Effects of herbivory and pathogen infection on plant-pollinator interactions
Data files
Nov 30, 2023 version files 36.29 KB
Abstract
Plant enemies can indirectly affect pollinators by modifying plant traits, but simultaneous tests of herbivore and pathogen effects are lacking, and the role of floral volatiles has seldom been mechanistically assessed.
In this study, we tested for indirect effects of insect herbivores and pathogens on pollinator attraction via altered floral volatile emissions, and its consequences for plant fitness in Brassica rapa. Plants in the field were exposed to either no damage or damage by caterpillars (Mamestra brassicae), aphids (Brevicoryne brassicae), a leaf fungus (Sclerotinia sclerotiorum), or a bacterium (Xanthomonas campestris pv. campestris). We recorded pollinator visits, and measured floral traits (flower number, volatiles) and plant fitness-correlates. We additionally performed a greenhouse experiment with artificial emitters to test for effects of target floral volatiles on pollinator attraction.
In the field experiments, plants subjected to herbivory by the aphid B. brassicae (but not the other enemies) exhibited a marked reduction in the emission of two VOCs (nonanal and 2-butyl-1-octanol), experienced lower pollinator visits, and produced seeds of lower quality in terms of seed biomass and germination rate, while flower output itself was not affected. Artificial emitters with reduced amounts of these compounds were less attractive to pollinators under greenhouse conditions.
Synthesis: These results provide strong evidence for volatile-mediated indirect interactions between plant enemies and pollinators ultimately impacting plant fitness, and further point at enemy and compound specificity underlying such effects.
README: Herbivore and pathogen effects on pollination
https://doi.org/10.5061/dryad.rbnzs7hjh
In this study, we tested for indirect effects of insect herbivores and pathogens on pollinator attraction via altered floral volatile emissions, and its consequences for plant fitness in Brassica rapa. Plants in the field were exposed to either no damage or damage by caterpillars (Mamestra brassicae), aphids (Brevicoryne brassicae), a leaf fungus (Sclerotinia sclerotiorum), or a bacterium (Xanthomonas campestris pv. campestris). We recorded pollinator visits, and measured floral traits (flower number, volatiles) and plant fitness-correlates. We additionally performed a greenhouse experiment with artificial emitters to test for effects of target floral volatiles on pollinator attraction.
Description of the data and file structure
| variable | location | units |
|---|---|---|
| height | Table1/Fig1 | cm |
| fruit_set | Table1/Fig1 | proportion |
| seeds_number | Table1/Fig1 | mean number of seeds per silique |
| seed_weight | Table1/Fig1 | mean seed weight in g |
| seed_germination | Table1/Fig1 | proportion |
| height | Table2 and 3/Fig2 | cm |
| VOCs | Table2 and 3/Fig2 | nanograms per hour |
| height | Table 4/Fig 3 | cm |
| flowers | Table 4/Fig 3 | total number of flowers |
| pollinator_visits | Table 4/Fig 3 | total number of bumblebee visits |
| frui_set | Table 4/Fig 3 | proportion |
| seed_number | Table 4/Fig 3 | mean number of seeds per silique |
| seed_weight | Table 4/Fig 3 | mean seed weight in g |
| seed_germination | Table 4/Fig 3 | proportion |
| pollinator_attraction | Fig 4 | "0" for not landing vs. "1" for landing on an inflorescence |
| height | Fig 5 | cm |
| fruit_set | Fig 5 | proportion |
Note: N/A means no values
Code/Software
Table 1 and Figure 1. We analyzed the effects of leaf damage treatment, plant accession, and their interaction (all fixed factors) on flower number, fruit-set, mean number of seeds per silique, mean seed weight, and seed germination (i.e., proportion of germinated seeds) using linear mixed effect models with PROC MIXED in SAS 9.4. We included the cages as a random factor and plant height as a covariate to account for differences in plant size affecting reproductive output and success. We log-transformed proportions of germinated seeds to achieve normality of residuals.
Tables 2 and 3 and Figure 2. We analyzed the effects of leaf damage treatment and plant accession (all fixed factors) on the emission of individual floral VOCs and on total VOC emission using linear models with PROC GLM in SAS 9.4. We also ran a permutational multivariate analysis of variance (PERMANOVA) to test for an effect of leaf damage treatment on floral VOC composition (using abundances of each compound). This analysis was based on 10,000 permutations and was performed with the ‘vegan’ package in R ver. 4.0.2 software. To visualize these results, we conducted a Principal Coordinates Analysis based on Bray-Curtis pairwise dissimilarities, and graphed the centroids of each leaf damage treatment effect. We also identified floral VOCs that correlated strongly (R2 > 0.60) with the first two ordination axes (using ‘envfit’ in vegan), and displayed these relationships using biplot arrows with length scaled to R2 values
Table 4 and Figure 3. We analyzed the effects of leaf damage treatment on total number of pollinator visits, fruit-set, mean number of seeds per silique, mean seed weight, and seed germination (i.e., proportion of germinated seeds) using linear mixed effect models with PROC MIXED in SAS 9.4. For all these models, we included the effect of plant accession as a random factor and plant height as a covariate. Moreover, to test for leaf damage effects on pollinator visits we tested for the leaf damage treatment as well as plant accession as a random effect and also included the total number of flowering stalks (across the two surveys) as a covariate. We log-transformed number of pollinator visits and fruit-set to achieve normality of residuals.
Figure 4. For each dual-choice assay type, we analyzed the effect of VOC blend type (two levels: control vs. B. brassicae blend, fixed factor) on the odds of bumble bees’ first choice using a generalized linear mixed model with a binomial distribution and logit-link function (PROC GLIMMIX in SAS 9.4). Odds ratio values are the ratio between successful and unsuccessful events (i.e., bees landing vs. not landing on an inflorescence of a given type, respectively) for each treatment level, i.e., a likelihood of a bee being attracted to the control or B. brassicae blend treatments. In both cases, we included the effect of the choice testing device or set-up (i.e., cylinder or flight cage) as a random factor to control for non-independence of each pair of inflorescences per replicate.
Figure 5. We assessed pollen limitation (i.e., difference in fruit-set for hand-pollinated vs. open-pollinated inflorescences) and whether it varied in enclosures vs. open field conditions by testing for the effect of stalk treatment (open vs. hand-pollinated), environment type (enclosure vs. open field), and their interaction on fruit-set using a linear mixed model with PROC MIXED in SAS 9.4. We additionally ran the same model on total number of pollinator visits per plant to assess differences in pollinator activity between environments potentially related to effects on fruit-set. For both models, we included the total number of flowering stalks per plant as a covariate to account for an effect of overall flowering intensity in testing for the pollination treatment on the focal stalks. In addition, both models included the effects of plant accession and individual plant (to control for non-independence of paired stalks per plant) as random factors, as well as plant height as a covariate to account for residual variation in plant size affecting reproductive output and success. We log-transformed number of pollinator visits to achieve normality of residuals.