Plant and insect material
Solidago altissima is a North American goldenrod (Asteraceae) that reproduces both sexually by flowering and asexually through the production of rhizomes [29]. All rhizome material derived from a single S. altissima clone that was originally collected near Bellefonte, Pennsylvania (see Table S1 for coordinates) in March 2015 by E.C.Y. The specimen was collected in a public right of way, which did not require a collection permit or licence. The species was identified based on plant characters and the presence of E. solidagnis galls, which are found almost exclusively on S. altissima in this part of its range [29]. Based on the unlikeliness of an erroneous identification, we have not deposited a voucher specimen. Rhizomes of S. altissima usually grow less than 30 cm per year [38], so many adjacent ramets in the field are likely competing for resources within the same genet, a scenario we replicated by using a single genotype of goldenrod.
The tephritid fruit fly E. solidaginis is a specialist gall-making herbivore on S. altissima and S. gigantea [29]. We used crude extract of the fly emission to prime plants. To obtain the extract, we collected galls from several locations near State College, Pennsylvania (see Table S1 for coordinates). We placed galls at room temperature to allow flies to pupate and emerge as adults. We used headspace aeration to collect the volatile emission following established protocol [6]. Using dichloromethane as a solvent, we pooled samples to create a uniform mixture and measured the concentration of the emission components in the mixture using GC-FID and a nonyl acetate internal standard.
Fitness effects of priming the field
To examine the costs and benefits of priming under competition, we performed a two-by-two factorial treatment manipulating priming and herbivory. First, we planted 240 ramets in pairs matched for size (two plants per pot) and randomly assigned pairs to either the primed treatment, where one ramet was primed and the other unprimed, or the unprimed treatment, where both ramets were unprimed. Ramets were paired to compare plants with primed and unprimed neighbours and determine if priming influences competitive ability. If primed plants are consistently better or worse competitors compared to unprimed plants, then differences in fitness should be greater when an unprimed plant is paired with a primed neighbour (an unequal competitor) than unprimed neighbour (an equal competitor). To ascertain that paired ramets were indeed competing under our experimental conditions, we planted an additional 30 ramets singly to serve as competition-free controls (see Supplementary Methods for further details on planting). These single ramets were not primed, so although we tested whether priming influences competitive ability (by comparing fitness between primed and unprimed neighbours), we did not test whether competition influenced the effects of priming (i.e. we did not compare single primed plants to paired primed plants) or a statistical interaction between priming and competition. We randomly assigned half of all pots (both single and double ramets) to the insecticide treatment to remove herbivores, with the other half assigned to a water spray control.
To prime plants in the priming treatment, we randomly selected one ramet to be exposed to the fly emission, while the other ramet was exposed to a solvent-only control. The other half of double pots (the unprimed treatment) had both plants exposed to only the solvent. To expose ramets to the emission, we enveloped the tip of each ramet with a 1 L plastic bag, sealed tightly around the stem with plastic-coated wire pressed into a cushion of modelling clay. We cut a small hole in the top of the bag through which we dropped rubber septa that had been filled with either 35 μg of the crude emission extract for primed ramets or the equivalent volume of the solvent (dichloromethane) for all other ramets. To minimize the effects of exposure to dichloromethane, we allowed both the emission extract and the solvent control to evaporate for approximately 1 min before placing the septa in plastic bags. After adding the septa, these holes were then sealed with wire. After 6 h, we removed the septa, refilled them with either crude extract or solvent, as appropriate, and placed them back in the bags, so that all ramets received 2 doses of the treatment, and primed ramets received a total of 70 μg of crude extract, which is the average amount emitted by a single male fly over 24 h [7]. This priming procedure was repeated over three days from 13 to 15 June to match the priming regimen of previous experiments [7, 25]. Ramet heights did not differ among the treatments at the time of priming (all p > 0.30).
After priming (16 June 2016), we placed all 150 pots into a former agricultural field (see Additional file 1 Table S1), embedded within a naturally growing S. altissima patch. In central Pennsylvania, E. solidaginis adults normally emerge in mid to late May and persist for about two weeks [30]. We purposefully delayed our experiment until flies were no longer expected in the field to ensure that naturally occurring E. solidaginis males did not prime our ramets. To standardize the vegetation surrounding each pot, we mowed six lanes in the naturally occurring goldenrod field, each lane separated by 1 m. We placed 25 pots spaced 1 m apart into each lane. Each lane received an equal number of pots of each treatment (5 pots each of the 2 × 2 factorial treatment and 5 no competition pots), but placement within the lane was randomized. At this time, we also measured plant height and any damage incurred during transport. Following placement in the field, pots in the insecticide treatment were sprayed with the synthetic pyrethroid esfenvalerate (Asana XL), diluted to 0.0033% in water and applied until runoff. This insecticide has been used extensively on S. altissima [9, 31, 37, 39] and was believed to have no physiological effect on the plant [31]. In addition, it breaks down quickly and has little effect on soil nutrients or microbes [40]. The insecticide was applied again one week later to match the approximately two weeks of protection gained from exposure to the fly emission [24].
Damage, growth, and fitness measurements in the field
Following placement in the field, we recorded leaf damage and plant height once a week for 4 weeks. As noted above, E. solidaginis was no longer active at the start of the experiment. Although priming is presumably directed against the herbivore generating the priming cue (i.e. E. solidaginis), in this experiment flies were no longer present in the field; thus, to assess plant defence, we measured the number of leaves damaged by the general assemblage of herbivores, as recorded in previous studies [6, 24]. We categorized leaf damage as chewing, spotting or leaf mining. We also recorded the number of leaves per ramet, including only those leaves that were fully separated from the apical or lateral buds and excluding leaves senescing at the base of the stem. To measure ramet fitness, we returned to the field on 23 September, when all of our experimental plants were in flower, and recorded final height and clipped the flower heads at the highest point below all flower-bearing branches. We placed the flower heads in drying ovens for 7 days and then weighed them to the nearest mg. Flower head mass strongly correlates with flower number [24]. Starting 26 September, we removed ramets from their pots and separated the below-ground biomass. We measured the length and mass (to the nearest 0.01 g) of each ramet’s rhizomes.
We note that although we are interested in the costs of priming, our fitness measures reflect the consequences of priming, induced defences, and herbivory combined. However, by placing plants into the field in a randomized block design, priming treatment groups were equally exposed to naturally occurring herbivores. Differences between treatments can therefore be attributed to our manipulation of priming, specifically.
Effects of esfenvalerate on plant growth in the greenhouse
Despite previous research suggesting no physiological effect by esfenvalerate on plant physiology [31], the insecticide applied to S. altissima in the field affected growth and flower production and interacted with priming in unexpected ways (see Results). To further explore these effects, we performed a separate experiment in a pest-free greenhouse to test for physiological effects of esfenvalerate on plant growth. We used the same S. altissima clone as in the field experiment with same growing methods, except that all ramets were planted singly into 2 L pots (9 Aug. 2017) and maintained in a pest-free greenhouse for the duration of the experiment. Eighty-six ramets were randomly selected to be treated with esfenvalerate, and 85 ramets were controls. As in our field experiment, we measured plant height 3 d before the application of the insecticide, the day of application, and once a week thereafter for 4 weeks. The insecticide was applied a second time after one week, and greenhouse lights were turned off for 4 h after each application to prevent burning.
Statistical analyses
To measure the effects of priming and insecticide on plant damage, growth, and fitness in the field, we constructed a single mixed model for each response variable (proportion of damaged leaves, ramet growth, flower mass, rhizome mass). We included insecticide treatment, priming, whether the ramet was paired or single, and week as fixed effects. Because herbivore damage correlated with growth, we included leaf damage as a fixed effect predicting weekly growth. We also analysed our data without pots with single ramets to see if these data were driving our results. As the results obtained were very similar (see Additional file 1), we have included single pots to maximize our sample size. To account for repeated measures over weeks and for pairing ramets within pots, we included ramet nested within pot as a random effect. Models assumed a normal distribution, and data were log or root transformed as necessary to normalize residuals. For final height, flower mass, and rhizome mass, we did not include week as a fixed effect and only included pot as a random effect because these data were measured only once at the end of the season. Interactions among the fixed effects were included in the models if they explained a significant amount of the variance. To measure the relative competitiveness between paired ramets, we calculated the absolute value of the difference in values between ramets in each pot (single ramets were excluded). We similarly constructed single mixed models per response variable, but without pot as a random effect, as we had only one response value per pot. If priming improves or impairs competitive ability, the difference in growth or fitness between neighbouring ramets should be greater when one ramet is primed than when both are unprimed. Models were constructed using the “nlme” package in R.
Row (each of six rows in the field) and placement within each row were included in our models if they explained a significant amount of the variance. We also detected a position effect relative to insecticide spraying. We had assumed that 1 m separation would allow each pot to be sprayed with insecticide independently, but ramets near sprayed pots tended to respond similarly to sprayed ramets. For example, one week after placement in the field, there was a significant correlation between ramet growth and the number of adjacent pots treated with insecticide (Mixed model: t = 2.44, n = 135, p = 0.017). Not every response appeared to be equally influenced by neighbouring insecticide treatment, but enough of our data showed correlations (see results) that we accounted for the neighbour effects of insecticide using an insecticide gradient in our models: 0 indicated that a non-sprayed pot had no adjacent sprayed pots; 1 indicated it was adjacent to 1 sprayed pot; 2 indicated it was adjacent to 2 sprayed pots, and 3 indicated the pot was sprayed directly.
To examine the effect of esfenvalerate on ramet growth in an herbivore-free environment, we constructed a single mixed model with insecticide treatment, initial growth rate prior to pesticide application, and their interactions with week as fixed effects and plant ID as a mixed effect to account for repeated measures over time. We used ANCOVAs (with initial growth rate as a covariate) within each week to clarify which data were driving overall patterns in the mixed model.
Note, that we provide the most relevant statistics of our models in our Results, while the full statistics are provided in the Supplementary Results.