Ecological traits interact with landscape context to determine bees’ pesticide risk

Widespread contamination of ecosystems with pesticides threatens non-target organisms. However, the extent to which life-history traits affect pesticide exposure and resulting risk in different landscape contexts remains poorly understood. We address this for bees across an agricultural land-use gradient based on pesticide assays of pollen and nectar collected by Apis mellifera, Bombus terrestris and Osmia bicornis, representing extensive, intermediate and limited foraging traits. We found that extensive foragers (A. mellifera) experienced the highest pesticide risk—additive toxicity-weighted concentrations. However, only intermediate (B. terrestris) and limited foragers (O. bicornis) responded to landscape context—experiencing lower pesticide risk with less agricultural land. Pesticide risk correlated among bee species and between food sources and was greatest in A. mellifera-collected pollen—useful information for future postapproval pesticide monitoring. We provide foraging trait- and landscape-dependent information on the occurrence, concentration and identity of pesticides that bees encounter to estimate pesticide risk, which is necessary for more realistic risk assessment and essential information for tracking policy goals to reduce pesticide risk.

resources and are thus more reliant on seminatural habitats to provide continuous forage. Therefore, limited foragers may be less exposed if seminatural habitats are available and provide non-contaminated forage (compare ref. 31 ). However, limited foragers may become disproportionately more exposed in intensively managed agricultural landscapes, where there is an increased likelihood of contamination in the few seminatural habitats ( Fig. 1b; line slope).
To test whether foraging traits alter exposure and risk for bees in different landscape contexts, we assayed pesticide residues in pollen and nectar collected by A. mellifera, B. terrestris and O. bicornis, representing extensive, intermediate and limited foragers, respectively, across three sequentially blooming crops ( Figs. 1 and 2). In doing so, we integrate multiple domains of pesticide exposure usually restricted to single studies: landscape context (for example, ref. 32 ), pollinator species (for example, ref. 33 ), crops (for example, ref. 15 ) and food sources (for example, ref. 34 ). We predicted that pesticide exposure and risk would increase with (1) the proportion of agricultural land and (2) the extent of foraging traits. Furthermore, we expected (3) limited foragers to experience greater pesticide exposure and risk than more extensive foragers with an increasing proportion of agricultural land. Additionally, we expected (4) that mass-flowering crops were the primary source of pesticide exposure, particularly for extensive foragers and that there may be crop-specific risks based on crop-specific pest management recommendations (Supplementary Table 1). Finally, we expected (5) pesticide exposure and risk to correlate between the pollen and nectar loads of bees, with potential application to postapproval pesticide monitoring. With expected drastic changes to pesticide regulation to meet current sustainability goals (for example, ref. 35 ) and calls for environmental risk assessment to become more accurate, reliable and holistic 36 , it is essential to understand why different cropping  , as demonstrated in low-intensity (c) and high-intensity (d) landscapes, whereby extensive (grey square) and limited (grey triangle) foragers move between habitat types within their respective foraging ranges (concentric circles relative to X, the central nests). Our baseline assumption (b, black circles) is that pesticide exposure and risk will increase with agricultural intensification, proportional to the area of agricultural land within the foraging range of bees (c and d, concentric circles). We expect bees with the largest foraging range, 'extensive' foragers, to receive the highest pesticide exposure and risk independent of landscape context (b, line intercept; c and d, grey squares). However, as agriculture intensifies, the proportion of agricultural land within the foraging range of bees increases and the likelihood of foraging on contaminated food increases. Therefore, we expect 'limited' foragers to be disproportionately more at risk from pesticide exposure as agricultural land expands (b, line slope; c and d, grey triangles). NA, not applicable.

Discussion
The pesticide exposure of bees arises from their activity intersecting pesticide use 12 . Thus, pesticide exposure and its correlated risk (additive toxicity-weighted concentrations) to bees are likely to be affected by their life-history traits 37 , particularly foraging habits 23,26,38 and land-use and pesticide-use patterns, especially in bee-attractive crops 39,40 . Using an ecological approach to pesticide risk, we found that extensive foragers (A. mellifera) experienced the greatest risk irrespective of the proportion of agricultural land in the landscape. Although risk correlated among bee species, both limited foragers (O. bicornis)  In addition, risk correlated between sample materials and was greatest in pollen. Consequently, A. mellifera-collected pollen can cautiously predict pesticide risk for bees, not accounting for residue intake, compared to nectar and pollen collected by other bee species, independent of landscape context. Thus, the A. mellifera-collected pollen-based pesticide risk indicator may be a promising metric for postapproval pesticide monitoring in terrestrial systems, generally proposed by ref. 41 and with parallels in aquatic systems 42 . Agricultural landscapes expose A. mellifera to multiple pesticides 15,27,32,[43][44][45] . However, we know less about the resulting pesticide risk, especially between bee species and in different landscape contexts (but see refs. 22,26,46,47 ). We found that increasing the proportion of agricultural land increased the risk for B. terrestris and O. bicornis but not for A. mellifera. We suggest that these landscape-dependent differences in risk result from species-specific activity patterns 23,38 . Different crop pollen use between the three species somewhat supports this: uniform collection by A. mellifera and B. terrestris compared to increasing collection by O. bicornis with an increasing proportion of agricultural land, consistent with findings in apple for A. mellifera 32 and O. cornifrons 28 . Consequently, mass-flowering crops appear to be a predominant food source for A. mellifera across agricultural landsca pes 32,38,40,45 . In contrast, despite access to mass-flowering crops, O. bicornis favours non-crop, predominantly woody, pollen resources when available 48,49 . These different preferences for crop pollen are evidenced by others finding that the collection of focal crop pollen positively correlates to the proportion of that crop in the landscape for A. mellifera (apple 32 ) and B. impatiens (blueberry 27 ) but not O. bicornis (oilseed rape 31,48 ). Therefore, sets of foraging traits (for example, large colony size and advanced communication) and foraging preferences probably drive the prevalence of A. mellifera in mass-flowering crops. In intensively managed agricultural landscapes with scarce seminatural habitats and high pesticide use, O. bicornis is increasingly likely to forage in less-preferred mass-flowering crops and seminatural habitats adjacent to arable land 31 and thus increase their pesticide exposure and risk. Consequently, populations of O. bicornis and similar, limited foragers may be disproportionately affected by agricultural intensification as their traits compound the combined effects of habitat loss and increased pesticide exposure 26 . Our use of O. bicornis as a sentinel allowed us to estimate exposure and risk of a limited forager in landscapes where they may not naturally occur, which, combined with the relatively generalised diet of Osmia spp. 26,48,50 , means that our estimates for limited foragers are probably precautionary among solitary bee species.
The focal crop (oilseed rape, apple or clover) was an important driver of pollen-derived exposure and risk for all bee species, independent of the proportion of agricultural land. For example, all bee species experienced the highest exposure and risk at apple sites, followed by oilseed rape and clover sites. These results mirror the approved number of active ingredients in plant protection products recommended for use in the three focal crops, with most in apple and least in clover. Apple and other fruit crops generally have higher pesticide use 51 and resulting bee exposure than annual arable crops or permanent grasslands 15 . We also found that the composition of pesticides in pollen differed between the three crops, identifying pest management strategies for specific crops and even specific compounds as determinants of landscape-level exposure and risk. Pollen pesticide risk was greater during crop bloom than after bloom across all three investigated crops. However, it did not correlate with either agricultural or focal crop pollen collection, possibly pointing toward the treated crop and associated flowering plants affected by drift as sources of pesticide exposure 20,22,26,27 . Focusing on spatiotemporally matched pollen and nectar samples from A. mellifera and B. terrestris, we found that exposure and risk were higher in pollen than in nectar, although this does not account for the uptake of residues by bees for example via consumption which is unequal between pollen and nectar 33 . Nonetheless, we found that risk but not exposure positively correlated between pollen and nectar; thus, pollen may be a precautionary material for estimating the pesticide risk of bees and, more generally, pesticide contamination of terrestrial environments 34,52 .
Pollen pesticide mixture composition differed the most between A. mellifera and O. bicornis, while B. terrestris overlapped the two. The three species shared two of the riskiest compounds, indoxacarb and acetamiprid, while the following most risky compounds were unique to each species: thiacloprid for A. mellifera, tebuconazole for B. terrestris and imidacloprid for O. bicornis. Nevertheless, risk positively correlated among the three species, suggesting that risk estimates for one species can, to some degree, inform the risk to other bee species. The generally low maximum cumulative ratio (MCR) values indicate that the pesticide mixture risk, independent of bee species and focal crop, was driven by one or a few high-risk compounds (similar to ref. 53 ). High-risk compounds were mainly neonicotinoid insecticides (acetamiprid, imidacloprid and thiacloprid), previously identified as high-risk to bees 33,54 but the riskiest compound was indoxacarb, an oxadiazine insecticide. Reduced exposure to these high-risk compounds would substantially decrease the risk for the three bee species. In the EU, pesticide restrictions (imidacloprid 2018, thiacloprid 2021 and indoxacarb 2022) are regulatory moves in this direction [55][56][57] , even if residues persist (like imidacloprid in our study 58 ) or new compounds with similar risk profiles enter the market in the future 59,60 .
Pesticide risk assessment primarily focuses on A. mellifera, partly because of its economic value, ease of management and a greater understanding of the species' biology [61][62][63] . However, risk assessment is becoming more holistic 36 , with a greater emphasis on non-Apis species 64 in recognition of wild bee diversity and their contribution to pollination services 65 . However, this change requires a better understanding of how pesticide risk varies among bee species and landscape contexts. We found that the pesticide risk estimated from A. mellifera-collected pollen was generally higher than or similar to B. terrestris and O. bicornis, particularly in landscapes with less agricultural land. Thus, whilst bee traits regulate pesticide exposure and risk, there is potential to extrapolate risk among bee species and exposure sources, with higher and thus precautionary risk estimates based on A. mellifera-collected pollen. However, pesticide exposure and our ecological indicator of pesticide risk do not account for species-specific processes past the pesticide use-bee activity intersection, such as consumption within the nest or indirect effects that could affect the fitness of the bees-important considerations when moving from exposure to effect in environmental risk assessment 63 . Using our trait-based approach, we conclude that landscape context modifies pesticide risk but only for limited and intermediate foragers (here, O. bicornis and B. terrestris, respectively). These findings highlight the potential for seminatural habitats to buffer pesticide-related risks for wild bees 26,46,66 . We also conclude that A. mellifera-collected pollen can predict environmental pesticide risk for other species and is precautionary, particularly in less agriculturally dominated landscapes. We, therefore, suggest that an A. mellifera-collected pollen-based pesticide risk indicator is a promising metric for postapproval pesticide monitoring in terrestrial systems (compare ref. 41 ). However, questions remain as to how this exposure affects individuals and, ultimately, populations of bees-tasks for a more holistic and realistic environmental risk assessment that aims to capture exposure to pesticide mixtures and risks within the diverse bee community 67 .

Field site system and sentinel bees
We centred 24 sites on three bee-attractive flowering crops: oilseed rape (8 sites), apple (8 sites) and red clover grown for seed production (8 sites) in southern Sweden (Fig. 2). These crops bloom sequentially: oilseed rape during April-May, apple during May-June and red clover during June-August (Fig. 2c) and are affected by different pests and therefore have different pest management strategies. The national pest management recommendations for 2019 included 26 active ingredients in oilseed rape, 32 in apple and 14 in clover seed and included acaricide (2 active ingredients), fungicide (20), herbicide (20) and insecticide (13) products (Supplementary Table 1). We selected sites on the basis of their surrounding proportion of agricultural land (2 km radius) to ensure an even gradient (for each crop type) of agricultural land and, therefore, anticipated pesticide use 15,16,68 . The average (± s.d.) proportion of agricultural land was 74 ± 24% (range 29-95%) for oilseed rape, 52 ± 29% (6-85%) for apple and 66 ± 20% (44-93%) for clover. All sites were >6 km apart, except for two clover sites, 2 km apart. Southern Sweden is characterised by annual crop production and nationally high pesticide use 69 . Farmers managed crops conventionally, except for one field of each focal crop, which was managed organically.
In 2019, we placed sentinel bees at focal crop fields at the onset of flowering and allowed them to forage freely without supplemental food. At each field, we placed: (1) two or three nationally produced, standardised and conventionally managed A. mellifera colonies, (2) six commercial B. terrestris colonies (Biobest Biological Systems) in two large ventilated wooden boxes and (3) three solitary bee trap nest units (at the oilseed rape and apple sites) each seeded with 50 male and 50 female O. bicornis cocoons (Wildbiene & Partner) (Supplementary Methods). We did not place O. bicornis in clover fields as their phenologies do not overlap (Fig. 2c).

Quantification of pesticide residues in pollen and nectar
We sampled pollen from (1) A. mellifera using pollen traps attached to two hives for 24 h, (2) B. terrestris by capturing foragers (~20 across all six colonies) and killing them on dry ice as they returned to their colonies and (3) multiple O. bicornis brood cell pollen provisions collected by females over the second half of the bloom period. We sampled pollen from A. mellifera and B. terrestris at two sampling intervals, coinciding with (1)  To compare residues between nectar and pollen, we sampled additional returning foragers of A. mellifera (n ≈ 100 individuals per sample) and B. terrestris (n ≈ 20 individuals per sample) 1-2, 4-6 and 12-16 days after a known pesticide application at four oilseed rape, two apple and seven clover sites (Supplementary Table 4). Corbicular pollen and nectar stomach content were collected from these foragers to produce paired pollen and nectar samples for each site and collection time point (n = 54).
We froze pollen and bee samples, before nectar extraction, at −20 °C before screening for 120 pesticide compounds included in the Swedish national monitoring scheme (Supplementary

Pollen identification
Part of each pollen sample was analysed to determine the pollen use of the three bee species at each site. First, we pooled pollen samples per site, bee species and bloom period in a 5 ml tube and agitated them in 5 ml of 70% ethanol before pipetting 2 µl of the pollen suspension onto a microscope slide stained and set using fuchsin gel under a coverslip. Next, we identified (using a pollen reference library at the Department of Biology (Lund) and ref. 70 ) and counted >400 pollen grains per slide (7-20 rows, 163 µm wide across the slide) using ×400 magnification. On the basis of this, we quantified the proportional use of all agricultural-type pollen and focal crop pollen by bees and categorised the latter into a Brassicacae group (including oilseed rape; Brassica napus), Malus group (including apple; Malus domestica) and Trifolium pratense group (including red clover; T. pratense) (Supplementary Table 6).

Landscape classification
We analysed the landscape surrounding our sites at multiple spatial scales (1,000, 1,500 and 2,000 m, corresponding to the average foraging capacities of bees (Fig. 1a)) on the basis of the IACS Spatial Data Layer provided by the Swedish Board of Agriculture. We classified land cover categories into two groups: agricultural land (all types of agricultural use, such as annual crops, orchards, leys and seminatural grasslands) and non-agricultural land (including forest, urban areas and water bodies). This distinction is because our focus was on the pesticide exposure and risk to bees from agricultural pesticide use and the pesticide exposure of bees is higher in rural compared to urban areas 22 . We also calculated the proportion of the focal crop in the radii and the average field size. We confirmed that the proportion of agricultural land was consistent (Supplementary Fig. 8) and correlated ( Supplementary Fig. 9) across the three spatial scales for each crop type and consequently used the landscape information at the largest scale (2,000 m) in all subsequent analyses.

Risk calculations
We use toxicity-weighted concentrations (TWC) as a basis for indicating pesticide risk for bees 26 , where the TWC of each compound (TWC i ) is the ratio between the concentration (c i ) of a detected compound in bee-collected pollen or nectar and its respective acute toxicity endpoint (LD 50i -the dose required to cause 50% mortality in the test population) 71 . Then, following a concentration addition approach-the recommended default for mixture environmental risk assessment 72 (even though some compound classes may synergize 73 ), we summed TWCs, to calculate the additive toxicity-weighted concentration of all compounds within a sample per site and bee species (TWC mix ): Henceforth, we refer to this metric, an indicator of pesticide-related risk, as 'risk'.
We averaged the acute oral and contact LD 50 (ref. 71 ) of each compound to provide an overall indicator of toxicity, reflective of how bees encounter pesticides in the landscape and their multiple exposure routes 37 . We used LD 50 for adult A. mellifera because there are incomplete toxicity data for other bee species and life stages and, where there are data, LD 50 for other bee species correlate with the corresponding A. mellifera LD 50 (refs. 53,74 ). Furthermore, in using the same LD 50 across bee species, we disentangle the ecology of bees from toxicology to explore relative differences in the activity patterns of bees in intersection with pesticide use. Finally, we used the tested dose for LD 50 based on limit tests 71 (used when a compound is expected to be low in toxicity or there are issues with solubility 75 ), which can overestimate the toxicity of a compound. Three of these compounds ranked highly for compound-specific risk due to their high concentrations and frequency of detection rather than toxicity (Table 1).
We also calculated the factor by which the mixture risk (TWC mix ) was greater than its composite most risky compound (max(TWC i )) using an MCR 76 . Thus, an MCR close to one indicates that a single compound dominates risk. The MCR did not vary among bee species or focal crops ( Supplementary Fig. 1).
Finally, we also calculated compound-specific risk (Table 1 and  Supplementary Table 2) to identify high-risk compounds by multiplying TWC i by its bee-specific detection frequency 33 .

Data analyses
We conducted four primary analyses to understand agricultural pesticide risk to bee species, followed by supporting multivariate analyses of the compound compositions. We performed analyses and data visualization using R v.4.1.1., constructed linear mixed-effects models (LMMs) with the lme4 package 77 and analysed compound composition with the vegan package 78 . For the primary analyses, risk data were log transformed and the proportion of crop pollen was logit transformed to meet assumptions of normality and homogeneity of variance. Upon detecting significant main effects, we examined the significance and difference of individual factor levels via pairwise comparisons of estimated marginal means using Tukey's method with the emmeans package 79 . Finally, we evaluated models for overdispersion and checked residuals for normality and homoscedasticity using diagnostic functions in the performance package 80 . We report marginal R 2 values calculated following the methods of ref. 81 .

Risk and pollen use with landscape context, focal crop and bee species.
We used LMMs to explore (1) risk from pollen and (2) use of agricultural pollen, with focal crop and bee species interacting with the proportion of agricultural land as fixed effects and site as a random intercept. We included an interaction between bee species and crop for both analyses but this was non-significant and thus removed. Additionally, we used a similar model, including the focal crop, bee species interaction and site as random intercept, to relate focal crop pollen to the proportion of that focal crop in the landscape.
Risk with sampling round and focal crop. We tested whether risk varied between the different sampling rounds using an LMM with sample round, focal crop and bee species included as fixed effects and site as a random intercept. Finally, we tested if risk related to the proportion of focal crop pollen, bee species and focal crop, with focal crop pollen interacting with bee species, as fixed effects and site as a random intercept.
Risk among bee species. We examined risk relationships among the site-specific pollen collection of bee species using three linear models, one for each species. We included the remaining bee species and focal crop as fixed effects; however, the focal crop was non-significant in all models (P > 0.05).
Risk between sample materials. We used data from the paired pollen-nectar collections to test for a difference in risk between sample materials (pollen versus nectar), using LMMs with sample material, focal crop and bee species as fixed effects, and sampling round nested in the site as a random intercept. In addition, we examined risk relationships among sample material collections, using an LMM with nectar risk specified as the response variable and pollen risk, focal crop and bee species as fixed effects, and sampling round nested in the site as a random intercept.
Differences in compound composition. We used PERMANOVA to compare the composition of compounds between focal crops and bee species using a Bray-Curtis dissimilarity index based on a Hellinger standardised community matrix of risk values using the adonis2() function in vegan. We used non-metric multidimensional scaling (NMDS) to visualise different clusters of compounds. We tested for differences in dispersion between focal crops or bee species using the betadisper() function in vegan. We detected no differences in the dispersion of compounds between crops. However, we found different dispersion of compounds between bee species (P = 0.03); therefore, we should interpret these community differences cautiously.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Corresponding author(s): Jessica Knapp
Last updated by author(s): Dec 17, 2022 Reporting Summary Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided

Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted

Software and code
Policy information about availability of computer code Data collection No software was used to collect data.

Data analysis
We used no custom computer code or algorithms for data analysis. All data were analysed in R version 4.1.1.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.

Data
Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A description of any restrictions on data availability -For clinical datasets or third party data, please ensure that the statement adheres to our policy The datasets generated and analysed during the current study are available from the corresponding author on reasonable request.

March 2021
Field-specific reporting Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.

Sample size
Describe how sample size was determined, detailing any statistical methods used to predetermine sample size OR if no sample-size calculation was performed, describe how sample sizes were chosen and provide a rationale for why these sample sizes are sufficient.

Blinding
Describe whether the investigators were blinded to group allocation during data collection and/or analysis. If blinding was not possible, describe why OR explain why blinding was not relevant to your study.

Behavioural & social sciences study design
All studies must disclose on these points even when the disclosure is negative.

Study description
Briefly describe the study type including whether data are quantitative, qualitative, or mixed-methods (e.g. qualitative cross-sectional, quantitative experimental, mixed-methods case study).

Research sample
State the research sample (e.g. Harvard university undergraduates, villagers in rural India) and provide relevant demographic information (e.g. age, sex) and indicate whether the sample is representative. Provide a rationale for the study sample chosen. For studies involving existing datasets, please describe the dataset and source.

Sampling strategy
Describe the sampling procedure (e.g. random, snowball, stratified, convenience). Describe the statistical methods that were used to predetermine sample size OR if no sample-size calculation was performed, describe how sample sizes were chosen and provide a rationale for why these sample sizes are sufficient. For qualitative data, please indicate whether data saturation was considered, and what criteria were used to decide that no further sampling was needed.

Data collection
Provide details about the data collection procedure, including the instruments or devices used to record the data (e.g. pen and paper, computer, eye tracker, video or audio equipment) whether anyone was present besides the participant(s) and the researcher, and whether the researcher was blind to experimental condition and/or the study hypothesis during data collection.

Timing
Indicate the start and stop dates of data collection. If there is a gap between collection periods, state the dates for each sample cohort.

Data exclusions
If no data were excluded from the analyses, state so OR if data were excluded, provide the exact number of exclusions and the rationale behind them, indicating whether exclusion criteria were pre-established.

Non-participation
State how many participants dropped out/declined participation and the reason(s) given OR provide response rate OR state that no participants dropped out/declined participation.

Randomization
If participants were not allocated into experimental groups, state so OR describe how participants were allocated to groups, and if allocation was not random, describe how covariates were controlled.

Ecological, evolutionary & environmental sciences study design
All studies must disclose on these points even when the disclosure is negative.

Study description
We selected 24 sites centred on three bee attractive flowering crops common in the region and flowering sequentially: oilseed rape (8 sites), apple (8 sites) and red clover (8 sites). We selected sites based on their surrounding proportion of agricultural land (2km radius) to ensure an even gradient (for each crop type) of agricultural land and, therefore, anticipated pesticide use.

March 2021
Within the replicated sites, we sampled pollen from (1) Apis mellifera using pollen traps attached to two hives for 24 hours, (2) Bombus terrestris by capturing foragers (~20 across all six colonies) and euthanising them on dry ice as they returned to their colonies, and (3) multiple Osmia bicornis brood cells collected by females over the second half of the bloom period. We sampled pollen from A. mellifera and B. terrestris at two sampling intervals, coinciding with (1) the peak of crop bloom and (2) after crop bloom and for O. bicornis only at the end of crop bloom (evenly from all the available pollen). In total, we collected 48 samples (595 g) of A. mellifera-, 44 samples (11 g) of B. terrestris-, and 16 samples (70 g) of O. bicornis-collected pollen. During and after bloom samples were pooled for both A. mellifera and B. terrestris, resulting in 24 samples of A. mellifera-, 22 samples of B. terrestris-(all colonies died at two sites), and 16 samples of O. bicornis-collected pollen. We did not pool O. bicornis pollen over the bloom period since this species already combined pollen provisions on our behalf. In addition, we collected additional samples of corbicular pollen and and foraging bees for nectar extraction to produce paired pollen and nectar samples 1-2, 4-6 and 12-16 days after known pesticide applications at a selection of focal sites (n = 54 site and time point combinations for each material).
We conducted three primary analyses to understand agricultural pesticide risk to bee species. 1) Linear mixed models (LMMs) explore risk from pollen and use of agricultural pollen, with focal crop and bee species interacting with the proportion of agricultural land as fixed effects and site as a random intercept. 2) three linear models explore risk relationships among bee species' site-specific pollen collection, one for each species with the remaining bee species and focal crop as fixed effects. 3) LMMs to explore risk between sample materials, with sample material, focal crop and bee species as fixed effects, and sampling round nested in the site as a random effect.

Research sample
We conducted our study in Scania, Southern Sweden, an intensively farmed European production region. The selected sites cover multiple mass-flowering crops and agricultural contexts representing temperate agricultural landscapes and different pesticide use patterns. The selected three bee species exemplify different life-history traits that determine their activity. Both these aspects are important since bees are predicted to encounter pesticides as their activity intersects pesticide use patterns.

Sampling strategy
No pre-study sample size calculation was performed. Instead, we relied on previous experience from the region on the needed replication for pesticide exposure and risk evaluations for bees. We have previously used a replication of 6-8 sites per factor level for bee pesticide exposure studies. In this study we used a replication of 8. We selected 24 sites in total centered on three bee attractive flowering crops common in the region and flowering sequentially: oilseed rape (8 sites), apple (8 sites) and red clover (8 sites). We selected sites based on their surrounding proportion of agricultural land (2km radius) to ensure a gradient (similar for each crop type) of agricultural land and, therefore, anticipated pesticide use, similarly to how we have selected sites in previous studies.
Within these sites, we collected a sufficient amount of pollen to represent pollen plant source and for pesticide residue quantification; we collected 48 samples covering the 24 sites and 2 time points (in total 1210 g) from A. mellifera and 44 samples covering 22 sites (all colonies were lost at two apple sites which excluded sampling) and 2 time points (in total 1309 pollen loads) from B. terrestris. Sampling was restricted beyond this to exclude negatively affecting colony functioning. Furthermore, we collected all pollen possible from O. bicornis, which were 16 samples (71 g).

Data collection
The authors and paid research assistants collected the data and samples of bees and bee-collected pollen from April to August 2019, using pollen traps and insect nets. Sample per bee species, type of material, site and sampling time point were separately stored in tubes on ice until return to the laboratory at the end of the day, after which samples were frozen (-20C). Pollen samples and bee bodies were sent to the analytical laboratory for nectar extraction and pesticide residue analysis. Half of the sample pollen was used to identify plant species origin. Data were noted in spreadsheets on sample origin and amount of material along with pesticide identity and concentration, as well as plant species origin for pollen samples. Land use data was based on the IACS Spatial Data Layer provided by the Swedish Board of Agriculture for the study year (2019) and extracted using R. Data on pesticide properties and toxicity information were extracted from the Pesticide Properties Database (PPDB) hosted by the University of Hertfordshire.
Timing and spatial scale We collected samples of pollen and bees from each focal crop field during and after bloom. Our crops bloomed sequentially: oilseed rape from April-May, apple from May-June and red clover from June-August, so that in addition to covering focal crop blooms we also covered most of the relevant season for the focal bee species. Samples at oilseed rape sites were collected 14-24 May (midbloom) and 22 May-7 June (end-post bloom), samples at apple sites were collected 20 May-2 June (mid-bloom) and 21 May-11 June (end-post bloom) and samples at clover sites were collected 25 June-18 July (mid-bloom) and 12-26 July (end bloom). In addition, we collected additional samples coinciding with a known pesticide application (before or during flowering) at a selection of focal sites during 14 May-25 July, all in 2019. Thus all samples were collected from May to July 2019. We conducted our study across Scania, Southern Sweden, an intensively farmed European production region covering about 100 x 100 km.

Data exclusions
No data were excluded.

Reproducibility
All data will be made freely available upon publication, and our manuscript describes our methodology in full. We have made no attempts to repeat this experiment.

Randomization
Sites centered on the three focal crops were selected to cover similar gradients of agricultural land and, therefore, anticipated pesticide use in the landscape. Proportion agricultural land was checked for consistency across bee relevant scales (1000, 1500, 2000 m radius). We matched honeybee hives and bumblebee colonies by their strength and randomly allocated them and Osmia cocoons to sites to ensure similar foraging efforts.

Blinding
Blinding during field sampling was not possible since focal bee species, focal crop and landscape context were clearly visible.
Did the study involve field work?
Yes No nature portfolio | reporting summary

March 2021
Ethics oversight was required and explain why not.
Note that full information on the approval of the study protocol must also be provided in the manuscript.

Animals and other organisms
Policy information about studies involving animals; ARRIVE guidelines recommended for reporting animal research

Laboratory animals
We collected samples from three managed bee species at focal sites: Apis mellifera, Bombus terrestris, and Osmia bicornis.

Wild animals
Our study did not involve wild animals.
Field-collected samples We stored pollen and bee samples at -20C before screening for pesticide compounds included in the Swedish national monitoring scheme, following established protocols at the analytical laboratory.

Ethics oversight
No ethical approval is required for insect collections in Sweden.
Note that full information on the approval of the study protocol must also be provided in the manuscript.

Human research participants
Policy information about studies involving human research participants

Recruitment
Describe how participants were recruited. Outline any potential self-selection bias or other biases that may be present and how these are likely to impact results.

Ethics oversight
Identify the organization(s) that approved the study protocol.
Note that full information on the approval of the study protocol must also be provided in the manuscript.

Clinical data Policy information about clinical studies
All manuscripts should comply with the ICMJE guidelines for publication of clinical research and a completed CONSORT checklist must be included with all submissions.
Clinical trial registration Provide the trial registration number from ClinicalTrials.gov or an equivalent agency.

Study protocol
Note where the full trial protocol can be accessed OR if not available, explain why.

Data collection
Describe the settings and locales of data collection, noting the time periods of recruitment and data collection.

Outcomes
Describe how you pre-defined primary and secondary outcome measures and how you assessed these measures.

Dual use research of concern
Policy information about dual use research of concern

Hazards
Could the accidental, deliberate or reckless misuse of agents or technologies generated in the work, or the application of information presented in the manuscript, pose a threat to: Experiments of concern Does the work involve any of these experiments of concern: No Yes Confirm that both raw and final processed data have been deposited in a public database such as GEO.
Confirm that you have deposited or provided access to graph files (e.g. BED files) for the called peaks.

Data access links
May remain private before publication.
For "Initial submission" or "Revised version" documents, provide reviewer access links. For your "Final submission" document, provide a link to the deposited data.

Files in database submission
Provide a list of all files available in the database submission.
Genome browser session (e.g. UCSC) Provide a link to an anonymized genome browser session for "Initial submission" and "Revised version" documents only, to enable peer review. Write "no longer applicable" for "Final submission" documents.

Methodology Replicates
Describe the experimental replicates, specifying number, type and replicate agreement.

Sequencing depth
Describe the sequencing depth for each experiment, providing the total number of reads, uniquely mapped reads, length of reads and whether they were paired-or single-end.

Antibodies
Describe the antibodies used for the ChIP-seq experiments; as applicable, provide supplier name, catalog number, clone name, and lot number.
Peak calling parameters Specify the command line program and parameters used for read mapping and peak calling, including the ChIP, control and index files used.

Data quality
Describe the methods used to ensure data quality in full detail, including how many peaks are at FDR 5% and above 5-fold enrichment.

Software
Describe the software used to collect and analyze the ChIP-seq data. For custom code that has been deposited into a community repository, provide accession details.

Flow Cytometry
Plots Confirm that: The axis labels state the marker and fluorochrome used (e.g. CD4-FITC).
The axis scales are clearly visible. Include numbers along axes only for bottom left plot of group (a 'group' is an analysis of identical markers).
All plots are contour plots with outliers or pseudocolor plots.
A numerical value for number of cells or percentage (with statistics) is provided.

Sample preparation
Describe the sample preparation, detailing the biological source of the cells and any tissue processing steps used.

Instrument
Identify the instrument used for data collection, specifying make and model number.