Chronic High Glyphosate Exposure Delays Individual Worker Bee (Apis mellifera L.) Development under Field Conditions

Simple Summary Glyphosate-based herbicides (GBH) can be found worldwide throughout conventional agroecosystems due to their unique and effective mode of action. Their use is generally not considered harmful to honey bees, and, consequently, foragers may encounter food sources that are potentially contaminated with GBH residues. However, recent studies found GBH to cause sublethal effects in bees, and therefore give rise to concern. While most related research has addressed such effects under laboratory conditions, field-realistic approaches under free-flying conditions are scarce. Here, we explore if GBH influences several important performance parameters at the colony level using standard and modified regulatory testing methods. Colony conditions (i.e., colony weight gain, individual worker bee survival, and overwintering) were not affected when subjected to chronic GBH exposure in a realistic range (high and low). In line with previous laboratory results, the high range of treatments revealed a delayed brood development of workers and reduced hatching weight of adults when compared with the control group. However, we concluded that more drastic effects on honey bee health did not seem to appear, as a broad range of performance parameters remained completely unaffected. In future research, the underlying mechanisms of the developmental delay that was confirmed here should be carefully investigated. Abstract The ongoing debate about glyphosate-based herbicides (GBH) and their implications for beneficial arthropods gives rise to controversy. This research was carried out to cover possible sublethal GBH effects on the brood and colony development, adult survival, and overwintering success of honey bees (Apis mellifera L.) under field conditions. Residues in bee relevant matrices, such as nectar, pollen, and plants, were additionally measured. To address these questions, we adopted four independent study approaches. For brood effects and survival, we orally exposed mini-hives housed in the “Kieler mating-nuc” system to sublethal concentrations of 4.8 mg glyphosate/kg (T1, low) and 137.6 mg glyphosate/kg (T2, high) over a period of one brood cycle (21 days). Brood development and colony conditions were assessed after a modified OECD method (No. 75). For adult survival, we weighed and labeled freshly emerged workers from control and exposed colonies and introduced them into non-contaminated mini-hives to monitor their life span for 25 consecutive days. The results from these experiments showed a trivial effect of GBH on colony conditions and the survival of individual workers, even though the hatching weight was reduced in T2. The brood termination rate (BTR) in the T2 treatment, however, was more than doubled (49.84%) when compared to the control (22.11%) or T1 (20.69%). This was surprising as T2 colonies gained similar weight and similar numbers of bees per colony compared to the control, indicating an equal performance. Obviously, the brood development in T2 was not “terminated” as expected by the OECD method terminology, but rather “slowed down” for an unknown period of time. In light of these findings, we suggest that chronic high GBH exposure is capable of significantly delaying worker brood development, while no further detrimental effects seem to appear at the colony level. Against this background, we discuss additional results and possible consequences of GBH for honey bee health.

Method S1: Brood development and photographic assessment adopted from OECD 2007 and based on Schur et al. 2003 Method S2: Analytical method and validation for glyphosate and AMPA Supplementary Table S1: Color code of labeled cell contents (BFD) Tab. S1 Brood area Fixing Day (BFD) assessments during the course of the chronic glyphosate exposure in experiement 1, modified after Schur et al. (2003). The numbers for the brood index were assigned to the cell content, respectively. If the expected brood stage was met at the specified BFD, the cell was labeled "1", if not as terminated (Brood termination = 0). To differentiate brood stages in the photographic assessment, a color code was used (see Fig. S1).  (Fig. S3, Method S1). Figure S2: Experimental setup scheme (experiment 1)

Fig. S2
Experimental setup illustrating the exposure and monitoring period from experiment 1.
Control and T1 were comprised of five mini-hives, T2 four. Colonies were exposed to glyphosate for 26 days. Subsequently, one ready-to-hatch brood frame per mini-hive was removed and placed together group-wise in an incubator for 24 h. After hatching, a total of C (n=152), T1 (n=149) and T2 (n=141) bees were marked and subdivided. Approximately 30-40 worker bees per treatment were introduced into four untreated mini-hives, each, and monitored for 25 d. Supplementary Figure S4: BTR, Brood Index and Compensation Index from BFD+13 and BFD+21

Fig. S4
Here, full details from the brood assessment in experiment 1 are presented modified after Schur et al. (2003) to complement Fig. 3 (see also Method S1). In groups C and T1, successful development was observed in the majority of the marked brood cells. Assuming that at the first assessment only eggs will be marked, the index is 1.0. An increase of the brood index (see paragraph 40) during the following assessment can be observed if a normal development of the brood is presumed. This increase is caused by the development from eggs to larval stages, to the pupae, and finally to the adult, emerged bee, and due to the rising numbers which are assigned to the brood stages (OECD, 2007; with ns = P > 0.05 and * = P < 0.05, t-test, pairwise).

Fig. S5
A Cox proportional hazards model was applied in experiment 1 to determine the hazard ratio (HR) displayed as forest plot. Significant differences within those groups (treatment) were revealed close to the statistical threshold (global P = 0.047, log-rank test). A pairwise comparison, however, did not confirm these differences between the respective groups. With an HR of 0.93 for T1 and 1.43 for T2, the treated bees were not at risk of dying sooner when compared to the control (T1-C: P = 0.73, T2-C: P = 0.051, Log-rank test) Supplementary Figure S6: Survival probability of untreated mini-hive replicates to justify pooling

Fig. S6
To justify pooling bees from the same groups but different mini-hives for survival analysis in experiment 1, these hives were evaluated separately treated as replicates. The test showed no significant differences (P > 0.05, Log-rank test).  Supplementary Method M S1: Brood development and photographic assessment adopted from OECD 2007 and based on Schur et al. 2003 Brood termination rate Based on the brood termination-rate the failure of individual eggs or larvae to develop is quantitatively assessed. For the calculation of the brood termination rate the observed cells are split into two categories:  The bee brood in the observed cell reached the expected brood stage at the different assessment days or was found empty or containing an egg after hatch of the adult bee on BFD +22 → successful development  The bee brood in the observed cell did not reach the expected brood stage at one of the assessment days or food was stored in the cell during BFD +5 to +16 → termination of the bee brood development For the final calculation the number of cells, where termination of the bee brood development was recorded, is summed up for each treatment and colony, is multiplied by 100 and divided by the number of cells observed to obtain the brood termination rate in %.

Brood index
The brood index is an indicator of the bee brood development and facilitates comparison between different treatments. The brood-index is calculated for each assessment day and colony. Therefore, the brood development in each cell will be checked starting from BFD 0 up to BFD +22. The cells are classified from 1 to 5 as described in paragraph 33 (Tab. S1, OECD 2007) if the cells contain the expected brood stage at the different assessment days. If a cell does not contain the expected brood stage or food is stored in the cell during BFD +5 to +16 (see Table 4, OECD 2007) the cell has to be counted 0 (see Table 5, OECD 2007) at that assessment day and also on the following days, irrespective whether the cell is filled again with brood. This might require a further transformation of a value as described in paragraph 33. For the final calculation the values of all individual cells in each treatment, assessed on the same day, are summed up and divided by the number of observed cells to obtain the average brood index.

Compensation index
The compensation index is an indicator of the recovery of the colony and will also be calculated for each assessment day and colony. The cells are classified from 1 to 5 as described in paragraph 33 (Tab. S1, OECD 2007), solely based on the identified growth stage on the assessment days. By that, the compensation of bee brood losses will be included in the calculation of the indices. For the final calculation the values of all individual cells in each treatment, assessed on the same day, are summed up and divided by the number of observed cells to obtain the average compensation index.

Preparation of feeding solution
A sample of 500 mg (approx. 400 µl) was weighed in a plastic tube (15 ml) and 9600 µl of the extracting agent (50 mM acetic acid/10 mM Na2EDTA) were added. The tubes were closed and shaken thoroughly. Depending on the active substance concentration in the feeding solutions, these solutions were measured undiluted (control samples) or diluted to different extents. The dilutions were made with the extracting agent while adding the internal standards.

Preparation of honey samples
A sample of approx. 1 g was weighed in a plastic tube (15 ml) and a surrogate standard solution (20 μl Glufosinate (conc.: 2.5 ng/µl, corresponding to 10 pg/µl in the measuring solution)) and 4.3 ml of the extracting agent (50 mM acetic acid/10 mM Na2EDTA) were added to the sample. The tubes were closed and after homogenization using a Vortex-mixer left to stand for 30 minutes. Afterwards the tubes were shaken for one minute by hand, further 10 minutes with a horizontal shaker, and then centrifuged for 5 minutes (1690 g). The entire supernatant was removed and filtered respectively cleaned using a Solid Phase Extraction (SPE) cartridge (OASIS HLB 6cc, 200 mg; Waters), to retain parts of the sugar in the sample. Before use, the SPE cartridge was conditioned with 2 ml methanol and 2 ml extracting agent and let run dry to not dilute the sample. 1000 µl of the sample extract were filled into a vial for measuring and 20 µl of an internal standard solution (glyphosate 13 C2 15 N, AMPA 13 C 15 N, conc.: 1 ng/µl each, corresponding to 20 pg/µl in the measuring solution) were added. The measurements were started immediately after the completion of the extracts.

Preparation of pollen samples
The samples were analyzed in the same way as described for honey with only one difference. 5.0 ml of the extracting agent (50 mM acetic acid/10 mM Na2EDTA) were added to the sample.

Preparation of plant samples
A sample of 10 g was weighed in a plastic cup (100 ml) and a surrogate standard solution (100 μL Glufosinate (conc.: 5 ng/µl, corresponding to 10 pg/µl in the measuring solution) and 50 ml of the extracting agent (50 mM acetic acid/10 mM Na2EDTA) were added. The cups were closed, left to stand for 30 minutes and then shaken for 20 minutes using a horizontal shaker.
Subsequently, the plant samples were crushed using a disperser and simultaneously extracted for 3 minutes. Then the samples were filtered using folded filters (particle retention 5 -8 µm) into a 50 ml tube, or alternatively, the plant extracts were centrifuged (10 min, 1.690 g). 5 ml of the filtered respectively centrifuged samples were transferred onto a conditioned (see the preparation of honey samples) SPE-cartridge and filtered again respectively further cleaned. 1000 µl of the extract were prepared for measurements as described above (preparation of honey samples).

Identification and quantification of the residues in the samples
LC-MS/MS was used for the identification and quantification of the target substances in the samples. Three multiple reaction monitoring (MRM) transitions were monitored for each analyte in order to confirm compound identity.
Reference standards in matrix and/or extracting agent were used for quantification, which was carried out according to the method of the internal standard. A large number of standards with the following concentrations were measured to create the calibration function: 0.1, 0.2, 0.5, 1, 5, 10, 25, 50, 100, 200, 500 and 1000 pg/µl). The results shown for the samples are averages of duplicate injections of sample extracts.
In undiluted and 1:10 diluted samples, the analyte contents were determined using matrixmatched standards. If samples had to be diluted 1:100 or 1:1000, the analytes were quantified using reference standards in extracting agent as matrix effects were sufficiently reduced by dilution.
The results for the surrogate standard glufosinate were used to control the analysis and were not included in the calculation of the analyte content in the samples.

LC-MS/MS
The system used was a Nexera X2 HPLC system (SHIMADZU Corp., Kyoto, Japan) coupled to a triple quadrupole mass spectrometer Q TRAP 6500+ (SCIEX, Framingham, MA, USA) equipped with an electrospray ionization (ESI) source.
The mass spectrometric parameters were as follows: The chromatographic separations were performed on an Acclaim Trinity Q1 column (3.0 x 100 mm; 3 µm, Thermo Fisher Scientific) with a pre-column SecurityGuard C 18 (3.0 x 4 mm, Phenomenex). The column oven temperature was set to 35 °C and the autosampler tray temperature was set to 15°C.
First, the samples were analyzed with the mobile phases (A) acetonitrile and (B) ultrapure water (0,055 µS/cm) with 50 mM ammonium formiate (adjusted to pH=2.9 with formic acid). The injection volume was 10 µl. The flow rates and gradient I and are shown in Tab. MS1.
Later the chromatographic conditions were optimized (Chamkasem & Vargo, 2017) and the samples analyzed with the mobile phases (A) ultrapure water and (B) ultrapure water with 50 mM ammonium formiate (adjusted to pH=2.9 with formic acid). A diverter valve between the LC column and the MS interface was used to direct the LC eluent to waste just before the AMPA peak (1.9 min) and after the glyphosate peak (3.5 min). The injection volume was 20 µl. The flow rates and gradient II are shown in Tab

Method validation -glyphosate
The validation study was performed to evaluate recoveries (REC), detection limits (LOD) and quantification limits (LOQ). Control samples of honey (used for honey stomach as well), pollen and plant material (phacelia) were fortified at different levels and 5 (4 respectively 7 in the case of plants) replicates extracted as described above. For the determination of recovery rates, detection and quantification limits, reference standards in solvent and matrix were prepared with the following concentration levels: 0.1, 0.25, 0.5, 1, 2.5, 5, 10, 25, 50, 100, 250 and 500 pg/µl.
The LOD was determined as the lowest concentration at which at least two MRM were detected, the peak signals of which were three times higher than the background noise of the chromatogram and the ratio of which was in the range of the required criteria (SANTE, 2019). The next highest concentration of the calibration standards above the detection limit was set as LOQ.
The results of the method validation procedure with honey, pollen and plant material are summarized in the following tables.