Skip to main content
Dryad logo

Honeybee optomotor behaviour is impaired by chronic exposure to insecticides


Parkinson, Rachel H; Fecher, Caroline; Gray, John R (2022), Honeybee optomotor behaviour is impaired by chronic exposure to insecticides, Dryad, Dataset,


Honeybees use wide&[ndash]field visual motion information to calculate the distance they have flown from the hive, and this information is communicated to conspecifics during the waggle dance. Seed treatment insecticides, including neonicotinoids and novel insecticides like sulfoxaflor, display detrimental effects on wild and managed bees, even when present at sublethal quantities. These effects include deficits in flight navigation and homing ability, resulting in decreased survival of exposed worker bees. Neonicotinoid insecticides disrupt visual motion detection in the locust, resulting in impaired escape behaviours, but it had not previously been shown whether seed treatment insecticides disrupt wide&[ndash]field motion detection in the honeybee. Here, we show that sublethal exposure to two commonly used insecticides, imidacloprid (a neonicotinoid) and sulfoxaflor, results in impaired optomotor behaviour in the honeybee. This behavioural effect correlates with altered stress and detoxification gene expression in the brain. Exposure to sulfoxaflor led to sparse increases in neuronal apoptosis, localized primarily in the optic lobes, however there was no effect of imidacloprid. We propose that exposure to cholinergic insecticides disrupts the honeybee&[nprime]s ability to accurately encode wide&[ndash]field visual motion, resulting in impaired optomotor behaviours. These findings provide a novel explanation for previously described effects of neonicotinoid insecticides on navigation and link these effects to sulfoxaflor for which there is a gap in scientific knowledge. --


\subsection{Bees and insecticide exposures}
Worker bees (\textit{Apis mellifera}) were collected at the entrance to outdoor hives located at Falmouth Academy (Falmouth, MA) between July and August 2019. To reduce the inclusion of bees affected by age or illness, bees were discarded if they displayed low activity during the first hour after collection or appeared otherwise damaged. Bees were housed at the Marine Biological Laboratory (Woods Hole, MA) for 5 days in groups of 2-3 bees in 5 cm\textsuperscript{3} acrylic boxes in a ventilated incubator at 32ºC and 55\% humidity, which is within the range of thermoregulation of honeybee hives \cite{Stabentheiner2010}. A 1.5 M sucrose solution was provided ad libitum to control bees (CTL), and treatment group bees were provided with the 1.5 M sucrose solution containing 50 ppb (195 nM) imidacloprid (IMD, Toronto Research Chemicals, Toronto, Canada), 50 ppb (180 nM) sulfoxaflor (SFX, Toronto Research Chemicals, Toronto, Canada), or a mixture of 25 ppb (97.5 nM) imidacloprid and 25 ppb (90 nM) sulfoxaflor (MIX). Treatments were sublethal during the 5 day exposure, and mortality of treated versus control bees did not vary over this period.

\subsection{Optomotor assay}
Using an open-loop, virtual reality arena that allows tethered honeybees to walk on an air-supported ball, we tested the effects of a 5 day chronic exposure to the neonicotinoid imidacloprid (IMD), novel insecticide sulfoxaflor (SFX) and a mixture of the two insecticides (MIX), versus a sucrose control on the optomotor response. The optomotor arena consisted of the air-supported ball, a USB camera module (ELP) positioned 15 cm from the front of the ball, and two computer monitors (CUK Bionic B25GM, 24.5”, 240 Hz refresh rate) oriented at a right angle, such that the bee was facing the apex of the two monitors, 15 cm away from each screen (Figure 1A). Computer monitors displayed vertically-oriented translating black sine wave bars, to give the illusion of rotational optic flow. Wide-field visual motion elicited the optomotor response in tethered walking bees. The resulting rotation of the air-supported ball was video recorded and the fictive walking path was extrapolated using FicTrac \cite{Moore2014} (Figure 1B, Supplemental Information Video 1).

We temperature-anesthetized the bees in a refrigerator at 4ºC until unresponsive (10-15 minutes). Once anesthetized, the dorsal side of the thorax was gently abraded with a blade to remove hairs, and a 1 ml pipette tip was affixed to the dorsal surface of the thorax using UV light cured dental glue (Prevest DenPro). Bees were then offered a drop of 1.5 M sucrose solution and allowed to recover in an incubator at 32ºC for 30 minutes. After recovery, bees were positioned with a micromanipulator (World Precision Instruments) over the air supported ball in the optomotor arena. A terrarium lamp was positioned to warm the bee until it began to walk unprompted. Bees were excluded from the experiment if they did not start walking within 15 minutes. This accounted for approximately 15\% of the population, irrespective of treatment. In total, n=23 CTL, n=30 IMD, n=37 SFX, and n=25 MIX bees were included in the behavioural assay, and the response of all of these bees to the 0.0625 cpd, 16 Hz grating stimulus were recorded. For some stimuli, a smaller subset of bees were tested, to a minimum of n=13 CTL, n=11 IMD, n=15 SFX, and n=11 MIX for any given stimulus. 

\subsection{Visual stimuli}
Visual stimuli were programmed in MATLAB R2019b (MathWorks, Natick, MA) by adapting code from the Psychophysics Toolbox Version 3 \cite{Brainard1997}. Translating vertical black and white sinusoidal bars were displayed in 10-second intervals, first leftwards on both screens then instantly changing direction to rightwards. This sequence was repeated 4 times, summing to a total of 80 seconds per stimulus. Eleven visual stimuli were tested, by varying either the spatial or temporal frequency, and were presented in a randomized order with a 3 minute inter-stimulus interval. Stimuli covered spatial frequencies of 0.5, 0.25, 0.125, 0.0625, and 0.0039 cycles per degree (cpd), where a smaller value represents wider gratings, which were displayed at a constant temporal frequency of 16 Hz. In addition, varying temporal frequencies of 4, 8, 16, 24, 32, 40, and 48 Hz were tested at a constant spatial frequency of 0.0625 cpd. The spatial frequencies were selected based on Ibbotson et al \cite{Ibbotson2017}, which demonstrated that descending neurons respond most robustly to spatial frequencies between 0.03 and 0.05 cpd. A subset of animals were also tested with vertical black bars that were stationary (“no optic flow” stimulus) to measure differences in spontaneous behaviour between treatments. 

\subsection{Fictrac video analysis}

Optomotor behaviour tracking was performed online using FicTrac \citep{Moore2014}. A program was custom-written in Matlab to align the FicTrac data to the visual stimuli and extract the integrated yaw path (yaw radians over time) and total movement across all axes (Figure 1B). The integrated yaw was reset to 0 at the initiation of each stimulus and each stimulus direction switch (e.g., leftward to rightward) such that the output for leftward turns was positive, while rightward turns produced negative values (Figure 1C). 

We fit linear regressions to the 8 normalized response replicates (4 replicates per direction) for a given animal-stimulus pair. The slopes of the linear regressions for each of the four leftward and rightward stimuli were averaged to assess the symmetry of the behavioural responses. Absolute slopes of 0 represented perfectly symmetrical responses, while slopes of increasing magnitude signified greater response asymmetry. The R\textsuperscript{2} values for each linear regression were also averaged to produce a measure of path tortuosity. Responses that switched direction slowly or those that wavered from left to right produced lower R\textsuperscript{2} values than responses that quickly switched direction to follow the visual stimulus and maintained a consistent yaw for the duration of the stimulus. Movement distance (across all rotational axes) and diameter of the air supported ball were used to calculate the walking speed and this was averaged across the 4 leftward and 4 rightward stimulus presentations (80\,s of walking behaviour).

To compare the magnitude of leftward and rightward optomotor responses, we changed the sign of the rightward responses so that both directions of responses were on the same scale (i.e., both positive slopes, Figure~1D). We calculated the response magnitude (cumulative yaw) as the maximum integrated yaw (\textit{ymax}) averaged over stimulus presentations in both directions. Cumulative yaw was also compared between the leftward and rightward stimuli to determine if the responses were asymmetrical. Average \textit{ymax} values were calculated separately for leftward and rightward responses, and cumulative yaw asymmetry (\textit{y.sym}) was obtained by subtracting the smaller cumulative yaw value (\textit{ymax}\,b) from the larger value (\textit{ymax}\,a), and dividing by the larger value (\textit{ymax}\,a) to constrain measurements between 0 and 1. \textit{Y.sym} values close to zero represented symmetrical responses, while highly asymmetrical responses had \textit{y.sym} values closer to 1. Responses were labeled "biased" in the direction that elicited the larger response and "anti-biased" in the other direction. 

\subsection{Relative quantification of gene expression}
A separate set of bees were chronically exposed to IMD, SFX, MIX, or sucrose (CTL) for 5 days, as described above (n = 25 bees per treatment). Freshly dissected honeybee brains were snap frozen on dry ice and stored at -80ºC until further use (<2 weeks). The material from 5 bee brains was pooled to one sample and 5 samples were generated per treatment group. Total RNA was isolated from each sample using the Maxwell® RSC miRNA Tissue Kit (Promega) and the Maxwell® RSC 48 Instrument (Promega) according to the manufacturer’s instructions. RNA concentration was measured using the QuantiFluor® RNA System (Promega), and total RNA was quantified by Quantus Fluorometer (Promega) following manufacturer protocol, yielding between 120 - 210 ng/µl RNA and a total of 7.2 – 12.6 µl. 

Genes of interest were superoxide dismutase 1 (SOD1), catalase (CAT), cytochrome P450 enzymes (CYP9Q2 and CYP9Q3), and the reference gene GAPDH. Forward and reverse primer sequences and sequence accession numbers are described in Supplemental Information Figure 1A. Primers were either designed using NCBI Primer BLAST (NCBI) with 100\% specificity for the genes of interest or taken from previous studies \cite{Collins2004, Reim2013}. Quantitative reverse transcription PCR (RT-qPCR) was performed with a 1-step method combining reverse transcription and amplification in a single tube using the GoTaq 1-Step RT-qPCR System (Promega). A 20 µl final volume contained 10 µl GoTaq qPCR Master Mix, 0.4 µl GoScript 1-step RT Master Mix, 0.5 uL of 10 µM forward and reverse primer mix (Genewiz, New Jersey), 7.1 µl Nuclease Free Water, 0.8 µl MgCl\textsuperscript{2} and the RNA sample. Samples were loaded into a QuantStudio 5 (ThermoFisher Scientific) and run with the following cycling conditions: 45ºC for 15 minutes, 95ºC for 10 minutes, 40 cycles of 95ºC for 15 seconds and 60ºC for 30 seconds (data collection step), and finally 72ºC for 60 seconds followed by the melt curve. RT-qPCR amplification was assessed with the normalized fluorescent signal (Delta Rn) from each sample (Supplemental Figure 1B). We used 1 – 0.0001x dilution series of each RNA sample and the standard curve was used to demonstrate that the presence of PCR inhibitors was unlikely (Supplemental Figure 1C). A reaction without RT enzyme was used as reaction control for assessing the absence of DNA in all samples: no amplification could be detected in these samples. 

We validated qPCR specificity using the melt curve analysis (Supplemental Figure 1D). Samples resulting in qPCR efficiency below 90\% were omitted from the analysis (3 of 20 samples). Data analysis was performed using QuantStudio5 (ThermoFisher Scientific) using the Cq threshold method \cite{Nolan2006, Pfaffl2001}. RNA amplification of genes of interest were normalized to the amplification of GAPDH for each sample (delta Ct), and then compared to the mean amplification of the control samples (delta delta Ct). 

\subsection{TUNEL assay}
An additional cohort of honeybees (n = 4 per group) were treated with IMD, SFX or sucrose (CTL) for 5 days as described above. For tissue processing, cold-anesthetized honeybees were decapitated, the brain was dissected under cold PBS, pH 7.4 (as previously described by Groh and  Rössler \cite{Groh2011}) and immersion fixed in 4\% PFA/PBS, pH 7.4 at 4°C overnight. The tissue was washed in PBS for 12 hours at 4°C and incubated in 30\% sucrose/PBS at 4°C overnight. Prior to sectioning, tissue was immersed in Tissue-Tek O.C.T. compound (Sakura) and frozen on dry ice. Brains were sectioned at 12 µm using a cryostat, collected on SuperFrost Plus adhesion slides (Thermo Scientific), and stored at -20°C until further use. Two sections per brain (120 µm cutting distance) were mounted on each slide.

For TUNEL assay, tissue was defrosted and washed 5 times with PBS, pH 7.4. Sections were permeabilized in 0.1\% Triton X-100, 0.1M sodium citrate pH 6.0 at RT for 30 minutes. After 4 washes with PBS, sections were treated with TUNEL reagent (In situ cell death detection kit, fluorescein; Roche) in a dark, humidified 37°C chamber for 3 hours according to the manufacturer’s instructions. As negative control, no enzyme solution was applied during this step. As positive control, sections were treated with DNase I (1.5 U/µl) in 10 mM Tris-HCl, pH 7.4, 10 mM MgCl\textsuperscript{2}, 1 mg/ml BSA at 37°C for 60 minutes prior to TUNEL development. All sections were washed 3 times with PBS and incubated with Hoechst 33342 (1:1000) in PBS for 10 minutes. After 2 washes in PBS, sections were mounted with Fluoromount aqueous mounting medium (Sigma-Aldrich).

Images were acquired using an upright FV1000 confocal microscopy system (Olympus) equipped with a 20x/N.A. 0.85 oil immersion objective and using the laser lines 405 nm for Hoechst and 488 nm for TUNEL with standard filter sets. Image analysis was further performed using the open source software ImageJ/Fiji \cite{Schindelin2012}. We selected areas o interest (MB, La, Lo, Me) in the honeybee brain and quantified TUNEL+ cells only for areas rich in nuclei (Hoechst staining); hence, excluding the neuropil (Supplemental Figure 4A). Under blinded conditions, the cells within this area were defined for 2-4 sections per bee. Sections with poor Hoechst staining intensity were excluded from the analysis.

\subsection{Consensus clustering and other statistical analyses}
We performed all statistical analyses in R version 4.0.3 \cite{RCoreTeam2016}. Optomotor response types were clustered using a k-means algorithm to group the features described above measured from the behavioural responses of bees across treatments and visual stimuli. Prior to clustering, response features were z-scored. The silhouette method \cite{Rousseeuw1987}, predicted an optimal cluster count of 3 resulting from the k-means algorithm. This cluster count was confirmed by a consensus clustering method \cite{Wilkerson2010}, an unsupervised learning technique that aggregates the results of multiple clustering runs. For each clustering iteration, we included 90\% of the total responses and 80\% of the features (4 of 5). The clustering algorithm was repeated for 2000 iterations, and the output from each clustering run was compared to define a consensus matrix. The proportion of behavioural responses assigned to each cluster was calculated for each treatment and visual stimulus. The features of the behavioural responses within each cluster were then compared to provide a quantitative illustration of the behavioural response types across the visual stimuli and treatments. 

Data were tested for normality using the Shapiro-Wilk normality test and equal variance using Levene’s test from the car package \cite{Fox2019}. In the lmerTest package \cite{Kuznetsova2017} we used a linear mixed effects (LME) model to compare features of behavioural responses across treatments and stimuli, with animal as a random effect to account for repeated measures. Post hoc multiple comparisons were performed using the emmeans package \cite{Lenth2018} with the Kenward-Roger degrees-of-freedom method and Tukey P value adjustments. Relative gene expression, and spontaneous (no stimulus) behavioural responses were quantified with Kruskal-Wallis rank sum tests with Wilcoxon pairwise comparisons.  Statistical significance was assessed at p < 0.05. In figures, asterisks denote statistical significance from post hoc multiple comparisons (< 0.0001 = ‘***’;  < 0.001 = ‘**’; < 0.05 = ‘*’), or letters show significant effects between groups.


Grass Foundation