Study system

Our study took place on five pumpkin farms in central Ohio, USA, from 23 July – 16 August 2019 (S1 Table). Our sites were located within a predominantly agricultural area of temperate northcentral North America (mean maximum daily temperature during the study period =  29.4 ±  1.7 C, mean daily rainfall =  1.3 ±  3.8 mm, S2 Table). The farms were all conventional, commercial-scale working farms, with pesticide application decisions reflecting typical management regimes for our study region (ranging from prophylactic applications to integrated pest management). All growers used seeds treated with FarMore® FI400, which contains three fungicides (mefenoxam, fludioxonil, azoxystrobin) and the neonicotinoid insecticide thiamethoxam. Farmers grew multiple cultivars per field to appeal to their customers at their small local farm stands. We did not have information regarding the cultivar planted in each row and by peak flowering the pumpkin vines were often intertwined across rows. Foliar sprays were applied mid-season during flowering to manage insect pests like cucumber beetles (vectors of bacterial wilt), squash vine borers, aphids, and squash bugs [47] and fungal diseases like powdery and downy mildew, black rot, and Plectosporium blight [48] (S1 and S3 Tables). Growers indicated that they applied foliar sprays in the evening when pollinator activity was low and flower buds were unopened. Since previous studies have documented the translocation of systemic insecticides applied as a seed treatment to squash pollen and nectar [1,2,20], we chose to focus instead on the translocation of mid-season non-systemic foliar-applied pesticides to floral resources and bees. Application dates and pesticides used at each site are listed in S1 Table.

Pumpkin flowers in our system are visited by free-living bumble bees and squash bees [24]; nevertheless, four of the five farms also contracted or kept honey bees to supplement pollination. As in other cucurbit crops, pumpkin flowers are short-lived, with each flower opening around sunrise and lasting only several hours before wilting. The ephemeral nature of these flowers limits the routes through which non-systemic foliar-applied chemicals may enter pollen and nectar; the flowers we surveyed for residue analysis (see below) opened for the first time on the day we collected the pollen and nectar.

We received verbal permission from the land owners (who were also the growers) to access the field sites; no permits were required.

Data collection

To determine the level of hazard to bees foraging in pumpkin fields after mid-season evening application of foliar sprays, we analyzed residues of six agricultural chemicals (three insecticides and three fungicides) in pumpkin leaves, pollen and nectar from flowers, and bee visitors to flowers. Growers decided independently what chemicals to apply and when. At each farm, we opportunistically assayed pumpkin tissues one day before, as well as one day, three days, and seven days after foliar spray events that involved one of our target agrochemicals (but not the day of the spray event), based on planned application schedules communicated by our collaborating growers. During each survey, we collected 20 pumpkin leaves as well as pollen and nectar from about 40 male flowers, from individual plants haphazardly selected across the entire field. We also collected up to 10 individuals of each of the following types of bees foraging on pumpkin flowers one day before and one day after spray events: bumble bee workers, honey bee workers, and male and female X. pruinosa squash bees. We prioritized bee collections on the day before and the day after application, when exposure to the pesticides was most likely. We decided not to collect additional bees on Day 3 and 7 after spray. Sampling at fewer time points allowed us to screen more samples on the day when contamination was most likely and to evaluate risk to three bee taxa, instead of honey bees only.

We collected the pumpkin leaves into resealable plastic bags. Bees were collected from flowers into individual microcentrifuge vials. We collected pollen and nectar by harvesting entire male flowers: anthers were collected into a 50-ml centrifuge tube, and nectar was extracted from the unopened nectar well (i.e., not depleted or contaminated by foraging bees) using a clean 50-μl micropipette tip glued onto the needle adapter of a 5-ml syringe. All samples were kept on ice and transferred to a -80°C freezer for storage.

Pesticide residue analysis was conducted on 1 g samples of leaf, pollen (half pollen and half anther tissue, hereafter referred to as ‘pollen’), nectar, and bee tissue. We cut thin latitudinal strips from stacked pumpkin leaves at the widest part of the leaves near the center. These strips were chopped into small pieces using a sterile razor blade and from these pieces, a 1-g sample of tissue was weighed and placed in a 1.5-mL microcentrifuge tube. Pollen samples were scraped from the anthers with a sterile spatula to obtain a 0.5-g sample to which we added 0.5 g of tissue removed from the outer layer of 2 - 3 anthers due to the difficulty of obtaining 1 g of pure pollen. A 1-mL sample of nectar was measured into a clean 1.5-mL microcentrifuge tube. For the bee samples, it was necessary to combine 1 - 5 bees of the same species and sex collected from the same farm on the same day in order to make a 1-g sample (mean =  2.6 bees, SD =  1.4).

The tissue samples were kept at -80°C until shipped overnight on ice to a commercial agricultural chemical testing laboratory – SGS North America (Brookings, SD, USA). There, tissues were homogenized and residues extracted using the QuEChERS method (EN15662) [50], which has been commonly used to determine concentration of pesticide residues in leaf, pollen, and nectar tissue and has suitable recovery rates [15,51]. Chemical residues were analyzed using protocols outlined in the Association of Official Agricultural Chemists method [52]. Carbaryl and triflumizole residues were analyzed using liquid chromatography and tandem mass spectrometry (LC-MS/MS). Chlorothalonil, cyhalothrin (total), permethrin (cis and trans), and 5,7-dichloro-4-(p-fluorophenoxy) quinoline (i.e., quinoxyfen) were analyzed using gas chromatography mass spectrometry (GC-MS/MS) and concentrated from 2 mL to 1 mL. The concentrations of each residue were determined by comparing to standard curves including triphenyl phosphate (TPP) as the internal standard. Chemical concentrations were recorded in μg/ kg (parts per billion, hereafter ‘PPB’). The lower limit of quantification for most chemicals was 10 μg/ kg, but was 20 μg/ kg for cyhalothrin and triflumizole.

Data analysis

As growers decided independently what and when to spray (S1 and S3 Tables), the number of samples tested and detected was tallied separately for each chemical (and for each of two categories – insecticide and fungicide) using only the data from farms where each was used. We present the percent detection of each chemical in each tissue type over time (from 1 d before through 7 d after spray). All analyses were conducted using the R programming language, version 4.3.2 [53].

To assess risk to bees from contact with pumpkin leaves or consumption of pollen and nectar after foliar spray, three risk indices were calculated – Hazard Quotient (HQ), a variation of Hazard Quotient (HQ_PPB), and US EPA BeeREX Risk Quotient (RQ). First, Hazard Quotient (HQ) was calculated for each tissue type as the concentration of a chemical residue in the sample in μg/ kg divided by LD₅₀ (the dose in μg/ bee that would kill 50% of a test honey bee population) [54]. Acute oral LD₅₀ was used to calculate HQ for pesticide concentrations in pollen and nectar, contact LD₅₀ for leaf, average of contact and oral LD₅₀ for bee samples. Honey bee LD₅₀ was used as LD₅₀ was not consistently available for bumble and squash bees (Table 1). Second, we created a variation of HQ (hereafter referred to as HQ_PPB) that first converts LD₅₀ into the same units as the pesticide concentrations (μg/ kg, or PPB, instead of μg/ bee) to allow for direct comparisons between pesticide load and toxicity. LD₅₀ values were converted from μg/ bee to μg/ kg of bee tissue by dividing LD₅₀ values by 0.00012 kg (the average weight of a honey bee worker [55]). This index expresses pesticide concentration in the sample as the proportion of the LD₅₀ that a honey bee forager would have reached if it eats its own body weight in the contaminated nectar or pollen. The following is an example of how we calculated and interpreted HQ_PPB versus HQ. The maximum concentration of carbaryl we detected in nectar 1 d after spray was 44.1 μg/ kg. The oral LD₅₀ of carbaryl is 0.21 μg/ bee (Table 1), or 1750 μg carbaryl per kg of nectar after conversion to PPB. Therefore, HQ_PPB =  0.03, which indicates that concentration would not pose a high risk to bees because it represents only a small proportion of LD₅₀. In terms of risk to individual bees, if an adult honey bee foraging for pollen consumed 0.0000435 kg of pumpkin nectar the day after spray [56] at a concentration of 44.1 μg carbaryl/ kg nectar, the bee would have ingested 0.002 μg of carbaryl, which is 1% of the LD₅₀ of 0.21 μg/ bee. By comparison, for the same carbaryl concentration HQ =  210, which would be interpreted as a potential risk to bees using a common level of concern (LOC) of HQ >  50 [57,58].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary of agrochemical sprays analyzed in pumpkin leaf, pollen, and nectar, and in bee visitors.

https://doi.org/10.1371/journal.pone.0311634.t001

In addition to HQ and HQ_PPB, we calculated contact and dietary Risk Quotient (RQ_contact, RQ_pollen, RQ_nectar) using the U.S. Environmental Protection Agency’s BeeREX tool, which is intended for foliar sprays applied to crops in bloom [56]. Risk quotient has the advantage of taking into account the amount of the contaminated substance consumed or encountered by a typical honey bee forager [57]. The BeeREX tool calculates dietary RQ separately for adult honey bee pollen and nectar foragers and compares empirical values of residue concentrations in pollen and nectar, multiplied by the typical amount of pollen and nectar consumed by a honey bee nectar or pollen forager, to oral acute LD₅₀. Contact RQ is calculated by comparing the chemical application rate, multiplied by a constant that represents the typical amount of chemical encountered by a honey bee forager if it flies through a cloud of spray, to the contact acute LD₅₀. Ideally, to assess bee risk of pesticide contamination from leaf contact, this index should be further modified to also take into account the proportion of a bee’s body that comes into contact with the plant when it is resting on foliage, but that data was not available. For further details on how the BeeREX tool calculates risk quotient, see the US EPA document Guidance for Assessing Pesticide Risks to Bees, Appendix 3 [59].

To interpret risk to bees associated with each tissue type and chemical, risk index values were compared to a pre-determined level of concern (LOC) for each index and interpreted as ‘low risk’ if below the LOC or ‘low risk cannot be concluded’ if exceeding LOC [57]. HQ values above 50 [57,58] and RQ values above 0.4 [59] are considered a potential risk to pollinators. A threshold LOC has not yet been established for HQ_PPB, but a simple and direct way to interpret it would be that values greater than 1 indicate a significant likelihood of acute bee toxicity (because if HQ_PPB >  1, the quantity of pesticide in a pollen or nectar sample that has the same mass as an average honey bee worker exceeds honey bee LD₅₀).

Detection of foliar-applied pesticides in pumpkin tissues and pollinators in the week after spray

There are two likely explanations for how non-systemic pesticides sprayed on foliage still reached pumpkin pollen, nectar, and foraging male squash bees, even though flowers were unopened during application. First, chemical-laden droplets may have leaked into flowers during spray and directly contaminated pollen and nectaries if the buds were not tightly sealed. Similarly, spray droplets may have also accumulated in wilted flowers where male squash bees were resting for the night, although only one chemical out of six (carbaryl) was detected in squash bees. Second, chemicals on the leaf surface or in stomal apertures may have crossed lipid boundaries, mobilized in phloem, and become incorporated into developing flowers. Environmental conditions, the addition of adjuvants (e.g., surfactants), and the chemical properties of pesticides such as water solubility, lipophilicity, ionizability, and molecular weight all contribute to their ability to infiltrate a plant’s epidermis and translocate to non-target regions [62–66]. These xenobiotic chemicals move through plants along both lipophilic and aqueous pathways, and undergo complex and transformative interactions [65]. Pesticides applied to crop plants are typically slightly lipophilic so that they can permeate the cuticle, but not be completely adsorbed in intra-cuticular non-crystalline wax [64]. They are also often weak electrolytes that can dissolve in water and be transported to different plant compartments [62,63]. Overall, pesticides that are somewhat lipophilic weak acids are most mobile; they are bioactive under certain conditions yet do not ionize at the pH levels of aqueous reservoirs inside plant cells and become ‘ion trapped’ before reaching phloem to disperse [62].

The lipophilicity and other properties of our target agrochemicals can partially explain their biological fate in pumpkin tissues in our dataset [62–65]. For example, three of the four pesticides we detected in pollen – permethrin, quinoxyfen, and triflumizole – are lipophilic (logK_OW >  4.5, Table 1), making them likely to adsorb onto lipid-rich tissues like the pollen coat, if they are not first trapped in leaf cuticle waxes. Another insecticide we detected in pollen – carbaryl – has only moderate lipophilicity (logK_OW =  2.4), giving it more potential to mobilize vascularly and be incorporated into developing floral tissue. Consistent with this reasoning, we recorded a five-fold increase in carbaryl concentrations in pollen from the first to the third day after spray (Table 5). However, our low sample sizes make us unable to resolve whether those values are within the typical range of variation for pollen or are evidence of systemic uptake. Another consideration for carbaryl is that it also has a low molecular weight and is a very weak acid (dissociates more readily at low pH levels), making it likely to cross membranes and to ionize and bind with compounds in relatively acidic compartments inside cells like vacuoles (pH ~  5) before it reaches phloem (pH ~  8) [62]. These properties contribute to its persistence in leaves, instead of translocation to pollen and nectar that bees eat. This example of carbaryl demonstrates the complexity of interactions pesticides undergo within crop plants; no one factor can fully explain their presence.

In addition to the properties of the agrochemicals, differences in the composition of plant tissues and the timing of their availability also influence their uptake of pesticides. For example, pollen has a higher lipid and protein content [67] and takes more time to develop than nectar, which may contribute to its greater pesticide load in ours and other studies [1,11,15,19]. We detected four out of six agrochemicals (two insecticides and two fungicides) in pollen after foliar application. Fungicide was detected almost as frequently in pollen as in leaves (Table 4). In contrast, only one insecticide and no fungicides were found in nectar. Nectar is secreted after flower anthesis [68], and so should not have been present in unopened flowers during spray.

Based on our detection of pesticides in pollen, it is likely that bee larvae fed on pumpkin pollen from these fields were exposed to multiple agrochemicals simultaneously, with the potential for additive, or even interactive toxicity [41,45]. However, in this field study, wild bee nest locations were unknown and we were not able to test larvae from the nests of the adult foragers we collected. Laboratory trials with contaminated pollen fed to larvae of honey bees or other agriculturally important pollinators could illuminate the effects of consuming pesticide residues on larval development and survival [39,41,69,70]. Population-level studies that investigate multiple exposure routes and outcomes for larvae and adult bees such as that by Willis-Chan and Raine [20] are best able to assess total risk.

Consumption of contaminated pollen and nectar is arguably a greater concern for bees than leaf contact exposure because it can impact larvae as well as adults and lead to population-level effects [14,15]. However, the persistence of the insecticide carbaryl in leaves between spraying events also raises concern from the pollinator perspective because it prolongs their risk of exposure. In addition, the high concentrations of pesticides in leaves during the week after foliar spray led to the highest bee risk quotient values in our dataset (Table 3). However, risk assessments that aim to quantify bee toxicity from leaf contact may overestimate the true risk to bees since they assume that a bee actually encountered the entire dose of the chemical present in the leaf sample. Instead, only the bee’s lower legs and tip of the abdomen contact a leaf on which it is resting (authors’ personal observations).

Bee contamination was infrequent in our dataset. We detected one insecticide (carbaryl, the most widely used by growers in this study) in two out of 69 total bee samples and at concentrations that were not acutely lethal. Both positive bee samples were from the same field on the same date and consisted of bees that we expected to have the highest risk of exposure – males of X. pruinosa, a Cucurbita pollen specialist bee species, which spend much of their adult lives resting in wilted squash flowers [1]. On that particular farm, carbaryl was also present in one out of three nectar samples the day after spray. Therefore, we are unable to distinguish whether the male squash bees were exposed directly while resting in wilted flowers at the time of spray, or via translocation of the pesticide through the plant to the nectar. Wild bees foraging or nesting in pumpkin fields are also at risk of exposure to agrochemicals through non-dietary media (e.g., contaminated soils or water in or near crop fields), which can contribute to multiple additive exposures, but these routes were beyond the scope of our study [13,19,20,71].

Since so few bees tested positive the day after spray application (when they would have come into contact with the greatest concentrations of pesticides), it seems unlikely that additional sampling on Days 3 and 7 would have revealed higher contamination rates. However, our data may be subject to survivorship bias; our sampling method only captured living bees that were still able to forage the day after a spray event, and not those that had encountered or accumulated a lethal dose. Monitoring nearby honey bee hives for worker losses before and after pesticide applications could have offered further insight [72].