Experimental Design: We performed chronic oral exposure toxicity studies on B. impatiens colonies using 7 x 2.5 x 2.5 m metal-framed mesh field cages (hereafter tents) at Waterman Agriculture and Natural Resources Laboratory Research Farms at The Ohio State University (Columbus, Ohio, 40.0136222, -83.0405457). We erected twelve tents on a mown grass plot with ground cloth to eliminate weed pressure and secured the tent fabric edges with cinderblocks and mulch to minimize bee escape. We checked tents daily for holes and repaired as necessary. Each tent was equipped with a quart sized modified honey bee gravity feeder containing either a control solution of 50% w/v sucrose: DI water or a treatment sucrose solution (see below), and a pollen feeder containing uncontaminated, commercially produced honey bee collected pollen ground via mortar and pestle (Figure 1). We conducted five tented foraging experiments during the summers of 2018 and 2019, each of which with 12 colonies per experiment, one per tent (N = 60 total colonies deployed). While experiments varied in length and metals tested (Table 1), the general procedure for each experiment is as follows: At the start of the experiment, we randomly assigned a naïve commercial bumble bee colony (Koppert Biological Systems, Howell, MI, USA) containing one queen and approximately 50 workers to each tent. We weighed each colony prior to deployment, placed it on a cinder block to reduce contact with ground moisture, and affixed a corrugated plastic roof to provide shade and rain protection. Bees were allowed to forage in their tent for a set amount of time, and all feeders were checked daily and refilled as necessary to ensure continuous food resources. Twenty-four hours prior to the end of each experiment, the entrance doors of the hives were positioned to allow returning bees to enter the hive, but no bees exit the hive. At the conclusion of the experiment, we collected the colonies, reweighed them, and froze them in a -20° freezer for five days. We then dissected the colonies to determine the number of individuals and the proportion dead in each life stage (egg, larva, pupa, adult, queen). Each life stage was sorted into living and dead based on visual confirmation of color and texture based on previous live colony dissections.
We tested four heavy metals commonly found in urban environments and correlated with lethal and sub-lethal effects on bees: As, Cd, Cr, and Pb (Jennings et al. 2002; Sharma et al. 2015; Sivakoff et al. 2020). As contaminated environments oftentimes have multiple heavy metals, we also evaluated a treatment of all heavy metals combined. Experiments were run for 15 days to allow larvae that arrived with the naïve colony to pupate and emerge and 30 days in order to encompass eggs laid by the queen through pupation and emergence, as it typically takes a bumble bee egg 25 to 37 days to emerge (Heinrich 2002). Since it was not feasible to run all the metal treatments in each round of the experiment, we included control colonies in each round to evaluate consistency and enable comparisons across all rounds. We ran one 15 and one 30-day length experiment for each treatment (Cd, Cr, As, Pb, All heavy metals combined), see table 1. Experiments 1 and 2 consisted of four Cd fed colonies, four Cr fed colonies, and four control colonies, for 15- and 30-day lengths, respectively. Experiments 3 and 4 consisted of four As fed colonies, four Pb fed colonies, and four control colonies for 15- and 30-day lengths, respectively. For Experiment 5 (all heavy metals combined), we deployed eight treatment and four control colonies on Day 1, collected four random treatment and two control colonies on Day 15, and collected the remaining four treatment and two control colonies on Day 30.
Concentrated test stock heavy metal solutions were prepared prior to experiments. Treatment test heavy metal concentrations were based on the highest metal concentrations found in bumblebee collected provisions from hives deployed in Cleveland, OH (Sivakoff et al. 2020). Test heavy metals included: arsenic (arsenic (III) oxide, As2O3; 0.894 ppm), cadmium (cadmium chloride, CdCl2; 0.276 ppm), chromium (chromium (VI) oxide, CrO3;0.245 ppm), and lead (lead nitrate, Pb(NO3)2 ; 0.265 ppm). We prepared treatment and control sucrose solution (hereafter, “nectar”) by creating a 50% w/v sucrose: DI water mixture and shaking until dissolved. Treatment nectar was created by adding stock heavy metal solution to 3.75 liters of prepared nectar and inverting 20 times, and control nectar was created identically to treatment but adding DI water in place of stock heavy metals. The nectar was placed in an inverted quart size jar with pinholes in the lid, placed on spacers, and secured to a cinderblock within a water moat to discourage ants.
Data Analysis
We analyzed data using R version 3.6.3 (R Core Team 2020). Experiment 4 (30 Day As, Pb, and control) was omitted from data analysis because of control colony death, likely a result of weather conditions throughout the experiment outside of bumble bee colony tolerance. In addition, one Cd colony and one control colony from Experiment 1 were omitted from data analysis due to missing data. To make comparisons among experiments, we first evaluated consistency among control colonies across the experiments and found no significant differences in caste abundance or mortality (15 day: F= 4.24, P = 0.07; 30 day: F= 0.01, P = 0.924). To analyze the effect of each heavy metal on the number of individuals in each life stage (eggs, larvae, pupae, adults), we used generalized linear models with the glm function using the MASS package (W.N.Venables et al. 2002) and modeled the counts using a negative binomial distribution. The model included treatment and experiment length as fixed effects. To evaluate whether exposure to heavy metals affected the likelihood that an individual was alive at the end of the experiment, we used generalized linear models with a binomial error distribution where the state of an individual at the conclusion of the experiment (alive or dead) was considered as a binary response variable. We used odds ratios with 95% confidence interval to calculate the expected difference in odds that a colony has dead brood given exposure, compared to the odds of dead brood occurring in the absence of exposure. For example, an odds ratio of four would indicate that the odds of a brood within a colony to be dead differ by a factor of four between treatment and control, whereas an odds ratio of one would indicate that the odds of a colony containing dead brood does not differ from the control treatment.