Model substances
We used the sulfoximine insecticide sulfoxaflor, the methoxy-acrylate fungicide Amistar (azoxystrobin 250 g/l, Suspension Concentrate, see supplementary methods, S1) and the glycine herbicide glyphosate (as active substance, RoundUp ProActive or RoundUp FL, see supplementary methods, S1) as model substances. Our choice was justified by their widespread use, regulatory status and systemic uptake in plants. Because of these characteristics, the likelihood of bees being exposed in the field was considered similarly plausible across model substances. Additionally, we are not aware of published evidence of the acute toxicity of these substances across castes and sexes of B. terrestris and O. bicornis.
Sulfoxaflor is a relatively novel insecticide55,56,57, developed to replace or complement the use of older chemical classes against which insect pest populations had developed resistance57. However, because of its risks to bees58, its uses have been recently restricted in the EU to indoor growing conditions. As a nicotinic acetylcholine receptor (nAChR) competitive modulator, sulfoxaflor targets the same neural receptor as the bee-harming neonicotinoid insecticides55,56,57. Despite evidence that sulfoxaflor may target the nAChR in a distinct way compared to recently banned neonicotinoids55,56,57, these substances were shown to be similarly lethal in acute exposure laboratory settings for individuals of Apis mellifera, B. terrestris and O. bicornis38. Additionally, sulfoxaflor was shown to reduce reproduction59,60,61 (but not learning62,63) in bumble bees under field-realistic laboratory settings. When applied pre-flowering in a semi-field study design, sulfoxaflor impacted colony growth, colony size and foraging in bumble bees64 but not honey bees65. Azoxystrobin is a broad-spectrum, systemic fungicide, which has been widely used in agriculture since its first marketing authorisation in 199666. Azoxystrobin shows low acute toxicity to honey bees67. Azoxystrobin residues were found in nectar and pollen from treated crops68,69 and subsequently in the bodies of wild bees70. In a semi-field experimental setting, foraging, but not colony growth, was significantly impaired in B. terrestris exposed to Amistar (azoxystrobin 250 g/L SC)64, while no lethal or sublethal effects could be observed in honey bees65 or in O. bicornis71. However, a recent study showed that, when formulated as Amistar this pesticide induced acute mortality in bumble bees at high doses, which was attributed to the dietary toxicity of the co-formulant C16-18 alcohol ethoxylates50.
Glyphosate is a broad-spectrum systemic herbicide and the most widely used pesticide in the world72. Products containing glyphosate may be applied to flowering weeds73 and contaminate their pollen and nectar54, thus driving bee contact and oral exposure. Glyphosate showed low lethal hazards in regulatory-ready laboratory74 and semi-field designs when dosed as pure active substance or as MON 52276 (SL formulation containing 360 g glyphosate/L)75. A recent study found ready-to-use consumer products containing glyphosate to be lethally hazardous to bumble bees73. However, this toxicity was attributed to co-formulants, rather than the active substance itself.
We characterised the acute oral and contact toxicity to B. terrestris and O. bicornis of sulfoxaflor, azoxystrobin and glyphosate as either pure active substances or formulation (see supplementary material S2 Table S1). Each test was repeated across castes and sexes of these two species. For bumble bees we used workers, males and gynes (i.e., unmated queens), hereby referred to as queens, whereas for O. bicornis we used males and females. Bumble bee experiments were designed following OECD protocols30,31, while O. bicornis was tested following published76 and ring-tested protocols32, as an OECD protocol for this latter species is not yet available.
We used a dose response design whenever the test item was found to drive significant mortality in the tested species. In all other cases, a limit test design using a single, high pesticide dose was used. Details on the methods and results of the limit tests are reported in the supplementary materials (S2 and S4).
Pesticide treatments
All dose response tests were performed with pure sulfoxaflor, while azoxystrobin was tested as a plant protection product (Amistar 250 g a.s./l, SC, Syngenta, UK) in all oral tests, as its solubility in water was insufficient (6.7 mg a.s./L, see EFSA, 2010) to achieve the desired concentrations. Amistar contains co-formulants with hazard classification (54 C16-18 alcohols, ethoxylated < 20% w/w; naphthalenesulfonic acid, dimethyl-, polymer with formaldehyde and methylnaphthalenesulfonic acid, sodium salt < 10% w/w and 1,2-benzisothiazol-3(2H)-one < 0.05% w/w) and other unknown compounds. Therefore, to reduce potential bias due to differences in composition, we used the same production batch of Amistar across all dose response experiments. Pesticide treatments were freshly prepared on each day of exposure by means of serial dilutions of concentrated stock solutions. At this stage, stock and treatment solutions were stored at − 20 °C for subsequent chemical analysis. A subset of these solutions was analysed for absolute quantification of active substance (Supplementary material S3). Each dose response included a minimum of 5 serial dilutions of the test item spaced by a factor ≤ 2, in addition to one untreated control, an acetone control (in cases where the test substance was dissolved in such solvent) and a positive control (dimethoate).
Further details on the test solutions and dosing regimen are given in the supplementary materials (S1 and S2).
B. terrestris: oral and contact toxicity tests
Oral exposure experiments were carried out in the UK using B. terrestris ssp. audax, while contact exposure experiments were conducted in Estonia using B. terrestris ssp. terrestris. Our design did not specifically investigate how responses to pesticide exposure may vary across these two subspecies.
Test organisms and conditions
For oral exposure, B. terrestris ssp. audax colonies were purchased from a local supplier (Agralan, UK) as queen-right standard hives (i.e., > 80 workers). Upon arrival, bees were screened for the most prevalent gut parasites (Apicystis bombi, Crithidia spp., and Nosema spp.) through microscopic examination of faecal samples (n = 5 per colony box) using a Nikon eclipse (50i) compound microscope at 400X magnification. No infections were detected. Males were either collected from the same commercial colonies described above or, when this was not possible, by direct purchase (Agralan, UK). All queens were collected from queen-right colonies provided pro bono by a commercial supplier (Koppert, Slovakia), which were screened using the same methods described above.
For the contact exposure experiments, B. terrestris ssp. terrestris were purchased as queen‐right standard hives (i.e., > 80 workers; A.M. Ozoli, Latvia), while males and queens of the same subspecies were obtained from queen-less boxes (A.M. Ozoli, Latvia). These bees were not microscopically examined for parasite infections. However, their health status was visually checked upon arrival.
For both exposure routes, bees were kept and tested in complete darkness in a rearing room at 26˚C and the humidity at ∼ 60%. Bee handling was undertaken under red light. Before and after exposure, orally exposed bees were fed ad libitum inverted syrup (45% w/w, Thornes, UK), while topically exposed bees were given ad libitum access to sucrose syrup (50% w/v). Prior to exposure, bees were given a provision of fresh-frozen, honey bee-collected pollen pellets (Agralan, UK and A. M. Ozoli, Latvia for oral and contact tests respectively). We could not analyse all pollen batches used across bumble bee acute oral exposure studies. However, screening of this pollen source found only low levels of miticides used in honeybee hives (results not shown). The pollen used in contact exposure experiments was obtained by a local supplier who distributes pollen from certified organic beekeeping. This pollen source was not analysed. However, considering it was sourced from organically managed apiaries, it is considered unlikely that this pollen source was contaminated with relevant concentrations of our model agrochemicals.
Experimental design
For both exposure routes and 1 day prior to their chemical exposure, bees of unknown age were weighed to the nearest milligram, before being individually housed in plastic cages (Nicot, Nicotplast, FR) for acclimatisation. At this stage, exceptionally large or small bees30,31 were visually excluded. On the following day, bees were allocated to treatments by colony of origin and body weight. When this was not possible (i.e., directly purchased males), the experimental design only controlled for body weight effects. Depending on the experiment, 3 to 9 colonies per test were used.
For oral exposure, following 4 h starvation, bees were given a 40 µL provision of pesticide-spiked or untreated sucrose syrup through a 2 ml syringe (Becton Dickinson, USA) with the tip removed. Four hours after exposure, syringes were visually inspected to ensure complete consumption of the treatment solution. At this stage, bees that did not consume the entire pesticide provision were removed. Across oral dose–response experiments, the average initial sample size per treatment group (i.e., test item and negative control) was 33 for workers, 34 for males and 43 for queens. Upon exclusion of unexposed bees, the average final sample size was 29 for workers (range: 20–35), 24 for males (range: 15–33) and 20 for queens (range: 7–36) (Supplementary material S2, Table S1). The lower limit of the range of sample sizes reported above corresponded to most concentrated treatment group in the Amistar experiments. For this compound we observed a dose-dependent feeding inhibition, which we attributed to the high dosing and viscosity of the treatment solution.
For contact exposure, following cold anesthetization, a droplet of treatment solution that was either spiked or not with the test solution, was applied to the dorsal side of the thorax (i.e., mesonotum) of each bee. The treatment volume was adjusted by bee size, with workers and males being exposed via a 2 µL droplet while queens were exposed to 4 µL. Upon topical exposure, bees were given ad libitum access to sucrose syrup through a 2 ml syringe with the tip removed (Terumo, Belgium).
Mortality was recorded at 24, 48, 72 and 96 h post-exposure. In dose response contact experiments, we tested 46 workers, 40 males and 30 queens per treatment group except for the untreated and solvent controls in the queen experiment, in which 15 individuals were tested per control group (Supplementary material S2, Table S1). Across experiments we tested 3685 bumblebees (workers: 1397; males: 1283 and queens: 1005; Supplementary material, S2).
Osmia bicornis: oral and contact toxicity tests
All O. bicornis experiments were carried out in Germany.
Diapausing O. bicornis males and females were purchased from a commercial rearing facility (Pollinature GhmB, Konstanz, DE). Upon arrival, cocoons were visually sorted by sex and stored in darkness at 4 °C in plastic bags.
To induce emergence from diapause, cocoons were placed in an incubator (Memmert, DE) at 21 ± 1 °C and 40% relative humidity. Cocoons were checked twice daily and, upon emergence, bees were transferred back to 4 °C to keep them dormant for maximum 4 days until test initiation.
Emerged males or females were allocated by day of hatching to rearing plastic boxes (27 * 14 * 16 cm) in groups of 10–20. Two feeders (Eppendorf 2 mL tubes with a 2 mm hole at the bottom) with a visual cue in the form of a petal (Brassica rapa or Diplotaxis tenuifolia) were provided in each cage. After ∼ 4 h of group housing in daylight, the now meconium-free bees were weighed to the nearest milligram and transferred to individual Nicot cages. Unusually small or large individuals were removed upon visual inspection and randomization was performed so that the treatments had and equal distribution of age classes (i.e., days after emergence) and body weights (mg). Bees were then left to starve at 21 °C overnight, after which each bee was presented with 20 µL of sucrose syrup containing pesticide or control treatment. The dose was presented in a cut-off tip of a 0.5 mL Eppendorf tube with a petal identical to the ones used in the hoarding cage in order to stimulate feeding behaviour. Using this method, we visually confirmed consumption, and found 74.5% of males and 88.3% of females (Supplementary material S2 Table S1; mean ratio across treatment groups) consumed the solution within 3 h. Bees that had not consumed the entire droplet within 3 h were considered non-feeders and were excluded. Consequently, across oral dose–response experiments, the average initial sample size per treatment group (i.e., test item and negative control) was 32 for females and 36 for males, while the final sample size was 27 for females [range 33–15], 245 for males [32–13]; Supplementary material S2, Table S1). Similar to bumble bees, the lower limit of the range of sample sizes reported above correspond to Amistar treatments. For this compound we observed feeding inhibition which we attributed to the high dosing and viscosity of the treatment solution.
For contact exposure, the above experimental procedure was repeated up to the allocation to treatments, after which bees were immediately cold-anaesthetized and topically exposed to the treatments. 1µL of solution was pipetted onto the dorsal part of the thorax between the wing-bases and bees were immediately returned to their individual Nicot holding cages. Across dose response contact experiments, a total of 30 individuals were included per treatment group dosed with the test item, and 20 individuals were included per control group (untreated control and solvent control; Supplementary material S2, Table S1). During the test phase, bees were kept in individual Nicot cages of the same design as described above. Bees were kept in an incubator at 21 °C, 16:8 h light:dark cycle for the remainder of the test. 50% w/v sugar solution was provided ad libitum in honey bee queen feeding cups (Nicot, Nicotplast, FR), covered by a metal mesh at the bottom of the cage. Mortality was recorded at 24, 48, 72 and 96 h post-exposure.
Across experiments we tested 1668 O. bicornis (females: 819 and males: 849; supplementary material, S2).
Data curation and statistical analysis
We defined the last observation timepoint at which increases in mortality were lower than 10% as a steady-state mortality level. We used this timepoint as baseline for hazard characterisation and comparative analysis across bee sexes, castes and species. This approach was preferred over the arbitrary selection of a fixed time point (e.g., 48 h) across experiment, as it enabled a more realistic characterisation of acute hazards. This choice was compliant with standard OECD methods30,31, which recommend extending the test duration beyond 48 h and up to 96 h whenever the mortality rate in the treated groups increases by ≥ 10% within a 24 h timeframe, whilst control mortality remains at acceptable (low) levels. The rationale behind the OECD recommendation is that the onset of lethal effects upon acute exposure may be delayed in time, in which case, selecting 48 h as timepoint for LD50 derivation may underestimate hazard.
However, when O. bicornis females were orally exposed to sulfoxaflor, the latest timepoint at which the control mortality was below 15% was 48 h. Therefore, for this latter test the LD50 was calculated at 48 h. For consistency, the same timepoint was selected for O. bicornis males exposed to sulfoxaflor38. Similarly, for Amistar, 48 h was selected as a valid timepoint for both sexes, as the control mortality of O. bicornis males exceeded 15% at 72 h.
To minimise statistical bias in comparative analyses, we used the same statistical model for dose response analysis across all experiments. After selection of mortality timepoints, dose responses were fitted using a log-normal model, based on which we estimated the median lethal doses (LD50s) and its asymptotic-based delta confidence interval. Analyses were first carried out by expressing doses as a function of the pesticide intake per bee (i.e., µg/bee) and then by normalising pesticide intake by fresh bee weight (i.e., ng/mg body weight). This enabled us to determine if, and to what extent, differences in bodyweight predicted the responses to pesticide exposure across bee castes, sexes and species. Doses were expressed as measured concentration whenever the chemical analysis of the test solutions deviated from the nominal concentration by more than 20%77 (Supplementary material S3). Mortality rates were corrected using Schneider-Orelli’s formula whenever control mortality at the steady-state timepoint exceeded 5%. Whenever acetone was used, we tested a solvent (acetone) control and a water control in parallel37, which were compared using a Fisher’s exact test. No difference in mortality between the two groups was observed.
Pesticide hazards across sexes and castes were compared with LD50 values (Table 1, Supplementary material S5 Fig. S3), which were used to calculate sensitivity ratios (SR) and relative confidence intervals using the comped function in drc33,34. Statistical analyses and data visualisation were performed in R78 using the packages dplyr79, drc71 and ggplot280.
For the limit tests, mortality rates were compared between controls and exposure groups using a Fisher’s exact test. Whenever appropriate, p-values were adjusted using the Benjamini–Hochberg correction.
Source: Ecology - nature.com
