Cody Prouty

We set up nine 10x10 ft screened canopy tents in a mowed field near the botany greenhouses at the University of Georgia. Each tent received 16 greenhouse grown milkweed plants of either the native (A. incarnata) or non-native (A. curassavica) variety. Of the experimental tents, four received A. incarnata and four received A. curassavica. Owing to space and plant limitations, the single control cage received A. curassavica only. Plants within each tent were given a unique number ID from 1-16 representing a specific position within the tent. This numbering system was consistent across all tents so that the same plant number always occupied the same position in each tent. At the center of each tent, we placed a wire stand to hold sponges with honey water for monarch feeding. We randomly assigned tent positions within the field, with all tents approximately 2m apart.

Monarchs were randomly assigned to a cage and each monarch was given a unique color ID to observe behavior. Colored tabs were attached prior to release into cages, and each cage differed in the variables in figure 1. Male and female monarchs were equal in each cage.

We conducted observations between 1100 and 1700 hrs. Each observation interval lasted for a maximum of 15 mins and ended early if no activity occurred for 5 minutes. We recorded weather condition data at the start of each interval, including temperature, cloud cover, precipitation, and shade created by the tent cover. During the behavioral observations, we recorded individual monarch-plant interactions, including the type of activity (landing, nectaring, or oviposition), the duration of interaction, and the contact height on the plant.

On days 2, 4, 7, and 9 of the experiment, we collected six leaves from four randomly selected plants in each cage. Leaves were chosen randomly from the top third of the plant and were placed into individual, pre-labeled bags. In the lab, we collected tape samples from four of the leaves and used the remaining leaves for the bioassay described below. Tape samples were collected by pressing clear packing tape firmly against the leaves and tracing the leaf outline so that we could obtain an estimate of leaf area. For each plant, we collected tape samples from the top side of two leaves and the underside of the other two. To estimate spore deposition on the leaves, tape samples were evaluated under a microscope and spores were counted across the entire leaf unless the observer reached 300 spores before scoring the entire leaf. Once 300 spores were counted a final estimate was obtained by extrapolating across the remaining leaf area.

For bioassay samples, we prepared petri dishes with damp filter paper and either a half or a whole experimental milkweed leaf so that each caterpillar received a roughly equivalent amount. The bioassay was conducted using late second instar caterpillars from a single lineage previously fed on A. incarnata. Caterpillars were given 24 hours to consume their leaf at which point we recorded the relative proportion of the leaf consumed and transferred each caterpillar into an individual pint container with clean greenhouse grown A. incarnata. Upon eclosion, monarchs with a pupal score of 2 or less were taped to confirm their infection status. We recorded the forewing length and mass of each monarch one day post eclosion unless the parasite resulted in extreme deformity, in which case they were immediately euthanized post eclosion.

We evaluated the spore load of infected monarchs from both the behavioral and bioassay components of the experiment that had either a positive pupal score or a tape score of 5. We first removed the wings and placed the monarch body in a vial with 5mL of water. Each vial was shaken for 5 minutes using a Vortex Genie 2 to dislodge the spores. After vortexing, the body was removed and the vial was vortexed for an additional 20 seconds to ensure spores were suspended in solution before pipetting the mixture into a hemocytometer slide. We counted spores in four cells within each well and obtained an average across 5 wells (only two wells were counted if each cell had more than 20 spores). This average spore count represented the number of spores in 1 ml of water. To obtain an estimate for the entire 5 ml, we multiplied this number by 50,000 and used the log transformed value in data analysis.

To test for significant effects of treatment variables (monarch density (6 or 12 monarchs), infection status (infected or uninfected), sex, and plant species (swamp or tropical)) on monarch visitation rate, we used generalized linear mixed models with a normal distribution and exponentiated linear predictors in the ‘lme4’ package (Bates et al. 2015). Monarch visitation rate was calculated by dividing the number of visits per monarch to leaves by the time interval (in hours) monarchs were observed. We used the model structure: [monarch visitation rate ~ sex × density × infection status + plant species + number of matings + genetic lineage (random effect) + cage number (random effect) + total time spent in cage (random effect)]. Although three-way interactions were tested for, only the three two-way pairs were reported to retain simplicity. Monarch contact rate with leaves was correlated with the total amount of time monarchs spent in cages (χ2 = 8, p = 0.004), and was included as a random effect in the full model.

To test for effects on the rate of visitation to plants, we again used generalized linear mixed models with a normal distribution and exponentiated linear predictors. The model structure used here was: [visitation rate to plants ~ density × infection treatment (IF:HM or HF:IM) + plant species + number of open flowers + number of leaves + plant location (random effect) + cage number (random effect)]. The plant location within cages was significant, so it remained in the final model.

To test the effects of treatment variables on the number of spores recovered in the bioassay, we first used generalized linear mixed models with binomial error structures to test effects on prevalence of spores in monarchs. We next tested for differences in the spore load of only infected individuals using a negative binomial distribution (theta = 0.6) and exponentiated linear predictors. For both response variables in the bioassay, the model structure was: [response variable ~ infection treatment (IF:HM or HF:IM) × monarch density × plant species + sex + day leaves were sampled + cage number (random effect)].

Finally, to test the effects of treatment variables on the number of spores on leaf samples, we first used binomial error structures to test effects on prevalence of spores on leaves. Next, we used generalized linear mixed models on the spore load for infected monarchs with negative binomial error structures (theta =1). Both analyses were performed with the model structure: [response variable ~ infection treatment (IF:HM or HF:IM) + day sampled + monarch density + plant species + number of open flowers + plant visitation rate + number of leaves on plant + plant number (random effect) + cage number (random effect)]. In the analyses for the bioassay and leaf taping, controls were removed and only treatments with infected individuals were included after an initial analysis with controls included. In all analyses, residual distributions were checked using the DHARMa package in R (Hartig, 2022).