Phenotypic plasticity, or the expression of different phenotypes based on the environment, can provide an evolutionary necessity: phenotypic variation. Diverse phenotypes are the raw material for natural selection to act on and can enable populations to reach environmental optima. We examined plasticity in 21 accessions of the globally distributed plant Arabidopsis thaliana under three experimental treatments: fertilizer, competition, and herbivory. Using a full factorial greenhouse design, we raised 902 plants and measured 11 traits to quantify plasticity. Our trait measurement dataset is provided along with the phenotypic plasticity dataset. We utilized F-ratios as our measure of phenotypic plasticity as they provide a robust, unitless metric of the strength of phenotypic responses across treatment levels relative to within-treatment variability, making them suitable for cross-trait comparisons. We identified the following model as the best fit for our trait data: Trait ~ Fertilizer * Competition, which includes the additive effect of fertilizer, the additive effect of competition, and their interaction. According to our model testing the herbivory treatment did not influence expression of the phenotypic traits we measured and therefore was not included in the model to generate our phenotypic plasticity dataset. F-ratios were calculated separately for each population, for each of the nine traits, and for the two relevant treatment gradients using a type II ANOVA. We employed a type II ANOVA rather than a type III as our data was not full rank and we were interested in the impact of the main effects while including the interaction rather than interpreting the interaction itself. We, therefore, excluded F-ratios representing the interaction of fertilizer and competition treatments as this was not biologically meaningful.

Plant Material

Arabidopsis thaliana is a highly self-fertilizing annual plant that has been shown to be phenotypically plastic in germination timing, shade avoidance responses, shoot branching, and the total number of rosette leaves. We selected 21 accessions throughout A. thaliana’s global range to compare phenotypic plasticity in traits. These accessions were obtained as single seed descent lines from the Arabidopsis Biological Resource Center (ABRC) and thus represent distinct genotypes. 

Seed bulking

From October 2020 to March 2021, we bulked seeds from each accession in the University of New Mexico Biology greenhouses to obtain enough seeds for the plasticity experiment (below) and to reduce any possible maternal effects (Kawecki et al., 2012). Plants were grown in Sungro Horticulture Metro Mix (Agawam, MA, USA) 360, treated with Marathon ® 1G (Hummert™ International), a granular insecticide, and fertilized with Osmocote ® 14-14-14 per the planting guidelines used by the ARBC. All seeds were vernalized on the soil surface for one week at 3.8° C before being placed in the greenhouse. Plants were misted daily or were bottom watered (in the event daily misting could not occur) to be given continuous access to water until all plants had germinated. Following germination, trays were bottom watered three times a week. We rotated the trays 180° after each watering to reduce microclimate effects. Plants were grown under supplemental light (16 hours a day; BML Spydr 600 LED Grow Light Grow-Max Spectrum – 120v, Austin, TX, USA) while the temperature was set to 23° C. Plants were allowed to self-fertilize and bulk seed was collected using the Arasystem (Betatech BVBA, GER), a system of tubes and cups that facilitate seed collection while preventing mixing of seeds from different plants.

Experimental Design

To initiate our experiment, plants were grown under the same soil and insecticide conditions outlined above. All plants were planted in trays of 36, 4.93cm x 5.66cm x 5.66cm cell length x width x depth, pots from Greenhouse Megastore (Danville, Illinois, USA). Seeds were germinated following the same procedures as seed bulking. Following germination, trays were bottom watered two to three times a week and we alternated the trays from the north to south end of the bench. Plants were grown under the same lighting conditions as detailed above while the temperature was set to a maximum of 21° C and minimum temperature of 10° C. To elicit plastic responses, we implemented a full-factorial experimental design manipulating three factors: competition (+/-), fertilization (+/-), and herbivory (+/-), yielding eight treatment combinations. Six plants per treatment combination from each of the 21 accessions produced a sample size of 1,008 plants.

The competition treatment was initiated immediately after vernalization. In two opposite corners of every pot assigned to the (+) level of the competition treatment, two seeds of Lolium multiflorum (annual ryegrass) were planted as the competitors. Lolium multiflorum is native to the same region as A. thaliana and has been widely introduced as a cover crop. Seeds were sourced from Papaws Garden Supply, L.L.C. (Seymour, Indiana, USA). The fertilizer treatment was applied 12 days after vernalization when all plants had germinated and had four leaves. One tablespoon of Peters Professional 20-20-20 General Purpose Fertilizer was dissolved per gallon to produce the solution. A Corning® Stripettor™ Ultra Pipet Controller (Corning, Inc, Arizona, USA) was used to apply 25mL of fertilizer. Plants that were assigned to the (-) level of the fertilizer treatment received 25mL of water to control for the water used in fertilizer application. The herbivory treatment was applied when plants reached 5.5 cm in rosette diameter or if plants remained <5.5 cm, when there was the development of an inflorescence bud. Trichoplusia ni (cabbage looper), known to feed on Arabidopsis and complete its life cycle on this plant, was selected as the herbivore. Second instar caterpillars were purchased from Benzon Research, Incorporated (Dearborn, Michigan, USA). Caterpillars were fed Benzon’s proprietary artificial diet prior to use in the experiment. A single cabbage looper was applied to a single rosette leaf using insect clip cages made from plastic petri dishes that measured 35mm diameter x 12 mm in height (NEST Scientific, Woodbridge, New Jersey, USA). Caterpillars were allowed to feed until the leaf was completely eaten. Each plant assigned to the (+) herbivory treatment had one leaf removed via this method.

Trait Measurements

Eleven traits relevant to colonization and/or fitness were measured. We counted the number of rosette leaves at 1 month. We lifted all top leaves to count any underlying rosette leaves. Rosette diameters at one and two months were measured using a ruler from the widest leaf-to-leaf tip to the nearest .1 cm. We recorded days vegetative by taking the difference between the germination day and surveyed flowering date (censused daily excluding one day on the weekends). Flowering was defined as the corolla being fully visible and the days flowering was the total number of days from initiation of flowering to flower senescence of the last flower. We counted the number of inflorescence branches at harvest and only included inflorescence branches that grew from the base of the rosette. Axillary branching was not recorded. We measured the maximum plant height from the soil to the tallest end of the inflorescence when stretched upright to its fullest extent. Measurements were recorded to the nearest millimeter using a meterstick. Silique counts were made monthly, and the maximum number of siliques recorded was utilized for analyses. In addition, we harvested three siliques from each plant at first dehiscence. These siliques were dissected, and seed numbers were recorded per silique. The average seed number of the three siliques represents our average seed number per fruit, which was multiplied by the maximum number of siliques per plant then averaged across all plants in a population to produce the average seed output. Longevity was the total number of days from germination to plant senescence.

Computation of Phenotypic Plasticity Dataset (Plasticity_24_Results.csv)

We first excluded the maximum number of siliques and the average seed number per fruit as these were used to calculate the total seed number. We aimed to utilize F-ratios as our measure of phenotypic plasticity as they provide a robust, unitless metric of the strength of phenotypic responses across treatment levels relative to within-treatment variability, making them suitable for cross-trait comparisons. F-ratios greater than one indicate the variation in a trait’s response is greater between the environments relative to the variation within a single environment and thus indicates phenotypic plasticity. F-ratios that are less than or equal to one do not represent phenotypic plasticity to an environment. F-ratio values that equal one indicates there was no difference in the trait variation between environments relative to within an environment. Finally, F-ratios less than one indicate that the variation within an environment is greater than the variation between environments. Here, the environment does not explain the variation in trait responses. Unlike coefficients of variation (CV), which cannot be reliably compared across traits or environments even when mean-scaled, F-ratios are independent of mean trait values and standardize the magnitude of plastic responses. This approach allowed us to identify which treatments elicited plastic responses in our multi-factorial experimental design.

To select which environments would serve as predictors for the ANOVA model, we used Akaike information criterion (AICc) that corrects for small sample sizes from the package AICcmodavg  to compare individual models that included the effects of treatments and their interactions for each trait in R version 4.0.3. We identified the following model as the best fit for our trait data: Trait ~ Fertilizer * Competition, which includes the additive effect of fertilizer, the additive effect of competition, and their interaction. According to our model testing the herbivory treatment did not influence expression of the phenotypic traits we measured and therefore was not included in the model to generate our phenotypic plasticity dataset. F-ratios were calculated separately for each population, for each of the nine traits, and for the two relevant treatment gradients using a type II ANOVA. We employed a type II ANOVA rather than a type III as our data was not full rank and we were interested in the impact of the main effects while including the interaction rather than interpreting the interaction itself. We, therefore, excluded F-ratios representing the interaction of fertilizer and competition treatments as this was not biologically meaningful. One population (Yo-0) was excluded from the fertilizer dataset due to low to no seed production for most plants. These computations yielded a dataset of 378 F-ratios.