Abstract
Predicting ecological effects of contaminants remains challenging because of the sheer number of chemicals and their ambiguous role in biodiversity-ecosystem function relationships. We evaluate responses of experimental pond ecosystems to standardized concentrations of 12 pesticides, nested in four pesticide classes and two pesticide types. We show consistent effects of herbicides and insecticides on ecosystem function, and slightly less consistent effects on community composition. Effects of pesticides on ecosystem function are mediated by alterations in the abundance and community composition of functional groups. Through bottom-up effects, herbicides reduce respiration and primary productivity by decreasing the abundance of phytoplankton. The effects of insecticides on respiration and primary productivity of phytoplankton are driven by top-down effects on zooplankton composition and abundance, but not richness. By demonstrating consistent effects of pesticides on communities and ecosystem functions and linking pesticide-induced changes in functional groups of organisms to ecosystem functions, the study suggests that ecological risk assessment of registered chemicals could be simplified to synthetic chemical classes or types and groups of organisms with similar functions and chemical toxicities.
Similar content being viewed by others
Introduction
Freshwater ecosystems are among the most biodiverse in the world1 and provide important ecosystem services2, yet many are imperiled by pesticide contamination3,4. Two major challenges, among many5, impede prediction of responses of freshwater ecosystems to pesticides. First, the extent to which individual pesticides have consistent effects on ecosystem functions and biodiversity is unknown. In the U.S. and Europe, tens of thousands of synthetic chemicals are registered, and in the U.S. >350 pesticides are applied annually6,7. If the effects of pesticides are consistent within ‘pesticide classes’ (those with similar chemical structures) or ‘pesticide types’ (those targeting similar pests), then the complexity in predicting impacts of pesticides could be markedly reduced8,9. Such consistency would improve efficiency of risk assessment and allow a greater focus on exceptions to general patterns. Second, the role of pesticides in biodiversity-ecosystem function relationships has not been elucidated10,11,12. Historically, random and direct manipulations of single-trophic level communities and measurement of associated ecosystem processes13,14 have established causality between biodiversity and ecosystem function15,16,17. However, this approach overlooks the importance of anthropogenic factors (e.g. climate change, nutrient enrichment, pesticide contamination), whose influences on communities are far from random18,19, alter multiple trophic levels14,20, and occur via direct and indirect pathways10.
In an effort to suggest improvements to risk assessment, the objectives of this study were to: (1) evaluate the consistency of effects across pesticide types, classes, and individual pesticides on ecosystem processes and communities, (2) assess whether the effects of pesticides on ecosystem processes and communities were the result of sublethal, non-target effects or changes in abundance of ‘targeted taxa’, and (3) test whether changes in composition, abundance, and/or richness of various functional groups mediate the effects of pesticides on ecosystem functions. We propose three hypotheses. First, ecosystem processes respond consistently to different pesticides within pesticide types because taxonomically related community members often have similar functional roles (redundancy) within the ecosystem. So, reductions in the abundance of taxa of a single group (e.g., green algae) might be specific to an individual pesticide or class, but these reductions would result in similar effects on ecosystem function overall (e.g., primary productivity)21. Second, communities respond consistently to pesticides within classes because of taxa-specific sensitivities to pesticides22,23. Third, disruptions in ecosystem processes caused by exposure to pesticides are mediated by changes to abundance, composition, and richness of functional groups.
Here, we show that ecosystem functions respond consistently to herbicides and insecticides, while communities respond somewhat less consistently. Pesticide-induced effects on ecosystem functions are driven by changes in the abundance and community composition of functional groups of organisms. For instance, herbicides reduced respiration and primary productivity by decreasing the abundance of phytoplankton, a bottom-up effect. Insecticides increased primary production of phytoplankton and respiration through top-down effects on zooplankton composition and abundance, but not richness. Our results suggest that predictions of the complex effects of pesticides on aquatic ecosystems can be simplified by considering effects of pesticide classes or types on groups of organisms with similar functions and chemical toxicities.
Results and discussion
Experimental design
While we recognize that an enormous challenge to predicting the effects of synthetic chemicals on complex natural systems is to understand the effects of mixtures of synthetic chemicals, this study focuses on the effects of single pesticides because scientists do not yet fully understand the effects of even single pesticides on complex ecosystems3,24,25,26. In addition, federal agencies are charged with evaluating ecological safety of synthetic chemicals one at a time as they come to market or need to be re-evaluated. Therefore, a framework must first be developed for understanding the effects of single synthetic chemicals before reliable predictions can be developed for the responses of communities and ecosystems to chemical mixtures27. Mesocosm studies are an efficient approach to toxicity testing as they provide toxicity data on multiple species simultaneously under environmentally realistic conditions. As such, we conducted a large-scale experiment using 72 outdoor mesocosms to evaluate the effects of two control treatments (water and solvent), four simulated-pesticide treatments, and 12 pesticides on tri-trophic temperate pond communities (Fig. 1a, b). The pesticide treatments were nested in four classes (organophosphates, carbamates, chloroacetanilides, and triazines) and two types (insecticides and herbicides) (Fig. 1a). The four pesticide classes in this study are representative of some of the most commonly used7 and detected pesticide classes in the US28,29. To represent pesticide runoff following rainfall, pesticides were applied singly at the beginning of the experiment at standardized environmentally relevant concentrations calculated using U.S. EPA software’s GENEEC v2 (see “Methods” section). Simulated-pesticide treatments were top-down or bottom-up food web manipulations that attempted to mimic direct (i.e. lethal) effects of actual herbicides and insecticides on algae and zooplankton abundances, respectively.
Effects of pesticide types on ecosystem function
Pesticide type explained 46% of the variation in ecosystem function associated with the pesticide treatments (Supplementary Table 1). Herbicides were associated with a decrease in primary production of phytoplankton that led to increased production of attached periphyton (benthic algal biofilms), probably through an increase in light availability (Fig. 1c–e). In addition, as production of phytoplankton decreased in response to herbicide exposure, respiration decreased. (Fig. 1c-e). These patterns are described in more detail in a structural equation model below. Herbicide exposure also lead to a decrease in pH (an increase in acidity), which might reflect the release of dissolved inorganic carbon as a result of decomposing phytoplankton.
In contrast to herbicide-exposed systems, insecticide-exposed systems exhibited an increase in production of phytoplankton, whose growth in the water column reduced light penetration, thereby shading and reducing the primary production and biomass of benthic periphyton (Fig. 1c–e), an effect shown in other studies22. Increases in phytoplankton were likely driven by replacement of cladocerans by copepod zooplankton, the latter of which are less efficient phytoplankton grazers (described below). The corresponding increase in phytoplankton also lead to a subsequent increase in respiration (Fig. 1c–e). As phytoplankton production increased, pH increased (acidity decreased), a possible result from an increase in phytoplankton removing inorganic carbon from the water column. While some variation in ecosystem responses was also explained by individual pesticides, it was small relative to variation explained by pesticide type (12 vs. 46% of variation in ecosystem responses, Supplementary Table 1). Decomposition of leaf litter did not appear to be strongly influenced by either herbicides or insecticides (Fig. 1c–e).
Effects of pesticide types and classes on communities
We tested for the effects of individual pesticides, classes, and types separately on the single-trophic-level zooplankton community (six zooplankton genera) and on the tri-trophic community (insect and salamander predators, snail and anuran herbivores, and periphyton and phytoplankton primary producers). Similar to ecosystem function, pesticide type explained the majority of the variance (44.2%) in the zooplankton community, followed by pesticide class (18.8%) (Supplementary Table 1). Distance-based redundancy analysis (dbRDA) showed that: (1) herbicide-treated and insecticide-treated mesocosms had distinct zooplankton communities, (2) within their respective pesticide types, organophosphate insecticides, chloroacetanilide herbicides, and triazine herbicides caused further distinction in zooplankton communities, and (3) there was relatively high multivariate dispersion within the carbamate class (Fig. 2a, b). In response to insecticides, cladoceran zooplankton experienced high mortality and were virtually eliminated, which perhaps led to competitive release of copepods (Fig. 2a–c)30. Cladocerans are more efficient phytoplankton grazers than copepods31, so it stands to reason that their declines potentially drove an increase in the relative abundance of phytoplankton in these treatments (Fig. 1). In contrast to the changes in community composition associated with insecticides, herbicides decreased zooplankton abundance with no apparent change in composition (Supplementary Fig. 3), likely by reducing phytoplankton (i.e., bottom-up effect32,). The stronger bottom-up effect of triazines compared to chloroacetanilide herbicides on zooplankton was probably because of longer environmental persistence (soil half-lives 110–146 days vs. 14–26 days, respectively [Pesticide Action Network Pesticide Database]). Thus, consistent with the ecosystem function results, these findings on the zooplankton community suggest that ecological risk assessment can be largely simplified to generalized effects of pesticide type or class.
In the tri-trophic community, variation explained by pesticides was about equally distributed among type, class, and individual pesticide (Supplementary Table 1). The dbRDA showed that: (1) herbicide-treated and insecticide-treated mesocosms had distinct responses of community members, (2) within herbicides, triazines classes caused further distinction in communities, and (3) there was relatively high multivariate dispersion in communities exposed to carbamate and organophosphate insecticides (Fig. 3a, b). Overall, survival of predators was low with insecticides33, except for aldicarb (Fig. 3a-c). Amphibian and snail prey generally had greater positive responses to insecticides compared to controls or herbicides (Fig. 3a-c), suggesting these organisms benefitted from predatory release, a trend found in other studies34. While multigenerational exposure of organisms to pesticides has the potential to alter vulnerability of organisms over time, it seems unlikely that it played a large role in our study given the length of the experiment, the persistence of the pesticides, and the generation time of the focal organisms of study (see Supplementary Discussion). In addition, none of the patterns on community or ecosystem-level responses could be explained by structural similarities among pesticide classes (see Supplementary Discussion).
Next, we used physical manipulation of communities in the application of simulated pesticide treatments in order to evaluate how direct vs. indirect effects of pesticides on aquatic communities influenced ecosystem functions. However, rarely were these treatments similar to actual pesticide classes (pair-wise comparisons Figs. 1–3). This lack of similarity was presumably because of difficulties in sustaining manipulations in the simulated pesticide treatments that matched the magnitude and specificity of actual pesticides on species with short generation times (see Supplementary Discussion and Supplementary Fig. 3 for details). While manipulating species composition has been critical historically in the study of biodiversity-ecosystem function, the same approaches are not well-suited for (1) species that can exhibit population dynamics in the timescale of an experiment because they may quickly recover from a manipulation and (2) studies of disturbance because matching the complexity of the effects of actual disturbance is challenging.
Linking pesticide induced changes in communities to ecosystem function
Finally, we used structural equation models acting on the natural variation created by pesticide exposures to evaluate how composition, abundance, and richness of different functional groups of organisms influenced whole-system respiration and primary production of phytoplankton and periphyton. Our results show that herbicide exposure decreased respiration and primary productivity through bottom-up effects; herbicides decreased the abundance of phytoplankton, which in turn drove respiration (Fig. 4a). In contrast, insecticides increase primary productivity of phytoplankton and respiration through top-down effects on zooplankton composition and abundance, but not richness (Fig. 4b).
Given that ecosystem functions respond more consistently to pesticides than communities and pesticide-induced changes to ecosystem functions were at times mediated by changes in abundance and composition of functional groups of organisms, the observed consistency in the responses of ecosystem functions to pesticides within pesticide types could be driven by functional redundancies of species. Further supporting this point, when we grouped responses of taxa by their functional roles in the community (algae, herbivores, and predators), the variance explained by pesticide type (29%) was nearly doubled compared to the variance accounted for by either pesticide class (17.6%) or individual pesticide (17.3%) (Supplementary Table 1). These results suggest that the complexity in predicting the effects of pesticides on communities could be reduced by evaluating responses of functional groups to pesticide classes or types rather than evaluating the responses of species or genera to individual pesticides. While this outcome presents an opportunity to simplify predictions, it also is a sobering illustration of the blanket detrimental effects pesticides can have on complex ecological systems.
Our results suggest that ecological risk assessment by regulatory agencies, made complex by tens of thousands of synthetic chemicals and diverse species assemblages, could be advanced in two ways. First, regulators could extend this study to additional pesticides and develop predictions based on consistencies of effects within chemical classes or types25. Additional pesticides of different types and classes might produce different sequences of direct and indirect effects of varying magnitudes on aquatic ecosystems. Nonetheless, we predict that, on average, ecosystem and community effects would still be consistent within pesticide classes and types, with more variation being explained by types. The generality of effects by class and type would be driven by the nestedness of the pesticides’ biologic activity; molecular modes of action are shared within classes (e.g., carbamates inhibit acetylcholinesterase) and targets in the environment are shared within types (e.g., insecticides are active against insects). A similar approach of using chemical structures of synthetic chemicals to predict toxicities is taken by Quantitative Structure-Activity Relationships (QSAR) analyses35. QSAR analyses on the present herbicides and insecticides did not recreate the same groupings of pesticides by class and by type as the current analyses (see Supplementary Discussion). The QSAR results were likely not similar to the present results because QSAR analyses are based on predictions from model organisms, which assumes that the responses of the model species are correlated with responses of all organisms that they are selected to represent. This assumption is problematic because of taxa-specific sensitivities to synthetic chemicals. In addition, QSAR analyses ignore species interactions and indirect effects, which we show are important considerations for predicting community-level and ecosystem-level responses.
Second, regulators could extend this study to additional communities and ecosystems and develop predictions based on the consistencies of effects within taxonomic or functional groups of organisms. Testing synthetic chemicals against entire communities and ecosystems incorporates natural complexity and allows for the evaluation of both direct and indirect effects. While we observed that organisms within functional groups organisms share similar direct and indirect toxicities to pesticide classes and types, it is possible that the genetic makeup36, community complexity and composition37, presence of other stressors38, and previous exposure history39 could at times obscure the generality of the effects of pesticides when additional communities and ecosystems are considered. Nevertheless, the addition of standardized tests across communities and ecosystems would allow regulators to assess just how important these factors are to their regulatory decisions. Although we fully expect that there could be occasional exceptions to the general patterns revealed and proposed in this study, by simplifying risk assessment in the ways we suggest, more time and resources would be available from regulatory agencies to detect any exceptions to these general patterns.
Methods
Experimental design and community composition
We conducted a randomized-block experiment at the Russell E. Larsen Agricultural Research Center (Pennsylvania Furnace, PA, USA) with replicated mesocosm ponds. Mesocosms were 1100-L cattle tanks covered with 60% shade cloth. The spatial block was distance from a tree line in our mesocosm field. Three weeks before pesticide application, these mesocosms were filled with 800 L water, 300 g mixed hardwood leaves, and inoculations of zooplankton, periphyton, and phytoplankton homogenized from four local ponds. Just before pesticide application on the same day, each tank received two snail, three larval anuran, one larval dragonfly, one water bug, one water beetle, one larval salamander, and one backswimmer species (11 Helisoma (Planorbella) trivolvis, 10 Physa gyrina; 20 Hyla versicolor, 20 Lithobates palustris, 20 Lithobates clamitans; 2 Anax junius; 2 Belostoma flumineum; 5 Hydrochara sp.; 3 Ambystoma maculatum; 6 Nototeca undulata) (Fig. 1b). These community members naturally coexist and were applied at naturally occurring densities40. Initial conditions of some mesocosms varied in simulated pesticide treatments (see below).
We randomly assigned 18 treatments (12 pesticides, 4 simulated pesticides, 2 controls) with four replicate mesocosms of each treatment, which resulted in 72 total mesocosms (Fig. 1a). The 12 pesticide treatments were nested; we included two pesticide types (insecticide, herbicide), two classes within each pesticide type (organophosphate insecticide, carbamate insecticide, chloroacetanilide herbicide, triazine herbicide), and three different pesticides in each of four classes (Fig. 1a). To represent runoff of pesticides into freshwater systems following a rainfall event, we applied single doses of technical grade pesticides at environmentally relevant concentrations at the beginning of the experiment. To ensure our exposures represented environmental relevance, we used estimated environmental concentrations of pesticides, calculated by U.S. Environmental Protection Agency’s GENEEC v2 software, Supplementary Table 2). Our design also included water and solvent (0.0001% acetone) controls (Fig. 1a). Pesticides were obtained from ChemService (West Chester, PA, USA). Nominal concentrations of pesticides (μg/L) were: 64 chlorpyrifos, 101 malathion, 171 terbufos, 91 aldicarb, 219 carbaryl, 209 carbofuran, 123 acetochlor, 127 alachlor, 105 metolachlor, 102 atrazine, 202 simazine, and 106 propazine. We collected composite water samples 1 h after application to mesocosms and shipped samples on ice to Mississippi State Chemical Laboratory to verify these nominal concentrations. Measured concentrations of pesticides (μg/L) were: 60 chlorpyrifos, 105 malathion, 174 terbufos, 84 aldicarb, 203 carbaryl, 227 carbofuran, 139 acetochlor, 113 alachlor, 114 metolachlor, 117 atrazine, 180 simazine, and 129 propazine.
The four simulated pesticide treatments were top-down or bottom-up food web manipulations intended to mimic effects of actual herbicides and insecticides on community members. These manipulations occurred once and were concurrent with the timing of pesticide applications. Top-down and bottom-up simulated insecticide treatments were designed to reduce densities of zooplankton, simulating effects of insecticides on zooplankton survival. For top-down simulated insecticides, we doubled the densities of zooplankton predators by including six total A. maculatum larval salamanders and 12 N. undulata backswimmers per mesocosm. For bottom-up simulated insecticides (i.e., direct manipulation of a lower arthropod trophic level), we removed zooplankton with a net. Top-down and bottom-up simulated herbicides were designed to reduce algae, simulating effects of herbicides on survival and growth of algae. For top-down simulated herbicides, we doubled the densities of large herbivores to increase grazing pressure by including 22 H. trivolvis snails, 20 P. gyrina snails, 40 H. versicolor larval anurans, 40 L. palustris larval anurans, and 40 L. clamitans larval anurans per mesocosm. For bottom-up simulated herbicides, we covered mesocosms in three sheets of 60% shade cloth in an attempt to block light and reduce photosynthesis. The experiment ran for four weeks, from June to July.
Measurements of experimental responses
During the experiment, we sampled periphyton using clay tiles (100 cm2) oriented perpendicularly along the bottom of the mesocosm. Each mesocosm had two periphyton measurements: ‘inaccessible periphyton’ taken from caged clay tiles that excluded herbivores and ‘accessible periphyton’ taken from clay tiles that were uncaged, allowing herbivore access. We sampled phytoplankton from water samples taken 10 cm below the water surface. Periphyton was scrubbed from tiles and phytoplankton from water samples (10 mL) were filtered onto glass fiber filters (under low vacuum pressure, <10 psi; Whatman EPM 2000, 0.3 μm, 47 mm) to estimate associated chlorophyll concentrations. The chlorophyll concentration of each filter was determined using an organic extraction procedure with a 50:50 mixture of 90% acetone to DMSO. We measured chlorophyll-a concentrations using a standard fluorometric technique. We scored water clarity, a metric of light availability, on a scale from one (clear) to five (opaque) blinded to treatment. We measured pH and dissolved oxygen (DO) at dusk and dawn on subsequent days using hand-held meters (YSI, Yellow Springs, OH, USA). We measured decomposition by taking the dry mass of hardwood leaf packets in each mesocosm at the beginning and the end of experiment. In addition, we sampled snail egg masses and hatchlings using two rectangular pieces of Plexiglass (465 cm2) in each mesocosm, one hung on the side and one on the bottom of the mesocosm. Zooplankton were collected from the entire water column by placing a PVC pipe (10 cm diameter, 60 cm height) upright in the center of each tank, capping the bottom, and pouring the water through a 20 μm Nitex mesh. We collected two samples of zooplankton from each mesocosm, and we combined and preserved the samples in 70% ethanol. Zooplankton were counted and identified in 5 mL subsamples for each mesocosm using a zooplankton counting wheel (Wildlife Supply Company, Yulee, FL, USA) and a dissecting microscope. At the end of the experiment, mesocosms were drained, and the remaining animals were counted, euthanized, and preserved. Two previous manuscripts, which use the same design as the current manuscript, also describe this experimental design and methods8,41. The research was reviewed and approved by the Penn State University Institutional Animal Care and Use Committee.
Statistical analyses
To test for the consistency of effects of type, class, and individual pesticide on aquatic ecosystem processes and communities and to attribute the variation explained to each pesticide level of organization while accounting for the nested structure of our experimental design (Fig. 1a), we completed permutational analyses of variance (PERMANOVA). For nested PERMANOVA models, the predictors were the following random categorical terms: type (insecticide, herbicide), class (carbamate, organophosphate, chloroacetanilide, triazine) nested within type, and pesticide (12 in total) nested within class within type. These models did not include controls or simulated pesticides because these treatments were not hierarchically nested (Fig. 1a). We evaluated 9999 permutations using residuals under a reduced model. Following nested PERMANOVAs, we used pair-wise multiple comparisons tests using PERMANOVAs to evaluate differences among controls, organophosphates, carbamates, top-down simulated insecticides, bottom-up simulated insecticides, chloroacetanilides, triazines, top-down simulated herbicides, and bottom-up simulated herbicides. In these pair-wise comparisons, we evaluated 9999 unrestricted permutations of raw data. All PERMANOVAs also included spatial block as a random predictor to account for variation in sunlight associated with distance from a tree line. Preliminary analyses showed that exclusion of the block did not change the results. In all PERMANOVAs, test statistics associated with Type III partial sums of squares were evaluated.
We conducted four nested PERMANOVAs. Our first nested PERMANOVA focused on ecosystem processes and included the following responses: pH taken at dawn, respiration (the difference between dissolved oxygen at dusk and dissolved oxygen at dawn of the subsequent day), decomposition (percent mass remaining of hardwood leaf packets), turbidity (water clarity scores from 1 to 5), and densities of phytoplankton and accessible periphyton (measured via chlorophyll-a). Since this analysis was focused on ecosystem-level responses, for periphyton we include accessible periphyton and not inaccessible periphyton because accessible periphyton better encompasses the totality of biologic (e.g. herbivore predation, shading from phytoplankton) factors that could influence periphyton biomass at the ecosystem level. Preliminary analyses included both accessible and inaccessible periphyton, and the results did not differ. The resemblance matrix for these responses was constructed using a Euclidean distance matrix of log-transformed and normalized values.
Our second and third nested PERMANOVAs focused on community structure. We separated community members into two statistical models based on the forms of response variables; those whose response variables were densities based on counts (zooplankton) and those community members whose response variables were survival, mass, reproductive rates, or density abstracted from chlorophyll measurements (insect predators, snail and tadpole herbivores, and algae; termed the tri-trophic community). The multivariate response for the zooplankton community included densities of Daphnia, Diaphanasoma, Chydorus, Bosmina, Diaptomus, and Cyclops. Zooplankton community analyses were based on square-root transformed densities using Bray-Curtis similarities. The multivariate response for the tri-trophic community model included: survival (0 to 1) of all amphibian, snail, and insect community members; average masses of surviving individuals for each amphibian species and H. trivolvis snails; average number of hatchlings and eggs per surviving H. trivolvis snail; and densities of phytoplankton and accessible and inaccessible periphyton. In this analysis, we include both accessible and inaccessible periphyton to account for any differential effects of excluding herbivores on the abundance of periphyton. Mass and reproductive rates were standardized to the number of surviving individuals to account for the different densities added to each tank at the beginning of the study (i.e. extra herbivores in top-down simulated herbicide treatment and extra predators in bottom-up simulated insecticide treatment). Finally, our fourth nested PERMANOVA evaluated a simplified tri-trophic community. We simplified the tri-trophic community responses into three functional roles within the community: algae, herbivores, and predators. Tri-trophic community responses of individua taxa were transformed and normalized as described previously, and then they were averaged according to functional group. We averaged densities of periphyton and phytoplankton into a single “algae” response, all amphibian and snail responses into a single “herbivore” response, and all insect and salamander responses into a single “predator” response. The simplified tri-trophic community model was based on Euclidean distances.
To visualize consistency of effects within type, class, and individual pesticides on multivariate ecosystem and community responses and to compare pesticide effects to simulated pesticides and controls, we used distance-based redundancy analyses (dbRDA) and two-way cluster diagrams. The dbRDAs were based on appropriate resemblance matrices for ecosystem and community responses as described above. The underlying categorical predictors in all models included: the spatial block, organophosphate, carbamate, chloroacetanilide, triazine, top-down simulated insecticide, bottom-up simulated insecticide, top-down simulated herbicide, bottom-up simulated herbicide, and control. In the dbRDA plots, we show the centroid values for the 18 experimental replicates.
As an alternative to the dbRDAs presented in the main text, we also visualized the consistency of effects within type, class, and individual pesticide on ecosystem, tri-trophic community, and zooplankton responses and compared pesticide effects to simulated pesticides and controls, using principal coordinates analyses (PCoA) (Supplementary Figs. 1, 2, 4). PCoAs were based on appropriate resemblance matrices as described previously. PCoAs were conducted in PERMANOVA+ for PRIMER and resulting data were exported. Point and vector plots were made using the exported data and the ‘ggplot2’ package in R. Ellipses on point plots represent 95% confidence intervals of groups based on standard errors and were made using the ordiellipse function in the ‘vegan’ package.
For the two-way cluster diagrams, clusters of pesticide treatments were based on centroid distances of the appropriate resemblance matrices. Clusters of multivariate responses were based on Euclidean distance resemblance matrices of averaged treatment responses. Before averaging, ecosystem and community responses were transformed and normalized as described previously. In clustering of treatments and responses, the cluster mode was the group average. In the PERMANOVAs for tri-trophic community responses, the effect of block was significant (Supplementary Table 1). Thus, we accounted for the effect of block by taking the residuals of simple linear regressions with individual tri-trophic community responses as the independent variable and block as the predictor in the generation of the shaded values of the two-way cluster diagrams. Then, we averaged these block-adjusted treatment responses with the ‘shade plot’ function in PRIMER. For the ecosystem responses and zooplankton community, the effect of block was not significant in the PERMANOVA model (Supplementary Table 1), so shaded values of the two-way cluster diagrams were simply the averaged treatment responses. All PERMANOVA models, pair-wise comparisons, dbRDAs, and two-way cluster diagrams were executed using PERMANOVA+ for PRIMER version 7 (PRIMER-E Ltd, Plymouth, UK). For ease of visualization of dbRDA and PCoA plots, data from PERMANOVA+ for PRIMER were exported, and plots were made using ‘ggplot2’ package in R.
To quantitatively test whether changes in composition, abundance, and/or richness of different functional groups mediate the effects of herbicides and insecticides on ecosystem function, we performed structural equation modeling and model comparison using the ‘piecewiseSEM’ package. We conducted separate analyses for insecticides and herbicides. Global models, which replicated the structures of periphyton and phytoplankton food webs (Fig. 1b), evaluated how herbicides or insecticides influence ecosystem functions through changes in composition, abundance, and richness of: (1) snail and tadpoles predators, (2) snails and tadpoles, (3) zooplankton predators, and (4) zooplankton. Ecosystem functions included primary production of accessible periphyton, primary production of phytoplankton, and whole system respiration. We did not include decomposition because ordination analyses showed decomposition was not dramatically influenced by pesticide exposures. Since this analysis was focused on ecosystem-level responses, we included accessible periphyton only (i.e. excluded inaccessible periphyton) because accessible periphyton better encompasses the totality of biologic (e.g. herbivore predation, shading from phytoplankton) factors that could influence periphyton biomass at the ecosystem level. For each functional group, composition was the locations of species or genera along the first axis of a principal coordinates analysis (PCoA) using a Bray-Curtis similarity matrix. Abundance was the total number of individuals of a functional group in a mesocosm, and richness was the number of species or genera within a functional group. The relationships among these variables in the global models are presented in Supplementary Fig. 5. Notably for herbicides, the global model considered the bottom-up direct effects of herbicides on periphyton and phytoplankton and the non-target direct effects of herbicides on zooplankton composition, abundance, and richness. Mesocosms treated with chloroacetanilide herbicides were excluded from the analysis because they did not influence algal densities (Fig. 3c). For insecticides, the global model considered the top-down direct effects of insecticides on composition, abundance, and richness of tadpole and snail predators, zooplankton predators, and zooplankton. These global models did not consider non-target direct effects of insecticides and herbicides on tadpoles and snails because community tri-trophic community analyses (Fig. 3) showed little evidence of non-target effects on these organisms. Phytoplankton, periphyton, and abundance of functional groups were all log-transformed in models. To facilitate comparisons among responses and clarify relationships among predictors, we reduced the global models by removing non-significant paths. The final models are presented in Fig. 4. All data supporting the results of the present manuscript are publicly available42. Full information regarding creation and attribution of silhouettes used throughout the main text and supplementary figures is included in Supplementary Table 3.
Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
Data availability
Data used in the generation of pesticide concentrations used in the present experiment are available within supplementary information. All other relevant data supporting the findings of this study are available at Figshare (https://doi.org/10.6084/m9.figshare.13224587).
Change history
14 January 2021
A Correction to this paper has been published: https://doi.org/10.1038/s41467-020-20854-1.
References
Dudgeon, D. et al. Freshwater biodiversity: importance, threats, status and conservation challenges. Biol. Rev. Camb. Philos. Soc. 81, 163–182 (2006).
Baron, J. S. et al. Meeting ecological and societal needs for freshwater. Ecol. Appl. 12, 1247–1260 (2002).
Bernhardt, E. S., Rosi, E. J. & Gessner, M. O. Synthetic chemicals as agents of global change. Front. Ecol. Environ. 15, 84–90 (2017).
Stehle, S. & Schulz, R. Agricultural insecticides threaten surface waters at the global scale. Proc. Natl Acad. Sci. USA 112, 5750–5755 (2015).
Liess, M., Schäfer, R. B. & Schriever, C. A. The footprint of pesticide stress in communities—Species traits reveal community effects of toxicants. Sci. Total Environ. 406, 484–490 (2008).
Baker, N. T. & Stone, W. W. Estimated Annual Agricultural Pesticide Use for Counties of the Conterminous United States, 2008-12: U.S. Geological Survey Data Series 907, 9 (U.S. Geological Survey, Reston, 2012).
Thelin, G. P. & Stone, W. W. Estimation of Annual Agricultural Pesticide Use for Counties of the Conterminous United States, 1992–2009: USGS Scientific Investigation Report 2013–5009, 54 (U.S. Geological Survey, Reston, 2019).
Rumschlag, S. L. et al. Effects of pesticides on exposure and susceptibility to parasites can be generalised to pesticide class and type in aquatic communities. Ecol. Lett. 22, 962–972 (2019).
Rohr, J. R., Kerby, J. L. & Sih, A. Community ecology as a framework for predicting contaminant effects. Trends Ecol. Evol. 21, 606–613 (2006).
McMahon, T. A. et al. Fungicide-induced declines of freshwater biodiversity modify ecosystem functions and services. Ecol. Lett. 15, 714–722 (2012).
Peters, K., Bundschuh, M. & Schäfer, R. B. Review on the effects of toxicants on freshwater ecosystem functions. Environ. Pollut. 180, 324–329 (2013).
Eisenhauer, N. et al. No interactive effects of pesticides and plant diversity on soil microbial biomass and respiration. Appl. Soil Ecol. 42, 31–36 (2009).
De Laender, F. et al. Reintroducing environmental change drivers in biodiversity–ecosystem functioning research. Trends Ecol. Evol. 31, 905–915 (2016).
Tilman, D., Isbell, F. & Cowles, J. M. Biodiversity and ecosystem functioning. Annu. Rev. Ecol. Evol. Syst. 45, 471–493 (2014).
Cardinale, B. J. et al. Effects of biodiversity on the functioning of trophic groups and ecosystems. Nature 443, 989 (2006).
Chase, J. M. & Leibold, M. A. Spatial scale dictates the productivity–biodiversity relationship. Nature 416, 427–430 (2002).
Hooper, D. U. et al. A global synthesis reveals biodiversity loss as a major driver of ecosystem change. Nature 486, 105–108 (2012).
Suding, K. N. et al. Scaling environmental change through the community-level: a trait-based response-and-effect framework for plants. Glob. Change Biol. 14, 1125–1140 (2008).
Díaz, S. et al. Functional traits, the phylogeny of function, and ecosystem service vulnerability. Ecol. Evol. 3, 2958–2975 (2013).
Petchey, O. L., McPhearson, P. T., Casey, T. M. & Morin, P. J. Environmental warming alters food-web structure and ecosystem function. Nature 402, 69 (1999).
Fleeger, J. W., Carman, K. R. & Nisbet, R. M. Indirect effects of contaminants in aquatic ecosystems. Sci. Total Environ. 317, 207–233 (2003).
Hua, J. & Relyea, R. Chemical cocktails in aquatic systems: pesticide effects on the response and recovery of >20 animal taxa. Environ. Pollut. 189, 18–26 (2014).
Ippolito, A., Todeschini, R. & Vighi, M. Sensitivity assessment of freshwater macroinvertebrates to pesticides using biological traits. Ecotoxicology 21, 336–352 (2012).
Gessner, M. O. & Tlili, A. Fostering integration of freshwater ecology with ecotoxicology. Freshw. Biol. 61, 1991–2001 (2016).
Rohr, J. R., Salice, C. J. & Nisbet, R. M. The pros and cons of ecological risk assessment based on data from different levels of biological organization. Crit. Rev. Toxicol. 46, 756–784 (2016).
Burton, G. A., Di Giulio, R., Costello, D. & Rohr, J. R. Slipping through the cracks: why is the U.S. environmental protection agency not funding extramural research on chemicals in our environment? Environ. Sci. Technol. 51, 755–756 (2017).
Halstead, N. T. et al. Community ecology theory predicts the effects of agrochemical mixtures on aquatic biodiversity and ecosystem properties. Ecol. Lett. 17, 932–941 (2014).
Stehle, S. et al. Aquatic pesticide exposure in the U.S. as a result of non-agricultural uses. Environ. Int. 133, 105234 (2019).
Stone, W. W., Gilliom, R. J. & Ryberg, K. R. Pesticides in U.S. streams and rivers: occurrence and trends during 1992–2011. Environ. Sci. Technol. 48, 11025–11030 (2014).
Brock, T. C. M., van Wijngaarden, R. & Van Geest, G. J. Ecological Risks of Pesticides in Freshwater Ecosystems; Part 2: Insecticides. (Alterra, Wageningen) https://www.researchgate.net/publication/40138928_Ecological_risks_of_pesticides_in_freshwater_ecosystems_Part_2_insecticides. (2000).
Sommer, U. & Sommer, F. Cladocerans versus copepods: the cause of contrasting top-down controls on freshwater and marine phytoplankton. Oecologia 147, 183–194 (2006).
Brock, T. C. M., Lahr, J. & van den Brink, P. J. Ecological Risks of Pesticides in Freshwater Ecosystems; Part 1: Herbicides. (Alterra, Wageningen) https://library.wur.nl/WebQuery/wurpubs/313292. (2000).
Schäfer, R. B. et al. Effects of pesticides monitored with three sampling methods in 24 sites on macroinvertebrates and microorganisms. Environ. Sci. Technol. 45, 1665–1672 (2011).
Peacor, S. D. & Werner, E. E. Predator effects on an assemblage of consumers through induced changes in consumer foraging behavior. Ecology 81, 1998–2010 (2000).
Mazzatorta, P., Benfenati, E., Lorenzini, P. & Vighi, M. QSAR in ecotoxicity: an overview of modern classification techniques. J. Chem. Inf. Comput. Sci. 44, 105–112 (2004).
Coors, A., Vanoverbeke, J., De Bie, T. & De Meester, L. Meester, Land use, genetic diversity and toxicant tolerance in natural populations of Daphnia magna. Aquat. Toxicol. https://doi.org/10.1016/j.aquatox.2009.08.004. (2009).
Rohr, J. R. & Crumrine, P. W. Effects of an herbicide and an insecticide on pond community structure and processes. Ecol. Appl. 15, 1135–1147 (2005).
Rumschlag, S. L. & Rohr, J. R. The influence of pesticide use on amphibian chytrid fungal infections varies with host life stage across broad spatial scales. Glob. Ecol. Biogeogr. 27, 1277–1287 (2018).
Oziolor, E. M., De Schamphelaere, K. & Matson, C. W. Evolutionary toxicology: meta-analysis of evolutionary events in response to chemical stressors. Ecotoxicology 25, 1858–1866 (2016).
Travis, J. & Wilbur, H. M. Ecological Communities: Conceptual Issues and Evidence. 113–122 (Princeton, New Jersey, 1984).
Rohr, J. R. et al. Agrochemicals increase trematode infections in a declining amphibian species. Nature 455, 1235–1239 (2008).
Rumschlag, S. L. et al. Dataset: consistent effects of pesticides on community structure and ecosystem function in freshwater systems. figshare. https://doi.org/10.6084/m9.figshare.13224587.v1. (2020).
Acknowledgements
Funding was provided by the National Institutes of Health (R01GM109499, R01TW010286-01), the National Science Foundation (EEID-1518681, EF-1241889, IOS-1754868, IOS-1651888), and the Department of Agriculture (NRI 2009-35102-0543).
Author information
Authors and Affiliations
Contributions
J.T.H., T.R.R., and J.R.R. designed the experiment, J.T.H., T.R.R., and H.J.C. conducted the experiment, S.L.R., M.B.M., and J.R.R. conducted the analyses, S.L.R., J.R.R., and P.J.H. interpreted the data, S.L.R. wrote the manuscript, and all authors contributed to editing the manuscript.
Corresponding author
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Peer review information Nature Communications thanks Ralf Schaefer and the other, anonymous, reviewer for their contribution to the peer review of this work.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Rumschlag, S.L., Mahon, M.B., Hoverman, J.T. et al. Consistent effects of pesticides on community structure and ecosystem function in freshwater systems. Nat Commun 11, 6333 (2020). https://doi.org/10.1038/s41467-020-20192-2
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41467-020-20192-2
This article is cited by
-
Exposure to an insecticide formulation alters chemosensory orientation, but not floral scent detection, in buff-tailed bumblebees (Bombus terrestris)
Scientific Reports (2024)
-
Connectivity mediates the spatial ecological impacts of a glyphosate-based herbicide in experimental metaecosystems
Oecologia (2024)
-
MOF-based Ag NPs/Co3O4 nanozyme for colorimetric detection of thiophanate-methyl based on analyte-enhanced sensing mechanism
Microchimica Acta (2024)
-
Population structure and insecticide response of Gammarus spp. in agricultural and upstream forested sites of small streams
Environmental Sciences Europe (2023)
-
Influence of River Disconnection on Floodplain Periphyton Assemblages
Wetlands (2023)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.