Neonicotinoid insecticides can pose a severe threat to grassland plant bug communities

Introduction

We are facing an unprecedented decline in biodiversity, which has been particularly evident for insects in recent years, with many studies showing global declines in insect biomass, abundance, and richness^1,2,3,4,5. The intensive use of agrochemicals is considered to be a major driver of insect decline^5,6,7. Non-target insects can be exposed to insecticides in both agricultural and non-agricultural areas. For terrestrial insects, direct contact and oral uptake are likely the main routes of exposure. Direct contact can occur through overspray or contact with plant or soil surfaces containing residues. Oral uptake can occur through the consumption of contaminated food, e.g. plants or other invertebrates⁸. In terrestrial non-target habitats (e.g. field margins, neighbouring meadows, hedges), such exposure is mainly caused by overspray and spray drift⁹. Although atmospheric transport of pesticides can lead to deposition even in remote areas such as alpine environments^10,11, areas close to the field such as field margins, are particularly affected, since the first metre of a field margin is exposed to both overspray and spray drift. Estimates of pesticide deposition here range from 30% to 58% of the field rate^9,12. At the same time, semi-natural habitats such as field margins make up a significant proportion of insect habitats in modern agroecosystems¹³. It is, therefore, all the more alarming that there is increasing evidence that modern agroecosystems, as well as non-agricultural landscapes and even insects in nature preserves, are contaminated with pesticides^14,15,16.

While the exposure of non-target insects to pesticides may seem obvious, its impact is poorly understood. Although modern insecticides should ideally be less toxic to beneficial insects such as the Western honeybee Apis mellifera than to the target pest (e.g. based on the German bee risk management levels B1–B4), sensitivity to insecticides can vary even between closely related species and is generally difficult to predict^17,18. The current EU insecticide registration protocol requires sensitivity testing for a limited number of non-target insects. These are typically the honeybee, parasitic wasps, predatory mites, and individual representatives of beetle families such as the Coccinellidae, Carabidae or Staphylinidae¹⁹. Furthermore, the vast majority of available ecotoxicological studies involve either aquatic invertebrates or terrestrial beneficial insects such as pollinators or predators^20,21,22. Herbivorous insects, which account for about 50% of all insect species globally, have been largely neglected^23,24. Another shortcoming in the ecotoxicological literature is the underrepresentation of community-level studies, i.e. the evaluation of a range of species with different sensitivities within the same habitat. The study of community-level effects in realistic exposure scenarios is an important cornerstone for assessing the true ecological consequences of insecticide contamination. Consistent community structure effects of insecticide exposure have already been shown for aquatic systems²¹. However, comparable studies for terrestrial systems are largely lacking, and the few that do exist tend to focus on soil invertebrates^25,26,27.

In this study, we address the existing research gap on insecticide exposure of non-target herbivorous insects, focusing on two main aspects: (1) realistic exposure scenarios, (2) community-level effects, i.e. differential sensitivity between closely related species and between sexes of the same species. We chose plant bugs (Heteroptera: Miridae) as a model group because they are one of the 20 most diverse insect families and a common component of non-target insect communities in agroecosystems^28,29. Plant bugs are highly abundant in meadows, but can also be found in grassy field margins, as many species are associated with Poaceae^28,30. The high diversity and abundance of this group suggest an important function for the ecosystem. However, this group of insects is understudied in terms of non-target effects of insecticides, with the existing literature focusing exclusively on agriculturally relevant beneficial species (i.e. predatory species)^31,32,33 but not on plant bug species primarily occurring in off-field environments. In addition, plant bugs are not considered in the risk assessment procedure for plant protection products. We argue that plant bugs are an ecologically relevant group to study. They are most certainly an important food source for birds and a wide range of predatory invertebrates³⁴. In addition, species within plant bug communities are often similar in many traits (e.g. morphology, habitat, diet, and phenology) and therefore offer the opportunity to study physiological differences in insecticide sensitivity at the species level due to the high diversity of this family²⁸.

We conducted field, greenhouse, and laboratory experiments to mimic insecticide exposure in different scenarios. We used the neurotoxic neonicotinoid insecticide Mospilan®SG with the active ingredient acetamiprid (200 g/kg) in all experiments. Mospilan®SG is applied by spraying and is used in field crops such as rapeseed and potato, in orchards, in wine growing and in floriculture, in particular against chewing-sucking pests. The active substance acetamiprid is both a contact and a systemic insecticide, as it can also be absorbed by plants and distributed throughout their tissues. Among other neonicotinoids, acetamiprid is used worldwide, but in the European Union, it is the only neonicotinoid still registered for open-field use. The global acetamiprid market size is estimated at USD 118 million in 2022 and is expected to grow significantly³⁵.

In the field experiment, we sprayed three concentrations of Mospilan®SG in semi-open field enclosures with wild plant bug populations to simulate in-field and off-field exposure scenarios (Fig. 1a). The enclosures were located in an extensively managed meadow in southern Germany. We selected treatments based on deposition estimates as follows: ‘100%‘ = typical field rate of 250 g/ha in 600 l of water (e.g. as used for potatoes, cabbage, or lettuce); ‘30%‘ = exposure estimate for 1 m field distance³⁶ (overspray + spray drift); ‘0.15%‘ = estimate for 20 m field distance³⁷ (spray drift) (Fig. 1b). In many cases, a specific safety distance from the field edge combined with the use of drift reducing nozzles must be maintained when using Mospilan®SG. However, there are several scenarios for use in Germany, for example, where these regulations do not apply³⁸. Therefore, the field edge exposure scenario chosen here (1 m field distance) is in principle applicable. We chose the 20 m estimate as representative of drift contamination in an adjacent meadow, but without the potential interference from overspray that would occur in close proximity to the field.

Neonicotinoid insecticides can pose a severe threat to grassland plant bug communities — **Fig. 1: Overview of the field experiment and concept of treatment choice.**

Plant bugs were sampled at 2, 9, 16, 24, and 30 days after application using a sweep net. We evaluated the total number of individuals of plant bug populations within the field enclosures and conducted species-level analyses focusing on the three most abundant species: Stenotus binotatus (Fabricius, 1794), Leptopterna dolabrata (Linnaeus, 1758), and Megaloceroea recticornis (Geoffroy, 1785). These three species overlap in phenology (May–August) and host plants (Poaceae) and differ in size (on average 6.6 mm S. b., 8.3 mm L. d., and 9.1 mm M. r.). We also conducted pesticide analyses of vegetation samples from the enclosures to monitor insecticide exposure and metabolism in plants over time.

We further assessed insecticide susceptibility in a greenhouse experiment using host plants (Poaceae from the same meadow) that were sprayed with Mospilan®SG two days earlier to simulate an indirect exposure, where non-target insects enter a habitat after insecticide deposition. In this experiment, we recorded the survival of the three focal species after four days of exposure (i.e. 6 days after plants were sprayed). We also analysed insecticide residues on host plants two days after application. Finally, we performed dose-response assays in the laboratory, where we applied different concentrations of Mospilan®SG topically on the dorsum of individual plant bugs and recorded mortality after 24 h. This approach allowed us to investigate intrinsic differences in contact sensitivity between species and between conspecific males and females without confounding factors, such as potential behavioural avoidance of exposure in the field.

This study reveals an overall high sensitivity of plant bugs to the neonicotinoid formulation, as well as species and sex-specific responses. The findings provide a better understanding of the impact of terrestrial insecticide contamination on non-target plant bug communities and identify potential weaknesses in the current European risk assessment.

Results

Field experiment

Neonicotinoid application significantly reduced total plant bug abundance after two days by 83% (100% treatment) and 82% (30% treatment) compared to the control based on a generalised linear mixed model followed by a Tukey post-hoc test (100%: p < 0.001; 30%: p < 0.001) (Fig. 2a). Nine days after exposure, abundance was reduced by 70% in the 100% treatment and by 57% in the 30% treatment (100%: p < 0.001; 30%: p < 0.001; Tukey test). This shift in proportions between treatments, and less pronounced differences in plant bug counts at later samplings, was probably due to a combination of repeated sampling and an overall reduction in individuals following the natural phenology of the most dominant species. No difference in mean counts was observed between the control and 0.15% treatments and between the 30% and 100% treatments at any of the five sampling events, except at the third sampling event (16 days after application). Here the mean counts in the 100% treatment are lower than in the 30% treatment (p = 0.012). A detailed report of the contrasts for each sampling is provided in Supplementary Table S1.

**Fig. 2: Exposure to Mospilan®SG reduces plant bug abundance and causes a species-specific response.**

Focusing on the three most abundant species, the mean counts for S. binotatus, L. dolabrata and M. recticornis in the 30% and 100% treatment were significantly reduced compared to the control two days after insecticide application (S. b. 100%: p < 0.001, S. b. 30%: p < 0.001; L. d. 100%: p = 0.002; L. d. 30%: p < 0.001; M. r. 100%: p = 0.031; M. r. 30%: p = 0.024) (Fig. 2b). A detailed report of the contrasts is given in Supplementary Table S2. As with the total number of plant bugs, the proportions between treatments for these three species shifted at later sampling dates, probably due to a combination of repeated sampling and a decline in overall abundance due to the natural phenology of the species. While these factors probably affected counts at later samplings, the population dynamics shown in Fig. 3 suggest that there has been no recovery at high treatment doses (e.g. by recolonization).

**Fig. 3: Population development of three focal species in the field trial.**

Pesticide analysis of vegetation samples from the enclosures (aboveground plant material including leaves, stems, and inflorescences of grasses and herbs) taken after 7, 17, and 30 days revealed residues of acetamiprid and its metabolite acetamiprid-N-desmethyl in the 100% and the 30% treatments (Table 1). No residues were found in the 0.15% treatment or in the control at any sampling time. Further information on pesticide analysis is provided in Supplementary Note 1.

Table 1 Insecticide residues in the field enclosures

Full size table

Greenhouse experiment

Mospilan®SG reduced the survival for all plant bug species on plants sprayed with 30% of the field application rate (=75 g Mospilan®SG/ha in 600 l water) to mimic field-margin exposure 48 h ahead of plant bug release. Specifically, we observed a reduction of 88% for S. binotatus, 39% for L. dolabrata, and 35% for M. recticornis compared to the respective controls four days after transfer on contaminated plants (Fig. 4). Notably, no individuals of S. binotatus survived the insecticide treatment in any of the nine replicates.

**Fig. 4: Survival of plant bugs after four days of exposure to plants treated with 30%-Mospilan®SG.**

Acetamiprid and the metabolite acetamiprid-N-desmethyl were found in all samples collected from treated plants two days after application (i.e. the time of plant bug release). The mean concentration was 267 (±78) ng/g plant material for acetamiprid and 62 (±33) ng/g for acetamiprid-N-desmethyl (dry weight; N = 9; rounded numbers). Further information on pesticide analysis is provided in Supplementary Note 1.

Dose-response assays

All three species showed high sensitivity to the Mospilan®SG treatment in the topical dose-response assays in the laboratory (Fig. 5). Comparing the observed lethal doses required to kill 50% of individuals (LD₅₀) for L. dolabrata (53.3 ng/µl), M. recticornis (39.0 ng/µl) and S. binotatus (4.1 ng/µl) with the typical field concentration of 416.0 ng/µl, the LD₅₀ values for L.d. would correspond to 12.8%, M.r. 9.4% and S.b. 1.0% of the field concentration. In addition, pronounced species-specific differences were observed, with S. binotatus being significantly more sensitive to the insecticide than the other two species based on LD₅₀ comparisons (L. dolabrata − S. binotatus: p = 0.001, t = 3.33; M. recticornis − S. binotatus: p = 0.021, t = 2.33).

**Fig. 5: Mospilan®SG dose-response curves.**

Besides the high toxicity of Mospilan®SG, we observed significant differences between males and females of two species. Based on LD₅₀ values, males of L. dolabrata and M. recticornis were >20 times more susceptible than females (L. d. t = 6.33, p < 0.001; M. r. t = 3.48, p < 0.001). Sensitivity was not significantly different between the sexes of S. binotatus based on the fitted generalised linear mixed model (GLMM). Therefore, no LD₅₀ values are reported from the dose-response model.

Discussion

Insecticides are considered a major cause of insect decline, but there is a lack of empirical evidence on the environmental effects on non-target species within and outside the cropping area. Using a neonicotinoid formulation, we studied the effect of terrestrial insecticide contamination on non-target plant bug communities. We conducted a series of field, greenhouse, and laboratory experiments to investigate community-, species- and sex-specific responses in this insect group. Plant bug species in the field experiment were highly sensitive to neonicotinoid exposure. We also found species-specific responses across different scales (i.e. field, greenhouse, and laboratory), and drastic differences in sensitivity between males and females.

Plant bug communities are highly sensitive to insecticide exposure

Consistent across experiments, our study shows a high susceptibility of plant bugs to insecticide exposure. In the field experiment, we aimed to simulate a scenario of overspray and direct drift contamination with specimens present in the off-field environment during spray application. A drastic reduction in plant bug abundance was observed in the scenario representing exposure within one metre of the field (30% of the typical field rate).

A second scenario of indirect drift contamination for the same field distance was simulated in the greenhouse experiment, as it is also possible that insects enter an area only after drift contamination occurs (i.e. after insecticides are applied). Residues of the compound that may still be present on surfaces (soil or plant material) hours or days later can still be absorbed through an insect’s cuticle, e.g. via the tarsi³⁹, or ingested during feeding. In addition, in systemic insecticides such as neonicotinoids, the active ingredient can be taken up by plants, providing an additional route of oral exposure for piercing-sucking herbivorous insects (e.g. plant bugs) over extended periods of time. This is not surprising as systemic insecticides are specifically designed to protect crops by controlling pests in this way. However, this mechanism also inevitably affects non-target insects. In our experiment, even two days after host plants were treated with 30% of the typical field rate of Mospilan®SG, the survival rates of plant bugs placed on these plants were significantly reduced and even zero in S. binotatus.

These results clearly demonstrate the detrimental effect of insecticide contamination on non-target plant bug populations located close to arable fields. Further, a generally high sensitivity to neonicotinoid insecticides of this ecologically relevant insect group that most likely occupies an important position in the trophic pyramid, is revealed^34,40. This pattern was further supported by the insecticide dose-response assays. Here, acute contact LD₅₀ values for three plant bug species, simulating direct exposure from spraying, were at concentrations corresponding to 1% of the typical field concentration. To put our findings into a broader context, these values can be compared to those of common model organisms such as the Western honeybee A. mellifera. However, LD₅₀ values from dose-response assays are typically given as a concentration of pure active ingredient per microliter or per individual. To establish comparability, the values we obtained using a commercial insecticide formulation can be adjusted to reflect the 20% active ingredient acetamiprid content. Based on the LD₅₀ values reported in Fig. 5, this corresponds to 10.63 ng acetamiprid /bug for L. dolabrata, 7.81 ng/bug for M. recticornis and 0.82 ng/bug for S. binotatus. For honeybees, an LD₅₀ of ~9260 ng active ingredient per bee is reported as contact exposure value, also based on a 20% acetamiprid formulation⁴¹. Accordingly, the sensitivity of the tested plant bug species is ~ 871 times (L. dolabrata) to ~ 11,293 times (S. binotatus) higher, which can be either explained by differences in species sensitivity or formulations. In general, several factors could cause differences in sensitivity between mirids and honeybees, such as body size, cuticle composition, body fat content, or the fact that the honeybee, as a dietary generalist, is potentially adapted to be exposed to a great variety of defensive secondary plant compounds (i.e. toxins) in nectar and pollen, whereas the plant bugs tested only feed on grasses.

However, comparisons between species involving different formulations should be made with caution. Nevertheless, the pronounced differences in LD₅₀ values show that plant bugs are very sensitive to neonicotinoid insecticides and consequently to environmental contamination at only a fraction of the applied field concentrations. This is consistent with existing literature suggesting that plant bugs are usually very sensitive to neonicotinoid insecticides^31,42,43. This group of insects should therefore be given greater consideration in risk assessment (see “Environmental risk assessment” below).

Species-specific sensitivity

The three plant bug species that were tested in this study can be considered representative non-target herbivorous insects. Specifically, these species are of particular interest because they share a similar ecology (e.g. feeding habits, phenology, and host plants²⁸) and yet showed markedly different insecticide susceptibility under the experimental exposure scenarios. Differences in species response were already suggested by the field experiment and further confirmed by the greenhouse experiment with treated plants and the dose-response assays in the laboratory. Therefore, they clearly represent physiological differences between species and are not related to feeding preferences or position on the plant during spraying. The species differences are also unlikely to be explained by age differences due to different phenologies, as adult specimens of the three species appeared simultaneously in the field and had overlapping phenologies (see phenogram Supplementary Fig. S1).

Physiological factors that have been proposed to influence sensitivity include, for example, differences in the activity and abundance of detoxification enzymes, or simply body size⁴⁴. Notably, the smallest (and lightest) species, S. binotatus, was always the most sensitive in our greenhouse and dose-response assays, suggesting that body size or mass may be an important factor. Consistent with this, Goulson⁴⁵ compared neonicotinoid LD₅₀ values reported in the literature and stated that much of the variation in sensitivity between different insect taxa could be explained by differences in body size or weight. For example, the pronounced difference in LD₅₀ values between the plant hopper Nilaparvata lugens and the Colorado potato beetle Leptinotarsa decemlineata for the neonicotinoid imidacloprid was explained by body mass (~1 mg vs ~130 mg). Here we found, that although L. dolabrata has more than twice the body mass of M. recticornis, the LD₅₀ values were almost identical when corrected for body mass (L. d.: 0.53 ng a.i./mg bug, M. r.: 0.66 ng a.i./mg bug; see Supplementary Tables S3 and S4). At the same time, the susceptibility of S. binotatus is still four times higher when corrected for body mass, suggesting either additional physiological factors or a non-linear relationship between body mass and susceptibility.

While the physiological mechanisms involved remain to be investigated, differences in species susceptibility could permanently alter local community structure under a constant or repeated exposure scenario. Changes in aquatic invertebrate community composition have been associated with pesticide contamination in the past^21,46, whereas the effects on terrestrial invertebrate communities are less well explored. A recent meta-analysis showed that pesticide contamination reduces soil invertebrate diversity²⁷ and studies on neonicotinoids show that they affect the abundance, diversity and community structure of soil invertebrates^25,26. Also, Main et al.⁴⁷ linked neonicotinoid soil contamination to reduced wild bee species richness in field margins. However, especially for insect herbivores, which play a central role in terrestrial food webs⁴⁸, more research is needed to understand the consequences of pesticide exposure at the landscape scale. Particular attention should be paid to univoltine species with short nymphal and adult phenologies that may overlap with common pesticide application regimes such as the plant bugs studied here.

Sex-specific sensitivity

In addition to differences in sensitivity at the species level, we found that males of L. dolabrata and M. recticornis were significantly more sensitive to the insecticide than the corresponding females. Sex-specific insecticide sensitivity is a pattern reported for many insects, with males often being more susceptible than females^49,50. However, there are also contrary reports in some species⁵¹.

Potential factors involved in sex-specific susceptibility partly overlap with those likely driving the observed species-specific differences. The most likely factor is again body mass, especially in sexually dimorphic species such as L. dolabrata and M. recticornis. In both species, males had a lower mean fresh body mass than females, with a 2.9-fold difference in L. dolabrata and a 2.2-fold difference in M. recticornis (N = 10 per species and sex). This is not directly reflected in the difference in LD₅₀ values between males and females, which is 22-fold for L. dolabrata and 21-fold for M. recticornis. However, as mentioned above, such relationships may not be necessarily linear.

Another source for differential susceptibility is at the enzymatic level. Frohlich⁵² showed that the activity of detoxification enzymes decreased with age in males of the solitary bee Megachile rotunda, but not in females. An age-dependent change in enzyme activity was also observed by Guirgus and Brindley⁵³ when they compared 1-day-old males of Megachile pacifica with 4-day-old males, both of which had been previously treated with a fungicide. Thus, age also appears to play a critical role in the susceptibility of insects to insecticides.

The higher control mortality observed in males in our dose-response study (which was considered in the models) could suggest an older initial age compared to females in the same cohort. Wheeler³⁴ noted that in most plant bug species, males usually appear before females and do not live as long. In addition, for L. dolabrata, Wachmann et al.²⁸ suggest that females live longer than males. A lower vitality of male specimens due to a shorter life cycle (i.e. a higher relative age at the time of sampling) could therefore explain a higher mortality in the controls and therefore a higher susceptibility to stressors such as insecticides. Tan et al.⁵⁴ have previously suggested that imidacloprid disproportionately affects males of the plant bug Apolygus lucorum beyond the sex differences observed in controls. However, in all three species, the interaction between ‘cohort’ and ‘sex’ did not explain any variance in the data when tested in the GLMM (see “Methods”). Therefore, in our dose-response assays, the age of the sampled field population (=cohort) is unlikely to influence the differential insecticide susceptibility between males and females.

Regardless of the underlying mechanisms, the observed differences of a full order of magnitude could lead to a shift in sex ratios, in this case towards females. This could ultimately lead to a reduction in abundance and, in the long term, a shift in species composition.

Environmental risk assessment

To assess the risk of insecticide contamination to non-target species, it is essential to determine the true vulnerability of these species in their natural environment. Vulnerability involves both physiological sensitivity and exposure to an insecticide⁵⁵. We have shown that plant bugs, as a non-target group, are physiologically highly sensitive to insecticides. Furthermore, the species tested here are likely to occur in highly exposed areas of the off-field environment, i.e. grass-dominated field margins or adjacent meadows²⁸. We also demonstrated that exposure to a systemic insecticide extends beyond the initial contamination by direct contact with spray droplets, as plant bugs placed on previously sprayed plants showed a drastic reduction in survival. Considering that in the field experiment, we were able to detect acetamiprid in host plant tissue up to 30 days after application, the threat to non-target plant bugs is long-lasting. It is therefore all the more surprising that the current risk assessment does not consider this group or other herbivorous insects in the non-target insect community. In addition, field margins less than 3 m wide are currently not considered to be terrestrial non-target habitats for pesticide application in Germany⁵⁶. Consequently, this habitat and the insects living there are not protected from pesticide contamination¹³.

Roß-Nickoll et al.⁵⁷ argue that the test procedures used in the EU risk assessment are insufficient because the test species used do not adequately reflect the non-target insect community. As a result, the disruption to these communities caused by pesticide exposure is not considered. This discrepancy is further illustrated by the drastic difference in LD₅₀ values between the honeybee model and the mirid species we tested, suggesting up to ~ 11,000 times higher toxicity to plant bugs compared to the honeybee. In addition, risk assessment is currently only done at the species level, and sex-specific sensitivity is not considered. However, the up to 22-fold difference in LD₅₀ values between sexes that we found could scale to detrimental long-term effects on plant bug community composition. This calls for an increase of the so-called uncertainty factors used to cover different species and sexes from currently 10 to at least 1000 (based on the least sensitive species, L. dolabrata). Consequently, and in light of the renewal of the authorisation of acetamiprid until 2033⁵⁸, we argue that the current European risk assessment system urgently needs to be adapted accordingly to adequately address the environmental risk of pesticide contamination.

Methods

Statistical analysis and graphical representation

Statistical analysis was conducted using R⁵⁹ (version 4.3.0) and R-Studio⁶⁰ (version 2023.6.1.524). Additionally used R-packages are cited in the following sections.

All graphs showing bar or stacked bar charts were generated using JMP®Pro⁶¹ (version 17.2.0). Dose-response curves were generated with R using the package ‘drc’⁶². All figures were processed with Inkscape⁶³ (version 1.3.2).