Changes in predator biomass may mask the negative effects of neonicotinoids on primary consumers in field settings

The non‐target effects of pesticides, particularly those that are long‐lasting and move easily through the environment, could have community‐level impacts on beneficial arthropods and hinder conservation efforts in agrolandscapes We assessed the impacts of a neonicotinoid insecticide, clothianidin, and possible synergisms with a fungicide by quantifying predator, herbivore, and pollinator biomass and morphospecies richness in simulated prairie restorations. Predator biomass was 66% lower in plots treated with clothianidin compared to controls and this effect persisted across the growing season. Herbivore biomass was 51% lower in clothianidin‐treated plots in June, but the effect waned over the growing season, and no difference was detected in July or August. There was a synergistic effect of clothianidin and fungicide in lowering herbivore morphospecies richness by 12%. Pollinators appeared unaffected by clothianidin. Instead, pollinator biomass increased by 71% with added fungicide in the absence of clothianidin. The results of this study underscore the complexity of pesticide effects in field settings. Additional studies are necessary to understand how pesticide dissipation and predator release may interactively affect late‐season herbivore populations.


INTRODUCTION
In landscapes dominated by agriculture, conservation of natural areas and the restoration of marginal cropland to more natural habitats are considered essential to support declining arthropod species while providing pest suppression and pollination ecosystem services in nearby crops (Bennett & Gratton, 2013;Haan et al., 2021;Samways et al., 2020;Seibold et al., 2019). However, past and ongoing pesticide use within agrolandscapes may hinder the effectiveness of such areas in supporting beneficial arthropods (Brittain et al., 2010;Kraus et al., 2021;Mogren & Lundgren, 2016). To best conserve arthropods in human-affected landscapes, we must evaluate how pesticides, both individually and in combination, affect communities in the field.
Pesticides that pose the greatest risk to non-target arthropod communities are those that degrade slowly, easily move through the environment, produce long-lasting effects, and interact with other toxic compounds (Bish et al., 2021;Carvalho, 2017;Mahdjoub et al., 2020;Topping et al., 2020). Neonicotinoid insecticides meet many of these criteria. They degrade slowly, move horizontally through environments as dust and in runoff, and can occur in biologically relevant concentrations in untreated areas Goulson, 2013;Hladik et al., 2018;Pecenka et al., 2021;Wood & Goulson, 2017). Additionally, neonicotinoids are often applied with fungicides, which can have synergistic effects that can have important implications for insects (Krupke et al., 2012;Main et al., 2020;Sgolastra et al., 2018). For example, microbial communities disrupted by fungicides can affect plant nutrition and shape plant communities (Koricheva et al., 2009;Smith et al., 2000;Yang et al., 2018). Such changes can influence the susceptibility of arthropods to pesticides (Gordon, 1961;Tosi et al., 2017). Despite the growing evidence of notable non-target effects, most toxicological research focuses on individual and population-level responses to pesticides and rarely examines broader community effects.
By examining the community-and feeding guild-level outcomes of pesticides, we will better understand the biological relevance of observed effects on individuals and populations. While toxicological research focused on sub-community levels provides insights into the mechanisms at finer scales, differences in species' sensitivity and reactions to pesticides can make it difficult to extrapolate effects on the broader community (Anderson & Harmon-Threatt, 2019, 2021Woodcock et al., 2016).
By examining community responses partitioned by feeding guild, we might gain a better and more comprehensive understanding of how pesticides affect arthropods and the ramifications for ecosystem services.
The goal of the current study was to investigate the effects of pesticides on arthropod communities in a simulated prairie restoration embedded within an agrolandscape. We focused on arthropod biomass and species richness as these metrics have been used to gauge the success of other types of restorations and they are related to the value of provided ecosystem services (Garibaldi et al., 2016;Garratt et al., 2014;Greenop et al., 2018;Schneider et al., 2022). We predicted that pesticides would have the largest impact on primary consumers (e.g. herbivores and pollinators) due to their more direct exposure to insecticides via contaminated food resources and fungicides altering plant nutrition and community structure Bredeson & Lundgren, 2018;Bromilow & Chamberlain, 1995;Koricheva et al., 2009). Predators were expected to be mainly impacted by changes in their prey communities and buffered from the direct effects of pesticides. As the active ingredients dissipate over the growing season, the magnitude of the pesticide effects was expected to decrease. We hope to illuminate the possible synergistic effects of pesticides on arthropod communities and the consequences for ecosystem services supplied by conserved and restored areas.

Site and experimental set-up
We collected arthropod data from a subset of the simulated prairie restorations at the University of Illinois Urbana-Champaign's Philips Tract in Urbana, Illinois (40 07 0 56 00 N, 88 09 0 06 00 W). In 2018, an agricultural field planted with a corn-soy rotation for at least the previous 50 years (J.L. Ellis, personal communication) was converted to 96 20 Â 20 m plots. The broader project goal was to determine the effects of neonicotinoids, microbial communities, flower diversity, and their interactions on native bee foraging and nesting. The treatments were applied using a split-plot design, with neonicotinoids applied to groups of eight neighbouring plots (i.e. 'whole plots') and the other treatments applied in a 2 Â 2 Â 2 Â 2 factorial design to individual plots (i.e. 'split plots') within each whole plot. For the present study, we sampled from 24 plots seeded with high diversity seed mixes of 34 native flowers and grasses in a 2 Â 2 factorial design to determine the main and interactive effects of the neonicotinoid insecticide clothianidin and the phthalimide fungicide captan on arthropod communities (i.e. six plots per treatment combination). Clothianidin, like other neonicotinoids, is incorporated into plant tissues as they grow (Bromilow & Chamberlain, 1995) and is a neurotoxin that binds to arthropod nicotinic acetylcholine receptors (nAChR; Casida & Durkin, 2013). It has a variety of effects on non-target arthropods at realistic field concentrations (reviewed by Pisa et al., 2015Pisa et al., , 2017. Clothianidin is widely used in seed coatings for row crops like corn and soy (Simon-Delso et al., 2015), has a relatively long half-life in the environment (148-6931 days, reviewed by Goulson, 2013), and moves horizontally as dust during seeding and in runoff (Nuyttens et al., 2013;Whiting et al., 2014). Captan was selected for the broader experiment due to its broad effect on microbial communities (Widenfalk et al., 2008). Captan has a short half-life (<20 days; Schoen & Winterlin, 1987), and we interpreted the effects of this treatment to be the result of changes in the microbial and plant communities, including plant nutrition, rather than direct toxicity (Laurin & Bostanian, 2007;Stanley & Preetha, 2016). The choice of clothianidin and captan for the broader project was not meant to mimic an existing cropping system. Rather, these pesticides were selected as general representatives with previously reported main effects on non-target arthropods and microbes (Pisa et al., 2015(Pisa et al., , 2017Widenfalk et al., 2008). Additionally, commercially available proprietary seed mixes can be challenging to recreate and limit our ability to explore interactive effects between compounds (Dubey et al., 2020)

Arthropod collection
We collected above-ground arthropods once per month (June to August) in 2020. We focused on above-ground arthropods because of their role in pest suppression (Schmitz & Barton, 2014) and pollination (Klein et al., 2007) ecosystem services. We sampled arthropods in two 20 m transects of 20 sweeps using a 38.1 cm diameter sweep net between 1200 and 1600. The direction of the transects (N-S or E-W) was randomly assigned during each sampling period. Transects were 6 m from the nearest parallel plot edge. Collected arthropods were placed in a 1 gallon Ziplock bag and placed on ice until they could be returned to the laboratory and stored at À20 C. Arthropods were sorted to feeding guild-predator, herbivore, or pollinator-based on individual life stage and the most common feeding type at the familylevel (Borror & White, 1970;Triplehorn et al., 2005;Whitfield & Purcell, 2014). Due to their complex position within food webs and their low abundance in our sample, omnivores were excluded. For predators, herbivores, and pollinators, we counted the number of morphospecies per plot-month combination, dried them at 60 C for

Statistical analysis
We used generalised linear mixed models (GLMM) to analyse the effects of captan and clothianidin on predator, herbivore, and pollinator dry biomass and morphospecies richness. Captan, clothianidin, and month were included as fixed effects and experimental block (i.e. 'whole plot', see section Site and experimental set-up) was included as a random effect. In the models for morphospecies richness, we added the natural log of the feeding guild's abundance as an offset term as sample abundance strongly influences observed species richness (Chase et al., 2019;Gooriah & Chase, 2020). Gaussian and Poisson distributions were used to model biomass and morphospecies richness, respectively. Models were validated visually, and biomass was natural-log transformed to meet model assumptions. In the case of a significant main effect of month or interaction, we conducted post hoc contrasts using Fisher's LSD. We chose an alpha of 0.1 for statistical significance due to sample size constraints created by the broader study design and to balance Type I and II errors. All analyses were performed in R v4.1.2 (R Core Team, 2021). We used the packages lme4 v1.1.27 (Bates et al., 2015) and car v3.0.10 (Fox & Weisberg, 2019) to fit our GLMMs and to run analysis of deviance with Type III Wald χ 2 tests, respectively. Post hoc analyses were performed using the package emmeans v1.6.1 (Lenth, 2020), and data figures were prepared using the packages ggplot2 v3.3.3 (Wickham, 2016) and patchwork v1.1.1 (Pedersen, 2020).

RESULTS AND DISCUSSION
Unexpectedly, predaceous arthropods, predominantly spiders (Order: Araneae), wasps and parasitic wasps (Order: Hymenoptera), beetles (Families: Coccinellidae and Carabidae), and true bugs (Families: Reduviidae and Anthocoridae), were the most dramatically impacted by clothianidin, and this effect did not lessen as the growing season progressed. Predator biomass was 66% lower in plots treated with clothianidin than the control (Table 1, Figure 1). This effect was consistent across months even as overall predator biomass was lowest in July and highest in August (t 45.1 = 2.413, p = 0.051). Predator morphospecies richness was not impacted by clothianidin or captan, but it did vary over time and was the lowest in July (jZj ≥ 3.503, p ≤ 0.001).
The simplest hypothesis for the dramatically lower predator biomass in clothianidin-treated plots is lower herbivore biomass (e.g. bottom-up controls on predators; Frederiksen et al., 2006). However, we only observed a matching decline in herbivore biomass for captan-free plots in June (Table 1, Figure 1). Thus, gross changes in herbivore community biomass are unlikely to account for the observed decline in predator biomass. Trophic interactions in the form of pesticide-mediated changes in herbivore community composition (Atwood et al., 2018) and

Predators
Herbivores Pollinators predator preferences were beyond the scope of the current study but deserve further exploration. In the absence of this information, long exposure to contaminated soils during vulnerable life stages-such as overwintering predators in semi-natural habitats that border agricultural fields (Clem & Harmon-Threatt, 2021)-may explain the observed loss of predator biomass. The toxicity of neonicotinoids compounds with exposure time (Charpentier et al., 2014;Hayasaka et al., 2019;Suchail et al., 2001) and chronic contact with contaminated soils has been suggested as an important yet understudied route of exposure for taxa with similar life histories (Anderson & Harmon-Threatt, 2019, 2021. Pesticide drift from early-season plantings in nearby agricultural fields may contaminate soils in restorations at a critical time for predators and have effects throughout the growing season. Other hypothesised routes of exposure for predators have been reviewed elsewhere (Tooker & Pearsons, 2021). Despite this large effect on predator biomass, some of the effects lower in the food web were inconsistent over the summer.
The negative effect of clothianidin on herbivore biomass appeared to wane as the growing season progressed while morphospecies richness was constantly lower in areas with both clothianidin and fungicide ( Figure 1, Table 1). Herbivores in this study were mainly grasshoppers and crickets (Order: Orthoptera), frit flies (Family: Chloropidae), true bugs (Families: Miridae, Pentatomidae, and Cicadellidae), F I G U R E 1 Drivers of arthropod biomass and morphospecies richness in simulated prairie restorations. Factors that were statistically significant are presented (p ≤ 0.1, see section Methods). There were no statistically significant effects of the included explanatory variables on pollinator morphospecies richness, so we show the full combination of factor levels. Capital letters indicate significant differences between months, lowercase letters between clothianidin and captan treatments within each month (herbivore biomass) and across months (herbivore morphospecies richness), and an asterisk between two groups. For clothianidin and captan, + indicates areas with the associated pesticide and À indicates areas without the associated pesticide. and leaf beetles (Family: Chrysomelidae). In June, herbivore biomass was 51% lower in plots with clothianidin when captan was absent (t 51.3 = 1.718, p = 0.092). This effect disappeared in July and herbivore biomass increased by 113% with captan in plots not treated with clothianidin (t 45.0 = 2.346, p = 0.023). Interestingly, there was no clear evidence that herbivores were released from predation when predator biomass was lower in clothianidin-treated plots. We hypothesise that the detectible negative effect of clothianidin on herbivore biomass shrank as the insecticide degraded until it was masked by the offsetting, positive effect of predator release. Additionally, herbivore populations are thought to recover more quickly than predators in response to disturbance events (Solé et al., 2002), which may help explain why the negative effect of clothianidin on herbivore biomass did not persist across the growing season as it did for predator biomass.
In terms of herbivore morphospecies richness, there was a synergistic interaction between clothianidin and captan, which resulted in 12% fewer herbivore morphospecies (Z = 1.828, p = 0.068) and this was stable across the growing season ( Figure 1, Table 1). By disrupting microbial communities, the fungicide may have lowered the nutritional quality of plant foods (Smith & Rice, 2000), which can compound the negative effects of neonicotinoids for some species (Tosi et al., 2017).
Reduced competition for resources for the remaining species could explain why we did not observe a corresponding decline in herbivore biomass.
The observed responses of predators and herbivores suggest that the prophylactic use of neonicotinoids may disrupt pest suppression services provided by semi-natural habitats, increase target and non-target pest damage to crops, and help explain the inconsistent economics of neonicotinoid seed treatments (Gore et al., 2014;Hurley & Mitchell, 2017;Krupke et al., 2017;Mourtzinis et al., 2019;Pecenka et al., 2021). By greatly reducing predator biomass, fewer predators may move from semi-natural areas into clothianidin-treated agricultural fields to control populations of pest species (Gagic et al., 2019). Additionally, by disrupting predator controls on herbivore populations within semi-natural areas, the use of clothianidin could lead to outbreaks of pest insect populations that are usually maintained under economic thresholds, causing increased crop damage and reducing yield (Douglas et al., 2015). Instead of prophylactic use, limited pesticide use within an integrated pest management (IPM) framework is more likely to be compatible with the pest suppression ecosystem services provided by semi-natural areas in agrolandscapes. The absence of a detected effect on predator morphospecies richness does provide some reason for optimism, as natural enemy diversity is often associated with pest suppression and may mean that some of the lost pest suppression ecosystem services are offset (Greenop et al., 2018;Schneider et al., 2022).
We did not detect a statistical effect of clothianidin on pollinators, primarily adult hoverflies (Family: Syrphidae) and wild bees (Clade: Anthophila) ( Figure 1, Table 1). However, the effect size of clothianidin on pollinator biomass was roughly equal to the 71% increase in response to captan-which was statistically significant (t 46.1 = 1.744, p = 0.088). The design of the broader study with clothianidin applied at the block level, the small size and proximity of the individual plots, and the high dispersal ability of pollinators may obscure effects in the current study. However, the raw data are consistent with previous studies that suggest pollinators are attracted to neonicotinoid-contaminated resources (Arce et al., 2018;Kessler et al., 2015;Tetlie, 2020). More targeted sampling around specific life stages and behaviour, such as foraging and nesting wild bees, will likely reveal the true effects on this feeding guild and if areas designated for pollinator conservation within agrolandscapes may cause long-term harm to this important group (Mogren & Lundgren, 2016).
Pesticide contamination of conserved and restored habitats in agrolandscapes poses a serious threat to our conservation goals for arthropod communities and desired ecosystem services. While toxicological risk assessments are important, extrapolating from single-species studies to full communities is likely inadequate in reflecting the ecological reality in the field. Negative effects on the most well-studied taxa-typically target pests and beneficial primary consumers (Pisa et al., 2015(Pisa et al., , 2017-may be compensated for by the release from predation described in the current study. Similarly, field studies that focus on a restricted range of taxa could underestimate the true negative non-target impacts of pesticides if other changes mask the effect of interest. Based on the results we present in this study, we urge caution when basing conservation policies on studies that report no effects of pesticide exposure until a more general community assessment is conducted.