Social signals and aversive learning in honey bee drones and workers

Abstract

The dissemination of information is a basic element of group cohesion. In honey bees (Apis mellifera Linnaeus 1758), like in other social insects, the principal method for colony-wide information exchange is communication via pheromones. This medium of communication allows multiple individuals to conduct tasks critical to colony survival. Social signaling also establishes conflict at the level of the individual who must trade-off between attending to the immediate environment or the social demand. In this study we examined this conflict by challenging highly social worker honey bees, and less social male drone honey bees undergoing aversive training by presenting them with a social stress signal (isopentyl acetate, IPA). We utilized IPA exposure methods that caused lower learning performance in appetitive learning in workers. Exposure to isopentyl acetate (IPA) did not affect performance of drones and had a dose-specific effect on worker response, with positive effects diminishing at higher IPA doses. The IPA effects are specific because non-social cues, such as the odor cineole, improve learning performance in drones, and social homing signals (geraniol) did not have a discernible effect on drone or worker performance. We conclude that social signals do generate conflict and that response to them is dependent on signal relevance to the individual as well as the context. We discuss the effect of social signal on learning both related to its social role and potential evolutionary history.

Keywords: Alarm pheromone; Associative learning; Drone; Honey bee; Semiochemicals; Social communication.

Fig. 1.

Correlation analysis of learning response in honey bee workers under continuous exposure to increasing doses of IPA. Here we show the correlative relationship between IPA dose exposure and response. Dots represent the group means, lines indicate each group's 95% confidence interval. The shaded region represent the mean of the control group (solid horizontal line) bounded by its 95% confidence interval (dashed lines). The solid diagonal line illustrates the linear relationship of the correlation between the log base 10 of the IPA sting equivalent dose and our response metric. Results show a highly significant negative association between the two variables. Statistical test was a linear correlation of the proportion of time on safe side following the first error and the log-transformed sting equivalent dose.

Fig. 2.

Effect of incremental doses of continuously presented IPA on worker learning. Performance is measured for each individual as proportion of time spent on the safe side of the apparatus during the first (acquisition) trial of aversive training following the first experience of punishment (mild shock). Bars indicate means and lines illustrate the 95% confidence interval for each group tested. Letters above the bars display statistical relationships between dosage groups. Numbers inside the bars are group sample sizes. Results show that the greatest statistical significance detected was between the 1 and 100, and the 10 and 100 Sting Equivalent Dose (SED) groups. Statistical test was a two-way analysis of variance on the logit transformed proportion of time on safe side following the first error. This was followed by a post-hoc Tukey's test of pairwise comparisons. Assessment is detailed in the ‘Data analysis’ section of the Materials and Methods.

Fig. 3.

Effect of variable dose levels of continuously presented IPA on drone response. Provided is a summary of the statistical relationship between IPA dose groups and drone response in our learning paradigm. The Y axis represents proportion of time spent on the safe side following the first error, the X axis outlines our treatment levels of IPA sting equivalent doses. Bars represent group means, lines the 95% confidence interval. Numbers inside the bars represent group sample sizes. Results show that there were no statistically significant differences in drone performance between any of the groups.

Fig. 4.

Learning performance of workers under continuous exposure to various odor cues and signals. The graph shows the statistical relationship of various continuously presented odors on the learning performance of honey bee workers. The Y axis shows the total time spent on the safe side of the electric shock avoidance assay following the first error. The X axis outlines the type of odor provided as well as the corresponding volume presented in brackets. Numbers inside the bars correspond to sample sizes. Results show that no significant differences in worker learning response were observed between control and treatment groups.

Fig. 5.

Effects of alternate odors on drone response during a punishment with discrimination learning assay. Here we summarize the statistical relationship in learning performance between groups under continuous exposure to various alternate odors. The Y axis represents our response metric, while the X axis provides odor presented and volume of presentation in brackets. Bars indicate sample means, lines represent the 95% confidence interval of each group. Numbers inside the bars correspond to sample sizes. Results show that drone performance was improved by the continuous presentation of cineole, a floral cue, and remained unaffected by the presentation of geraniol, a social homing signal. **P=0.023 by two-way analysis of variance on the logit-transformed proportion of time on safe side following the first error, followed by a post hoc Tukey's test of pairwise comparisons.

Fig. 6.

Response of control and IPA exposed honey bee workers during proboscis extension assay. Illustrated is the learned behavior of bees exposed to IPA (open circles, dashed line) and unexposed bees (filled circles, solid lines) during proboscis extension response conditioning assay. In this graph the Y-axis shows the proportion of individuals of each group that showed a conditioned response to the paired presentation of stimulus (antennal stroking) and reward (sugar water) over 12 learning trials (X axis). This is a conditioned response to the CS (not the proportion showing unconditioned responses to the US). Sample sizes for each group are provided in the boxed legend. Both IPA exposure was a significant predictors of response, with IPA treated bees being more likely to perform poorly when compared to control group (GEE, X²=7.37, d.f.=1, P=0.006) (see Supplemental data).

Fig. 7.

Diagram of electric shock avoidance assay with corresponding time course. This is a punishment with discrimination training paradigm that trains honey bees to avoid a color paired with a negative stimulus (mild shock). The figure illustrates the cassette of color background, intercalated grid, and individualized lanes. Additionally shown are the specific current and voltage settings for all tests. The figure also provides an illustrated breakdown of the 56 min learning challenge. In the illustrated time course, black bars denote periods of rest where no color or stimulus was experienced that occurred during pre-trial recovery from CO₂ anesthesia, or at inter trial intervals (ITI). The two-toned portions of the bar illustrate the times when color was presented in association with shock (Trial 1, Trial 2, highlighted the electrical warning symbol) and a one minute presentation of colors without the shock as our short term memory (STM) test of the established association. Video records are available upon request.