Data from: Sugar content of diet does not buffer against chronic oral imidacloprid exposure in the alfalfa leafcutting bee, Megachile rotundata (Hymenoptera: Megachilidae)|昆虫学数据集|农药暴露数据集

Experimental Design We acquired nest cells containing prepupae of alfalfa leafcutting bees, M. rotundata, originating in Canada (JWM Leafcutters, Inc., Nampa, ID) and stored them at 4°C until the experiment. We placed nest cells in an incubator at 30.3 ± 0.03°C and 57.2 ± 0.35% (mean ± SE) relative humidity to induce completion of development. Adult emergence began 21 d after the start of incubation and continued for several days. Adults were kept at the above temperature and humidity for the entire duration of the experiment. Upon emergence, each adult bee was isolated in its own container (Supp Fig. 1 [online only]) and randomly assigned to one of six treatments (Table 1). The day of emergence, assigned treatment, and sex were recorded upon isolation. Each container consisted of an inverted, clear 59-ml plastic soufflé cup with a 5.5-cm diameter circle of filter paper (Ahlstrom 610, 1.5-μm pore size, Thomas Scientific, Swedesboro, NJ) fitted inside as a substrate. Through a hole in the side of each cup, we inserted a 0.6-ml microcentrifuge tube with a 2-mm hole near the bottom. These tubes served as feeders through which we delivered each bee’s assigned diet. Each of the six treatments included 20 adult bees comprising both sexes (n = 120 bees). About 29% of bees were female, which reflects sex ratios in managed populations of M. rotundata that are biased towards males (Pitts-Singer and James 2005). In a fully crossed factorial experiment, six diets were prepared (Table 1). Sugar solutions were prepared as either 20 or 50% sucrose (m/m) solutions. These concentrations are well within the range of total sugar concentrations naturally found in floral nectars and reflect the lower and higher sugar diets, respectively, used in a similar experiment on honey bees (Tosi et al. 2017). Premeasured quantities of Marathon 1% Granular (Olympic Horticultural Products, Mainland, PA) were added to achieve one of three concentrations of imidacloprid in solution: 0 (control), 30 (low), or 300 ppb (high). This formulation is 1% imidacloprid and 99% inert ingredients (up to 9% of which is crystalline silica) by mass. We chose to test the same insecticide concentrations as a previous study (Abbott et al. 2008), though these authors exposed M. rotundata larvae to imidacloprid via pollen. Although imidacloprid concentrations in floral resources can vary widely due to numerous factors (formulation, time since application, etc.), a concentration of 30 ppb (μg kg−1 or ng g−1) is slightly above the ‘field-realistic’ average for nectars of imidacloprid-treated plants (Bonmatin et al. 2015, Wood and Goulson 2017) but commensurate with levels that elicit sublethal effects in other bee species (Tan et al. 2014, Gregorc et al. 2018). The 300 ppb treatment was intended as an upper bound for investigating rates of mortality and is closer to the maximum concentrations in nectar reported in reviews (Sánchez-Bayo and Goka 2014, Bonmatin et al. 2015). Diet stock solutions were stored in the dark at 5°C to minimize photodegradation of imidacloprid. We periodically removed 100-μl aliquots of each diet stock solution to confirm imidacloprid concentrations using a commercial ELISA kit (QuantiPlate, catalog #EP-006, EnviroLogix, Portland, ME). ELISA results were quantified by measuring absorbance at 450 nm wavelength using a Multiskan GO 1.01.10 (Thermo Fisher Scientific, Waltham, MA) and SkanIt Software 4.1 for Microplate Readers RE v. 4.1.0.43. Absorbance readings for our control (no imidacloprid) diets indicated a lack of matrix effects in this ELISA (Byrne et al. 2005). For our low and high diets, respectively, the ELISA returned concentrations of 27 ± 5 and 209 ± 34 ppb (mean ± SE of multiple aliquots from the same stock solution), and these numbers did not decline significantly over the duration of the experiment. Mean ELISA readings for our high pesticide diet were 91 ppb lower on average than we had intended, perhaps due to interference from the inert materials in the insecticide formulation, but were still over seven times higher than the low pesticide treatment. We added a drop of yellow food coloring (glycerin, polysorbate 80, and turmeric) to 10-ml diet aliquots to facilitate visually tracking consumption. Bees were supplied 100-μl doses of their respective diets daily for the first week after emergence, and three times a week for the remainder of the experiment (ad libitum) as consumption tended to decrease over time. Until the last individual’s death, bee survivorship and any abnormal behaviors were documented. Feeders obstructed with precipitated solid sucrose were cleared with a clean needle when necessary. Upon each bee’s death, we recorded their date of death and measured the distance between tegulae (wing bases) as a proxy for body size (Cane 1987). Statistical Analyses All statistical analyses were performed in R version 3.3.3 (R Core Team 2019). We ran survival analyses with the package survival (Therneau 2015) using a Cox proportional hazards model (functions ‘Surv’ and ‘coxph’) to determine the effect of the fixed effects (sucrose and imidacloprid concentrations and the interaction between these two factors, bee sex, and bee body size) on bee survivorship, calculated as the number of days between an individual’s emergence and death. We checked for multicollinearity among fixed effects using the function ‘vif’ in the car package (Fox and Weisberg 2011) and ensured all variance inflation factors were <2. We used the package survminer (Kassambara and Kosinski 2018) to generate a Forest plot (function ‘ggforest’) for this model. We examined residuals of this model using the function ‘cox.zph’ in survival and verified they met model assumptions (P > 0.05). We created survival curve plots using the functions ‘survfit’ (survival) and ‘ggsurvplot’ (survminer). To determine whether a bee exhibited a behavior (dependent variable) that was related to the same fixed effects, we performed a logistic regression using the function ‘glm’ with a binomial distribution and logit link function in the package lme4 (Bates et al. 2014). We conducted posthoc pairwise Tukey tests via the functions ‘emmeans’ and ‘pairs’ in the package emmeans (Lenth 2019). To determine whether the amount of time that passed until this behavior was exhibited, or how long it took for bees to die after exhibiting this behavior, was related to our fixed effects, we constructed two separate linear models (function ‘lm’). Type III sums of squares and F- and P-values for our logistic regression and linear models were obtained using the function ‘Anova’ in the car package (Fox and Weisberg 2011). We created bar plots using the function ‘ggplot’ in the package ggplot2 (Wickham 2016).