Development of a test design for a semi-field, colony feeding study for the common eastern bumble bee (Bombus impatiens [Hymenoptera:Apidae])

Ecological risk assessment is a key component of the regulatory process required for registration of crop protection products around the world. The western honey bee (Apis mellifera) is the model organism for pesticide risk assessments for bees, but there are uncertainties over whether it is predictive of risks to other bees. Consequently, efforts are underway to develop test methodologies for other non-Apis bees. We conducted a semi-field colony feeding study with Bombus impatiens colonies to develop a colony-level methodology for bumble bees. We exposed commercially available bumble bee colonies to diets consisting of four concentration treatments of dimethoate insecticide (0.05, 0.19, 0.75, and 3.0 mg a.i./L) via supplemental sugar solution for six weeks and compared exposed colonies to untreated controls. Each treatment group had 10 replicate colonies, with one replicate per treatment group represented at each of the 10 study rural locations. We collected data on various colony-level endpoints including production of female reproductive (gyne) offspring, colony weight, foraging activity, and consumption of provisioned sugar solution. Our results indicated that the test design could be used to derive concentration-response relationships for several endpoints including the most sensitive, colony mass (No Observed Adverse Effect Concentration = 0.05 mg a.i./L). Overall, our study provides the foundation for a semi-field, colony feeding study test design for bumble bees, thus adding to the growing body of studies that may be used to assess the protectiveness of the honey bee risk assessment framework for non-Apis bees exposed to pesticides.

Methods

Test Material

We used pure PestanalTM dimethoate (Sigma Aldrich, St. Louis, MO; Lot # BCBS9338, purity ≥98%) as the test material.

Bumble Bee Colonies

On 16 June 2020, we obtained 80 Excel-type B. impatiens colonies from Koppert Biological Systems Inc. (Michigan, USA). Excel-type commercial colonies as described by the vendor contained a minimum of 70 worker-caste females and a queen nesting on a plastic foundation within a cardboard hive box. The plastic foundation has a hive entry of limited size that acts as a queen excluder, allowing the traffic of worker and drones, but not queens. Each hive was fitted a plexiglass observation lid and without any cover to the nest by request to the supplier. Bees were fed ad libitum with a sugar syrup solution and honey bee-collected pollen for one week. Upon receipt, the weight of each colony was measured. Sixty queenright study colonies were selected for field deployment by removing the most extreme by weight colonies, i.e., the lightest and heaviest colonies. A quick visual assessment of the brood area was also conducted and those that had queen pupae were removed from consideration.

Experimental Design

The colonies were moved to the field on 26 June 2020. Colonies were randomly assigned to treatment group and site. We conducted the study at 10 sites established along the margins of hayfields (organic-certified or managed without pesticides) on working dairy farms in central Vermont. The sites were clustered at three farms (hereafter, “locations”) that were separated by > 500 m. In May 2020, an area that was approximately 3 × 65 meters in size was mowed at each site in preparation for the relocation of colonies from the test facility to the field. Each mowed area was surrounded by a solar-powered electric fence for protection from bears. We performed maintenance on fencing weekly and mowed fence lines as necessary during the season. At each site, six colonies were deployed in one row with a 10 m spacing. 

Each bumble bee colony was placed in an open-sided plastic box to prevent rain damage. Each hive was balanced on a stand in a second plastic box filled with 5 to 10 cm of water to prevent ants from entering the hive. Each hive was stabilized with a pair of grade stakes pounded into the ground.

Each of the six treatment groups, two controls (fed and unfed sucrose solution) and four dimethoate concentrations, were represented at each site by one experimental unit (colony). The “Fed” control colonies received 50% (w/v) supplemental sucrose solution whereas the “Unfed” control colonies did not receive supplemental sucrose solution. The four dimethoate treatment groups involved exposure to 0.05, 0.19, 0.75, or 3.0 mg dimethoate/L sucrose solution. In the dimethoate treatments, pure PestanalTM dimethoate powder was dissolved in measured volumes of 50% sucrose solution. One L of solution per week per colony was then provisioned to bees via supplemental nectar feeders. The feeding solutions were poured into the plastic bags, which acted as nectar feeders, with calibrated cylinders. The exposure period lasted six weeks starting on 2 July 2020 and ending 14 August 2020. This exposure period was selected to harmonize the test design to the honey bee colony feeding study design, and to represent a worst-case scenario exposure to an indeterminate blooming crop variety. Following the exposure period, colonies remained at the same locations for post-exposure assessments as described below.

During the field phase encompassing the exposure and post-exposure periods, colony condition assessments (hereafter “CCA”) were performed weekly starting at dusk after the forager bees had returned to the colony for the night. Headlamps provided enough light for the conduct of CCAs. Each hive was weighed (including the colony and the plastic hive insert box), and observations were made on queen survival and production of reproductive offspring (gyne pupae, new gynes, males). New gynes were differentiated from the foundress queen by the presence of intact wings, intact setae especially in the abdomen, and coloration (newly emerged queens have a grey complexion before sclerotization is complete). Presence of wax moths (adult and larva stages), ants, snails, beetles, wasps, mold, and bumble bees from a different species was recorded. At each CCA, the previous sugar solution provisions were removed and replaced with a fresh 1 L bag. We calculated the amount of solution consumed by each colony by subtracting the initial fresh mass of the bag from the mass of the bag remaining after one week in the field.

Foraging activity at each colony was determined on four dates during the exposure period. Hive entrances were observed for 10 minutes between the hours of 0900 to 1600, and the number of bees leaving and returning to the hive recorded. We also made notes on bee behavior and the presence of insects and other invertebrates in/on the colonies.

After the six-week exposure period, bees in the colonies continued to forage freely. Weekly CCAs continued until the experiment was terminated on 9 September 2020. At that time, approximately 75% of the surviving colonies had produced female reproductive offspring (hereafter “gynes,” those individuals who will found new colonies the following spring). Most other colonies had either died or were incapable of reproduction. After the field phase, we placed colonies in a freezer (≤-15 ºC) until dissection in the final CCA.

Statistical Analysis

Statistical analyses were conducted in the R statistical computing environment (Version 4.3.1) using tidyverse, DescTools, lme4, lmerTest, vegan, FedData, and other R libraries (Bates et al. 2015; Kuznetsova et al. 2017; Wickham et al. 2019, R Core Team 2021, Signorell 2023). To evaluate access of the colonies to floral resources across the study, we compared patterns of land use/land cover in a 1 km radius around sites. This radius is within the reported bumble bee foraging ranges of 250 m to 3000 m. The foraging range is affected by species differences in colony size and body size (Goulson 2010). Furthermore, workers from commercial B. impatiens colonies in lowbush blueberry fields foraged up to 400 m from the colony (Desjardins and De Oliveira, 2006). Using the R library, FedData (Bocinsky et al. 2023), the 2016 National Land Cover Dataset was downloaded for an area around the study site. The R library, ‘Raster’, was used to extract pixels within 1 km of each site. Proportional representations of 12 land cover classes at each site were then determined. We also explored land use patterns among sites using non-metric multi-dimensional scaling (NMDS) implemented with the R library, vegan.

Dimethoate treatment groups were compared to the Fed control for sucrose solution consumption, colony mass, numbers of gyne offspring produced, colony survival, and foraging activity. To determine the effect of supplementing hives with 50% sucrose solution, we compared responses between the two controls (Fed and Unfed). Foraging behavior was expressed as the number of bees returning to the hive per hour after normalizing from 10-minute counts. Data from one site were excluded because all six colonies were destroyed by bears between CCAs 2 and 3. In addition, dead colonies were excluded from CCA-specific analyses. For some response variables, such as colony mass, analyses were precluded at CCA 7 because the treatment colonies exposed to 0.75 and 3.0 mg/L dimethoate had poor survival. For final counts of gynes, we excluded colonies that were lost to bears, but included colonies that died before the end of the study. We excluded five hives from the analysis for gyne production because their plexiglass lids had failed, possibly affecting response variables. The Shapiro-Wilk Test was used to test for normality of residuals and Levene’s Test to test for homogeneity of variance among treatment groups. Dunnett’s multiple comparison procedure was used to determine the NOAEC and LOAEC for response variables that met the assumptions of parametric statistical tests. Otherwise, the non-parametric Kruskal-Wallis test was used.