Environmental levels of neonicotinoids reduce prey consumption, mobility and emergence of the damselfly Ischnura elegans

1. Freshwaters are among the most endangered ecosystems in the world as a result of anthropogenic interference such as pollution. Pollution in the form of neonico tinoids has been intensively studied, but data thus far is often conflicted by con trasting responses between laboratory and field experiments. In addition, toxicity data are scarce and contradictory for insects such as Odonates (dragonflies and damselflies) and a potential risk to them may therefore be overlooked. 2. We investigate the potential risk of neonicotinoids to Odonates by exposing nymphs of the blue-tailed damselfly Ischnura elegans to environmentally relevant concentrations of the neonicotinoid thiacloprid. We consider I. elegans as an indicator species for other Odonates


| INTRODUC TI ON
Pesticides have been indicated as a potential pollution threat to freshwater ecosystems (Beketov, Kefford, Schäfer, & Liess, 2013;Tilman et al., 2001), one of the most endangered ecosystems in the world (Dudgeon et al., 2006). Currently, neonicotinoid insecticides receive much of the attention as their usage have increased sharply over the past two decades (Simon-Delso et al., 2015) and they are now the most commonly used insecticides globally (Jeschke, Nauen, Schindler, & Elbert, 2011).
Neonicotinoids combat insect pest species by binding to the nicotinic acetylcholine receptors (nAChR), which induces continuous excitation of the neuronal membranes leading to paralysis (Simon-Delso et al., 2015). Like most agricultural chemicals, neonicotinoids are often found in surface waters due to processes such as run-off, drift and leaching (Morrissey et al., 2015). The effects of neonicotinoids on non-target freshwater organisms have been studied intensively in the laboratory (e.g., Raby, Zhao, Hao, Poirier, & Sibley, 2018;Roessink, Merga, Zweers, & Brink, 2013), but the data are actually scarce for many groups, such as Odonates (dragonflies and damselflies). Consequently, scientists that focused on the risk of neonicotinoids to Odonates (e.g. Morrissey et al., 2015;Vijver et al., 2017) largely had to interpolate their assessment on only two bioassays (Beketov & Liess, 2008;Jinguji, Thuyet, Uéda, & Watanebe, 2013). These studies showed that the survival and emergence of the dragonfly genus Sympetrum was negatively affected by concentrations that are higher than those typically found in natural environments. This contrasts with the study of van Dijk, Staalduinen, Sluijs, and Der, (2013) who indicated possible effects of current environmental levels of neonicotinoids on the abundance of the damselfly Ischnura elegans. Nevertheless, the low amount of data available and the contrasting findings indicate that the risk of neonicotinoids to Odonates is currently poorly understood.
On top of this, recent findings show that the toxicity of neonicotinoids to aquatic invertebrates is food quantity or quality dependent (Alexander, Luis, Culp, Baird, & Cessna, 2013;Barmentlo, Parmentier, Snoo, & Vijver, 2018;Ieromina et al., 2014). High food quality/quantity, as offered in standard laboratory experiments, may allow for higher energy expenditure to cope with the toxic stress (the Dynamic-Energy budget theory, see Nisbet, Muller, Lika, & Kooijman, 2010). However, in the field, a plethora of (indirect) effects on the aquatic communities can limit food quality/quantity and consequently species sensitivity in a laboratory setup can differ from that in the field (Barmentlo et al., 2018).
Approximately one of seven Odonates is threatened with extinction in Europe and currently 24% of the species have declining populations (Kalkman et al., 2010). Since conservation science of Odonates is limited in general (Bried & Samways, 2015) and toxicity data are scarce and contradictory for this group, a potential hazard of neonicotinoids to Odonates could be overlooked. Therefore, we aimed to investigate if environmentally relevant concentrations of the neonicotinoid thiacloprid could affect a widespread, eurytopic, Odonate: the blue-tailed damselfly I. elegans. We investigated the effects of thiacloprid on several endpoints (survival, consumption, growth, molting, mobility and emergence) using an environmentally realistic set-up in naturally colonized ditches. Additionally, we investigated toxicity by either feeding the damselfly nymphs with laboratory-cultured prey to have culture-fed specimens versus nymphs that hunted natural aquatic invertebrates to account for the potential indirect effects affecting the food quantity/quality that in turn can alter toxicity. Finally, we compared our toxicity results by monitoring the emergence of natural populations of I. elegans that were exposed to the same spikes of thiacloprid.

| Test species
We selected I. elegans (Zygoptera: Coenagrionidae) for multiple reasons. First and most importantly, it is the most common damselfly in large parts of Europe (Boudot & Kalkman, 2015), including the Netherlands (Van Swaay et al., 2018). Because of its high abundance and widespread distribution, I. elegans has the potential to be used as an indicator species for damselflies in general. Second, I. elegans is eurytopic (i.e., it can adapt to a wide range of ecological conditions) and found in a range of lowland fresh and brackish waters; it is often the dominant damselfly in highly anthropogenic habitats (Brockhaus et al., 2015). As such, this species is likely to be found in the vicinity of anthropogenic disturbances such as neonicotinoid runoff from the agricultural landscape. Third, because of its preference for slow flowing or standing water and the fact that they can be identified at an early instar, it is a suitable test species for caged bioassays. Fourth and finally, I. elegans typically lives in the water up to nearly a year and has a very long emergence period in the Netherlands, from late April till late September (NVL, 2002). Therefore, I. elegans has high ecological relevance for both the aquatic and terrestrial food webs as it is an active opportunistic hunter and also because it is common prey for invertebrates, fish, birds and bats (Corbet, 1962). population trends of this species, these results indicate neonicotinoids play a central role in the Odonate decline in general.

| Test location
The experiment was conducted at the outdoor research facility of Leiden University: 'the Living Lab' (The Netherlands). The site consists of 36 experimental ditches (1,750 L per ditch) with a length of 10 m, a width of 0.8 m and a depth of 0.3 m that are connected to an adjoining lake that provides a natural input of flora and fauna (see Appendix Figure S1 for a photographic overview). After a 5month colonization period and prior to the start of the experiment, these ditches were hydrologically isolated from the lake by placing a 1,000 × 500 × 2 mm acrylic plate firmly into the ditch banks and 15 cm deep into the sediment at the end of every ditch.

| In situ conditions
The damselfly nymphs used in the in situ experiment were caught from the adjoining lake of the experimental site by sweeping with a dipping net (mesh size 0.5 mm). The selected nymphs had an average length of 8 mm (SD 1.6 mm). Each nymph was placed in a cylindrical cage (ø: 5.5 cm, height: 6.8 cm), in an experimental ditch. Inoculation was performed 3 days prior to the start of the treatments (see Section 2.32.3) so that the nymphs could acclimatize to the cage. We also placed a thin twig in each cage for the nymphs to rest and hold onto.
All cages contained one nymph and each experimental ditch received two cages. These two cages were nearly identical; high-density polyethylene cages with a 3.75 cm opening on one side (in the lid) covered with mesh. One of the cages was closed with a larger mesh size of 1,000 µm and the other with a finer mesh size of 150 µm. Smaller invertebrates, such as Daphniidae, Cyclopidae and Diptera larvae, living in the experimental ditches could disperse freely in and out of the cages with the larger mesh, but not the finer. We did not actively feed the nymphs in the cages with the larger mesh. Instead, these nymphs fed on the natural invertebrates living in the experimental that freely dispersed into the cage. Nymphs in the cages with fine mesh were fed daily with <3 day old daphnids (Daphnia magna) retrieved from a longstanding culture. These two food conditions will hereafter be referred to as either the 'Free-feeding' or the 'Fed' treatment, respectively. During the acclimatization period, we fed the nymphs of the "Fed" treatment 5 daphnids per day. At the start of the experiment (the first thiacloprid spike, see Section 2.32.3), we started feeding the nymphs 10 daphnids daily and increased this number by 5 daphnids every week to fill the need of the growing nymphs, maintaining socalled ad libitum food conditions. However, if uneaten daphnids were observed, we subtracted the number of uneaten daphnids from the number of daphnids that we fed.

| Exposure conditions
After the initial acclimatization period of the nymphs, thiacloprid was spiked in the experimental ditches. Thiacloprid in Dutch surface waters is increasingly detected starting in April-May (see Barmentlo et al., 2018) with the highest numbers in June and July. This closely equals the season when I. elegans emerge. In May, we exposed the nymphs to four nominal concentrations of thiacloprid; 0 (control), 0.1, 1 and 10 µg/L with nine replicate ditches per concentration.
These concentrations have all been measured in Dutch surface waters in the period 2013-2017, with the lower concentrations being observed more commonly (see Table 1, Leiden University & Rijkswaterstaat-WVL, 2018;Vijver, Van 'T Zelfde, Tamis, Musters, & De Snoo, 2008). However, the octanol-water coefficient for thiacloprid is low (log K ow = 1.26 for thiacloprid, USEPA, 2003) and as such thiacloprid is removed rapidly of the aqueous phase and transported to the sediment. Consequently the chance of detecting the maximum concentrations (1.8% of all detects, as shown in Table 1) by grab sampling is low. The findings reported in Table 1  purity, purchased from Sigma-Aldrich) in demineralized water. These stocks were then diluted in a 10 L bottle with water originating from the ditch to which the thiacloprid was added and carefully poured into each ditch while spreading evenly. As >95% of the initial thiacloprid concentrations degrades after 2 weeks in our test system (DT90 = 11.1 days; shown by Barmentlo et al., 2018), a second spike was applied after two weeks in order to maintain experimental concentrations for one month. Water quality was monitored on a weekly basis by measuring water temperature, pH, oxygen concentration and conductivity using a portable hq 40 days electronic multiparameter meter (Hach). We also monitored thiacloprid concentrations over time of exposure by collecting water samples 5 cm below the surface level of each experimental ditch daily during the first week after a spike and three times a week thereafter. These samples were then analysed using liquid chromatography-tandem mass spectrometry (Agilent Technologies; see Roessink et al., 2013 for the detailed procedure). Note: Note that the detection limit was often approximately 10 ng/L.

| Caged individuals
All nymphs were monitored daily on survival, consumption and moulting. Nymphs of the 'Free-feeding' treatment ( Figure 1b) were also monitored for emergence while the 'Fed' treatment (Figure 1a) was terminated after six weeks of inoculation. Food consumption was monitored in the 'Fed' treatment by counting the number of eaten daphnids that were offered the previous day. For the 'Freefeeding' treatment, we counted all invertebrates that were present in the cage and determined them as prey or non-prey (as according to Thompson, 1978). Non-edible species (such as snails) were removed from the cages. Moulting was monitored by collecting and counting the shed chitin moults in the cages. When animals were nearly ready to emerge -indicated by the presence of fully developed and non-transparent wing-sheaths, 50% of the water in the cage was removed. The twig was replaced with a broader shoot of dry Typha latifolia, being a naturally preferred substrate for emergence of I. elegans.
Cages were returned to the experimental ditch at a slight downward angle so that the mesh remained completely covered with water, but an air pocket was created in the enclosure where the damselflies could emerge on the shoot of T. latifolia. Emergence was monitored daily for 5 weeks since the start of emergence. In addition to these daily observations, growth of the nymphs was determined on a weekly basis until the first emergence by temporarily and carefully removing the nymphs from their cages and photographing them on a Petri dish using an eScope DP-M17 USB-microscope camera. Body length was then determined from the top of the head to the start of the caudal gills using ImageJ (version 1.45S). Fleeing behaviour of the nymphs was tested on a weekly basis; swimming responsiveness was recorded from three weeks after the first spike of thiacloprid which was from the controls sixth instar onwards (Brochard & van der Ploeg, 2014). Swimming responsiveness was tested as activity occurring within 15 s of gentle stimulation using a plastic pipette, following the OECD guidance, for example, daphnids (OECD, 2012).

| Natural populations
To quantify the results from our controlled in situ setup to actual natural populations of I. elegans, natural emergence was monitored by placing one emergence trap above the middle of each ditch ( Figure 1c). The traps consisted of a 60:60:74 (length:width:height) pyramid-form stainless steel frame fitted with No-See-Um netting (300 holes per cm 2 ) and the invertebrate collection system as according to Cadmus, Pomeranz, and Kraus (2016). The emergence traps were tightly secured to bamboo sticks that were placed horizontally in the ditch banks to hold them in place during the entire F I G U R E 1 Graphical overview of the experimental setup. Shown is a lateral cross section of an experimental ditch with experiment (a) caged bioassay with Daphnia magna culture-fed nymphs, (b) caged bioassay with nymphs living of the natural food supply and (c) monitoring of emerged I. elegans of naturally colonized populations, exposed to four different concentrations of thiacloprid (0, 0.1, 1 and 10 µg/L). Note that invertebrates are not scaled to the equipment experimental period. The emergence traps were installed 3 cm below the water surface level so no emerging insects could escape the trap (or vice versa). Naturally emerging insects were subsequently continuously caught in 100 ml of 80% EtOH and collected from the trap two times per week (every Monday and Thursday) for the duration of three months. All emerged I. elegans were counted subsequently.

| Statistical analyses
We tested for the effect of thiacloprid on all different endpoints and the emergence of natural populations of I. elegans by using linear mixed-effect models (lme, package 'nlme') with the actual concentration of thiacloprid, time and their possible interaction as explanatory variables. As we remeasured the same individuals, we nested the nymphs within the respective ditch they resided in. We used a similar model for the emergence of the caged nymphs, but accounted for the binomial distribution of the data (glmer, package 'lme4'). For the models investigating growth, moulting and emergence we added the initial length per nymph, and all possible interactions, as this could potentially affect the growth rate and total number of moults. We assessed normality of the model and random variable residuals using QQ-plots and tested for sphericity using the function ezANOVA (package 'ez'). Prior to these analyses, we removed one individual from the data due to an accidental release. To further investigate the responsiveness of the nymphs in respect to mobility (swimming), we

| Treatments
Actual concentrations closely followed the nominally applied concentrations (see Table 2). Thiacloprid was rapidly removed from the aqueous phase as DT50 and DT90 values, calculated by first order kinetics, were 3.6 and 12.0 days, respectively. These values are very similar to those of the study of Barmentlo et al. (2018; DT50 = 3.3 and DT90 = 11.1 days), who also followed the degradation of thiacloprid in the same experimental ditches.  Table S1). These alterations in physicochemical characteristics were likely the result of altered primary production as also observed earlier in our experiments (see Barmentlo et al., 2018). The average water temperature was 22.4°C during the experimental period and was not influenced by thiacloprid (p > 0.05; see Appendix Table S1). In addition, oxygen concentrations were never below 9.65 mg/L, so no oxygen deficiency occurred.

| Effects on survival
Survival was 100% in all cases apart from one death individual in the 'Fed' as well as in the 'Free-feeding' treatment. Thus no significant effects of thiacloprid on survival were observed.

| Effects on consumption
The food consumption by the nymphs was clearly affected by thiacloprid. Nymphs that were fed consumed significantly less daphnids with increasing concentrations and this effect inter-

| Effects on growth and molting
We observed no effect of thiacloprid on the growth of nymphs in the 'Fed' treatment (p > 0.05, Figure 3a). However, growth of the nymphs in the 'Free-feeding' treatment was significantly delayed by increasing concentrations of thiacloprid interacting with time and the initial length per nymphs (marginal R 2 = 0.71, F 1,32 = 4.0, p = 0.047, Figure 3b).
The cumulative growth was slowest three weeks after the start of the treatments (one week after the second spike), by 28% at 9.86 µg/L thiacloprid. There was no significant difference in total growth at the end of the measurements (p > 0.05). Similar to the results for growth, we observed no effect of thiacloprid on the cumulative number of molts of the culture-fed nymphs (p > 0.05, see Appendix Figure S2a), while molting of nymphs in the 'Free-feeding' treatment was significantly delayed (interacting with the initial length; marginal R 2 = 0.61, F 1,32 = 19.4, p < 0.001, see Appendix Figure S2b). There was no significant difference in the total number of molts at the end of the measurements (p > 0.05).

| Effects on mobility
Swimming ability of the nymphs declined in both feeding treatments with increasing concentrations (Figure 4a,b). In nearly all cases, the animals that started swimming did so within 1-2s after gentle stimulation (observation not shown). The LOEC of nymphs within the 'Free-feeding' treatment was 0.84 µg/L (Dunnett's test: t = −3.1, p = 0.010), which was close to our obtained 50% Effect Concentration (EC 50 ) value of 1.04 µg/L (95% CI: 0.38-1.71). After this initial decline, the swimming ability increased again over time (marginal R 2 = 0.40, z = 2.2, p = 0.031, Figure 4a,b). The decline was weaker in the culture-fed nymphs, for which maximally 44% reduction in swimming ability at the highest test concentration was observed versus 100% in the 'Free-feeding' treatment. In addition, the EC 50 within the 'Fed' treatment was higher than the highest spike concentration (>9.7 µg/L) and no statistically significant LOEC could be determined.

| Effects on emergence
At the end of the experiment, 89% of all caged damselflies within the control treatment emerged (Figure 5a). Only one individual within the control did not emerge as this individual lost two of its caudal appendages during the experiment. These appendages were regrown during the course of the experiment but resulted in a delayed emergence. Exposure to thiacloprid significantly reduced the cumulative emergence of damselflies within the 'Freefeeding' treatment (marginal R 2 = 0.66, χ 2 = 7.9, df = 3, p = 0.047).

| D ISCUSS I ON
Many alarming reports are currently published on the decline of insects (e.g. Hallmann et al., 2017) with (agricultural) pollution often being recognized as one of the predominant drivers in intensively used environments (Dudgeon et al., 2006). This also holds true for neonicotinoids as it has been acknowledged that they are causing environmental risks higher than lab-based studies previously indicated (Barmentlo et al., 2018;EASAC, 2015;Goulson, 2013;Vijver et al., 2017). We have tested the effects of environmentally occurring thiacloprid concentrations on the Odonate I. elegans systematically using various endpoints, which allows to understand the mechanistic pathway of toxicity. If such eurytopic and widespread species is affected or even in decline, this is indicative that other Odonates are potentially threatened as well. We found clear effects of environmentally relevant neonicotinoid concentrations on both caged I. elegans and natural populations. All sublethal parameters selected were affected by thiacloprid to some degree, but the severity of effects was stronger in caged individuals feeding on natural food supplies compared to fed individuals. The severity of effects increased even more so when considering the emergence of natural populations, which decreased strongly due to neonicotinoid exposure in a dose-dependent manner.
In contrast with the observed strong effects for sublethal parameters, survival was not impacted in this study, which is consistent with current literature that indicate lethal effects of neonicotinoids on other Odonates at levels that are at least a threefold higher than our highest experimental concentration (Beketov & Liess, 2008;Jinguji et al., 2013;Sugita, Agemori, & Goka, 2018). nymphs had more natural prey leftover at the highest test concentration. These observations are underpinned by the mode of action of neonicotinoids to invertebrates, namely by blocking the nicotinic acetylcholine receptors within the nervous system (Simon-Delso et al., 2015). This can also explain the significant delays in both growth and moulting rate of the nymphs living of the natural food supplies. As well as that, these effects can also explain the reduced swimming performance of the nymphs. The observed strong decline in swimming ability can also be attributed to a physiological response; I. elegans nymphs are excellent swimmers, but are only able to do so starting from the sixth instar (Brochard & van der Ploeg, 2014). As all nymphs could eventually swim, it is likely that the delay in moulting subsequently led to a delay in reaching the sixth instar, making the nymphs temporally incapable of swimming. Irrespective of the causal mechanism, swimming behaviour of I. elegans and many other Odonate species is generally used to avoid predators such as fish or other invertebrates (Corbet, 1962). The temporarily reduced swimming capacity can thus have obvious negative effects for natural populations because the nymphs are more susceptible to predation. Fitness reduction throughout the life cycle of the damselflies can also explain the observed delay in emergence time in the 'Free-feeding' treatment. All our tested endpoints are connected through either energy transfer or a physiological response and hence the combined sublethal effects offer a mechanistic explanation for the delay in emergence of the damselflies.
Nymphs feeding on the natural food supplies consistently exhibited more toxic pressure than nymphs that were fed. This is shown by the number of endpoints that were affected; namely all sublethal endpoints for the 'Free-feeding' versus two for the 'Fed' nymphs.
For example, growth rate was reduced only within the 'Free-feeding' treatment. The severity of effects for endpoints that were affected in both treatments also differed markedly; swimming ability was more affected in the 'Free-feeding' treatment. Such differences in neonicotinoid-induced toxicity due to differences in food quantity/ quality has been observed earlier and explained as a compensatory feeding response to cope with toxic stress (see Alexander et al., 2013;Ieromina et al., 2014;Barmentlo et al., 2018). In our study, the food quantity/quality was indirectly determined by information on how many potential prey were not eaten on a daily basis. Strikingly, this number decreased by 24% at the lowest test concentration compared to the control within the 'Free-feeding' treatment, indicating that these nymphs were possibly feeding more and thus indicating compensatory feeding (Alexander et al., 2013;Barmentlo et al., 2018). Alternatively, a plethora of indirect effects can occur within the communities of the experimental ditches explaining the difference in toxicity between the treatments. The most obvious would be a secondary exposure to thiacloprid through contamination of the natural food supply. However, this is not that likely as neonicotinoids are highly soluble (USEPA, 2003) and thus generally considered to F I G U R E 5 Average cumulative emergence of Ischnura elegans (±SE, n = 8-9) per concentration of thiacloprid. (a) In situ 'Free-feeding' treatment, caged experiment and (b) natural collected individuals from the emergence traps. Note that for (a) x-axis was set to t = 30 for clarification purposes and no individuals emerged before this time and for (b) we caught 99% of the generations' emergence during the exposure period have a low bioaccumulation potential. More likely are thiaclopridinduced alterations in the communities residing in the experimental ditches, which altered the amount of natural food supply that entered the cages. Thus it is possible that we observed less prey in the cages at the lowest test concentration (0.1 µg/L) due to less supply from the ditch community rather than due to compensatory feeding. This is supported by the observation that the delay in emergence of caged individuals was greatest at this lowest test concentration.
In any case, the outperformance of the 'Fed' nymphs importantly showcases that optimal food conditions can alter neonicotinoid-induced toxicity. As the common laboratory-based ecotoxicological approach is to feed animals ad libitum (according all OECD regulations, for example OECD, 2012), the already scarce toxicity data for Odonates as collected in the laboratory setting may currently be underestimating the actual risks in the field.
The total emergence of the natural populations of I. elegans declined strongly (39%-65%) with increasing thiacloprid concentrations, whereas within our cage setup we mostly observed a delay in emergence time. The most likely explanation for this discrepancy is that in our cage setup many biotic interactions are excluded, for example, there was no predation pressure. These biotic interactions are obviously present within the natural damselfly populations. It is recognized that the effects of contaminants depend on ecological context (Clements, Hickey, & Kidd, 2012;Trekels, Meutter, & Stoks, 2011  veillance data for all Odonates shows an initial increase in Odonates between 1990 and 2008. This is likely because of improved habitat and water quality in that period (Termaat, Grunsven, Plate, & Strien, 2015). However the data also shows that total Odonate distribution in the Netherlands is steadily declining, like for I. elegans, since the late 2000s (see Appendix Figure S3). Given that our test species is one of the most widespread and eurytopic Odonates in Europe (Boudot & Kalkman, 2015), the observed neonicotinoid-induced toxicity may thus be indicative for this overall decline of Odonates.

| CON CLUS IONS
Clear effects of environmentally relevant concentrations of the neonicotinoid thiacloprid on the life cycle of the Odonate I. elegans were shown. While no direct effects on mortality were observed at environmental relevant concentrations, all sublethal endpoints tested were affected. Our results strongly depended on the food offered, which indicates that current laboratory assessments performed at ad libitum food underestimate neonicotinoid toxicity in the actual environment. In addition, it appears that even our realistic exposure scenario using caged individuals in experimental ditches also underestimates toxicity as the emergence of natural populations was more strongly affected. This is likely because biotic pressures such as predation add to toxicity and these pressures are not included within the caged experiment nor the common laboratory approaches. Finally, our observed reduced fitness during the nymph stage and the strong decline in natural emergence can be indicative for neonicotinoids adding to the ongoing I. elegans decline.

ACK N OWLED G EM ENTS
We thank Justin Knetsch, Jo-Anne Bartels and Janneke van der Horst for their assistance with the experimental work.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data is available via Dryad Digital Repository https ://doi.org/10. 5061/dryad.0vh187j (Barmentlo, Vriend, Grunsven, & Vijver, 2019).