Apis mellifera cognition as a tool for scientific engagement

Honeybees are a well-established model for the study of learning and cognition. This is mostly due to the existence of well established conditioning protocols in this group. Although most conditioning protocols are straightforward, their implementation can be challenging due to the presence of multiple procedural steps necessary to ensure learning. Therefore, the use of volunteers to assist in data collection, can be a valuable resource to those who study animal cognition. Here, we developed and implemented a protocol to safely study the effects of caffeine and dopamine on learning performance in honeybees. Using a classical conditioning protocol, we demonstrated that caffeine, but not dopamine, significantly reduced the number of trials required for a successful conditioning response. Simple Summary Global scientific literacy can be greatly improved through widespread and effective community engagement by researchers. However, a significant part of citizen-driven science are projects with compelling narratives and relevance. We propose Apis mellifera (honeybee) as an excellent engagement tool due to widespread awareness of colony collapse and the bees’ importance in food production. Moreover, their cognitive abilities provide a system for meaningful experimentation that can be performed economically. Using proboscis extension response as a proxy for honeybee learning, a group of non-specialist high-school-aged participants demonstrated that caffeine, but not dopamine, improved learning. Given the importance of learning for hive health, this demonstrates that this experimental system, with non-specialist participants, could rapidly identify potential factors that shape learning


Introduction
We currently face an unprecedented combination of events from SARS-CoV-2 and climate change to desertification of arable lands and the extinction of predators, the challenges whose understanding and overcoming demands broad scientific expertise. Thus, it is imperative that communities not only understand the scientific method, but also engage with it [1][2][3]. A key approach to drive this engagement is through citizen science activities. Although there is not a single concise definition of citizen science, it can typically be thought of as involving non-scientists in the scientific process. However, many of the most significant programs either engage with community members already scientifically literate or are much more goal focused without necessarily improving the understanding of the scientific method [4][5][6]. Therefore, effective community engagement and citizen science needs to include both tentpole and smaller activities with community members that are underrepresented in the scientific process.
Although citizen science and scientific outreach take many forms, it is most effective when participants can meaningfully engage with the topic [4,7,8]. The central activity, experiment or scientific question needs to incite enthusiasm but also needs to be presented in a framework that ensures participants gain the most from the experience [4,7,8]. There is potential for broad, long-term impact on the participants. Ultimately, citizen science programs are not only beneficial for students, but also help to build positive relationships between universities that sponsor them and the local community. Due to the broad appeal, external funding from private donors and corporate grants, with more flexible spending parameters, are often available and can be easier to secure.
The Ryukyu Girls scientific engagement program seeks to engage female high school students from the Japanese Okinawa prefecture in the scientific process. As part of this program, we developed an experiment that sought to teach the scientific method as a way of interacting with the world. We employed clear and simple language when describing the scientific processes to ensure it was accessible to non-native English speakers. The program was taught in two languages, English and Japanese, with simultaneous translation provided. We have also gone beyond the language when our presenter made purposeful choices to forge a human connection by sharing relatable life experiences. As a result, the participants were not only more deeply engaged with the content, but also developed a level of trust in the presenter, who they clearly viewed as a role model.
In addition, a critical component of our approach was in highlighting the importance of participation in the scientific process. By drawing a connection between their participation as a part of a larger effort to increase underrepresented voices in science, participants became emotionally invested and, as a result, experienced meaningful engagement in the experiment.
Apis mellifera (european honeybees) provides an excellent tool for scientific engagement, of academic and lay public alike, while furthering our understanding of this critical agricultural pollinator. Honeybees have complex social interactions driven by intra-hive learning and communication [9]. They are capable of not only learning the location of food resources but also communicate to their nestmates through their waggle dance [9,10]. In the laboratory, honeybees are exceptional model organisms to study cognition, memory and communication. Honeybees are capable of complex cognitive processes. For instance, honeybees can memorise locations, patterns, faces and even understand conceptual relationships, such as above/below and same/different [11,12]. Despite this great potential for honeybees as models in cognitive neuroscience, its use is still limited when compared to Drosophila melanogaster [13].
Classical conditioning is a form of conditioning on which a subject learns to associate a neutral stimulus, called conditioned stimulus (CS), with a stimulus of biological significance, the unconditioned stimulus (US), such as sucrose [14]. Over time, animals start to associate the initial neutral stimulus to the US, acquiring the capacity to elicit a conditioned response. Although classical conditioning is considered to be a basic learning process, it has become the foundation of cognition and memory studies in animals, especially in insects [15][16][17][18][19][20]. Amongst insects, Apis mellifera is considered one of the most robust organisms for the study of classical conditioning [21][22][23][24]. Such success is mainly due to the presence of several powerful conditioning protocols [18,19,23,[25][26][27][28].
Honeybees extend their proboscis when their chemoreceptors enter in contact with sucrose. When sucrose is paired with another stimulus, such as a distinctive scent or a visual pattern, honeybees can learn to anticipate the sugar reward when exposed to the stimulus, extending their proboscis. In conditioning, a naive bee (i.e. a honey bee without any previous experience to the stimulus) is exposed to a neutral stimulus (i.e. a scent or an image it hasn't being exposed to before), CS, followed by a sucrose reward, US [18,19,23,25,26]. During conditioning, the honey bee learns to associate the initially neutral stimulus, CS, to the US (i.e. sucrose). Although honeybees are a smart, cost-effective, relatable animal, relatively easy to study and ubiquitous worldwide, conditioning protocols rely on the implementation of several procedural steps (Scheiner et al. 2013). Such steps, such as training trials, can be laborious and require large sample sizes and can be an impediment on the advance of research (Poddar et al. 2013). Therefore, the presence of a system that provides large-scale data in a semi-structured system, such as citizen science, can help to overcome these challenges.
Herein we describe the use of A. mellifera learning as an experimental system to investigate learning performance of bees affected by dopamine and caffeine. Our approach, implemented within the Ryukyu Girls engagement program as an example of outreach and citizen science effort, demonstrated that caffeine treated bees learn faster than dopamine and control bees. Based on this experience, we propose extension and expansion of the programme to enable large-scale observation and data collection to characterise factors that shape bee learning by citizen scientists.

Honeybee Sourcing and Handling
All bees used were from a single hive. Newly emerged bees were obtained by removing two frames containing capped larvae from the hive, brushing off adult bees and placing the frames in a small hive box for 6 hours. To maintain optimal conditions for honeybees, small hives were kept inside the incubator, with constant temperature of 34ºC and 60% humidity. After 6 hours all emerged honeybees were collected and harnessed for experimentation. The two frames were returned to the hive as soon as possible after removing the newly emerged bees.

Harnessing
After being anesthetized using ice [29], bees were harnessed using plastic drinking straws, approximately 14mm in diameter, and cut to lengths of approximately 2cm with a diagonal section removed from one end, creating a V, to allow the bee's head to protrude as described by [30]. Anesthetized bees were placed inside the straw piece, restricting the movement of their arms. The head was fixed in place with small pieces of masking tape that allowed movement of the antennae and proboscis. Although this process is well described by Scheiner et al. (Scheiner et al. 2013), harnessing requires patience and practice to effectively perform. Harnessed bees were divided in three groups, providing three bees per treatment group to the students. To facilitate handling and reduce confusion associated with the treatment, harnessed bees were placed in small holes present in a styrofoam tray, containing a color coded indication and names of the treatment groups ( Figure 1a). To reduce mortality and stress, bees were harnessed 6 hours prior to the practical by the authors without the support of the Ryukyu Girls class.

Quantifying Learning
Despite honeybees being a well established model organism, there is mixed evidence to whether newly emerged bees are able to show significant learning [32]. Therefore, a pilot experiment was conducted to ensure newly emerged honeybees would be capable of associative learning. Learning performance was measured as the number of trials required for the honey bee to learn a new stimulus. In this case, bees were exposed to a lemon scent while offering the sugar solution. When offered a sugar syrup reward, bees extend their proboscis. Over a few trials (i.e. odour paired with sugar reward) the bee learns to associate the odour to the reward, extending their proboscis when exposed to the odor alone. Fewer trials indicates higher learning performance.

Pilot Training
Pilot was performed in 18 bees (6 bees per treatment group) and allowed to establish the presence of associative learning in newly emerged honeybees and if there was a clear trend according to the treatment group.

Pre-Experimental Preparation
Because dopamine and caffeine can take a few minutes to hours to have an effect, depending on the organism, all bees were fed three times prior to the experiment. Feeding was done at 2, 1 and 0.5 hours before the experiment with a sucrose solution mixed with the proposed treatment. Bees were fed according to their experimental group, with sucrose (sucrose solution only: control) or sucrose mixed with either caffeine (sucrose + caffeine: caffeine) or dopamine (sucrose + dopamine: dopamine) solution. The feeding was not paired with odour cue. Sucrose solution was prepared by mixing 200ml water with 100g of sucrose and stirred until all sugar is dissolved. The sucrose solution was divided in three equal parts. One part was mixed with 1mg of caffeine (caffeine treatment) and another with 1mg of dopamine (dopamine treatment). Both solutions were stirred until the chemical was dissolved. All solutions were allocated in 1ml eppendorf tubes to be used during the practice.
To create the odour stimulus during the trials, one drop of lemon essential oil in a cotton ball was used.
To prevent evaporation and further contamination, the cotton ball was stored inside an eppendorf tube. All solutions used during the pre-preparation stage and during experimental training were prepared 18h prior to the practice. All material used in the experimental training was labelled and colour coded, following the same styrofoam tray scheme (Figure 1).

Experimental Training
Prior to the experiment the Ryukyu Girls Class were provided with an explanation of the motivation, literature review (i.e. effects of dopamine in humans and other animals) and what a hypothesis is. Following, they were shown the experimental setup and divided in groups of three. Each group received 3 copies of the "laboratory notebook" (see supplemental material for a copy of the laboratory notebook), eppendorf tubes containing the treatment solutions, the odour scent and multiple swabs. They were then asked to formulate a hypothesis and a prediction for the experiment. They were informed about the risks related to honeybees and how to proceed in case they escape their harness. The training trial consisted of approximating the eppendorf tube containing the lemon scent near the honey bee antennae followed by lightly touching the antennae until the bee extends its proboscis, then feed the bee using the cotton swab ( Figure 2). Each trial was then repeated across all groups following the diagram present in the Figure 1b. They were instructed to repeat the trial multiple times. At the third training trial, students were instructed to delay the sucrose reward for a few seconds to allow visualization of the associative learning. Such small delay, would allow students to visualise proboscis extension without impairing further training protocol. We considered that the bee learned when they exposed their proboscis to the lemon scent prior to the offer of the sucrose solution. The number of trials per bee per group was recorded in the lab notebook and supplied to the authors for posterior analysis.

Cost
Around 100 dollars for the hive supplied. The use of newly emerged bees for the practice did not cause significant damage to the hive and it could be used in further experiments.

Student Group and Engagement
on se To facilitate learning and increase student engagement, we developed a powerpoint presentation and a lab notebook. Both the presentation and lab notebook contained the research background and the protocol for the experiment (lab notebook and presentation can be found on the supplemental materials).

Risk Management
When employing an experimental system that poses a stinging and subsequent allergy risk diligence is required. We employed several strategies to ensure no participants were stung. Although we conducted an initial survey of participants regarding their allergy status to honeybees, this is insufficient to minimise risk. Where possible, double containment was employed. We used only european newly emerged bees, less than 20 hours old, as they are both less aggressive and do not produce venom [33]. We also developed a harness-like system (see Harnessing) that prevents contact with the bee abdomen. We ensured staff levels were such that every experimental group could be supervised by a demonstrator so if a bee did escape, it could be quickly caught. As a final precaution we also kept an adrenaline injector (EpiPen) in our medical kit.

Data analysis
Behaviour is one of the most labile phenotypes and thus use of an appropriate statistical analysis is required to accommodate experimental variation. To minimise issues related to handling, repeated measures ANOVA was employed to analyse the effect of caffeine and dopamine on learning performance. Repeated measures ANOVA compares the difference between means across the treatment groups that are based on repeated observations. Differences between groups were estimated using Multiple Comparisons of Means (Tukey Contrasts). All data analysis was performed in R (R version 4.0.2) and can be viewed at https://github.com/marivelasque/HoneybeeOutreach.git.

Hypothesis Development
To facilitate engagement and understanding, considerable classroom time focused on the process of developing a hypothesis and subsequent testing. Although due to time and logistical constraints, hypotheses could not be generated spontaneously within the classroom. However, after describing the context, materials available and providing guidance in the form of classroom discussion, participants developed hypotheses nearly identical to those we sought to test.

Data Reliability, Collation and Analysis
A total of twelve student groups performed each experiment in triplicate. However, due to loss of bees or experimental error, the groups averaged 2.4 bees per treatment. Each experimental treatment comprises, on average, 28.6 data points. Of the total 108 data points collected, 22 failed. This data loss was predominantly related to harnessing of the bees (see Harnessing). Newly emerged bees have softer cuticles than adults [34] and thus, improper handling during harnessing and training could be partially responsible for the high mortality. Subsequent to completing the experimental work, data collation and a brief analysis was conducted within a classroom setting. This was closely linked to the aforementioned hypothesis generation and thus provided participants with a further understanding in hypothesis testing.

Group results
Caffeine, but not dopamine, was found to significantly reduce the number of trials required for a successful conditioning response (Table 1; Figure 3; Supplementary Table 1). Each of the 12 groups conducted the control, caffeine treatment and dopamine treatment, in triplicate, by measuring the number of trials required for a conditioning response. Although the average was lower for both caffeine and dopamine treatments, only caffeine had a statistically significant difference (caffeine p = 0.038, dopamine p = 0.252; repeated measures anova).  Caffeine-treated bees showed a higher learning performance than the control, requiring less training trials until conditioned to the stimulus. Significance was calculated using repeated measures anova and differences between groups were estimated using Multiple Comparisons of Means (Tukey Contrasts).

Discussion
Community-wide understanding of the scientific process can have enormous public health and economic benefits but requires careful engagement, education and participation from a wide and diverse group of participants. Citizen science and outreach activities provide a compelling way to bridge the divide between scientists and the public at large. By focusing on a highly relevant and engaging model system we sought to bridge the divide between scientists and a subset of the community. Through this engagement, we demonstrated the power citizen science and engagement can have to generate compelling scientific data.

Bee cognition
Caffeine is the most consumed psychoactive in the world, being used by different cultures and social groups to promote wakefulness. Similar to other psychoactive drugs, caffeine also affects dopamine signaling, by blocking dopamine transporters, stimulating its release from terminals and reducing reuptake [35][36][37]. Dopamine is a neurotransmitter with multiple functions, such as the control of rewardmotivated behavior. Therefore, dopamine is an essential component in conditioning, memory and learning [38][39][40], being present in most multicellular animals [41]. Using the group-generated data, we were able to identify caffeine-induced improvements in honeybee learning when compared to dopamine alone. This was potentially due to caffeine increasing the dopamine production and reducing reuptake [35][36][37]. As a result, caffeine treated bees likely had more available dopamine than dopamine treated bees.
In honeybees, improved learning is associated with improved hive health [42][43][44]. Therefore, we propose that further examination of caffeine as a potential hive booster be pursued. Although neither practical nor appropriately appraised as a hive supplement, the inclusion of caffeine-producing plants such as coffee trees, may provide benefit to pollinators. As such, it would provide a potentially fertile avenue of further research. As our climate continues to change and agriculture is impacted by colony collapse, research of this nature will become critical. As it is typically laborious, with large sample numbers needed, engagement activities like this potentially facilitate discovery of compounds improve bee health.

Insect-based engagement
We propose using honeybees as a hook to highlight the importance of insects to global ecology and economic prosperity. Here we describe a straightforward and engaging activity that can be widely deployed to facilitate this. We do however appreciate that the sourcing and handling of bees is a potential issue in conducting this experiment. A potential solution to this is partnership with a local aparist or an apiarist society. Insects are an essential part of most, if not all, ecosystems [45][46][47][48]. They provide multiple ecological functions, ranging from breaking down organic matter in the soil to pollination and the control of insect and plant pests [49][50][51]. Their diversity and abundance is directly related to the state of conservation of the environment, with more natural and undisturbed areas having a higher diversity and abundance than disturbed areas [52][53][54][55]. However, insects are now facing an unprecedented threat [56]. Worldwide their numbers have plummeted, but because of their size, and relative unimportance to the average citizen, scientists don't know the exact extent of their decline [56]. Educating the public by demystifying their presence, function and importance is imperative to solve this crisis [57,58]. Projects that stimulate contact and promote mutual respect between humans and invertebrates, such as citizen science, are more important today than ever. [45] [57,58] [45] Amongst insects, the European Apis mellifera is an ideal candidate for citizen science studies. They are likable, relatively docile (when carefully handled) and, because of their biology, they can have direct parallels with humans. For instance, they live in a society, they share food, communicate locations and even the necessity of grooming through the grooming dance [59]. Furthermore, their relatively larger brain, compared to another laboratory staple, the fruit fly Drosophila melanogaster, makes learning experiments simpler and easier to conduct with non-specialists [9,13].

Data Quality and Loss
The 20 percent loss in data observed in our experiment was well within what we consider acceptable given the complex experimental system employed. This loss is compensated by the increase in data points a project like this achieves. The experiment was conducted on a relatively small scale and thus, a limited amount of samples per experimental group was provided (3 bees per experimental group). Given the simplicity and low cost related to the experimental setup, a larger number of replicates could have been provided to the Ryukyu Girls group to compensate for any data loss.

Absence of adverse events
No adverse events, in particular bee stings, occurred during this activity. Honeybees pose an additional risk when compared to other model systems but when effectively managed this risk is greatly minimised. However, it is paramount that any honeybee-based engagement activities are mindful of the risks posed and take steps to mitigate these risks. Although strong advocates for the inclusions of honeybees in activities involving individuals untrained in their handling, we also recognise the need for restraining the bees.

Conclusion
With the ongoing worldwide concerns regarding Colony Collapse Disorder [60], we hope the use of A. mellifera in outreach, citizen science and education raises awareness and is instrumental in communities adopting more bee-friendly policies. We also hope that through engagement with sections of the communities typically absent from scientific discourse, we amplify this awareness and also foster lifelong critical learning.

Funding
JAD, IK and MV were supported by the Okinawa Institute of Science & Technology. Experimental materials were supported by the Okinawa Institute of Science & Technology and MV. Subsequent analysis and manuscript collation was supported by JSPS KAKENHI Grants 19K06795 and 19K16205 awarded to JAD and MV respectively. The Ryukyu Girls program and participants were supported by Okinawa Institute of Science & Technology Graduate University, University of the Ryukyus, Okinawan Prefectural Board of Education, Okinawan Prefectural Government, the Okinawa Institute of Science and Technology Graduate University Promotion Council and the 'FY2018 GST Female Jr. high and high school student support program..

Author Contributions
MV designed and developed the honeybee experimental framework with input from all authors. JAD, IK, MT & MV conducted the engagement activity. All authors wrote and edited the paper.