Plant ontogeny determines strength and associated plant fitness consequences of plant‐mediated interactions between herbivores and flower visitors

Abstract Plants show ontogenetic variation in growth–defence strategies to maximize reproductive output within a community context. Most work on plant ontogenetic variation in growth–defence trade‐offs has focussed on interactions with antagonistic insect herbivores. Plants respond to herbivore attack with phenotypic changes. Despite the knowledge that plant responses to herbivory affect plant mutualistic interactions with pollinators required for reproduction, indirect interactions between herbivores and pollinators have not been included in the evaluation of how ontogenetic growth–defence trajectories affect plant fitness. In a common garden experiment with the annual Brassica nigra, we investigated whether exposure to various herbivore species on different plant ontogenetic stages (vegetative, bud or flowering stage) affects plant flowering traits, interactions with flower visitors and results in fitness consequences for the plant. Effects of herbivory on flowering plant traits and interactions with flower visitors depended on plant ontogeny. Plant exposure in the vegetative stage to the caterpillar Pieris brassicae and aphid Brevicoryne brassicae led to reduced flowering time and flower production, and resulted in reduced pollinator attraction, pollen beetle colonization, total seed production and seed weight. When plants had buds, infestation by most herbivore species tested reduced flower production and pollen beetle colonization. Pollinator attraction was either increased or reduced. Plants infested in the flowering stage with P. brassicae or Lipaphis erysimi flowered longer, while infestation by any of the herbivore species tested increased the number of flower visits by pollinators. Our results show that the outcome of herbivore–flower visitor interactions in B. nigra is specific for the combination of herbivore species and plant ontogenetic stage. Consequences of herbivory for flowering traits and reproductive output were strongest when plants were attacked early in life. Such differences in selection pressures imposed by herbivores to specific plant ontogenetic stages may drive the evolution of distinct ontogenetic trajectories in growth–defence–reproduction strategies and include indirect interactions between herbivores and flower visitors. Synthesis. Plant ontogeny can define the direct and indirect consequences of herbivory. Our study shows that the ontogenetic stage of plant individuals determined the effects of herbivory on plant flowering traits, interactions with flower visitors and plant fitness.

herbivores (Tukey's post hoc test, P < 0.001 and P < 0.001 respectively) or root herbivores (Tukey's post hoc tests, P < 0.001 and P < 0.001 respectively). Plants exposed in the vegetative stage to chewing herbivores terminated flowering later if compared with plants exposed to sapfeeding or root herbivores (Tukey's post hoc test, P < 0.001 and P = 0.007 respectively). We observed herbivore-species-specific effects of P. brassicae and B. brassicae for the duration of the bud stage, days till flowering, and termination of flowering (Fig. S1b,c,e). Effects of herbivory on flowering time depended on HFG when plants were infested in the bud stage, whereas herbivore-species-specific effects were observed for plants exposed to herbivory in the vegetative and flowering stage (Fig. S1d, Table S2). Plants exposed in the bud stage to root herbivores flowered longer than plants exposed to sap-feeding herbivores (Tukey's post hoc test, P = 0.002).
For example, pollinator community composition differed for plants exposed to D. radicum in the bud stage compared to plants exposed to these root-feeding herbivores in the vegetative stage (χ 2 test, P = 0.004) or flowering stage (χ 2 test, P = 0.006). We found herbivore-speciesspecific effects on pollinator community composition for plants exposed to herbivores in the bud stage (χ 2 test, χ 2 = 29.15, df = 18, P = 0.047), but pairwise testing only revealed marginally insignificant differences between herbivores (Fig. S2). All pollinator groups responded to the herbivore treatments and their responses were specific for timing of herbivore attack during plant ontogeny (Fig. S2, Table S5). For instance, solitary bees visited plants infested with the root herbivore D. radicum in the bud stage less frequently compared to the expected community, but bumblebees visited these plants more frequently. In contrast, solitary bees visited plants infested in the flowering stage with the root herbivore D. radicum more frequently compared to the expected community, but bumblebees visited these plants less frequently (Table S5).

Effects of herbivore infestation and plant ontogeny on floral mutualistscorrelations between numbers of pollinators and flowers
The total number of pollinators positively correlated with the number of inflorescences one and two weeks after the start of flowering ( Fig. S6 and S7, r = 0.25, t = 3.19, df = 157, P = 0.002 ; r = 0.49, t = 5.08, df = 82, P < 0.001 respectively). In both cases, plant exposure to herbivores affected this correlation ( Fig. S6 and S7). One week after plants had started flowering, we found a positive correlation between the number of inflorescences and the total number of pollinators for plants exposed in the flowering stage to herbivores (r = 0.49, t = 3.85, df = 46, P < 0.001), especially to L. erysimi aphids (r = 0.79, t = 3.19, df = 6, P = 0.019) and the root herbivore D. radicum (r = 0.85, t = 3.95, df = 6, P = 0.008). In contrast, we found a positive correlation for plants exposed to herbivores in the vegetative stage two weeks after the start of flowering stage (r = 0.79, t = 6.75, df = 28, P < 0.001). Due to low number of observations two weeks after the start of flowering, we could only analyse each herbivore species for all three plant ontogenetic stages combined. We found a positive correlation between the number of inflorescences and the total number of pollinators for uninfested plants (r = 0.76, t = 3.29, df = 8, P = 0.011), and plants exposed to the sawfly A. rosae (r = 0.66, t = 2.79, df = 10, P = 0.019) or the nematode H. schachtii (r = 0.74, t = 3.62, df = 11, P = 0.004).

Effects of herbivore infestation and plant ontogeny on floral mutualistsvisitation times and flower visits
Plant ontogeny determined the effects of plant exposure to herbivores on pollinator visits: the number of flowers visited per visit (for honeybees (HB) and syrphid flies (SF)), the time spent per flower (for HB and SF), visitation duration (for SF); and effects depended on both herbivore identity and HFG. (Fig. S11 -S16, Tables S3 and S4). One week after plants had started flowering, syrphid flies visited more flowers per visit on plants exposed to herbivores in the flowering stage compared to plants exposed in the bud stage (Tukey's post hoc test, P = 0.044), and this was especially true for plants exposed to larvae of the sawfly A. rosae (Tukey's post hoc tests, P = 0.044) or the nematode H. schachtii (Tukey's post hoc tests, P = 0.039). Syrphid flies spent more time per flower on plants exposed to herbivores in the bud stage compared to plants exposed in the vegetative stage (Tukey's post hoc tests, P = 0.002), whereas honeybees spent more time per flower on plants exposed in the vegetative stage compared to plants exposed in the flowering stage (Tukey's post hoc tests, P = 0.033). In the case of honeybees, this effect was also observed two weeks after plants had started flowering (Tukey's post hoc tests, P = 0.020), but then honeybees also visited fewer flowers of plants that had been exposed to herbivores in the vegetative stage compared to plants exposed in the flowering stage (Tukey's post hoc tests, P = 0.015). Effects of individual herbivore species on the visitation behaviour of honeybees and syrphid flies varied over plant ontogeny (Table S2). Specific effects of HFG were observed for plants exposed to herbivory in the vegetative stage ( We detected such changes at the first time point -one week after plants had started flowering (Table S4). Syrphid flies visited more flowers of plants exposed in the vegetative stage to chewing or root herbivores compared with plants exposed to sap-feeding herbivores (Tukey's post hoc tests, P < 0.001 and P < 0.001 respectively) or uninfested plants (Tukey's post hoc tests, P < 0.001 and P < 0.001 respectively). Syrphid flies spent, however, less time on plants infested by chewing herbivores than on uninfested plants if plants had been infested in the vegetative stage (Tukey's post hoc tests, P = 0.033). When plants were exposed in the bud stage to herbivory, syrphid flies visited more flowers of uninfested plants compared with plants exposed to sap-feeding herbivores (Tukey's post hoc test, P < 0.001), but less compared with plants exposed to chewing herbivores (Tukey's post hoc test, P = 0.002). Honeybees spent more time per flower of plants exposed in the bud stage to chewing herbivores than on plants exposed to sap-feeding herbivores (Tukey's post hoc test, P = 0.031). When flowering plants were exposed to herbivores, syrphid flies visited more flowers of plants exposed to chewing or root herbivores than on plants exposed to sap-feeding herbivores (Tukey's post hoc tests, P < 0.001 and P < 0.001 respectively) or uninfested plants (Tukey's post hoc tests, P < 0.001 and P < 0.001 respectively). Herbivore-species-specific effects were observed for plants exposed in the vegetative stage (visitation duration SF, number of flowers visited by HB and SF), bud stage honeybees, herbivore-species-specific effects on the number of flowers visited per visit and the time spent per flower were observed for plants exposed in the vegetative and flowering stage, and for number of flowers visited per visit also for plants exposed in the bud stage ( Fig. S12 and S13).

Effects of herbivore infestation and plant ontogeny on a floral antagonistcorrelations between numbers of pollen beetles and flowers
The number of pollen beetle adults per plant positively correlated with the number of inflorescences per plant for all three time points: one week after plants had started to produce buds (Fig. S17, τ = 0.46, z = 32.73, P < 0.001), and one and two weeks after the start of flowering ( Fig. S18 and S19, τ = 0.45, z = 33.87, P < 0.001; τ = 0.53, z = 37.46, P < 0.001 respectively). We found a positive correlation between the number of inflorescences and the number of pollen beetle adults for all treatments combined, for plants exposed in different ontogenetic stages to herbivores, and for each herbivore species and plant ontogenetic stage combination ( Fig. S17 -S19). Table S1. Output of generalized linear (mixed) models showing the effects of different fixed (herbivore species, plant ontogenetic stage, and herbivore functional group) factors on plant phenological traits. All random factors were initially included in the model, and factors which explained less than 3 percent variation or with a P -value above 0.05 were excluded from the model.  Letter groups (a -c, k -l, x -y) above bars indicate significant differences (P ≤ 0.05) between herbivore species within a plant ontogenetic stage based on Tukey's post hoc tests. Greek letters above lines indicate significant differences at (P ≤ 0.05) between plant ontogenetic stages based on Tukey's post hoc tests.           Observations lasted for 10 min and were made at two time points: between 7 -9 days and between 14 -16 days after plots had started flowering. For 7 -9 days after plots had started flowering, the number of replicates per herbivore treatment varied between 7 and 9, and 16 for the control treatment. For 14 -16 days after plots had started flowering, the number of replicates per herbivore treatment varied between 2 and 6, and was 10 for the control treatment. Letter groups (a -d, k -n, x -z) above bars indicate significant differences at (P ≤ 0.05) between herbivore species within a plant ontogenetic stage based on Tukey's post hoc tests, and small or capital letters were used for different time-points. Greek letters above lines indicate significant differences at (P ≤ 0.05) between plant ontogenetic stages based on Tukey's post hoc tests, whereas ns indicates no differences. Outliers are represented by circles (1.5 times the interquartile range below the 1 st or above the 3 rd quartile). Observations lasted for 10 min and were made at two time points: 7 -9 and 14 -16 days after plots had started flowering. For 7 -9 days after plots had started flowering, the number of replicates per herbivore treatment varied between 5 and 8, and 12 for the control treatment. For 14 -16 days after plots had started flowering, the number of replicates per herbivore treatment varied between 2 and 5, and was 10 for the control treatment. Letter groups (ab, k -l) above bars indicate significant differences at (P ≤ 0.05) between herbivore species within a plant ontogenetic stage based on Tukey's post hoc tests, and small or capital letters were used for different time-points. Greek letters above lines indicate significant differences at (P ≤ 0.05) between plant ontogenetic stages based on Tukey's post hoc tests, whereas ns indicates no differences. Outliers (1.5 times the interquartile range below the 1 st or above the 3 rd quartile) are represented by circles. Observations lasted for 10 min and were made at two time points: between 7 -9 days and between 14 -16 days after plots had started flowering. For 7 -9 days after plots had started flowering, the number of replicates per herbivore treatment varied between 0 and 6, and was 8 for the control treatment. For 14 -16 days after plots had started flowering, the number of replicates per herbivore treatment varied between 0 and 4, and was 5 for the control treatment.   Observations lasted for 10 min and were made at two time points: between 7 -9 days and between 14 -16 days after plots had started flowering. For 7 -9 days after plots had started flowering, the number of replicates per herbivore treatment varied between 7 and 9, and was 16 for the control treatment. For 14 -16 days after plots had started flowering, the number of replicates per herbivore treatment varied between 2 and 6, and was 10 for the control treatment. Letters group (k -l) above bars indicate significant differences at (P ≤ 0.05) between herbivore species within a plant ontogenetic stage based on Tukey's post hoc tests. Greek letters above lines indicate significant differences at (P ≤ 0.05) between plant ontogenetic stages based on Tukey's post hoc tests, and small or capital letters were used for different time-points. Observations lasted for 10 min and were made at two time points: between 7 -9 days and between 14 -16 days after plots had started flowering. For 7 -9 days after plots had started flowering, the number of replicates per herbivore treatment varied between 7 and 9, and was 16 for the control treatment. For 14 -16 days after plots had started flowering, the number of replicates per herbivore treatment varied between 2 and 6, and was 10 for the control treatment. Letters group (k -l) above bars indicate significant differences at (P ≤ 0.05) between herbivore species within a plant ontogenetic stage based on Tukey's post hoc tests. Greek letters above lines indicate significant differences at (P ≤ 0.05) between plant ontogenetic stages based on Tukey's post hoc tests and small or capital letters were used for different time-points. Observations lasted for 10 min and were made at two time points: between 7 -9 days and between 14 and 16 days after plots had started flowering. For 7 -9 days after plots had started flowering, the number of replicates per herbivore treatment varied between 2 and 6, and was 8 for the control treatment. For 14 -16 days after plots had started flowering, the number of replicates per herbivore treatment varied between 0 and 4, and was 6 for the control treatment. Letter groups (a -d, k -l, x -z) above bars indicate significant differences at (P ≤ 0.05) between herbivore species within a plant ontogenetic stage based on Tukey's post hoc tests. ns indicates no differences between plant ontogenetic stages based on Tukey's post hoc tests. Observations lasted for 10 min and were made at two time points: between 7 -9 days and between 14 -16 days after plots had started flowering. For 7 -9 days after plots had started flowering, the number of replicates per herbivore treatment varied between 7 and 9, and was 16 for the control treatment. For 14 -16 days after plots had started flowering, the number of replicates per herbivore treatment varied between 2 and 6, and was 10 for the control treatment. Observations lasted for 10 min and were made at two time points: between 7 -9 days and between 14 -16 days after plots had started flowering. For 7 -9 days after plots had started flowering, the number of replicates per herbivore treatment varied between 7 and 9, and was 16 for the control treatment. For 14 -16 days after plots had started flowering, the number of replicates per herbivore treatment varied between 2 and 6, and was 10 for the control treatment. Letter groups (a -b, k -l, x -z) above bars indicate significant differences at (P ≤ 0.05) between herbivore species within a plant ontogenetic stage based on Tukey's post hoc tests, and small or capital letters were used for different time-points. Greek letters above lines indicate significant differences at (P ≤ 0.05) between plant ontogenetic stages based on Tukey's post hoc tests, whereas ns indicates no differences. Boxplots show median (line), 1 st and 3 rd quartiles, minimum and maximum. Outliers (1.5 times the interquartile range below the 1 st or above the 3 rd quartile) are represented by circles. Observations lasted for 10 min and were made at two time points: between 7 -9 days and between 14 -16 days after plots had started flowering. For 7 -9 days after plots had started flowering, the number of replicates per herbivore treatment varied between 2 and 6, and was 8 for the control treatment. For 14 -16 days after plots had started flowering, the number of replicates per herbivore treatment varied between 0 and 4, and was 6 for the control treatment. Letter groups (a -c, x -y) above bars indicate significant differences at (P ≤ 0.05) between herbivore species within a plant ontogenetic stage based on Tukey's post hoc tests. ns indicates no differences between plant ontogenetic stages based on Tukey's post hoc tests. Observations lasted for 10 min and were made at two time points: between 7 -9 days and between 14 -16 days after plots had started flowering. For 7 -9 days after plots had started flowering, the number of replicates per herbivore treatment varied between 2 and 6, and was 8 for the control treatment. For 14 -16 days after plots had started flowering, the number of replicates per herbivore treatment varied between 0 and 4, and was 6 for the control treatment.
Letter groups (a -d, k -m, x -y) above bars indicate significant differences at (P ≤ 0.05) between herbivore species within a plant ontogenetic stage based on Tukey's post hoc tests. Greek letters above lines indicate significant differences at (P ≤ 0.05) between plant ontogenetic stages based on Tukey's post hoc tests, whereas ns indicates no differences. Observations lasted for 10 min and were made at two time points: between 7 -9 days and between 14 -16 days after plots had started flowering. For 7 -9 days after plots had started flowering, the number of replicates per herbivore treatment varied between 2 and 6, and was 8 for the control treatment. For 14 -16 days after plots had started flowering, the number of replicates per herbivore treatment varied between 0 and 4, and was 6 for the control treatment.
Letter group (x -y) above bars indicate significant differences at (P ≤ 0.05) between herbivore species within a plant ontogenetic stage based on Tukey's post hoc tests.
Greek letters above lines indicate significant differences at (P ≤ 0.05) between plant ontogenetic stages based on Tukey's post hoc tests, whereas ns indicates no differences.     Outliers (1.5 times the interquartile range below the 1 st or above the 3 rd quartile) are represented by circles. For central plants, the number of replicates per herbivore treatment varied between 14 and 17, and was 28 for uninfested plants. For side plants, the number of replicates per herbivore treatment varied between 29 and 35, and was 63 for uninfested plants. Letters above bars indicate significant differences at (P ≤ 0.05) between herbivore species within a plant ontogenetic stage based on Tukey's post hoc tests. Greek letters above lines indicate significant differences at (P ≤ 0.05) between plant ontogenetic stages based on Tukey's post hoc tests, whereas ns indicates no differences.