A European CMV isolate, PV0187, which is 99% identical in RNA sequence to CMV-Fny and which encodes a 2b VSR that is identical in amino acid sequence to that of CMV-Fny , was used for these experiments in order to comply with UK quarantine and containment regulations. CMV-PV0187 had similar effects on growth of tomato plants as CMV-Fny and volatiles emitted by tomato plants infected with CMV-PV0187 were attractive to bumblebees in the free choice assay . When CMV-infected and mock-inoculated tomato plants were exposed to bumblebees, a higher proportion of the insects made their initial floral visits to CMV-infected plants and spent longer sonicating the flowers of CMV-infected plants . As had been seen for artificial buzz-pollination , when bumblebees buzzed flowers, seed yield was increased . For CMV-infected plants, when bees did not visit flowers or where flowers were on plants not exposed to bees , the seed yield was significantly decreased . However, although CMV infection decreased seed number in fruits derived from unvisited flowers, buzz-pollination by bumblebees negated this effect; indeed, bee-pollinated flowers on CMV-infected plants yielded fruit that contained seed numbers similar to those found in fruit that developed from bee-pollinated flowers on mock-inoculated plants . The results imply that there was greater buzzing activity on flowers of CMV-infected plants , 25 liter pot plastic resulting in a greater amount of seed production.We have seen that under controlled conditions CMV infection made tomato plants more attractive to bumblebees .
We also found that although infected plants yielded fewer seeds, simulated buzz-pollination could to some extent rescue seed production and when bees were allowed access to CMV-infected plants they caused a greater increase in seed production per fruit compared to simulated buzz-pollination . The results obtained with this domesticated plant under controlled conditions prompted us to wonder what would be the consequences for a wild buzz-pollinated plant growing under natural conditions, if virus infection resulted in greater pollinator visitation and/or seed production and whether this might result in any benefits for the host plant or the virus. To investigate this idea further we developed a mathematical model to test whether increased pollinator service to virus-infected plants could allow genes for virus susceptibility to persist in a host plant population, despite the significant fitness cost of infection for plants as female parents . Our model tracks the long-term dynamics of the interaction between resistant and susceptible phenotypes in a population of annual plants . We focused on resistance as a dominantly inherited trait and attached no fitness penalty to the presence of resistance, which is a conservative approach given that recessive resistance is a commonly observed antiviral defense mechanism and that resistance may incur fitness costs in the absence of infection. We assume infected susceptible plants produce fewer seeds, with the parameter δ controlling the proportionate number of viable seeds produced per fertilized ovary on a virus-infected plant. However, we also assume that, all other things being equal, an individual visit by a pollinator is ν times more likely to be to a flower on an infected versus an uninfected plant. This pollinator bias makes infected plants more likely to reproduce as both male and female parents, potentially out-weighing the deleterious effect of infection on seed production.
We focus initially on the trade-off between pollinator bias and reduction in seed production , for different levels of pollinator service . In indicative examples of both low and high pollination regimes, parameter space can be divided into three regions: resistance takes over in the long-term, susceptibility takes over in the long-term, or resistant and susceptible plants coexist . For both values of γ, at high values of ν and δ , then genes conferring susceptibility will take over in the plant population. For low values of ν and δ the situation is reversed, and resistance is favored. At intermediate values of ν and δ, resistant and susceptible plants coexist. For fixed baseline values of ν = 3.0 and δ = 0.5, the proportion of susceptible alleles in the population first increases then decreases as the level of pollinator service is increased . At very low values of γ, although virus-infected plants benefit from additional pollinator service on both male and female sides, the vast majority of fertilizations do not involve pollinator visits . The cost to susceptible plants of reduced seed production as female parents is therefore more important than increased pollinator visitation, and so virus resistance takes over. As γ is increased, the proportion of fertilizations caused by pollinators goes up, which allows the benefits to virus-infected plants on both male and female sides to outweigh the cost of infection, and so the genes for susceptibility are favored. As γ is increased still further, the benefit on the female side becomes smaller , but on the male side proportionately more pollen still comes from infected plants. For these values of the parameters, alleles conferring virus susceptibility persist in the plant population, but at reduced density. The maximum density of susceptible genotype plants is therefore realised at intermediate pollinator densities.
The broad pattern of a rise then fall in the proportion of plants carrying the susceptible allele is repeated for a range of values of the proportion of susceptible plants that are infected . However, for our default parameterization at low levels of infection the eventual fall with increasing pollinator levels is not apparent, and susceptible plants exclude resistant plants even for very high values of γ. A full sensitivity scan around default parameter values ν = 3.0, δ = 0.5, γ = 1.0, σ = 0.5, φ = 0.75 and α = 0.5, shows the behaviour of the model over large regions of parameter space. The susceptible genotype is able to persist under many combinations of parameters. Our model therefore suggests preferential visitation of infected plants by pollinators could in principle provide a robust mechanism allowing susceptible genotype plants to be retained in the host population for a wide range of conditions.Infection with CMV altered the volatile profile of tomato plants and made them more attractive to bumblebees, indicating that these insects possess an innate preference for the blend of volatile compounds emitted by CMV-infected tomato. Although bumblebees showed no innate preference for CMV-infected or mock-inoculated Arabidopsis plants, differential conditioning experiments showed that bumblebees were able to perceive alterations in volatiles emitted by these plants. Experiments with the 2b gene deletion mutant virus, CMVΔ2b, in tomato and Arabidopsis, and with 2b-transgenic and ago1 and dcl1 mutant Arabidopsis plants, implicate small RNA pathways in the regulation of the production of bee-perceivable volatile compounds. The inability of bees to learn to effectively distinguish between volatiles emitted by ago1 and dcl1 mutant plants causes us to conclude that miRNAs are the predominant class of small RNAs involved in regulating the metabolism of bee-perceivable compounds. The rationale for this conclusion is that AGO1, a target for the CMV 2b VSR, utilizes both short-interfering RNAs and miRNAs to guide RNA cleavage, while DCL1 is involved in miRNA biogenesis but is not involved in production of short-interfering RNAs . As far as we are aware, an effect of miRNAs on plant volatile production has not been previously reported. The work also indicates that olfactory signals emitted by non-floral tissue may have a more important effect than previously thought in plant-bee interactions and may play roles in bee attraction, presumably at longer ranges than visual features such as the optical or tactile qualities of flowers. Thus, foliar volatile signals may affect bee choices or synergize with and reinforce visual floral cues, 25 litre plant pot as has been seen with floral odors. How do changes in the output of volatiles increase the attractiveness of CMV-infected plants for bumblebees? Much of the existing bee perception literature is focused on the effects of visual stimuli , whereas the effects of olfactory stimuli have been relatively neglected. But it is known, for example, that the VOC output from flowers decreases after they have been pollinated. Pollination can also trigger qualitative changes in the volatile blend. For instance, following pollination by bees, blueberry flowers emit an increased proportion of their volatiles as -caryophyllene. It is thought that decreased volatile emission by pollinated flowers decreases their saliency to bees and prevents damage from over-visitation and a similar explanationwas offered by Rodriguez-Saona and colleagues to explain the post-visit increase in -caryophyllene emission. In the case of tomato plants infected with CMV, it may be that the virus is both ‘turning up the volume’ of plant volatile emission whilst ‘tuning’ volatile blend composition so as to diminish levels of a signal -caryophyllene, that at higher levels might indicate a previous bee visitation, and levels of β-phellandrene and 2-carene that might discourage visitation. When the bumblebees were allowed access to flowering tomato plants under glasshouse conditions we found that buzz-pollination by bumblebees was more effective at enhancing seed yield on CMV-infected plants.
This result suggests that additional foliar or floral cues, for example visual or tactile stimuli, do not negate the effects on the bees of CMV-induced changes in volatile emission. It is possible that our findings may have implications for transmission of viruses vectored by bees. However, pollinators transmit very few viruses and CMV is not one of them . Nevertheless, is it possible that a virus that is not bee transmitted gains some advantage by re-paying a susceptible host by altering its volatile cues to attract pollinators? In our mathematical model it was assumed that a hypothetical population of wild plants included some hosts that possessed genetic resistance to the virus. It might then be assumed that pathogen-imposed selection pressure would favor the takeover of the plant population by any plants possessing one or more resistance genes. This outcome, causing a decrease in the population, or even the extinction, of susceptible plants would clearly not be beneficial either for the pathogen or for the susceptible hosts. However, our mathematical model shows that in the case where pollinators show increased bias towards pathogen-infected plants, the increased reproductive success of infected plants means that the outcome might be different. Thus, it is plausible that if the attractiveness of infected plants to pollinators is increased, this might inhibit or negate the selective advantage of resistant individuals and prevent them from taking over the population . If such an effect occurred in a wild plant population, it is conceivable that this may give infected, susceptible plants a further advantage over resistant or uninfected plants in the competition for limited pollinator services. Indeed, there are examples in which earlier flowering increases pollination and enhances yield. However, the relationship between flowering time and pollination is complex and there may be environments in which it is more advantageous for plants to flower in a concerted fashion. However, in certain contexts earlier flowering may result in flowers being produced before pollinators are available . At this stage, it may be imprudent and premature to propose that increased pollinator attraction to infected, susceptible hosts represents some sort of specific viral strategy to inhibit selection for resistance, and there are difficulties in envisaging how this might initially arise. However, it seems plausible to suggest that in principle increased pollinator attraction to virus infected plants could favor the persistence of susceptible plants in the environment and this could be seen as payback or compensation to the host. It is worth noting that other forms of payback by viruses to their hosts have been observed in a number of systems. This has led to the suggestion that our general view of viruses has been overly colored by their pathogenic properties and that we should view them as symbionts in the classical sense . For plant viruses it has been shown that virus infection can enhance the endurance of susceptible host plants to drought or in one case to cold and that plants of several species were protected from herbivory by virus infection . It has been suggested that resistance to drought is a conditional phenotype that could act as a payback to the host. In the case of CMV-induced drought resistance in Arabidopsisand other plants and in the present study, in which CMV enhances a tomato plant’s attractiveness to bumblebees, we may be seeing examples of ‘extended phenotypes’. An extended phenotype emerges from the action of a parasite gene when it alters the phenotype of a host; potentially to the benefit of the parasite.