1. Phenotypic variability results from interactions between genotype and environment and is a major driver of ecological and evolutionary interactions. Measuring the relative contributions of genetic variation, the environment, and their interaction to phenotypic variation remains a fundamental goal of evolutionary ecology.

2. In this study, we assess the question: How do genetic variation and local environmental conditions interact to influence phenotype within a single population? We explored this question using seed from a single population of common milkweed, Asclepias syriaca, in northern Michigan. We first measured resistance and resistance traits of 14 maternal lines in two common garden experiments (field and greenhouse) to detect genetic variation within the population. We carried out a reciprocal transplant experiment with three of these maternal lines to assess effects of local environment on phenotype. Finally, we compared the phenotypic traits measured in our experiments with the phenotypic traits of the naturally-growing maternal genets to be able to compare relative effect of genetic and environmental variation on naturally-occurring phenotypic variation. We measured defoliation levels, arthropod abundances, foliar cardenolide concentrations, foliar latex exudation, foliar carbon and nitrogen concentrations, and plant growth.

3. We found a striking lack of correlation in trait expression of the maternal lines between the common gardens, or between the common gardens and the naturally-growing maternal genets, suggesting that environment plays a larger role in phenotypic trait variation of this population. We found evidence of significant genotype-by-environment interactions for all traits except foliar concentrations of nitrogen and cardenolide. Milkweed resistance to chewing herbivores was associated more strongly with the growing environment. We observed no variation in foliar cardenolide concentrations among maternal lines but did observe variation among maternal lines in foliar latex exudation.

4. Overall, our data reveal powerful genotype-by-environment interactions on the expression of most resistance traits in milkweed.

Overall Experimental Design

To explore how milkweed genotype and environment influence resistance phenotypes, we designed three experimental groups: a field common garden, a greenhouse common garden, and a reciprocal field transplant. The seeds for all experiments came from fourteen genets of A. syriaca, growing in a 5-acre old-field at the University of Michigan Biological Station (UMBS) in Pellston, MI (45.558605, -84.677488). Within the old-field, the growing environments of the genets vary in plant community cover type, proximity to structures and dirt roads, and proximity to the forest edge (M.D. Hunter, personal observations). The old-field is maintained by semi-annual mowing and hosts 32 genets that have been studied annually since 2007 (M.D. Hunter, personal observations). The 14 genets that we selected for our experiments were those that produced enough seed in 2018 for all three experiments in 2019. The distance between neighboring genets varies from approximately 5 meters – 20 meters. The seeds were at least half-siblings (multiple seed pods from unknown fathers for each genetic mother). Seeds and seedlings were classified by their maternal genotype and are referred to as “maternal lines” hereafter. 

The field common garden and greenhouse common garden were both randomized block designs, and the same experimental design was replicated in the field and greenhouse. The reciprocal field transplant consisted of three maternal lines, each grown “at home” and “away”. 

We measured defoliation, arthropod abundances, and plant size (height, leaf number, stem diameter) weekly from all common garden plants (12 weeks from June 3^rd – August 21^st 2019) and reciprocal transplant plants (10 weeks from June 19^th – August 21^st 2019) and monthly from the naturally-growing maternal genets. We collected samples for foliar chemistry from all plants once in mid-July, in the middle of the growing season (methods below). This set of traits represents known drivers of insect performance, but we acknowledge that many other plant traits we did not sample likely also contribute to the insect abundance recorded in this study.

Common Gardens

(i) Growing the Plants

Plants for the field and greenhouse common gardens were grown from seed for one month at the University of Michigan in Ann Arbor before transfer to UMBS. Seeds were cold stratified for 6 weeks, treated with household bleach (5%), germinated in petri dishes for one week, and then planted in Sungro Metro-mix^® 360 potting soil in Deepots^®. Seedlings were grown in a controlled growth room (14:10 L:D, mean temperature 78°F) for the month of April 2019. Seeds were planted in April to ensure that plants were large enough to withstand field conditions by June, when local ramets emerge at UMBS. We transported plants from Ann Arbor to UMBS on May 1^st, 2019 to complete an additional month of greenhouse growth while outside conditions were still too cold. Plants were then either maintained in the greenhouse (greenhouse common garden) or transferred outside to the field common garden on June 1^st, 2019. This timing matches the typical phenology of the local milkweed population at UMBS (M.D. Hunter, personal observations).

(ii) Experimental Setup

The randomized block design of 18 blocks, each containing one individual of 14 maternal lines resulted in 252 plants total per common garden. Each plant was grown in an 18 cm x 16 cm pot held on benches (greenhouse) or set into the ground such that the topsoil of the pot was level with the ground (field). Each block consisted of two rows of 7 plants. Within each block of the field common garden, plants were spaced 1 m apart and 1.5 m separated each block. A 12.68 m x 32.53 m fenced exclosure surrounded the field common garden to protect plants from deer and rabbit browsing. The greenhouse common garden plants were arranged on benches so that plants were not touching. Plants were watered ad libitum and fertilized using Osmocote controlled release fertilizer (14:14:14  N:P:K) (ICL Specialty Fertilizers, Dublin, OH) once in May and once in July. From each A. syriaca plant we measured weekly arthropod abundance and plant height, leaf number, defoliation, and base stem diameter (12 weeks total). In mid-July, we measured foliar latex exudation and collected tissue to measure foliar concentrations of cardenolides, carbon, and nitrogen. 

Reciprocal Transplant Experiment

We chose three of the 14 maternal genets to provide seed for reciprocal field transplants. We chose genets that spanned the known range of nutritional quality in A. syriaca at UMBS and also originated from spatially separated locations (at least 50 meters apart) within the UMBS population. Based on the past 10 years of sampling, genet 14 has relatively high foliar nitrogen concentrations (3.60% N, 42.96% C), genet 20 has relatively low foliar nitrogen concentrations (2.72% N, 43.00% C), and genet 44 has relatively high foliar carbon concentrations (2.99% N, 44.34% C) (M.D. Hunter, unpublished data). Seeds from each of the three maternal lines were planted in the soil and location of all three of the original maternal genets. We chose to use only 3 maternal lines and 5 replicates at each location to minimize disturbance on the milkweed population growing in the old-field. At each of the three maternal genet growing locations, 5 seedlings from each of the three maternal lines (15 seedlings total) were grown in soil from that maternal location (i.e. maternal location includes the maternal soil). Therefore, at each maternal location, we grew offspring plants from the “matching” maternal line (“at home” seedlings) and two “non-matching” maternal lines (“away” seedlings).

Because we wanted to use soil from each maternal genet location, the reciprocal transplant experiment started later than the common garden experiments. Seeds were planted in 18 cm x 16 cm pots on May 13^th, 2019 at UMBS in the soil of their reciprocal transplant destination. Seedlings were grown in the greenhouse until large enough to withstand outside conditions and were placed in the field on June 19^th. Replicate seeds per maternal line were established within the spatial boundaries of each of the maternal genets. Each transplant location (maternal genet location) hosted 5 replicate plants of each of 3 maternal lines, totaling 15 plants at each maternal genet location (45 plants total in the experiment) in a 3 x 5 plant grid with 0.5 m separating each plant. We randomized the order of the plants at each of the three locations. Plants were protected by a wire open-top cage to block deer and rabbit browsing but allow access by insects. We measured arthropod abundance and plant height, leaf number, and defoliation weekly (10 weeks from June 19^th – August 21^st 2019) for each A. syriaca plant; foliar chemistry samples were collected once on August 21^st. Stem diameter was not measured due to the small size of plants.

Maternal Genet Sampling

To measure trait variation in the naturally-growing milkweed population, we sampled 5 individual ramets (randomly selected in June) from each of the 14 maternal genets (70 ramets total) on three dates (mid-June, mid-July, and mid-August, 2019). We measured size (height, leaf number, stem diameter), defoliation, and arthropod abundance for each ramet. We collected foliar chemistry samples and measured latex exudation once in mid-July. 

Estimate of Defoliation

To assess the contributions of genetic variation and environment to resistance to herbivory, we estimated defoliation by chewing herbivores from each plant in the common gardens, the reciprocal transplant experiment, and the maternal genets. We visually categorized each leaf longer than 1 cm into one of the following defoliation levels: no defoliation, 0-5%, 5-30%, 30-50%, 50-70%, 70-90%, >90% defoliated. To estimate the overall percentage of defoliation per plant, we multiplied the number of leaves in each defoliation level by the median value of the level (2.5, 17, 40, 60, 80, 95), and summed the values. This sum was then divided by the total number of leaves on that plant. The final value represents the overall estimation of percent defoliation for that plant. This method has a long history in the literature, and correlates strongly with independent estimates of defoliator activity (Hunter 1987; Hunter et al. 1997; Meier & Hunter 2019).

Plant Chemical Analyses 

We performed chemical analyses (cardenolides, C:N) on foliar samples from half of the blocks (blocks 10 – 18) in the two common gardens. We analyzed a subset of the samples due to project time constraints. We analyzed foliar chemistry in July because insect diversity and density are highest during July and this month represents the time period during which milkweed chemistry is most likely responsive to plant-herbivore interactions (Agrawal 2004a) (Appendix A, Table A1).

We analyzed foliar cardenolide concentrations using established methods (Zehnder & Hunter 2007; Decker et al. 2018). We cut 6 leaf disks with a hole puncher from the fifth leaf pair of each plant and placed the disks in 1 mL of methanol. Samples were stored at -10 °C for later cardenolide analysis. We took 6 additional disks from the same leaves to estimate the dry mass of the cardenolide samples. To extract cardenolides, we finely ground the leaf disks in methanol, sonicated the mixture for 1 hour at 60 °C, and centrifuged for 6 minutes. We transferred the supernatant to new 1 mL Eppendorf tubes and evaporated the samples under vacuum at 45 °C until dry. We resuspended the sample in 300 mL of methanol and used reverse-phase ultra-performance liquid chromatography (UPLC) on a Waters Acquity UPLC with an Acquity BEH C18 column (1.7 μm, 2.1 x 50 mm, Waters Inc., Milford, MA, USA). We separated and quantified cardenolides with a 0.15 mg/mL digitoxin internal standard (Sigma Chemical Company, St. Louis, Missouri, USA). Each 2 μL injection sample was eluted for 9 minutes at a constant flow rate of 0.7 mL per minute under a mobile phase of 20% acetonitrile (ACN): 80% water for 3 minutes followed by a gradient increasing to 45% ACN: 55% water over the remainder of the run. Cardenolides were quantified using a diode array detector scanning between 200 and 300 nm and we identified cardenolides as peaks with symmetrical absorbance between 216-222 nm. To calculate cardenolide concentrations, we took the sums of all separated peak areas, corrected by the concentration of the internal digitoxin standard and estimated by the dry sample mass.

We measured milkweed latex exudation by collecting latex from the 6 holes cut for the cardenolide samples on pre-weighed paper disks (Vannette & Hunter 2011), ensuring no latex was lost and excluding the leaf midrib from the hole punches. Disks were dried in a drying oven at 45º C for 24 hrs and then weighed. We measured latex exudation in all 18 blocks in both common gardens.

To analyze foliar carbon and nitrogen concentrations, we collected 2-3 leaves from each plant. Leaves were dried in a drying oven at 45º C and finely ground. Leaf powder was dried again for 24 hrs before 2 µg of each sample was transferred to a tin capsule. Carbon and nitrogen concentrations were measured on a ThermoScientific EA 1112 elemental analyzer. We used 99.7% caffeine powder as an external standard.