Pesticide Residues in Honey Bee (Apis mellifera) Pollen Collected in Two Ornamental Plant Nurseries in Connecticut: Implications for Bee Health and Risk Assessment

Abstract

Honey bees (Apis mellifera L.) are one of the most important managed pollinators of agricultural crops. While potential effects of agricultural pesticides on honey bee health have been investigated in some settings, risks to honey bees associated with exposures occurring in the plant nursery setting have received little attention. We sought to identify and quantify pesticide levels present in honey bee-collected pollen harvested in two ornamental plant nurseries (i.e., Nursery A and Nursery B) in Connecticut. From June to September 2018, pollen was collected weekly from 8 colonies using bottom-mounted pollen traps. Fifty-five unique pesticides (including related metabolites) were detected: 24 insecticides, 20 fungicides, and 11 herbicides. Some of the pesticide contaminants detected in the pollen had not been applied by the nurseries, indicating that the honey bee colonies did not exclusively forage on pollen at their respective nursery. The average number of pesticides per sample was similar at both nurseries (i.e., 12.9 at Nursery A and 14.2 at Nursery B). To estimate the potential risk posed to honey bees from these samples, we utilized the EPA’s BeeREX tool to calculate risk quotients (RQs) for each pesticide within each sample. The median aggregate RQ for nurse bees, was 0.003 at both nurseries, well below the acute risk level of concern (LOC) of ≥ 0.4. We also calculated RQs for larvae due to their increased sensitivity to certain pesticides. In total, 6 samples had larval RQs above the LOC (0.45 – 2.51), resulting from the organophosphorus insecticide diazinon. Since 2015, the frequency and amount of diazinon detected in pollen increased at one of our study locations, potentially due to pressure to reduce the use of neonicotinoid insecticides. Overall, these data highlight the importance of considering all life stages when estimating potential risk to honey bee colonies from pesticide exposure.

1. Introduction

Pollinators provide critical ecological services essential to maintaining our food supply and valued natural habitats. Honey bees (Apis mellifera L.) are one of the most important agricultural pollinators, pollinating at least 90 commercial crops in North America (The White House 2014). Among these commercial crops are ornamental plants, which represent a major industry in the U.S., with annual sales over ~$10 billion as of 2019 (USDA-NASS, 2020). Although the potential for ornamental plants and plant nurseries to provide resources for pollinators is being increasingly understood (Baldock et al., 2019, Erickson et al., 2019, Erickson et al., 2021, Frankie et al., 2005, Garbuzov et al., 2015, Lowenstein et al., 2019, Mach et al., 2018, Potter and Mach, 2022), relatively few studies have been conducted in the ornamental plant nursery setting.

In recent decades, there has been a significant decline in populations of some insect pollinators in the U.S. and other geographies (Goulson et al., 2015). It is widely recognized that honey bees are often stressed by multiple factors including genetic diversity, habitat loss, pests/pathogens, poor nutrition, and pesticides (Goulson et al., 2015, Potts et al., 2010). Honey bees can be exposed to pesticides through a variety of mechanisms including particles in air (dust and spray), nectar, pollen, mud/soil, honeydew, wax, water, guttation fluid, plant surfaces, and propolis/resin (Gradish et al., 2019). Although many ornamental plants are unattractive to insect pollinators (Garbuzov et al., 2017), some ornamental plants are utilized by bees as nutritional resources. Lentola and colleagues reported a wide variety of pesticides were detected in the nectar and pollen of bee-attractive ornamental plants sold in the UK (Lentola et al., 2017).

Concerningly, some samples analyzed in the previous study contained concentrations of neonicotinoids which may be sufficient to cause sublethal effects in honey bees and bumble bees (Lentola et al., 2017). Experimental studies have demonstrated that similar pesticide concentrations can cause reductions in growth, queen production, and foraging ability at the colony level and immunosuppression in individual bees (Di Prisco et al., 2013, Feltham et al., 2014, Goulson, 2015, Whitehorn et al., 2012). Reflecting different pest management goals, ornamental plants in a nursery setting can be treated with insecticides not registered for use on food crops and with higher levels of systemic insecticides (potentially 120 times higher) to meet commercial aesthetic standards or to control pests to meet phytosanitary standards in order to ship plants across state or national boundaries (Bethke and Cloyd, 2009, Brown et al., 2013, Hopwood et al., 2012). Consequently, bees utilizing ornamental plants as a source of nectar and pollen forage may be at greater likelihood of harm from exposure (i.e., risk) than when foraging in other agricultural environments (Lentola et al., 2017). Reflecting these points, some consumers are selecting “pollinator friendly” ornamental plants (Khachatryan and Rihn, 2020), some ornamental plant producers have taken steps to reduce the presence of pesticide residues in the nectar and pollen of ornamental plants.

While potential effects of agricultural use of pesticides on honey bee health have been investigated in some settings, risks associated with exposures occurring in the plant nursery setting have received much less attention than in agricultural settings (Cowles and Eitzer, 2017, Mach et al., 2018, Stoner et al., 2019). The limited amount of data on pesticide applications to ornamental plants and to nursery crops and on pesticide residues in the pollen and nectar of these plants is of particular concern to many stakeholders (e.g., regulators, nursery operators, pesticide applicators). Pesticide exposure in this setting has not been adequately characterized, and exposure data are needed to understand the potential risks for bees exposed in this environment.

In previous analyses of pesticide residues in pollen, we have evaluated the potential pesticide risk to honey bee colonies using the pollen Hazard Quotient (HQ; pesticide concentration in parts per billion (ppb; μg/kg) divided by acute oral LD₅₀ for the pesticide) (Stoner et al., 2019, Stoner and Eitzer, 2013). The use of HQs as a measure of pesticide risk has recently come under criticism because it does not account for the amount of pesticide consumed and there is no experimentally established level of concern (LOC) for HQs (Thompson, 2021). Furthermore, broad-based surveys of pesticide contamination and colony health have found only weak associations between HQs and measures of colony health (Carlson et al., 2022).

Calculating Risk Quotients (RQs; i.e., ratio of daily exposure through pesticide consumption divided by acute oral LD₅₀ or similar ratio for contact exposure) using the U.S. Environmental Protection Agency's (USEPA) BeeREX model has emerged as an alternative to the use of HQs (USEPA, 2015). BeeREX is a screening-level tool for determining whether exposure to individual pesticide active ingredients reach a threshold LOC (USEPA, 2012). A LOC ≥ 0.4 for acute pesticide exposure can trigger additional refinements in exposure estimates, colony-level studies to evaluate potential impacts, and/or mitigation measures to reduce exposure to honey bees (USEPA, 2014). BeeREX is a flexible tool with the ability to estimate pesticide acute and chronic risks to individual bees in different life stages, castes, and worker tasks in the honey bee colony (USEPA, 2015). Although the LOC is based on laboratory testing of effects of individual pesticide active ingredients on individual bees and does not apply to colony-level study data, the BeeREX model and the associated LOC of 0.4 have recently replaced the pollen HQ as a method of evaluating pesticide risk to honey bees in field measurements of mixed pesticide residues and exposures (Demares et al., 2022, Graham et al., 2022).

Connecticut is home to a thriving horticulture industry, including large plant nurseries and greenhouse operations generating $230 million in annual sales as of the 2019 (USDA-NASS, 2020). In a previous 2015 study conducted at commercial ornamental plant nurseries in Connecticut, we reported differences in the levels and types of pesticides detected in pollen trapped from honey bee colonies compared to those found in pollen trapped in other environments in the state (Stoner et al., 2019). In the present study, we collected pollen at two ornamental plant nurseries in Connecticut in 2018. Our objectives were to quantify pesticide residues present in honey bee-collected pollen, evaluate trends between 2015 and 2018, and to put our findings in context, calculate pesticide risk to adults and larvae using the USEPA's BeeREX model (USEPA, 2015).

2. Materials and Methods

2.1. Study Sites and Pollen Collection

Honey bee colonies (Apis mellifera L., Carniolan race) were established, maintained, and sampled as described previously (Stoner et al., 2019). Four established colonies were placed at each of the study locations (i.e., Nursery A and Nursery B) in areas which were close to a water source and had minimal risk of a direct pesticide spray. To avoid introducing external pesticides in the environment, no miticides were applied to the colonies during this study. Hive entrance activity was observed weekly, and general colony condition was assessed every two weeks. No significant colony health issues were noted. The two nurseries are 59 km apart; therefore, the hives at each nursery had unique (i.e., non-overlapping) foraging areas. Nursery A is a 183-hectare nursery in north-central Connecticut surrounded by agricultural fields, suburban development, and forested area. Nursery B is similarly sized (168 hectares) but located in eastern Connecticut and primarily surrounded by agricultural fields mixed with forested area. A more detailed analysis of the land cover of the areas surrounding the nurseries is available in Stoner et al. (2022).

To estimate the colonies’ pesticide exposure at each nursery, pollen was trapped using Sundance bottom-mounted pollen traps (Ross Rounds, Inc., Canandaigua, NY) and collected on a weekly basis from June to September 2018. The individual colonies were set to trap pollen on a biweekly rotating basis, so that each hive had sufficient opportunity to collect pollen for its own use. After collection, pollen samples were stored in −18°C freezers until analysis. Pollen samples analyzed for pesticide residues were derived from the same samples collected and subjected to palynology in Stoner et al. (2022).

2.2. Pollen Analysis

Pollen samples were analyzed in-house at the Connecticut Agricultural Experiment Station and by the USDA Agricultural Marketing Service National Science Laboratory (AMS-NSL) in Gastonia, NC. A complete list of all pesticides which were screened for (217 in total) and their limits of detection is contained in Supplemental File A. In all cases, samples were prepared for analysis using the QuEChERS extraction protocol (Anastassiades et al., 2003). To detect a wide range of pesticides and relevant metabolites, extracted samples were analyzed using liquid (LC) and gas chromatography coupled with single or tandem mass spectrometry (MS). The LC-MS was performed using an Agilent 1200 LC with a Hypersil GOLD aQ C18 column (100 mm length × 2.1 mm width, 1.9 μm particle size) interfaced to a Thermo Exactive Mass spectrometer, operated in both positive and negative electrospray modes. LC-MS/MS was performed using an Agilent 1290 Infinity LC system spectrometer with a Zorbax Eclipse Plus C18 Rapid Resolution HD column (50 mm length x 2.1 mm width, 1.8 μm particle size) interfaced with Agilent 6420A triple quadrupole mass spectrometer. Gas chromatography (GC) was done using an Agilent 7890 GC system with an Agilent DB-5MS+DG with Integra-guard (30 m length x 0.25 mm inner diameter, 0.25 μm film thickness) column interfaced with either an Agilent 5975 mass spectrometer in negative chemical ionization mode (for GC-MS/NCI) or an Agilent 7010 triple quadrupole mass spectrometer (for GC-MS/MS). Pesticide quantification (in ppb) and associated detection limits were determined using external standards with a known concentration. In cases where pesticides were detectable by multiple methods, only the results from the most sensitive method were reported.

2.3. Data Analysis

Descriptive statistics for each sample (e.g., pesticide counts) and each pesticide (e.g., percent detection, mean, standard deviation) were calculated using the nursery as the unit of analysis. Central tendency measures were calculated using samples with positive detections only; otherwise, non-detects were treated as zeros for analysis. When samples are presented individually, the hive the sample was collected from is also included. Risk quotients (RQs) were calculated with EPA’s BeeREX tool (v1.0), using ingestion rates of 9.6 mg of pollen/day for adult in-hive workers (nurse bees) and 3.6 mg of pollen/day for larval workers (consumption estimates based on 5-day-old worker larvae). The Pesticide Properties Database (University of Hertfordshire) was used for acute toxicity values (i.e., oral LD₅₀ values) (Lewis et al., 2016) and, when necessary, supplemented with values reported in Sanchez-Bayo and Goka (2014). In the absence of acute toxicity data for pesticide metabolites, toxicity values for the parent compound were used. Contact LD₅₀ values were substituted when oral LD₅₀ values were not available. Larval LD₅₀ values for insecticides were collected from Chmiel et al. (2020), which reported median toxicity values calculated from those reported in the literature; not all of these underlying studies used standardized methodology, thus these data are likely to be more variable than the adult toxicity values. Toxicity values used for detected pesticides are listed in Supplemental File A. Insecticides were further categorized using the Insecticide Resistance Action Committee’s (IRAC) mechanism of action classifications (Sparks and Nauen, 2015). All data transformation and visualizations were performed using R (version 4.0.3), and scripts will be made available upon request.

3. Results

3.1. Sample Composition

A total of 25 pollen samples from Nursery A and 24 pollen samples from Nursery B were collected between June 7^th and September 6^th of 2018 and subsequently tested for pesticide residues. Fifty-five unique pesticides (including related metabolites) were detected: 24 insecticides, 20 fungicides, and 11 herbicides. A minimum of 6 pesticides were detected in each of the 49 samples analyzed. The average number of pesticides per sample was similar at both nurseries (12.9 at Nursery A (SD = 3.2) and 14.2 at Nursery B (SD = 6.2)). Four samples from Nursery B contained greater than 20 individual pesticides, while samples from Nursery A contained no more than 19 unique pesticides (Figure 1A). When detections were further analyzed by pesticide class, fungicides constituted 41% and 46% of detections per sample at Nursery A and Nursery B, respectively, while insecticides and herbicides contributed 25-30% each (Figure 1B).

Figure 1.

Total number pesticides and pesticide classes detected in honey bee (Apis mellifera) collected pollen sampled at two commercial plant nurseries in Connecticut. Panel A depicts the distribution of the total number of pesticides detected in each sample. The vertical dotted lines correspond to the collective mean number of pesticides at each nursery. Panel B shows the average number of unique fungicides, herbicides, and insecticides in each sample. Bars represent the mean ± SD (n = 24-25).

Descriptive statistics (including detection frequencies, means, standard deviations, and maximums) of each insecticide, fungicide, and herbicide at each nursery are displayed in Tables 1, 2, and 3, respectively. The neonicotinoids acetamiprid and imidacloprid were the most frequently detected insecticides at Nursery A and Nursery B, respectively. The organophosphate insecticide chlorpyrifos and the butenolide insecticide flupyradifurone were the second and third most frequently detected insecticides at Nursery A but were not prominently represented at Nursery B. In contrast, 58% of samples at Nursery B contained organophosphorus insecticide diazinon which was not detected at Nursery A. Both nurseries had several fungicides that were detected in >50% of all samples (Nursery A = 4; Nursery B = 5); however, the anilino-pyrimidine fungicide cyprodinil was the only fungicide that was detected in a majority of samples at both nurseries (and nearly ubiquitously at Nursery B). The azole fungicide myclobutanil was a particularly prominent constituent at Nursery A (in 92% of samples), but at Nursery B it was found less frequently and in smaller amounts. At both nurseries, the dinitroaniline herbicide pendimethalin, the triazine herbicide atrazine, and the pyridine herbicides dithiopyr were the top three herbicides detected.

Table 1.

Summary of insecticides detected in honey bee (Apis mellifera) collected pollen sampled from two nurseries in Connecticut. Units in parts per billion (ppb; μg/kg).

Insecticide	LOD	Nursery A					Nursery B
		% Detections	Median	Mean	SD	Max	% Detections	Median	Mean	SD	Max
Acephate	2	16	4.8	5.3	1.9	8	29	5.3	39.8	65.7	183.3
Acetamiprid	0.5	64	8.2	14.3	17.4	55	21	2.2	2.1	1.2	3.4
Bifenazate	2	0	—	—	—	—	4	58.9	58.9	—	58.9
Bifenthrin	25	24	36	42.3	22.7	85	8	85	85	0	85
Carbaryl	1	8	5.6	5.6	3.4	8	0	—	—	—	—
Chlorantraniliprole	1	32	3.3	5.1	4.8	15.8	0	—	—	—	—
Chlorfenapyr	13	0	—	—	—	—	8	245	245	217.8	399
Chlorpyrifos	1	60	10.3	18.5	25.7	102.5	8	2.7	2.7	0	2.7
Diazinon	0.1	0	—	—	—	—	58	8.1	13.2	17.6	60.9
Diflubenzuron	1	4	2.6	2.6	—	2.6	29	13	21.8	22.1	60.9
Dinotefuran	0.2	0	—	—	—	—	12	0.4	0.4	0.1	0.5
Etoxazole	1	0	—	—	—	—	4	3	3	—	3
Flonicamid	2	0	—	—	—	—	29	4.2	27.7	57.4	157.6
Flupyradifurone	5	56	38	66.2	80.1	305	0	—	—	—	—
Hexythiazox	0.5	0	—	—	—	—	12	0.8	1.1	0.6	1.8
Imidacloprid	0.2	32	0.6	1.2	1.9	5.9	71	0.8	1.1	1	4.3
Malathion	0.2	0	—	—	—	—	38	0.4	12.3	30.8	93.8
Methamidophos*	0.5	24	1.8	2	1.2	3.8	25	3.4	5.7	6	15.5
Methoprene	2,500	4	3,780	3,780	—	3,780	8	5,500	5,500	42.4	5,530
Phosmet	0.1	16	0.5	0.6	0.4	1.1	21	0.3	0.5	0.4	1.1
Piperonyl Butoxide†	1	4	1.5	1.5	—	1.5	0	—	—	—	—
Resmethrin	2	0	—	—	—	—	4	2.9	2.9	—	2.9
Spiromesifen	2	4	14.1	14.1	—	14.1	17	95.5	131.1	143.8	325.6
Thiamethoxam	0.5	28	2.6	2.1	1	3.2	8	2.6	2.6	1	3.3

^*Could be an active ingredient or a metabolite of acephate

^†Insecticide synergist

Table 2.

Summary of fungicides detected in honey bee (Apis mellifera) collected pollen sampled from two commercial plant nurseries in Connecticut. Units in parts per billion (ppb; μg/kg).

Fungicide	LOD	Nursery A						Nursery B
		% Detections	Median	Mean	SD	Max	% Detections	Median	Mean	SD	Max
Azoxystrobin	0.2	60	5.3	36.8	80.9	318.8	25	1	5.7	11.1	28.2
Boscalid	0.5	72	9.6	84	146.9	491	42	7.8	14.8	22.5	74.9
Captan	125	28	314	429.6	267.8	725	0	—	—	—	—
Carbendazim*	25	0	—	—	—	—	21	98	115	86.8	265
Chlorothalonil	10	0	—	—	—	—	58	189	1,035.1	2,877	11,000
Cyprodinil	0.1	84	1.8	8.9	16.8	58.7	96	1.7	26.4	91.3	422.7
Dimethomorph	0.2	0	—	—	—	—	21	2.6	9.5	14.8	35.8
Fludioxonil	1	0	—	—	—	—	12	44.2	189.6	289.7	523.2
Fluopicolide	3	8	3.5	3.5	0.7	4	0	—	—	—	—
Iprodione	0.2	0	—	—	—	—	67	3.1	7.6	9	25.1
Metalaxyl	0.5	0	—	—	—	—	29	0.8	2.1	2.4	6.7
Myclobutanil	0.1	92	10.9	17.8	20.7	82.2	38	1	1.5	1.2	4.3
4-OH-chlorothalonil†	0.5	28	1	1.3	0.7	2.7	79	4	36.2	106.1	462.1
Propiconazole	0.5	0	—	—	—	—	29	4.1	28.1	54.6	150.4
Pyraclostrobin	0.5	48	7.8	22	24.5	60.4	21	3	5.4	4.8	13.5
Tebuconazole	5	0	—	—	—	—	25	18.1	29.6	27.2	70.9
Tetrahydrophthalimide‡	125	24	174.5	244.8	159.2	550	0	—	—	—	—
Thiophanate-methyl	4	32	0.6	1.8	2.8	8.3	79	2	107.7	353.5	1,488.7
Triadimefon	8	0	—	—	—	—	21	377	268.8	236.2	506
Trifloxystrobin	0.2	40	39.3	33.5	21.5	67.5	0	—	—	—	—

^*Could be an active ingredient or metabolite of thiophanate-methyl

^†Chlorothalonil metabolite

^‡Captan metabolite

Table 3.

Summary of herbicides detected in honey bee (Apis mellifera) collected pollen sampled from two commercial plant nurseries in Connecticut. Units in parts per billion (ppb; μg/kg).

Herbicide	LOD	Nursery A					Nursery B
		% Detections	Median	Mean	SD	Max	% Detections	Median	Mean	SD	Max
Atrazine	0.1	72	0.2	0.3	0.2	0.9	67	1.5	2.8	4	15.3
2,6-Dichlorobenzamide*	5	4	5	5	—	5	0	—	—	—	—
Dimethenamid	0.1	40	0.4	1.3	1.8	5.7	25	3.1	5.2	5.4	12.6
Dithiopyr	0.2	84	1.6	2.4	2	6.9	50	1	2.4	4.3	15.8
Flumioxazin	2	12	4.2	4	0.7	4.6	4	12	12	—	12
Metolachlor	0.2	32	0.8	1.4	1.7	5.3	42	2.1	4	4.4	13.1
Oxyfluorfen	2	0	—	—	—	—	17	5.8	6.1	4.2	10.6
Pendimethalin	0.2	92	4.2	10.1	14.7	61.4	96	3.6	8.5	9.4	34.5
Prodiamine	10	8	13.6	13.6	1.3	14.5	0	—	—	—	—
Simazine	0.5	52	3.5	13.1	17.1	55.2	0	—	—	—	—
Trifluralin	13	0	—	—	—	—	38	33	43.4	33.3	124

^*Dichlobenil metabolite

3.2. Origin of Pesticides in Pollen

Applications of insecticides and fungicides were documented by the nurseries and provided to us, and a summary of active ingredient totals applied during the study period is provided in Supplemental File A. Using these data, Figure 2 shows the intersection of insecticides which were detected in pollen and/or applied at the nurseries. At Nursery B, only 3 out of 19 detected insecticides were not directly applied by the nursery. At Nursery A, 8 out of 14 detected insecticides were not directly applied at the nursery. Most interestingly, acetamiprid was not applied by Nursery A despite it being found in 64% of samples (Table 1). Phosmet and methoprene were detected at both nurseries but not applied by either. A comparison of detected versus applied fungicides is provided in Supplemental File B (Figure S1).

Figure 2.

Comparison of insecticides that were only applied (left circles), only detected (right circles), or applied and detected (overlap) at each commercial plant nursery in Connecticut. Red text denotes compounds that were not included in our pesticide screen and therefore were not possible to detect.

3.3. Pesticide Trends from 2015 to 2018

We also analyzed insecticide and fungicide residues in pollen across time, using data previously collected from these nurseries in 2015 (Stoner et al., 2019). In some cases, the limits of detection (LOD) decreased (between 2- and 50-fold) between the 2015 and 2018 samples, which skewed the observed differences for a few pesticides. Therefore, when comparing percent detections, we only considered data above the less sensitive LOD. For example, the absolute number of samples containing imidacloprid increased at Nursery B in 2018, but samples with detections above 2 ppb (i.e., the 2015 LOD for imidacloprid) decreased from 32% to 8% (Figure 3A). Furthermore, the median imidacloprid values in this study (0.6 – 0.8 ppb) are below what were detectable for the 2015 samples. Total phosmet detections also increased at both nurseries, solely due to samples containing levels of phosmet which were not detectable in 2015 (<5 ppb). The median values of most individual insecticides in samples decreased between 2015 and 2018; however, the median values of three insecticides (i.e., acetamiprid, chlorpyrifos, and diazinon) increased (Figure 3B). Median acetamiprid residues increased modestly at Nursery B (39%), and median chlorpyrifos residues at both nurseries roughly doubled. Importantly, there was more than a 900% increase in the median value of diazinon at Nursery B. Diazinon was detected in 16% of Nursery B pollen samples from 2015, and in 2018, 58% of samples at Nursery B contained diazinon at levels above 0.5 ppb (i.e., the 2015 LOD for diazinon). Insecticides which were first detected in 2018 (i.e., acetamiprid and neonicotinoid thiamethoxam for Nursery A; the keto-enol insecticide spiromesifen and phosmet for Nursery B) are not shown in Figure 3B. With the exception of phosmet, these new detections cannot be solely attributed to increases in the detection sensitivity. Changes in the frequency and amount of fungicide residues detected in pollen from 2015 to 2018 are shown in Supplemental File B (Figure S2).

Figure 3.

Panel A shows the detection frequency of insecticides at the nurseries using honey bee (Apis mellifera) collected pollen sampled from commercial plant nurseries in Connecticut in 2015 (Stoner et al. 2019) and 2018. Due to differences in analytical instrument sensitivity between these datasets, only data which were detectable at both time points were considered (i.e., data which were above the less sensitive level of detection (LOD)). The lower limit used for each insecticide is shown in parentheses, and detections below this limit were not included in this figure. Panel B shows the percent change in the reported median concentrations of pesticides detected in both 2015 and 2018 samples.

3.4. Risk Determination

Using the EPA’s BeeREX model, we determined RQ values for each pesticide within each sample using adult nurse bees, the adult stage that consumes the most pollen, as the model organism. The RQs for each pesticide within a sample were then summed to create the sample’s Aggregate RQ (ARQ). At both nurseries, the median ARQ was 0.003, and the highest ARQ for an individual sample was an order of magnitude higher than the median (i.e., 0.044) (Figure 4). These values are all well below BeeREX’s acute risk LOC of 0.4. Figure 5 shows the relative contribution of each pesticide type to the ARQ for each sample. Unsurprisingly, insecticides account for the majority of the ARQs, so this category was subdivided into three IRAC mechanism of action categories (i.e., nicotinic acetylcholine receptor modulators, acetylcholinesterase inhibitors, and sodium channel modulators) as well as an “Other” category for insecticides with mechanisms of action outside of our three primary categories. In all but four samples (one at Nursery A and three at Nursery B), a single pesticide type accounted for a majority of the risk. Nicotinic acetylcholine receptor (nAChR) modulators were the primary contributer to the ARQ in 48% (Nursery A) or 46% (Nursery B) of the pollen samples. At Nursery B, insecticides belonging to this class were all neonicotinoids, but at Nursery A the butenolide flupyradifurone was also a contributing factor. Acetamiprid contributed the least to RQ out of all nAChR modulators. For 28% (Nursery A) or 21% (Nursery B) of samples, acetylcholinesterase inhibitors (primarily organophosphates) were the major contributor, making them the second most prominent pesticide type. Although sodium channel modulators (e.g., synthetic pyrethroids) were infrequently detected, when present they contributed between 20% and 98% of observed risk. Spikes in the contribution of “Other” insecticides were a result of the juvenile hormone analogue methoprene (Sample 4.3 at Nursery A) and the pyrrole chlorfenapyr (Samples 5.5 and 7.4 at Nursery B). Fungicides and herbicides were generally a minor contributor to the overall risk, except in a handful of samples which contained minimal or no insecticides.

Figure 4.

Box plot of aggregate adult risk quotients (ARQ) based on pesticide residues in pollen sampled each commercial plant nursery. Each sample’s ARQ is the sum of the RQs from individual pesticides detected in that sample. For each nursery, the first quartile, second quartile (i.e., median), and third quartile correspond to the bottom, middle, and top lines of the box, respectively. Individual samples are represented as points on the plot, and its colony of origin is indicated by color. Samples with ARQs in the top 25% overall are also labeled with their Sample IDs, which indicates both the hive the sample was collected from and the collection order (e.g., Sample ID 4.2 indicates the second sample from hive 4).

Figure 5.

Heatmap of the percent contribution of each pesticide type to the ARQ of individual samples. Because insecticides are the primary contributer to risk, this category is subdivided into categories based on their mode of action (acetylcholinesterase (AChE) Inhibitors, nicotinic acetylcholine receptor (nAChR) modifiers, sodium (sodium channel modifiers), and others). Individual samples are labeled with their Sample ID, which indicates both the hive the sample was collected from and the collection order (e.g., Sample ID 4.2 indicates the second sample from hive 4). Blank squares indicate that the pesticide type was not detected within that sample, while squares showing 0 indicate that the pesticide type was detected but contributed less than 0.05% to the ARQ for that sample.

In addition to the adult ARQs, we calculated larval RQ values for selected insecticides (i.e., insecticides which had larval LD₅₀ values available). Larvae consume less pollen than adult nurse bees (3.6 mg/day for larvae versus 9.6 mg/day for adults) but may be more sensitive to certain pesticides. No samples at Nursery A had larval RQ values above the acute risk LOC of 0.4. Six samples from three hives at Nursery B had larval RQs above the LOC (Table 4). In each case, the elevated risk was a result of the organophosphorous insecticide diazinon. Three consecutive pollen samples from Hive 6 were above the LOC for acute risk to larvae, and two consecutive samples from Hive 8. Out of the insecticides we evaluated, diazinon had the largest difference between the reported adult and larval LD₅₀ values.

Table 4.

Samples from Nursery B with larval honey bee (Apis mellifera) Risk Quotient (RQ) values above the acute risk level of concern (LOC =0.4).

Date	Hive	Diazinon Detected (ppb)	BeeRex Risk Quotient
			Adult1	Larval2
8/9/2018	6	11.00	0.001	0.45
8/23/2018	7	44.98	0.005	1.86
8/30/2018	8	60.85	0.006	2.51
8/30/2018	6	12.99	0.001	0.54
9/6/2018	8	13.00	0.001	0.54
9/6/2018	6	10.11	0.001	0.42

¹Adult LD₅₀ = 0.09 μg/bee

²Larval LD50 = 0.000087 μg/bee

4. Discussion

Public concern about the impact of pesticide exposure on honey bee health has been increasing in recent years. Previous studies have found insecticides in various matrices (e.g., leaves, stems, flowers, nectar, and pollen) of plants sold at large retailers (Brown et al., 2013, Kegley et al., 2016, Lentola et al., 2017). Public concern regarding such residues in pollinator-attractive plants has resulted in several retail chains committing to reducing neonicotinoid use on ornamental plants. This paper is the final installment in a series of publications using pollen as a vehicle to better understand how honey bees interact with ornamental plant nurseries. In previous work, we determined the floral sources of pollen collected by honey bees deployed on the grounds of the same ornamental plant nurseries sampled in this study (Sponsler et al., 2020, Stoner et al., 2019, Stoner et al., 2022). Here, we sought to characterize the risk posed to honey bee colonies located at two ornamental plant nurseries located in Connecticut, using honey bee-collected pollen as the testing matrix.

The present work follows up a similar study conducted at these nurseries three years prior (Stoner et al., 2019). The diversity of insecticides, fungicides, and herbicides detected at both nurseries increased between the 2015 and 2018 sampling seasons, but this could be at least partially attributed to methodological improvements (i.e., more pesticides tested for, greater sensitivity). All pesticides were applied according to standard practices at both nurseries. Compared to Nursery A, Nursery B applied a greater diversity of insecticides and fungicides, which could be a reflection of the specific plant assemblages cultivated there (e.g., more diversity). Surprisingly, this did not result in an increase in the average number of pesticides detected per pollen sample, although the maximum number of pesticides in a sample was greater at Nursery B (27) compared to Nursery A (19). Focusing on use patterns for specific pesticides, the organophosphate insecticide methamidophos (also a metabolite of acephate) and imidacloprid were the most frequently detected insecticides at both nurseries in 2015. Detections of these insecticides remained common in 2018; however, their median concentrations decreased between 43%–89%. Generally, the presence of neonicotinoids, particularly higher levels, within pollen samples decreased in 2018 (apart from acetamiprid), which could be a result of the consumer pressure against neonicotinoids. At the same time, however, the detection frequency and median concentration of the organophosphates chlorpyrifos and diazinon increased at both nurseries and Nursery B, respectively. It is not clear if these increases are an indirect effect of the effort to reduce neonicotinoid usage. Although the method for calculating risk differed between the two papers, no samples at either nursery exceeded the acute risk LOC when calculated using adult oral LD₅₀ values.

While our goal was to characterize pesticides in ornamental plants, some pesticides detected in pollen were not applied by their respective nursery, suggesting that the bees foraged on plants not cultivated by the nurseries. For example, honey bees at both nurseries collected pollen containing methoprene and phosmet residues, presumably from non-nursery plants. The use of these pesticides would not be unusual in the surrounding areas. Phosmet is used for pest control on orchard crops (e.g., apples, peaches), while methoprene is a general use pesticide (i.e., available to the public) with multiple uses including control of mosquito larvae. Detections of acetamiprid in pollen increased at both nurseries, although it was not applied at Nursery A during our study window. Acetamiprid is less acutely toxic to honey bees (Zhu et al., 2015) and produces fewer sublethal effects in bumble bees (Camp et al., 2020a, Camp et al., 2020b, Weitekamp et al., 2022) than other neonicotinoids; therefore, it is subject to fewer restrictions in Connecticut (An Act Concerning Pollinator Health, 2017), which may explain its increased use in recent years within the bees’ foraging area. In a separate analysis of the plant types contributing to the pollen we collected, Stoner et al. (2022) showed that while bees do collect pollen from nursery plants, they also foraged heavily from off-site plants.

Overall, 57% of insecticides in the pollen collected by bees at Nursery A were not applied by the nursery compared to only 16% at Nursery B. This could suggest that the area around Nursery A, which has more developed land and more land in cultivation in the primary foraging range around the hives (Stoner et al., 2022), generally has more pesticide use. Alternatively, because Nursery B applied a greater diversity of insecticides, this finding could simply reflect that most insecticides which were encountered off-site were also used at the nursery. Detecting pesticides not applied to focal crops in pollen collected by honey bees from colonies located in proximity to those crops is not unprecedented (Graham et al., 2022, McArt et al., 2017). To that point, McArt et al. (2017) found that pesticide risk was associated with pesticides that were not sprayed on focal crops and, for that reason, pesticide risk was likely attributed to contaminated wildflowers and other sources. Consequently, it would not be unusual to find these residues in the pollen of other plants in the area, and detections of specific pesticides can only be putatively linked to nursery applications.

Importantly, EPA’s acute risk LOC was exceeded in six pollen samples for honey bee larvae despite posing minimal or no risk to adult bees. This finding was driven by the presence of the organophosphorus insecticide diazinon, which was detected in 58% of samples collected at Nursery B (8.1 ppb, median; 60.9 ppb, max). Diazinon was also detected in honey bee-collected pollen samples harvested at the same nursery in 2015 (Stoner et al., 2019), albeit at much lower concentrations (i.e., 0.8 ppb, median; 2.1 ppb, max) and in a smaller proportion of the samples (i.e., 16%) indicating that use of diazinon in and/or around Nursery B has increased in recent years. Diazinon is a restricted use pesticide that has been not registered for residential use since 2004 in the United States, but in addition to its uses on ornamental plants, diazinon may be used on a variety of fruit and vegetable crops.

Although larvae consume less pollen than nurse bees, the larval risk will be greater if the ratio of adult to larval LD₅₀ values is greater than 2.67. In the case of diazinon, the ratio of adult to larval LD₅₀ values used in this study is greater than 1,000. To a lesser extent, the calculated risk was greater in larvae than in adults for two additional insecticides (acetamiprid and diflubenzuron), despite not exceeding the acute risk LOC. Importantly, the accuracy of these comparisons relies solely on the underlying accuracy of the adult and larval toxicity data. Most of the larval toxicity data used herein are derived from single publications, which frequently predate the implementation of standardized methodology (i.e., OECD 237). Furthermore, for many pesticides we evaluated larval LD₅₀ values were not readily available to allow for this comparison. A previous analysis of the U.S. EPA’s ECOTOXicology Knowledgebase found that it contains 1,681 studies of 167 pesticides evaluating adult toxicity and only 75 studies of 58 pesticides for larval toxicity (Farruggia et al., 2022). Although honey bee colonies are less sensitive to larval mortality than mortality of in-hive workers, sustained and significant larval mortality may still result in colony failure (Rumkee et al., 2015). These observations highlight the need for more data relating to the toxicity of pesticides to honey bee larvae.

Compared to recent studies evaluating honey bee-collected pollen in other settings, our findings at the ornamental plant nurseries are consistent with respect to estimated exposure. Demares et al. (2022) analyzed the pesticide content of honey bee-collected pollen harvested in urban and suburban environments across California, Florida, Michigan, and Texas. In that study, 94% of pollen samples contained less than four pesticides and, depending on the state the samples originated in, insecticides or herbicides were the most common type detected. Three samples collected in Michigan exceeded EPA’s acute risk LOC resulting from the pyrethroid insecticide deltamethrin. Another study conducted at blueberry farms in Michigan found that the average pollen sample contained 22 pesticides, and three samples containing the neonicotinoid clothianidin exceeded the acute risk LOC (Graham et al., 2022). In a survey conducted in Maine, pollen samples from non-agricultural sites averaged 1.3 pesticides, while pollen collected nearby agricultural sites averaged significantly more (either 2.5 or 6.3 pesticides per sample) (Drummond et al., 2018). Regardless of the site, fungicides followed by herbicides accounted for most of the pesticides detected, but insecticides contributed most to the estimated risk (calculated using the HQ approach). We similarly detected fungicides with the highest frequency, and the average number of pesticides in our samples fell within range of the comparable studies; however, none of our samples exceeded the acute risk LOC when calculated using adult oral LD₅₀ values. None of these studies estimated risk to larvae. Overall, the specific pesticides and relative abundance of insecticides, fungicides, and herbicides detected has significant heterogeneity between studies, as did the level of risk associated with the pollen samples.

Given that there are more than 900 pesticide active ingredients used in the United States (Martin et al., 2011) and that our analytical methods can only detect a subset of these chemistries, it was not possible for us to test for every possible pesticide used in and around the plant nurseries. Among the pesticides that fell outside the range of our analytical capabilities was abamectin. Abamectin, a member of the avermectin family of insecticides, was applied by both plant nurseries. McArt et al. (2017) detected abamectin residue in beebread, but, due to the lack of oral LD₅₀ toxicity data, the investigators could not estimate the potential risk posed to bees feeding on it. Abamectin is a mixture of avermectin B1a and avermectin B1b. Avermectin B1a has been detected in honey bee-collected pollen (Graham et al., 2022) and is highly toxic to honey bees (oral LD₅₀ value: 0.009 μg/bee) (Wislocki et al., 1989). These findings highlight the importance of assaying for abamectin residues in pollen and the need to account for this highly toxic pesticide when calculating RQs.

The pollen sampled herein contained mixtures of up to 27 pesticides (average 13-14), and when calculating the ARQ for a sample, we assumed an additive effect by simply adding the risk posed by each pesticide individually. Using this additive approach, we determined that adult bees in this study were at little to no acute risk. Risk assessment calculations are designed to evaluate worst-case exposure scenarios, so there is a reasonable margin of safety on this conclusion. However, this conclusion is also based on several assumptions. First, that the pollens only contained the pesticides we screened for (217 in total). Second, pesticides in a mixture can interact, potentially deviating from our assumption of additive effects (Hernandez et al., 2017). To this point, multiple studies have demonstrated that pesticide mixtures display a range of interactions (i.e., additive, synergistic, and antagonistic) on adult honey bees (Iwasa et al., 2004, Liu et al., 2017, Thompson et al., 2014, Wade et al., 2019, Zhu et al., 2017), and larval honey bees (Wade et al., 2019, Zhu et al., 2014). Wang et al. (2020) evaluated mixtures of the neonicotinoid acetamiprid with up to seven additional pesticides of differing types (i.e., one organophosphate, three pyrethroids, two avermectins, and one azole fungicide) and determined that synergistic interactions were more common than additive and antagonistic interactions (45%, 6%, and 9%, respectively). The same study also found that mixtures of 5-8 pesticides resulted in a synergistic interaction more frequently than mixtures of 2-3 pesticides. These studies indicate that the additive approach used herein to characterize these complex pesticide mixtures may underestimate the combined risk; however, a recent meta-analysis suggested that many of the synergistic interactions reported in the literature do not occur at environmentally-realistic concentrations (Belden, 2022). Moreover, we do not know if the residues detected in pollen translate proportionally to exposure (i.e., consumption). Bees store pollen in the form of beebread; therefore, at the time of consumption, it is possible that individuals are only exposed to a subset of the detected pesticides, which may further reduce the potential for synergistic interactions to occur. More research should be done to determine the effects of complex, field-realistic pesticide mixtures.

In this study, we calculated the likelihood of lethality from the consumption of contaminated pollen, which has several limitations. Although pollen is a widely used and easily collectable matrix to estimate dietary pesticide exposure in bees, pollen constitutes a small portion of a bee’s diet, and the primary dietary component, nectar, may also contribute significantly to the total dietary exposure. The distribution of systemic pesticides between pollen and nectar is influenced by several factors (e.g., plant type, physiochemical properties of the pesticide, environmental conditions), which make it difficult to predict nectar contamination from the pesticide concentrations measured in pollen alone (Gierer et al., 2019). Additional avenues of exposure, including contact exposure, either through direct spray or plant surface residues, and oral exposure through water sources (e.g., puddles, guttation fluid) were also not accounted for (Gradish et al., 2019, Samson-Robert et al., 2014). Finally, while we focused on individual lethality risk, the lower levels of pesticide contamination found herein may cause sublethal effects (e.g., mobility, learning, memory, orientation, thermo regulation, foraging performance, and homing) which can impair colony function and indirectly contribute to colony decline (Desneux et al., 2007, Gill et al., 2012, Schneider et al., 2012, Williamson and Wright, 2013).

Conclusions

The present investigation aimed to develop a better understanding of how honey bees interact with ornamental nursery crops and the risks associated with pesticide exposure in this understudied environment. We found that honey bees collected pollen from nursery and non-nursery plants, exposing the bees and their brood to a broad range of pesticides. Our data suggest that acute risk to adult honey bees from pesticide residues in pollen collected in and around the ornamental plant nurseries is generally low, at least at the locations we sampled. Importantly, however, the risk to honey bee larvae was higher as a result of the insecticide diazinon in some of the pollen samples. Compared to adult RQs, larval RQs could only be calculated on a subset of the detected insecticides, due to the relative inaccessibility of larval toxicity data. This highlights the need for increased availability of larval toxicity data and the importance of considering all life stages when estimating potential risk to honey bee colonies from pesticide exposure. These findings may inform risk mitigation strategies at ornamental plant nurseries, but additional studies are needed to determine if the findings herein are generalizable to other ornamental plant nurseries.

Abbreviations:

AChE

acetylcholinesterase

ARQ

aggregate risk quotient

GC

gas chromatography

IRAC

Insecticide Resistance Action Committee

LC

liquid chromatography

LD₅₀

lethal dose which causes the death of 50% of a group of bees

LOC

level of concern

LOD

limits of detection

MS

mass spectrophotometer

NCI

negative ion chemical ionization

nAchR

nicotinic acetylcholine receptor

QuEChERS

quick, easy, cheap, effective, rugged, and safe

RQs

risk quotients

SD

standard deviation