Journal of Insect Physiology 121 (2020) 104002
Contents lists available at ScienceDirect
Journal of Insect Physiology journal homepage: www.elsevier.com/locate/jinsphys
Selection of key ﬂoral scent compounds from fruit and vegetable crops by honey bees depends on sensory capacity and experience
Flore Masa, , Rachael M. Hornera, Sam Brierleya, Ruth C. Butlera, David M. Sucklinga,b a b
The New Zealand Institute for Plant and Food Research Limited, Gerald St, Lincoln 7608, New Zealand School of Biological Sciences, University of Auckland, Auckland, New Zealand
A R T I C LE I N FO
A B S T R A C T
Keywords: Learning Social insect Olfaction Preference Electrophysiology
Flowers have complex odours often comprising hundreds of volatile compounds. Floral scents are species-speciﬁc, and vary also among populations, varieties, sexes or lines, as well as with phenology. Honey bees, Apis mellifera, generally associate only a few key compounds among the complex ﬂoral scent with the food reward which guides their foraging choices. How these key compounds are selected remains partially unexplained, despite their crucial role in inﬂuencing foraging. Using electrophysiological techniques and behavioural assays, we identiﬁed the key bioactive compounds that bees detected with their antennae and that were associated with appetitive responses from four fruit crops and three vegetable crops. Three quantities of identiﬁed key volatile compounds were assayed with the two methods in each of four diﬀerent seasons with experienced foragers. Whether the selection of these key compounds is determined by the sensory capability of the bee or inﬂuenced by its foraging experience was assessed by comparing experienced and naïve bees. Our results showed that experienced foragers were electrophysiologically-sensitive to a speciﬁc set of key compounds for each crop, independent of variation in quantity among several varieties. Experienced foragers responded to these compounds in all seasons, with increased electrophysiological amplitude with increasing quantities. Behavioural appetitive responses varied amongst compounds and seasons, revealing preferences based on associative learning. Naïve bees that were exposed to compounds and subsequently learned them, tended to be overall more sensitive. We discuss our results based on the identity of each bioactive compound and their presence in nature. Preferences for speciﬁc ﬂoral compounds based on sensory biases exist and associative learning may reinforce behavioural attraction depending on foraging experience in each season.
1. Introduction Honey bees, Apis spp., are the most commonly used pollinators in agriculture worldwide because they can be managed in large colonies, and deployed in crops to increase yields (Free, 1993; Klein et al., 2007). Certain crops have been shown to be better pollinated by other visitors, such as the bumble bee (Bombus terrestris) for beans (Garratt et al., 2014), the leafcutting bee (Megachile rotundata) for alfalfa, and mason bees (Osmia spp.) for fruit trees (Bosch and Kemp, 2002; Delaplane and Mayer, 2000). Also, not all ﬂowering plants (i.e. angiosperms) have coevolved with honey bees, meaning that their ﬂoral traits have not been selected to be attractive to honey bees. In a recent evolutionary experiment, Gervasi & Schiestl (2017) demonstrated that Brassica rapa ﬂowers pollinated only by bumble bees evolved traits more attractive to them, such as taller size and increased ultraviolet reﬂection, whereas ﬂowers reproducing under the selection of hover ﬂies became shorter
and less fragrant. Thus depending on the evolutionary context of a plant, ﬂoral traits may become more attractive to speciﬁc pollinators or may stay variable under the divergent selection of diverse pollinators (Schiestl and Johnson, 2013) as well as herbivores or pathogens (Raguso, 2008a, 2009). Plant breeders have been selecting plant traits for improving their performance and resistance to certain environmental conditions such as drought or plant pathogens, but for 30% of our crops, yield ultimately relies on successful insect pollination (Klein et al., 2007). Some crops suﬀer from pollination limitation, even though pollinators are supplied in enough numbers (Aizen et al., 2008; Ashman et al., 2004). Although honey bees are the most commonly used pollinators, we still know little about their ﬂoral preferences, apart from colour, which has been well studied (Giurfa et al., 1994; Giurfa et al., 1995; Menzel, 1968, 1985). Floral scent attracts pollinators, and honey bees associate ﬂoral odours with the ﬂoral food reward which guides their foraging
Corresponding author at: Plant & Food Research, PB 4704, Christchurch 8140, New Zealand. E-mail address: ﬂ[email protected]
https://doi.org/10.1016/j.jinsphys.2019.104002 Received 15 July 2019; Received in revised form 16 December 2019; Accepted 19 December 2019 Available online 21 December 2019 0022-1910/ © 2019 Elsevier Ltd. All rights reserved.
Journal of Insect Physiology 121 (2020) 104002
F. Mas, et al.
2.1.2. Floral volatile collection We carried out dynamic ﬂoral headspace collection in the ﬁeld from the diﬀerent varieties of fruit crops: apricots (n = 2), cherry (n = 5), plum (n = 3) and avocado (n = 4), as well as the hybrid vegetable crops: Chinese cabbage (n = 5), radish (n = 7) and carrot (n = 9). The ﬂoral headspace collection involved encasing in situ ﬂowers from male and female plants inside plastic oven bags. A constant ﬂow rate of 125 mL/min was used, with air entering the chamber through a charcoal ﬁlter and the ﬂoral headspace pulled out the chamber by passing through an adsorbent ﬁlter composed of 60 mg of Tenax®-GR 35/60 (Grace Davison Discovery Sciences, VIC, Australia). The Tenax was conditioned at 250 °C for 3 h under nitrogen before use. Charcoal ﬁlters were also used upstream to clean the air pumped into the bag. Volatile collections were made for 24 h. For each crop and variety, at least three ﬂoral collections were made, plus an air or leaf control collected at the same time. Immediately after headspace collection, Tenax were eluted with 1 mL of n-hexane (96% HPLC grade, Scharlab S.L., Sentmenat, Spain). A 200 µL subsample of each extract was concentrated down to 80 µL under a gentle stream of argon to use for chemical analysis.
(Free, 1963; Menzel, 1985, 1993; Menzel and Erber, 1978; Menzel and Greggers, 2013; Menzel et al., 1990; Raguso, 2004, 2008b; Reinhard et al., 2004a; Reinhard et al., 2004b; von Frisch, 1923). Reinhard et al. (2010) showed that honey bees learn only a few key compounds from a blend of ﬂoral volatiles, but the basis of selection of those compounds is still unexplained. Therefore a better understanding of the key ﬂoral compounds from the major crops may help to improve honey bee pollination. Honey bees can learn a wide range of odours, even man-made odours such as TNT (trinitrotoluene) (Bromenshenk et al., 2015). Thus associative learning during foraging may strongly inﬂuence each individual’s preferences and perception. Yet, learning varies with seasons, with summer bees generally being poor learners compared to winter bees (Ray and Ferneyhough, 1997). Ramirez et al. (2016) also demonstrated that pre-exposure to odours during the pre-imaginal stage can aﬀect honey bees’ memory at a young age (3–5 days old), but this eﬀect was no longer detectable at a later age (17–19 days old). A few studies have linked the behavioural choice of bees to ﬂoral volatiles and their electrophysiological perception (Henning and Teuber, 1992; Kobayashi et al., 2012; Pham-Delegue et al., 1997; Pham-Delegue et al., 1992; Twidle et al., 2015; Wadhams et al., 1994). It is well known that bees can learn to associate an odour stimulus with a sugar reward, as demonstrated by extension of the proboscis when presented with the trained odour in the Proboscis Extension Reﬂex (PER) paradigm (Matsumoto et al., 2012). This can be indirectly used to measure behavioural attraction to an odour. In contrast, perception and function of the olfactory pathways in insects can be measured via electro-antennography (EAG) by measuring the neuronal electric signal of an insect antenna for a given odour. However, EAG does not record integration of diﬀerent odours in the brain and therefore cannot translate directly into behavioural output. Furthermore, research by Claudianios et al. (2014) showed that learning of some compounds inﬂuences the regulation of olfactory receptor expression in the antennae. Little is known about how foraging experience over the seasons inﬂuences olfactory neuroplasticity in honey bees. More work is needed to separate plasticity from learning and sensory bias from olfactory sensors inﬂuencing ﬂoral trait preferences. To gain a better understanding of the selection of the key compounds driving foraging choices of bees, we aimed to answer the following questions: 1) what are the key compounds from commercial crops and do these compounds vary with varieties?; 2) are there differences in olfactory perception and behavioural response to diﬀerent key compounds depending on their quantity and availability varying with the seasons? 3) are the selection of key compounds the result of foraging experiences or sensory bias?
2.1.3. Honey bees Honey bees were sourced from the Plant and Food Research apiary at Lincoln (-43.640052°,172.472080°). Returning foragers of unknown age were collected from eight diﬀerent hives. These bees were considered to be “experienced foragers”, as they were active foragers that had been exposed to ﬂowers and odours from the hive and the environment. 2.1.4. Identiﬁcation of key compounds via gas chromatography – Electroantennogram detection (GC-EAD) By coupling the chemical analysis of a ﬂoral extract performed with a GC to an EAD system, we are able to simultaneously record the electrophysiological response of an antennae (i.e. the summation of all olfactory neurons’ electrical signal within an antennae) to each chemical compound being separated on the column of the GC. Thus for each compound in the ﬂoral extract, an electrophysiological response can be measured in millivolts (Schiestl and Marion-Poll, 2002; Struble and Arn, 1984; Thiery et al., 1990). Bees were anaesthetised with CO2 and a whole head mount with antennae was prepared. The base of the head was inserted into one electrode in a glass capillary in gel (Parker Laboratories Inc., Fairﬁeld, NJ, USA) and the distal part of the antenna remained intact and was mounted to the tip of second glass capillary containing Ringer’s saline solution. The mount was placed in front of a Varian 3800 GC (Walnut Creek, CA, USA) outlet coupled to an Syntech EAD (Syntech Research and Equipment, Hilversum, The Netherlands). A 1 µL aliquot of each pooled ﬂoral headspace extract was injected into the GC with a DB5 MS column (J&W Scientiﬁc, Folsom, CA, USA), 30 m × 0.25 mm i.d. × 0.25 μm ﬁlm thickness running a temperature programme from 40 to 280 °C at a rate of 10 °C/min, with a 2 min delay at the start and a 5 min hold at the end, lasting a total duration of 31 min. The column eﬄuent was split in a 1:1 ratio between the antenna and a ﬂame ionisation detector, which recorded the GC output for comparison with the GC coupled to a mass spectrometer (GC–MS) output (see below for description). At least 15 bees were tested for each pooled extract per crop, and only compounds that triggered an electrophysiological response from at least half of the bees tested were considered as bioactive. We conﬁrmed the identiﬁcation of each bioactive compound by injecting the corresponding synthetic compound and conﬁrming that there was an EAD response at the corresponding retention time recorded previously with the ﬂoral extract.
2. Material and methods 2.1. Floral volatile collection and identiﬁcation of key compounds 2.1.1. Crops Seven crops commercially grown in New Zealand and ﬂowering at diﬀerent seasons were selected. Four spring-ﬂowering (September to November) fruit trees of diﬀerent varieties were selected: ‘Sundrop’ and ‘Clutha Gold’ apricot (Prunus armeniaca L.), ‘Sweetheart’, ‘Sonnet’, ‘Summit’, ‘Stella’ and ‘Lapins’ cherry (Prunus avium L.), ‘Black Doris’, ‘Omega’ and ‘Angelino’ plum (Prunus persica L.) and ‘Bacon’, ‘Hass’, ‘Fuerte’, and ‘Zutano’ avocado (Persea americana). Three summerﬂowering (December to February) vegetable crops were also selected: nine hybrids of carrot (Daucus carota var. sativus), seven hybrids of round red radish (Raphanus sativus) and 5 hybrids Chinese cabbage (Brassica rapa L subsp. chinensis). All vegetable crops were hybrids generated via cytoplasmic male sterility with pollen-donor ﬂowers represented by the male-fertile lines (MFL) and the pollen-receiver ﬂowers represented by the male-sterile lines (MSL).
2.1.5. Chemical analysis One µl of each concentrated ﬂoral headspace extract from each variety of crop was injected into a GC–MS (Agilent 7890B GC and 5977A MSD, Agilent Technologies Inc., Santa Clara, CA, USA). The 2
Journal of Insect Physiology 121 (2020) 104002
F. Mas, et al.
mix, then runs for the six compounds, ﬁnishing with the solvent mix, then a sugar run. There were three sequential runs for each compound, one for each of three quantities (0.1/10/1000 µg), with these tested in ascending order. These concentrations were based on the results of our chemical analysis and Mas et al. (2018). The quantities are thus a ‘splitplot’ treatment nested within the compounds (‘main plots’, six per day), but with the split plot treatments not randomised. Individual bees were placed 5 cm in front of a syringe containing 10 μL of the test compound on ﬁlter paper. A constant airﬂow of 200 mL/min was presented through a second syringe containing 10 μL of the solvent mixture on ﬁlter paper. Bees had 15 s of settling time before being presented with 30 mL of air containing the odour for 4 seconds. The PER response was recorded as a binary response (no/yes). All bees responding to the solvent at testing, or not responding to the sugar were excluded.
injector port of the GC was set at 250 °C and equipped with a DB-5 MS non-polar column (30 m × 0.25 mm i.d., 0.25 μm ﬁlm thickness; J&W Scientiﬁc Folsom, CA, USA). The temperature program of the oven was set at 40 °C, held for 2 min, then raised to 280 °C at a rate of 4 °C/min and held for 5 min. Injections were splitless for 0.6 min and helium gas carried a constant ﬂow at 1 mL/min. Spectra were recorded at 70 eV over a mass range from 20 to 499 m/z. Only compounds that triggered an electrophysiological response via GC-EAD (see Figure S1 and S2) were identiﬁed from the GC–MS trace, comparing retention time and mass spectra to the NIST 2008 MS library and conﬁrmed with synthetic compounds when available. Eleven compounds were selected and conﬁrmed with synthetics purchased from Sigma-Aldrich with a minimum of 98% purity: 4-allylanisole, 4oxoisophorone, linalool, methyl-p-anisate, nonanal, 2-aminobenzaldehyde, benzaldehyde, decanal, methyl salicylate, 2-phenylethanol, and phenylacetaldehyde. One compound from the cherry crop could not be conﬁrmed, so instead we chose to add (E,E)-α-farnesene to our selection of test compounds, as it was reported by us to be honeybee-key compound from the green kiwifruit, Actinidia chinensis (Twidle et al., 2015). These twelve test compounds were used in all experiments below. The relative proportion for each key compound identiﬁed for each sampled plant within each crop was calculated by dividing the area of the peak from that compound measured on the chromatogram by the total peak area of all key compounds in that plant and multiplying by 100. We have displayed mean relative proportions per variety per crop instead of absolute quantity in Fig. 1 to show variation in the ratio of key compounds independently from overall absolute quantity since each plant and variety may vary in size and thus quantity of odours released.
2.2.4. EAG Five bees per season per compound group (groups 1 and 2) were assessed (i.e. N = 40 bees), with one bee from each compound group tested per day and with each bee tested with each of the six compounds in their group. The running order of the six compounds per group per day was determined using an incomplete Latin square, generated with CycDesign 5.1 . Each compound was tested in ascending quantity from 0.1 µg, 10 µg to 1000 µg with a run for the solvent mix before and after each compound run. Solvent runs were used to normalise responses between compounds to control for fatigue over time. Thus, quantity (including 0 for the solvent) is a ‘split-plot’ treatment nested within the compounds (‘main plots’, six per day, equivalent to the running order of the compounds), but with the split plot treatments not randomised. Bees were anaesthetised with CO2 gas, and the antenna was cut oﬀ at the ﬂagellomere I segment and mounted to the tip of a glass capillary tube ﬁlled with Ringer’s saline solution. The distal part of the antenna remained intact and was mounted to the tip of the second glass capillary. Ag-AgCl wire was used to conduct electrical signals from the electrodes to a high-input impedance head-stage pre-ampliﬁer (universal AC/DC probe; Syntech, Buchenbach, Germany). Signals from the pre-ampliﬁer were ampliﬁed and processed with a PC-based signal processing system (IDAC-4; Syntech, Buchenbach, Germany). The test compounds were prepared individually in a 50:50 mixture of n-hexane and dichloromethane with 10 μL pipetted onto ﬁlter paper 25 mm × 5 mm (Whatman No. 1) inside a Pasteur pipette.
2.2. Seasonal responses of experienced foragers 2.2.1. Honey bees Returning experienced foragers of unknown age were collected from a single hive per day in the morning of each behavioural (PER) and electrophysiological (EAG) experiment. The same ﬁve hives were used across the seasons, with one hive out of ﬁve randomly selected each day for each season. Experiments were run in summer (19 January −8 February 2018), autumn (17–23 April 2018), winter (19–26 July 2018) and spring (1–5 October 2018). 2.2.2. Key compounds The 12 test compounds were split into two groups to avoid fatiguing the bees by presenting them with too many compounds consecutively. Group 1 consisted of six compounds mainly found in the vegetable crops: 2-aminobenzaldehyde, benzaldehyde, decanal, methyl salicylate, 2-phenylethanol and phenylacetaldehyde (see Figure S1). Group 2 consisted of six compounds mainly found in the fruit crops: 4-allylanisole, 4-oxoisophorone, (E,E)-α-farnesene, linalool, methyl-p-anisate and nonanal (see Figure S2). Each compound was prepared at three diﬀerent quantities (0.1, 10 and 1000 µg), diluted in a solvent mixture of 50:50 hexane and dichloromethane.
2.3. Responses of naïve bees 2.3.1. Honey bees Bees used in these experiments were 21-day-old bees from a single hive that were reared under laboratory conditions from capped brood stage. These bees were not exposed to the outside environment nor to food from nurse bees after emergence and were only pre-exposed to odours from the frame that they emerged from. The frame of capped brood was placed into a humidiﬁed nucleus hive inside an incubator maintained at 35 °C. Relative humidity was maintained at 70% by an automated sensor-humidity unit. Emerged bees were collected and placed into hoarding cages in groups of 15–30 bees. These cages consisted of a 250 mL party cup with venting holes on the sides. Bees were provided with 50% sugar syrup via 3 mL syringes with the syringe tip cut oﬀ to allow for drip feeding. Protein was provided via a commercial pollen substitute (Apifeed® Pollen Substitute Roll, MSugar, New Zealand) formulated into a patty in a queen rearing cup. Bees were reared until 21 days old, when they would be expected to be of forager age (Winston, 1991). The PER-EAG method described below was undertaken eight times, each on a diﬀerent day; four times in November 2018 and four times in January 2019, with a separate frame of bees from the same hive for each month. This resulted in 160 bees undergoing PER training and, selected from those results, 48 bees undergoing EAG testing (16 naïve bees of forager age tested per treatment group).
2.2.3. PER testing There were a total of eight trials across four seasons, with each season having two groups of compounds tested as above. For each of the eight trials, ﬁve groups of approximately 20 bees were assessed. Each set of 20 bees was exposed to the six compounds in their group per day, with the order of the compounds determined using an incomplete Latin square, generated with CycDesign 5.1 (VSN International Ltd, 2013). Each day, the two groups of 20 bees were harnessed in small cylindrical 3D-printed tubes of 8 mm diameter. Bees were fed 4 μL of 50% sugar syrup and left in a humidiﬁed box for two hours before testing. Before testing, bees’ antennae were gently touched for 1 s with a toothpick soaked in 50% sucrose solution and any bees not exhibiting PER were excluded. The ﬁrst run was sugar, followed by the solvent 3
Journal of Insect Physiology 121 (2020) 104002
F. Mas, et al.
Fig. 1. Key bioactive compounds identiﬁed from each crop. Each black dot represents the mean relative proportion (%) of peak area of each compound for each variety and line per crop. Red dots are overall mean per compound per crop A) Vegetable crop, B) Fruit crop. (For interpretation of the references to colour in this ﬁgure legend, the reader is referred to the web version of this article.)
paper. This concentration was selected to ensure that a maximum number of bees will learn during the PER test. Before training, bees’ antennae were gently touched for 1 s with a toothpick soaked in 50% sucrose solution and any bees not exhibiting PER were excluded. For the training, bees had 15 s of settling time before being presented with 30 mL of air containing the odour mix. This corresponds to the conditioned stimulus (CS). After 3 s of odour exposure, bees’ antennae were gently touched with a toothpick soaked in 50% sucrose solution and once the proboscis was extended in response, a sugar reward was given corresponding to the unconditioned stimulus (US). The association in
2.3.2. PER training and testing Twenty bees per day were used for this experiment. The general methodology based on Pavlovian conditioning and reviewed by Matsumoto et al. (2012) was followed. As described in the PER method above, naïve bees were harnessed and fed, then left in a humidiﬁed box for two hours before training. Five bees remained in the humidiﬁed box and did not undergo PER training, but received four unassociated sugar rewards separated by 10 min and were categorised as “not trained”. The remaining ~15 bees were presented with 10 μL of 100 mg/mL solution of the mix of all 12 test compounds described in section 2.2 on ﬁlter 4
Journal of Insect Physiology 121 (2020) 104002
F. Mas, et al.
(‘eﬀects’) between a mean and that for the solvent is the normalised response, once this diﬀerence has been back-transformed. Therefore both the maximum responses and the normalised responses are provided for comparison. For the experienced bees, data for each of the eight group by season sets were analysed separately. The log-maximum responses were analysed with analyses of variance, with blocking structure Bee/Main Plot (i.e. random model), and treatments (i.e. ﬁxed model) compound by quantity, where quantity included the solvent run that preceded each of compound set. Data points for which the |standardised residual| > 2.5 were excluded. Fixed eﬀects were assessed with F-tests. For the naïve bees, the log-maximum PER response was analysed with Residual Maximum likelihood (REML, Payne et al., 2017) to allow adjustment for all of the possible random eﬀects (the design was not fully balanced). The ﬁxed model comprised treatment, compound and the interaction of these. The random model used was Week/ (Day/ (Main Plot*Bee)). Random components were assessed with a χ2 test of the change in REML deviance on dropping a term. All components other than week and day were highly signiﬁcant, so were included in the ﬁnal analysis. Fixed eﬀects were assessed using F-tests of Wald statistics, using the Kenward and Roger method for estimating the denominator degrees of freedom (Kenward and Roger, 1997). For all EAG analyses, the normalised responses were obtained as part of the analysis. They are presented along with 95% conﬁdence limits, which were obtained on the log-scale and back-transformed. For Figs. 2 and 3, to reduce visual clutter and enable the patterns in the means to be seen more clearly, conﬁdence limits are not shown in the main part of the ﬁgure. Instead, the smallest, largest and a mid-range mean were re-drawn to the side of the ﬁgure, with their associated conﬁdence limits. Conﬁdence limits for other means can approximately be interpolated from the three that are shown. All statistical analyses and manipulation of the data were carried out with Genstat (Payne et al., 2017).
time of the CS with the US is expected to trigger the formation of a memory of the odour cues with the presence of the reward and prompt a conditioned response (CR). Each bee was trained four times consecutively, with 10 min intervals between each training association. Then, bees were tested for short-term memory 1 h after training with the odour mix. In the absence of a sugar reward, bees’ responses to the presentation of the odour mix were recorded as positive if they showed a PER or as negative if they did not show a PER. Bees were again tested for long-term memory after 18 h. All bees responding to the solvent at testing were excluded. Bees were categorised into three categories: “trained and did not learn” (i.e. did not respond in the memory tests after 18 h), “trained and learned” (i.e. responded to the 18 h test) and “not trained” (i.e. were not subjected to PER training as described above). 2.3.3. EAG of naïve bees following PER Based on the 18 hr PER memory test results, two bees from each of the three treatment groups “trained and learned”, “trained and did not learn”, and “not trained”, were selected for EAG testing and the rest of the bees were discarded. Each bee was tested with each of the 12 compounds at a single quantity of 1000 µg, which is the same concentration as in the PER training. The testing order of the compounds for each bee and the order of the bees was determined according to a split-plot design, with bee as the main plot, and compounds as the split plot treatment. The order of the bees in each day and the order of the compounds were generated using Latinised designs generated with CycDesign. 2.4. Statistical analysis Data for the diﬀerent seasons were analysed within separate analyses, since there was no replication of the season over several years. Thus, results for the diﬀerent seasons have been compared informally through graphical presentation only.
3. Results 2.4.1. PER trials: Experienced bees Data for the eight groups by season sets were analysed separately. Data for any bees that died during the trial and any bees that responded negatively to a sugar solution run or positively to a solvent run were excluded. The data were converted into number of bees that responded for each replicate of each compound and quantity out of the total bees per replicate and treatment. The data were analysed using a hierarchical generalised linear model (HGLM, Lee et al., 2006), with a binomial distribution for the ﬁxed eﬀects and a beta distribution for random eﬀects, with logit links for both, and dispersion estimated. The ﬁxed eﬀects comprised a factorial set of compounds by quantity (and contrasts within this set), and possible random eﬀects included replicate (day), individual bees, main plot (six per bee), and exposure order (six main plots by three quantities). The importance of ﬁxed or random eﬀects was assessed using a χ2 test of the change in likelihood, as implemented in Genstat’s HGFIXED and HGRANDOM procedures (GenStat Committee, 2015). Both the replicate and main plot random eﬀects were important for at least some of the eight trials, so these two eﬀects were included in the ﬁnal analyses. The percentage of bees responding and associated 95% conﬁdence limits were obtained on the link (logit) scale, and back-transformed for presentation.
3.1. Key bioactive compounds With the GC-EAD of the vegetable crop samples, we identiﬁed eight consistently-bioactive compounds, previously reported in Mas et al. (2018) (Fig. 1A and S1). From the four fruit crops, we identiﬁed nine bioactive compounds (Fig. 1B and S2). Within each crop, three to four compounds were usually consistently bioactive. Some bioactive compounds occurred in more than one crop. For example, nonanal was found in all vegetable crops and two fruit crops, and linalool was found in all fruit crops except apricot. Others were speciﬁc to a crop: decanal triggered an EAD response only in cabbage, methyl salicylate only in carrot, and 2-phenylethanol only in radish, even though they were also present in other crops (data not shown). There were strong variations in relative quantities of each bioactive compound between varieties for a crop as well as between crops. But each combination of bioactive compounds was speciﬁc to each crop. 3.2. Seasonal responses of experienced foragers to key compounds 3.2.1. PER For both compound group 1 (arable) and compound group 2 (fruit), there were signiﬁcant interactions between compound and quantity in autumn for group 1 (p = 0.04) and in the winter and possibly summer for group 2 (p < 0.001 and p = 0.06), indicating that for these cases, the changes with quantity varied between the compounds, Fig. 2. However, the interaction was not signiﬁcant for the other cases (group 1: p = 0.22, p > 0.99; p = 0.11 for spring, summer and winter; group 2: p = 0.63, p = 0.11 for spring and autumn respectively), indicating that the changes with quantity were fairly similar for all compounds. On average over quantities, the percentage of bees responding varied
2.4.2. EAG trials Analysis of the maximum electrophysiological responses included data for the solvent. The data were log transformed before analysis, primarily to make the variances more homogeneous. However, the logtransformation also allows the maximum responses normalised by the maximum responses to the solvent to be obtained and assessed within the analysis, since on the log-scale, the ratio of a value to the value of the solvent becomes a diﬀerence. Thus in the analysis, the diﬀerences 5
Journal of Insect Physiology 121 (2020) 104002
F. Mas, et al.
Fig. 2. Mean Proboscis Extension Response (PER) by experienced foragers for each compound and quantity in each season. Error bars are 95% conﬁdence limits, and are shown for three selected means from within the plots only (smallest, largest and a mid-range mean from within the plot). For clarity, these means have been redrawn to the side, along with their errors (see methods for an explanation). 6
Journal of Insect Physiology 121 (2020) 104002
F. Mas, et al.
Fig. 3. Mean normalised electroantennographic (EAG) response (in mV) by experienced foragers for each compound and quantity in each season. Error bars are 95% conﬁdence limits, and are shown for three selected means from within the plots only (smallest, largest and a mid-range mean from within the plot). For clarity, these means have been re-drawn to the side, along with their errors (see methods).
Journal of Insect Physiology 121 (2020) 104002
F. Mas, et al.
3.3. EAG responses of naïve bees following PER training
with compound in the spring, autumn and winter for group 1 (p < 0.001) and in the spring and autumn and winter for group 2 (p < 0.001, 0.04 and < 0.001 respectively), but varied much less in the summer for both groups of compounds (group 1: p = 0.21, group 2: p = 0.08). Similarly, the response averaged over compounds varied with loading quantity in the spring autumn and winter for group 1 (p < 0.001), but only a little in the summer (p = 0.06). For group 2, the quantity response varied in all four seasons (p < 0.001, < 0.001, 0.01 and 0.02 for spring, summer, autumn and winter). For all seasons, the highest PER response for the group 1 was for phenylacetaldehyde and for group 2 linalool and 4-oxoisophorone. In comparison, for all seasons the lowest PER response for group 1 was decanal and for group 2, methyl-p-anisate and (E,E)-α-farnesene.
The responses for naïve bees from the three groups are shown in Fig. 4. The (E,E)-α-farnesene and methyl-p-anisate triggered the lowest EAG responses on average, while benzaldehyde and nonanal triggered the highest EAG responses, irrespective of the bee treatments. The mean maximum response to hexane was substantially higher for the trained and learned bees (mean of 0.1) than for the untrained (0.06) and trained and not learned (0.05) bees, and this was true also for each chemical (Fig. 4, left). This may imply that training and learning results in a generally enhanced response when exposed to a chemical, regardless of what that compound is. Thus, normalising may potentially be obscuring important eﬀects. There was substantial variation in the maximum response between the chemicals (p < 0.001), but, despite the consistently higher maximum responses for trained bees, there was no signiﬁcant eﬀect relating to the bee treatments (p = 0.34 for the bee treatment main eﬀect and p = 0.95 for the interaction with chemical). Since the interaction between chemical and training is essentially assessing the normalised response to training, there was thus also little diﬀerence between the bee treatments in the normalised response to the various chemicals (p = 0.93 for the chemical by treatment interaction for chemicals excluding hexane).
3.2.2. EAG For both groups of compounds, there were very strong increases in the maximum response with increasing quantities for all four seasons (p < 0.001 for the quantity main eﬀect in all four seasons), Fig. 3. The maximum response, and thus the normalised responses, increased with increasing quantities of compound, regardless of the compound. For group 1, there were signiﬁcant compound main eﬀects in spring and winter (p = 0.01, p = 0.04 respectively), and signiﬁcant compound by quantity interactions in the summer and autumn (p = 0.02 and p = 0.04 respectively), with the interaction indicating that the changes with quantity varied with compound. These eﬀects were at least partly because the normalised responses to benzaldehyde were consistently higher at all quantities than the responses to the other chemicals, except for 1000 µg in winter, where the response was similar to decanal. For group 2, there were signiﬁcant compound by quantity interactions in all four seasons (p = 0.03, p < 0.01, p < 0.001, p < 0.001 respectively). Thus, the changes with increasing quantity varied noticeably among the chemicals in all four seasons. In the spring, the normalised responses increased more rapidly for nonanal than for the other compounds, and the response to increasing quantities of this compound was also the strongest in the summer and winter, but similar to linalool in summer. The responses to methyl p-anisate and (E,E)-α-farnesene were weaker than for the other chemicals in the summer, autumn and winter.
4. Discussion We present for the ﬁrst time the key bioactive compounds that bees detected from four fruit crops (cherry, plum, apricot and avocado), and from previously published key compounds of three vegetable crops (carrot, radish and cabbage) (Mas et al., 2018). These results show that bees responded to a speciﬁc set of key compounds for each crop independent of their quantity present in each variety. Some key compounds are common across some crops and some are speciﬁc to a crop. Nonanal and phenylacetaldehyde, which are present in some fruit crops, were also common bioactive compounds across the three vegetable crops. Linalool was also a common bioactive compound among fruit crops. Experienced foragers responded to every compound in any given season, but with varying electrophysiological intensity and behavioural
Fig. 4. Mean maximum (left) and normalised (right) electroantennographic response (in mV) to each of 12 compounds and hexane by naïve bees of forager age reared in the laboratory. Bees were trained on a mix of all 12 compounds and classiﬁed as trained that learned, trained that did not learn, or not trained. Compounds are in increasing order of the mean maximum response to a compound. Error bars are 95% conﬁdence limits. 8
Journal of Insect Physiology 121 (2020) 104002
F. Mas, et al.
(Gervasi and Schiestl, 2017), suggesting a preference for this key volatile. Linalool also triggered high PER responses in spring and winter, when bees may have been exposed to this compound produced by fruit crops blooming at this season. Interestingly, Nouvian et al. (2015) found that ﬂoral compounds like linalool and 2-phenyethanol can overtake the decision of recruitment triggered by the bee alarm pheromone, demonstrating the power of appetitive odour. In contrast, methyl-p-anisate and (E,E)-α-farnesene triggered the lowest PER responses at any quantity across the four seasons, as they did with EAG for experienced and naïve bees of forager age. Therefore, for these compounds, poor sensory responses may predict poor behavioural preferences. But also, as discussed previously, these two compounds may not be strongly associated with food reward, or only be active at much higher quantities. Finally, we did not detect any substantial diﬀerences in normalised EAG responses between naïve bees of forager age, whether or not they had been trained or had learned. However, naïve bees that were trained and had learned had overall larger maximum EAG responses for all compounds before normalisation with the hexane solvent. Thus, the lack of a treatment eﬀect may be due to normalisation to the solvent compound that bees could have learned as well as the odour. Also, for practicality, we trained all the naïve bees of forager age to the mix of the 12 compounds at a single quantity, but tested them on individual compounds. It is possible that some compounds could have masked some others; but we know from our previous results that bees are able to detect each of these compounds individually. Claudianos et al. (2014) showed that learning actually lowers EAG responses and induces the down-regulation of olfactory receptors (ORs), as they have shown in particular with linalool. With our results, there were some tendencies for naïve bees of forager age that were trained and learned to have lower normalised EAG responses overall for some compounds, but this was not statistically signiﬁcant. In their paper, Claudianos et al. (2014) actually compared experienced foragers that were exposed to odours and rewarded with sugar to bees exposed to odours but not rewarded. Therefore, these authors tested the eﬀect of associative learning on OR expressions. We tested the diﬀerence among naïve bees of forager age between trained (i.e. exposed) and not trained (i.e. no previous experience) as well as comparing naïve bees of forager age that were exposed but did not learn. More work is needed to conﬁrm how associative learning in bees relates to various compounds, so that generalisations can be made on the plasticity of neuroperception in bees.
appetitive responses depending on the compound and the season. For all 12 compounds, EAG responses varied signiﬁcantly with quantities with the highest quantity triggering the highest responses. This is generally expected with dose response studies in electrophysiology. Although we did not reach an asymptotic value with the quantities used, this suggests that higher quantities may trigger even higher responses. (E,E)-α-farnesene, which was identiﬁed in a previous study on kiwifruit (Twidle et al., 2015), triggered the lowest EAG responses at every season, independent of the quantities. This sesquiterpene is common across several plant families (Knusden et al., 2006; Schiestl, 2010). It has been also associated with herbivore-induced responses (Kugimiya et al., 2010; Suckling et al., 2012). Therefore, this compound may be too common in the environment, and variation in its abundance may not aﬀect bees unless a higher quantity is perceived, which we did not test. Benzaldehyde and nonanal triggered the largest EAG responses overall in every season for experienced foragers, as well as for naïve bees of forager age, whether they were trained or not. These two compounds are ubiquitous in plants (Knusden et al., 2006; Schiestl, 2010). However, increased nonanal was previously correlated negatively with seed yield in the vegetable crops (Mas et al., 2018). We also observed decreased PER responses with the highest quantity of nonanal tested with winter and summer experienced foragers, but increased PER responses with higher quantities for autumn and spring experienced foragers. Therefore, this compound may vary with ecological situations and foragers may learn to be attracted or repelled depending on the associative learning conditions. In contrast, benzaldehyde almost always triggered higher PER responses with higher quantities, except for experienced foragers in summer. Thus, compounds triggering the same level of electrophysiological responses can produce diﬀerent behavioural responses, so no generic behavioural conclusions can be drawn from EAG results only. Overall, there was no strong variation in patterns of EAG responses to each compound across seasons, and variation between compounds was more detectable at higher quantities than lower quantities. Thousands of micrograms (equivalent to mg) of a compound may be more associated with odour production by a group of ﬂowers than only one ﬂower, which may be more detectable and attractive to bees in the environment. For the PER responses, on average over the quantities, the percentage of experienced foragers responding varied with compound in most season, but varied much less in the summer for both groups of compounds. We expected variations of the PER responses among the seasons, since the PER varies with associated learning through foragers’ experience. Bees tested at each season may have had a diﬀerent foraging experience depending on the season. Changes in behavioural responses with seasons have been previously reported by Ray and Ferneyhough (1997) who found that bees learned the best in winter and the worst in summer. Our results from experienced foragers conﬁrmed that pattern. Yet, Scheiner et al. (2003) showed that there was a strong variation in water and sucrose responsiveness during the foraging season, aﬀecting proboscis extension learning in honey bees. We hypothesise that bees do not need to associate speciﬁc plant odours with food reward as much in summer when there is plenty of food available. Also, foragers live for a shorter time and there is consequently a faster turnover of foragers during this season. During their four weeks of longevity in summer, one forager may experience only one type of crop. In fact, vegetable crops usually bloom over four weeks during summer, compared with a few days of blooming in spring for fruit crops. Phenylacetaldehyde was the compound triggering the largest number of experienced foragers to respond with appetitive responses, particularly at the highest quantity tested, except in summer. This compound was previously found to be positively correlated with the food reward quantity and used as an honest signal by bumble bees (Knauer and Schiestl, 2015). Also, it was demonstrated experimentally that Brassica rapa will evolve higher phenylacetaldehyde production when pollinated only with bumble bees over several generations
5. Conclusion Our results show that experienced foragers respond electrophysiologically to a small subset of ﬂoral compounds for each crop. Each of these key bioactive compounds can be perceived by bee antennae at any quantities and any time in the year, but the associated appetitive responses vary with each compound and across seasons. Some compounds such as linalool and phenylacetaldehyde triggered the strongest appetitive response, although they were not the compounds triggering the highest EAG responses. On the contrary, (E,E)-αfarnesene and methyl-p-anisate had low PER responses overall and triggered the lowest EAG, particularly at high quantities. Therefore, there was no strong correlation between olfactory sensitivity (EAG responses) and appetitive responses (PER responses) across compounds. Nevertheless, knowing which compounds can be detected by the bee’s olfactory system (i.e. sensory bias), thereby triggering behavioural responses – either with associative learning (i.e. acquired preference) or without previous experience (i.e. innate preference) – may reveal preferences for ﬂoral compounds that may be exploited to enhance honey bee pollination. Acknowledgement We thank South Paciﬁc Seeds for access to the vegetable crops in the 9
Journal of Insect Physiology 121 (2020) 104002
F. Mas, et al.
Canterbury plains, as well as many other growers managing fruit crops in the North and South Islands.
Matsumoto, Y., Menzel, R., Sandoz, J.-C., Giurfa, M., 2012. Revisiting olfactory classical conditioning of the proboscis extension response in honey bees: a step toward standardized procedures. J. Neurosci. Methods 211, 159–167. Menzel, R., 1968. Das Gedächtnis der Honigbiene für Spektralfarben [The memory of the honeybee for spectral colours]. Zeitschrift für vergleichende Physiologie 60, 82–102. Menzel, R., 1985. Learning in honey bees in an ecological and behavioral context. Experimental Behavioral Ecology 31, 55–74. Menzel, R., 1993. Associative learning in honey-bees. Apidologie 24, 157–168. Menzel, R., Erber, J., 1978. Learning and memory in bees. Sci. Am. 239, 102–110. Menzel, R., Greggers, U., 2013. Guidance by odors in honeybee navigation. J. Comp. Physiol. A 199, 867–873. Menzel, R., Wittstock, S., Sugawa, M., 1990. Chemical Codes of Learning and Memory in Honey-Bees. Biology of Memory 23, 335–355. Nouvian, M., Hotier, L., Claudianos, C., Giurfa, M., Reinhard, J., 2015. Appetitive ﬂoral odours prevent aggression in honeybees. Nat. Commun. 6, 10247. Payne, R., Welham, S., Harding, S., 2017. A Guide to REML in Genstat, 19th ed. VSN International, Hemel Hempstead, UK. Pham-Delegue, M.H., Blight, M.M., Kerguelen, V., Metayer, M.l., Marion-Poll, F., Sandoz, J.C., Wadhams, L.J., 1997. Discrimination of oilseed rape volatiles by the honeybee: combined chemical and biological approaches. Entomologia Experimentalis et Applicata 83, 87-92. Pham-Delegue, M.H., Metayer, M.l., Picard-Nizou, A.L., Wadhams, L., 1992. Screening of oilseed rape volatiles behaviorally active in honeybees: coupled gas chromatographybehavioral assays: Summary. Apidologie 23, 481-483. Raguso, R.A., 2004. In: Advances in Insect Chemical Ecology, pp. 151–178. Raguso, R.A., 2008a. Start making scents: the challenge of integrating chemistry into pollination ecology. Entomol. Exp. Appl. 128, 196–207. Raguso, R.A., 2008b. Wake up and smell the roses: The ecology and evolution of ﬂoral scent. Annu. Rev. Ecol. Evol. Syst. 39, 549–569. Raguso, R.A., 2009. Floral scent in a whole-plant context: moving beyond pollinator attraction. Funct. Ecol. 23, 837–840. Ramirez, G., Fagundez, C., Grosso, J.P., Argibay, P., Arenas, A., Farina, W.M., 2016. Odor Experiences during Preimaginal Stages Cause Behavioral and Neural Plasticity in Adult Honeybees. Front. Behav. Neurosci. 10, 105. Ray, S., Ferneyhough, B., 1997. Seasonal variation of proboscis extension reﬂex conditioning in the honey bee (Apis mellifera). J. Apic. Res. 36, 108–110. Reinhard, J., Sinclair, M., Srinivasan, M.V., Claudianos, C., 2010. Honeybees learn odour mixtures via a selection of key odorants. PLoS ONE 5, e9110. Reinhard, J., Srinivasan, M.V., Guez, D., Zhang, S.W., 2004a. Floral scents induce recall of navigational and visual memories in honeybees. J. Exp. Biol. 207, 4371–4381. Reinhard, J., Srinivasan, M.V., Zhang, S., 2004b. Olfaction: scent-triggered navigation in honeybees. Nature 427. Scheiner, R., Barnert, M., Erber, J., 2003. Variation in water and sucrose responsiveness during the foraging season aﬀects proboscis extension learning in honey bees. Apidologie 34, 67-72. Schiestl, F., Marion-Poll, F., 2002. Detection of physiologically active ﬂower volatiles using gas chromatography coupled with electroantennography. Analysis of taste and aroma. Springer 173–198. Schiestl, F.P., 2010. The evolution of ﬂoral scent and insect chemical communication. Ecol. Lett. 13, 643–656. Schiestl, F.P., Johnson, S.D., 2013. Pollinator-mediated evolution of ﬂoral signals. Trends Ecol. Evol. 28, 307–315. Struble, D.L., Arn, H., 1984. Combined gas chromatography and electroantennogram recording of insect olfactory responses. Techniques in Pheromone Research. Springer 161–178. Suckling, D., Twidle, A., Gibb, A., Manning, L., Mitchell, V., Sullivan, T., Wee, S., ElSayed, A., 2012. Volatiles from apple trees infested with light brown apple moth larvae attract the parasitoid Dolichogenidia tasmanica. J. Agric. Food. Chem. 60, 9562–9566. Thiery, D., Bluet, J.M., Pham-Delegue, M.H., Etievant, P., Masson, C., 1990. Sunﬂower aroma detection by the honeybee. Study by coupling gas chromatography and electroantennography. J. Chem. Ecol. 16, 701–711. Twidle, A.M., Mas, F., Harper, A.R., Horner, R.M., Welsh, T.J., Suckling, D.M., 2015. Kiwifruit ﬂower odor perception and recognition by honey bees, Apis mellifera. J. Agric. Food. Chem. 63, 5597–5602. von Frisch, K., 1923. Uber die'Sprache'der Bienen. Zoologische Jahrbucher Abteilung fur Allgemeine Zoologie und Physiologie der Tiere 40, 1–186. VSN International Ltd, 2013. CycDesigN 5.1 A package for the computer generation of experimental designs., 4.0 ed. VSN International Ltd, Hertfordshire, England. Wadhams, L.J., Blight, M.M., Kerguelen, V., Lemetayer, M., Marionpoll, F., Masson, C., Phamdelegue, M.H., Woodcock, C.M., 1994. Discrimination of oilseed rape volatiles by honey-bee - novel combined gas-chromatographic electrophysiological behavioral assay. J. Chem. Ecol. 20, 3221–3231. Winston, M.L., 1991. The biology of the honey bee. Harvard University Press.
Funding This work was supported by the New Zealand Ministry of Business, Innovation and Employment (Grant No: C11X1309). Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.jinsphys.2019.104002. References Aizen, M.A., Garibaldi, L.A., Cunningham, S.A., Klein, A.M., 2008. Long-term global trends in crop yield and production reveal no current pollination shortage but increasing pollinator dependency. Curr. Biol. 18, 1572–1575. Ashman, T.-L., Knight, T.M., Steets, J.A., Amarasekare, P., Burd, M., Campbell, D.R., Dudash, M.R., Johnston, M.O., Mazer, S.J., Mitchell, R.J., 2004. Pollen limitation of plant reproduction: ecological and evolutionary causes and consequences. Ecology 85, 2408–2421. Bosch, J., Kemp, W.P., 2002. Developing and establishing bee species as crop pollinators: the example of Osmia spp. (Hymenoptera: Megachilidae) and fruit trees. Bull. Entomol. Res. 92, 3–16. Bromenshenk, J.J., Henderson, C.B., Seccomb, R.A., Welch, P., Debnam, S., Firth, D., 2015. Bees as biosensors: Chemosensory ability, honey bee monitoring systems, and emergent sensor technologies derived from the pollinator syndrome. Biosensors 5, 678. Claudianos, C., Lim, J., Young, M., Yan, S., Cristino, A.S., Newcomb, R.D., Gunasekaran, N., Reinhard, J., 2014. Odor memories regulate olfactory receptor expression in the sensory periphery. Eur. J. Neurosci. 39, 1642–1654. Delaplane, K.S., Mayer, D.F., 2000. Crop pollination by bees. Cabi, Cambridge. Free, J.B., 1963. The ﬂower constancy of honeybees. J. Anim. Ecol. 32, 119–131. Free, J.B., 1993. Insect pollination of crops. Academic press. Garratt, M.P., Coston, D.J., Truslove, C.L., Lappage, M.G., Polce, C., Dean, R., Biesmeijer, J.C., Potts, S.G., 2014. The identity of crop pollinators helps target conservation for improved ecosystem services. Biol. Conserv. 169, 128–135. GenStat Committee, 2015. GenStat Reference Manual Release 18 - Part: Procedures. VSN International, Oxford, pp. 1351. Gervasi, D.D., Schiestl, F.P., 2017. Real-time divergent evolution in plants driven by pollinators. Nat. Commun. 8, 14691. Giurfa, M., Nunez, J., Chittka, L., Backhaus, W., 1994. Odour and colour information in the foraging choice behaviour of the honeybee. J. Comp. Physiol. A. 175. Giurfa, M., Núñez, J., Chittka, L., Menzel, R., 1995. Colour preferences of ﬂower-naive honeybees. J. Comp. Physiol. A. 177, 247–259. Henning, J.A., Teuber, L.R., 1992. Combined gas chromatography-electroantennogram characterization of alfalfa ﬂoral volatiles recognized by honey bees (Hymenoptera: Apidae). J. Econ. Entomol. 85, 226–232. Kenward, M.G., Roger, J.H., 1997. Small sample inference for ﬁxed eﬀects from restricted maximum likelihood. Biometrics 983–997. Klein, A.-M., Vaissière, B.E., Cane, J.H., Steﬀan-Dewenter, I., Cunningham, S.A., Kremen, C., Tscharntke, T., 2007. Importance of pollinators in changing landscapes for world crops. Proceedings of the Royal Society of London, Series B: Biological Sciences 274, 303-313. Knauer, A.C., Schiestl, F.P., 2015. Bees use honest ﬂoral signals as indicators of reward when visiting ﬂowers. Ecol. Lett. 18, 135–143. Knusden, J.T., Eriksson, R., Gershenzon, J., Stahl, B., 2006. Diversity and distribution of ﬂoral scent. Bot. Rev. 72, 1–120. Kobayashi, K., Arai, M., Tanaka, A., Matsuyama, S., Honda, H., Ohsawa, R., 2012. Variation in ﬂoral scent compounds recognized by honeybees in Brassicaceae crop species. Breeding Science 62, 293–302. Kugimiya, S., Shimoda, T., Tabata, J., Takabayashi, J., 2010. Present or past herbivory: a screening of volatiles released from Brassica rapa under caterpillar attacks as attractants for the solitary parasitoid, Cotesia vestalis. J. Chem. Ecol. 36, 620–628. Lee, Y., Nelder, J.A., Pawitan, Y., 2006. Generalized Linear Models with Random Eﬀects: Uniﬁed Analysis via H-likelihood. Chapman & Hall/CRC Press, London, pp. 416. Mas, F., Harper, A., Horner, R., Welsh, T., Jaksons, P., Suckling, D.M., 2018. The importance of key ﬂoral bioactive compounds to honey bees for the detection and attraction of hybrid vegetable crops and increased seed yield. J. Sci. Food Agric. 98, 4445–4453.