To successfully persist in a chemosensory environment, animals must receive and interpret cues and signals of ecologically important information, such as the quantity and quality of resources potentially available to them [1]. This is especially true of pollinators such as bumblebees, which integrate multi-modal signals, including form, colour and scent, to accurately identify rewarding flowers [2]. Like other food resources, flowers host varied microbial species and communities [3,4], which produce metabolites that may act as cues of resource availability and quality, with consequences for pollinator foraging [5,6]. Indeed, insect pollinators are highly sensitive to shifts in volatile organic compound abundance and identity [7][8][9], with scents known to both influence foraging preferences and mediate learning [10]. However, the role of microbial volatile organic compounds (mVOCs), as well as those that are non-volatile, in mediating pollinator attraction and foraging decisions still remains largely unclear.
In standing crop nectar, bacteria and fungi colonize between 20 and 70% of individual flowers and can reach densities exceeding 10 7 and 10 5 cells ml 21 , respectively [3,4]. Upon colonization, these microbes metabolize sugars and & 2019 The Author(s) Published by the Royal Society. All rights reserved.
amino acids [5,11], affecting pollinator foraging and plant reproduction [5,12,13]. Intense competition between microbes in nectar often results in flowers that are dominated by either yeast or bacteria [14]. Yeasts and bacteria differ in mVOC composition and acceptance to pollinators [6], but also differentially influence non-volatile nectar traits [15] and shift pollinator perceptions of nectar quality [16]. Predicting microbial effects on pollinator foraging and behaviour requires examining responses to olfactory (headspace mVOCs) and gustatory (dissolved chemicals) cues.
Here, we test the hypothesis that yeasts and bacteria differentially influence bumblebee attraction and feeding. Bumblebees (Bombus impatiens) represent an ideal animal system for testing this hypothesis, owing to their close ecological and evolutionary relationships with yeasts [17,18] and bacteria [19,20]. In this study, we addressed the following research questions. First, can bumblebees perceive mVOCs? And if so, how do they affect preference? Second, how do nectar-inhabiting microbes influence bumblebee gustation? Finally, how does gustation experience influence bumblebee preferences for mVOCs? Through the use of electroantennography (EAG), olfactometer (Y-tube) bioassays, and choice and no-choice gustation assays, we discovered that bumblebees exhibit distinct preferences for mVOCs versus gustatory cues, with microbial metabolites informing foraging decisions in both a species-specific manner and modality.
We used three colonies of the generalist bumblebee Bombus impatiens (Koppert Biological Systems, Inc.; Howell, MI, USA) and a single strain each of the nectar-inhabiting yeast Metschnikowia reukaufii (Metschnikowiaceae; GenBank ID: MF319536) and bacterium Asaia astilbes (Acetobacteraceae; GenBank ID: KC677740). Both Metschnikowia and Asaia are commonly isolated from floral nectar [21] and pollinators [22], but are known to differentially influence nectar chemistry and scent [23]. With respect to nectar chemistry, prior work has revealed that Asaia can cause greater reductions in nectar pH than Metschnikowia, while also simultaneously increasing glucose and fructose concentrations to a greater degree [22]. As for scent, the mVOC blend emitted by Metschnikowia is characterized by esters, including ethyl butyrate, 2-methylpropyl acetate and 3-methylbutyl acetate, and the alcohols 2-butanol and 3-ethoxy-1-propanol, and a relatively greater abundance of ethyl acetate, and alcohols 3-methyl-1butanol, 2-methyl-1-butanol, 3-methyl-3-buten-1-ol, ethanol and 2-phenylethanol. Asaia emits many of these same compounds, albeit at significantly lower concentrations. The Asaia mVOC profile is also distinguished by the presence of the metabolite 2,5-dimethylfuran [6].
(b) Experiment 1: can bumblebees perceive microbial volatile organic compounds?
We examined responses of antennal olfactory neurons to mVOCs produced by Metschnikowia and Asaia (table 1) by puffing each metabolite (0.4 mmol) over excised B. impatiens antennae (N ¼ 6/ metabolite), following an established protocol [6]. Recorded antennal responses were standardized using responses to both blanks and a positive control stimulus (0.4 mmol geraniol). For additional details, see electronic supplementary material, S1 Methods. To assess which mVOCs were detected by bumblebees, we used t-tests with false discovery rate correction to examine if normalized EAG responses were significantly different from zero (i.e. no detectable response). All analyses here and below were performed in R (v. 3.5.2) [24].
(c) Experiment 2: how do microbial volatile organic compounds influence bumblebee preference?
To assess whether bumblebees exhibit an innate preference when exposed to mVOCs, we used an olfactometer assay (Y-tube; electronic supplementary material, figure S1), with the assay performed under red light. Naive bumblebees housed at the University of California, Davis were starved for 6 h, then released individually into the Y-tube. For each bee, both initial choice and time spent in each arm were recorded, with the assay repeated twice for each bee. Across assays, the treatment assignment for each arm was reversed, with preference measured over a 5 min period. Treatments consisted of synthetic nectar (3% w/v sucrose; 6% w/v each of glucose and fructose; 0.1 mM each of glycine, L-alanine, L-asparagine, L-aspartic acid, L-glutamic acid, L-proline and L-serine) [25,26], inoculated with each respective microbe at an initial density of 10 3 cells ml 21 from actively growing subcultures, and incubated at 298C for 4 days. This initial density is a magnitude or more below averages typically observed in flowers in the field [3,4]. Bees tested were both fed and treated similarly to those used for EAG assays. Overall, a total of 32 bees were tested, and sourced from two colonies. For additional experimental details, see electronic supplementary material, S1 Methods. To determine if bees have a preference for different microbes, data were analysed with a binomial mixedmodel for first choice, implemented with the lme4 package [27]. Bee identity and source colony were treated as random effects.
A linear mixed-effect (LME) model was used for time spent in each arm, implemented with the nlme package [28], with microbial treatment as a fixed effect, and bee individual and colony source as random effects.
To first assess gustatory preferences of bumblebees (N ¼ 42 bees from two colonies) for nectar colonized by microbial taxa, we used a no-choice feeding assay. In this assay, bees were housed in individual vials with modified lids that accommodated a feeding apparatus (electronic supplementary material, figure S2). The feeding apparatus consisted of attached vials, filled with 1 ml of either Asaia-or Metschnikowia-treated nectar. Before presentation to foragers, these vials were weighed, and then bees were allowed to feed for 24 h, after which tubes were re-weighed to determine consumption. Additional vials without bees were included to account for potential differences in evaporation among nectar treatments (N ¼ 3 per treatment). For details, see electronic supplementary material, S1 Methods. We used a t-test to assess how nectar consumption was affected by the nectar treatment.
(e) Experiment 4: how does gustation experience influence bumblebee preferences for microbial volatile organic compounds?
Because bees exhibited marked differences in preference for mVOCs versus gustatory cues (see Results below), we also assessed how exposure to gustatory cues influenced bee preference for mVOCs (N ¼ 24 bees from two colonies). Individual foragers were subjected to the olfactometer assay (electronic supplementary material, figure S1), then a gustatory choice assay where individual bees were housed in a feeding chamber, consisting of approximately 9 cm of perforated tubing, with feeding vials on either royalsocietypublishing.org/journal/rsbl Biol. Lett. 15: 20190132 end of the chamber (electronic supplementary material, figure S3) for 24 h, each containing a different microbial-conditioned nectar. Vials were weighed to determine nectar consumption. Bees were then subjected to a second olfactometer assay. In order to determine whether olfactory preferences changed before and after gustation experience, we fit an LME model with proportion of time spent in olfactometer arms as the response variable, nectar treatment and choice test order as fixed effects, and bee individual and source colony as random effects. An interaction between choice test order and nectar treatment was also included as a term in the model. Bumblebee feeding preferences were also analysed with an LME model, with amount consumed as the response variable, nectar treatment as a fixed effect, and bee individual and source colony as random effects.
(a) Experiment 1: can bumblebees perceive microbial volatile organic compounds? royalsocietypublishing.org/journal/rsbl Biol. Lett. 15: 20190132 approximately 50% more Metschnikowia-conditioned nectar when presented with a choice (F 1,46 ¼ 12.29, p ¼ 0.001), mirroring results observed in Experiment 3. After experiencing these microbial olfactory and gustatory cues however, bees reduced the frequency (albeit non-significant) with which they chose the Asaia mVOC blend in the second olfactometer test (figure 1c; x 2 ¼ 0.83, p ¼ 0.41), with many foragers making 'no choice' at all (i.e. remained active in the main channel of the olfactometer). These bees also increased the amount of time spent in the Metschnikowia arm of the olfactometer, but this increase was not significant (F 1,163 ¼ 0.11, p ¼ 0.74).
Collectively, our results indicate that volatile and non-volatile microbial metabolites can shape interspecific, plant-pollinator signalling. More specifically, microbial VOCs were both perceived by and could influence bee preference. Across behavioural assays, bees were more attracted to the mVOC blend produced by the bacterium Asaia over yeast Metschnikowia. We hypothesize that the metabolite 2-ethyl-1-hexanol, which elicited the strongest EAG depolarization response, may play a role in mediating this response: Asaia emits nearly twice the amount of this mVOC [6]. The gustatory preference exhibited by bees, however, was distinct from that observed in olfactory tests. Across all feeding assays performed, bumblebees consistently consumed more Metschnikowia-conditioned nectar. We suspect that this aversion to the taste of Asaia-conditioned nectar may be driven by metabolites dissolved in nectar, such as acetic acid. Asaia is known to significantly reduce nectar pH [22,23], likely through production of this organic acid. Though volatile, acetic acid was not detected in our previous screening of Asaia mVOCs [6] and we believe that due to its high aqueous solubility, it remained primarily dissolved in nectar tested and presented to foragers. Finally, our results suggest that bees may integrate experiences of volatile and non-volatile metabolites to inform future foraging decisions; however, future experimentation is required that explicitly disentangles exposure to these metabolites and how they collectively influence associative learning in bee pollinators.
In natural systems, bees must navigate a chemosensory landscape partly shaped by microbial associates of floral hosts. Though bumblebees may display innate preferences for particular mVOCs, as suggested in our naive forager olfaction test, foragers have potential to develop learned preferences for microbial metabolites through repeated exposure to both the scent and taste of yeast or bacterial-colonized nectar. Such preferences may manifest to affect patterns of floral constancy and the quantity and quality of benefits exchanged in these mutualistic interactions. It remains to be determined however whether pollinators benefit from microbial-derived cues, such as improved foraging efficiency through localization of resources. Alternatively, these cues may be more exploitative, and benefit microbes that rely upon pollinator dispersal to reach new floral habitats [29]. Such outcomes may hinge on both the identity and density of the microbial species encountered, where varied immigration histories can give rise to divergent microbial communities both within flowers of a host and among other species. Our results demonstrate that future investigations on the evolutionary ecology of floral signalling should consider the multiple ways in which microbes influence host phenotype and the innate and learned response of pollinators.