Global warming impairs chemical communication between plants and pollinators

Global warming is expected to have a great impact on oral scents and consequently, on the attraction of pollinators. While there is evidence that temperature affects the biosynthesis and emission of oral scents, the effects on chemically mediated pollinator behavior have not been assessed. Here, we show by chemical analytical, electrophysiological, and behavioral approaches that increased air temperatures affect the chemical communication between strawberry (Fragaria x ananassa) and its bee pollinators (Apis mellifera, Bombus terrestris, Osmia bicornis). Plants cultivated at increased temperatures released smaller amounts and a different composition of oral scent than plants cultivated at physiological optimum temperatures, which translated into a reduced attractiveness to pollinators. Our study highlights for the rst time that increased temperatures negatively affect the chemical communication between plants and their pollinators. This raises important ecological and agricultural questions, as impaired communication between plants and their pollinators might result in insucient pollination with negative consequences for ecosystem functioning and crop yields.

Here, we used a combination of chemical analytical, electrophysiological, and behavioral approaches to quantify the effects of increased air temperatures on oral scent emissions of strawberry (Fragaria x ananassa Duch -Rosaceae) and tested whether temperature-induced shifts in oral scent affect chemical communication with its main bee pollinators (Apis mellifera Linnaeus, Bombus terrestris Linnaeus, Osmia bicornis Linnaeus) 23 . Strawberry is an economically important crop cultivated worldwide 24 . To obtain good yields, pollination by bees is essential 23,25 . Strawberry plants were grown in plant-growth chambers under two temperature scenarios: physiological optimum (for growth and owering) 26 and 5 °C higher than optimal temperatures. Such an increase in temperature is predicted by the end of this century by the IPCC global warming scenario SSP-8.5 1 . Sampling of scent samples was performed by dynamic headspace, and samples were analyzed using GC/MS (gas chromatography/mass spectrometry). After analyzing the oral scent pro les of the two temperature scenarios, the identi ed scent compounds were tested on the antennae of the pollinators by GC/EAD (gas chromatography coupled with electroantennographic detection) to evaluate the compounds eliciting physiological responses in the antennae of the bees. Physiologically active compounds were used to build scent mixtures that resembled the scent of plants grown at the different temperature scenarios. These mixtures were used for behavioral assays that tested for their attractiveness to bee pollinators.
Our results showed that temperature strongly affected the amount of scent released, with ninefold less scent emission at warmer than optimum temperatures (9.55 ± 2.16 ng/ ower/hour versus 1.06 ± 0.59 ng/ ower/hour; Z= -3.14, N=19, p=0.001) (Table 1, Figure 1A). Temperature also in uenced the number and composition (Pseudo-F 1,18 =12.84, p<0.001) of compounds released by strawberry owers, since plants grown at optimum temperatures emitted ve compounds (four aromatic compounds and one sesquiterpene), whereas only 2 compounds (two aromatics) were detected in samples collected from plants cultivated at warmer temperatures. Plants in the optimum scenario released mainly the aromatic compound p-anisaldehyde (91%), whereas plants in the warmer scenario released this compound to a lesser extent, but mainly the aromatic compound 1,4-dimethoxybenzene (83%), a compound not detected in plants grown in the optimum scenario (Table 1, Figure 1B). All these compounds are widespread among bee-pollinated owers 8 . Our ndings are consistent with other studies that also found that oral scents change under heat stress 6,27-29 . In one study it was demonstrated that such changes are due to temperature dependent activities of enzymes and genes involved in the biosynthesis of scent compounds 14 . Overall, the oral scent strongly differed between the two different scenarios to an extent known not only within species but also among plant species 8,30,31 , suggesting that increased temperatures will strongly affect the communication between strawberry owers and their pollinators.
Indeed, our electrophysiological recordings revealed that all three pollinator species are sensitive to at least ve of the six ower scent compounds detected in the samples (Figure 2; see also 32-34 ). Benzyl tiglate, which was not tested in our analyses, can also be sensed by bees, as previously shown by physiological measurements that tested the scent of apple owers on honey bees 35 . Assuming that different receptors are involved in the detection of the different compounds, these ndings altogether suggest that the bees very likely perceive the scent of the plants grown at the two different temperature scenarios differently.
In the behavioral experiments, all three bees preferred both optimum scent and warmer scent against negative controls and optimum scent against warmer scent ( Figure 3). Two (A. mellifera, B. terrestris) of the three bee species were offered the same stimuli in a choice setting (optimum against optimum scent) and did not show a side bias. Therefore, the choices recorded in the tests, such as the reduced attractiveness of the warmer scent compared to the optimum scent, were indeed due to the stimuli, and not due to a side bias of the bees (Figure 3).
In a wild strawberry species, F. virginica, which releases 1-2 ng of scent per hour and produces hermaphrodite and female owers, scent was also found to be important in pollinator attraction 36 . In this wild species, hermaphrodite owers (ca. 1.8 ng/h) emitted slightly more scent than female owers (ca. 1.4 ng/h), mainly due to the emission of 2-phenylethanol from stamens. This "small" difference was enough that female owers received in the eld only half as many bee pollinator visits as hermaphrodite owers 36 . Thus, we predict that the observed strong decrease in the scent emission of cultivated strawberry grown under heat stress will result in the decreased attractiveness of the owers to the pollinators, and it will have strong negative effects on the overall yield and fruit quality 23,25 .

Conclusion
Our study shows for the rst time that increased temperatures predicted by global warming have not only strong negative effects on oral scent emissions but also on the attraction of pollinators. Together with the trend that modern crops emit fewer oral scents and are less attractive to pollinators than wild relatives 37-39 , this raises the important agricultural question of whether modern crops will continue to be su ciently attractive to pollinators. It also raises the question of how increased temperatures will affect pollinator attraction in natural settings and in uence ecosystem functioning. Currently, the molecular and biochemical processes in the biosynthesis of oral scent affected by heat stress are mostly unknown 5,14,40 , and therefore, more efforts are needed to elucidate the effects of increased temperatures on oral scent production and emission at a molecular level. This will help to develop breeding programs to increase the oral scent emissions of less heat-tolerant crops to avoid a mismatch with their pollinators and strong yield losses.  Strawberry is a gender-dimorphic species with hermaphroditic and female owers, but in contrast to wild strawberry 36 , we did not nd an effect of ower sex on scent emission (total amount of scent: Z= -0.32, N= 10, p= 0.761; relative scent composition: Pseudo-F 1,9 =1.12, p=0.323) and thus did not discriminate between the ower sexes in the present study.
The requirements of strawberry plants for relative humidity (range between 60-70%), light conditions, and water availability were controlled following literature data 26 . Plants were cultivated at 16 h light/08 h darkness. The light intensity was 2000 lx (via cool white led lamps, model VT-5959 LED-Flutlicht, V-TAC, 50 W). To keep the soil at comparable moisture levels during the development of the plants and between the different treatments, as measured by a tensiometer (model FDA 602 TM2, ALMEMO ® , Germany), the water supply varied according to the age of the plants and the temperature scenario. When they were sown, the amount was, independent of the scenario, 15 ml/plant/day, in the mid-ages 60 ml/plant/day (optimum scenario) and 90 ml/plant/day (warmer scenario), and, during the owering phase, 120 ml/plant/day (optimum scenario) and 170 ml/plant/day (warmer scenario).
The plants were grown under two temperature scenarios: physiological optimum and 5 °C higher than optimal temperatures (according the global warming scenario SSP-8.5, IPCC 2021). The physiological mean optimal temperature for the growth and owering of cultivated strawberry (F. ananassa) plants is 20 °C 26 . Considering the mean daily thermal amplitude in Central Europe 41,42 and the optimal physiological temperatures, strawberry plants were cultivated in the optimal scenario at day and night with temperatures of 23 °C and 13 °C (mean 20 °C, when considering the length of the day and night periods), respectively, and in the warmer scenario, the temperatures were 28 °C and 18 °C (mean 25 °C) during day and night, respectively. For each temperature scenario, 12 individual plants were cultivated.

Sampling and analysis of ower scents
Sampling of scent samples was performed inside the growth chambers by dynamic headspace. From 12 individuals cultivated for each scenario, 10 individuals of the optimum scenario and nine individuals of the warmer scenario produced owers during the experiment and were sampled. The samples were always obtained from owers on their rst day of anthesis and between 10 am and noon. A single ower per sample was enclosed in a polyester oven bag (Toppits ® ). After bagging, two small adsorbent tubes were inserted into the bag: one was used to trap the oral scent, and the other was used to insert clean air from outside of the growth chambers (to avoid internal air contamination). This sampling lasted 30 min using membrane pumps (G12/01 EB; Gardner Denver Thomas GmbH, Fürstenfeldbruck, Germany). The ows of both pumps were adjusted at 200 ml/min with the help of owmeters. The adsorbent tubes (quartz vials, length: 25 mm, inner diameter: 2 mm) were lled with 1.5 mg Tenax-TA (mesh 60-80) and 1.5 mg Carbotrap B (mesh 20-40, both Supelco). The adsorbents were xed in tubes using glass wool. Dynamic headspace samples of green leaves (N=3 samples per scenario) were collected with the same method to discriminate between vegetative (not considered for subsequent analyses) and ower-speci c scent components. Samples from empty oven bags (N=3) were collected to identify potential contaminants.
Scent samples were analyzed using GC/MS (gas chromatography/mass spectrometry). The system consisted of an automated thermal desorption system (model TD-20, Shimadzu, Japan) coupled to a QP2010 Ultra EI GC/MS (Shimadzu, Japan) equipped with a Zebron™ ZB-5 fused silica column (5% phenyl 95% dimethylpolysiloxane; 60 m long; inner diameter 0.25 mm; lm thickness 0.25 μm; Phenomenex), as described previously 43 . The GC/MS data were processed using GCMSsolution (Version 4.41, Shimadzu Corporation 2015). The tentative identi cation of compounds was carried out using the mass spectral libraries Wiley 9, Nist 2011/FFNSC 2, and Adams 44 , as well as the database available in MassFinder 3. The identity of all compounds was con rmed by a comparison of mass spectra and retention times with those of authentic standard compounds available at the Plant Ecology lab of the Paris-Lodron University of Salzburg. For estimation of total absolute scent emission per ower, we followed 45 .
Following identi cation and quanti cation of the oral scent compounds, synthetic mixtures (compounds that contributed at least 1% to the scent compositions of the two scenarios) that resembled the absolute and relative scent composition of the plants grown at the different temperature regimes were prepared to be used for electroantennographic detection. The mixture of the optimum scenario was composed of four compounds (99% of relative amount emitted; benzyl alcohol, methyl salicylate, p-anisaldehyde, (E,E)-αfarnesene), the mixture of warmer scenario of two compounds (100% of relative amount emitted; 1,4-dimethoxybenzene, p-anisaldehyde). Benzyl tiglate, which was additionally released by plants under the optimum scenario, was not included in the mixture, as it was only found in trace amounts in some samples.

Electroantennographic detection
The synthetic scent mixtures of the different scenarios were tested on the antennae of the three bee pollinators (N = 7 worker bee individuals each of A. mellifera and B. terrestris, and 5 females and 4 males of O. bicornis) by GC/EAD (gas chromatography coupled with electroantennographic detection) to evaluate the compounds eliciting antennal responses. The GC/EAD system, the same as that used by 46  , and connected to silver wires as described previously 47 .
A oral compound was considered EAD-active in a bee species when it elicited a depolarization response in at least four individuals. After identi cation of the EAD-active compounds, synthetic mixtures considering only these compounds were prepared for behavioral experiments. When pipetted on lter papers (diameter 3.7 cm; Whatman), sampled by dynamic headspace, and analyzed by GC/MS, these mixtures resembled the relative and absolute scent composition of the plants grown at the different temperature regimes. The synthetic scent mixtures were prepared with compounds available in the reference collection of the Plant Ecology lab of the Paris-Lodron University of Salzburg in the highest purity available (>90%). The solvent to dilute the scent mixtures and used as a negative control (see below) was acetone (Sigma-Aldrich, 99.8%).

Behavioral experiments
Behavioral assays were performed to test whether the differences in ower scent emissions from strawberry plants grown at the different temperature scenarios affect their olfactory attractiveness to bee pollinators. These assays were conducted outdoors (A. mellifera, B. terrestris) and indoors (O. bicornis) at the Paris-Lodron University of Salzburg. Tests with O. bicornis were performed indoors, as weather conditions during their ight period did not allow testing outdoors.
The outdoor behavioral experiments were performed in a ight cage (wooden construction clamped with white gauze of 8 x 4 x 2.2 m) in the Botanical Garden, the same as that successfully used with bees before 35 . In this ight cage, Reseda lutea (Resedaceae) continuously owers and is used as a pollen and nectar resource. The scent mixtures were offered on arti cial owers in dual-choice assays, with a distance of 1.50 m between them. The arti cial owers were made of blue bond paper of 7 cm diameter and tied on wooden sticks. White lter paper of 1 cm diameter (Whatman), onto which a mixture was applied, was placed in the middle of the arti cial ower (Supplementary Figure 1A and 1B). For each pollinator, we tested the optimum and warmer scents against negative controls (acetone) and against each other. To test for a potential side bias, scents of the optimum scenario were also tested against itself. The amounts of scent mixtures offered to pollinators were equivalent to 100 owering plant individuals, i.e., representing a small crop area (Supplementary Table 1).
A dual-choice assay lasted for 1 h, whereas the position of the arti cial owers/ lter papers were changed and the scent mixtures were renewed after 30 min. Bees that landed on arti cial ower/ lter paper were recorded and marked with a nontoxic pen (Posca ® -Tokyo, Japan) to avoid counting an individual bee twice.
A speci c assay was replicated until a minimum of 15 bees responded. Only in some assays with warmer scents versus the negative control was the number of responding bees smaller than 15, although the number of replicates was the same as for tests with the optimum scenario ( Figure 3).

Data analysis
Mann-Whitney U tests (PAST Version 2.17c) 48 and PERMANOVA (based on Bray-Curtis similarities of percentage contribution of single compounds to total scent 49 ; Primer 6 Version 6.1.15 & Permanova Version 1.0.5) were used to test for differences in the total amount and relative composition, respectively, of scent per ower between the different temperature scenarios. SIMPER was used to determine the compounds most responsible for relative differences in scent between the two temperature scenarios

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.