QMP, the mechanisms involved remain unclear. One component of QMP (homovanillyl alcohol, HVA) has been found to suppress aversive learning in young worker bees [15], and to activate the D2-like dopamine receptor, AmDOP3 [57]. The functional consequences of AmDOP3 receptor activation by HVA have yet to be fully resolved, but it is possible that this QMP component is involved also in suppressing stress responses induced by IPA in young worker bees. The DOP3 receptor has recently been found to be expressed in the antennae [58], as well as in antennal lobes and mushroom bodies of the brain [59], [60]. As IPA and components of QMP are detected by olfactory receptor neurons housed in the antennae, QMP modulation of responses to IPA could potentially occur at this level [61], [62], but why this is age dependent remains unclear.

Conclusions and predictions arising from this study

Our results show that IPA induces stress responses in young worker bees that can be blocked by QMP. Responses to these pheromones allow us to make several important predictions relating to the cellular and molecular mechanisms that support reward learning in honey bees. We predict firstly, that responses to acute stress modulate neural circuits involved in STM recall, but leave intact processes that underpin the formation and retrieval of MTM. We predict also that in young bees, QMP lowers responsiveness to IPA, a shift that should be detectable at the level of the antennal lobes. It will be interesting in future studies to test the strength of these predictions.