Bee and flower maintenance
Petunia integrifolia and Antirrhinum majus MTP plants were grown from seed in the GroDome at the University of Bristol at a 16:8 day:night cycle at 20°C. Where experiments were conducted at Rothamsted, plants were transported from Bristol and housed in the Rothamsted greenhouses with a natural light cycle and kept at 22°C. Bumblebees (Bombus terrestris audax L.) were obtained from Koppert, UK, and were housed in the laboratory and trained to forage in a Perspex arena (100 × 75 × 40 cm) under a 16:8 day:night. Bees were provided with ad lib pollen (Bee Pollen Mixed Polifloral, The Happy Health Company, UK) and 30% sucrose solution.
Dynamic headspace collection of floral volatiles
Volatiles were collected from both P. integrifolia and A. majus MTP by dynamic headspace collection (air entrainment), using Pye volatile collection kits (Kings Walden, Herts, UK). Intact flowers on potted and lightly watered P. integrifolia plants, and inflorescences of stem-cut A. majus plants placed in a conical flask containing water, were used throughout. To prepare headspace extracts for gas chromatography (GC) and GC-Mass Spectrometry (GC-MS) analyses, the flowers were individually enclosed in roasting bags (28cm × 30cm; Sainsbury's Supermarkets Ltd, UK), which were connected with a charcoal-cleaned air source, supplying an inflow of 600 mL/min. The air was then drawn through a Porapak Q trap, consisting of 50 mg 50/80 mesh polymer (Supelco, Bellefonte, PA) sandwiched between glass wool plugs in a 24-mm inner diameter glass tube, at 500 mL/min at the air outlet for 2 h, with the Porapak Q tube placed at the floral opening 5 mm from petals. A room control was done without flowers present to identify peaks relating to potential contaminants. Only peaks that were reliably present in the floral samples, but not in the room control, were analysed and identified. Prior to use, roasting bags were baked for 2 h at 140°C, and Porapak Q tubes were conditioned by washing each with 4 mL diethyl ether and heating at 132°C under a stream of nitrogen. The volatiles were eluted from the polymer tubes by flushing them with 750 μL redistilled diethyl ether. The samples were then concentrated to 50 μL and stored at -20°C until analysis.
For experiments requiring electrical stimulation, the flower needed to be accessed by an electrical stimulus, so encapsulation inside an inert container was impractical. As such, the Porapak Q tube was placed very close (<2 mm) to the flower of interest, but the flower or inflorescence was not enclosed. Air was subsequently drawn through the polymer at 500 mL/min for 2 h. To control for environmental contamination, control samples from the room without the flowers present were taken and analysed before and after the experiment. The floral compounds previously identified from enclosed flowers were not present in the room controls (Fig. S1). Any compounds present in the room controls were not analysed in the floral samples.
GC and GC-MS
For the identification of the compounds present in P. integrifolia and A. majus MTP, a Hewlett-Packard 5890 series II GC fitted with a capillary HP-1 GC column (50 m × 0.32 mm i.d., 0.52 μm film thickness; J&W Scientific, Folsom, CA) and equipped with a cool on-column injector was directly coupled to a mass spectrometer (Hewlett-Packard 5972 mass-selective detector). Ionisation was by electron impact at 70 eV, 220°C. The oven temperature was maintained at 40°C for 1 min and then programmed at 5°C/min to 250°C (hold time 17.2 min). The carrier gas was helium. Tentative identification by GC-MS was confirmed by comparing retention index of the unknown peak with that of synthetic compounds and by GC peak enhancement by co-injection with an authentic sample (Pickett 1990), using an Agilent 6890N GC equipped with a cool on-column injector, flame ionisation detector and a 50 m × 0.32 mm i.d., 0.52 μm film thickness HP-1 column. The oven temperature was maintained at 30°C for 1 min and then programmed at 5°C/min to 150°C for 0.1 min, then at 10°C/min to 250°C for 20 min. The carrier gas was hydrogen. Compounds were quantified using the single point external method with an n-alkane (C7-C22) mixture. Authentic standards were purchased from Sigma-Aldrich UK and were >95% pure according to the supplier`s guidelines. (E)-Ocimene was synthesized in our laboratory as previously described (Hassemer et al. 2016).
Measuring the electric charges on bees and the change in VOC emission from P. integrifolia
Bombus terrestris bumblebees were trained to visit P. integrifolia flowers in a laboratory foraging arena. A bumblebee flight arena was split into two (Fig. 1a). Both sides were connected to a bumblebee colony via polyurethane tubes, which contained doors that could be closed and opened to control bee access to each side of the arena. Each side contained a ring charge sensor [RCS, described by Montgomery et al. 2019] consisting of an identical metal ring connected to a picoammeter. Each RCS was calibrated with a Faraday pail (JCI 141, Chilworth Global, Southampton, UK) to measure, in a non-contact manner, the charge on bees approaching the flower. Bees were initially trained to fly through each RCS to access a sugar reward.
During trials, a P. integrifolia flower was placed underneath each RCS, so that the bees would have to fly through the RCS to reach the flower (Fig. 1A). All bees were removed from the arena and volatiles were collected from both flowers for 2 h. The Porapak Q tubes were then refreshed and bees were allowed to forage in one side of the arena (and visit the experimental flower) but were excluded from the other side of the arena, so that only one flower could be visited by bees (Fig. 1A). Volatiles were collected from both flowers for a further 2 h. The charge on each bee visiting the experimental flower over the 2 h period was measured. Whenever a bee visited the experimental flower, the control flower was touched with a grounded rod to control for the mechanical stimulus. The increase in the amount of benzaldehyde produced by each flower was compared over the 2 h period before and after adding bees, using Wilcoxon signed rank tests for the experimental and control flowers. All statistical tests were conducted using R (version 3.5.1). One experimental and control flower was removed from analysis due to bees severing the flower 10 minutes after being released into the arena.
Measuring bee charge using the RCS
The RCS comprised 2 concentric conductive aluminium rings based on the sensor described by Colin et al. (1991). These are insulated from each other by a layer of polycarbonate. The outer ring was electrically grounded and acted as an electrical shield, whilst the inner ring was connected to a picoammeter. When a charged object moved through the inner ring, it induced a current in the ring, the integral of which was proportional to the charge on the object passing through. Two RCSs were used to measure the charge on bees visiting P. integrifolia flowers. Each RCS was calibrated in situ by dropping charged polyurethane cubes (1 cm × 1 cm × 1 cm) through the RCS into a Faraday pail (JCI 141, Chilworth Global, Southampton, UK). The charge measured by each RCS and by the Faraday pail had a direct linear correlation with R2 values of 0.92 and 0.97.
Manual electrical stimulation of flowers
To distinguish between the effects of electrical and mechanical stimulation, volatile emissions were measured from P. integrifolia flowers whilst either electrically stimulated by touching with a charged nylon ball, or mechanically stimulated by touching with an electrically grounded metal rod. Plants were randomly allocated to the control group (touched with electrically grounded rod) or the experimental group (electrically stimulated by touching with a positively charged rod). Plants with flowers of the same age were randomly paired into control and experimental groups. Flowers were used at 2–4 days post anthesis corresponding with the likely peak VOC emission period. All experiments took place between 9:00 and 17:00. To account for temporal variation, measurements were always taken from control and experimental plants simultaneously. During each trial a control and an experimental plant were placed at opposite ends of a room. Using a portable dynamic headspace sampling kit (Pye volatile collection kit, Kings Walden, Herts, UK), volatiles were collected from the control and experimental flowers for 2 h at a flow rate of 500 mLmin-1 by placing a Porapak Q tube at the opening of the flower 5 mm from the petals. The soil at the base of the plant was lightly watered before volatile collection took place. Volatiles were collected from both plants whilst undisturbed for 2 h. After this time, the soil was lightly watered again and the plants were electrically grounded by piercing the soil at the base of the plant with a grounded metal wire. The volatiles were collected for a further 2 h, during which the experimental flower was electrically stimulated every 10 min by lightly touching the flower with a positively charged ball. The stimulus carrier consisted of a nylon ball (diameter 10 mm) fixed to a wooden rod which was given an electric charge of approximately 1 nC by rubbing the ball with polystyrene. The charge on the ball was measured using a JCI 147 Faraday pail with a JCI 140 voltmeter (Chilworth Global, Southampton, UK) before and after touching the plant. The control flower was touched at the same 10 min intervals with a metal rod that was electrically grounded. The charge on the nylon ball dissipated rapidly. To estimate the charge on the ball at the point of contact with the flower, the charge decline on the ball was measured by charging the ball positively by triboelectrification and holding the ball in a Faraday pail (n = 5). An exponential decay curve was fitted to the data and used to estimate the charge on the ball at a point in time given the starting charge (Fig. S2). The increase in benzaldehyde produced by the flowers was compared using a Student’s paired t-test. With the low-charge experiment, the distribution of results was non-normal, so Wilcoxon-Mann-Whitney was used to compare the volatile emissions before and after stimulation.
For the electrical stimulation of A. majus MTP, 2 inflorescences were cut from each plant and placed in conical flasks containing water. A strip of aluminium foil connected to a grounding point was also placed in the water to electrically ground the base of the stem. Flowers of a similar age on each inflorescence were randomly allocated to be touched with the grounded rod or the experimental charged ball. The volatiles were then collected from the control and experimental inflorescences over a 2 h period, during which every 10 min, the outer lobe of the flower was touched with the grounded rod or charged ball. This experiment was done with separate inflorescences at both <1000 pC and <100 pC of charge. The rods were charged in an identical manner to the experiments with P. integrifolia and the charge was measured the same way. The amount of each volatile produced by the charged and the control flowers was compared. The amount of each volatile was highly correlated within each flower, so volatiles were combined for each flower and the total volatile emissions were compared.
Behavioural responses of bumblebees to benzaldehyde
GC and GC-MS identified benzaldehyde as the primary compound produced by P. integrifolia. The ability of bumblebees to sense benzaldehyde was tested using the proboscis extension reflex (PER) and by coupled gas chromatography–electroantennography (GC-EAG). The PER experiment is a common behavioural experiment used to test memory and learning in insects. PER involves pairing a scent (conditioned stimulus) with a sugar reward (unconditioned stimulus). Over a series of trials, the bee is taught to associate the scent with the reward. During a trial, the bee is presented with the scent and given the opportunity to extend its proboscis (unconditioned response). The antenna of the bee is then touched with a tissue containing 30% sugar solution, causing the bee to extend its proboscis and the bee is allowed to drink from the sugar solution. Once the association is learnt, the bee will extend its proboscis in anticipation of the reward upon detecting the scent (conditioned response). An overview of PER in bumblebees is found in Laloi et al. (1999).
The PER experiment exposed bumblebees to the scent of benzaldehyde administered as a puff of air from a pipette containing a filter paper onto which 2 μL of pure benzaldehyde was applied. Bees were starved of sugar water 12 h prior to the experiment. One bee was anaesthetised using CO2 and placed in an enclosure formed from the head of a pipette, where the end had been removed to allow the head and tongue to protrude out the front of the enclosure. The bee enclosure and the end of the stimulus pipette were held down with plasticine modelling clay (TTS, UK). The stimulus pipette was placed so the tip was 1 cm away from the head of the enclosure. The reward was administered as a drop of 30% sugar water on cotton wool rolled around a wooden rod.
Sixteen bees were conditioned through 10 trials to associate the puff of air containing benzaldehyde with a reward (administered as a small drop of 30% sugar water on tissue paper wrapped around a wooden rod). Each trial consisted of slowly depressing the stimulus pipette for 12 seconds ensuring flow of scented air past the head of the bee. During the first 6 s of this period, the bee was observed for proboscis extension. During the second 6 s, the bee was presented with a sugar solution by lightly touching the antenna with the solution and allowed to drink.
The bee was left for 5 min between trials to allow the benzaldehyde scent to dissipate. After 10 conditioning trials, 3 control trials (Trial 11, 12 and 13) were administered, where the stimulus pipette was replaced by a control pipette not containing filter paper. In all but one case, these failed to elicit a PER response from the bee. After the 3 control trials, a final stimulus trial was conducted with the original benzaldehyde scent stimulus. The purpose of the control and final stimulus trials was to confirm the bee was responding to the scent of benzaldehyde and not just to the mechanical stimulus of the puff of air.
Electrophysiological responses of bumblebees to floral volatiles
Volatiles were collected from enclosed P. integrifolia and A. majus MTP flowers by dynamic headspace collection (air entrainment). To locate the compounds that bumblebees responded to in headspace extracts from P. integrifolia and A. majus MTP, coupled GC-electroantennography (GC-EAG) was used. The system has been described previously (Wadhams 1990). EAG electrodes were constructed using borosilicate glass capillaries (2 mm outer diameter, 1.6 mm inner diameter) using a Narishige electrode puller (Optical Instrument Services Ltd, Croydon, UK). These were filled with electrolyte solution (7.55 gL-1 sodium chloride, 0.64 gL-1 potassium chloride, 0.22 gL-1 calcium chloride, 0.86 gL-1 sodium bicarbonate, 1.73 gL-1 magnesium chloride, 0.61 gL-1 sodium orthophosphate). The electrodes were attached to a holder (Ockenfels Syntech GmbH, Kirchzarten, Germany) on a micromanipulator (Leica Microsystems, Milton Keynes, UK) and threaded on so that a silver wire connected to the circuitry was inside the electrolyte.
A worker bumblebee was anaesthetised by cooling on ice, and an antenna was excised below the scape, also making a slit in the tip to ensure better contact between the electrolyte and the antenna. Either end of the excised antenna was placed in the tip of the electrodes. A glass tube with a hole in the side was placed 10 mm in front of the antenna, through which charcoal-filtered and humidified air was passed at a constant flow of 1 L/min. The effluent from the GC was split (1:1) between the flame ionisation detector (FID) and a heated GC transfer line (250°C) connected to the humidified air flowing towards the antennal preparation. The signals were passed through a high-impedance Syntech amplifier. Separation of VOCs collected from flower headspaces was achieved on a GC (6890N; Agilent Technologies, Santa Clara, CA) equipped with a cool-on-column injector and an FID, using a 50 m × 0.32 mm i.d. × 0.52 μm film thickness non-polar HP-1 column. The oven temperature was maintained at 30°C for 2 min and then programmed at 5°C/min to 250°C. The carrier gas was helium. The outputs from the EAG amplifier and the FID were monitored simultaneously and analysed using a customised software package (Syntech). One μL aliquots of pooled headspace samples were injected. A compound was identified as EAG-active if it evoked an antennal response in three coupled runs.