Impacts of neonicotinoids on the bumble bees Bombus terrestris and Bombus impatiens examined through an AOP framework lens

Abstract

Bumble bees (Bombus sp) are important pollinators for agricultural systems and natural landscapes and have faced population declines globally in recent decades. Neonicotinoid pesticides have been implicated as one of the reasons for the population reductions in bumble bees and other pollinators due to their widespread use, specificity to the invertebrate nervous system, and toxicity to bees. Adverse outcome pathways (AOPs) are used to describe the mechanism of action of a toxicant through sequential levels of biological organization to understand the key events that occur for a given adverse outcome. Here, we used the AOP framework to organize and present the current literature available for the impacts of neonicotinoids on bumble bees. This review focuses on Bombus terrestris and B. impatiens, the two most commonly studied bumble bees due to their commercial availability. The present review does not seek to describe an AOP for the molecular initiating event shared by neonicotinoids, but rather aims to summarize the present literature and highlight data gaps for the Bombus research community to address. Overall, we highlight a great need for additional studies, especially those examining cellular responses and organ responses in bumble bees exposed to neonicotinoids.

INTRODUCTION

Insects are necessary for plant pollination in both agricultural settings and natural landscapes (Potts et al. 2010). In recent decades, pollinator populations have experienced declines, stemming from several interacting factors including pesticide use (Blacquière et al. 2012), habitat loss and fragmentation (Brown and Paxton 2009), pathogens (Doublet et al. 2015), climate change (Soroye et al. 2020) and parasites (Goulson et al. 2015). Honey bees (Apis mellifera) are a prominent insect pollinator due to their widespread use commercially and their economic importance globally (Gallai et al. 2009; Losey and Vaughan 2006). Like other bees, honey bee population declines have been documented widely (Aizen and Harder 2009; Ellis et al. 2010).

Not only are honey bees the most visible and well-understood pollinator, they have also been used extensively in toxicity testing. Honey bees are the representative insect pollinator for ecological risk assessment and pesticide registration decisions within the United States (EPA 2004), and are one of the pollinators used for European Union guideline decisions (EFSA 2013). In the United States, the suitability of honey bees as a surrogate species for risk assessment has been questioned since they are not native, and almost exclusively exists as managed populations, and many aspects of their life history are much different from native pollinators like the bumble bee (Gradish et al. 2019; Stoner 2016). For instance, honey bees are perennial and overwinter as an intact colony, while bumble bee nests die out in the fall and newly mated queens are the sole overwintering caste. Bumble bee queens exiting hibernation are then responsible for establishing new nests in the spring and regenerating the active worker and drone populations each year (Goulson 2010). Additionally, their documented differences in patterns of food consumption, colony sizes, development times, suitable nest sites, and physiology that dictate variable sensitivity to stressors (e.g., pesticide exposure), further blur the relevance of extrapolation (Gradish et al. 2019).

Bumble bees (Bombus sp) are critical for commercial and non-commercial pollination and also face population declines (Cameron et al. 2010; Colla and Packer 2008; Goulson et al. 2008; Grixti et al. 2009; Losey and Vaughan 2006; Lozier et al. 2011). Bumble bees are used to pollinate crops such as blueberries (Drummond 2012; Stubbs and Drummond 2001), tomatoes (Morandin et al. 2001; Velthuis and van Doorn 2006) and peppers (Shipp et al. 1994). The United States has listed several species of bees as endangered or threatened, including one bumble bee (Gorman 2017). Despite their clear importance to agricultural and ecosystem function, bumble bees are relatively understudied as compared to the non-native honey bee.

One of the major stressors threatening bumble bee populations are neonicotinoid pesticides. Since their introduction to the market in the 1990s, neonicotinoids have become widely used in agricultural settings and their use has steadily increased globally (Simon-Delso et al. 2015). As of 2015, neonicotinoids comprised nearly 25% of the global market share of pesticides for agricultural use (Bass et al. 2015). While effective at controlling a variety of crop pests, neonicotinoids also have the ability to adversely impact pollinators (Blacquière et al. 2012; Goulson et al. 2015). They have been implicated as one of the reasons for the widespread population declines of bees observed globally (Cameron et al. 2010; Goulson et al. 2015; Lozier et al. 2011; Potts et al. 2010). Due to concerns about their toxicity to pollinators, the European Union has taken steps to restrict the use of several neonicotinoids (EU 2013).

In order to organize and review the available data on neonicotinoid pesticides and bumble bees, the framework of the adverse outcome pathway (AOP) is useful. AOPs provide a structured means by which relevant information relating to an adverse outcome can be compiled and organized in order to identify data gaps and guide future research efforts. Introduced by Ankley and colleagues (2010), AOPs describe the complete mechanism of action of a toxicant from a molecular initiating event to population-level effects. The pathway outlines sequential steps that occur within a toxicological response at different levels of biological organization. In the case of bees and other wildlife, an AOP can support ecological risk assessment and inform test guidelines put forth by the Organization for Economic Co-operation and Development (OECD) and other regulatory bodies (OECD 2018).

The AOP framework will be briefly described here (for a more detailed description, see (Ankley et al. 2010). The first section of the AOP is the toxicity pathway, which includes the toxicant chemical properties (e.g., solubility, metabolites), the relevant macro-molecular interactions (e.g., receptor ligand interactions, DNA binding), and cellular responses (e.g., gene activation, protein production, altered signaling). The anchoring event within the toxicity pathway are the macro-molecular interactions that initiate the subsequent responses (i.e., molecular initiating event). The second part of the pathway includes organ responses (e.g., altered tissues development or function, altered physiology), organism responses (e.g., lethality, impaired development, impaired reproduction), and finally population responses (e.g., population structure, recruitment). In the second portion of the pathway, the anchoring responses are organism and population responses (Figure 1).

Figure 1. Diagram of the adverse outcome pathway (AOP) framework.

The AOP begins with the molecular initiating event (MIE) when a chemical interacts with the biological target within the organism. Chemical binding triggers a series of sequential intermediate events that culminate in an adverse outcome that is relevant to risk assessment. The AOP is anchored by the MIE and adverse outcome at the organism-or population-level.

AOPs have previously been described for neonicotinoid-like stressors impacting honey bees for the adverse outcome of colony death/failure (LaLone et al. 2017). However, the purpose of the current review is not to describe a complete Bombus sp AOP, but rather to use the AOP framework as an organizational tool by which to present what is currently known about the impacts of neonicotinoid pesticides on two species of bumble bee, Bombus terrestris and B. impatiens. We will focus on the effects of neonicotinoid exposure rather than concentrations that elicit adverse effects or relevancy to concentrations found in the environment. We highlight opportunities for advancing our understanding of Bombus sensitivity to this common class of insecticides, as well as key areas that lack adequate research and methodology. Our hope is that Bombus researchers will use this review to guide future research efforts so that a robust a Bombus AOP can be developed for neonicotinoids.

METHODS

We conducted a scoping review of literature. Scoping reviews provide a mechanism to assemble all of the available literature on a particular topic, thereby revealing important knowledge gaps, trends and stimulating additional research (Munn et al. 2018). The literature search was conducted from 1980, encompassing a period before neonicotinoids were first introduced, through December of 2019, to ensure all relevant literature was captured. Web of Science, ScienceDirect, Proquest Agricultural, and Environmental Science databases were searched with the search terms “bumble bee” or “bumblebee” or “Bombus” and “neonicotinoid” or “imidacloprid” or “thiacloprid” or “thiamethoxam” or “clothianidin” or “acetamiprid” or “nitenpyram” or “dinotefuran” or “nithiazine”. Of the 182 references that were generated, duplicates, conference abstracts, and non-peer reviewed publications were removed to ensure rigor of the research included. Then, references that were not relevant (e.g., did not generate bumble bee-specific data), reviews, and modeling-only papers were removed. The final list of references included 78 citations, with 56 references focusing on B. terrestris, and 22 references pertaining to B. impatiens that conducted primary research (Supplementary Table 1). We did not find papers for dinotefuran or nithiazine that fit our criteria for this review. The publications included in the review by year are shown in Figure 2, and their distribution within the AOP framework is shown in Figure 3.

Figure 2: Publications included in this review by year.

Publications that met the criteria for this review are depicted here by year. Notably, the majority of relevant publications have been published since 2015.

Figure 3. Literature search results organized by biological responses.

Headings indicate the biological level of organization in the AOP. Boxes below contain biological responses documented to occur following neonicotinoid exposure, but not all perturbations are known to lead to an adverse outcome. Numbers in parenthesis indicate the number of peer-reviewed publications available for each biological response discussed in this systematic review.

FIRST SECTION OF AOP

Toxicant properties:

Neonicotinoid pesticide development began in the 1970s and 1980s by Shell and Bayer Crop-Science to develop pesticides with lower mammalian toxicity and without the problems of resistance that had developed with other common insecticides (e.g., organophosphates and pyrethroids) (Denholm et al. 1998; Jeschke and Nauen 2008; Tomizawa and Casida 2003). All neonicotinoid chemicals structurally resemble nicotine, a naturally occurring plant alkaloid (Tomizawa and Casida 2005). The first neonicotinoid to be brought to market was imidacloprid in 1991 by Bayer Crop-Science (Jeschke and Nauen 2008).

Neonicotinoid compounds are commonly categorized by either their functional groups (cyano-group containing or nitro-group containing) or their structure (ring versus open) (Table 1). The cyano-group includes thiacloprid, acetamiprid, and the less commonly used dinotefuran, while the nitro-group chemicals include imidacloprid, clothianidin, thiamethoxam, and the less commonly used nitenpryam and nithiazine (Casida 2011; Jeschke and Nauen 2008) (Table 1). To see their structure and similarity to nicotinoids, see work by Tomizawa and colleagues (2005).

Table 1.

Physical attributes of the commonly used neonicotinoids

Neonicotinoid	Class	Structurea,b	Water solubility (g/L at 20°C)c	log KOW (at 25°C)c
Imidacloprid	Nitro-containing	Ring	0.61	0.57
Clothianidin	Nitro-containing	Open	0.327	0.7
Nitenpryam	Nitro-containing	Open	840	−0.64
Thiamethoxam	Nitro-containing	Ring	4.1	−0.13
Acetamiprid	Cyano-containing	Open	4.2	0.8
Thiacloprid	Cyano-containing	Ring	0.185	1.26
Dinotefuran	Cyano-containing	Open	54.3	−0.64

^aJeschke and Nauen 2008,

^b Casida 2011

^c Jeschke et al 2011

High water solubility and low octanol/water partition coefficient (log K_OW) has led to the massive popularity and efficacy of these chemicals (Table 1). This property allows for systemic transport of the pesticide into soluble plant compartments and has enabled the successful deterrence of a wide range of herbivorous insects, spanning many orders of insects including coleoptera, lepidoptera, diptera, homoptera, and hemiptera (Jeschke et al 2011). As a result of their versatility and broad reach as insecticides, neonicotinoids have been approved for use on hundreds of crops (Jeschke et al. 2011). The water solubility of these pesticides has facilitated the method of seed coating crops, which has greatly decreased the amount of pesticide needed per plant. Integration into the growing plant via plant tissues and fluids also creates a means of exposure for non-target insects including pollinators, since the pesticides also move into pollen and nectar compartments of the plants. Some neonicotinoids are applied via foliar spray (e.g., acetamiprid), presenting another means of pollen and nectar contamination (Elbert et al. 2008). For the purposes of this review, we will focus on the 5 commonly used neonicotinoids: imidacloprid, clothianidin, thiamethoxam, acetamiprid and thiacloprid.

Macro-molecular interactions (molecular initiating event)

The macro-molecular interaction element of the AOP outlines the molecular initiating event (MIE) within the toxicity pathway (Ankley et al. 2010). In the case of neonicotinoids, the MIE that initiates the toxic response in insects is thought to be well established. Neonicotinoids interact with the nicotinic acetylcholine receptor (nAchR), an ionotropic post-synaptic receptor in the nervous system of insects. The nAchR is a membrane spanning, five subunit receptor of heterodimeric composition (Tomizawa and Casida 2001). These receptors are located at neuromuscular junctions, as well as in the central nervous system of insects (Tomizawa and Casida 2001), however, the neural nAChR is thought to be more abundant (Thany et al. 2010). The muscular and neuronal versions of the receptor differ in their properties as well as their subunit composition. This MIE is the working MIE for the honey bee colony death/failure AOP (LaLone et al 2017).

Under normal conditions, the neurotransmitter acetylcholine activates the nAchR, resulting in an open ion channel that allows ion flow through the channel. The primary ions that move through the channel are sodium, potassium, and calcium. Calcium is responsible for a variety of post-synaptic cellular events including the Ca²⁺ /calmodulin response pathway that leads to changes in gene expression within the neuron (Wiegert and Bading 2011). Then, acetylcholinesterase rapidly deactivates the acetylcholine molecule such that stimulation of the receptor ceases, the ion channel closes, and the acetylcholine can be recycled and used again by the pre-synaptic cell. Neonicotinoids activate the receptor, but cannot be broken down by acetylcholinesterase, allowing for continual stimulation of the post-synaptic cell. This can lead to overstimulation and desensitization of the post-synaptic neuron. At sublethal exposures, these effects may result in a variety of downstream effects including altered signaling and transcriptional changes, while higher exposures can lead to neurological impairments including hyperactivity, paralysis, and death.

The MIE of nAchR activation is widely accepted for honey bees, and there is supporting evidence within the bumble bee literature. Importantly, researchers have shown that neonicotinoids (Table 2; imidacloprid and clothianidin) reach the brain of bumble bees after exposure (Moffat et al. 2015). Further, studies with imidacloprid and clothianidin have demonstrated that exposing neurons (Kenyon cells) to these chemicals results in calcium ion influx, providing evidence that these compounds act at the target receptor (Moffat et al. 2016). As supporting genomic evidence, both B. impatiens and B. terrestris genomes contain genes for 11 nAchR subunits (Sadd et al. 2015). Finally, it is important to note that there is limited evidence that neonicotinoids may be able to elicit effects through a separate molecular initiating event, since Bebane and colleagues exposed B. terrestris to imidacloprid and found altered DNA methylation (2019).

Table 2.

Molecular Initiating events

Neonicotinoid	Caste	Effect	Citations
Imidacloprid	Workers	Receptor-ligand interactions	Moffat et al 2015, Moffat et al 2016
		DNA	Bebane et al 2019
Clothianidin	Workers	Receptor-ligand interactions	Moffat et al 2015, Moffat et al 2016
		DNA	Bebane et al 2019

Cellular responses

Within the AOP framework, cellular responses include a variety of changes at the cellular level such as gene activation, protein production, and altered signaling. Within the honey bee draft AOP, nAchR desensitization, mitochondrial dysfunction, and altered calcium/calmodulin signal transduction are the three cellular responses thought to be the most likely events leading to colony death/failure (LaLone et al 2017). However, in this review, we included all responses that pertained to cellular function in order to highlight all available research to date. This data is compiled and organized in Table 3 and topics of interest will be discussed below.

Table 3.

Cellular Responses

Neonicotinoid	Caste	Effect	Citations
Imidacloprid	Workers	Mitochondrial function	Moffat et al 201, Colgan et al 2019, Powner et al 2016
		Gene activation	Colgan et al 2019, Simmons and Angelini 2017
		Protein production	Erban et al 2019
		Altered signaling	Czerwinksi and Sadd 2017, Wanderdorff et al 2018
Clothianidin	Workers	Mitochondrial function	Moffat et al 2015, Colgan et al 2019
		Gene activation	Colgan et al 2019, Mobley and Gegear 2018, Samson-Robert et al 2015
		Protein production
		Altered signaling
	Drones	Gene activation	Mobley and Gegear 2018
Thiamethoxam	Workers	Gene activation	Samson-Robert et al 2015

Mitochondrial function.

Evidence from bumble bees suggests that exposure to neonicotinoids impacts mitochondrial function in multiple ways. In B. terrestris, clothianidin exposure resulted in rapid depolarization of neuronal mitochondria in vitro, while imidacloprid did not elicit as robust a response (Moffat et al. 2015). This study provides a plausible connection between nAchR receptor activation and subsequent mitochondrial effects. There is also evidence that imidacloprid and clothianidin alter gene expression relating to mitochondrial function in B. terrestris, as RNA-seq experiments revealed that alanine-glyoxylate aminotransferase and phosphoenolpyruvate carboxylase were both differentially expressed with clothianidin exposure (Colgan et al. 2019). Additionally, it is known that the cytochrome C oxidases in the complex IV of mitochondria can absorb 670 nm light (Karu 2008), and a study with B. terrestris has shown that exposure to 670nm light can improve the survival after imidacloprid exposure (Powner et al. 2016). Taken together, there are several lines of evidence supporting mitochondrial dysfunction after neonicotinoid exposure.

Gene expression.

In the same work described above by Colgan and colleagues, comparisons between worker and queen gene expression were made. Notably, they found that more transcripts were differentially expressed after exposure in workers (55) than in queens (17), and only one common gene was differentially expressed in both castes (ejaculatory bulb-specific protein 3) (Colgan et al. 2019). Since queens are responsible for reestablishing the population each spring, their sensitivity is particularly critical for predicting the impacts of neonicotinoids. Another gene expression study that highlighted caste differences found significant differences in response to clothianidin between worker and drone B. impatiens. At the concentrations assessed, exposure resulted in increased expression of 11 detoxification genes in workers as compared to drones, and 9 detoxification genes with increased expression in drones as compared to workers (Mobley and Gegear 2018). While worker efficiency is essential for nest success, healthy males are required for successful bumble bee queen mating and overwintering (Baer and Schmid-Hempel 2005).

Additional studies with B. impatiens have examined gene expression alterations in workers. Exposure to imidacloprid significantly increased relative mRNA expression of three antimicrobial peptide transcripts, abaecin, apidaecin, and hymenoptaecin (Simmons and Angelini 2017), suggesting that pesticide exposure may increase antimicrobial peptide levels within the exposed individuals. Further, work examining the nAchR and downstream effects from this receptor found that neonicotinoids also disrupt elements of the acetylcholine transmission pathway, including acetylcholinesterase (AchE) activity. B. impatiens subject to field-relevant exposures of clothianidin and thiamethoxam had significantly increased AchE relative expression as compared to control bees (Samson-Robert et al. 2015). The impact of neonicotinoids on AchE expression may be a compensatory response from the dysregulation of normal transmitter function as the result of neonicotinoids acting on the nAchR.

Immune function.

Several studies have documented the effects of neonicotinoid exposure on bumble bee immune system. Prophenoloxidase and the activated phenoloxidase enzyme is responsible for the melanization response that occurs after the insect cuticle has been punctured or perforated (Söderhäll and Cerenius 1998). Exposure to imidacloprid resulted in decreased phenoloxidase activity in B. impatiens (Czerwinksi and Sadd 2017). In a dual imidacloprid and immune challenge assay, Czerwinski and colleagues also found that imidacloprid dampened the immune response, as measured by antimicrobial peptide activity of bumble bee hemolymph (2017). Additional work by Walderdorff and colleagues measured other immune endpoints in B. terrestris after imidacloprid exposure and they found that phagocytosis by hemocytes was reduced, and further reduced by immune challenge (2018). Also, hydrogen peroxide generation was significantly decreased, while nitric oxide showed mixed effects with the combination of imidacloprid and immune challenge (Walderdorff et al. 2018).

Protein production.

A proteomics analysis of B. terrestris brains revealed that imidacloprid exposure downregulated the mevalonate and terpenoid backbone biosynthesis pathway, which are responsible for fatty acid synthesis and the production of juvenile hormones (Erban et al. 2019). Juvenile hormones regulate a variety of insect processes including development, molting, metamorphosis, and reproduction (Bellés et al. 2005).

Other cellular responses.

Research conducted on B. terrestris neurons found that neonicotinoids were unable to activate gustatory neurons, nor did they inhibit sucrose-sensitive neurons (Kessler et al. 2015). This suggests that bumble bees are unable to detect neonicotinoids in their food source, and do not interpret them as sweet. This information helps explain why several studies have determined that bumble bees do not avoid neonicotinoids when present in a food source (Gels et al. 2002; Havstad et al. 2019; Larson et al. 2013).

Taken together, there is evidence that bumble bees may face some of the same cellular-level impacts as a result of neonicotinoid exposure as honey bees. However, there are additional lines of evidence that suggest changes in gene expression impacting immune responses and hormone production may exert strong effects on individual and population health outcomes with neonicotinoid exposure.

SECOND SECTION OF AOP

Organ responses

The organ response key events within an AOP refer to endpoints that include altered physiology, disrupted homeostasis, altered tissue development, and altered tissue function. Often, these key events are more difficult to measure with a method that is accessible, reliable, reproducible, and accepted within the scientific community, which are the method criteria proposed by the OECD AOP guidelines (OECD 2018). For instance, tissue histology that requires a specialist may not meet the accessibility criteria. The general difficulty in assessing organ-level responses is evident within the bumble bee literature and demonstrates a need for more simplified or accessible methodology. This data is compiled and organized in Table 4 and topics of interest will be discussed below.

Table 4.

Organ responses

Neonicotinoid	Caste	Effect	Citations
Imidacloprid	Workers	Altered tissue development	Laycock et al 2012, Wilson et al 2013
		Altered physiology	Potts et al 2018
Thiamethoxam	Queens	Altered tissue development	Baron et al 2017a

Tissue development.

Two studies to date have examined the impact of neonicotinoids on ovary development. Baron and colleagues (2017b) exposed B. terrestris queens to thiamethoxam and observed a significant reduction in terminal oocyte length. Oocyte length is a measure of energetic investment in offspring; however, it is difficult to parse whether thiamethoxam was directly or indirectly responsible for reductions in oocyte size since feeding rates were also impacted by exposure. Further, it is possible that the detoxification processes associated with thiamethoxam were partially responsible for the reduced investment in oocytes. In another study examining neonicotinoid impacts on B. terrestris ovary development, Laycock and colleagues (2012) found no change in mean oocyte size at field realistic concentrations of imidacloprid, however, high concentration reduced mean size.

Altered tissue development has been observed with in vitro studies examining a subclass of bumble bee neurons called Kenyon cells (Wilson et al. 2013). B. impatiens neuronal cultures were exposed to imidacloprid and depending on the age of the bee from which the cells were derived, exposure inhibited neurite growth. Older worker Kenyon cells were more sensitive to imidacloprid than newly eclosed bees, highlighting the potential for age-specific effects of neonicotinoid exposure.

Physiology.

There is also evidence of altered physiology in bumble bees after exposure to neonicotinoids. B. terrestris exposed to imidacloprid displayed impaired thermal physiology (Potts et al. 2018). Bees exposed to imidacloprid and thiamethoxam displayed dose-dependent thermogenesis impairments, resulting in slower rewarming times after chilling. Since thermogenesis is critical for brood incubation, this effect could result in developmental delays or mortality among developing brood.

Organism Responses

The body of literature on organism responses, part of the second anchor within the AOP, is much larger and more thorough than cellular or organ responses. As expected, there are many studies that evaluate the acute and chronic toxicity of neonicotinoids and report mortality. Also, within the organism response section are sublethal effects, which include reductions in food consumption and decreased body weight as well as behavioral changes, such as impaired foraging and homing, altered memory and learning, altered activity levels, altered food preference, and delayed nest initiation and egg laying. This data is compiled and organized in Table 5 and topics of interest will be discussed below.

Table 5.

Organism responses.

Neonicotinoid	Caste	Effect	Citations
Imidacloprid	Workers	Lethality	(Tasei et al. 2000, Gels et al. 2002, Gradish et al. 2010, Mommaerts et al. 2010, Moffat et al. 2016, Switzer and Combes 2016, Baines et al. 2017)
	Queens	Lethality	(Scholer and Krischik 2014, Leza et al. 2018, Wu-Smart and Spivak 2018)
	Workers	Sublethal effects	(Tasei et al. 2001, Gradish et al. 2010, Cresswell et al. 2012, Laycock et al. 2012, Cresswell et al. 2014, Feltham et al. 2014, Gill and Raine 2014, Kessler et al. 2015, Thompson et al. 2015, Switzer and Combes 2016, Crall et al. 2018, Phelps et al. 2018, Kenna et al. 2019, Muth et al. 2019, Muth and Leonard 2019)
	Queens	Sublethal effects	(Leza et al. 2018, Wu-Smart and Spivak 2018)
Clothianidin	Workers	Lethality	(Moffat et al. 2016, Baines et al. 2017, Heard et al. 2017, Sgolastra et al. 2017, Wintermantel et al. 2018)
	Queens	Lethality	(Scholer and Krischik 2014, Fauser et al. 2017, Mobley and Gegear 2018, Wintermantel et al. 2018)
	Drones	Lethality	(Mobley and Gegear 2018)
	Workers	Sublethal effects	(Fauser-Misslin et al. 2014, Scholer and Krischik 2014, Kessler et al. 2015, Thompson et al. 2015, Wintermantel et al. 2018, Dietzsch et al. 2019)
	Queens	Sublethal effects	(Fauser et al. 2017)
Thiamethoxam	Workers	Lethality	(Mommaerts et al. 2010, Laycock et al. 2014, Moffat et al. 2016, Baines et al. 2017, Raimets et al. 2018, Iverson et al. 2019)
	Queens	Lethality	(Fauser et al. 2017)
	Workers	Sublethal effects	(Elston et al. 2013, Fauser-Misslin et al. 2014, Laycock et al. 2014, Kessler et al. 2015, Stanley et al. 2015, Stanley and Raine 2016, Stanley et al. 2016, Whitehorn et al. 2017)
	Queens	Sublethal effects	(Baron et al. 2017b, Baron et al. 2017a, Fauser et al. 2017)
Thiacloprid	Workers	Lethality	(Mommaerts et al. 2010)
Acetamiprid	Workers	Lethality	(Baines et al. 2017)

Lethality.

Individual mortality data is commonly assessed for pesticides and is easily linked to population-level effects (e.g., population structure and recruitment). This type of data was the most comprehensive of any endpoint, with data present for all five major neonicotinoids. Taken together, it is clear that these chemicals can cause mortality at sufficient concentrations or doses. Most studies focused on workers, few were on queens, and very little work was identified for drones.

Sublethal effects.

Behavioral data are more difficult to link in a definitive way to population-level endpoints or whole nest health. However, there is a substantial amount of data showing that neonicotinoids impact a variety of behaviors. Behaviors relating to food acquisition, including foraging behavior (Feltham et al. 2014; Gill and Raine 2014; Kenna et al. 2019; Lämsä et al. 2018; Muth and Leonard 2019; Stanley et al. 2016; Tasei et al. 2001) and activity levels (Crall et al. 2018; Cresswell et al. 2014; Dietzsch et al. 2019; Leza et al. 2018) are perturbed by exposure to neonicotinoids. Reduced foraging may impact population growth by impacting gyne (unmated queen) production or drone production, and ultimately the number of overwintering queens and their nutritional status. Additionally, in the spring, when queens are establishing nests, queen foraging behavior is critical for successful nest initiation (Goulson 2010).

Other notable impacts on behavior were related to changes in learning and memory, and included changes in olfactory learning (Muth et al. 2019; Stanley et al. 2015), altered food preference (Phelps et al. 2018; Stanley and Raine 2016), and spatial working memory (Samuelson et al. 2016). Ultimately, many of the learning and memory changes observed with neonicotinoid exposure will likely impact foraging and food acquisition since bumble bees learn to locate valuable food resources and return to forage repeatedly. Both food location and foraging ability are important for the success of bumble bee nests to guarantee that nutritional limitations do not occur.

A sublethal effect observed in many studies was decreased food consumption (Cresswell et al. 2012; Cresswell et al. 2014; Fauser-Misslin et al. 2014; Gradish et al. 2010; Laycock et al. 2014; Laycock et al. 2012; Thompson et al. 2015). This effect may lead to some of the same resource limitation issues detailed above regarding foraging behavior or may be a symptom of another perturbation within the nest. For instance, reduced food consumption has been observed in queenless bumble bee microcolony assessments where larval morality occurs, because the nutritional needs of the nest fall commensurately with reduced brood survival (Camp et al. 2019).

Caste-specific effects.

Most research conducted on bumble bees focuses on workers, however of special importance are the effects observed when queens are exposed to neonicotinoids. Queens are not only the sole layers of fertilized eggs, which result in workers, but they are also responsible for reestablishing the population each spring after hibernation (Goulson 2010). Wu-Smart and colleagues (2018) observed a variety of significant impacts on queens after exposure to imidacloprid including mortality, increased time to egg-laying, decreased body weight, and ultimately reduced nest weight. Similar effects were observed in another study assessing imidacloprid (Leza et al. 2018). Notably, exposure to clothianidin has been found to have negative impacts on queen survival while overwintering and resulted in weight loss during hibernation (Fauser et al. 2017). Thiamethoxam exposure prior to overwintering had a similar effect, with reductions in the number of egg-laying queens as well as delays in colony initiation by the queens that did lay eggs (Baron et al. 2017a). As mentioned previously, the success of the bumble bee queen (i.e., successful overwintering and nest establishment in the spring) is one of the most critical aspects of bumble bee life history in predicting population success.

Drones are also important for population maintenance. Males can be produced by workers in the case of queen death, but are typically produced by the queen within a queenright colony (Alaux et al. 2004). Gynes mate prior to overwintering as a queen, and studies indicate that sperm can influence overwintering survival (Baer and Schmid-Hempel 2005). This suggests that drone health is important for population maintenance, however it is unclear what aspect of drone health is most critical. Few studies have looked at the effects of neonicotinoids on drones, in part due to the misconception that drones are only useful for mating. Other reasons for a lack of drone data may be because they are only present in the nest at later stages of nest development, and they leave the nest after a short time in order to find mates (Goulson 2010). The main endpoints that have been evaluated with drones pertain to their production within a nest, for instance within queenless microcolony studies, rather than direct impacts of pesticides on their health.

Population responses

Population responses are the final response within the AOP and encompass aspects of population dynamics that will affect population maintenance and sustainability. For instance, population structure is one response that may have notable impacts on overall population stability and encompasses sex ratios, age distribution, and within the context of social insects, caste allocation. Population recruitment is another measure that refers to the addition of new individuals in order to maintain or grow a population given the mortality rate. Here, recruitment included alternations in nest weight, brood production, and offspring production. Low recruitment will inevitably lead to population declines over time. And finally, other responses such as extinction are also captured within population responses, however, no studies in the present review captured this response.

Within the bumble bee literature, a fair amount of work has been conducted using experimental designs that evaluate population-relevant endpoints (Table 6). Experiments using field, semi-field, and laboratory assessments using queenright colonies frequently examine population-relevant questions. Within the honey bee AOPs developed by Lalone and colleagues (2017), the defined adverse outcome was colony failure rather than total population extinction. For bumble bees, nest failure would also be an appropriate adverse outcome. Nest failure could occur if any of the population-relevant measures sufficiently disrupted whole-nest food acquisition, egg-laying, brood-tending, or queen production. For instance, if brood production is low and young workers are not replacing older workers within the nest, resource foraging may decline such that additional brood would perish and population recruitment could drop perilously low.

Table 6.

Population responses.

Neonicotinoid	Caste/measure	Population Effect	Citation
Imidacloprid	Workers	Structure	(Gels et al. 2002, Gill et al. 2012)
	Queens	Structure	(Whitehorn et al. 2012)
	Drones	Structure	(Scholer and Krischik 2014)
	Whole nest	Recruitment	(Tasei et al. 2000, Tasei et al. 2001, Mommaerts et al. 2010, Whitehorn et al. 2012, Laycock and Cresswell 2013, Scholer and Krischik 2014, Moffat et al. 2016, Leza et al. 2018, Wu-Smart and Spivak 2018)
Clothianidin	Workers	Structure	(Cutler and Scott-Dupree 2014, Fauser-Misslin et al. 2014, Larson et al. 2014, Rundlöf et al. 2015, Arce et al. 2017)
	Queens/gynes	Structure	(Larson et al. 2013, Fauser-Misslin et al. 2014, Rundlöf et al. 2015, Arce et al. 2017, Wintermantel et al. 2018)
	Drones	Structure	(Fauser-Misslin et al. 2014, Scholer and Krischik 2014, Rundlöf et al. 2015, Arce et al. 2017, Woodcock et al. 2017, Wintermantel et al. 2018)
	Whole nest	Recruitment	(Scholer and Krischik 2014, Rundlöf et al. 2015, Moffat et al. 2016)
Thiamethoxam	Drones	Structure	(Fauser-Misslin et al. 2014)
	Queens/gynes	Structure	(Fauser-Misslin et al. 2014, Moffat et al. 2016, Baron et al. 2017b)
	Whole nest	Recruitment	(Mommaerts et al. 2010, Elston et al. 2013, Laycock et al. 2014, Moffat et al. 2016, Baron et al. 2017b)
Thiacloprid	Drones	Structure	(Ellis et al. 2017)
	Queens/gynes	Structure	(Ellis et al. 2017)
	Whole nest	Recruitment	(Ellis et al. 2017, Havstad et al. 2019)

^*Fauser-Misslin et al 2014 was a mixtures exposure experiment with thiamethoxam and clothianidin

Studies that utilized the microcolony model were included as population-relevant if they measured nest weight, brood production, or offspring production. The microcolony model leverages the tendency of workers in isolation to establish a false queen and lay unfertilized eggs which will develop into drones. Microcolonies recapitulate many aspects of full-sized queenright colonies such as worker behavior, nest initiation, brood production and brood tending, as well as offspring production (i.e., drones) (Klinger et al. 2019). While useful for a variety of experimental applications, microcolonies are especially useful for brood development assays since they are more easily replicated than full size queenright colony assessments. Further, they require fewer resources and are easily conducted in the laboratory therefore minimizing the variables typically associated with queenright colony assays (e.g., weather conditions).

Structure.

Many studies evaluated the impacts of neonicotinoids on population structure. Imidacloprid exposure to colonies resulted in reduced worker (Gels et al. 2002; Gill et al. 2012), queen (Whitehorn et al. 2012), and drone (Scholer and Krischik 2014) counts. Similar work conducted with clothianidin found that exposure impacted the population structure within colonies. Workers (Arce et al. 2017; Cutler and Scott-Dupree 2014; Fauser-Misslin et al. 2014; Larson et al. 2014; Rundlöf et al. 2015), queens (Arce et al. 2017; Fauser-Misslin et al. 2014; Larson et al. 2013; Rundlöf et al. 2015; Wintermantel et al. 2018), and drones (Arce et al. 2017; Fauser-Misslin et al. 2014; Rundlöf et al. 2015; Scholer and Krischik 2014; Wintermantel et al. 2018; Woodcock et al. 2017) all had reduced numbers after exposure. Additionally, thiamethoxam also disrupted population-related endpoints in bumble bee colonies and drone (Fauser-Misslin et al. 2014), and queen (Fauser-Misslin et al. 2014; Moffat et al. 2016) numbers within colonies were reduced.

Recruitment.

Nest weight, brood production, and offspring production are considered recruitment-related endpoints. Changes in nest weight have been observed for imidacloprid (Scholer and Krischik 2014; Wu-Smart and Spivak 2018), clothianidin (Rundlöf et al. 2015; Scholer and Krischik 2014), and thiamethoxam (Moffat et al. 2016). Due to the complexity of queenright colonies, measuring brood production can be difficult since bumble bees build nest structures on top of each other, however, researchers have used methods such as end-of-experiment nest dissection, and photographic quantification of observable brood to measure this endpoint. Reductions in brood production have been observed with imidacloprid (Laycock and Cresswell 2013; Leza et al. 2018; Moffat et al. 2016; Tasei et al. 2000), clothianidin (Moffat et al. 2016), and thiamethoxam (Elston et al. 2013; Laycock et al. 2014). Finally, reductions in offspring production have also been observed with neonicotinoid exposure. This measurement has been found with exposure to imidacloprid and thiamethoxam (Mommaerts et al. 2010).

Notably, after exposure to thiamethoxam, fewer queens initiated egg laying in an overwintering experiment by Baron and colleagues (Baron et al. 2017a). This type of study design, while rare, gives valuable insight into an aspect of queen reproductive potential that is understudied. One study documented effects on thiacloprid which found effects on whole nest measures (reduced weight) and the presence of reproductives (drones and queens) (Ellis et al. 2017). One additional study found that bumble bee colonies placed on thiacloprid-sprayed clover had reduced nest weights (Havstad et al. 2019).

DATA GAPS/DISCUSSION

While the body of research assessing neonicotinoids on bumble bees is growing rapidly, there are notable data gaps within the field to be addressed in order to have a full picture of the effects of these chemicals (Figure 4). We will highlight some areas most in need of additional research and discuss other important considerations for bumble bee researchers.

Figure 4. Important research data gaps.

Headings indicate the biological level of organization in the AOP. Boxes below highlight important data gaps identified during completion of this review. Additional data gaps that do not fall into a discrete category are listed separately.

MIE

Despite the consensus amongst scientists about the mechanism of action of neonicotinoids on neural targets within insects, this class of chemicals has the potential to exert a wide range of effects and impact a variety of endpoints. It remains unknown the exact means by which this occurs, and there is a need for additional research in this area. For instance, there is evidence that mitochondrial function may be altered with neonicotinoid exposure, however, it remains unverified as to whether bumble bee mitochondria possess surface receptors that are responsive to neonicotinoid pesticides. Additionally, radioligand binding assays and patch clamp studies have not been conducted with neonicotinoids and bumble bee neural tissues. Studies of this nature may provide insights into the differences in binding kinetics and agonist potency between honey bees and bumble bees with respect to these chemicals, however, given the evidence in the honey bee literature, it is likely that this level of confirmation is not the most pressing research need (Liu and Casida 1993; Matsuda et al. 2001).

Additionally, work with other insects has demonstrated that there are subtypes of nAchRs with different sensitivities. Researchers have identified nAchRs that are sensitive and insensitive to α-bungarotoxin, a competitive agonist of nAchR (Lapied et al. 1990). Within the insensitive group of nAchRs, there are the subtypes nAchR1 and nAchR2, and the sensitivity of these receptor subtypes to neonicotinoids varies. For instance, the nAchR1 is sensitive to imidacloprid, whereas nAchR2 is not (Courjaret and Lapied 2001). Further, Clothianidin is an agonist of both α-bungarotoxin sensitive and insensitive receptor subtypes (Thany 2009). This work, primarily conducted in cockroaches, reveals the important role of receptor subunit composition and its role in pesticide sensitivity. Equivalent work in bumble bees has not been conducted and may provide insights as to why different species exhibit different effects of neonicotinoid exposure.

Cellular and organ responses

An aspect of queen biology that warrants further investigation is the impact of neonicotinoids on reproductive physiology. Research conducted with bumble bee workers, as described in the cellular response section, found that neonicotinoids have the potential to downregulate biosynthesis pathways relating to juvenile hormone production, which is essential for a variety of reproductive processes (Bellés et al. 2005). If the same downregulation of these pathways occurred in queens, queen fitness and consequently the next season’s bumble bee population could be severely impaired. For instance, juvenile hormones are an important element of oogenesis, and queen juvenile hormone levels are much higher than in the reproductively suppressed workers (Röseler 1977). This cellular-level information may help explain some of the impacts on queens that have been observed, such as delays in nest initiation (Wu-Smart and Spivak 2018) and reductions in egg-laying queens as well as reduced oocyte length (Baron et al. 2017a) after exposure to neonicotinoids. Further studies are needed to explore this possible connection.

Additional studies are also needed to examine other aspects of queen ovary development, since this is likely one of the most critical lines of inquiry within the organ response category of the AOP. Insights into the consequences of early spring neonicotinoid exposure, when young bumble bee queen ovaries are rapidly developing, is of particular interest. Presently, measuring terminal oocyte length is one of the primary measures of queen ovary development, and additional fundamental research into whether this measure is effected by pesticides would be a useful first step within this realm (Bloch et al. 2000).

Organism and population responses

Queen survival, overwintering, and nest initiation are likely the most critical determinants of population stability for bumble bees, however, this area of research has few neonicotinoid studies. Obtaining suitable numbers of young queens for experimentation is challenging since they are not available commercially. Queenright colonies can be purchased, however, there is no guarantee of development stage or whether they will contain gynes or gyne cells. Further, ensuring gynes are fertilized prior to experiments provides yet another challenge. There is a need for refined methods or resources in this area to facilitate research efforts. Further, organizing queen-related data into the AOP framework is difficult, since queens/gynes are in a separate caste, and are also essential for population recruitment as the sole fertilized egg-layers each spring. Therefore, while changes to the numbers of queens was considered a population structure effect here, it can be argued that a decrease in queens is ultimately a population recruitment measure. Overall, clear data gaps remain on the effects of neonicotinoids on queen survival, mating, overwintering success, and subsequent spring nest re-establishment that should be addressed by bumble bee researchers.

Another topic area that has garnered little research attention is the role of drones and drone reproductive measures on queen survival, fitness, and ultimately population stability. As shown in Table 2, little work has been conducted directly on drones with respect to the effects of neonicotinoids. While drones have historically been thought of as a less important caste, there is growing evidence that drone health is tied to young queen health (Baer and Schmid-Hempel 2005; Belsky et al. 2020). If using the microcolony model, determining impacts on drones becomes a very accessible research question since microcolonies exclusively produce drones.

Other considerations

Bumble bee researchers have recognized the need for a clearer understanding of the species differences between B. impatiens versus B. terrestris in terms of their sensitivity to neonicotinoids and other pesticides. Currently, B. terrestris, which is unavailable to North American researchers, dominates the neonicotinoid literature, with 56 out of 78 primary research articles in this review examining effects on B. terrestris. One major reason for this discrepancy is that the European Union has incorporated non-Apis testing into their risk assessment processes, while the United States has not (EFSA 2013). It is possible that life history differences, or the need for method modification for some assays (Gradish et al. 2013), may preclude data extrapolation between these species. Overall, parallel studies with B. terrestris and B. impatiens would clarify any meaningful differences, and if minimal, then the existing data sets for one can be used to inform the other.

The majority of studies to date examined neonicotinoids individually, however, within the environment, bumble bees will encounter a mixture of pesticides throughout their lifespan. For instance, neonicotinoids are often co-applied with fungicides for certain crop applications. A few of the present studies evaluated the effects of these co-exposures and found the potential for synergistic effects on mortality, however, further research is needed to determine whether synergistic effects also occur for other pertinent endpoints, such as impaired brood development or queen survival.

In general, there are substantial data gaps for the two neonicotinoids that have lower acute toxicity, acetamiprid and thiacloprid. Their lower toxicity has been thought to be due to the presence of CYP450 detoxification enzymes, present in both honey bees and bumble bees, that can metabolize these two neonicotinoids. Research has shown that the subfamily CYP9Q are responsible for their metabolism (Manjon et al. 2018). Further characterization of this enzyme in B. terrestris has been conducted by Troczka and colleagues (2019). Between their lower acute toxicity, and the bans placed on some of the more toxic neonicotinoids (EFSA 2013), it is possible that use of acetamiprid and thiacloprid may increase in upcoming years, thus additional research attention for these compounds is warranted.

FUTURE DIRECTIONS

This review identified several important data gaps that could impact our ability to perform risk assessments on bumble bees. As an alternative to conducting tests in multiple species to fill these data gaps, some investigators are calling for increased use of in silico tools to support risk assessment performed on non-Apis bees (Thompson and Pamminger 2019). Seven bumble bee-specific models have been published (Banks et al. 2017; Cresswell 2017; Crone and Williams 2016; Haussler et al. 2017). Bumble-BEEHAVE, the most advanced model available, is designed to simulate colony growth and survival and to predict the effects of multiple stressors on bumble bee survival at the level of the individual, colony and population (Becher et al. 2018). Models like these have great promise for use in ecological risk assessment.

CONCLUSIONS

As bumble bees are further developed as a non-Apis test organism, the body of work describing the impacts of neonicotinoids will naturally expand given the concerns regarding neonicotinoid exposure to pollinators. However, this review demonstrates that there are several areas of research for which notable data gaps exist, and we recommend that research efforts are focused such that our understanding of the impacts of these chemicals on key biological aspects of bumble bees are expanded. It is important to reiterate that this scoping review focused on the types of effects of associated with neonicotinoid exposure and the identification of research gaps rather than the concentrations that elicit observed effects. Therefore, additional work is needed to tie observed effects to exposure to fully inform risk assessments. This scoping review could serve as the foundation for completion of a full systematic review to address this limitation. Overall, increasing the research efforts around Bombus sp will aid in their use for policy-related decisions that may be made around neonicotinoids.