In both examples, drought resistance in Arabidopsis and pollinator attraction in tomato , the parasite gene controlling these extended phenotypes is the CMV 2b gene. A potential result of these extended phenotypes would be to increase the odds of continued survival of susceptible host plant populations, which would be beneficial to both host and pathogen. Our mathematical modeling results indicated that, for the areas of the parameter space that are most salient to our experimental findings, the most likely outcome of long-term selection would be coexistence of resistant and susceptible genotypes, i.e. the long-term maintenance of R gene polymorphisms. Several mechanisms have been proposed that could contribute to the maintenance of balanced R gene polymorphisms such as the ratio of costs versus benefits of resistance, and diffuse interactions between hosts and attackers. Our data suggest that the enhanced attraction of pollinators to infected susceptible plants might add to these mechanisms and contribute to the long-term maintenance of R gene polymorphisms in insectpollinated species. Production of many important crops depends on bee-facilitated pollination. Worryingly, bee populations are endangered by disease, environmental change and, 30 litre plant pots bulk more controversially, by anthropogenic factors. To mitigate the ensuing loss of pollination activity requires among other things a deeper understanding of the mechanisms shaping bee-plant interactions.

Our data show that non-floral plant volatiles can be perceived by bumblebees and affect their behaviour and that emission by plants of bee-perceivable compounds is regulated in part by miRNA activity. This information may be useful in developing strategies to increase pollinator services for crops under conditions of cultivation, as well as for a better understanding of the interplay of plant pathogens, wild plants and pollinators under natural conditions.Plants used were Arabidopsis thaliana accession Col-0 and Solanum lycopersicum cv. Moneymaker . Plants were grown in a growth chamber at 22°C in M3 compost . Tomato and Arabidopsis plants were grown under 16hr light/8hr dark and 8hr light/16hr dark photoperiods, respectively. Fny-CMV, Fny-CMVΔ2b, the 2b-transgenic plant line 2.30F, and the dcl1-9, dcl2/4, and ago1-25 mutant plant lines have been described elsewhere. CMV isolate PV0187 was obtained from the German Collection of Microorganisms and Cell Cultures . RNAs1, 2 and 3 of CMV isolate PV0187 were sequenced and submitted to GenBank under accession numbers KP165580, KP165581, and KP165582, respectively. Inoculations were carried out at the seedling stage and were performed as described previously. Plants were used in experiments when the virus had spread systemically and infection was confirmed routinely by double-antibody sandwich enzyme-linked immunosorbent assays . Quantification of CMV and CMVΔ2b RNA accumulation was carried out as previously described. Leaf tissue from systemically infected leaves was harvested at 10 and 18 dpi. Total RNA for reverse transcription coupled polymerase chain reaction analysis was extracted using an RNeasy Plant Kit and treated with TURBO-DNase according to the manufacturers’ instructions.

Reverse transcription was carried out with 0.5 μg total RNA using Goscript with random hexamerprimers according to the manufacturer’s instructions. Following the reaction, cDNA was diluted 1/10 for subsequent use. Semi-quantitative PCR was performed using Biomix Red and products were separated electrophoretically on a 1.5% agarose gel. Reverse transcription coupled to quantitative polymerase chain reaction analysis was performed using SYBR Green JumpStart Taq ReadyMix in 15 μl reactions according to the manufacturer’s instructions. Reactions were performed in triplicate. Primers described in were designed against the conserved 3’ non-translated regions of the CMV genomic RNAs and the stable transcript elongation factor 1 alpha was used as the reference RNA. Data were analyzed using LinRegPCR to give Ct values. Relative viral RNA accumulation was calculated using ΔΔCt methodology, incorporating the EF1α transcript to control for variation in loading.Bombus terrestris colonies were connected by gated transparent tubing to flight arenas with the dimensions 72 x 104 x 30 cm containing 11 cm tall feeding towers formed from black card sitting within ‘Aracon’ bases , roofed by plastic mesh supporting a microcentrifuge tube lid containing sucrose solution. Tower height was selected because bumblebees cannot effectively resolve visual cues beyond 10 cm. Seven days prior to carrying out conditioning or free choice assays bees were allowed to feed on sucrose solution from cups placed on empty towers for three days to familiarize them with the arena. Foraging bees were marked on the thorax with water-soluble paint and used once.Initially, cups on towers offered 30% sucrose, conditioning bees to associate towers with a reward.

For differential conditioning and free-choice experiments, five plants per treatment group were individually covered by towers. For differential conditioning experiments, towers hiding plants from one treatment group provided 0.3 ml quinine hemisulfate , whilst the others offered 0.3 ml of 30% sucrose. Individual foraging bumblebees were released into the arena and allowed to forage until satiated. Aborts following landing or hovering over towers offering quinine or drinking on towers offering sucrose were scored as correct choices. Between foraging bouts, towers were re-arranged randomly to inhibit spatial learning and meshes cleaned to remove scent marks. One hundred choices for each bee tested for each pair-wise comparison were recorded. In free-choice preference assays towers covering plants from both treatment groups offered equal sucrose rewards and only the first feeding choice was recorded. The learning curve data were analysed using binomial logistic regression. The experimental protocol did not record individual choices made by the bees, but instead the number of ‘correct’ choices made by each bee was grouped into sets of 10 successive choices for ease of scoring.Artificial buzz-pollination was carried out using an electrically actuated toothbrush . Mean seed mass was obtained by dividing the mass of seeds by the total seed number for a total of five fruits per plant, with three plants per treatment group. Pollen viability was assessed by staining with fluorescein diacetate and pollen grains viewed under blue light and bright field using an epi-fluorescent microscope connected to a digital camera . For bumblebee pollination experiments two-week-old tomato seedlings were inoculated with CMV or mock-inoculated and grown in a controlled environment room for 4 weeks. At this time, the plants began flowering and were transferred to a glasshouse. Two weeks later single bumblebees were allowed to buzz pollinate flowers on three mock-inoculated and three CMV-infected tomato plants within a larger flight arena constructed from nylon netting . Two inflorescences of two to three flowers per plant were left accessible to the bee . When each bee had made 10 visits to flowers , any buzz-pollinated flowers were labeled with a jeweler’s tag and all plants that had been visited by the bee were removed from the arena and replaced with another. A new bee was then released from the small arena into the larger arena containing plants. In total, 8 bees freely pollinated flowers from 17 mock inoculated and 14 CMV-infected tomato plants. Bumblebee visitation to mock-inoculated versus CMV-infected plants was noted and, using a stopwatch, the duration of flower sonication was recorded for each bee. The plants were left in the greenhouse for a further 8 weeks to allow fruits to develop. Further flower development on the plants was permitted. To release seeds, fruits were harvested individually into 60 ml screw-cap pots and left to ferment for 1–2 weeks before washing and counting. Fruits were either from flowers that were not buzz-pollinated by a bumblebee or from flowers that were buzz-pollinated . A further category of fruit was from flowers that were not buzz-pollinated, but were adjacent to fruit from buzz-pollinated flowers . Fruits were also harvested from eight mock-inoculated and eight CMV-infected plants that were not exposed to bees in the flight arena, whoelsale plant containers but had otherwise experienced the same growth conditions as the plants used in the bee pollination experiment .Headspace volatiles were collected from tomato plants by dynamic headspace trapping over a period of 24 hours onto Porapak Q filters [50 mg, 60/80 mesh size, Supelco ] as described by Beale and colleagues.

The tomato plants were contained in a 1.0 liter bell jar clamped to two semi-circular metal plates with a hole in the center to accommodate the stem. Charcoal-filtered air was pumped in at the bottom of the container at a rate of 750 ml.min-1 and drawn out through the Porapak Q filter at the top, at a rate of 700 ml.min-1. Leaf fresh weight and dry weight were measured to enable normalization of the volatile abundance. Trapped organic chemicals were eluted from the Porapak Q filter with diethyl ether for analysis by gas chromatography coupled to mass spectrometry . For initial investigation of volatiles by principal component analysis, volatiles were separated on a capillary GC column . The injection volume was 1μl, the injector temperature was 200°C, and helium was used as the carrier gas at a constant flow rate of 2.6 ml min−1 in an oven maintained at 30°C for 5 minutes and then programmed at 15°C.min-1 to 230°C. The column was directly coupled to a mass spectrometer with a MS transfer line temperature of 240°C. Ionization was by electron impact with an ion source temperature of 250°C in positive ionization. Mass ions were detected between 30 and 650 m/z. Data were collected using Xcalibur software . Principal component analysis on the mass spectra was performed with MetaboAnalyst 2.0using binned m/z and per cent total ion count values. Confirmation of identities of specific organic compounds comprising the blends emitted by mock-inoculated and virus-infected plants was carried out by re-analysis of trapped organic compounds using a Thermo-Finnigan Trace GC directly coupled to a mass spectrometer equipped with a cold on-column injector. Two microliters of collected volatiles were separated on an HP1 capillary gas chromatography column in an oven maintained at 30°C for 5 min and then programmed at 5°C.min-1 to 250°C. The carrier gas was helium. Ionization was by electron impact at 70 eV at 220°C. Compounds were identified by comparison of spectra with mass spectral databases , as well as by coinjection with authentic standards on a Hewlett-Packard 6890 gas chromatograph with two different columns of different polarity .Our model tracks the interaction over evolutionary time between virus resistant and virus susceptible phenotypes in a population of diploid annual plants. The plant population size is assumed to be large and to remain constant over generations. Since CMV is a broad hostrange pathogen, we can reasonably make the simplifying assumption that within-generation pathogen prevalence is not affected by the density of resistance in the focal host plant species. The proportion of susceptible plants that become virus infected in each generation is therefore held constant as a parameter in our model. We model resistance as controlled by a single bi-allelic locus, with resistant and susceptible forms, and we assume R is dominant. We assume infected plants produce fewer seeds, with the parameter δ controlling the proportionate number of viable seeds produced per ovary on a virus-infected plant. We additionally assume that virus resistance carries no fitness penalty when compared to uninfected susceptible hosts. If the reduction in seed number were the only consequence of virus infection, resistance would certainly fix in the plant population under such a conservative assumption on the cost of virus resistance for the plant. However, we also assume that increased attractiveness to pollinators means infected plants are more likely to reproduce, as both male and female parents. In particular, we assume the pollinator density remains constant over generations, and that this pollinator density leads to an average of γ pollinator visits per flower averaged over all plants over the entire reproductive season. We assume that flowers visited by pollinators will certainly be pollinated: by cross-pollination or by self-pollination . Self-pollination after a visit by a pollinator can be due to either geitonogamous pollen transfer from flowers on the same plant, or via autogamous buzz-pollination . A proportion σ of the remaining ovules in flowers that are not visited by pollinators also go on to self-pollinate. We assume that an individual pollinator is ν times more likely to visit a flower on an infected vs. an uninfected plant than would be expected by chance alone. This potentially increases female fitness by making ovules on infected plants more likely to be fertilized, and male fitness by increasing rates of pollen transfer from infected plants.The behaviour of the model can most easily be characterised in terms of which genotypes persist in the long-term.