The Pennsylvania State University

The Graduate School

Department of Entomology

ECOLOGICAL TRADE-OFFS ASSOCIATED WITH INSECTICIDE USE,

FROM PENNSYLVANIA TO BANGLADESH

A Dissertation in

Entomology and

International Agriculture & Development

by

Margaret R. Douglas

© 2016 Margaret R. Douglas

Submitted in Partial Fulfillment

of the Requirements

for the Degree of

Doctor of Philosophy

August 2016

The dissertation of Margaret R. Douglas was reviewed and approved* by the following:

Dr. John Tooker Associate Professor of Entomology & Extension Specialist Dissertation Advisor Chair of Committee

Dr. Mary Barbercheck Professor of Entomology

Dr. Christopher Mullin Professor of Entomology

Dr. Armen Kemanian Associate Professor of Production Systems and Modeling

Dr. Gary Felton Professor of Entomology Head of the Department of Entomology

*Signatures are on file in the Graduate School

iii

ABSTRACT

Integrated Pest Management (IPM) requires an understanding of the interaction between chemical and biological control tactics. In recent decades, seed treatment with neonicotinoids has become increasingly widespread in field crop production, but the full ecological and agronomic effects of these insecticides are still far from understood. This dissertation helps to fill this knowledge gap by describing the use of neonicotinoid seed treatments in U.S. agriculture, and examining the influence of these insecticides on pest and predatory invertebrates. In Chapter 1, I synthesized data from various government sources and pesticide use labels to estimate the national use of seed-applied neonicotinoids in the United States. In Chapters 2 and 3, I used laboratory and field studies together with insecticide residue testing to explore the movement of seed-applied neonicotinoids through a crop-slug-beetle food chain and its agronomic consequences in

Pennsylvania no-till soybean and corn systems. I found that seed-applied neonicotinoids variably disrupt biological control of non-target pest slugs. Furthermore, field-collected slugs, earthworms, and caterpillars contained concentrations of neonicotinoids likely to harm predatory arthropods that consume them. In Chapter 4, I tested the generality of the effects of neonicotinoid seed treatments on arthropod natural enemies through a meta- analysis of field studies. By assembling almost 1,000 observations from North American and European experiments, I found that seed-applied neonicotinoids have a negative effect on insect natural enemies corresponding to a roughly 16% seasonal reduction in abundance. Taken together, my results shed light on the agro-ecological effects of

iv neonicotinoid seed treatments and inform their targeted use. Finally, Chapter 5 comprised the international component of my dissertation. In a lablab bean production system in

Bangladesh, I tested the hypothesis that biorational pesticides can replace broad-spectrum synthetic insecticides in insect pest management. Additionally, I used DNA barcoding to characterize emerging thrips pests and determine whether they need to be incorporated into lablab bean IPM programs. Biorational pesticides were variably effective against lablab bean insect pests, and my preliminary evidence suggests that flower thrips (mainly

Megalurothrips spp.) pose a significant threat to lablab bean production and should be included in insect control efforts in this crop.

v

TABLE OF CONTENTS

LIST OF FIGURES ...... ix

LIST OF TABLES ...... xiv

PREFACE ...... xvii

ACKNOWLEDGEMENTS ...... xviii

Introduction ...... 1

Dissertation objectives ...... 2

Chapter 1 Large-scale deployment of seed treatments has driven rapid increase in use of neonicotinoid insecticides and preemptive pest management in U.S. field crops ...... 5

Abstract ...... 5 Introduction ...... 6 Materials & Methods ...... 10 Pesticide data sources ...... 10 Seed treatments as a proportion of neonicotinoid use in major U.S. row crops ...... 12 Proportion of U.S. maize and soybean area planted with neonicotinoid- treated seeds ...... 13 Neonicotinoid seed treatments in U.S. maize and soybean pest management ...... 15 Results ...... 15 Neonicotinoid use in the U.S. by product type, crop, and active ingredient ...... 15 Neonicotinoids applied as seed treatments in major U.S. row crops ...... 16 Hectares planted with neonicotinoid-treated seed in major U.S. row crops ...... 17 Neonicotinoid seed treatments and pest management trends in U.S. maize and soybeans ...... 19 Discussion ...... 22 Acknowledgements ...... 29 Tables ...... 30 Figures ...... 39

vi Chapter 2 Neonicotinoid insecticide travels through a soil food chain, disrupting biological control of non-target pests and decreasing soybean yield ...... 41

Abstract ...... 41 Introduction ...... 42 Materials & Methods ...... 45 Laboratory experiments ...... 45 Field experiment ...... 48 Insecticide analyses ...... 50 Statistical analyses ...... 51 Results ...... 54 Laboratory experiments ...... 54 Field experiment ...... 55 Insecticide residues ...... 57 Discussion ...... 58 Acknowledgements ...... 63 Data accessibility ...... 64 Tables ...... 65 Figures ...... 69

Chapter 3 Neonicotinoid seed treatments move through a no-till corn food web but fail to influence biological control or yield ...... 78

Abstract ...... 78 Introduction ...... 79 Materials & Methods ...... 82 Laboratory experiments ...... 82 Field experiments ...... 85 Insecticide analyses ...... 88 Statistical analyses ...... 89 Results ...... 92 Laboratory experiments ...... 92 Field experiments ...... 94 Insecticide residues ...... 97 Discussion ...... 99 Acknowledgements ...... 103 Tables ...... 104 Figures ...... 108

Chapter 4 Meta-analysis reveals that neonicotinoid seed treatments and pyrethroids have similar negative effects on abundance of arthropod natural enemies ...... 114

Abstract ...... 114 Introduction ...... 116

vii Materials & Methods ...... 119 Searching the literature & building the dataset ...... 120 Defining the scope of the study ...... 121 Calculating effect size: Hedge’s d ...... 122 Addressing non-independence ...... 122 Fitting meta-regression models ...... 123 Assessing statistical assumptions, potential biases, and robustness of results ...... 126 Predator-prey ratios ...... 128 Results ...... 128 Results of the literature search & characteristics of the meta-analysis dataset ...... 128 Seed-applied neonicotinoids negatively affected natural enemies compared to no-insecticide controls ...... 129 Seed-applied neonicotinoids were not significantly less harmful to natural enemies than pyrethroid insecticides ...... 131 Statistical assumptions, potential biases, and robustness of results ...... 132 Effect of neonicotinoid seed treatments on predator-prey ratios in soybeans ...... 134 Discussion ...... 135 Conclusion ...... 141 Acknowledgements ...... 142 Tables ...... 143 Figures ...... 146

Chapter 5 Investigating biopesticides and DNA barcoding as tools to improve insect pest management in lablab bean (Lablab purpureus) in Bangladesh ...... 148

Abstract ...... 148 Introduction ...... 149 Materials & Methods ...... 152 Field experiments ...... 152 Characterizing the thrips community in lablab bean ...... 157 Results ...... 161 Field experiments ...... 161 Characterizing the thrips community in lablab bean ...... 164 Discussion ...... 167 Acknowledgements ...... 171 Tables ...... 173 Figures ...... 179

Conclusions ...... 187

Future research directions ...... 188

viii Appendix A Supplemental details on methods used in experiments on the effects of neonicotinoid seed treatments on a soybean-slug-beetle food chain ...... 191

Laboratory experiment ...... 191 Soybean-slug and slug-ground beetle bioassays ...... 191 Field experiment ...... 192 Study site and crop management ...... 192 Invertebrate activity-density ...... 192 Predation ...... 193 Insecticide analyses ...... 193

Appendix B Supplemental results from experiments on the effects of neonicotinoid seed treatments on a soybean-slug-beetle food chain ...... 196

Methods ...... 196 Analysis ...... 196 Results ...... 197

Appendix C Supplemental results from a meta-analysis of the effect of seed- applied neonicotinoids on natural enemies ...... 198

Works cited ...... 203

ix

LIST OF FIGURES

Figure 1-1. Neonicotinoid sales by product type (a), and use by crop (b) and active ingredient (c) from 1992 to 2011. Data on use (a) is based on sales data from the Minnesota Department of Agriculture (2014). Data on crops and active ingredients are for the entire U.S., from USGS (EPest-High estimate; Thelin & Stone 2013). Y-axes represent mass of neonicotinoid active ingredient in thousands or millions of kg...... 39

Figure 1-2. Estimated percent of U.S. maize (a-c) and soybean (d-f) hectares treated with neonicotinoid seed treatments (a,d), any pesticide seed treatment (b,e), and non-seed treatment (Non-ST) insecticides (c,f) from 1996 to 2012. Estimates for national neonicotinoid seed treatment use (a,d) were based on the USGS data (Thelin & Stone 2013) and rates indicated on product labels (Table 1-2), for scenarios with varying amounts of low-rate (LR) versus high- rate treated seed (see text for details). Estimates for overall pesticide seed treatment use (b,e) are for North Dakota (Glogoza et al. 2002; Zollinger et al. 1996; Zollinger et al. 2006; Zollinger et al. 2012; Zollinger et al. 1992; Zollinger et al. 2009)...... 40

Figure 2-1. A schematic representation of our hypothesis for the potential influence of seed treatments on the ecological community in no-till soybeans. The “+” and “-” signs indicate the anticipated effect (positive or negative) of the preceding factor on the following factor. We would expect this model to hold when slugs are the dominant early-season soybean herbivore. Based on previous findings that moderate early-season leaf damage has little effect on soybean yield (Hammond 2000), we expected slugs to reduce yield mainly by killing plants rather than by eating leaf tissue...... 69

Figure 2-2. Outcomes from laboratory experiments investigating the influence of neonicotinoid seed treatments on interactions between soybeans (G. max), slugs (D. reticulatum), and ground beetles (C. tricolor). Soybean seed treatments: U = untreated, F = fungicide-only, F+L = fungicide + low rate thiamethoxam, F+H = fungicide + high rate thiamethoxam. Error bars show ± one standard error. (a) Number of soybean seedlings (out of four) damaged by slugs over seven days (n = 34 microcosms/treatment; no statistical differences among treatments). (b) Slug mass gain (%) after seven days of feeding on soybean seedlings (n = 34 microcosms/treatment; no statistical differences among treatments). (c) Beetle symptoms after consuming slugs fed upon the four seed treatments; beetles exposed to insecticides via slugs suffered significantly higher frequency of impairment (Fisher’s exact test, P < 0.0001)...... 70

x Figure 2-3. Slug survival when confined individually with C. tricolor after seven days of feeding on soybeans with different seed coatings: U = untreated, F=fungicide-only, F+L= fungicide + low rate thiamethoxam, F+H = fungicide + high rate thiamethoxam. Seed treatment did not influence rates of beetle predation on slugs (Likelihood ratio test, D = 2.18, d.f. = 3, P = 0.54). .... 71

Figure 2-4. Partial regression plots for relationships among organisms in soybean plots planted with untreated (‘Control’) or thiamethoxam and fungicide- treated (‘Neonicotinoid’) seeds (n = 6 plots/treatment). P values test the significance of each partial correlation coefficient (Bonferroni-corrected α = 2 0.01). Rp is the proportion of non-block variation explained by the predictor 2 (squared partial correlation). Rsp is the proportion of total variation explained by the predictor (squared semi-partial correlation). 95% confidence bands are shown in gray. (a) Activity-density of potential slug predators (#/trap) was positively related to predation on sentinel prey (proportion prey killed). (b) Slug activity-density (#/trap) was negatively related to predation on sentinel prey. (c) Soybean population (10000/ha) was marginally negatively related to slug activity-density (#/trap). (d) Yield (t/ha) was positively related to soybean population...... 73

Figure 2-5. Animals per trap (mean ± SE) measured in pitfall traps (A. & B.) or shelter traps (C.) over the season in plots that were planted with untreated (Control) or thiamethoxam and fungicide-treated (THX) soybean seeds (n = 6 plots/treatment). Slugs in pitfall traps (A.) and refuge traps (C.) were numerically more abundant in THX than control plots over the season, whereas potential slug predators in pitfall traps (B.) showed a transient response to seed treatment, being depressed in treated plots on the first sample date. See main text for statistical results...... 74

Figure 2-6. Predation on sentinel waxworm caterpillars (mean proportion killed ± SE) during two 12-hour time periods (AM/PM) in plots that were planted with untreated (control) or thiamethoxam and fungicide-treated (THX) soybean seeds (n = 6 plots/treatment). Predation was reduced in treated plots in June when neonicotinoids concentrations were highest, but not in July or August when concentrations would have declined. See main text for statistical results...... 75

Figure 2-7. Neonicotinoid concentrations (mean ppb ± SE) in samples collected 12-169 days after planting (number of days for each sample noted in parentheses on the X-axis), from field plots planted with untreated (Control) or thiamethoxam and fungicide-treated (Neonic) soybean seeds (n = 3 plots except for earthworms, where n = 2 plots , listed separately). Thiamethoxam (THX) was the active ingredient applied to the seeds, while CLO is clothianidin + clothianidin TZMU [N-(2-chlorothiazol- 5-ylmethyl)-N-

xi methylurea], metabolites of THX that are also insecticidal. Earthworms (Worms) were only sampled in Neonic plots...... 76

Figure 2-8. Concentrations of neonicotinoids in soybeans, slugs (D. reticulatum), and ground beetles (C. tricolor) from laboratory and field experiments. Samples from the field were collected when soybeans were at the cotyledon stage. In the regression equation, “Setting” represents the effect of experiment location (lab vs. field), while “Level” represents the effect of trophic level (soybeans = 0, slugs = 1, beetles = 2)...... 77

Figure 3-1. Outcomes from laboratory experiments investigating the influence of neonicotinoid seed treatments on interactions between corn seedlings (Zea mays), slugs (Deroceras reticulatum), and ground beetles (Chlaenius tricolor). Corn seed treatments: U = untreated, F = fungicide-only, F+L = fungicide + low rate thiamethoxam, F+H = fungicide + high rate thiamethoxam. Error bars show ± one standard error; means with different letters above them are different at α = 0.05 based on post-hoc Tukey comparisons. (a) Corn seedling biomass (g) after 7 days of growth with or without slugs. (b) Slug mass gain (%) after 7 days of feeding on corn seedlings (n = 26-31 microcosms/treatment, no statistical differences among treatments). (c) Beetle symptoms after consuming slugs fed upon three seed treatments...... 108

Figure 3-2. Slug survival when confined individually with C. tricolor after seven days of feeding on corn with different seed coatings: U = untreated, F+L = fungicide + low rate thiamethoxam, F+H = fungicide + high rate thiamethoxam. Seed treatment did not significantly influence rates of beetle predation on slugs (Likelihood ratio test, D = 2.29, df = 2, P = 0.32)...... 109

Figure 3-3. Slug activity-density (mean ±SE) in field experiments comparing plots that were planted with neonicotinoid-treated corn seed to controls with only fungicides (2012) or no seed treatment (2013; n = 6 plots/treatment). Slug populations were measured using pitfall traps (A,B) or shelter traps (C,D). Asterisks mark significant comparisons at α = 0.05, based on post-hoc comparisons where there was a significant Date ×Treatment interaction...... 110

Figure 3-4. Activity-density (mean ±SE) of potential slug predators (A,B) and all arthropod predators (C,D) in field experiments comparing plots that were planted with neonicotinoid-treated corn seed to controls with only fungicides (A,C; 2012) or no seed treatment (B,D; 2013; n = 6 plots/treatment). There were no significant differences in predator activity-density between treatments...... 111

Figure 3-5. Predation (mean ±SE)on sentinel caterpillars (Galleria mellonella) over the season in field experiments comparing plots that were planted with

xii neonicotinoid-treated corn seed to control plots that were planted with fungicide-only (A; 2012) or untreated (B; 2013) seed. THX = thiamethoxam, CLO = clothianidin. Predation did not differ significantly by treatment on any sample date...... 112

Figure 3-6. Concentrations of neonicotinoids in corn shoots, pest herbivores (slugs Deroceras reticulatum and cutworms Agrotis ipsilon), and ground beetles Chlaenius tricolor from laboratory and field experiments. Samples from the field were collected in 2012 when corn seedlings were in the V2 stage...... 113

Figure 4-1. Confidence intervals (95%) for the effect of neonicotinoid seed treatments on natural-enemy abundance, relative to no-insecticide controls. Based on 607 observations from 56 site-years and 20 studies. See text for details on models used to generate these estimates...... 146

Figure 4-2. Confidence intervals (95%) for the effect of neonicotinoid seed treatments on natural-enemy abundance, relative to controls treated with foliar or soil-applied pyrethroids. Based on 384 observations from 15 site- years and 8 studies. Results are presented both with and without data from Ohnesorg et al. 2009, which had effect sizes quite different from the other studies. See text for details on models used to generate these estimates...... 147

Figure 5-1. Adult and larval bean thrips (Megalurothrips spp.) infesting the flowers (A), pods (B), and leaves (C) of lablab bean in a field experiment in Gazipur, Bangladesh ...... 179

Figure 5-2. Thrips abundance (A) and flowering activity (B) in response to insecticide treatments in Experiment 1 (mean ± standard error). OP = organophosphate...... 180

Figure 5-3. Pod-borer damage (A), aphid infestation (B), thrips abundance (C), mean inflorescences (D), and marketable yield (E) after insecticide treatments in Experiment 2. Aphid infestation and pod damage were logit- transformed for analysis, but untransformed means are shown here. Means (± standard error) reflect seasonal totals (yield, pod-borer damage), seasonal means (inflorescences), or a cumulative measure of seasonal abundance (aphids, thrips; see Materials & Methods for details). Based on Tukey’s HSD test, letters above means indicate significant differences at α = 0.05...... 182

Figure 5-4. Partial regression plots for relationships between pest abundance and lablab bean yield. The fitted model was used to generate the slopes (dotted lines) and their 95% confidence bands (shown in gray). Lablab bean yield was negatively related to thrips abundance (A) and pod-borer damage (B), but only marginally related to aphid abundance (not shown)...... 183

xiii Figure 5-5. Neighbor-joining tree based on the 5’ region of the mitochondrial cytochrome oxidase I gene of thrips samples collected from lablab bean plants in Bangladesh. Each specimen is labeled with its sequence number in the Barcode of Life Data System (BOLD). Reference specimens are in gray. Larval samples are in purple. One sample for which the morphological and molecular identifications conflict is in orange. Clades are labeled with Barcode Index Numbers (BINs), which have been proposed to be molecular taxonomic units similar to species. The tree is based on p-distances and was built with node support assessed with 1000 bootstrap iterations in Mega (version 6.06). Bootstrap values less than 70% are not shown...... 185

Figure 5-6. Neighbor-joining tree based on the 5’ region of the mitochondrial cytochrome oxidase I gene of Megalurothrips and related samples (BINs AAM8053, AAN6623, ACS4755) from the present study and the Barcode of Life Database (BOLD). Each specimen is labeled with its BOLD sequence number; the specimens from the current study start with “LLBT”. Colors correspond to morphological identifications. The tree is based on p-distances and was built with node support assessed with 1000 bootstrap iterations in Mega (version 6.06). Bootstrap values less than 70% are not shown...... 186

xiv

LIST OF TABLES

Table 1-1. Summary of the major data sources we relied upon for our U.S. pesticide use estimates, 1992-2012...... 30

Table 1-2. Neonicotinoid use rates of some common seed-treatment products on four major crops, based on information from pesticide labels...... 31

Table 1-3. Summary of neonicotinoids and their major agricultural uses in the U.S., based on regulatory documents from the EPA...... 33

Table 1-4: Estimated neonicotinoids applied (kg active ingredient), and neonicotinoid seed treatments (ST) as a percentage of neonicotinoids applied, for major U.S. field crops from 2000 to 2012. Dashes represent combinations for which data were unavailable...... 35

Table 1-5: Nationwide estimates of insecticides applied (kg active ingredient), and neonicotinoid seed treatments (ST) as a percentage of insecticides applied, for major U.S. field crops from 2000 to 2012. Dashes represent crop- year combinations for which data were unavailable...... 37

Table 1-6: Some insecticidal seed treatments labeled for maize, based on active ingredients other than neonicotinoids. We do not consider products for stored grain control...... 38

Table 2-1. Responses of plants, slugs, and predators in a field experiment comparing soybean plots planted with untreated (Control) or thiamethoxam and fungicide-treated (Neonic) seeds (n = 6 plots/treatment). The “predicted effects” listed here for each response variable are illustrated in Figure 2-1...... 65

Table 2-2 Activity-density of potential slug predators (mean ± SE) on individual sample dates and cumulatively for the season, in plots that were planted with untreated (control) or thiamethoxam and fungicide-treated (THX) soybean seeds (n = 6 plots/treatment)...... 66

Table 2-3. Neonicotinoid concentration (ppb) detected in samples from a laboratory experiment investigating soybean-slug-ground beetle interactions in the presence or absence of seed treatments...... 67

Table 2-4 Thiamethoxam as a percentage of total neonicotinoid residues (mean ± SE) in the soybean-slug-beetle food chain in the field and laboratory. Values with different letters are significantly different at α = 0.05 (Field: Kruskal-

xv 2 Wallis χ = 3.86, df = 1, P = 0.05; Lab: F2,5 = 405, P < 0.001, with post-hoc Tukey test)...... 68

Table 3-1. Responses of plants, herbivores, and predators in a 2012 field experiment comparing corn plots planted with fungicide-only (Control) or thiamethoxam and fungicide-treated (Neonic) seeds (n = 6 plots/treatment)...... 104

Table 3-2. Responses of plants, herbivores, and predators in a 2013 field experiment comparing corn plots planted with untreated (Control) or clothianidin-treated (Neonic) seeds (n = 6 plots/treatment) in a year with low slug abundance, at a site with a 1-year legacy of neonicotinoid or control treatments...... 105

Table 3-3. Neonicotinoid concentration (ppb) detected in samples from a laboratory experiment investigating corn-slug-ground beetle interactions in the presence or absence of seed treatments. Neonicotinoids were not detected in any samples from the untreated control (data not shown)...... 106

Table 3-4. Neonicotinoid concentration and abundance (mean ± SE) detected in samples from a field experiment investigating the movement of neonicotinoid seed treatments in a no-till corn ecosystem. Non-detections (ND) were treated as zero for the purpose of calculating means and standard errors...... 107

Table 4-1. LC50 results from two laboratory studies that compared imidacloprid toxicity to insect and arachnid predators...... 143

Table 4-2. Description of the dataset used in a meta-analysis of neonicotinoid seed treatment effects on natural enemies of crop pests...... 144

Table 4-3. Estimates and tests of significance for moderators in a meta-regression model testing the effect of neonicotinoid seed treatments on natural enemies, compared to controls treated with no insecticides (n = 607 observations from 56 site-years and 20 studies)...... 145

Table 5-1. Insecticide formulations and rates applied to lablab bean (Lablab purpureus) in field experiments in Bangladesh...... 173

Table 5-2. Reference specimens used in the molecular identification of thrips specimens ...... 174

Table 5-3. Pairwise genetic distances (p-distances) for thrips taxa collected from lablab bean (Lablab purpureus) in Bangladesh. Genetic distances are shown for only the specimens from this study (“This study”), as well as for the specimens from this study combined with all other records for each BIN in the Barcode of Life Data Systems (BOLD)...... 175

xvi Table 5-4. Collection and curation information for thrips specimens collected from lablab bean (L. purpureus) in Bangladesh and their associated coxI sequences...... 176

xvii

PREFACE

Chapter 1 of this dissertation was multiply authored with Margaret R. Douglas as the first author and John F. Tooker as the second author. MRD conceived the project, collected the relevant data and publications, conducted the analyses and drafted the manuscript. JFT provided feedback on the goals of the project, contributed to the literature search, and suggested edits to the manuscript.

Chapter 2 of this dissertation was multiply authored with MRD as the first author,

Jason R. Rohr as the second author, and JFT as the third author. MRD and JFT conceived the experiments, MRD executed the field experiments and collected the data, MRD analyzed the data with support from JRR, and MRD drafted the manuscript while JFT and JRR suggested edits.

Chapter 5 of this dissertation was multiply authored with MRD as the first author, and six coauthors. MRD conceived the project, managed the field experiments, collected and analyzed the data, and drafted the manuscript. Srinivasan Ramasamy helped conceive the project, provided feedback on the experiments, and suggested edits to the manuscript.

Jan Chang provided training on DNA barcoding. Kohinoor Begum provided support for field experiments and data collection in Bangladesh. Sevgan Subramanian performed morphological identifications of thrips specimens. JFT provided feedback on the experiments and suggested edits to the chapter. Syed Nurul Alam provided guidance and support for the field experiments in Bangladesh.

xviii

ACKNOWLEDGEMENTS

I am deeply thankful to my colleagues, mentors, friends, and family, who made writing this dissertation both possible and often enjoyable! Thank you, Dr. John Tooker, for your unrelenting support and encouragement as my advisor, and for granting me the freedom to pursue my research interests. I am a far better researcher and writer because of your constructive guidance. My committee has also been generous in sharing their time and expertise. Thank you, Dr. Chris Mullen, for enriching my understanding of insecticide chemistry and ecotoxicology. Thank you, Dr. Armen Kemanian, for your insightful critiques and help strengthening my statistical analyses. And thank you, Dr.

Mary Barbercheck, for challenging me to think critically about soil invertebrate communities and their ecological functions, not to mention broader topics like the place of science in agriculture and society. Thank you also, Dr. Gary Felton, for facilitating my support through the department and creating a vibrant and collegial research environment.

I also owe thanks to those people who made my work possible in Taiwan and

Bangladesh. Thank you Dr. Ed Rajotte, my ‘shadow’ committee member, for your instrumental support and encouragement in pursuing the international component of my research. Thank you Dr. Srini Ramasamy for co-developing the Borlaug proposal that allowed me to study lablab bean IPM, Jan Chang for showing me the ropes of DNA barcoding, and the AVRDC Entomology staff for lab and field assistance. Thank you Dr.

Syed Nurul Alam, Dr. Kohinoor Begum, Fatema Khatun, Jakir Hossain, and the entire

xix BARI Entomology crew not only for your research help and guidance but also for your incredible hospitality that made me feel at home in a most unfamiliar place.

I thank the members of the Tooker lab for their camaraderie, helpful feedback on experiments and manuscripts, and general good humor. Eric Bohnenblust, Ian

Grettenberger, Anjel Helms, Marion Le Gall, Anna Busch, Anthony Vaudo, Kevin Rice,

Eric Yip, Elizabeth Rowen, and Kirsten Pearsons, thank you for being such excellent lab- mates. A special thank you to Andrew Aschwanden for helping to order supplies and keeping the lab running. Over the years a number of fantastic undergraduate research assistants provided invaluable field and laboratory help, and they are acknowledged by name in the relevant chapters. Also, thank you Scott Smiles for your willingness to make anything possible at the research farm, and for doing it with a smile.

Finally, I thank my family and friends. Thank you, Karen and Doug Douglas, and

Laura Douglas and Katie Kindle, for your love and enthusiastic cheerleading. Thank you to the students in the Entomology and INTAD graduate programs, and broader Penn State sustainable agriculture research community, for sharing your impressive combined knowledge and experiences. And finally, thank you Bill Freese, for being my Gabe Oak.

xx

DEDICATION

This dissertation is dedicated to Scott Smiles, who truly lived up to his name.

1

Introduction

Insects are both friends and foes in crop production; this simple fact has far- reaching consequences for the judicious use of insecticides. Since at least the 1950’s, researchers have observed that while insecticides undoubtedly suppress insect pests and protect yield in many situations, they can simultaneously disrupt biological control by predators and parasitoids (i.e. natural enemies), leading to resurgence of target or non- target pest populations, eroding or even reversing the benefits of insecticide use (Stern et al. 1959). Understanding the complex ecological trade-offs between chemical and biological control is a central project of research in integrated pest management (IPM), and one that remains relevant as long as insecticide active ingredients and crop production practices continue to evolve.

One important change in North American field crop production over the past several decades has been a shift toward greater dependence on seed-based pest management strategies. Through transgenic traits and seed-applied pesticides, many of a farmer’s pest management decisions have been made before the first seed hits the soil

(Gray 2011). Neonicotinoid insecticides, applied to the seed and taken up systemically into crop tissues, have enjoyed an especially rapid expansion in recent years (Simon-

Delso et al. 2015). While these novel insecticide products are now used over a very large area, their fate in the agricultural food webs and non-target effects on natural enemies and biological control are still far from understood.

2 Another recent change in pest management has been the increasing availability of narrow-spectrum insecticides, many of them based on biological organisms or their products ("biopesticides"; Glare et al. 2012; Sparks 2013). Some biopesticides show promise to help farmers manage pests while avoiding non-target harm to natural enemies, wildlife, and farm workers. These benefits could be especially large in countries like

Bangladesh, where farmers often spray conventional, broad-spectrum insecticides frequently and with little protective equipment (Dasgupta et al. 2005). However, one of the challenges of employing narrow-spectrum insecticides, like some biopesticides, is that they require a detailed understanding of pest assemblages; unlike broad-spectrum products that often control multiple pests simultaneously, narrow-spectrum products by definition will suppress only some pest species.

In this dissertation, I seek to advance scientific understanding of the ecological basis of pest management in diverse cropping systems, from no-till corn and soybeans in

Pennsylvania to lablab bean in Bangladesh. The specific objectives are described below.

Dissertation objectives

In Chapter 1, I synthesized evidence from government agencies, pesticide labels, and the peer-reviewed literature to i) describe the extent of seed-applied neonicotinoid use in major U.S. field crops (corn, soybean, wheat, cotton), ii) estimate the contribution of these products to overall insecticide use, and iii) outline their role in U.S. pest management. This chapter provides the necessary background to motivate and

3 contextualize my further work on the ecological trade-offs associated with neonicotinoid seed treatments, as reported in the following three chapters.

In Chapter 2, I used laboratory and field studies together with insecticide residue testing to explore the movement of seed-applied neonicotinoids through a soybean-slug- beetle food chain and its agronomic consequences in a Pennsylvania no-till cropping system. I tested the hypothesis that seed-applied neonicotinoids can disrupt biological control of non-target pest slugs in soybeans, ultimately decreasing yield. Furthermore I provided the first field-based estimates of neonicotinoid concentrations in organisms in the soil food web.

In Chapter 3, I conducted a similar study in a no-till corn system over two field seasons. I tested the hypothesis that seed-applied neonicotinoids would benefit corn production when non-target slug pests were scarce, but impose a cost when slugs were abundant. I also continued describing the movement of neonicotinoids through the soil food web and examined whether the higher insecticide rates associated with corn production lead to higher concentrations of insecticide residues in soil organisms.

In Chapter 4, I tested the generality of the effects of neonicotinoid seed treatments on natural enemies through a meta-analysis of available field studies. After searching the literature and assembling almost 1,000 observations from North American and European experiments, I conducted a meta-regression to test whether seed-applied neonicotinoids reduce natural enemy populations relative to control plots either i) treated with no insecticides or ii) treated with broadcast applications of pyrethroids.

In Chapter 5, I conducted field experiments in a lablab bean production system in in Bangladesh, to test the hypothesis that biopesticides can replace broad-spectrum

4 synthetic insecticides in insect pest management in this crop. Additionally, I used DNA barcoding and field experiments to characterize emerging thrips pests and to determine whether they need to be incorporated into IPM strategies for this crop.

5

Chapter 1

Large-scale deployment of seed treatments has driven rapid increase in use of neonicotinoid insecticides and preemptive pest management in U.S. field crops

This chapter has been published in Environmental Science & Technology with the citation:

Douglas, M. R. and J. F. Tooker. 2015. “Large-scale deployment of seed treatments has

driven rapid increase in use of neonicotinoid insecticides and preemptive pest

management in U.S. field crops.” 49(8): 5088-5097.

Abstract

Neonicotinoids are the most widely used class of insecticides worldwide, but patterns of their use in the U.S. are poorly documented, constraining attempts to understand their role in pest management and potential non-target effects. We synthesized publicly available data to estimate and interpret trends in neonicotinoid use since their introduction in 1994, with a special focus on seed treatments, a major use not captured by the national pesticide-use survey. Neonicotinoid use increased rapidly between 2003 and 2011, as seed-applied products were introduced in field crops, marking an unprecedented shift toward large-scale, preemptive insecticide use: 34-44% of soybeans and 79-100% of maize hectares were treated in 2011. This finding contradicts recent analyses, which concluded that insecticides are used today on fewer maize

6 hectares than a decade or two ago. If current trends continue, neonicotinoid use will increase further through application to more hectares of soybean and other crop species and escalation of per-seed rates. Alternatively, our results, and other recent analyses, suggest that carefully targeted efforts could considerably reduce neonicotinoid use in field crops without yield declines or economic harm to farmers, reducing the potential for pest resistance, non-target pest outbreaks, environmental contamination, and harm to wildlife, including pollinator species.

Introduction

Since their introduction in the 1990s, neonicotinoids have become the most widely used class of insecticides in the world (Sparks 2013), but patterns of their use in the United States (U.S.) remain largely undefined. It is unclear when, where, and how neonicotinoids are used in U.S. agriculture, because they are often applied as seed treatments (Jeschke et al. 2011), a use that is not captured by the major national pesticide survey conducted by the National Agricultural Statistics Service (NASS; U. S.

Department of Agriculture NASS 2014a). Fortunately, seed-applied neonicotinoids are included in an independent dataset that recently became available through the U.S.

Geological Survey (USGS; Thelin & Stone 2013). This dataset shows that aggregate neonicotinoid use has increased dramatically over the past decade, but it does not report the percentage of cropland treated or the relative importance of different modes of application (e.g., seed treatments vs. foliar sprays). Anecdotally, extension entomologists have noted that seed for some field crop species, such as maize (Zea mays), in the U.S. is

7 now routinely treated before sale with neonicotinoids (Gray 2011; Krupke et al. 2012), suggesting that neonicotinoid seed treatments (NSTs) are being used over a very large area.

Because seed-treatment use has not been captured in the national pesticide survey, recent analyses that relied on this survey to understand pest management trends in the

U.S. (Benbrook 2012; Osteen & Fernandez-Cornejo 2013) appear to have missed an important aspect of insecticide use, including how seed-applied insecticides relate to other pest-management approaches. In particular, it is unclear whether NSTs have displaced other insecticide applications, and how they relate to transgenic, insect-resistant crops. The void of information on seed treatments also challenges researchers and regulators seeking to assess environmental contamination and potential non-target effects associated with neonicotinoids, areas of increasing concern (Bonmatin et al. 2015;

Chagnon et al. 2015; Hallmann et al. 2014; Hladik et al. 2014; Krupke et al. 2012; Main et al. 2014). The potential for lethal and sublethal effects of neonicotinoids on pollinators has attracted particular attention, prompting the European Union to suspend the use of neonicotinoids on bee-attractive crops (European Commision 2013), and accelerating the review of neonicotinoids by the U.S. Environmental Protection Agency (U. S.

Environmental Protection Agency 2015). Characterizing the risk posed by neonicotinoids to non-target species obviously requires understanding where and how these compounds are used. We therefore see an opportunity to synthesize the available information to estimate trends in neonicotinoid use in U.S. agriculture since their introduction twenty years ago.

8 Documenting patterns of neonicotinoid use is an important first step toward elucidating their role in U.S. pest management, but a full consideration also requires a broader context, including an understanding of their physical properties and importance of target pests in particular cropping systems. The physical properties of neonicotinoids have been well documented; they generally have high acute toxicity to insects, low acute toxicity to mammals, and are systemic in plants, though the degree of these characteristics varies among neonicotinoid active ingredients (Jeschke et al. 2011;

Tomizawa & Casida 2005). Depending on the cropping system and target pest(s), they are applied using a variety of methods including foliar sprays, soil drenches, chemigation systems, trunk injection, and seed treatment. Seed-applied neonicotinoids are taken up into plant tissues, where they provide up to several weeks of protection against insect pests (McCornack & Ragsdale 2006; Seagraves & Lundgren 2012), and persist at low concentrations for up to several months (Bonmatin et al. 2015). Their efficacy against target pests depends on the active ingredient being present when and where pests are feeding, an important consideration given their variable uptake into different crop tissues and declining concentrations as plant growth continues (Sur & Stork 2003). Finally, it is unclear how important many of the pest species being targeted by NSTs really are; there is a general dearth of information on prevalence and impact of most labeled pests, many of which are considered ‘secondary’ in U.S. field crops because they occur sporadically in space and time (Steffey & Gray 2000).

To contribute to a broader understanding of the role of neonicotinoids in U.S. agriculture since their introduction in 1994, in this paper we synthesize publicly available information and the entomological literature to address the following questions:

9 1) Which uses, crops, and active ingredients accounted for the most neonicotinoid

use?

2) On major field crops (maize, soybean [Glycine max], wheat [Triticum spp.],

cotton [Gossypium spp.]), what was the contribution of NSTs to total

neonicotinoid use and total insecticide use?

3) What proportion of the area of maize and soybeans was planted with

neonicotinoid-treated seed?

4) What role did neonicotinoids play in pest management in maize and soybean

production?

To address questions one through three, we made simple calculations based on data from a variety of sources, primarily the USGS and U.S. Department of Agriculture

(USDA), as well as state-level sources and pesticide label information. For question four, first we compared trends in neonicotinoid use to trends in use of other insecticides and transgenic insect-resistant crops, and then we integrated our findings with entomological literature on neonicotinoids in maize and soybeans. This last section of the results is necessarily interpretive given the nature of our question. We gave special attention to maize and soybeans because they are the highest area crops in the U.S. (U. S. Department of Agriculture NASS 2014c), our results indicate that they account for the majority of neonicotinoid use, and they benefit from more research than many other crops.

10 Materials & Methods

Pesticide data sources

We used five main sources for data on the use of insecticide active ingredients in the U.S. (Table 1-1). Our primary source of information on neonicotinoid use was the

Pesticide National Synthesis Project of the U.S. Geological Survey

(http://water.usgs.gov/nawqa/pnsp). These pesticide-use estimates were derived from proprietary farm surveys (conducted by GfK Kynetec, Inc.) within Crop Reporting

Districts (CRDs; Thelin & Stone 2013). For each crop-pesticide combination, from 1992 to 2011, the dataset includes two estimates for pesticide use, which differ in how they treat missing values within surveyed CRDs. For the “EPest-low” estimate, researchers treated these values as zero, while for the “EPest-high” estimate, researchers treated these values as unsurveyed, and extrapolated pesticide-crop use rates from nearby CRDs

(Thelin & Stone 2013). We used both sets of estimates for many of our analyses, and when we did not, we relied on the EPest-high estimate but noted how using the EPest-low estimate would have influenced our results. Importantly, the proprietary dataset used by

USGS captures all agricultural pesticide use, including seed treatments. The dataset we obtained from USGS (courtesy of Wes Stone) reported at the national level total mass applied for each neonicotinoid active ingredient by crop and year. This dataset reports pesticide use on cropland, and so does not account for other uses of neonicotinoids such as homeowner or veterinary use.

The other major source of pesticide data we used was the Agricultural Chemical

Use Survey, originally administered by USDA’s National Agricultural Statistics Service

11 (NASS) and now administered jointly by NASS and the USDA’s Economic Research

Service (ERS) as part of the Agricultural Resource Management Survey (U. S.

Department of Agriculture NASS 2014a). Importantly, this survey excludes pesticides applied as seed treatments, and so provides an estimate of non-seed-treatment pesticide use (U. S. Department of Agriculture NASS 2014a). The NASS survey focuses on major production states (called “Program States”) for each crop, usually covering >80% of the national area. We estimated national, non-seed-treatment neonicotinoid use for each crop and year by dividing the total mass of active ingredient applied in the program states by the proportion of area surveyed.

For maize, we found an additional document providing insight into use of neonicotinoid-treated seed, an environmental report accompanying a petition for non- regulated status of a genetically engineered maize variety submitted to USDA by Pioneer

Hi-Bred International, Inc., one of the largest suppliers of maize seed in the U.S. (Pioneer

Hi-Bred International Inc. 2012). This report contains estimates of the percentage of maize seed treated with insecticides from 2004 to 2011, providing a comparison for our estimates of neonicotinoid use in maize. We assumed that neonicotinoids were the dominant class of insecticides used to treat maize seed (see Results for a detailed treatment of this issue).

Two state-level sources provided complementary pieces of information to understand neonicotinoid use and its role in pest management. First, North Dakota State

University, in partnership with the North Dakota Department of Agriculture and NASS, has surveyed farmers on their use of pesticides six times since 1992 (Glogoza et al. 2002;

Zollinger et al. 1996; Zollinger et al. 2006; Zollinger et al. 2012; Zollinger et al. 1992;

12 Zollinger et al. 2009). These surveys estimate the percentage of field-crop area in North

Dakota that was planted with pesticide-treated seed and how much of that seed was treated on-farm (vs. by seed suppliers before sale). It is important to note that these estimates are not specific to neonicotinoids or even to insecticides; they include all kinds of pesticide seed treatments (e.g. fungicides). Nonetheless, they set an upper bound for the percentage of seed that was treated with insecticides, and by whom (i.e., seed suppliers or farmers). Second, to gain insight into the relative importance of different use sites for neonicotinoids (e.g. cropland, animal care, home gardens, buildings, etc.), we drew on pesticide sales data from Minnesota (reported in mass of active ingredient), generated by the Minnesota Department of Agriculture from information submitted by pesticide registrants (Minnesota Department of Agriculture 2014). We do not suggest that these state-level data sources are precise estimates of national trends, but they are the only data sources we know of that address these particular questions, and so they provide valuable insight.

Because many of our data sources were reported in imperial units, we converted all of our results to metric units for reporting here.

Seed treatments as a proportion of neonicotinoid use in major U.S. row crops

To estimate the proportion of neonicotinoids that were applied as seed treatments, we took advantage of a key difference between the USGS and NASS datasets. Because

USGS estimates all pesticide use, including seed treatments, while NASS estimates all non-seed treatment use, we estimated neonicotinoids applied as seed treatments by

13 subtracting the NASS estimate from the USGS estimates for each crop. We could only make these estimates for those crop-year combinations for which both datasets were available (Table 1-1). Although the two datasets were generated independently, they both employed farmer surveys for data collection and used sampling designs to ensure their results represented U.S. agriculture as a whole.

Our analysis focused on the top four row crops by area: maize, soybeans, wheat, and cotton (U. S. Department of Agriculture NASS 2014c). We focused on these crops for two reasons: i) they now account for a large proportion of neonicotinoids applied in the U.S. (Figure 1-1), and ii) crop-specific data were available for these crops in both the

USGS and NASS datasets, while for many other crops the data were aggregated into larger groupings (e.g. “vegetables and fruit”) that obscured crop-specific use.

Proportion of U.S. maize and soybean area planted with neonicotinoid-treated seeds

Next, we conducted an additional analysis for maize and soybeans, the two dominant crops in U.S. agriculture (U. S. Department of Agriculture NASS 2014c). We estimated possible ranges for the area of each crop planted with neonicotinoid-treated seed in each year, using the following equation:

+, %../0"1 !"#$%&"' = () () +, 2%$" ℎ%

kg appliedST = the estimate for kg neonicotinoids applied as seed treatment (see above)

Rate (kg/ha) = the neonicotinoid use rate, in kg per ha per year

14 Possible use rates were derived from pesticide labels for seed treatment products (Table

1-2). To translate these application rates (active ingredient [a.i.] per seed or per seed-kg) into per hectare rates (a.i. per ha), we needed to estimate seeding rates (seeds/ha or seed- kg/ha). For this we used a combination of USDA-ERS data on per hectare seed costs (U.

S. Department of Agriculture ERS 2014a), and USDA-NASS data on per unit seed costs

(U. S. Department of Agriculture NASS 2014b), based on previous work (Benbrook

2009). We checked these estimates against extension recommendations to ensure they were plausible.

Because most NST products can be applied at more than one rate, we also needed to make some assumptions about rates to calculate ranges for maize and soybean area planted with treated seed. We assumed that low rates were much more common than high rates based on marketing material from several major seed suppliers indicating that the low rate is “standard” (Dow AgroSciences 2014; Pioneer Hi-Bred International Inc.

2014a; Pioneer Hi-Bred International Inc. 2014b), agreeing with our own experience interacting with farmers and seed companies. We therefore calculated values based on the following scenarios: (i) 100% of hectare-treatments at the low rate, (ii) 90% of hectare- treatments at the low rate, 10% at the high rate, (iii) 80% of hectare-treatments at the low rate, 20% at the high rate. We fit each of these scenarios using the EPest-high and EPest- low estimates for neonicotinoid use, so that our calculated range of values incorporates uncertainty associated with both overall neonicotinoid use and application rates.

Once we obtained estimates for hectares planted with treated seed under each of our scenarios, we divided each value by total crop area (U. S. Department of Agriculture

15 NASS 2014c) to estimate the proportion of planted area that was planted with treated seed for each crop.

Neonicotinoid seed treatments in U.S. maize and soybean pest management

To understand the relationship of NSTs to other pest management approaches, for maize and soybeans we compared the temporal trend in percentage of hectares planted with neonicotinoid-treated seed to the percentage of hectares treated with non-seed- treatment insecticides (U. S. Department of Agriculture NASS 2014c), and the percentage of hectares planted with seed treated with any pesticide, as reported in North

Dakota surveys (Glogoza et al. 2002; Zollinger et al. 1996; Zollinger et al. 2006;

Zollinger et al. 2012; Zollinger et al. 1992; Zollinger et al. 2009). For maize, we also looked at the temporal trend in the use of transgenic, insect-resistant crops expressing insecticidal toxins from the bacterium Bacillus thiuringensis (i.e. Bt crops; U. S.

Department of Agriculture ERS 2014b), because these crops have played a major role in maize pest management since their introduction in 1996.

Results

Neonicotinoid use in the U.S. by product type, crop, and active ingredient

The first neonicotinoid (imidacloprid) was registered in the U.S. in 1994, and neonictoinoids now have over 500 registered uses (Table 1-3). Neonicotinoid use increased dramatically from 1994 to 2011, especially after 2003 (Figure 1-1). Insecticide

16 sales data from Minnesota suggest that most neonicotinoids were applied to crops: from

1996 to 2011, 93% of neonicotinoid active ingredients were sold in crop-use products while the remainder were sold mainly in turf/ornamental (4%), structural (1.4%), and lawn/garden (1.2%) products (Figure 1-1a). Within crops, neonicotinoid use was fairly constant in fruits and vegetables, whereas use in field crops increased after 2003. By

2011, just three field crops (maize, cotton, and soybeans) accounted for the vast majority

(~80%) of neonicotinoid use (Figure 1-1b). The active ingredient imidacloprid was dominant for the first half of the study period (Figure 1-1c), but after 2003 the increase in neonicotinoid use coincided with the newer active ingredients clothianidin and thiamethoxam.

Using USGS EPest-low data did not change any of the conclusions about the relative importance of different crops or active ingredients as contributors to neonicotinoid use. The EPest-low and -high estimates generally came to resemble each other over time, with the greatest difference between estimates in 1994 (low estimate

28% lower than high estimate) and the smallest differences in 2010 and 2011 (low estimate 4% and 5% lower than high estimate, respectively).

Neonicotinoids applied as seed treatments in major U.S. row crops

From 2000 to 2012, virtually all neonicotinoids applied to maize, soybeans, and wheat were applied as seed treatments (Table 1-4). In cotton, seed treatments accounted for an estimated 7 to 72% of neonicotinoid use, with estimates less variable after 2004, ranging from 60 to 70% (Table 1-4). Per-seed application rates varied widely, with

17 highest rates on maize and lowest rates on wheat, but per-hectare application rates were more comparable among crops (Table 1-2).

From 2000 to 2011, seed-applied neonicotinoids accounted for a growing proportion of the total mass of insecticide active ingredient applied to maize, soybeans, and wheat (Table 1-5). Neonicotinoid seed treatments accounted for roughly 43% of insecticide mass applied to maize by 2010, 21-23% of mass applied to soybean by

2011/2012, and 25-29% of mass applied to wheat by 2011/2012 (Table 1-5). Cotton was quite different; neonicotinoids accounted for less than 4% of insecticide mass applied from 2000 to 2011 (Table 1-5), reflecting the larger volume of other insecticides applied to this crop.

Hectares planted with neonicotinoid-treated seed in major U.S. row crops

In maize, the percent of hectares planted with neonicotinoid-treated seed increased rapidly after 2003, and by 2011 had reached ≥ 79% under all three scenarios we simulated (Figure 1-2a). Data from Pioneer (Pioneer Hi-Bred International Inc. 2012) most closely matched our 90% low-rate scenario and suggested that 87% of maize hectares were planted with insecticide-treated seed by 2011 (Figure 1-2a). Clearly our

100% low-rate scenario did not reflect reality, because by 2008 it estimated that over

100% of maize hectares were planted with treated seed (Figure 1-2a). The 90% low-rate scenario had also (barely) exceeded 100% of maize hectares by 2011, suggesting that neonicotinoid rates on maize seed were likely increasing toward the end of the study period.

18 In soybeans, the percent of hectares planted with neonicotinoid-treated seed increased steadily starting in 2006 (Figure 1-2d). By 2011, we estimate that NSTs were used on 34-44% of soybean hectares, depending on the prevalence of the low and high application rates and whether the EPest-low or EPest-high estimate was used. Our estimate is consistent with a recent analysis based on proprietary data by the U.S.

Environmental Protection Agency (Myers & Hill 2014), which reported that NSTs were used on an average of 31% of soybean hectares from 2008 to 2012. There may be important regional variability in NST use in soybeans, as suggested by estimates from the

Southern U.S. (ranging from 0% to 75% of hectares across seven states in 2011; Musser et al. 2012) and the Corn Belt (73% of hectares in Iowa in 2009; Hodgson et al. 2012).

With our dataset we were unable to estimate the cotton area planted with NSTs, but some insight can be gleaned from a 2010-2013 survey that asked agricultural professionals working in cotton to estimate the prevalence of insecticidal seed treatments in their regions (Williams 2014). The results suggest that NSTs were used on 52 to 77% of national cotton area over those four years, with significant regional variability (in 2011 ranging from a low of 17% in Arizona, to a high of 96% in Tennessee).

Summing the area planted with neonicotinoid-treated seed for maize, soybean, and cotton, we conservatively estimate that at least 42 million hectares of cropland (57% of the total area) were planted with NSTs by 2011 in these three crops alone, an area roughly the size of California.

19 Neonicotinoid seed treatments and pest management trends in U.S. maize and soybeans

Two trends coincided with rapidly increasing use of NSTs in maize after 2003: i) increased planting of transgenic Bt hybrids, from 29% of hectares in 2003 to 80% of hectares in 2014, and ii) decreased application of non-seed treatment insecticides, from

29% of hectares in 2003 to 12% of hectares in 2010 (Figure 1-2c). Importantly, the introduction of NSTs closely followed introduction of Bt hybrids targeting corn rootworms (Diabrotica spp.), a pest complex that has historically driven insecticide use in U.S. maize (Fernandez-Cornejo et al. 2000). In the 1990s, chemical control for the rootworm complex was dominated by soil-applied insecticides (mainly organophosphates and pyrethroids; Fernandez-Cornejo et al. 2000; U. S. Department of Agriculture NASS

2014c), which may have also protected against some secondary soil pests (e.g. wireworms, grubs, maggots). Because Bt hybrids do not control most secondary pests, and because low- and mid-rates of NSTs do not control the rootworm complex (Obopile et al. 2013), the two technologies are potentially complementary. Importantly, however,

NSTs are now used on almost triple the area historically treated with non-seed treatment insecticides (Figure 1-2a,c); therefore, NSTs (together with Bt hybrids) have more than displaced non-seed treatment insecticide use on an area basis. This finding supports the apparent shift toward an ‘insurance’ paradigm of pest management in maize (Gray 2011), in which transgenic crops and NSTs are deployed even when target pest densities are expected to be low. This notion is also supported by a recent survey, in which 39% of maize growers using NSTs were not targeting any particular pest (Hurley & Mitchell

2014a).

20 In soybeans, use of both seed-applied and other insecticides have intensified over the past several decades (Figure 1-2d,f), a development that can be partly, but not entirely, explained by changing pest pressure. Prior to introduction of soybean aphid

(Aphis glycines) into North America around 2000, soybeans in the Midwestern U.S. were only sporadically challenged by insect pest populations (Boethel 2004; Fernandez-

Cornejo et al. 2000), explaining the historically low insecticide use in this crop (< 1% of area treated, Figure 1-2f). Soybean aphid changed this situation, and is now the most economically important soybean insect pest, often controlled in outbreak years with foliar sprays (Olson et al. 2008; Ragsdale et al. 2011). Seed-applied neonicotinoids rarely displace foliar sprays against soybean aphid, because in most regions the period of insecticidal activity wanes prior to aphid attack (Johnson et al. 2009; McCornack &

Ragsdale 2006; Seagraves & Lundgren 2012), though there are exceptions (Magalhaes et al. 2009). In contrast, NSTs can sometimes displace other insecticide applications against the overwintered generation of bean leaf beetle (BLB, Cerotoma trifurcata) (Bradshaw et al. 2008). Populations of BLB in the Midwest increased dramatically in the late 1990s and early 2000s (Bradshaw & Rice 2003; Giesler et al. 2002), and NSTs were first used on soybeans in Iowa and Wisconsin under an emergency exemption for BLB concerns

(U. S. Environmental Protection Agency 2003). BLB populations in the Midwest have returned to more typical lower levels (O'Neal & Hogsdon 2014), and only 12% of U.S. farmers reported ‘actively managing’ beetles (including BLB) in soybeans (Hurley &

Mitchell 2014b). Nevertheless, use of NSTs on soybeans continues to rise. This trend may reflect that an ‘insurance’ approach to managing insects is also prevalent in

21 soybeans; indeed, 47-65% of farmers using NSTs on soybeans reported they are not targeting any particular pest (Hurley & Mitchell 2014a; Myers & Hill 2014).

An important question is whether NSTs have displaced older seed-applied insecticides, or whether they represent a truly new trend. In North Dakota soybeans, pesticide seed treatments appear to have been uncommon into the early 2000s (Figure

1-2e). In maize, seed-applied insecticides have a longer history, but available evidence suggests that they were not very prevalent prior to neonicotinoid introduction. Dieldrin and lindane (organochlorines) were used to treat seed on significant maize area as early as the 1950s (Lange 1959; Lilly 1956), but dieldrin was discontinued in 1974 because of environmental and human health concerns (Agency for Toxic Substances & Disease

Registry 2002). Lindane was used as a seed treatment as late as the 2000s (Table 1-6), but was used on only ~6% of maize area in 2002 (Brassard & Yusuf 2002). Similarly, some organophosphates and carbamates were available as seed treatments in maize (Table 1-6), but were used on only ~5% of the combined area of maize, soybean, and cotton in the early 1980s (Smith 1987). A few pyrethroid-based seed treatments were introduced in the

1990s (Table 1-6), but were replaced with neonicotinoids because the latter are easier to handle and have systemic activity (Pedersen 2014). Finally, North Dakota data suggest that supplier-applied seed treatments were already widespread in maize before neonicotinoids were introduced (Figure 1-2b), but most of these applications likely involved only fungicides, because most of the older insecticidial seed treatments were applied exclusively on-farm (Table 1-6). Based on the evidence available, we conclude that seed-applied insecticides were uncommon in maize and soybeans before the advent of NSTs.

22 Discussion

Neonicotinoid use in the U.S. increased dramatically after 2003 and was driven by seed treatments on field crops such as maize, soybean, wheat, and cotton. The significant

(> 20%) contribution of neonicotionids to mass of insecticide active ingredient applied on maize, soybeans, and wheat is all the more striking because these insecticides are used at relatively low rates due to their high insect toxicity (e.g., in corn: 18-90 g ai/ha for clothianidin versus 84-185 g ai/ha for tefluthrin [Force® 3G] and 526-1052 g ai/ha for chlopyrifos [Lorsban® Advanced]). If current trends continue, neonicotinoid use could increase considerably further through use of seed treatments on additional crop area (e.g. on soybeans or wheat), or through higher per-seed application rates. In 2013, mid- or high-rate products were apparently widely used (Hurley & Mitchell 2014b) and this year at least one seed company has announced that its ‘standard’ treatment for maize seed will now include the highest labeled rate of NST (1.25 mg ai/seed, five times the low rate)

(Beck's Hybrids 2014).

Our results lead to very different conclusions than analyses that did not consider insecticidal seed treatments. For instance, a recent summary of U.S. pesticide-use trends based solely on USDA-NASS data estimated that neonicotinoids accounted for a maximum of 6% of insecticide hectare-treatments and 1% or less of insecticide quantity in 2005 and 2010.8 In contrast, our estimates suggest that neonicotinoids accounted for

>10% of insecticide quantity and probably far more than 6% of hectare treatments over this period, given their widespread use in large-area crops. Furthermore, analyses based on USDA-NASS data suggested that insecticide use on maize has been declining, with only 12% of maize hectares treated with insecticides in 2010, a trend attributed mainly to

23 Bt hybrids (Benbrook 2012; Osteen & Fernandez-Cornejo 2013; Stokstad & Grullón

2013). Our results are critically different; we too found that quantities of insecticides (i.e., mass of active ingredients) applied to maize declined during the period when Bt hybrids became prevalent, but at the same time the extent of insecticide use almost tripled, to

>75% of maize hectares treated with NSTs by 2011. Our findings are consistent with previous research based on U.S. agricultural census data, which found that insecticide use in maize increased in extent between 2002 and 2007, despite widespread use of Bt hybrids (Fausti et al. 2012). Several analyses on the influence of Bt crops on pesticide-use patterns do not seem to have considered seed treatments (Klümper & Qaim 2014;

Naranjo 2009), and so may have overstated reductions in insecticide use (especially “area treated”) associated with this technology. Clearly, seed treatments should be considered in future assessments of pest management trends.

It would be easier to account for seed-applied pesticides if these products were included in major pesticide use surveys. It is remarkable that almost the entire area of the most widely grown crop in the U.S. (i.e., maize) is now treated with an insecticide, yet we have no public survey data reflecting this trend (USGS data are based on proprietary surveys and do not report the key metric of percent area treated). We made the most accurate estimates we could with publicly available data, but it is clear that significant uncertainty remains in some of our estimates, and additional uncertainty may have been introduced by differences in methodology in our two major data sources. Given the rising prevalence of pesticide seed treatments, not only insecticides but also fungicides and nematicides, it would be valuable for USDA-NASS to update its survey methodology to include seed treatment use. Survey questions could be modified to quantify the percent of

24 crop area planted with treated seed, and the active ingredients and rates applied. If this is impossible due to resource constraints, at a minimum the agency could make it clearer in its data products that seed treatment use is excluded from pesticide use estimates, to ensure that users are aware of limitations of the data.

The rapid rise of NSTs in the U.S. is ultimately a social phenomenon, and several factors may have facilitated this trend. Seed suppliers rather than farmers typically apply neonicotinoids, meaning that both groups have a role in ‘adopting’ this technology, though their relative contributions are unclear. According to the seed-chemical industry, the unique properties of neonicotinoids shifted the perspective of seed suppliers away from seeing seed treatments as a cost of production and toward seeing them as a profit center (CropLife Foundation 2013), creating incentives to strongly market these products to farmers. At the same time, during the recent ethanol boom (2004-2011), U.S. maize and soybean prices more than doubled (U. S. Department of Agriculture NASS 2014c), and seed costs rose dramatically with increasing use of transgenic seed (U. S. Department of Agriculture ERS 2014a), contributing to a perception of NSTs as relatively cheap insurance for expensive seed. This perception may have been bolstered by arrival of soybean aphid, and transient outbreaks of BLB in soybean and Stewart’s wilt, vectored by flea beetles, in maize, which led the U.S. EPA to issue emergency exemptions, bringing these products to the maize and soybean seed markets (U. S. Environmental

Protection Agency 1998; U. S. Environmental Protection Agency 2003). Neonicotinoid seed treatments may also have ‘tagged along’ with other technologies that were attractive to farmers. They are usually one component of larger packages, that, for instance in maize, can include germplasm (i.e., crop variety), up to eight transgenes, and up to six or

25 more different seed treatments (fungicides, nematicides, and insecticides). While many

U.S. maize and soybean farmers appear to value NSTs, a significant minority (21% for maize, 15% for soybean) would reduce or eliminate their use of these products if the same seed were available without them (Hurley & Mitchell 2014b).

The extensive use of NSTs raises the question of how these products relate to

Integrated Pest Management (IPM), the guiding framework for U.S. pest management policy since the 1970s (Kogan 1998; Norris et al. 2002; U. S. Department of Agriculture

National Institute of Food and Agriculture 2013). In an IPM framework, insecticide applications are reserved for situations where monitoring reveals that pest populations have reached levels of economic concern (Norris et al. 2002). Preemptive use of insecticides (as with seed treatments) can be justified within IPM rarely, when two conditions are satisfied. The first is that rescue treatments cannot keep pests under the economic injury level. This does not apply to most foliar pests targeted by NSTs (e.g., bean leaf beetle on soybean), for which scouting protocols and economic thresholds are well developed (Hadi et al. 2012). Detection, however, is challenging for many secondary soil pests targeted by NSTs because their belowground activity is difficult to observe before economic damage occurs, though pre-plant scouting procedures do exist for some soil pests (e.g., wireworms; Simmons et al. 1998). The second condition for preemptive insecticide use is that target pests have a high probability of causing economic damage.

The prevalence of secondary soil pests is poorly documented in most of the U.S., but

‘secondary pests’ are not consistently troublesome. The sporadic nature of these pests may explain why peer-reviewed studies from across the U.S. have not found consistent yield benefits of NSTs under ‘typical’ pest pressure in maize (Cox et al. 2007a; Cox et al.

26 2007b; Jordan et al. 2012; Wilde et al. 2007; Wilde et al. 2004), or soybeans (Cox &

Cherney 2011; Esker & Conley 2012; Gaspar et al. 2014; Johnson et al. 2008; Johnson et al. 2009; Magalhaes et al. 2009; Reisig et al. 2012; Seagraves & Lundgren 2012; Tinsley et al. 2012). These findings, together with the patterns of neonicotinoid use we documented, suggest that NSTs are being used on many hectares where they do not deliver an economic return and cannot be considered part of an IPM approach (Myers &

Hill 2014). This phenomenon is also apparently common in European maize systems

(Furlan & Kreutzweiser 2015). Note that this conclusion should not be extrapolated to other cropping systems, where neonicotinoid use can be more consistent with IPM and where these compounds may more often displace insecticides that are acutely hazardous to human health (e.g., for managing insect vectors in cucurbit production; Fleischer et al.

1998).

The possible unintended consequences of overusing insecticides were summarized in the classic contribution of Stern et al. (1959) : i) evolution of insecticide- resistance in target pests, ii) outbreaks of non-target pests, iii) resurgence of target pests, and iv) negative effects on human health or wildlife. For neonicotinoids, some, but not all, of these unintended consequences are possible or already occurring. History and common sense dictate that the use of NSTs in many fields year after year will select for resistant pest populations. Indeed, resistance to imidacloprid and/or thiamethoxam was recently detected in populations of tobacco thrips (Frankliniella fusca) in the Southern

U.S. (Herbert & Kennedy 2015). Neonicotinoid seed treatments can also exacerbate some non-target pests (spider mites and slugs; Douglas et al. 2015; Smith et al. 2013;

Szczepaniec et al. 2013), suggesting that they are not always ‘cost-free’ insurance. On the

27 other hand, pest resurgence does not appear to have been documented with NSTs, and one of the important, attractive features of neonicotinoids is their low acute mammalian toxicity (e.g. high LD50) relative to older insecticide classes such as organophosphates

(Tomizawa & Casida 2005). Neonicotinoids do have potent activity against many invertebrates, including pollinators (Krupke et al. 2012; Pisa et al. 2015; Whitehorn et al.

2012) and natural enemies of crop pests (Hopwood et al. 2013; Pisa et al. 2015). The risk of neonicotinoids to wild and managed pollinator species is a topic of current debate, and could be modeled in the U.S. by combining our estimates on the extent of neonicotinoid use in various crops with data on bee visitation to specific crop species and the concentrations bees would likely encounter through planting dust and/or floral products.

The influence of neonicotinoids on wildlife is not well investigated in the U.S., but in the

Netherlands neonicotinoid concentrations in surface water have been correlated to declines in aquatic invertebrates and insectivorous birds (Hallmann et al. 2014; Van Dijk et al. 2013). These findings are concerning in light of recent detections of neonicotinoid residues in 76% of samples taken from streams in the U.S. Corn Belt, a substantial increase over insecticide detections in the region in the 1990s (Hladik et al. 2014), consistent with our finding that insecticide use in field crops has expanded dramatically with NSTs.

In conclusion, NSTs are a recently developed insect-pest-management tactic that has become very widespread in U.S. agriculture since the mid-2000s. This development remained anecdotal because it was partially obscured by the lack of information on pesticide seed treatments in the nation’s major pesticide-use survey. Our synthesis of publicly available information indicates that NSTs are being used over a very large area

28 (>40 million hectares), and that in major crops (maize and soybeans) these products are often used as part of an insurance-based approach to pest management that may be reinforced in the seed market by limited availability of neonicotinoid-free seed. This pattern of use may have unintended consequences, namely resistance in target pests, outbreaks of non-target pests, and pollution with detrimental effects cascading to wildlife.

As noted above, some of these effects have already emerged. Rather than seeing neonicotinoids as an ‘either-or’ issue, we believe there is an opportunity to judiciously decrease use of these powerful products, and their attendant risks, using the well- established framework of integrated pest management (IPM). At a minimum, such an approach would include identifying which pests are present, monitoring those pests for their potential to cause injury, and keeping records to identify fields where problems are likely or unlikely in the future. A further step would be to weigh short-term yield benefits of insecticide application against documented risks to non-target organisms and the long- term health of the agroecosystem. Entomologists can support such efforts by better characterizing risk factors for early-season pests targeted by NSTs and developing crop- and region-specific, decision-support tools for neonicotinoid use (Ontario Ministry of

Agriculture Food and Rural Affairs 2014), which are surprisingly scarce. Seed companies could help by increasing availability of neonicotinoid-free seeds. Nonetheless, recent history suggests that IPM will not be widely adopted in U.S. field crops given current incentives and disincentives (as detailed above) for farmers and seed suppliers, which appear to strongly favor an insurance-based approach.

29 Acknowledgements

We thank Wes Stone and the U.S. Geological Society for sharing the data that made this effort possible. We thank Matt O’Neal, Christian Krupke, Bill Freese, Dave

Mortensen, and five anonymous reviewers for insightful comments that improved this paper.

30

Tables

Table 1-1. Summary of the major data sources we relied upon for our U.S. pesticide use estimates, 1992-2012.

Source Response variable (units) Region Crops Years NASSa Non-seed treatment insecticides applied (mass a.i.) Major Maize 1992-2003, 2005, 2010 growing Cotton 1992-2001, 2003, 2005, 2007, 2010 regions Soybeans 1992-2002, 2004, 2006, 2012 Wheat 1992-1998, 2000, 2002, 2004, 2006, 2009, 2012 USGSb Total neonicotinoids applied (mass a.i.) U.S. Maize 1992-2011 Cotton 1992-2011 Soybeans 1992-2011 Wheat 1992-2011 MNDAc Neonicotinoid sales (mass a.i.) Minnesota All crops 1996-2011 Pioneerd Area planted with insecticide-treated seed (%) U.S. Maize 2004-2011 NDSU/NDDAe Area planted with pesticide-treated seed (%) North Dakota Maize 1992, 1996, 2000, 2004, 2008, 2012 Soybeans 1992, 1996, 2000, 2004, 2008, 2012 Wheat 1992, 1996, 2000, 2004, 2008, 2012 a. U. S. Department of Agriculture NASS 2014a; U. S. Department of Agriculture NASS 2014c b. Thelin & Stone 2013 c. Minnesota Department of Agriculture 2014 d. Pioneer Hi-Bred International Inc. 2012 e. Glogoza et al. 2002; Zollinger et al. 1992; Zollinger et al. 1996; Zollinger et al. 2006; Zollinger et al. 2009; Zollinger et al. 2012

31

Table 1-2. Neonicotinoid use rates of some common seed-treatment products on four major crops, based on information from pesticide labels.

Year Range of rates

Crop Active ingredient Product (Registrant) registered e on cropd g/kg seed mg a.i./seed g a.i./ha Maizea Imidacloprid Gaucho 480 (Gustafson, now Bayer) 2000 - 0.16 – 1.34 11 – 96 Gaucho 600 (Bayer) 2003 - 0.16 – 1.34 11 – 96

Clothianidin Poncho 600 (Bayer) 2003 - 0.25 – 1.25 18 – 90

NipsItInside (Valent) 2008 - 0.25 – 1.25 18 – 90

Poncho/VOTiVO (Bayer) 2010 - 0.5 36

Thiamethoxam Cruiser 5FS (Syngenta) 2002 - 0.25 – 1.25 18 – 90 Cottona Imidacloprid Gaucho 480 (Gustafson, now Bayer) 1994 1.88 - 2.5 0.16 – 0.21 24 - 32 Gaucho 600 (Bayer) 2003 - 0.375 56

Clothianidin Poncho/VOTiVO (Bayer) 2011 - 0.424 63

Thiamethoxam Cruiser 5FS (Syngenta) 2002 - 0.3 - 0.375 44 – 56

Soybeanb Imidacloprid Gaucho 480 (Gustafson, now Bayer) 2005 0.63 - 1.25 0.095 – 0.19 40 – 79 Gaucho 600 (Bayer) 2003 0.63 - 1.25 0.095 – 0.19 40 – 79

Clothianidin NipsItInside (Valent) 2009 0.5 0.076 32

Poncho/VOTiVO (Bayer) 2010 0.86 - 1.3 0.13 55

Thiamethoxam Cruiser 5FS (Syngenta) 2004 0.5 - 1 0.076 – 0.15 32 – 63

CruiserMaxx (Syngenta) 2006 0.5 0.076 32 Wheatc Imidacloprid Gaucho 480 (Gustafson, now Bayer) 1995 0.05 - 0.94 0.0019 – 0.035 6 – 114 Gaucho 600 (Bayer) 2003 0.05 - 0.94 0.0019 – 0.035 6 – 114

Clothianidin NipsItInside (Valent) 2010 0.098 - 0.7 0.0037 – 0.026 12 – 85

NipsItSuite (Valent) 2011 0.1 - 0.15 0.0038 – 0.0057 12 – 18

Thiamethoxam Cruiser 5FS (Syngenta) 2002 0.29 - 0.52 0.011 – 0.020 36 – 63

CruiserMaxx Cereals (Syngenta) 2008 0.1 - 0.038 0.0038 12 a. Rates were typically reported on a per seed basis. When reported on a seed weight basis, we assumed 11685 seeds/kg to estimate per seed rates.

32 b. Rates were typically reported on a seed weight basis; we assumed 661387 seeds/kg to estimate per seed rates c. Rates were typically reported on a seed weight basis; we assumed 2645549 seeds/kg to estimate per seed rates d. Based on a search in the Pesticide Product Label System (http://iaspub.epa.gov/apex/pesticides/f?p=PPLS:1) by EPA registration number. The year of registration may not be the year the product was commercially available. e. Assumed the following seeding rates: maize, 71630 seeds/ha; cotton, 148200 seeds/ha; soybean, 419900 seeds/ha; wheat, 3211000 seeds/ha

33

Table 1-3. Summary of neonicotinoids and their major agricultural uses in the U.S., based on regulatory documents from the EPA.

Registrations (#) Year first Information last Special Application Leading agricultural Active ingredient Standard Seed treatment uses registered updated by US-EPA local methods uses (section 3) need Acetamiprida 2002 2012 36 14 Spray, seed Apples, cotton, Canola, mustard, potato treatment pears, oranges Clothianidinb 2003 2011 29 2 Spray, soil Maize Maize, small grains, application, canola, cotton, soybeans, trunk many root/tuber injection, vegetables, sorghum, seed millet, sugar beet, treatment broccoli Dinotefuranc 2004 2011 37 16 Spray, soil Grapes, cantaloupes, n/a application, rice, tomatoes, chemigation watermelons Imidaclopridd 1994 2008 397 38 Spray, seed Maize, potatoes, Maize, canola, cotton, treatment, cotton, lettuce, soybeans, small grains, soil oranges, pecans, many root/tuber application, grapes, broccoli, vegetables, sorghum, trunk apples, tomatoes, sugar beets injection, tobacco, soybeans chemigation Thiaclopride 2003 2012 3 - Spray Apples, pears n/a (cancelled 2014) Thiamethoxamf 1999 2011 55 10 Spray, soil Soybeans, maize, Maize, small grains, application, cotton beans (incl. soybean), seed canola, cotton, many treatment, vegetables, melons, peas, chemigation rice, potato, sorghum, sugarbeet, sunflower a. U. S. Environmental Protection Agency 2014a b. U. S. Environmental Protection Agency 2014b

34 c. U. S. Environmental Protection Agency 2014c d. U. S. Environmental Protection Agency 2014d e. U. S. Environmental Protection Agency 2014e f. U. S. Environmental Protection Agency 2014f

35

Table 1-4: Estimated neonicotinoids applied (kg active ingredient), and neonicotinoid seed treatments (ST) as a percentage of neonicotinoids applied, for major U.S. field crops from 2000 to 2012. Dashes represent combinations for which data were unavailable.

Crop Estimate 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011/12 a Maize USGS, Neonics. (low)b 1 0 68 2 268868 404248 331379 494426 611349 684513 588007 933983 USGS, Neonics. (high)b 1 0 1056 2 353854 452126 353522 543572 661445 694275 594036 947457

NASS, Non-ST neonics c 0 0 0 0 - 0 - - - - 0 -

ST Neonics. (low)d 1 0 68 2 - 404248 - - - - 588007 -

ST Neonics. (high)d 1 0 1056 2 - 452126 - - - - 594036 -

% ST Neonics. (low – high)e 100 - 100 100 - 100 - - - - 100 -

Soy USGS, Neonics. (low)b 0 0 0 0 0 0 73020 83692 184172 300980 593038 535581 USGS, Neonics. (high)b 0 0 0 0 0 0 95358 117693 232589 351661 612878 582474

NASS, Non-ST neonics c 0 0 0 - 0 0 0 - - - - 15120

ST Neonics. (low)d 0 0 0 - 0 0 73020 - - - - 520461

ST Neonics. (high)d 0 0 0 - 0 0 95358 - - - - 567355

% ST Neonics. (low – high)e ------100 - - - - 97

Cotton USGS, Neonics. (low)b 13183 16841 25684 31247 103692 110809 161828 127025 97054 81927 127566 276078 USGS, Neonics. (high)b 17086 35352 36962 49244 115260 118578 173075 133039 102554 91629 131778 290941

NASS, Non-ST neonics c 8203 9957 - 29030 - 42301 - 41499 - - 36287 -

ST Neonics. (low)d 4980 6884 - 2217 - 68508 - 85526 - - 91279 -

ST Neonics. (high)d 8883 25395 - 20214 - 76277 - 91541 - - 95491 -

% ST Neonics. (low – high)e 38-52 41-72 - 7-41 - 62-64 - 67-69 - - 72 -

Wheat USGS, Neonics. (low)b 0 0 0 2 0 2 0 0 0 52216 61462 66717 USGS, Neonics. (high)b 0 0 0 2 0 2 0 0 0 77358 79455 82404

NASS, Non-ST neonics c 0 - 0 - 0 - 0 - - 0 - 0

ST Neonics. (low)d 0 - 0 - 0 - 0 - - 52216 - 66717

ST Neonics. (high)d 0 - 0 - 0 - 0 - - 77358 - 82404

% ST Neonics. (low – high)e ------100 - 100 a. For soybeans and wheat, we used NASS data from 2012 and USGS data from 2011 because data for both years were not available in both datasets

36 b. Data supplied by Wes Stone, from the Pesticide National Synthesis Project; “low” = EPest-low estimate; “high” = EPest-high estimate (Thelin & Stone 2013) c. USDA NASS data (U. S. Department of Agriculture NASS 2014c) was adjusted to reflect national totals (see Methods) d. Calculated by subtracting non-ST neonicotinoids (NASS) from total neonicotinoids (USGS) e. Calculated by dividing ST neonicotinoids from total neonicotinoids

37

Table 1-5: Nationwide estimates of insecticides applied (kg active ingredient), and neonicotinoid seed treatments (ST) as a percentage of insecticides applied, for major U.S. field crops from 2000 to 2012. Dashes represent crop-year combinations for which data were unavailable.

Crop Estimate 2000 2001 2002 2003 2004 2005 2006 2007 2009 2010 2011/12a NASS, Non-ST Maize insecticidesb 4785152 4391551 2743185 3680504 - 2365019 - - - 795493 - Total insecticides (low)c 4785152 4391551 2743253 3680506 - 2769267 - - - 1383500 - Total insecticides (high)c 4785152 4391551 2744240 3680506 - 2817144 - - - 1389529 - % ST Neonics. (low – high)d 0 0 0 0 - 15-16 - - - 43 - NASS, Non-ST Soy insecticidesb 141689 154605 503627 - 278315 1218073 1263443 - - - 1918316 Total insecticides (low)c 141689 154605 503627 - 278315 1218073 1336463 - - - 2438777 Total insecticides (high)c 141689 154605 503627 - 278315 1218073 1358801 - - - 2485671 % ST Neonics. (low – high)d 0 0 0 - 0 0 5-7 - - - 21-23 NASS, Non-ST Cotton b insecticides 19716776 13170763 - 6183366 - 7466940 - 3832852 - 2740149 - Total insecticides (low)c 19721756 13177646 - 6185583 - 7535447 - 3918379 - 2831428 - Total insecticides (high)c 19725658 13196157 - 6203580 - 7543217 - 3924393 - 2835640 - % ST Neonics. (low – high)d 0 0.1-0.2 - 0.0-0.3 - 0.9-1.0 - 2.2-2.3 - 3.2-3.4 - NASS, Non-ST Wheat insecticidesb 306888 - 402185 - 421385 - 182559 - 395512 - 201050 Total insecticides (low)c 306888 - 402185 - 421385 - 182559 - 447728 - 267767 Total insecticides (high)c 306888 - 402185 - 421385 - 182559 - 472870 - 283454 % ST Neonics. (low –

high)d 0 - 0 - 0 - 0 - 12-16 - 25-29 a. For soybeans and wheat, we used NASS data from 2012 and USGS data from 2011 because data for both years were not available in both datasets b. USDA NASS data (U. S. Department of Agriculture NASS 2014c) was adjusted to reflect national totals (see Methods). Cotton values are for upland cotton, > 97% of U.S. cotton hectares from 2000 to 2012. Wheat values are for winter, spring, and durum wheat combined. c. Calculated by adding ST neonicotinoids (Table 1-4) to Non-ST insecticides for each crop d. Calculated by dividing ST neonicotinoids (Table 1-4) by total insecticides for each crop

38

Table 1-6: Some insecticidal seed treatments labeled for maize, based on active ingredients other than neonicotinoids. We do not consider products for stored grain control.

Year Insecticide active Year Class Producta registered Application methodb ingredient(s) cancelledb by EPAb Organochlorine Lindane Isotox Seed Treater (F) ® 1971 2005 Planter-box Grain Guard Plus 1992 2005 Planter-box Organochlorine + Lindane + Diazinon Kernel Guard ® 1986 2005 Planter-box Organophosphate Agrox Premiere ® 1995 2005 Planter-box Germate Plus ® 1987 2004 Planter-box Vitavax ® 1987 2004 Planter-box Organophosphate Diazinon Diazinon 50 WP 1983 2005 Planter-box Chlorpyrios Lorsban ® 30 Flowable 1987 2004 Liquid/slurry treaters Pyrethroid Permethrin Kernel Guard ® Supreme 1998 - Planter-box Pounce ® 25 STD 1998 2010 Commercial seed treaters Tefluthrin Force ® ST/30 CS 1995 - Commercial seed treaters a. Major maize seed treatment products were identified based on key references (Brassard & Yusuf 2002; CropLife Foundation 2013; Glogoza 2005; Smith 1987). One source listed several carbamate insecticides that were little-used as seed treatments (Smith 1987), but we could not find trade names for these products or other evidence of their use.

39

Figures

Figure 1-1. Neonicotinoid sales by product type (a), and use by crop (b) and active ingredient (c) from 1992 to 2011. Data on use (a) is based on sales data from the Minnesota Department of Agriculture (2014). Data on crops and active ingredients are for the entire U.S., from USGS (EPest-High estimate; Thelin & Stone 2013). Y-axes represent mass of neonicotinoid active ingredient in thousands or millions of kg.

40

Figure 1-2. Estimated percent of U.S. maize (a-c) and soybean (d-f) hectares treated with neonicotinoid seed treatments (a,d), any pesticide seed treatment (b,e), and non-seed treatment (Non-ST) insecticides (c,f) from 1996 to 2012. Estimates for national neonicotinoid seed treatment use (a,d) were based on the USGS data (Thelin & Stone 2013) and rates indicated on product labels (Table 1-2), for scenarios with varying amounts of low-rate (LR) versus high-rate treated seed (see text for details). Estimates for overall pesticide seed treatment use (b,e) are for North Dakota (Glogoza et al. 2002; Zollinger et al. 1996; Zollinger et al. 2006; Zollinger et al. 2012; Zollinger et al. 1992; Zollinger et al. 2009).

41

Chapter 2

Neonicotinoid insecticide travels through a soil food chain, disrupting biological control of non-target pests and decreasing soybean yield

This chapter has been published in Journal of Applied Ecology with the citation:

Douglas, M. R., J. R. Rohr, and J. F. Tooker. 2015. “Neonicotinoid insecticide travels

through a soil food chain, disrupting biological control of non-target pests and

decreasing soybean yield.” 52(1): 250-260.

Abstract

Neonicotinoids are the most widely used insecticides worldwide, but their fate in the environment remains unclear, as does their potential to influence non-target species and the roles they play in agroecosystems. We investigated in laboratory and field studies the influence of the neonicotinoid thiamethoxam, applied as a coating to soybean seeds, on interactions among soybeans, non-target molluscan herbivores and their insect predators. In the laboratory, the pest slug Deroceras reticulatum was unaffected by thiamethoxam, but transmitted the toxin to predaceous beetles (Chlaenius tricolor), impairing or killing >60%. In the field, thiamethoxam-based seed treatments depressed activity–density of arthropod predators, thereby relaxing predation of slugs and reducing soybean densities by 19% and yield by 5%. Neonicotinoid residue analyses revealed that insecticide concentrations declined through the food chain, but levels in field-collected

42 slugs (up to 500 ng g-1) were still high enough to harm insect predators. Our findings reveal a previously unconsidered ecological pathway through which neonicotinoid use can unintentionally reduce biological control and crop yield. Trophic transfer of neonicotinoids challenges the notion that seed-applied toxins precisely target herbivorous pests, and highlights the need to consider predatory arthropods and soil communities in neonicotinoid risk assessment and stewardship.

Introduction

Neonicotinoid insecticides are the most widely used class of insecticides worldwide (Sparks 2013), and mounting evidence suggests that they can undermine populations of non-target animals in natural and agricultural ecosystems (van der Sluijs et al. 2015). Neonicotinoids are neurotoxins that can have sublethal effects on bees

(Goulson 2013), and their concentrations in surface waters have been negatively correlated with abundance of aquatic invertebrates and insectivorous birds (Hallmann et al. 2014; Van Dijk et al. 2013).

Despite this recent scrutiny, major gaps remain in our knowledge about the fate of neonicotinoids and their consequences for animal communities, even for invertebrates in agroecosystems where these compounds are most commonly used (Goulson 2013; van der Sluijs et al. 2015). Neonicotinoids are routinely applied as seed coatings to large- acreage crops, such as corn and soybeans, and so are used preventatively on millions of hectares of farmland annually (U. S. Geological Survey 2014) to counter early season insect pests, many of which are sporadic in space and time. Neonicotinoid seed coatings are absorbed systemically into crop tissues, and then decline over the season (Laurent &

43

Rathahao 2003; Sur & Stork 2003). Non-target effects of these coatings might be particularly important for soil organisms, given their proximity to the insecticides

(Goulson 2013). Populations of soil-dwelling arthropod predators have been depressed by neonicotinoid seed coatings in some field studies (e.g. Leslie et al. 2010), but effects have been variable and exposure pathways remain obscure. It is also virtually unknown what significance these effects may have for biological control of pests. There is, therefore, a strong need to investigate more closely the fate of neonicotinoid seed treatments in the soil environment and their influence on predatory arthropods that provide pest-control benefits to farmers.

Molluscan pests (slugs and snails) are often overlooked but in many cropping systems are among the most challenging pests farmers face (South 1992). In Great Britain alone, slugs cost wheat and rapeseed farmers upwards of ₤9 million annually in control costs (U. K. DEFRA 2010). Slugs are also a mounting problem in grain and forage production the Mid-Atlantic U.S., where their populations have increased with adoption of conservation-tillage farming techniques (Douglas & Tooker 2012). The introduced species Deroceras reticulatum Müller and its native congener D. laeve Müller are the major pest slugs in the mid-Atlantic region, where they feed upon emerging seedlings and compromise crop establishment, sometimes requiring costly replanting (Hammond &

Byers 2002).

Slugs are likely to consume neonicotinoids when they feed early in the growing season upon seedlings grown from coated seeds, but as molluscs they may not be sensitive to these insecticides. In Silent Spring, Carson noted “For some reason, snail-like mollusks seem to be almost immune to the effects of insecticides” (1962, p. 257). This

44 rule-of-thumb appears to hold for imidacloprid, which has low acute toxicity to D. reticulatum (Simms et al. 2006; but see effects on freshwater snails, Van Dijk et al.

2013). If slugs ingest neonicotinoids without dying, they may serve as toxic prey for predators that attack slugs, potentially disrupting biological control. Such a phenomenon would bear a strong resemblance to the protection some insect herbivores derive from toxic secondary metabolites, such as nicotine, that they acquire from their host plants

(Kumar et al. 2014; Thorpe & Barbosa 1986). Notably, neonicotinoids and nicotine share a common mode of action, but neonicotinoids are roughly 10,000 times more toxic to insects (Jeschke & Nauen 2008). Predators of slugs in temperate agroecosystems include epigeal beetles, especially certain species of ground beetles (Carabidae; Symondson

2004), which are physiologically susceptible to neonicotinoids (Mullin et al. 2005).

Here we examine the influence of neonicotinoid seed treatments on slug pests and the potential for these insecticides to disrupt slug predators through dietary transfer of the toxin. We studied soybeans Glycine max L. Merr. coated with thiamethoxam because they are the most popular no-till crop in the U.S. (50% of acres; Horowitz et al. 2010), and thiamethoxam is one of two neonicotinoids commonly used on soybeans (U. S.

Geological Survey 2014). We began our investigation with laboratory experiments to test whether: i) slugs are susceptible to thiamethoxam applied as a seed coating, and ii) thiamethoxam and its metabolites move from slugs to their predators. After finding that slugs could transfer neonicotinoids to their predators, we tested whether thiamethoxam influences trophic relationships among soybeans, slugs, and predators in the field. Our primary hypothesis was that thiamethoxam would disrupt predation of slugs, fostering larger slug populations that would in turn hinder soybean establishment, potentially

45 decreasing yield (Figure 2-1). We complemented our experiments with neonicotinoid sampling to quantify insecticide residue transfer through the food chain under laboratory and field conditions. To our knowledge, the results we present here are the first to describe the flow of neonicotinoids through any food chain, and to rigorously investigate the potential for neonicotinoid seed treatments to disrupt biological control under field conditions.

Materials & Methods

Laboratory experiments

Seeds, slugs, and beetles

To explore the influence of soybean seed treatments on slug–predator interactions, we used a single soybean variety (A1016495, FS HiSOY® RR2, Growmark,

Bloomington, IL) treated in one of four ways to represent a range of commercially available seed treatments: 1) untreated control; 2) fungicide-alone (ApronMaxx®, active ingredients [a.i.]: mefenoxam, ~ 0.0068 mg ai seed-1; fludioxonil, ~ 0.0045 mg ai seed-1;

Syngenta, Basel, Switzerland); 3) fungicides plus low rate insecticide (CruiserMaxx®, a.i.: thiamethoxam, 0.0756 mg ai seed-1; Syngenta); and 4) fungicides plus high rate insecticide (thiamethoxam, 0.152 mg ai seed-1). Neonicotinoids on soybean seed are virtually always combined with fungicides.

We collected gray garden slugs D. reticulatum in State College, PA (+40.78, -

77.87) in areas free from insecticide use, primarily an old field and a residential

46 backyard, from early spring to early summer. In our area this species has a roughly annual life cycle with juveniles hatching in spring (Douglas 2012). Because slugs do not have distinct growth stages, we standardized our experiments by slug mass (see details below). We kept slugs at room temperature in covered plastic boxes lined with moist potting soil, and fed them organic cabbage.

We collected adults of the ground beetle Chlaenius tricolor Dejean from crop fields at Russell E. Larson Agricultural Research Farm (LARF; Pennsylvania Furnace,

PA; +40.71, -77.95) using dry pitfall traps and hand collection. This carabid was previously identified as an important slug predator in the eastern U.S. (Eskelson et al.

2011). We housed beetles individually in 16-oz plastic containers (Reynolds Del-Pak®,

Lake Forest, IL) with moist potting soil, in a growth chamber (21oC, 14:10 L:D). Beetles were fed dry kitten food (Purina® ProPlan® Selects®; Nestlé Purina PetCare, St. Louis,

MO) that we moistened with water.

Soybean–slug and slug–ground beetle bioassays

To determine whether seed treatments alter D. reticulatum feeding, in fall 2011 we conducted a factorial experiment with the four types of seed treatments crossed with presence or absence of slugs. The no-slug treatment accounted for possible direct effects of thiamethoxam on plant growth. On day zero, we planted four soybean seeds in 16-oz clear plastic containers, and on day one added one juvenile slug (0.22 ± 0.09 [SD] g) per container and placed them in a growth chamber (21oC, 14:10 L:D; n = 34 containers per treatment with slugs; n = 24 containers per treatment without slugs). For a week, we recorded daily the status of seedlings and slugs. On day eight, we recovered slugs,

47 weighed them and held them for use in ground beetle assays. This experiment was blocked into three consecutive trials due to space limitations.

We next investigated whether D. reticulatum can transmit seed-applied insecticides from seedlings to ground beetle predators. On day eight, we transferred previously weighed slugs to new 16-oz plastic containers with ~1 cm of moist potting soil

(one slug per container), and introduced C. tricolor (starved for six days, 47% male, one beetle per container), randomly assigning beetles to containers (n = 17–19 container per treatment). Six days is within the normal range of starvation for carabids (Bilde & Toft

1998). We tracked the status of slugs and beetles closely for the first 3.5 hours in the evening when beetles were introduced, and then daily for one week when the containers were stored in a growth chamber (21oC, 14:10 L:D). Because neonicotinoids can impair motor control at sublethal doses (Goulson 2013), we recorded beetle flip-time to assess beetle coordination (Lundgren & Wiedenmann 2002). For each beetle, we flipped the beetle on its back using forceps and used a stopwatch to record the time necessary for the beetle to right itself, ending a trial after 30 seconds if the beetle failed to flip over (four trials per beetle per day to reduce variability). Beetle flip-time was bimodal with most beetles flipping either in <1 s or not at all, so we considered a beetle “normal” if it had an average flip time ≤1 s, and “impaired” if it had an average flip time >1 s. From day 8 to

16, the one slug we provided was the only food available to each beetle. Starting on day

16, we maintained beetles with kitten food (Purina® ProPlan® Selects®) in a growth chamber until day 24 of the experiment, when all beetles had either recovered (defined as flip time ≤ 1s) or died. See Appendix A for further details on bioassays.

48

Field experiment

Study site, experimental design, and crop management

To explore the effects of seed-applied insecticides on interactions among plants, slugs, and predators, we conducted a field experiment in 2012 at LARF where plots were arranged in a replicated Latin square design (2 x 6 array), in a field that had been farmed using no-till practices for seven years. Using a four-row planter with graphite as a seed lubricant, we planted a regional soybean variety on May 18th at a rate of 444,600 seeds ha-1 (76-cm row spacing), either with commercially applied fungicide (mefenoxam and fludioxonil, concentrations as above) and insecticide (thiamethoxam, 0.152 mg ai seed-1; n = 6 plots) or without a seed coating (untreated control; n = 6 plots). We only used two experimental treatments to improve statistical power, and because our lab-based results demonstrated that fungicides did not influence slug-soybean or slug-predator interactions

(see Results). Plots (27 × 40 m) abutted one another but we collected all samples in a central area in each plot (15 × 22 m), leaving a buffer of at least 6 m to adjacent plots or edges. We managed weeds in all plots by spraying glyphosate on May 2nd, paraquat on

May 17th, and glyphosate again on June 14th. See Appendix A for more details on crop management and experimental design.

Stand establishment, early season herbivory, and yield

To assess the influence of the seed treatment on crop establishment and productivity, we measured soybean plant populations and herbivore damage during three early soybean growth stages (cotyledon, one-trifoliate, and three-trifoliate stages). On

49 each sampling date, we counted the number of plants in 3-m stretches of row (6 stretches plot-1, random locations). We also examined the first 15 seedlings in each sample for evidence of herbivory, recording damage to each cotyledon (0: none; 1: some; 2: cotyledon missing) and the approximate percentage of leaf area removed (on true leaves) using a four-point scale (0: 0%; 0.4: <10%; 1: 10–25%; 2: 25–50%; 3: 50–75%; 4:

>75%). We harvested soybeans on 9 November (at 12.9% moisture), taking yield samples 30.5-m long and four rows wide (2 samples plot-1).

Invertebrate activity–density

To assess the influence of the treatments on activity–density of slugs and their predators, we installed pitfall traps (4 plot-1; 15 and 26 m down rows 16 and 24 of each plot) that we opened monthly for 72 hours from June–September. Depending on the taxa, we identified captured specimens to genus, family, or order (see Appendix A). To provide an additional measure of slug activity–density, we used square-foot pieces of roofing material (Owens Corning Rolled Roofing, color: Shasta White) as artificial slug shelters (6 plot-1, random locations). We checked shelters in the morning, weekly from planting through harvest, and identified slugs to species in the field (Chichester & Getz

1973; McDonnell et al. 2009).

Predation

In addition to measuring predator activity–density, we more directly measured the prey-consuming function of the generalist predator community by deploying waxworm

50 caterpillars Galleria mellonella L. as sentinel prey. While sentinel slugs would have been more relevant to our study, their lack of exoskeleton makes restraining them difficult and impractical. In our previous work (Douglas 2012), predation on waxworm caterpillars was positively related to activity–density of large (>9 mm) ground beetles, which are thought to be among the most important predators of slugs in agroecosystems (Ayre

2001; Symondson 2004). Within several days of pitfall sample dates (June 8th, July 12th,

August 15th), we deployed sentinel caterpillars (0.21 ± 0.06 g, 10 per plot, equally spaced from 10 to 30 m along the 14th and 22nd row of each plot) in two 12-hour periods (day and night). We restrained each waxworm to a clay ball with an insect pin through its last abdominal segment, and placed each waxworm in the field under a wire mesh cage (mesh size: 1.3 cm) to exclude vertebrate predators (after Lundgren et al. 2006; Appendix A).

Insecticide analyses

To further understand the potential movement of neonicotinoid residues through the plant–slug–beetle food chain, we tested both lab- and field-exposed organisms for neonicotinoid insecticides and their major degradates. We deposited samples into pre- weighed 50-ml tubes and stored them at -80o C before shipping them on dry ice to the

USDA’s National Science Laboratory (Gastonia, NC) for analysis with LC/MS-MS

(methods adapted from Kamel 2010). In June 2012, we repeated a subset of our laboratory experiments to describe neonicotinoid concentrations in organisms in the low and high thiamethoxam treatments. Replication was minimal (n = 2 per treatment for soil and soybeans; n = 1 per treatment for slugs and beetles) because of the expense of insecticide analyses and the need to pool numerous organisms to generate the mass

51 required for an acceptable limit of detection. In the field study previously described, we collected soil, soybean seedlings, and slugs D. reticulatum for insecticide analysis (n = 3 plots per treatment, pooling subsamples within plots). At the cotyledon stage, we sampled the above-ground portion of soybean seedlings and collected soil from cores centered on soybean stems (10 cm deep, 10.8 cm diameter). In the course of our soil sampling, we also found and collected several earthworms (rinsed in water to remove soil particles).

We collected slugs from plants at night at the cotyledon and one-trifoliate stages, and from under refuge traps shortly before soybean harvest, when plants had senesced. See

Appendix A for more details on pooling our subsamples and for reasons why we did not sample predators in the field.

Statistical analyses

We performed all statistical analyses in R 3.1.0 (R Core Team 2014), using the

‘lm’ function for fixed-effects models and the ‘lme’ function for mixed-effects models

(Pinheiro et al. 2013). For repeated measures analyses, we chose among candidate covariance structures using Akaike Information Criteria. We report results based on type

II sum of squares for models with multiple fixed effects. Because our blocking factors

(trial in the laboratory; ‘row’ and ‘column’ blocks in the field) had relatively few levels and were not sampled randomly from a larger population of blocks, we treated them as fixed effects.

52

Laboratory experiments

To test whether seed treatment influenced slug mass gain, we used analysis of variance (ANOVA) with seed treatment and trial as fixed effects. For slug damage to soybeans over time we fit a similar model, but with a random effect of microcosm to account for repeated measurements, and an AR(1) covariance structure.

In the slug–ground beetle experiment, to test whether slug-feeding history influenced likelihood of attack by C. tricolor, we fit a Cox proportional hazards regression model on slug survival, stratified by trial (Therneau 2014). We compared numbers of impaired and normal beetles across treatments using a Fisher’s exact test.

Field experiment

We expected seed treatments to influence mainly early season trophic interactions, so our primary analyses focused on response variables measured during the first two weeks after soybeans emerged (~21–35 days after planting). Because we sampled many different predatory taxa, we created a variable for potential slug predators by summing Carabidae, Lampyridae, Staphylinidae, and Opiliones, the major arthropod groups at our site that include slug predators (Barker 2004). With the early season dataset, we first conducted ANOVAs to test whether seed treatment affected each response in the hypothesized direction (Figure 2-1). Then, to see if seed treatment effects could have been caused by our proposed mechanisms, we fit linear regressions between i) predator activity–density and predation, ii) predation and slug activity–density, iii) slug activity–density and soybean damage, iv) slug activity–density and soybean populations,

53 and v) soybean populations and grain yield, using a Bonferroni-corrected α = 0.01 to account for five interdependent regression analyses. For soybean damage, we created a factor (loadings for leaf damage: 0.89; loadings for cotyledon damage: 0.89) to represent variation in overall soybean damage. All analyses included blocking factors as fixed effects to help account for environmental variation unrelated to treatments.

To examine whether seed treatment had lasting effects on activity–density of slugs and their potential predators, we analysed the remaining sample dates. We fit mixed-effects models with fixed effects of blocks, seed treatment, and their interactions with time, and a random effect of plot to account for repeated measurements.

Insecticide analyses

To describe changes in neonicotinoid residues across trophic levels and test for differences between laboratory and field experiments, we fit a regression model to the combined residue data, treating trophic level as a numeric predictor (soybeans = 0; slugs

= 1; ground beetles = 2) and setting as a categorical predictor (lab, field). We treated trophic level as a numeric predictor because this allowed us to test for a consistent change in neonicotinoid concentration across trophic levels. Laboratory data were from both low and high thiamethoxam treatments and field data were from treated plots in the cotyledon stage.

54

Results

Laboratory experiments

Soybean–slug bioassays

Slugs readily attacked soybean seedlings grown from each of the four seed treatments, with no significant differences in the number of seedlings damaged over time

(Figure 2-2A; Seed treatment: F3,116 = 0.26, P = 0.85; Day*Seed treatment: F18,708= 0.54,

P = 0.94). Slug survival was similar in all treatments (85–94%; Fisher’s exact test, P =

0.46), and there was no evidence that seed treatment influenced slug mass gain (Figure

2-2B; F3,116 = 0.41, P = 0.75). These results suggest that fungicidal and insecticidal seed- coatings did not alter slug herbivory, survival, growth, or behaviour, a finding that we also confirmed in an additional experiment with smaller slugs (Appendix B).

Slug–ground beetle bioassays

Overall, 75% (54 of 72) of C. tricolor beetles killed the slug with which they were confined. Beetles attacked slugs at a similar rate regardless of slug feeding history

(Figure 2-3; Likelihood ratio test, D = 2.18, d.f. = 3, P = 0.54). All beetles appeared normal after eating slugs from the untreated and fungicide-only treatments, while the majority of beetles that fed upon slugs from the low and high thiamethoxam treatments were impaired (Figure 2-2C; Fisher’s exact test, P < 0.0001). Symptoms of beetle poisoning ranged from twitching and mild motor difficulty, to partial paralysis (especially of hind legs), extensive paralysis, and death. Of the sixteen beetles impaired in the two

55 insecticide treatments, seven died (3 high, 4 low thiamethoxam); the rest eventually recovered. Beetles that recovered took several days to do so (4.3 ± 0.4 days [SEM], n =

9).

Field experiment

Experimental conditions and community composition

Slug populations and damage were intense across our region in Spring 2012. We found three slug species at our field site: Deroceras reticulatum, D. laeve, and Arion fasciatus. As expected, the two Deroceras species were dominant, comprising 99% of individuals captured in pitfall samples over the season. Under shelter traps, D. reticulatum and D. laeve accounted for 69% and 19% of observations respectively, while

A. fasciatus accounted for 12%. We observed few above-ground, non-slug herbivores during the early growth stages of soybean except for occasional caterpillars and bean leaf beetles Cerotoma trifurcata Forster (<1 beetle per 10 plants). Diverse natural enemies were represented in our pitfall samples (3,861 individuals), about a quarter of which were potential slug predators (1,052 individuals), represented by Carabidae (49%),

Staphylinidae (36%), Opiliones (12%), and Lampyridae (4%; Table 2-2).

Effects of seed treatment on early season trophic dynamics and yield

Consistent with our predictions, seed treatment had the following significant effects during the first 35 days after planting: reduced activity–density of potential slug

56 predators by 31%, reduced predation on sentinel prey by 33%, increased slug activity– density by 67%, decreased soybean population by 19% (P < 0.02 for all; Table 2-1). Seed treatment also reduced grain yield by 5% (Table 2-1). Furthermore, regression analyses

(with Bonferroni-corrected α = 0.01) revealed that predation was positively associated with activity–density of potential slug predators, and slug activity–density was negatively related to predation (Figure 2-4). In turn, slug activity–density was marginally, negatively related to soybean plant population, which was positively related to grain yield (Figure

2-4). Slug activity–density was not significantly related to soybean leaf and cotyledon damage (partial R2 = 0.52; P = 0.11). With the exception of damage, all of these results were consistent with our hypothesis that seed treatments would disrupt a trophic cascade, indirectly fostering slugs and reducing soybean yield (Figure 2-1).

Effects of seed treatment on slugs and predators over the season

Seed treatment had lasting effects on slug activity–density; slug captures were greater in pitfall traps in treated plots through the end of the season (Figure 2-5;

Treatment: F1,4 = 9.74, P = 0.04; Treatment*Date: F1,4 = 0.08, P = 0.79). Slug activity– density under shelter traps was also consistent with these findings (Figure 2-5). In contrast to slugs, potential slug predators appeared to rebound quickly, with no significant differences in activity–density between treatments after the first sampling date

(Figure 2-5; Treatment: F1,4 = 0.31, P = 0.61; Treatment*Date: F2,8 = 0.48, P = 0.64).

Predation on sentinel prey also seemed to recover quickly (Figure 2-6; Treatment: F1,4 =

0.0, P = 1.0; Treatment*Date: F1,4 = 0.07, P = 0.80).

57

Insecticide residues

From our laboratory-collected samples, we confirmed that neonicotinoid residues travelled up the food chain from soybeans to slugs to beetles (Table 2-3). As in our previous experiment, slugs fed upon thiamethoxam-treated soybeans were poisonous to ground beetles. Of the beetles that ate slugs, 84% in the high-insecticide treatment and

89% in the low-insecticide treatment were impaired the morning after slugs were introduced.

Neonicotinoids were also transferred from soybeans to slugs in the field, and were detected in earthworms (Figure 2-7). Soil, soybean seedlings, and slugs in thiamethoxam- treated plots had neonicotinoid residues several orders of magnitude greater than in control plots.

Neonicotinoids declined exponentially along the food chain, at a similar rate in the laboratory and field (Figure 2-8; Site*Trophic Level: F1,10 = 1.1, P = 0.32).

Concentrations of neonicotinoids were higher in the laboratory than in the field (F1,11 =

22.2, P < 0.001), and, for trophic level, the estimated slope (-3.25) suggests that neonicotinoids declined by ~96% per trophic level (F1,11 = 363.8, P < 0.001; Figure 2-8).

Even so, field-collected slugs contained 500 ppb of neonicotinoids at the cotyledon stage

(~13 ng slug-1), dropping to 177 ppb at the one-leaf stage (~6 ng slug-1), and finally to non-detections by the end of the season (Figure 2-8).

As expected, thiamethoxam was the dominant neonicotinoid residue in our samples, though we also found substantial quantities of its major degradates, especially clothianidin and related metabolites, in all sample types (Figure 2-7). As neonicotinoids moved through the food chain, clothianidin and its metabolites became more prominent

58 as a proportion of total neonicotinoids (Figure 2-7, Table 2-4). Earthworms were the only organisms containing the neonicotinoid imidacloprid (25 and 23 ppb in the two samples, respectively). Total neonicotinoid concentrations in earthworms were 54 and 279 ppb, corresponding to ~16 and 126 ng worm-1.

Discussion

Neonicotinoid seed treatments are intended to maintain yield by protecting plants from target insect pests (e.g. C. trifurcata, Delia platura, scarab larvae), but we have discovered that they can also indirectly decrease yield by disrupting biological control of non-target pests. Moreover, one mechanism contributing to this trophic disruption appears to be a novel phenomenon of slugs passing neonicotinoids from treated plants to their predators. Our findings suggest that i) benefits and costs of neonicotinoids for crop production are likely mediated by the relative importance of target and non-target pests in particular cropping systems, and ii) more broadly, neonicotinoids can move through soil food webs with important consequences for agriculture.

In our field experiment with heavy slug infestations, thiamethoxam seed coatings decreased predator activity–density and predation in the early season, increased slug activity–density, and reduced soybean establishment by 19% and grain yield by 5%

(Table 1-1). Regressions among populations of predators, slugs, and soybeans (Figure

2-4) supported the hypothesis that seed treatments dampened a trophic cascade, fostering larger slug populations that hindered soybean establishment and ultimately decreased yield. Given that insecticidal seed treatments enhanced slug populations and decreased plant populations, it may seem puzzling that leaf and cotyledon damage were similar in

59 treated and control plots. We suspect this result was influenced by a small population (< 1 per 10 plants) of bean leaf beetles (BLB) attacking plants in untreated plots. BLB damages leaves and cotyledons similarly to slugs, but should have been controlled by neonicotinoids in the treated plots (Johnson et al. 2008). The other important insect pest of soybeans in our area is soybean aphid Aphis glycines Matsumura, but we did not quantify this pest because its densities at our site were very low. At a neighbouring site in the same year, soybean aphid did not arrive until 8 August, and peak densities (150 ± 47 aphids plant-1) were well below the economic injury level (McCarville et al. 2014), suggesting that this pest had negligible influence on yield. For slugs, our findings agree with previous evidence that slugs decrease soybean yield mainly by killing plants and thus preventing their establishment, rather than by removing leaf tissue of already established plants (Hammond 2000). Peak slug activity–density in our experiment was roughly four times higher than that measured nearby in the previous two years (Douglas

2012), consistent with reports from no-till farmers that slugs tend to be a problem in our region every two to three years (Douglas & Tooker 2012).

In the laboratory, thiamethoxam did not alter D. reticulatum survival or feeding behaviour, but slugs that fed upon thiamethoxam-treated soybeans were poisonous to the majority of C. tricolor individuals that consumed them, with symptoms ranging from poor coordination to death. These effects on beetles were not driven by fungicides because slugs from the fungicide-only treatment did not poison C. tricolor, consistent with previous toxicity data for mefenoxam and fludioxonil (Mullin et al. 2005).

Concentrations of neonicotinoids in our laboratory experiment were high relative to the field; however, our laboratory experiment was fairly conservative in that predators ate

60 only a single neonicotinoid-exposed slug. Rates of carabid food intake in the field are not well quantified, but many ground beetle species can eat close to their weight in prey each day (Thiele 1977), suggesting that generalist predators could be chronically exposed to neonicotinoids when slugs are abundant.

Some authors have argued that neonicotinoid seed treatments should have negligible influence on natural enemies, because the insecticide is “targeted” in the plant, only reaching herbivorous species (Jeschke et al. 2011). Natural enemies, however, can encounter seed-applied neonicotinoids through omnivory (e.g. Seagraves & Lundgren

2012), and now we have found that they can be exposed via prey, consistent with previous studies where neonicotinoids were applied by other methods (e.g. Szczepaniec et al. 2011). Prey-mediated exposure through non-target pests could be especially disruptive to biological control, because it affects precisely those natural enemies that eat the pests.

Notably, tritrophic movement of neonicotinoids is similar to the mobility of the related plant toxin nicotine. Nicotine from solanaceous plant species is known to influence interactions between caterpillars and their enemies, for instance by reducing parasitoid fitness (Barbosa et al. 1986) and protecting caterpillars from some generalist predators (e.g. Kumar et al. 2014). The analogy to nicotine is intriguing and distribution of nicotine within plants and across natural systems may even suggest clues to the sustainable use of neonicotinoids, but it is important to recognize that effects of neonicotinoids may differ from nicotine because these synthetic compounds are far more toxic to insects (Jeschke & Nauen 2008).

61

The extreme potency of neonicotinoids helps explain why their tritrophic movement negatively influenced predators despite concentrations declining by ~96% per trophic level. Dietary toxicity of neonicotinoids to predators such as C. tricolor and other ground beetles is poorly characterized, but their toxicity to the similar-sized honey bee

(Apis mellifera) is well documented. Lethal doses (i.e., LD50) of thiamethoxam and clothianidin are 4.4 and 3.5 ng bee-1, based on acute oral exposure (Laurino et al. 2013), and doses of thiamethoxam as low as 1.34 ng bee-1 can impair foraging (Henry et al.

2012). For comparison, juvenile slugs in our field experiment contained up to 13 ng of neonicotinoids slug-1. While not the focus of our study, our finding of neonicotinoids in earthworms (16 to 126 ng worm-1) is also concerning for biological control, because earthworms are known to be important prey for generalist predators when other prey are scarce (Symondson et al. 2000). Notably, earthworms were the only organisms at our site that contained imidacloprid, which had not been used for at least one year. Given their burrowing behaviour, earthworms may be more likely to encounter and ingest neonicotinoid residues in soil, but this deserves further study.

An important implication of our study is that neonicotinoid seed treatments may worsen slug problems in managed ecosystems where neonicotinoids and molluscan pests overlap. Deroceras reticulatum is native to western Europe and has invaded North and

South America, Asia, South Africa, and Oceania, causing economic damage to cereals, legumes, canola/rapeseed, strawberries, and myriad ornamental and vegetable species

(McDonnell et al. 2009; South 1992). Neonicotinoids are commonly used in many of these crops and regions (Jeschke et al. 2011). While neonicotinoid seed treatments are currently suspended on bee-attractive crops in the European Union (European

62

Commision 2013), they can still be used on some crops (e.g. wheat) that are prone to slug damage. We predict that neonicotinoids will most likely exacerbate slug problems when their use coincides with the small juvenile stage of slugs, because small slugs are acceptable to a wider range of predatory insects (Ayre 2001). Future research could explore whether neonicotinoids can also flow to predators through caterpillars, which are often not well controlled by neonicotinoid seed treatments (e.g. Kullik et al. 2011), or other insect pest species that tolerate neonicotinoids, including those that have evolved resistance. While our results appeared to be driven solely through top-down mechanisms, it is worth noting that herbivorous mites, another non-target herbivore, can be facilitated by neonicotinoids through both top-down and bottom-up ecological pathways (e.g.

Szczepaniec et al. 2011).

Pesticide regulatory authorities, agricultural organizations, researchers, and the public are struggling to weigh costs and benefits of neonicotinoid seed treatments both within and outside of agriculture. So far, these discussions have focused mainly on managed and wild pollinators, and have only recently widened to include other wildlife such as aquatic invertebrates and birds (e.g. Hallmann et al. 2014; Van Dijk et al. 2013).

Our findings highlight the importance of considering species that contribute to biological control, an ecosystem service conservatively valued at over $200 million per year for

U.S. soybean production alone (Landis et al. 2008). Effects on biological control may help explain why neonicotinoid seed treatments have only inconsistently improved yield in soybeans and other crops. For instance, neonicotinoid-fungicide seed treatments changed soybean yield compared to an untreated control by -8% to +13% across 28 environments, but the factors producing this variability were unknown (Gaspar et al.

63

2014). Our research shows that a better understanding of ecological interactions among target and non-target pests and their natural enemies should allow us to better predict yield responses to neonicotinoid use.

Pest management scientists have long known that pesticides can impose trade-offs in agricultural production, and in fact, such discoveries were part of the impetus behind developing Integrated Pest Management (IPM) as a knowledge-based alternative to the indiscriminate use of pesticides (Stern et al. 1959). In most cropping systems, neonicotinoid seed treatments are being used outside of an IPM framework (e.g. Gray

2011), and, as we show here, this indiscriminant use can have unintended consequences, with measurable costs for farmers. Using neonicotinoids only when and where they are needed, guided by a strong understanding of the underlying ecology, provides potential to harness their strengths and limit their weaknesses to achieve more sustainable pest control.

Acknowledgements

We thank Scott Smiles, Christy Rose Aulson, Amber Delikat, Andrew

Aschwanden, and Janet Teeple for their assistance with experiments, Jonathan Barber and Roger Simonds for their assistance with insecticide testing, and the Tooker Lab,

Franklin Egan, Armen Kemanian, Chris Mullin, Heather Karsten, Mary Barbercheck, and three anonymous reviewers for insights that improved our experiments and this paper.

This work was supported by the USDA’s Northeast Sustainable Agriculture Research and

Extension program (GNE11-014 & LNE09-291), the Pennsylvania Department of

Agriculture (44123709), the Maryland Grain Producers Utilization Board, and the

64

Pennsylvania State University, Department of Entomology. J.R.R. was supported by grants from the US Department of Agriculture (NRI 2006-01370 and 2009-35102-0543) and the US Environmental Protection Agency (CAREER 83518801).

Data accessibility

Data available from the Dryad Digital Repository:

http://doi.org/10.5061/dryad.7s403 (Douglas, Rohr, & Tooker 2014)

65

Tables

Table 2-1. Responses of plants, slugs, and predators in a field experiment comparing soybean plots planted with untreated (Control) or thiamethoxam and fungicide-treated (Neonic) seeds (n = 6 plots/treatment). The “predicted effects” listed here for each response variable are illustrated in Figure 2-1.

Days after Control Neonic Predicted Response (units) %Δ η 2 η 2 F P planting mean ± SE* mean ± SE* effect† ST p sp 1,4 Potential slug predators 22 7.79 ± 0.57 5.38 ± 0.69 - -31 0.80 0.42 15.9 0.016 (#/trap/72 hrs) Predation 21 0.33 ± 0.04 0.22 ± 0.03 - -33 0.85 0.34 24.1 0.008 (prop. killed/24 hrs) Slugs (#/trap/72 hrs) 22 3.71 ± 0.71 6.21 ± 0.67 + +67 0.85 0.40 23.1 0.009 Soybean leaf damage 24 0.28 ± 0.01 0.24 ± 0.02 + -14 0.53 0.17 4.46 0.10 (prop. area removed) Soybean cotyledon damage 25 0.96 ± 0.04 0.89 ± 0.05 + -7 0.38 0.07 2.42 0.19 (rating on 0-2 scale) Soybean establishment 35 17.0 ± 0.23 13.8 ± 0.51 - -19 0.94 0.76 63.8 0.001 (10000 plants/ha) Yield (t/ha) 176 3.56 ± 0.08 3.37 ± 0.24 - -5 0.97 0.05 118.4 < 0.001 * We report raw means and standard errors, but F-tests were based on residual error once variation due to blocks was removed. %ΔST is (meanTHX - meanControl)/meanControl X 100, the percent change due to seed treatment. 2 ηp is partial eta-squared, the percent of non-block variation explained by seed treatment. 2 ηsp is semi-partial eta-squared, the percent of total variation explained by seed treatment.

66

Table 2-2 Activity-density of potential slug predators (mean ± SE) on individual sample dates and cumulatively for the season, in plots that were planted with untreated (control) or thiamethoxam and fungicide-treated (THX) soybean seeds (n = 6 plots/treatment).

Individual samples (#/72 hrs) Life Season total Taxon stage(s) Treatment 12-Jun 16-Jul 21-Aug 24-Sep (#/4 samples)

Opiliones Opiliones spp. J, A THX 0.3 ± 0.1 1.8 ± 0.4 0 ± 0 0.3 ± 0.2 2.3 ± 0.6 Control 0.2 ± 0.1 2.3 ± 0.4 0.1 ± 0.1 0.3 ± 0.1 2.9 ± 0.5 Coleoptera Staphylinidae A THX 2.5 ± 0.4 4.1 ± 0.3 0.3 ± 0.1 0.4 ± 0.1 7.4 ± 0.4 Control 4 ± 0.5 3.6 ± 0.4 0.8 ± 0.3 0.2 ± 0.1 8.6 ± 0.9 Lampyrid larvae L THX 0.5 ± 0.3 0 ± 0 0 ± 0 0 ± 0 0.6 ± 0.4 Control 0.6 ± 0.2 0.1 ± 0.1 0 ± 0 0.3 ± 0.2 1 ± 0.4 Carabid adults A THX 1.9 ± 0.2 1.5 ± 0.6 3.1 ± 0.8 3.5 ± 1.5 10 ± 1.1 Control 2.7 ± 0.2 1.5 ± 0.4 2.2 ± 0.6 4.6 ± 1.6 11 ± 1.7 Carabid larvae L THX 0.2 ± 0.1 0.2 ± 0.1 0.1 ± 0.1 0.4 ± 0.1 0.9 ± 0.3 Control 0.2 ± 0.1 0.1 ± 0.1 0.3 ± 0.2 1 ± 0.3 1.6 ± 0.2 All potential slug J, L, A THX 5.4 ± 0.7 7.7 ± 1.0 3.5 ± 0.7 4.6 ± 1.5 21.2 ± 1.9 predators Control 7.8 ± 0.6 7.6 ± 0.6 3.3 ± 0.6 6.4 ± 1.7 25.1 ± 1.0

67

Table 2-3. Neonicotinoid concentration (ppb) detected in samples from a laboratory experiment investigating soybean-slug-ground beetle interactions in the presence or absence of seed treatments.

LOD (ppb) Sample CLO CLO Treatment Sample type Replicate THX CLO IMI Total description THX, CLO, CLO TZMU, TZMU TZNG IMI TZNG* Low rate Soil† 1 ~ 9 g soil 1 63 28535 403 ND ND ND 28939 THX 2 ~ 9 g soil 1 63 24632 274 ND ND ND 24905 Soybeans 1 7 seedlings 1 63 91100 6600 51 331 8 98090 2 8 seedlings 1 63 50200 1300 ND 193 4 51697 Slugs 1 11 slugs 4 188 458 1750 ND 117 ND 2325 Beetles 1 18 beetles 4 188 23 121 ND ND ND 144 High rate Soil† 1 ~ 9 g soil 1 63 65970 547 ND ND ND 66518 THX 2 ~ 9 g soil 1 63 44088 522 ND ND ND 44609 Soybeans 1 8 seedlings 1 63 161000 11500 95 485 10 173089 2 7 seedlings 1 63 111000 9200 81 618 9 120908 Slugs 1 7 slugs 4 188 2130 5230 51 328 ND 7739 Beetles 1 19 beetles 4 188 51 183 ND ND ND 234 LOD = limit of detection; THX = thiamethoxam; CLO = clothianidin; IMI = imidacloprid; CLO TZMU = CLO metabolite, N-(2-chlorothiazol- 5-ylmethyl)-N- methylurea; CLO TZNG = CLO metabolite, N-[(2-Chloro-5-thiazolyl)methyl]-N’-nitroguanidine; ND = not detected. Tested for but not detected: CLO MNG, TMG; IMI metabolites; Acetamiprid; 6-Chloronicotinic acid, Thiacloprid. * The LOD for metabolites (e.g. CLO TZMU, TZNG) is often higher than for parent compounds (Kamel 2010). †Values are for dry soil. Soil moisture (by mass) was 79% at sampling.

68

Table 2-4 Thiamethoxam as a percentage of total neonicotinoid residues (mean ± SE) in the soybean-slug-beetle food chain in the field and laboratory. Values with different letters are 2 significantly different at α = 0.05 (Field: Kruskal-Wallis χ = 3.86, df = 1, P = 0.05; Lab: F2,5 = 405, P < 0.001, with post-hoc Tukey test).

% THX/total neonics Field Lab Soybeans 98 ± 0.2a 94 ± 1A Slugs 65 ± 0.5b 24 ± 4B Beetles - 19 ± 3B

69

Figures

Figure 2-1. A schematic representation of our hypothesis for the potential influence of seed treatments on the ecological community in no-till soybeans. The “+” and “-” signs indicate the anticipated effect (positive or negative) of the preceding factor on the following factor. We would expect this model to hold when slugs are the dominant early-season soybean herbivore. Based on previous findings that moderate early-season leaf damage has little effect on soybean yield (Hammond 2000), we expected slugs to reduce yield mainly by killing plants rather than by eating leaf tissue.

70

Figure 2-2. Outcomes from laboratory experiments investigating the influence of neonicotinoid seed treatments on interactions between soybeans (G. max), slugs (D. reticulatum), and ground beetles (C. tricolor). Soybean seed treatments: U = untreated, F = fungicide-only, F+L = fungicide + low rate thiamethoxam, F+H = fungicide + high rate thiamethoxam. Error bars show ± one standard error. (a) Number of soybean seedlings (out of four) damaged by slugs over seven days (n = 34 microcosms/treatment; no statistical differences among treatments). (b) Slug mass gain (%) after seven days of feeding on soybean seedlings (n = 34 microcosms/treatment; no statistical differences among treatments). (c) Beetle symptoms after consuming slugs fed upon the four seed treatments; beetles exposed to insecticides via slugs suffered significantly higher frequency of impairment (Fisher’s exact test, P < 0.0001).

71

Figure 2-3. Slug survival when confined individually with C. tricolor after seven days of feeding on soybeans with different seed coatings: U = untreated, F=fungicide-only, F+L= fungicide + low rate thiamethoxam, F+H = fungicide + high rate thiamethoxam. Seed treatment did not influence rates of beetle predation on slugs (Likelihood ratio test, D = 2.18, d.f. = 3, P = 0.54).

72

73

Figure 2-4. Partial regression plots for relationships among organisms in soybean plots planted with untreated (‘Control’) or thiamethoxam and fungicide-treated (‘Neonicotinoid’) seeds (n = 6 plots/treatment). P values test the significance of each partial correlation coefficient (Bonferroni- 2 corrected α = 0.01). Rp is the proportion of non-block variation explained by the predictor 2 (squared partial correlation). Rsp is the proportion of total variation explained by the predictor (squared semi-partial correlation). 95% confidence bands are shown in gray. (a) Activity-density of potential slug predators (#/trap) was positively related to predation on sentinel prey (proportion prey killed). (b) Slug activity-density (#/trap) was negatively related to predation on sentinel prey. (c) Soybean population (10000/ha) was marginally negatively related to slug activity-density (#/trap). (d) Yield (t/ha) was positively related to soybean population.

74

Figure 2-5. Animals per trap (mean ± SE) measured in pitfall traps (A. & B.) or shelter traps (C.) over the season in plots that were planted with untreated (Control) or thiamethoxam and fungicide-treated (THX) soybean seeds (n = 6 plots/treatment). Slugs in pitfall traps (A.) and refuge traps (C.) were numerically more abundant in THX than control plots over the season, whereas potential slug predators in pitfall traps (B.) showed a transient response to seed treatment, being depressed in treated plots on the first sample date. See main text for statistical results.

75

Figure 2-6. Predation on sentinel waxworm caterpillars (mean proportion killed ± SE) during two 12-hour time periods (AM/PM) in plots that were planted with untreated (control) or thiamethoxam and fungicide-treated (THX) soybean seeds (n = 6 plots/treatment). Predation was reduced in treated plots in June when neonicotinoids concentrations were highest, but not in July or August when concentrations would have declined. See main text for statistical results.

76

Figure 2-7. Neonicotinoid concentrations (mean ppb ± SE) in samples collected 12-169 days after planting (number of days for each sample noted in parentheses on the X-axis), from field plots planted with untreated (Control) or thiamethoxam and fungicide-treated (Neonic) soybean seeds (n = 3 plots except for earthworms, where n = 2 plots , listed separately). Thiamethoxam (THX) was the active ingredient applied to the seeds, while CLO is clothianidin + clothianidin TZMU [N-(2-chlorothiazol- 5-ylmethyl)-N-methylurea], metabolites of THX that are also insecticidal. Earthworms (Worms) were only sampled in Neonic plots.

77

Figure 2-8. Concentrations of neonicotinoids in soybeans, slugs (D. reticulatum), and ground beetles (C. tricolor) from laboratory and field experiments. Samples from the field were collected when soybeans were at the cotyledon stage. In the regression equation, “Setting” represents the effect of experiment location (lab vs. field), while “Level” represents the effect of trophic level (soybeans = 0, slugs = 1, beetles = 2).

78

Chapter 3

Neonicotinoid seed treatments move through a no-till corn food web but fail to influence biological control or yield

Abstract

Field corn in the U.S. is almost universally planted with neonicotinoid seed treatments, mainly as ‘insurance’ against secondary soil pests. Despite their ubiquity, their agronomic benefits and trade-offs for biological control are not well studied in many regions, such as the Mid-Atlantic U.S. where no-till soil management is common. On the basis of previous experiments in soybeans, I predicted that seed-applied neonicotinoids would increase corn yield when non-target slug pests were rare but decrease corn yield when slug pests were abundant. In laboratory experiments, I confirmed that slugs are immune to usual doses of neonicotinoids and able to pass these insecticides to their predators. I also found that field-collected slugs, as well as black cutworms, contain concentrations of neonicotinoids that are high enough to affect insect predators (~29 ng/organism). Nonetheless, I found no evidence that neonicotinoid seed treatments disrupted biological control, in a high or low slug year. Neither did neonicotinoids increase yield even when slug populations were low. These results highlight the challenge of predicting the effects of seed-applied neonicotinoids in the face of variability in abiotic conditions and pest communities, and reinforce the importance of monitoring target pests to guide economic use of these products.

79

Introduction

Neonicotinoid seed treatments are widely used in field crops, particularly in corn

(Zea mays), where virtually all seed in the U.S. is treated each year (Douglas & Tooker

2015). The main active ingredients used in this crop are the second-generation neonicotinoids clothianidin and thiamethoxam (U. S. Geological Survey 2014), which are labelled for use against an array of secondary soil pests including grubs (Coleoptera:

Scarabaeidae), wireworms (Coleoptera: Elateridae), seedcorn maggot (Delia platura), and black cutworm (Agrotis ipsilon), and, at higher rates, corn rootworm, a primary pest and one of the most important pests of corn worldwide (Diabrotica spp.; Bayer

CropScience 2010; Gray et al. 2009; Syngenta 2016). It has also been suggested both in the scientific literature and in marketing materials that neonicotinoids provide a physiological ‘vigor effect’ that confers tolerance to abiotic stress, thereby benefiting crop production even in the absence of insect pests (Afifi et al. 2014; Elbert et al. 2008).

Nonetheless, there are few peer-reviewed studies reporting the performance of seed- applied neonicotinoids in corn production, especially across the wide range of climates and management practices present in North America. The few studies that exist have found an inconsistent yield benefit, which emerges most often when target pests occur at economic densities, but not otherwise (Cox et al. 2007a; Cox et al. 2007b; Jordan et al.

2012; Wilde et al. 2007; Wilde et al. 2004).

To my knowledge, there are no peer-reviewed studies exploring the value of neonicotinoid seed treatments in corn under no-till management, which is increasingly prevalent in the Mid-Atlantic U.S. and some other regions (e.g. 66% of corn acres in

Pennsylvania as of 2014, USDA NASS 2014). The stable soil environment and surface

80 residue characteristic of no-till fields can shift the pest community, favoring slugs

(mainly Deroceras spp.) and black cutworms while reducing the prevalence of problems with seedcorn maggot (Stinner & House 1990). Slugs are serious but sporadic pests in the

Mid-Atlantic region, particularly in warm and wet springs when they can reduce corn yield both by reducing plant stands and by eating foliage of young seedlings (Byers &

Calvin 1994). In no-till soybeans, I previously found that slugs are immune to typical rates of thiamethoxam, but can pass the insecticide to insect predators, disrupting biological control and ultimately decreasing yield (Douglas et al. 2015). Using meta- analysis, I also found that natural-enemy populations tend to be reduced by about 16% in fields planted with neonicotinoid seed treatments (Chapter 4). Such reductions in natural- enemy populations could constitute a trade-off of neonicotinoid seed treatments in years when target pest populations are low; that is, they may provide a benefit when target pests are abundant, but a liability when target pests are absent or non-target pests, like slugs, dominate. Concurrent with the soybean study, I conducted a series of similar experiments in corn to better understand the influence of neonicotinoid seed treatments on key pests and natural enemies in the no-till context.

In addition to the agronomic uncertainties that still surround seed-applied neonicotinoids, their environmental fate remains unclear. Roughly 20% of imidacloprid applied as a seed treatment was taken up into corn plants grown in a greenhouse (as reported in Sur & Stork 2003), but I am not aware of any studies that quantify concentrations of second-generation neonicotinoids in corn seedlings, or of any neonicotinoid under field conditions. Neonicotinoid concentrations in soil, water, pollen, and nectar following seed treatment are increasingly well documented (Bonmatin et al.

81

2015), but other than honeybees (Blacquière et al. 2012; Codling et al. 2016), few data report their concentrations in non-target animals. In my previous experiment in soybeans,

I found concentrations of thiamethoxam and its metabolite clothianidin >100 ppb (>10 ng/individual) in slugs and earthworms during the early season in fields planted with thiamethoxam-treated seeds (Douglas et al. 2015), concentrations that are within the range of those associated with lethal and sublethal effects on insects (Blacquière et al.

2012; Pisa et al. 2015).

In this study, I conducted a series of experiments to test the influence of neonicotinoid seed treatments on crop production and biological control in a

Pennsylvania no-till corn system. Based on my laboratory findings and my previous results in soybeans, I predicted that the value of seed treatments in corn would depend on the relative abundance of target and non-target pests; I predicted that seed treatments would have a benefit in the absence of slugs but a cost in the presence of slugs.

Furthermore, I continued to explore the movement of neonicotinoid residues in the soil community and its relevance for pest management and biological control. Given the higher rates of neonicotinoids that are used in corn versus soybean, I predicted that concentrations of the insecticides in field-collected soil organisms would be higher in this crop. Taken together, my results provide insight into the benefits and costs of seed- applied neonicotinoids in Mid-Atlantic no-till corn systems.

82

Materials & Methods

Laboratory experiments

Seeds, slugs, and beetles

To explore the influence of corn seed treatments on slug–predator interactions, I used a single corn variety (TA451-00, T.A. Seeds) treated in one of four ways to represent a range of commercially available seed treatments: 1) untreated control; 2) fungicide-alone (Apron XL®, Maxim® 4FS, and Dynasty®; active ingredients [a.i.]: mefenoxam, 0.005 mg ai seed-1; fludioxonil, 0.0064 mg ai seed-1; and azoxystrobin,

0.0025 mg ai seed-1, respectively; Syngenta, Basel, Switzerland); 3) fungicides plus low rate insecticide (CruiserExtreme®, a.i.: thiamethoxam, 0.25 mg ai seed-1; Syngenta); and

4) fungicides plus high rate insecticide (CruiserExtreme® plus Cruiser® 5FS, thiamethoxam, 1.25 mg ai seed-1). Neonicotinoids on corn seed are virtually always combined with fungicides.

I collected gray garden slugs D. reticulatum in State College, PA (+40.78, -77.87) in areas free from insecticide use, primarily a residential backyard, from early spring to early summer 2013. Because slugs do not have distinct growth stages, I standardized my experiments by slug mass (see details below), and used juvenile slugs of a size that is typical in the region during corn planting season. I kept slugs at room temperature in covered plastic boxes lined with moist potting soil, and fed them organic cabbage.

I collected adults of the ground beetle Chlaenius tricolor Dejean from crop fields at Russell E. Larson Agricultural Research Farm (LARF; Pennsylvania Furnace, PA;

+40.71, -77.95) using dry pitfall traps and hand collection. Beetles were housed

83 individually in 16-oz plastic containers (Reynolds Del-Pak®, Lake Forest, IL) with moist potting soil, in a growth chamber (21oC, 14:10 L:D). Beetles were fed dry kitten food

(Purina® ProPlan® Selects®; Nestlé Purina PetCare, St. Louis, MO) that was moistened with water.

Corn–slug and slug–ground beetle bioassays

To determine whether seed treatments alter D. reticulatum feeding, in summer

2013 I conducted a factorial experiment with the four types of seed treatments crossed with presence or absence of slugs. The no-slug treatment accounted for possible direct effects of thiamethoxam and fungicides on plant growth. On day zero, I planted one corn seed ~3.8cm deep in each 32-oz clear plastic container (W. Y. Industries, Inc., North

Bergen, NJ) with 5cm of sifted potting soil. Three days later, into some of the containers

I introduced two juvenile slugs (0.038 ± 0.013 [SD] g) and placed all containers in a growth chamber (21oC, 14:10 L:D; n = 32 containers per treatment with slugs; n = 28 containers per treatment without slugs). After letting slugs feed for seven days, I recovered them, weighed them, and held them for use in ground beetle assays or insecticide residue testing. I also recovered plants, rinsed their roots to remove soil, and either saved them for insecticide analysis or dried them at 65oC for at least 7 days, and then weighed the dry biomass. This experiment was blocked into two consecutive trials due to space limitations.

I next investigated whether D. reticulatum can transmit seed-applied insecticides from corn seedlings to ground beetle predators, as I previously found for soybeans

(Douglas et al. 2015). After slugs had fed on corn plants for seven days, I transferred

84 them to new 16-oz plastic containers with ~1 cm of moist potting soil (one slug per container), and introduced C. tricolor (starved for six days, 69% male, one beetle per container), randomly assigning beetles to containers (n = 14 containers per treatment).

Because fungicides alone had no effect on beetles in my previous experiments (Douglas et al. 2015), for this experiment I included only the untreated, fungicide plus low-rate insecticide, and fungicide plus high-rate insecticide treatments. I tracked the status of slugs and beetles closely for the first 3.5 hours in the evening when beetles were introduced, then stored them in a growth chamber (21oC, 14:10 L:D) and checked them a final time at 13 hours after the start of the experiment. I then recorded beetle flip-time to assess potential sublethal effects of neonicotinoids on beetle coordination (Lundgren &

Wiedenmann 2002). For each beetle, I flipped the beetle on its back using forceps and used a stopwatch to record the time necessary for the beetle to right itself, ending a trial after 30 seconds if the beetle failed to flip over (four trials per beetle per day to reduce variability). As in my past study, I considered a beetle “normal” if it had an average flip time ≤ 1 s, and “impaired” if it had an average flip time > 1 s. If a beetle could not right itself within 30 s I considered it “immobilized”. Because I harvested beetles from this experiment for insecticide analyses after assessing them at 13 hours, I was not able to determine how many of the immobilized beetles would have eventually recovered.

However, in my past experiment immobilization was strongly associated with eventual mortality (Douglas et al. 2015).

85

Field experiments

Study site, experimental design, and crop management

To explore the effects of seed-applied insecticides on interactions among plants, slugs, and predators, I conducted field experiments in 2012 and 2013 at LARF. In both years, plots were arranged in a replicated Latin square design (2 × 6 array) in fields that had been farmed using no-till practices for at least 7 years.

In 2012, using a two-row planter with graphite as a seed lubricant, I planted a conventional corn variety (TA451-00, T.A. Seeds) on May 4th at a rate of 74,132 seeds ha-1 (76-cm row spacing), either with commercially applied fungicide plus insecticide seed coating (THX; Cruiser Extreme® 1250, a.i.: mefenoxam, 0.005 mg ai seed-1; fludioxonil, 0.0064 mg ai seed-1; azoxystrobin, 0.0025 mg ai seed-1; and thiamethoxam,

1.25 mg ai, seed-1; Syngenta, Basel, Switzerland) or with only the fungicide components of the above (Control; n = 6 plots). Plots (30.5 × 32 m) abutted one another but I collected all samples in a central area in each plot (18 × 18 m), leaving a buffer of at least

6 m to adjacent plots or edges. This site had not received insecticide applications

(including seed treatments) for at least one year. Weeds were managed in all plots by spraying glyphosate on April 18th, and atrazine and metolachlor on May 11th. The corn was fertilized with manure on March 19th and side-dressed with urea-ammonium nitrate on June 19th.

In 2013, using a four-row planter with graphite as a seed lubricant, I planted a conventional corn variety (TA477-18, T.A. Seeds) on May 7th at a rate of 74,132 seeds ha-1 (76-cm row spacing), either with commercially applied insecticide seed coating

86

(CLO; Poncho®; a.i.: clothianidin, 1 mg ai, seed-1; Bayer CropScience, Monheim am

Rhein, Germany) or with no seed coating (Control; n = 6 plots). I had initially planned to repeat the treatment comparisons from 2012, but the seed dealer was no longer able to supply these treatments. This experiment was conducted at a site where my similar soybean experiment was conducted the previous year (Douglas et al. 2015), and so the plots had a 1-year history of either neonicotinoid use or no insecticides. Plots (27 × 40 m) abutted one another but I collected all samples in a central area in each plot (15 × 22 m), leaving a buffer of at least 6 m to adjacent plots or edges. Weeds in all plots were managed by spraying glyphosate on April 25th, and gramoxone, atrazine, and S- metolachlor on May 10th. I fertilized the corn at planting (10-30-10, N-P-K), and side- dressed with urea-ammonium nitrate on July 1st.

Stand establishment, early season herbivory, and yield

To assess the influence of the seed treatment on crop establishment and productivity, I measured corn plant populations and herbivore damage during three early corn growth stages (when the plants had one [V1], three [V3], or five [V5] expanded leaves). On each sampling date, I counted the number of plants in 3-m of row (6 row areas plot-1, random locations). I also examined the seedlings in each sample for evidence of herbivory, recording the approximate percentage of leaf area removed on each plant in the sample using a four-point scale (0: 0%; 0.4: <10%; 1: 10–25%; 2: 25–50%; 3: 50–

75%; 4: >75%). Corn was harvested on 28 November 2012 and 19 November 2013, taking yield samples two rows wide (2 samples plot-1). Yield measurements were standardized to 15.5% moisture.

87

Invertebrate activity–density

To assess the influence of the treatments on activity–density of slugs and their predators, I installed pitfall traps (4 plot-1; Reynolds Del Pak ® 16-oz plastic container).

They were located 11 and 20 m from the plot edge in the 14th and 26th row of each plot in

2012, and 15 and 26 m from the plot edge in the 12th and 24th row of each plot in 2013.

Using 50% propylene glycol as a killing agent, traps were opened monthly from May to

September, for 48 hours in 2012 and 120 hours in 2013. Upon collection, I strained samples (1-mm mesh) and stored the collected specimens in 80% ethanol. I identified slugs to genus (some characters cannot be assessed on preserved specimens; Chichester

& Getz 1973; McDonnell et al. 2009) and most predators to family or order (Triplehorn

& Johnson 2005; Ubick et al. 2005). To provide an additional measure of slug activity– density, I used square-foot pieces of roofing material (Owens Corning Rolled Roofing, color: Shasta White) as artificial slug shelters (6 plot-1, random locations). I checked shelters in the morning, weekly from planting through the end of October, and identified slugs to species in the field (Chichester & Getz 1973; McDonnell et al. 2009).

Predation

In addition to measuring predator activity–density, I more directly measured the prey-consuming function of the generalist predator community by deploying waxworm caterpillars Galleria mellonella L. as sentinel prey. I previously found this method to be a good indicator of slug predation (Douglas et al. 2015). Three times over the season

(2012: May 25th, June 26th, July 30th; 2013: June 4th, July 3rd, August 1st), sentinel

88 caterpillars were deployed (10 per plot) in two 12-hour periods (day and night).

Caterpillars were spaced 3 m apart from 9 to 21 m along the 16th and 24th row of each plot in 2012, and 4 m apart from 10 to 30 m along the 14th and 22nd row of each plot in

2013. I restrained each waxworm to a clay ball with an insect pin through its last abdominal segment, and placed each waxworm in the field under a wire mesh cage (mesh size: 1.3 cm) to exclude vertebrate predators (after Lundgren et al. 2006). After the first

12 h, I replaced any attacked, missing, or compromised caterpillars. I recorded caterpillars as whole and alive, partially eaten, or missing. Dead caterpillars showing no sign of predation were excluded from analyses.

Insecticide analyses

To further understand the potential movement of neonicotinoid residues through the plant–slug–beetle food chain, I tested both lab- and field-exposed organisms for neonicotinoid insecticides and their major degradates. I deposited samples into pre- weighed 50-ml tubes and stored them at -80o C before shipping them on dry ice to the

USDA’s National Science Laboratory (Gastonia, NC) for analysis with LC/MS-MS

(methods adapted from Kamel 2010). I saved a subset of organisms generated in the laboratory experiments to describe neonicotinoid concentrations along the food chain in the low and high thiamethoxam treatments and to verify the absence of neonicotinoids in the untreated control. Replication was minimal (n = 2 per treatment for soil and corn seedlings; n = 1 per treatment for slugs and beetles) because of the expense of insecticide analyses and the need to pool numerous organisms to generate the mass required for an acceptable limit of detection (43-46 slugs/sample, 13 beetles/sample).

89

In the field study previously described, I collected soil, corn seedlings, and invertebrate pests for insecticide analysis (n = 3-4 plots per treatment, pooling subsamples within plots). When the plants had two fully expanded leaves (i.e. growth stage V2), I sampled the aboveground portion of corn seedlings (4 seedlings/plot) and collected soil from cores centered on corn stems (10 cm deep, 10.8 cm diameter). In 2013

I separately saved the belowground portion of the corn seedlings after rinsing them in water to remove soil particles. In 2012, both slugs (Deroceras reticulatum) and black cutworms (Agrotis ipsilon) were present at the site and so I sampled both to test for insecticide residues. When the corn was at V2 I collected slugs from plants at night

(approximately 9 pm to 12 am; 11-91 slugs/plot), and collected cutworms by searching for them in areas around cut plants (2-8 caterpillars/plot). I also collected adult slugs from under refuge traps shortly before corn harvest (4 slugs/plot). In 2013, herbivore populations were insufficient to repeat this sampling effort.

Statistical analyses

I performed all statistical analyses in R 3.2.3 (R Core Team 2015), using the ‘lm’ function for fixed-effects models and the ‘lme’ function for mixed-effects models

(Pinheiro et al. 2013). For repeated measures analyses, I chose among candidate covariance structures using Akaike Information Criteria (Littell et al. 2006). Because the blocking factors (trial in the laboratory; ‘row’ and ‘column’ blocks in the field) had relatively few levels and were not sampled randomly from a larger population of blocks, I treated them as fixed effects (Newman et al. 1997). Where necessary, I transformed data before analysis to conform to parametric assumptions.

90

Laboratory experiments

To test whether seed treatment influenced slug survival, I used Fisher’s exact test on the number of slugs surviving and dead in each treatment at the end of the experiment.

For those replicates in which both slugs survived, I tested whether seed treatment influenced slug mass gain (pooled for the two slugs in each replicate) using analysis of variance (ANOVA) with seed treatment and trial as fixed effects. To test whether seed treatments protected corn seedlings from slugs, I used ANOVA with final plant biomass as the response variable and fixed effects of trial, seed treatment, presence of slugs, and the interaction between seed treatment and slugs. The realized designs for the two

ANOVAs were slightly unbalanced, so I report results based on Type II sum of squares.

In the slug–ground beetle experiment, to test whether slug-feeding history influenced likelihood of attack by C. tricolor, I fit a Cox proportional hazards regression model on slug survival, stratified by trial (Therneau 2014). I compared numbers of impaired, immobilized, and normal beetles across treatments using Fisher’s exact test.

Field experiments

I expected seed treatments to mainly influence early season trophic interactions, so the primary analyses focused on response variables measured during the early part of the season (~20–40 days after planting). I created a variable for potential slug predators by summing Carabidae, Lampyridae, Staphylinidae, and Opiliones, the major arthropod groups at the research site that include slug predators (Barker 2004). With the early season dataset, I conducted ANOVAs to test whether seed treatment affected each

91 response in the hypothesized direction. My predictions were contingent on the composition of the pest community, as follows. In the presence of heavy slug populations, I expected neonicotinoids to reduce the activity-density of slug predators, increase the activity-density of slugs and their associated leaf damage, and ultimately decrease plant yield, as I found previously for soybean (Table 3-1). In the absence of heavy slug populations, I expected neonicotinoids to have a modest negative influence on predators as suggested by my meta-analysis (Chapter 4). However, I also expected neonicotinoids to reduce plant damage from various insect pests and improve stand establishment, ultimately increasing yield (Table 3-2).

To examine whether seed treatments had seasonal effects on activity–density of slugs and their potential predators that were not apparent in the early season sample, I conducted a seasonal analysis that included all of the sample dates. I fit mixed-effects models with fixed effects of blocks, seed treatment, and their interactions with time, and a random effect of plot to account for repeated measurements. Where significant interactions occurred, I tested for significant differences between treatments within dates through pairwise comparisons using ‘lsmeans’ (Lenth 2016). For sentinel prey data, there was an additional factor of time (day vs. night) in the analysis. I fit a repeated measures model similar to the one for the activity-density data, but with time and its interactions with treatment and date included.

Insecticide analyses

I report much of the insecticide residue data descriptively. Additionally, to describe changes in neonicotinoid residues across trophic levels and test for differences

92 between laboratory and field experiments, I fit a regression model to the combined residue data, treating trophic level as a numeric predictor (corn = 0; slugs and cutworms

= 1; ground beetles = 2) and setting as a categorical predictor (lab, field). Laboratory data were from both low and high thiamethoxam treatments and field data were from 2012 in thiamethoxam-treated plots in the two-leaf stage (V2). I fit the model to the natural log of neonicotinoid concentration because this exponential model appeared to be a better fit than a linear model.

Results

Laboratory experiments

Corn–slug bioassays

Slugs reduced plant biomass during the experiment (Figure 3-1A; Slugs F1,201 =

11.1, P = 0.001), and there was no evidence that seed treatments protected corn seedlings from slug damage (Seed treatment × Slugs F3,201 = 0.67, P = 0.57). Regardless of whether slugs were present or not, plant biomass was higher in the treatments containing fungicides (Figure 3-1A; Seed treatment F3,201 = 8.68, P < 0.001). Slug survival was high in all treatments (91-98%; Fisher’s exact test, P = 0.19), and in those microcosms in which all slugs survived (n = 26-31 replicates/treatment), there was no evidence that seed treatment influenced slug mass gain (Figure 3-1B; F3,107 = 0.96, P = 0.42). Notably, slugs gained little or no mass on average over the 7 d of the experiment (Figure 3-1B), suggesting that corn may not be a very good food plant for this species. Overall these

93 results suggest that fungicidal, but not insecticidal, seed treatments influenced plant growth, but none of the seed treatments killed slugs or deterred them from feeding upon corn seedlings.

Slug–ground beetle bioassays

Overall, 93% (39 of 42) of C. tricolor beetles killed the slug with which they were confined, with most beetles attacking slugs within the first two hours after they were introduced (Figure 3-2). There was no evidence that slugs were more or less likely to be attacked if the slug had eaten neonicotinoid-treated plant tissue (Figure 3-2; Likelihood ratio test, D = 2.29, d.f. = 2, P = 0.32). All beetles appeared normal after eating slugs that fed upon untreated corn seedlings, while the majority of beetles eating slugs from the low and high thiamethoxam treatments were impaired (Figure 3-1C; Fisher’s exact test, P <

0.0001). Outcomes for the low and high thiamethoxam treatments differed significantly from the untreated control (Bonferroni-corrected pairwise Fisher’s exact tests, U vs. F+L:

P < 0.01, U vs. F+H: P < 0.0001). While there was a trend for more severe impairment for beetles in the high thiamethoxam treatment (Figure 3-1C), the thiamethoxam concentrations did not differ significantly in their effects on beetles (Bonferroni-corrected pairwise Fisher’s exact tests, F+L vs. F+H: P = 0.19).

94

Field experiments

Experimental conditions and community composition

The two field seasons in this study represented opposite extremes in the range of spring precipitation in the region. Compared to 20-year averages for May rainfall (9.1 cm; NOAA 2016), May 2012 was more than twice as wet (19.5 cm of rain) while May

2013 was quite dry (4.8 cm of rain; according to weather data collected at LARF). In

2012, the field site flooded so that five experimental plots (out of twelve) were under water for roughly two weeks in late May and early June. During this period pitfall trapping was impossible, so the first sample was delayed.

The slug community was dominated by D. reticulatum at both sites (70% and

65% of slugs under shingle traps in 2012 and 2013, respectively), with D. laeve the next most abundant species (29% and 19% in 2012 and 2013, respectively), and relatively few individuals of Arion subfuscus (1% in 2012) or A. fasciatus (16% in 2013). In 2012, I also observed significant populations of and damage from black cutworm (Agrotis ipsilon; see results below).

In 2012, the pitfall samples comprised 3,000 arthropod natural enemies, dominated by Formicidae (28%), Staphylinidae (23%), parasitic Hymenoptera (15%), and Carabidae (14%). In 2013, I expanded the sampling window and so collected more natural enemies (7,841 individuals), dominated by Formicidae (51%), Staphylinidae

(14%), and Carabidae (14%). Potential slug predators comprised 38% of the natural enemy community in 2012 and 34% of the natural enemy community in 2013.

95

Effects of seed treatments on early season trophic dynamics and yield

Not surprisingly given the ample rainfall, 2012 was characterized by heavy slug damage, with corn seedlings losing roughly 40% of their leaf area by V3 (Table 3-1).

However, contrary to my hypothesis, predator abundance, predation, slug abundance, and foliar damage did not differ significantly between neonicotinoid-treated and control plots

(Table 3-1). The experimental plots also experienced economic damage by black cutworms, which did not differ by insecticide treatment; both thiamethoxam-treated and fungicide-only control plots experienced > 4% cut plants at growth stage V3 (Table 3-1).

Seed treatment with thiamethoxam did not significantly influence crop establishment or yield, and the numerical trend was for neonicotinoid treatment to increase plant establishment but lower yield (Table 3-1).

In 2013, slug activity was generally low during corn emergence and early crop growth, and foliar damage was < 1% of plant tissue at V3 (Table 3-2). There was also no cutworm damage in either treatment (Table 3-2), and I did not see evidence of any other early season insect pests. Once again, neonicotinoid seed treatment did not significantly influence predator activity-density or crop establishment (Table 3-2). There was a numerical but not significant trend for yield to be higher in the neonicotinoid-treated plots (Table 3-2).

Seasonal effects of seed treatments on slugs

In 2012, consistent with the early season results, slug activity-density over the season was similar across seed treatments (Figure 3-3A, C). Slug activity-density

96 measured by shelter traps did not differ between thiamethoxam-treated and fungicide- only control plots on any date, or overall (Date × Treatment F5,20 = 0.77, P = 0.59;

Treatment F1,4 = 1.55, P = 0.28). In the pitfall trap data, there was a significant interaction between seed treatment and date (F3,12 = 9.18, P = 0.002), driven by greater activity- density of slugs in the fungicide-only control in early August (t = -5.0, P = 0.007; Figure

3-3C). By September, however, slug activity-densities in the two treatments were almost identical (Figure 3-3C).

In 2013, slug activity-density tended to be greater where plots were planted with clothianidin-treated seed (Figure 3-3B, D). Slug activity-density as measured by shelter traps was higher in clothianidin-treated plots than in untreated control plots (Treatment

F1,4 = 17.4, P = 0.014), and this pattern did not change significantly over time (Date ×

Treatment F5,20 = 1.22, P = 0.34). A similar pattern was seen in slug activity-density measured by pitfall trapping, with numerically higher activity-density in clothianidin- treated plots on four out of five sample dates (Figure 3-3B). Statistically, given low mean values and relatively high variability, this trend was significant on only one of the dates

(Figure 3-3B; Treatment F1,4 = 1.15, P = 0.34; Date × Treatment F4,15 = 4.16, P = 0.018).

Seasonal effects of seed treatments on predators and predation

In 2012, seed treatment with thiamethoxam did not significantly influence activity-density of slug predators (Figure 3-4A; Treatment F1,4 = 0.00, P = 0.94; Date ×

Treatment F3,12 = 0.87, P = 0.48), or all arthropod predators (Figure 3-4C; Treatment F1,4

= 0.27, P = 0.63; Date × Treatment F3,12 = 0.28, P = 0.84). Predation on sentinel caterpillars was higher at night than in the day (Time F1,38 = 68.7, P < 0.0001) and

97

increased over the season (Date F2,38 = 116.2, P < 0.0001), but did not differ significantly between thiamethoxam-treated plots and fungicide-only control plots (Treatment F1,4 =

0.19, P = 0.69; Date × Treatment F2,38 = 1.15, P = 0.32; Date × Treatment ×Time F2,38 =

0.42, P = 0.66; Figure 3-5A).

Similarly, in 2013, seed treatment with clothianidin did not significantly influence activity-density of slug predators (Figure 3-4B; Treatment F1,4 = 0.00, P = 0.99; Date ×

Treatment F4,15 = 0.69, P = 0.61), or all arthropod predators (Figure 3-4D; Treatment F1,4

= 1.54, P = 0.28; Date × Treatment F4,15 = 0.55, P = 0.70). Predation on sentinel caterpillars was higher at night than in the day (Time F1,38 = 6.3, P = 0.016) and increased over the season (Date F2,38 = 196.1, P < 0.0001), but did not differ significantly between clothianidin-treated plots and untreated control plots (Treatment F1,4 = 0.21, P = 0.67;

Date × Treatment F2,38 = 0.83, P = 0.44; Date × Treatment ×Time F2,38 = 0.18, P = 0.83;

Figure 3-5B).

Insecticide residues

From the laboratory-collected samples, I confirmed that neonicotinoid residues travelled from corn to slugs to beetles (Table 3-3). As insecticides moved up this food chain, the metabolite clothianidin increased as a percentage of total neonicotinoids (Table

3-3), suggesting either continued metabolism of the parent thiamethoxam at each trophic level, and/or differential excretion of the two compounds.

Neonicotinoids were also transferred from corn to slugs and cutworms in the field

(Table 3-4). During collection, cutworms were alive and showed no obvious signs of poisoning, and yet contained an average of 100 ppb neonicotinoids (29 ng/caterpillar;

98

Table 3-4). Slugs collected at V2 contained an average of > 300 ppb neonicotinoids (29 ng/slug), but at the end of the season neonicotinoids were not detectable (Table 3-4). In

2013, analysis of the aboveground and belowground portions of the corn seedling showed that 98% of the total neonicotinoid in the plant was in the belowground portion, which included the remnant seed coat (Table 3-4). Taken together, the belowground and aboveground portion of the corn plant at V2 (23 d after planting) contained 17,813 ng clothianidin/seedling, less than 2% of the amount originally applied (1 mg/seed).

Neonicotinoid residues in soil, corn seedlings, slugs, and cutworms in thiamethoxam- treated or clothianidin-treated plots were several orders of magnitude greater than in control plots (Table 3-4).

The regression model fit to the laboratory and 2012 field insecticide data showed that neonicotinoid concentrations declined exponentially along the food chain (Figure

2 3-6; overall F3,16 = 21.2, P < 0.0001, R adj = 0.76), and this decline was faster in the laboratory than in the field (Site×Trophic Level: F1,16 = 5.0, P = 0.039). Based on the fitted model, the change in neonicotinoid residues along the food chain can be estimated by the equations:

Laboratory: ln(Neonic ppb) = 8.42 – 2.63Trophic level

Field: ln(Neonic ppb) = 6.38 – 1.22*Trophic level

While the initial concentrations of neonicotinoids in corn shoots were higher in the laboratory than in the field, they declined by 93% per trophic level in the laboratory but only 70% per trophic level in the field. This meant that concentrations of neonicotinoids in pest herbivores (slugs and cutworms) were similar in laboratory and field-collected samples (Figure 3-6).

99

Discussion

The results from the laboratory experiment confirmed my earlier findings that the pest slug D. reticulatum is largely unaffected to neonicotinoids, and can pass these insecticides to beetle predators. Seed treatment with thiamethoxam did not alter slug survival or feeding damage to corn seedlings. In contrast, beetles fed on slugs containing neonicotinoids in the laboratory contained ~1-4 ng of combined thiamethoxam and clothianidin, and this level of toxin exposure was associated with a majority of beetles displaying neurotoxic symptoms ranging from poor motor coordination to complete immobilization at 13 hours after slug prey were introduced. The concentration of neonicotinoids in slug prey in my laboratory experiment (60-350 ppb) was similar to the dietary LD50 of thiamethoxam (in water) for the predatory stinkbug Podisus nigrispinus of 50-60 ppb (Torres & Ruberson 2004). That study and studies on honeybees

(Blacquière et al. 2012) have found that neonicotinoids tend to be more toxic by ingestion than by contact exposure, suggesting that the presence of neonicotinoids in prey or other food poses a special risk to natural enemies.

The results from this study combined with those of my soybean study highlight the importance of crop architecture in mediating the fate of neonicotinoid seed treatments in plants and the environment. Despite the higher rates of thiamethoxam applied to corn

(0.25 and 1.25 mg ai/seed) versus soybean (0.07 and 0.15 mg ai/seed), the aboveground portions of laboratory-grown and field-grown corn seedlings had concentrations of neonicotinoids an order of magnitude lower than similarly grown and sampled soybean seedlings. This could be because in soybeans, half or more of the neonicotinoid residues in young treated seedlings are in the cotyledons (Stamm et al. 2015), whereas in corn, a

100 monocot, most of the neonicotinoid residues from our field-collected seedlings were belowground, where the seed coat remained. This difference in distribution likely influences the degree to which different organisms come into contact with neonicotinoid seed treatments and their metabolites. For instance, slugs tend to prefer feeding on fleshy cotyledons, a behaviour which could increase their exposure to neonicotinoid residues in dicotyledonous crops. On the other hand, earthworms may encounter greater concentrations of neonicotinoids in the rhizosphere of monocotyledonous crops.

Interestingly, however, slugs collected in the field from corn seedlings had higher concentrations of neonicotinoids than those collected in soybean. This discrepancy could be explained by slugs feeding belowground as well as aboveground on corn seedlings; under the extremely wet conditions of our experiment we did observe slug damage to belowground plant parts.

The effects of neonicotinoid seed treatments on slug populations in the field experiments differed in the two years of the experiment. In 2012, contrary to my hypothesis, I did not detect a decrease in predator activity or an increase in slug activity where thiamethoxam seed treatment was used. Interpreting this outcome is challenging because five out of the twelve experimental plots (3 control, 2 thiamethoxam) were underwater during the crucial first few weeks after corn emergence. Even predatory arthropods that inhabit natural floodplains (including some species of carabid beetles) mostly rely on immigration into and out of flooded areas to survive (Rothenbücher &

Schaefer 2006). It is therefore possible that the standing water drove slug predators out of fields temporarily, precluding a strong test of my hypothesis. Alternatively, the excess water could have diluted the insecticides or leached them away from plants (Bonmatin et

101 al. 2015). This seems an unlikely explanation for an absence of neonicotinoid effect, because neonicotinoid concentrations in slugs and cutworms collected from corn seedlings were still greater than concentrations detected in slugs in our soybean study, where neonicotinoids did disrupt biological control (Douglas et al. 2015).

In 2013, slug abundance was higher in clothianidin-treated plots than untreated control plots, despite similar predator activity-density and predation. There are at least two potential explanations for this outcome: i) predation on slugs was in fact disrupted, but I did not detect it because of limited statistical power or isolated effects on components of the predator community, or ii) higher slug abundance was a legacy of the preceding year’s soybean field experiment, where I found strong evidence that thiamethoxam disrupted biological control of slugs (Douglas et al. 2015). The latter hypothesis seems more likely given that the trend for greater slug abundance in clothianidin-treated plots began before corn plants had emerged, suggesting that transient disruptions of the predator community by neonicotinoid seed treatments can have lasting effects on pest communities.

Another unexpected result in the 2012 field experiment was that thiamethoxam seed treatment did not protect corn plants from damage by black cutworm, and in fact cutworms in a range of instars were apparently healthy despite their close association with treated seedlings. The high rate of thiamethoxam that I tested (1.25 mg a.i./seed) is labelled for black cutworm control (Syngenta 2016), and so I expected at least a reduction in cutworm damage in the thiamethoxam-treated plots. A previous field study using artificial infestation of black cutworms found that thiamethoxam was sometimes effective against this species, though it was generally not as effective as clothianidin

102

(Wilde et al. 2007). Even high rates of clothianidin (1.25 mg a.i./seed), however, killed only ~30% of black cutworm third instars in laboratory trials (Kullik et al. 2011). I conclude that seed-applied neonicotinoids are an unreliable pest management tactic for black cutworm infestations. Additionally, generalist predators can play an important role in suppressing black cutworms (Brust et al. 1985). The insecticide residue results show that cutworms, in addition to slugs and earthworms, can provide a source of contaminated prey in no-till environments and can serve as a conduit of neonicotinoid exposure to predators.

Neonicotinoid seed treatments failed to increase corn yield in either year of the study, under conditions of high or low slug abundance. This finding is consistent with results from peer-reviewed field studies in other regions, which have shown yield benefits only under particular circumstances. In Virginia, seed treatment with the high rate of clothianidin (but not the low rate) tended to increase corn yield only in fields with economic populations of white grubs (Jordan et al. 2012). Neonicotinoid seed treatments rarely increased corn yield in Kansas, but target pests were scarce across the sites and years of the experiments (Wilde et al. 2007; Wilde et al. 2004). In conventionally tilled silage corn in New York, the high rates of clothianidin and thiamethoxam tended to improve yield in continuous corn, where they reduced damage by corn rootworm, but not in corn following soybeans (Cox et al. 2007a; Cox et al. 2007b). In contrast, an industry- sponsored meta-analysis that included private field trial data estimated the average yield benefit of neonicotinoid seed treatments in corn at > 17% (Mitchell 2014). The large discrepancy between peer-reviewed results and this meta-analysis is challenging to explain without access to the original studies included in the meta-analysis. In the

103 absence of more information, we argue that the publicly available results reinforce the keystone concept of IPM, that using insecticides economically depends on monitoring and identifying significant pest populations.

Overall, we found that seed-applied neonicotinoids move from corn seedlings to slugs and cutworms in a no-till system, in concentrations that pose a risk to natural enemies consuming these pests during the early season. However, we found no evidence that neonicotinoid seed treatments disrupted biological control by generalist predators, a finding that may have been related to the influence of extreme abiotic conditions at our research site. Neonicotinoid seed treatments did not control black cutworm or significantly increase yield in either a high or low slug year, reinforcing the need to monitor or predict populations of target pests to guide economic use of these insecticides.

Acknowledgements

I thank Scott Smiles, Christy Rose Aulson, Amber Delikat, Emily Newman, Katie

Speicher, Lacy Heberlig, and Andrew Aschwanden for their assistance with experiments,

Jonathan Barber and Roger Simonds for their assistance with insecticide testing. This work was supported by the USDA’s Northeast Sustainable Agriculture Research and

Extension program (GNE11-014 & LNE09-291), the Pennsylvania Department of

Agriculture (44123709), the Maryland Grain Producers Utilization Board, and the

Pennsylvania State University, Department of Entomology.

104

Tables

Table 3-1. Responses of plants, herbivores, and predators in a 2012 field experiment comparing corn plots planted with fungicide-only (Control) or thiamethoxam and fungicide-treated (Neonic) seeds (n = 6 plots/treatment).

Days after Control mean Neonic mean Predicted Response (units) % Δ η2 η2 F P planting ± SE ± SE effect ST p sp 1,4 Potential slug predators 37* 11.5 ± 1.4 9.5 ± 1.7 - -17 0.16 0.07 0.78 0.43 (#/trap/48 h) Predation (prop. killed/24 h) 21 0.26 ± 0.05 0.22 ± 0.04 - -15 0.13 0.04 0.58 0.49 Slugs (#/trap/48 h) 37* 5.2 ± 1.7 2.9 ± 0.6 + -44 0.37 0.14 2.44 0.19 Corn leaf damage 22 42.7 ± 11.0 36.8 ± 8.9 + -14 0.13 0.02 0.59 0.48 (% area removed) Cut plants (%) 22 4.1 ± 1.6 4.6 ± 1.5 - +12 0.01 0.004 0.05 0.83 Corn establishment 31 64.1 ± 1.6 65.8 ± 1.5 - +2.7 0.12 0.05 0.54 0.50 (1,000 plants/ha) Yield (t/ha) 208 8.5 ± 0.5 8.2 ± 0.7 - -3.2 0.06 0.01 0.26 0.63 * These samples had to be delayed due to extensive flooding in five out of twelve plots %ΔST is (meanNeonic - meanControl)/meanControl X 100, the percent change due to seed treatment. 2 ηp is partial eta-squared, the percent of non-block variation explained by seed treatment. 2 ηsp is semi-partial eta-squared, the percent of total variation explained by seed treatment.

105

Table 3-2. Responses of plants, herbivores, and predators in a 2013 field experiment comparing corn plots planted with untreated (Control) or clothianidin-treated (Neonic) seeds (n = 6 plots/treatment) in a year with low slug abundance, at a site with a 1-year legacy of neonicotinoid or control treatments.

Days after Control mean Neonic mean Predicted Response (units) % Δ η2 η2 F P planting ± SE ± SE effect ST p sp 1,4 All predators 34 48.3 ± 2.1 57.3 ± 9.6 - +19 a a 0.85a 0.40a (#/trap/120 h) Predation (prop. killed/24 h) 28 0.18 ± 0.05 0.21 ± 0.03 - +17 0.17 0.06 0.82 0.42 Slugs (#/trap/120 h) 34 0.21 ± 0.16 0.46 ± 0.29 + +119 0.17 0.05 0.82 0.42 Corn leaf damage 24 0.75 ± 0.23 0.46 ± 0.14 - -39 0.21 0.10 1.07 0.36 (% area removed) Cut plants (%) 24 0 ± 0 0 ± 0 - 0 Corn establishment 35 64.2 ± 0.6 66.0 ± 2.0 + +2.8 0.11 0.07 0.51 0.51 (1,000 plants/ha) Yield (t/ha) 196 12.2 ± 0.4 12.5 ± 0.5 + +2.5 0.37 0.03 2.34 0.20 a Because of unequal variance between treatments, these results are based on a Kruskall-Wallis test with 1 and 5.5 degrees of freedom. As a 2 2 result, I did not calculate ηp and ηsp %ΔST is (meanNeonic - meanControl)/meanControl X 100, the percent change due to seed treatment. 2 ηp is partial eta-squared, the percent of non-block variation explained by seed treatment. 2 ηsp is semi-partial eta-squared, the percent of total variation explained by seed treatment.

106

Table 3-3. Neonicotinoid concentration (ppb) detected in samples from a laboratory experiment investigating corn-slug-ground beetle interactions in the presence or absence of seed treatments. Neonicotinoids were not detected in any samples from the untreated control (data not shown).

ng a.i./ Treatment Sample type Replicate Description LOD THX CLO IMI Total % CLO organism Low rate THX Soil 1 ~ 17 g soil 1 433 0.00 ND 433.0 0 - 2 ~ 19 g soil 1 1280 18.8 ND 1298.8 1 - Corn 1 3 shoots 2 1280 203 ND 1483 14 1744 2 3 shoots 2 3210 273 ND 3483 8 3622 Slugs 1 46 slugs 4 86.9 252 ND 339 74 11.3 Beetles 1 13 beetles 4 ND 18.6 ND 18.6 100 1.4 High rate THX Soil 1 ~ 17 g soil 1 3630 31.7 ND 3661.7 1 - 2 ~ 18 g soil 1 4910 26.2 ND 4936.2 1 - Corn 1 3 shoots 2 15800 416 3.2 16219 3 13700 2 3 shoots 2 9120 415 2.4 9537 4 10250 Slugs 1 43 slugs 4 61.2 28.4 ND 61.2 46 2.8 Beetles 1 13 beetles 4 8.5 48.3 ND 56.8 85 4.2 LOD = limit of detection; THX = thiamethoxam; CLO = clothianidin; IMI = imidacloprid; ND = not detected. Tested for but not detected: CLO metabolites; IMI metabolites; Acetamiprid; 6-Chloronicotinic acid, Thiacloprid, Dinotefuran.

107

Table 3-4. Neonicotinoid concentration and abundance (mean ± SE) detected in samples from a field experiment investigating the movement of neonicotinoid seed treatments in a no-till corn ecosystem. Non-detections (ND) were treated as zero for the purpose of calculating means and standard errors.

Concentration (ppb) n plots Year Treatment Sample type DAP LOD* THX CLO CLO MNG Total ng/organism (# positive) 2012 THX Soil 20 1/63 227 ± 66 21 ± 4 ND 248 ± 71 - 3(3) Corn shoots 20 1/63 549 ± 126 113 ± 30 ND 661 ± 154 784 ± 157 4(4) Slugs 20 3/125 209 ± 9 152 ± 75 ND 361 ± 81 29 ± 8 4(4) 187 2/94 ND 2 ± 2 ND 3 ± 3 1 ± 1 3(3) Cutworms 20 7/375 71 ± 21 26 ± 2 ND 97 ± 21 29 ± 5 4(4) Control Soil 20 1/63 1 ± 1 3 ± 3 ND 4 ± 2 - 3(3) Corn shoots 20 1/63 22 ± 12 6 ± 4 ND 28 ± 16 32 ± 18 4(3) Slugs 20 4/188 1 ± 1 ND ND 1 ± 1 0.04 ± 0.04 4(1) 187 2/94 ND ND ND ND ND 3(0) Cutworms 20 3/125 ND ND ND ND ND 2(0) 2013 CLO Soil 23 1/63 ND 77 ± 5 ND 77 ± 5 - 3(3) Corn roots 23 1/63 ND 9900 ± 2626 ND 9900 ± 2626 17402 ± 5058 3(3) Corn shoots 23 1/63 ND 352 ± 99 20 ± 20 372 ± 115 411 ± 152 3(3) Control Soil 23 1/63 ND 2 ± 2 ND 3 ± 1 ND 4(3) Corn roots 23 1/63 1 ± 1 14 ± 5 ND 16 ± 5 19 ± 6 3(3) Corn shoots 23 2/94 ND ND ND ND ND 3(0) THX = thiamethoxam; CLO = clothianidin; CLO MNG = CLO metabolite, N-methyl- N’-nitroguanidine. Imidacloprid (IMI) was detected at low levels (< 2 ppb) in three samples (not shown). Tested for but not detected: CLO metabolites TZMU, TZNG, TMG; IMI metabolites; Acetamiprid; 6- Chloronicotinic acid, Thiacloprid, Dinotefuran. * LOD = limit of detection. The first value is the LOD for THX, CLO, and IMI. The second value is for the CLO metabolite CLO MNG.

108

Figures

Figure 3-1. Outcomes from laboratory experiments investigating the influence of neonicotinoid seed treatments on interactions between corn seedlings (Zea mays), slugs (Deroceras reticulatum), and ground beetles (Chlaenius tricolor). Corn seed treatments: U = untreated, F = fungicide-only, F+L = fungicide + low rate thiamethoxam, F+H = fungicide + high rate thiamethoxam. Error bars show ± one standard error; means with different letters above them are different at α = 0.05 based on post-hoc Tukey comparisons. (a) Corn seedling biomass (g) after 7 days of growth with or without slugs. (b) Slug mass gain (%) after 7 days of feeding on corn seedlings (n = 26-31 microcosms/treatment, no statistical differences among treatments). (c) Beetle symptoms after consuming slugs fed upon three seed treatments.

109

100

80 U F+L 60 F+H

40 % Survival

20

0 0 2 4 6 8 10 12 Hours

Figure 3-2. Slug survival when confined individually with C. tricolor after seven days of feeding on corn with different seed coatings: U = untreated, F+L = fungicide + low rate thiamethoxam, F+H = fungicide + high rate thiamethoxam. Seed treatment did not significantly influence rates of beetle predation on slugs (Likelihood ratio test, D = 2.29, df = 2, P = 0.32).

110

2012 2013 A B 7 ● Control 1 ● Control THX CLO 6 0.8 ● 5 *

0.6 4

3 0.4

Slugs/Pitfall trap Slugs/Pitfall 2 ● ● 0.2 ● 1 * ● ● ● ● ● 0 0 June 12 June 29 Aug 2 Sept 13 May 21 June 10 July 9 Aug 7 Sept 10 C D 7 ● Control ● Control THX 2 CLO 6 ● 1.8

5 1.6 ● Treatment P = 0.01 1.4 ● 4 1.2 1 3 0.8 ● Slugs/Shelter trap 2 0.6 ● 0.4 1 ● ● ● ● 0.2 ● 0 ● ● 0 May June July Aug Sept Oct May June July Aug Sept Oct

Figure 3-3. Slug activity-density (mean ±SE) in field experiments comparing plots that were planted with neonicotinoid-treated corn seed to controls with only fungicides (2012) or no seed treatment (2013; n = 6 plots/treatment). Slug populations were measured using pitfall traps (A,B) or shelter traps (C,D). Asterisks mark significant comparisons at α = 0.05, based on post-hoc comparisons where there was a significant Date ×Treatment interaction.

111 2012 2013 A B 16 ● Control 24 ● Control THX CLO 14 20 ● 12 ● 16 10 ●

8 12

6 ● ● 8 ● ● ● 4

● Potential slug predators/trap Potential 4 2

0 0 June 12 June 29 Aug 2 Sept 13 May 21 June 10 July 9 Aug 7 Sept 10 C D 28 ● Control 70 ● Control THX CLO 24 ● 60

20 50 ●

16 ● 40

● ● ● 12 30 ● ●

All predators/trap 8 ● 20

4 10

0 0 June 12 June 29 Aug 2 Sept 13 May 21 June 10 July 9 Aug 7 Sept 10

Figure 3-4. Activity-density (mean ±SE) of potential slug predators (A,B) and all arthropod predators (C,D) in field experiments comparing plots that were planted with neonicotinoid-treated corn seed to controls with only fungicides (A,C; 2012) or no seed treatment (B,D; 2013; n = 6 plots/treatment). There were no significant differences in predator activity-density between treatments.

112 A May 25 June 26 July 30

1.0 ●

● Control 0.8 THX

0.6 ●

● 0.4 ● Proportion killed 0.2 ●

● 0.0 AM PM AM PM AM PM

B June 5 July 2 Aug 2

1.0

● ● ● Control ● ● 0.8 CLO

0.6

0.4

Proportion killed ● 0.2

●

0.0 AM PM AM PM AM PM

Figure 3-5. Predation (mean ±SE)on sentinel caterpillars (Galleria mellonella) over the season in field experiments comparing plots that were planted with neonicotinoid-treated corn seed to control plots that were planted with fungicide-only (A; 2012) or untreated (B; 2013) seed. THX = thiamethoxam, CLO = clothianidin. Predation did not differ significantly by treatment on any sample date.

113 105

R 2= 0.76

● Lab 104 ● ● ● Field

●

103 ● ● ● ● ●

● ● ● 102 ● ● Total neonicotinoids (ppb) Total ● ● ●

● 101

Corn Slugs+Cutworms Beetles

Figure 3-6. Concentrations of neonicotinoids in corn shoots, pest herbivores (slugs Deroceras reticulatum and cutworms Agrotis ipsilon), and ground beetles Chlaenius tricolor from laboratory and field experiments. Samples from the field were collected in 2012 when corn seedlings were in the V2 stage.

114

Chapter 4

Meta-analysis reveals that neonicotinoid seed treatments and pyrethroids have similar negative effects on abundance of arthropod natural enemies

Abstract

Seed-applied neonicotinoids are widely used in agriculture, yet their effects on non-target species remain incompletely understood. One important group of non-target species is arthropod natural enemies (predators and parasitoids), which contribute considerably to suppression of crop pests. We hypothesized that seed-applied neonicotinoids reduce natural-enemy abundance, but not as strongly as alternative insecticide options such as soil- and foliar-applied pyrethroids. Furthermore we hypothesized that seed-applied neonicotinoids affect natural enemies through a combination of direct toxicity and prey scarcity.

To test our hypotheses, we compiled datasets comprising observations from randomized field studies in North American and Europe that compared natural-enemy abundance in plots that were planted with seed-applied neonicotinoids to control plots that were either i) managed without insecticides (20 studies, 56 site-years, 607 observations) or ii) managed with pyrethroid insecticides (8 studies, 15 site-years, 384 observations). Using the effect size Hedge’s d as the response variable, we used meta- regression to estimate the overall effect of seed-applied neonicotinoids on natural-enemy abundance and to test the influence of potential moderating factors such as crop species and natural-enemy functional or taxonomic group.

115

Neonicotinoid seed treatments reduced the abundance of arthropod natural enemies compared to untreated controls (d = -0.30 ± 0.10 [95% confidence interval]), and as predicted under direct toxicity this effect was stronger for insect than for non-insect taxa (QM = 8.70, df = 1, P = 0.003). Moreover, seed-applied neonicotinoids affected the abundance of arthropod natural enemies similarly to soil- or foliar-applied pyrethroids (d

= 0.16 ± 0.42 or -0.02 ± 0.12; with or without one outlying study). Effects of seed- applied neonicotinoids were surprisingly consistent across both datasets (I2 = 2.7% for no-insecticide controls; I2 = 0% for pyrethroid controls), suggesting little influence on the results of crop species, neonicotinoid active ingredients, or methodological choices.

Our meta-analysis of nearly 1,000 observations from North American and

European field studies revealed that seed-applied neonicotinoids reduce the abundance of arthropod natural enemies, and that they are generally no safer for these important non- target species than broadcast applications of pyrethroid insecticides. These findings suggest that substituting pyrethroids for neonicotinoids, or vice versa, is unlikely to cause a significant change in natural enemy abundance. Furthermore, consistent with laboratory toxicity assays, we found that seed-applied neonicotinoids are more harmful to insects than spiders and mites, which can contribute substantially to biological control in many agricultural systems. Finally, our ability to interpret the negative effect of neonicotinoids on natural enemies is constrained by a lack of knowledge relating natural-enemy abundance to biological control function; this is an important area for future study.

116

Introduction

Arthropod natural enemies (predators and parasitoids) contribute considerable value to agriculture by suppressing pests that attack crop plants. For example, biological control of the soybean aphid (Aphis glycines) is estimated to be worth at least $84 million per year in just five U.S. states (Zhang & Swinton 2012). Given their importance to pest management, it is essential to understand how agricultural practices influence natural- enemy communities and their ability to suppress crop pests. Insecticide use is one common agricultural practice that can influence natural-enemy populations and biological control. Insecticides are of course used to manage pests, however in some cases they also disrupt biological control, leading to unintended outbreaks of target or non-target pests (Geiger et al. 2010; Settle et al. 1996; Stern et al. 1959). Elucidating how insecticides and natural enemies interact can be useful in identifying insecticide products that have greater selectivity for pests versus natural enemies, applying them in ways that are less harmful to natural enemies, and in some cases even using insecticides to enhance the efficacy of biological control (Croft & Brown 1975; Hull & Beers 1985).

One use of insecticides that is increasingly widespread is planting seeds coated with neonicotinoid insecticides, especially in large-acreage field crops where these seed coatings are ubiquitous in some crops and regions (Simon-Delso et al. 2015). Seed coatings account for at least 80% of neonicotinoids applied in the U.S. (Douglas &

Tooker 2015) and accounted for roughly 87% in Britain prior to recent restrictions

(Simon-Delso et al. 2015). Through their systemic activity, seed-applied neonicotinoids target some early-season soil and foliar pests, and their relatively low cost and ease of handling makes them an attractive option as ‘insurance’ against sporadic pest populations

117

(Douglas & Tooker 2015; Jeschke et al. 2011). The popularity of neonicotinoids is also related to their toxicological profile; their binding behavior at nicotinic acetylcholine receptors confers potent toxicity against a broad array of pest insect species, and simultaneously low acute toxicity to mammals (Tomizawa & Casida 2005).

Despite the broad toxicity of neonicotinoids to insects, some researchers reasoned that by applying them in relatively small doses ‘targeted’ to crop seed, neonicotinoid seed treatments should have high ecological selectivity for crop pests and low potential to harm beneficial insects (Jeschke et al. 2011). The environmental safety of these products has since been called into question by evidence that non-target species, especially bees, can be exposed to seed-applied neonicotinoids via contaminated soil, planting dust, floral resources, and guttation droplets (Godfray et al. 2014; Godfray et al. 2015). This pollinator controversy has spurred new regulations in several jurisdictions (European

Commision 2013; Government of Ontario 2015), in turn raising the question of whether other pest management tactics that might replace neonicotinoids, including conventionally applied insecticides, are more or less harmful to non-target species. We argue that fully understanding the trade-offs associated with seed-applied neonicotinoids requires attention not only to pollinators but also to natural enemies that suppress crop pests.

Neonicotinoid seed treatments can potentially reduce populations of arthropod natural enemies through at least two mechanisms: direct toxicity and prey scarcity.

Neonicotinoids are toxic to many natural-enemy species (Cloyd & Bethke 2011;

Hopwood et al. 2013), but the question remains as to whether under field conditions natural enemies encounter meaningful doses of toxins from seed treatments. What

118 constitutes a ‘meaningful dose’ is likely to vary by taxon. In the laboratory, insects are orders of magnitude more susceptible to neonicotinoids than arachnids (Table 4-1), and at least some evidence suggests that this difference is based on the molecular structure of arachnid acetylcholine receptors (Meng et al. 2015). If natural-enemy populations were reduced by seed-applied neonicotinoids through direct toxicity, we would expect insects to be more strongly affected than arachnids. Alternatively, neonicotinoids could exert indirect negative effects, for instance by reducing the abundance of prey, leading to less aggregation, persistence, or reproduction of natural enemies in crop fields (Croft &

Brown 1975). Under this prey-scarcity scenario, we would expect insects and arachnids to be similarly affected, but functional groups to differ relative to their degree of dependence on pest prey (parasitoid > predator > omnivore). Direct and indirect effects of insecticides on natural enemies are not mutually exclusive, and both can interfere with biological control (Johnson & Tabashnik 1999). Clarifying the mechanisms at play can nonetheless guide researchers and pest managers toward more successful integration of chemical and biological control.

Field experiments on the influence of seed-applied neonicotinoids on natural enemies have reached mixed conclusions. Some have found no statistically significant effects (Al-Deeb et al. 2003), some a mix of null and negative effects (Albajes et al.

2003), and others more consistent negative effects (Hallett et al. 2014). Some of this variability may be related to the small sample sizes of most field studies, which limits their statistical power. Conversely, variability in study results may reflect real differences in the effects of seed-applied neonicotinoids across crop species, natural-enemy taxa, active ingredients, and other factors. One powerful tool to make sense of such apparently

119 mixed results and to untangle the various factors influencing study outcomes is meta- analysis (Cumming 2012). This approach has been fruitfully applied in similar situations, for instance, to estimate the influence of Bt transgenic crops on non-target organisms

(Marvier et al. 2007; Naranjo 2009; Wolfenbarger et al. 2008). One of the salient benefits of meta-analysis in controversial areas is that it provides a rigorous, quantitative, and replicable method of synthesizing evidence for researchers, policy-makers, and the public

(Marvier et al. 2007).

Here, we report results from a meta-analysis of studies investigating under field conditions the influence of neonicotinoid seed treatments on arthropod natural enemies.

We used a meta-regression approach to test the hypotheses that seed-applied neonicotinoids: i) negatively affect natural-enemy abundance relative to untreated controls; ii) are less harmful to natural enemies than conventional foliar- or soil-applied insecticide treatments; and iii) affect natural enemies through a combination of direct toxicity and prey scarcity. Our results taken together should allow researchers and pest managers to better predict the compatibility of seed-applied neonicotinoids and natural enemies, and to more effectively weigh seed-applied neonicotinoids against alternative pest-management options.

Materials & Methods

Meta-analysis is a method for synthesizing observations from independent yet similar studies to characterize the size and variability of an effect – in this case the influence of seed-applied neonicotinoids on natural enemies of crop pests. Our meta- analysis was guided to some extent by previous meta-analyses that characterized the

120 influence of Bt crops on populations of non-target arthropods (Marvier et al. 2007;

Naranjo 2009; Wolfenbarger et al. 2008). We departed from these previous studies in our statistical approach. In particular, as described below, we capitalized on advances in statistical programs over the past several years to better account for the hierarchical nature of the dataset. All of our analyses were conducted within the R statistical program

(R Core Team 2015).

Searching the literature & building the dataset

Using four databases (ISI Web of Science, Agricola, CAB abstracts, and ProQuest

Dissertations & Theses Database), we searched for studies on the influence of neonicotinoids on arthropod natural enemies. We used the following search phrase, adjusting the syntax as necessary for different search engines: “(neonic* OR imidacloprid

OR clothianidin OR thiamethoxam) AND (preda* OR enem* OR parasit*) AND seed”.

We combed the resulting studies and published reviews (Chagnon et al. 2015; Hopwood et al. 2013; Lundgren 2009; Pisa et al. 2015) for additional references, and found one additional unpublished thesis because a colleague mentioned it at a scientific meeting.

Our final literature search for this analysis was conducted on August 7, 2015.

We used the following criteria to include a study in the dataset: (i) it compared field plots that were planted with neonicotinoid-treated seed with control plots that were planted with neonicotinoid-free seed of the same crop variety. There were two types of control plots: those that were not treated with any insecticides (testing whether neonicotinoids have any effect on natural enemies), and those that were treated with an alternative insecticide product (testing whether neonicotinoids are more or less harmful to

121 natural enemies than alternatives). Studies also had to (ii) measure abundance or activity- density of one or more taxonomic groups of arthropod natural enemy, (iii) be replicated,

(iv) report the data necessary to calculate effect size (means, sample sizes, and standard errors or standard deviations), and (v) be available in English. Where studies met the first three criteria but did not report some necessary data, we contacted authors to obtain that data, although not all responded. Where necessary we extracted data from published figures using the software GraphClick 3.0.3 (Arizona Software).

To build the database, we recorded for each study the means and variability for natural-enemy abundance in each treatment group, along with a wide variety of supporting information such as author and affiliation, study location and year, crop species, active ingredient of the seed treatment, size of each plot, number of replicates, and other methodological details. For each taxon, we recorded sampling method, life stage, habitat (soil/epigeal or foliar/aerial), functional group, and taxonomic information to the lowest level provided. Many natural enemies consume plant products (pollen, nectar, seeds, etc.) in addition to live prey; we assigned taxa to functional groups using an existing classification described below (Wolfenbarger et al. 2008), and filling in gaps where necessary based on the scientific literature.

Defining the scope of the study

We defined ‘natural enemies’ to include the following functional groups

(Wolfenbarger et al. 2008): mixed, omnivore, predator, and parasitoid. The ‘omnivore’ group comprises taxa that are believed to rely equally on prey and non-prey foods (e.g.,

Formicidae, Gryllidae). The ‘mixed’ group refers to taxonomic units that contain species

122 in multiple functional groups (e.g., Carabidae). While our initial intent was to include studies from all geographic regions, we restricted the current analysis to North America and Europe because most of the studies from other regions (8 of 11 studies, all from

South Asia) lacked sufficient details for us to interpret reported measures of variation.

For the part of our analysis comparing neonicotinoids to alternative insecticides, we restricted the analysis to pyrethroids, the only insecticide class that was compared to seed-applied neonicotinoids in at least three independent studies (Table C-1).

Calculating effect size: Hedge’s d

The response variable in our analysis is Hedge’s d, the mean abundance in the control group minus the mean abundance in the treatment group, divided by the pooled standard deviation, and corrected for small sample size (Koricheva et al. 2013). In this case d measures the difference in natural-enemy abundance between control plots and plots planted with neonicotinoid-treated seed, with negative values reflecting lower abundance of natural enemies in neonicotinoid-treated plots compared to controls. We used the ‘escalc’ function (‘metafor’ package) to calculate d and its associated variance for each observation in the dataset.

Addressing non-independence

Typical studies in our dataset contributed numerous observations resulting from multiple taxa, sampling methods, life stages, site-years, and other factors. As a result we could not assume that all of the observations were statistically independent. As in past

123 meta-analyses (Marvier et al. 2007), we addressed this problem in part by eliminating redundant observations, as follows. When results were reported at varying levels of taxonomic resolution, we used only the results at the finest taxonomic level. When multiple life stages were sampled for a given taxon, we used only the observations from the least mobile, but still feeding life stage (Wolfenbarger et al. 2008). If taxa were sampled repeatedly over a single season, we used seasonal summary data when available; otherwise we requested seasonal data from the authors. If this failed, we used peak values. When a given taxon was sampled in multiple ways, we included results from the sampling method with the highest precision (lowest relative standard deviation). We made an exception to this rule for studies that sampled soil and foliar habitats for

Araneae, “predatory mites” and Carabidae, because for these broad taxonomic groups, these habitats are likely to contain mostly non-overlapping taxa. Even after taking these steps to reduce redundancy, our dataset still contained numerous observations per study as a result of multiple taxa, site-years, and crossed factors such as crop varieties and insecticide active ingredients. We accounted for the remaining non-independence through the structure of the meta-regression models as described in the next section.

Fitting meta-regression models

To estimate the influence of neonicotinoid seed treatments on natural enemies and to test the influence of agroecological or methodological moderating variables on the size of this effect, we employed a mixed effects meta-regression approach using the package

‘metafor’ in R (R Core Team 2015; Viechtbauer 2010). Meta-regression is analogous to multiple regression, but differs in that observations are weighted relative to their

124 precision (typically 1/variance). The strength of the meta-regression approach is that it allows us to investigate the influence of multiple moderators at once, while also using random effects to control for the hierarchical nature of the dataset (observations nested in site-years nested in studies).

We split our larger dataset into two parts: one for observations that compared seed-applied neonicotinoids to an insecticide-free control and a second for observations that compared seed-applied neonicotinoids to pyrethroids. For each of these datasets we fit three models, all estimated using restricted maximum likelihood in ‘metafor’. First we fit a ‘null’ model that did not include random effects of site-year or study, nor fixed effects of moderators. This model mainly serves as comparison to results of previous meta-analyses on non-target effects of agricultural technologies, many of which did not account for nesting of observations within studies. We next fit a ‘site-year/study’ model that included only random effects of site-year and study, but no fixed effects (i.e., no moderators). From these two initial sets of models we generated 95% confidence intervals for the overall influence of neonicotinoid seed treatments on natural enemies, and also characterized variability in the effect sizes using ‘heterogeneity’ as measured by the Q statistic. Q is the weighted sum of squared differences of effect sizes from the mean, and can be used to test whether variability among effect sizes is greater than would be expected by sampling error alone (Viechtbauer 2007b). In addition to the Q test, we calculated I2 (100% X (Q-df)/Q), which estimates the percentage of variability in effect sizes that is due to true heterogeneity rather than sampling error (Higgins et al. 2003).

Finally, to assess whether it was necessary to include the study and site-year effects, we

125 examined the variance components associated with these effects in the ‘site-year/study’ model using profile plots.

The third model we fit for each dataset was a ‘moderator’ model designed to test whether agroecological or methodological variables influenced the effect of seed-applied neonicotinoids on natural enemies. Along with random effects of site-year and study, these models included fixed effects of potential moderating variables that we identified a priori: broad taxonomic group (insect or non-insect), functional group (predator, parasitoid, omnivore, or mixed), habitat (soil-associated or foliage-associated), neonicotinoid active ingredient group (imidacloprid or clothianidin/thiamethoxam), crop species (maize, soybean, or other), publication status (peer-reviewed or dissertation/thesis/other), pyrethroid application method (where applicable; soil/granular or foliar/spray), plot size, and proportion of samples collected during the first 40 days of crop growth, when neonicotinoids typically have activity against target pests (Magalhaes et al. 2009; Seagraves & Lundgren 2012). For the pyrethroid analysis, to reduce the risk of drawing spurious conclusions we left out crop species, publication status, and functional group, because some levels of these moderators were not supported by at least three independent studies ( Table 4-2). We transformed continuous moderators (plot size, proportion early samples) where necessary and centered them on a mean of zero to facilitate interpretation. Categorical variables were converted to effects coding by employing the ‘contr.sum’ option in ‘contrasts’. This made the intercept of the fitted model reflect the mean value across the means of all moderator variables, and the slopes reflect the difference associated with the level of each moderator from the overall mean.

We first tested for the significance of all the moderators combined (using an omnibus Q

126 test for moderators); when that was significant, we went on to test the significance of individual moderators.

We tested the moderators that we expected to have the greatest likelihood of influencing effect size, while limiting the total number of moderators to preserve the power of the analysis. As described in the introduction, effects of taxonomic group and functional group have implications for whether natural enemies are affected by neonicotinoids through direct toxicity (insects > arachnids) and/or prey scarcity

(parasitoid > predator > omnivore). Habitat, active ingredient, and crop species may mediate the influence of neonicotinoids on natural enemies because these factors likely correspond to differences in exposure, toxicity, and prey communities. We included publication status because a relationship between publication status and effect size could indicate publication bias (Koricheva et al. 2013). Finally, we included plot size, early- season sampling, and pyrethroid application method in the model to control for methodological variables that we suspected might influence research outcomes.

Assessing statistical assumptions, potential biases, and robustness of results

As in multiple regression, in meta-regression correlations among moderator variables can render estimates and tests of model parameters unreliable (Kutner et al.

2005; Viechtbauer 2007a). Before fitting the ‘moderator’ models, we first calculated and examined pairwise correlations among all our moderators. We also examined generalized variance inflation factors (GVIF) by fitting linear models using the ‘lme’ function (nlme package) with our moderators in the models as fixed effects and site-year nested in study as random effects. We used the ‘vif’ function (car package) to calculate GVIF values.

127

To further assess the fit of our models, we examined the standardized residuals versus fits and inspected the normality of the residuals using a QQ plot. To screen for influential observations, we plotted leverage values on their own and relative to residuals.

When we found potential outliers, we re-fit our models without them to assess their influence on our conclusions.

We tested for publication bias in part through the ‘publication status’ moderator in the meta-regressions, as discussed above. Additionally we examined a weighted histogram of the effect sizes for evidence of ‘missing’ observations near zero, and used the ‘trimfill’ and ‘funnel’ functions (‘metafor’ package) on our null models to generate funnel plots and test for ‘missing’ observations in the distributions. Finally to test the robustness of the overall effect size estimates, when they differed from zero, we calculated Rosenberg’s ‘fail safe N’, that is, the number of null observations necessary to make the observed effect size non-significant. We note that most of these methods for testing publication bias do not take into account the hierarchical nature of our dataset; there do not appear to be tools available that explicitly account for this data structure.

To test the sensitivity of our results to inclusion of particular studies, we conducted a ‘leave one out’ analysis in which we removed the observations associated with each study from each dataset one by one, and then re-fit all three models. We assessed the consistency of the confidence intervals for the overall effects, as well as the fitted slopes and hypothesis tests for the ‘moderator’ models. Where eliminating a study changed the interpretation of our analysis, we noted this in the results.

128

Predator-prey ratios

Because insecticides can affect pest and predator populations differently, predator-prey ratios may be more reflective of biological control function than predator abundance alone (e.g. Croft & Nelson 1972; Naranjo 2005; Ooi 1982). Unfortunately most studies did not report predator-prey ratios or sufficient data to calculate them, but we were able to perform a preliminary summary based on soybean studies, which more often reported cumulative abundance of both pests and relevant predators. For the soybean studies that reported both the cumulative abundance of a pest taxon and the cumulative abundance of the predator guild for that pest, we calculated predator-prey ratios for neonicotinoid treatments and controls. Without access to the original data it was impossible to estimate a variance for the predator-prey ratios, so we did not perform a formal meta-analysis; instead we discuss them qualitatively to lend preliminary insight into the relative impact of neonicotinoids on predator and pest populations. To facilitate this summary, we calculated the percent change in the predator-prey ratio in each neonicotinoid treatment relative to each control. Negative values indicate that seed- applied neonicotinoids reduced the predator-prey ratio relative to the control, while positive values indicate the opposite.

Results

Results of the literature search & characteristics of the meta-analysis dataset

In total we screened 518 titles and abstracts, yielding 68 candidate reports and

1,965 potential observations. After filtering for relevant functional groups and geographic

129 regions (North America and Europe), and reducing redundant observations, our final dataset for the no-insecticide controls comprised 607 observations collected over 56 site- years and 20 independent studies. For the pyrethroid controls, our final dataset comprised

384 observations collected over 15 site-years and 8 independent studies. Corn (Zea mays) and soybean (Glycine max) were the dominant crops, and insects were better represented in the dataset than non-insect arthropods (arachnids and chilopods; Table 4-2).

Unsurprisingly, given the focus of our study, predators were the dominant functional group (67-68% of observations), although parasitoids, omnivores, and mixed functional groups were also represented. Observations were spread fairly equally among soil and foliar habitats, different active ingredients, and pyrethroid application methods, and most observations were derived from peer-reviewed studies (76-93%; Table 4-2). Plot size ranged widely from 1 to 110,000 m2, and proportion of early season sampling ranged from zero to 100 percent (Table C-1). Most studies had a sample size of three to six replicate plots per treatment (Table C-1).

Seed-applied neonicotinoids negatively affected natural enemies compared to no- insecticide controls

Consistent with our hypothesis, seed-applied neonicotinoids reduced the abundance of arthropod natural enemies relative to untreated plots (Figure 4-1). The mean effect size (d) was -0.30 or -0.26, for the ‘site-year/study’ and ‘null’ models respectively. For context, an effect size of -0.30 would correspond to an approximate reduction of 16% in natural-enemy abundance (-0.30 × median relative standard deviation [RSD] of 0.53). The estimates of the variance components from the fitted

130 model suggested that site-year explained most of the shared variation among observations within studies (study σ2 = 0.0035, site-year σ2 = 0.08). The negative effect of seed- applied neonicotinoids on natural enemies appeared to be homogenous (Q = 622.5, df =

606, P = 0.31), with an I2 indicating that all but 2.7% of variation in effect sizes could be explained by random sampling error.

Despite the low heterogeneity identified in the initial analysis, we proceeded with the ‘moderator’ analysis to test the influence of various factors on the effect of seed- applied neonicotinoids on natural enemies. We did this because we had planned the moderator analysis a priori, and because the Q test, despite being the most powerful test available, still has low power to detect heterogeneity in datasets like ours where the within-study sample sizes are small (Viechtbauer 2007b). Indeed, when we fit the meta- regression model with our eight moderator variables, the omnibus test suggested that the moderators taken together did explain significant variation in effect size (QM = 21.5, df =

11, P = 0.029). Broad taxonomic group apparently drove this result, as it was the only moderator that was significant when tested individually (Table 4-3). As predicted under direct toxicity, the negative effect of neonicotinoid seed treatments on natural enemies was stronger for insects than for non-insect taxa (mostly spiders and mites; Figure 4-1).

When estimated separately, the negative effect of neonicotinoids on insects remained significant, while the effect on non-insect taxa did not differ significantly from zero

(Figure 4-1). Although functional group did not significantly moderate the influence of seed-applied neonicotinoids on natural enemies (P = 0.13), the fitted slopes for this moderator were fairly large and followed a trend consistent with indirect effects via prey scarcity (parasitoid > predator > omnivore, Table 4-3). Surprisingly, effect size was not

131 influenced to a significant extent by crop species, neonicotinoid active ingredient, habitat, or any of the methodological variables (Table 4-3).

Seed-applied neonicotinoids were not significantly less harmful to natural enemies than pyrethroid insecticides

Contrary to our prediction, seed-applied neonicotinoids were not significantly safer for natural enemies than sprayed or granular pyrethroids (i.e. the effect size for pyrethroids versus neonicotinoid seed treatments did not differ significantly from zero;

Figure 4-2). The sensitivity analysis revealed that one study (Ohnesorg et al. 2009) had a large influence on effect sizes and confidence intervals, so we present results both with and without this study (Figure 4-2).

With all studies in the dataset, the mean effect size (d) was 0.16 or 0.07, for models that did or did not include random effects of site-year nested in study (Figure

4-2). Including random effects for site-year and study made the confidence intervals very wide, because of the influence of Ohnesorg et al. (2009). Similarly, the estimates of the variance components from the fitted model suggested that study explained most of the shared variability among observations (study σ2 = 0.58, site-year σ2 = 0.04). The effects of seed-applied neonicotinoids on natural enemies compared to pyrethroid controls appeared to be homogenous (Q = 369.6, df = 383, P = 0.68), with an I2 indicating that

100% of the variation in effect sizes could be explained by random sampling error.

When Ohnesorg et al. (2009) was excluded from the dataset, the confidence intervals were smaller but the conclusion remained the same: neonciotinoid seed treatments and pyrethroids had similar influences on natural-enemy populations (Figure

132

4-2). The overall mean effect size was very close to zero, -0.02 or 0.03, for models that did or did not include random effects of site-year nested in study. The variance components for study and site-year were small in this model (study σ2 = 0.01, site-year σ2

= 0.0008). Consistent with the first analysis, there was no evidence of heterogeneity in these effects (Q = 316.9, df = 375, P = 0.99, I2 = 0).

Again, we proceeded with the ‘moderator’ model to test whether various factors influenced the magnitude of effect size. In this case, the omnibus test suggested that the moderators did not explain significant variation in effect size (all studies: QM = 4.50, df =

6, P = 0.61; excluding Ohnesorg et al. 2009: QM = 3.40, df = 6, P = 0.76). This result is not surprising given the zero estimate of heterogeneity in this dataset, and suggests that the effect size of seed-applied neonicotinoids compared to pyrethroids is fairly consistent across the dataset, except for the observations associated with one outlying study.

Statistical assumptions, potential biases, and robustness of results

We found little evidence of collinearity among our moderators. Pairwise correlations among moderators were centered near zero and mostly small (No-insecticide control: 84% < 0.2, median: -0.03, mean: -0.003, range: -0.56 to 0.52; Pyrethroid control:

80% < 0.2, median: -0.05, mean: -0.05, range: -0.51 to 0.34). Generalized variance inflation factors for moderators in both analyses were all less than two, again suggesting that collinearity among our predictors was minimal (Kutner et al. 2005).

For each of the datasets, diagnostic plots identified a handful of outliers with large standardized residuals (absolute value > 3). However, these outliers had little leverage,

133 and removing them did not appreciably change parameter estimates or the outcome of significance tests (data not shown).

We found no evidence of publication bias in our datasets. The distributions of effect sizes were bell-shaped with no evidence of an absence of observations near zero

(Appendix C; Figure C-1, C-2). Furthermore, publication status was not a significant moderator of effect size in the no-insecticide comparison (Table 1-1). Rosenberg’s fail- safe N suggested that over 10,000 null observations would be necessary to render non- significant the difference between seed-applied neonicotinoids and insecticide-free controls. The ‘trimfill’ analysis estimated zero missing observations for each of the datasets, lending further support to the absence of publication bias.

The ‘leave one out’ analyses showed that our results for the no-insecticide comparison were fairly robust to the exclusion of particular studies. The estimated intercepts, slopes, and confidence intervals were quite similar across the analyses, and the overall effect of neonicotinoids on natural enemies was consistently negative (data not shown). In two out of twenty cases (Ohnesorg et al. 2009; Sotelo-Cardona 2010), leaving a study’s observations out of the analysis changed the omnibus test of moderators from significant to non-significant. This is perhaps not surprising given that the heterogeneity in this dataset was generally low. Overall, the sensitivity analysis suggested that no particular study was overly influential in the finding of a negative effect of neonicotinoid seed treatments on natural enemies compared to insecticide-free controls, but that the difference in this effect between insects and other taxa should be tested in future studies.

As discussed previously, for the effect of neonicotinoid seed coatings versus pyrethroids, the ‘leave one out’ analysis revealed that one study (Ohnesorg et al. 2009)

134 had a fairly large influence on the width of confidence intervals. Nonetheless, we reemphasize that regardless of the inclusion of this study or the model used to estimate effect sizes, the confidence interval for this comparison always enclosed zero, suggesting little to no difference in the influence of seed-applied neonicotinoids and foliar pyrethroids on natural-enemy abundance. It is notable that there were fewer studies available that investigated pyrethroid insecticides versus no-insecticide controls (8 studies versus 20 studies), and very few studies investigating other insecticide classes

(Table S1), a discrepancy that could be addressed in future research.

Effect of neonicotinoid seed treatments on predator-prey ratios in soybeans

Seven soybean studies reported sufficient information to calculate predator-prey ratios. The focal prey in five of the studies was the soybean aphid (Aphis glycines), while a sixth study focused on herbivorous thrips and a seventh focused on pest slugs (mainly

Deroceras spp.). Aphids and thrips are listed on the neonicotinoid label for soybeans and so could be considered ‘target pests’, though in practice soybean aphids are often not controlled sufficiently with seed-applied neonicotinoids (Myers & Hill 2014). Slugs are non-target pests because they are generally not susceptible to neonicotinoids (Douglas et al. 2015; Simms et al. 2006).

For studies focusing on soybean aphids, plots planted with neonicotinoid-coated seeds had numerically lower predator-prey ratios than plots treated with foliar insecticides (neonicotinoids, pyrethroids, or pymetrozine) in 13 out of 16 comparisons

(Figure C-3). In contrast, plots planted with neonicotinoid-coated seeds had numerically higher predator-prey ratios than untreated controls in 11 out of 16 comparisons. For the

135 studies focusing on non-aphid prey (thrips or slugs), all three predator-prey ratios were numerically lower in neonicotinoid-treated plots than untreated controls (Figure C-3).

These results suggest that for aphids, seed-applied neonicotinoids can sometimes enhance predator-prey ratios, and the foliar insecticides tested in available studies appear to enhance them even more strongly. On the other hand, the limited data available for non- aphid pests suggest that neonicotinoid seed treatments reduce predator-prey ratios. We caution that these conclusions are based on relatively few studies and pest/predator combinations, and lack an estimate of variability. Moreover, most of them are based on the ratio of a focal pest to the summed abundance of a relevant guild of generalist predators, and so do not take into account differences between natural enemy taxa in predation rates.

Discussion

We performed a meta-analysis of field studies to determine the influence of seed- applied neonicotinoids on arthropod natural enemies of crop pests in North America and

Europe. After gathering and synthesizing results from almost 1,000 observations gleaned from 20 studies, we found that seed-applied neonicotinoids i) reduced natural-enemy abundance and ii) were not generally safer for natural enemies than foliar or soil-applied pyrethroids. Furthermore, the influence of seed-applied neonicotinoids on natural enemies differed by broad taxonomic group: insects were more strongly affected than non-insect taxa such as spiders and mites. This last result suggests that reductions in natural-enemy populations associated with seed-applied neonicotinoids are at least partly a result of direct toxicity, rather than prey scarcity alone.

136

Seed-applied neonicotinoids negatively affected natural enemies relative to no- insecticide controls, with an effect size (d = -0.30) corresponding to roughly 16% reduced abundance. This result was robust to different modeling choices and surprisingly consistent across crop species and neonicotinoid active ingredients. For comparison, the mean effect of organic farming (versus conventional farming) on predatory insect abundance was estimated to be d = 0.49 (Bengtsson et al. 2005). Both effect sizes suggest that insecticides can undermine natural-enemy populations, but the consequences of these reductions for ecosystem services are hard to predict given a lack of research relating predator abundance to biological control function and its economic value (Naranjo et al.

2015). The one study in our dataset that explicitly related predator abundance to crop yield was our previous study in a no-till soybean system (Douglas et al. 2015). In that study, a 31% reduction in early season abundance of slug predators in neonicotinoid- treated plots corresponded with a 67% increase in slug abundance and an eventual 5% reduction in soybean yield. Incidentally, the season-long reduction in slug-predator abundance was 16%, very similar to the mean effect identified in this meta-analysis, suggesting that a reduction of this magnitude can have economic consequences. Future efforts to relate natural-enemy abundance to crop yield could also make use of the concept of ‘natural-enemy units’, which help to consolidate diverse natural enemies into a single measure of pest-suppression potential (Bahlai et al. 2010; Hallett et al. 2014).

Ultimately, it would be valuable to build the knowledge base necessary to fit models relating natural-enemy abundance to pest abundance and ultimately crop productivity, analogous to those recently developed for pollination services (Koh et al. 2016).

137

Our finding that seed-applied neonicotinoids can in some cases increase predator- prey ratios further highlights that natural-enemy abundance is not equivalent to biological control function. We stress that a formal analysis of predator-prey ratios was not possible, but generally seed-applied neonicotinoids tended to have a smaller effect on natural enemies than on pest aphids, and a relatively larger effect on natural enemies than on other pest taxa (slugs and thrips). This pattern is further supported by case studies in the literature. We are not aware of any systems in which seed-applied neonicotinoids have been associated with resurgence of target pests such as aphids; however, there are several examples where these seed treatments have been associated with increased abundance and sometimes economic outbreaks of non-target pests, including spider mites (Smith et al. 2013), slugs (Douglas et al. 2015), and late-season stem-boring caterpillars (Pons &

Albajes 2002).

Our finding that insects were more strongly affected by seed-applied neonicotinoids than were non-insect groups (mainly spiders and mites) suggests that direct toxicity is at least partly responsible for the overall negative effect we observed, and raises the question of how insect natural enemies are being exposed to these seed- applied toxins. Neonicotinoids can poison natural enemies through ingestion as well as contact with sprays or residues (Lucas et al. 2004; Torres & Ruberson 2004; Wang et al.

2008). Possible exposure pathways include contact with soil or planting dust (Goulson

2013), ingestion of contaminated prey (Douglas et al. 2015; Szczepaniec et al. 2011), and for some natural enemies, ingestion of pollen, nectar, or other plant tissues (Lundgren

2009; Moser & Obrycki 2009). The relative importance of the various exposure pathways in the field is unclear, but we did see a non-significant trend for soil-dwelling taxa to be

138 more strongly affected than foliar-dwelling taxa. Typically ~ 90% of seed-applied neonicotinoids remain in soil (Goulson 2013), and recent findings reveal a layer of elevated residues on the soil surface where many species are active (Limay-Rios et al.

2016). Soil exposures, therefore, appear to be an important area for future research, particularly because previous research has leaned toward foliar-dwelling taxa.

In contrast to direct toxicity, there was insufficient evidence to conclude that prey scarcity contributed to reductions in natural-enemy abundance by seed-applied neonicotinoids, although we emphasize that this result may change with additional research. While functional group was not a significant moderator of natural-enemy response to seed-applied neonicotinoids, there was a trend in the direction we would expect if prey scarcity were involved (parasitoid > predator > omnivore). The prey scarcity hypothesis is also supported by a case study on the multicolored Asian lady beetle, Harmonia axyridis. This species is an important predator of the soybean aphid, and its population dynamics in the American Midwest over the past two decades correlated with changes in abundance of its soybean aphid prey, which in turn correlated with use of seed-applied neonicotinoids (Bahlai et al. 2015). More generally, although seed-applied neonicotinoids do not always provide economic control of aphids, they do sometimes reduce their seasonal populations (Hallett et al. 2014; Heidel-Baker 2012;

Johnson et al. 2009; Ohnesorg et al. 2009; Tinsley et al. 2012). In turn, aphids are key prey for many generalist predators in agricultural systems (Donaldson et al. 2007;

Symondson et al. 2002). Future research could test the relative importance of prey scarcity versus direct toxicity through field studies that manipulate prey density independently of neonicotinoid treatment.

139

We expected seed-applied neonicotinoids to be less harmful to natural enemies than foliar or soil-applied pyrethroids, but aside from one outlying study (Ohnesorg et al.

2009), this was not the case. This finding is consistent with previous meta-analyses

(Naranjo 2009; Wolfenbarger et al. 2008) that found a negative effect of pyrethroids on predatory arthropods (versus transgenic Bt varieties) of similar magnitude to the negative effect we found for seed-applied neonicotinoids (versus untreated controls). Pyrethroids are the second most widely used class of insecticides in the world after neonicotinoids

(Sparks 2013), and are important alternatives to seed-applied neonicotinoids in North

American and European field crops (Budge et al. 2015; Douglas & Tooker 2015; Furlan

& Kreutzweiser 2015). Their use is therefore likely to increase if, when, and where neonicotinoid use is restricted. Foliar and some soil-applied pyrethroids have the advantage that they can be applied in response to economic pest populations, and can therefore be more compatible with integrated pest management than seed treatments, which are typically applied to the seed months before planting (Furlan & Kreutzweiser

2015; Johnson et al. 2009). It is also worth noting that pyrethroids have even lower acute toxicity to mammals than neonicotinoids (Tomizawa & Casida 2005), although spray or granular applications would likely also entail different exposures than seed applications.

Additionally, in some cases seed-applied neonicotinoids may be replaced by cultural management tactics or nothing at all. While the full economic, human health, and environmental trade-offs of neonicotinoids versus pyrethroids and other pest management strategies are beyond the scope of this study, our results do suggest that seed-applied neonicotinoids are neither uniquely harmful nor uniquely safe to an important group of non-target invertebrates.

140

Prior to our meta-analysis, the statistical results within and across studies in our dataset appeared highly variable, and a narrative review of these findings could characterize them as mixed. In fact, their measured effects were largely consistent with one another, as reflected in the low heterogeneity of effect sizes across our datasets. This apparent contradiction results from the modest size of the effect combined with the high variability of measurements in field studies, and emphasizes the importance of considering statistical power during ecological risk assessment. Detecting a 20% reduction in natural-enemy abundance with 80% probability requires at least 15 plots per treatment for many predatory arthropod taxa (Prasifka et al. 2008), far more than most studies contributing to our dataset (typically three to six replicates per treatment). While increasing sample size is an obvious solution, logistical and funding constraints make this a challenge. We suggest that researchers interpret null results conservatively in light of statistical power. Periodic meta-analyses may be useful for drawing broader conclusions, as has been the case for transgenic Bt crops (Marvier et al. 2007; Naranjo 2009). The datasets we compiled for this study and a dataset of neonicotinoid effects on bees (Lundin et al. 2015) together provide a foundation for ongoing meta-analyses on the influence of neonicotinoids on non-target species.

There are several important limitations of our meta-analysis that stem from constraints of the dataset we compiled. Because most studies measured natural-enemy abundance within a single field season, our results do not address the influence of neonicotinoid seed treatments on other important metrics like species diversity, sublethal effects on behavior, reproduction of long-lived species, or long-term effects on natural- enemy populations associated with chronic exposure. Furthermore, our dataset comprises

141 manipulative plot studies that by their nature do not account for the movement of natural enemies across landscapes. By focusing on seasonal mean abundance, we may have underestimated important but transient effects that occur only during the period soon after planting. And finally, our study was not able to address the influence of seed-applied neonicotinoids in cropping systems outside of North American and Europe. It is our hope that future meta-analyses will benefit from increased research in these areas.

Conclusion

Using meta-analysis to synthesize the results from field studies in North

American and Europe, we found that seed-applied neonicotinoids reduce natural-enemy populations and are generally no safer for natural enemies than foliar- or soil-applied pyrethroids. The negative effect of neonicotinoids on natural enemies was d = -0.30 ±

0.10 [95% CI], corresponding to a reduction of ~16%. The patterns we observed suggest that seed-applied neonicotinoids exert their effects mainly on insect (versus arachnid) natural enemies, at least partly through direct toxicity. If restrictions on neonicotinoid use encourage substitution with pyrethroids, our results suggest that natural-enemy populations are unlikely to be worse off. In fact, the results of neonicotinoid restriction for natural enemies are likely to be complex, particularly since some pyrethroids can more easily be saved for those situations where economically damaging pest populations occur. Finally, translating natural-enemy abundance into biological control function is not possible given current knowledge, and is an important area for future study.

142

Acknowledgements

We thank the following researchers for sharing data or other information that contributed to this study: Aqeel Ahmad, Ramon Albajes, Christine Bahlai, Galen Dively,

Peter Krauter, Christian Krupke, Jonathan Lundgren, Steven Naranjo, Matthew O’Neal,

Wayne Ohnesorg, and Madeline Spigler. We thank Matthew Reimherr for insights into the statistical analysis, and Mary Barbercheck, Armen Kemanian, Chris Mullin, and members of the Tooker lab for comments that improved the manuscript. This research was supported by grant 2013-41530-21473 from USDA-NIFA, Northeast Regional

Integrated Pest Management Program.

143

Tables

Table 4-1. LC50 results from two laboratory studies that compared imidacloprid toxicity to insect and arachnid predators.

Study Class Order Family Species Life stage LC50 (ppm) Mizell & Arachnida Mesostigmata Phytoseiidae Neoseiulus collegae Adult females >12744 Sconyers Phytoseiulus macropilus Adult females 3561 1992* Proprioseiopsis mexacanus Adult females >1274 Insecta Coleoptera Coccinellidae Olla v-nigrum Adults 3.07 Last instar larva 2.62 Hemiptera Miridae Deraeocoris nebulosus Adults 0.0163 Neuroptera Chrysopidae Chrysoperla rufilabris Adults (pop 1) 190 Adults (pop 2) 155 Eggs 20.2 Tanaka et al. Arachnida Araneae Linyphiidae Gnathonarium exsiccatum 1st instar nymphs 801 2000§ Ummeliata insecticeps 1st instar nymphs 995 Lycosidae Pardosa pseudoannulata 1st instar nymphs 440 Tetragnathidae Tetragnatha maxillosa 1st instar nymphs 136 Insecta Hemiptera Miridae Cyrtorhinus lividipennis Adult females 0.36 Hymenoptera Dryinidae Haplogonatopus apicalis Adult females 0.12 * Residual toxicity; predators were exposed to imidacloprid residues on petri dishes for 48 to 72 hours § Contact toxicity; predators were immersed in insecticide solution and mortality was measured after 24 to 48 hours

144

Table 4-2. Description of the dataset used in a meta-analysis of neonicotinoid seed treatment effects on natural enemies of crop pests.

No insecticide control Pyrethroid control Variable Levels Studies Site-years Obs. (%)* Studies Site-years Obs. (%)* Taxonomic group Insect 20 56 493 (81%) 8 15 313 (82%) Non-insect 14 30 114 (19%) 6 11 71 (18%) Habitat Soil-associated 11 26 189 (31%) 6 10 156 (41%) Aboveground 15 48 418 (69%) 5 11 228 (59%) Functional group Omnivore 6 13 39 (6%) 4 8 41 (11%) Mixed 12 32 79 (13%) 5 8 46 (12%) Predator 17 48 408 (67%) 8 15 262 (68%) Parasitoid 7 27 81 (13%) 2 6 35 (9%) Active ingredient IMI 11 29 336 (55%) 6 12 279 (73%) CLO/THX 13 35 271 (45%) 6 10 105 (27%) Crop species Corn 7 20 300 (49%) 4 10 244 (64%) Soybeans 7 22 200 (33%) 2 5 114 (30%) Other 6 14 107 (18%) 2 3 26 (7%) Publication status Peer-reviewed journal 13 36 459 (76%) 6 12 358 (93%) Diss./Thesis/Other 7 20 148 (24%) 2 3 26 (7%) Pyrethroid application Soil (granular) - - - 5 8 159 (41%) Foliar (spray) - - - 4 10 225 (59%) TOTAL 20 56 607 (100%) 8 15 384 (100%) * Number of observations in each category, followed by the percentage of values in the dataset in that category.

145

Table 4-3. Estimates and tests of significance for moderators in a meta-regression model testing the effect of neonicotinoid seed treatments on natural enemies, compared to controls treated with no insecticides (n = 607 observations from 56 site-years and 20 studies).

Moderator Level β QM df P value Intercept - -0.23 - - - Taxonomic group - 8.70 1 0.003 Insect -0.11 Non-insect (Arachnida, Chilopoda) 0.11 Habitat - 1.42 1 0.23 Aboveground 0.057 Soil-associated -0.057 Functional group - 5.61 3 0.13 Omnivore 0.186 Mixed 0.049 Predator -0.071 Parasitoid -0.164 Crop species - 0.79 2 0.67 Corn (Zea mays) 0.072 Soybean (Glycine max) 0.002 Other -0.074 Active ingredient - 0.99 1 0.32 Imidacloprid 0.043 Clothianidin/Thiamethoxam -0.043 Publication type - 0.51 1 0.56 Peer-review journal 0.062 Dissertation/Thesis/Other -0.062 ln(Plot size) - -0.016 0.34 1 0.56 ln(Early sampling+0.1) - -0.076 0.61 1 0.44

146

Figures

Null model

Site−year/Study model

Moderator model (intercept)

Insects

Non−Insects

−0.6 −0.5 −0.4 −0.3 −0.2 −0.1 0.0 0.1 0.2 Effect size (d)

Figure 4-1. Confidence intervals (95%) for the effect of neonicotinoid seed treatments on natural- enemy abundance, relative to no-insecticide controls. Based on 607 observations from 56 site- years and 20 studies. See text for details on models used to generate these estimates.

147

Null model

Site−year/Study model

Moderator model (intercept)

Null model (− Ohnesorg)

Site−year/Study model (− Ohnesorg)

Moderator model (intercept, − Ohnesorg)

−0.4 −0.2 0.0 0.2 0.4 0.6 0.8 1.0 Effect size (d)

Figure 4-2. Confidence intervals (95%) for the effect of neonicotinoid seed treatments on natural- enemy abundance, relative to controls treated with foliar or soil-applied pyrethroids. Based on 384 observations from 15 site-years and 8 studies. Results are presented both with and without data from Ohnesorg et al. 2009, which had effect sizes quite different from the other studies. See text for details on models used to generate these estimates.

148

Chapter 5

Investigating biopesticides and DNA barcoding as tools to improve insect pest management in lablab bean (Lablab purpureus) in Bangladesh

Abstract

Lablab bean (Lablab purpureus) is a popular vegetable crop in Bangladesh, but farmers growing this crop experience significant losses to insect pests despite heavy reliance on conventional insecticides. Lablab bean is also under continuous threat of emerging pests such as unidentified species of flower-inhabiting thrips. We conducted field studies to improve pest management in lablab bean by i) testing biopesticides as alternatives to conventional insecticides for the control of pod-borers (Maruca vitrata) and aphids (Aphis craccivora), and ii) characterizing flower-inhabiting thrips as an emerging pest in this crop. In field experiments, spinosad was the most promising biopesticide we tested, suppressing pod-boring caterpillars more effectively than thiamethoxam or quinalphos. In contrast, azadirachtin (neem) did not significantly suppress any of the insect pests we measured, although the target aphid populations were generally low at our research site. Using DNA barcoding at the coxI locus combined with morphological identification, we found eight thrips taxa inhabiting lablab bean flowers, dominated by Megalurothrips usitatus and M. distalis/peculiaris. A preliminary regression analysis indicated that a seasonal abundance of 25 thrips/blossom correlated to a 20% reduction in lablab bean yield. Our results suggest that spinosad may be useful within lablab bean IPM programs, and that these programs will likely need to incorporate

149 tactics against thrips to effectively protect yield. Finally, we found that DNA barcoding was a valuable tool to characterize the thrips community in an understudied crop and region, but that to reach its full potential will require an investment in more comprehensive reference libraries.

Introduction

Lablab bean (Lablab purpureus) is an adaptable, multi-purpose legume that has been widely grown across Africa and Asia for over a thousand years (Maass et al. 2010).

In Bangladesh, it is popular as a vegetable and pulse, grown in both commercial fields and home gardens (Atikullah et al. 2013; Islam et al. 2010). Insect pests are a major challenge in this crop, and the most important pest species are pod borers (primarily

Maruca vitrata) and bean aphids (Aphis craccivora) (Ahmed et al. 2004; Kabir et al.

1996). Farmers in Bangladesh use insecticides intensively against these pest species, sometimes spraying more than once a week during the growing season, and often still do not achieve satisfactory pest control (Hoque et al. 2002; Kabir et al. 1996). At the same time, this indiscriminate use of insecticides, often organophosphates, negatively influences farmer health, water quality, and wildlife (Dasgupta 2003; Dasgupta et al.

2005). There is a clear need to develop integrated pest management (IPM) strategies for lablab bean that are more effective and safer for farmers and the environment.

To contribute to development of an IPM program for lablab bean in Bangladesh, we conducted a field study to test the efficacy of biologically derived insecticides

(hereafter ‘biopesticides’) against key pests. In previous studies, sprays based on Bacillus thuringiensis (Bt), Maruca vitrata multiple nucleopolyhedrovirus (MaviMNPV), and the

150 actinomycete-derived spinosad showed promise to control pod borers (Lee et al. 2007;

Sreekanth & Seshamahalakshmi 2012; Srinivasan 2008; Srinivasan et al. 2012), while botanical insecticides based on neem (Azadirachta indica) showed promise against bean aphids (Dalwadi et al. 2008; Das et al. 2008). There is little information on the efficacy of these biopesticides in lablab bean production under field conditions. We hypothesized that pest management programs combining MaviMNPV, Bt, or spinosad for pod borers and neem for aphids would protect yield as well or better than broad-spectrum insecticides typically used by farmers.

As our field studies progressed, we observed relatively high densities of thrips inhabiting lablab bean flowers. A previous report identified flower thrips (mainly

Megalurothrips distalis and Frankliniella schultzei) infesting lablab bean as potential pollinators in India (Velayudhan et al. 1985), but there is little research on thrips as pests in this crop. M. distalis was listed as an ‘important pest’ of lablab bean in Gujarat, India

(Dalwadi et al. 2008), and scientists in Bangladesh have observed that flower thrips have become more abundant in lablab bean over the past several years. In other legume crops, flower-inhabiting thrips can cause large yield reductions through premature bud or flower abscission (Ananthakrishnan 1993; Tamò et al. 1993) and/or by transmitting plant viruses

(Jones 2005; Pappu et al. 2009). Since the taxonomic identity of thrips species in

Bangladesh is not clearly defined, we expanded our study to characterize the thrips community in lablab bean in Bangladesh and to gain preliminary insights into the damage potential of thrips in this crop.

Thrips are challenging to characterize as crop pests because their diversity and small size make them difficult to identify, and their mobility means that host-plant

151 relationships cannot be inferred solely from presence of adults (Mound 2013; Tyler-

Julian et al. 2014). DNA barcoding may help to address both of these problems, explaining why it is quickly becoming an important tool in thrips research (Iftikhar et al.

2016; Kadirvel et al. 2013; Macharia et al. 2015; Rebijith et al. 2014). The ‘DNA barcode’ for animals is the sequence of the mitochondrial cytochrome c oxidase subunit I gene (coxI), which tends to show < 2% sequence divergence within species and > 2% sequence divergence between species (Hebert et al. 2003; Meyer & Paulay 2005). When used for identification, DNA barcodes from specimens of interest are matched to those of reference specimens (Collins & Cruickshank 2013). The Barcode of Life Data System

(BOLD, http://www.boldsystems.org) serves as a public repository for curated sequences and includes several tools for characterizing genetic diversity at the coxI locus

(Ratnasingham & Hebert 2007). One of these is the Barcode Index Number (BIN), a taxonomic unit thought to align closely with species (Ratnasingham & Hebert 2013), which in BOLD is automatically assigned to each sequence that is high quality and >

500 base pairs. In addition to assisting adult identification, DNA barcodes can be used to match life stages within a species. This may be especially valuable for thrips, because the larvae are often responsible for crop damage (Ananthakrishnan 1993; Tamò et al. 1993), but are even more challenging to identify than adults.

Our study had two major objectives, both related to improving IPM in lablab bean in Bangladesh. First, we tested the hypothesis that biopesticides could be effective alternatives to conventional synthetic insecticides in managing key insect pests of lablab bean. To do this we measured pest abundance and crop yield in two field experiments that compared biopesticides to conventional insecticide regimes. Second, we

152 characterized the thrips community inhabiting lablab bean flowers using morphological keys combined with DNA barcoding. We were particularly interested in using molecular tools to match unknown larval specimens to adult specimens, thereby establishing which species are breeding in lablab bean flowers. We also gleaned from our field experiments preliminary insights into the influence of flower-inhabiting thrips on lablab bean yield.

Materials & Methods

Field experiments

To evaluate biopesticides for their use in IPM programs in lablab bean, we established two field experiments at the Bangladesh Agricultural Research Institute

(BARI; Gazipur, Bangladesh; 23.992, 90.415).

Experiment 1 (early winter)

This experiment was designed as a randomized block design with six treatments

(n = 3 plots/treatment). The planned treatments were: 1) Untreated control, 2)

MaviMNPV once/week, 3) MaviMNPV once/week + Neem once/two weeks, 4) B. thuringiensis subsp. kurstaki once/week, 5) B. thuringiensis subsp. kurstaki once/week +

Neem once/two weeks, and 6) Organophosphate once/week (for formulations and rates see Table 5-1). MaviMNPV and B. thuringiensis subsp. kurstaki were intended to control pod borers while neem was intended to control aphids. The organophosphate treatment represents a common farmer practice and is meant to control all insect pests; the

153 organophosphate product we chose, quinalphos, is common in the area. Under an IPM framework, we began all treatments when the pests reached a working threshold (10% of inflorescences infested with aphids; 10% of pods infested with pod borers). However, pod borers never reached this threshold and so MaviMNPV and B. thuringiensis subsp. kurstaki were not applied, reducing the experiment to three treatments: untreated control

(n = 9 plots), neem (n = 6 plots), and organophosphate (n = 3 plots). Lablab bean seeds of the variety BARI Sheem 2 were planted on October 10, 2014 in 5 m 2.5 m plots (6 hills per plot, 3-4 plants per hill). Neem and organophosphate treatments were initiated on December 18, 2014. Lablab bean pods were harvested roughly weekly from December

24, 2014 to March 4, 2015.

Experiment 2 (late winter)

This experiment was designed as a randomized complete block design (n = 3 plots/treatment) with each plot receiving one of the following treatments: 1) Untreated control, 2) Spinosad, once/week, 3) Neem, once/week, 4) Alternating Spinosad and Neem once/week, 5) Quinalphos, once/week, 6) Thiamethoxam once/week (for formulations and rates see Table 5-1). This experiment served two purposes: i) to test insecticides for their potential to control insect pests, and ii) to generate a range in thrips densities to better understand the effect of thrips on lablab bean flowering and yield. We expected thiamethoxam to provide the greatest thrips control because of its systemic activity

(Jeschke et al. 2011), given that thrips were fairly well concealed inside lablab bean flowers. Lablab bean seeds of the variety BARI Sheem 2 were planted on October 19,

2014 in 4 m 2.5 m plots (4 hills per plot, 3-4 plants per hill). All treatments were

154 initiated on January 19, 2014 and lablab bean pods were harvested roughly weekly from

January 15, 2015 to March 23, 2015.

Measuring insect pest populations, flowering activity, and yield

To measure populations of aphids and thrips, we began sampling 4-16 inflorescences per plot weekly at bloom initiation. Fewer inflorescences were measured when insect populations were high. First, we examined each stalk and recorded the presence or absence of aphids. Then, we measured thrips abundance by holding the same inflorescence over a white piece of paper and tapping it vigorously to dislodge insects, recording thrips abundance in the field. This method does not capture absolute thrips densities but the results of plant tapping have correlated with absolute thrips densities in earlier studies (Shipp & Zariffa 1991). This method predominantly samples adult thrips.

In Experiment 1, we sampled insects weekly from December 15, 2014 to January 6,

2015, after which most plots had insufficient flowers to continue sampling. In

Experiment 2, we sampled insects weekly from January 13, 2015 to March 23, 2015, although differences in flowering activity across treatments meant that not all plots were sampled each week.

Partway through Experiment 1 we noticed that the treatments were affecting flowering activity; thus, we recorded the number of inflorescences per plot five times between January 21, 2015 and March 4, 2015. In Experiment 2, we recorded the number of inflorescences per plot weekly from February 1, 2015 to March 23, 2015. We counted those inflorescences that had at least one fully opened blossom. In Experiment 2 we also

155 recorded the number of fully opened blossoms per inflorescence for those inflorescences sampled by tapping.

To measure the influence of the treatments on yield, each week we harvested, counted, and weighed the marketable pods from each plot. We also recorded the number and mass of pods that were damaged by pod-boring caterpillars.

Statistical analysis of field experiments

We conducted all statistical analyses in R 3.2.3 (R Core Team 2015). To investigate the influence of insecticide treatments on aphid and thrips abundance and flowering activity, we used different approaches in the two experiments because of the way our sampling evolved over the course of the study. For Experiment 1, we tested the effects of treatments on these responses over time using linear mixed effects models in the ‘nlme’ package (‘lme’ command). We used the percentage of inflorescences infested as the response variable for aphids and the number of thrips per inflorescence as the response variable for thrips. The model for each response variable included insecticide treatment and its interaction with date as fixed effects, and a random effect of plot to control for the repeated nature of the data. We chose among candidate covariance structures using Akaike information criteria (Littell et al. 2006).

For Experiment 2, because flowering activity differed among treatments, pest data were missing for some treatments on some dates. We also had the benefit of additional data on inflorescence and bloom production. Therefore, we examined the influence of treatments on a seasonal measure of aphid and thrips abundance. For aphids, we first estimated the total number of inflorescences infested in each plot on each date by

156 multiplying the percentage of inflorescences infested by the number of inflorescences.

We then calculated seasonal infestation by dividing the seasonal sum of infested inflorescences by the seasonal sum of all inflorescences in each plot. For thrips, we first estimated the total thrips in each plot on each date by multiplying the mean number of thrips per inflorescence by the number of inflorescences. We also calculated the number of blossoms in each plot on each date by multiplying the number of open blossoms per inflorescence by the number of inflorescences. Finally, we calculated seasonal thrips per blossom by dividing the seasonal sum of thrips by the seasonal sum of blossoms. To look at the influence of treatments on flowering activity, we took an average of the number of inflorescences per plot across sample dates. To examine the influence of insecticide treatments on seasonal aphid and thrips abundance, and production of inflorescences, we subjected these seasonal response variables to one-way analysis of variance (ANOVA) using the ‘aov’ command, with insecticide treatment as the independent variable. Where a significant treatment effect occurred, we used Tukey’s Honestly Significant Difference

(HSD) post-hoc test to separate means.

In both experiments, to examine the influence of insecticide treatments on pod- borer damage and marketable yield, we first calculated for each plot a seasonal total for each response variable across all post-treatment sample dates. We then subjected these response variables to one-way ANOVA using the ‘aov’ command, with insecticide treatment as the independent variable. Where a significant treatment effect occurred, we used Tukey’s HSD post-hoc test to separate means. Where inspection of the residuals suggested heterogenous variances, we used Welch’s ANOVA (‘oneway.test’), which does not assume equal variances.

157

We performed an additional, exploratory analysis based on the results of

Experiment 2 to investigate the relationship between insect pests and lablab bean yield.

Using the ‘lm’ command, we fit a regression model with marketable yield as the response variable and aphid infestation, thrips abundance, and pod borer damage as predictor variables.

Throughout the analyses, data were transformed as necessary to meet parametric assumptions. We present untransformed means in the results, but note where statistical results correspond to transformed data.

Characterizing the thrips community in lablab bean

Sampling and initial sorting

To characterize the thrips community, we destructively sampled lablab bean inflorescences at the beginning and end of the two field experiments. We collected three inflorescences per plot and placed them into a polyethylene bag. At the lab, we dissected the blossoms and transferred all insects into 70% ethanol. In addition to the thrips collected from our experiments in Gazipur, we collected thrips from lablab bean flowers in Jessore, Bangladesh, a major lablab-growing region, in early February 2015. While we were not able to sample comprehensively in Jessore, these samples do provide preliminary insight into the consistency of the thrips community in lablab bean across regions of Bangladesh.

Thrips were initially sorted and counted by morphotype. We then selected representatives of each morphotype for morphological and molecular identification,

158 aiming for at least six individuals per morphotype. However, this was not possible for some rare morphotypes.

DNA extraction, amplification, sequencing, and alignment

To extract DNA, whole thrips were placed individually in 0.2 ml PCR tubes containing 20 µl of DNA extraction buffer (UniversAllTM Tissue Extraction/PCR Kit,

Yeastern Biotech Co., Ltd., Taipei, Taiwan) and incubated at 96oC for 6 min. This method left thrips individuals intact for later morphological identification. After extraction, 5 µl of each DNA sample was diluted in 20 µl of sterile distilled water and stored at -20oC.

To amplify coxI sequences, we used the degenerate primers dgLCO-1490 ([5’-3’]

GGT CAA CAA ATC ATA AAG AYA TYG G) and dgHCO-2198 ([5’-3’] TAA ACT

TCA GGG TGA CCA AAR AAY CA) (Meyer 2003). For each sample we assembled a

50 µl polymerase chain reaction (PCR) comprising: 8 µl template DNA, 2 µl each dgLCO and dgHCO primers, 10 µl pre-made master mix (5X PCR Dye Master Mix II,

GMbiolab Co., Ltd., Taichung, Taiwan), and 28 µl sterile distilled water. Touchdown

PCR (Korbie & Mattick 2008) was performed in an MJ Research thermal cycler (PTC-

200 Peltier Thermal Cycler, Bio-Rad Laboratories, Hercules, California, USA) with the following program: 94oC for 5 min; then nine cycles of 94oC for 30 s, 52oC to 48oC for

45 s [decreasing 0.5oC/cycle], 72oC for 1 min; then 29 cycles of 94oC for 30 s, 47oC for

45 s, 72oC for 1 min; then a final extension at 72oC for 7 min. To check the quality and size of the amplified products, we electrophoresed them on 1.5% agarose gels containing

EtB“Out” Nucleic Acid Staining Solution (Yeastern Biotech Co., Ltd., Taipei, Taiwan).

159

PCR products from successful reactions were sequenced in both the forward and reverse directions on an ABI 3730 XL DNA Analyzer by Genomics Biotechnology Company

Limited, Taiwan. We inspected DNA sequences for quality, aligned them with their complements, and edited them using Vector NTI Advance® 11.5.1. (Life Technologies

Corporation, Carlsbad, California, USA).

Thrips identification using morphological characters.

Following DNA extraction, we slide-mounted adult thrips specimens and identified them morphologically using taxonomic keys (Moritz et al. 2004; Mortiz et al.

2009; Mound & Sartiami 2013; Tyler-Julian et al. 2014). Adult specimens were identified to species with the exception of Phlaeothripidae, which in most cases could be identified to genus. Slide-mounted specimens are stored in the thrips IPM program laboratory at the

International Centre of Insect Physiology and Ecology (Nairobi, Kenya).

Phylogenetic analysis.

We submitted sequence data to BOLD under the project ‘Lablab bean thrips’

(LLBT). Through the BOLD system, sequences were automatically assigned to a

Barcode Index Number (BIN; Ratnasingham & Hebert 2013).

To identify larval specimens and to compare the results of morphological and genetic identifications, we built a neighbor-joining tree with coxI-5’ sequences of our specimens, together with reference sequences obtained from BOLD (Table 5-2).

Wherever possible, we chose reference specimens from peer-reviewed studies in which

160 taxonomic experts verified identification of thrips. We also sought reference specimens for the breadth of barcode index numbers (BINs) for those species that have been associated with multiple BINs, which may indicate cryptic species complexes (Iftikhar et al. 2016). For Phlaeothripidae, because we did not have species-level identifications based on morphology, we included for comparison as many reference taxa as possible from a list of species previously documented in Bangladesh (ThripsWiki 2016). We used the software MEGA version 6.06 (Tamura et al. 2013) to align consensus sequences for all thrips specimens and build a neighbor-joining tree of coxI-5’ variants based on p- distances (Srivathsan & Meier 2012) with node support assessed using 1000 bootstrapping iterations.

Our initial analysis revealed an ambiguity in the identification of Megalurothrips species, and so we constructed an additional neighbor-joining tree that included sequences from our specimens together with publicly available sequence data. Using

BOLD’s ‘Public Data Portal’, we searched for sequence data from the coxI-5’ region for

Megalurothrips and its associated BINs. As before, we used the software MEGA (version

6.06) to align consensus sequences for all thrips specimens and build a neighbor-joining tree of coxI-5’ variants using p-distances with node support assessed using 1000 bootstrapping iterations.

On the basis of BIN assignments and groupings within the neighbor-joining trees, we assigned larvae and unknown adult specimens to the lowest taxonomic unit possible.

We then used MEGA (version 6.06) to calculate within-group and between-group genetic distances for each BIN in our dataset, using complete deletion for gaps and missing data

(Fregin et al. 2012). We summarized these data by reporting mean and maximum genetic

161 distances within BINs, as well as distance of each BIN to its nearest neighbor in another taxon. We compared the results calculated from our dataset with genetic distances calculated for the same BINs including all worldwide sequences in BOLD; these measures were calculated within the BOLD system.

Results

Field experiments

As described in the introduction, we observed unexpectedly high thrips densities in the field experiments. Thrips were especially concentrated in lablab bean flowers, but also occurred on leaves and pods (Figure 5-1). Thrips were associated with characteristic stippling of flowers and pods (Figure 5-1).

Insecticide effects on pest abundance

In Experiment 1, insecticide treatments did not significantly influence the percentage of inflorescences infested by bean aphids (Treatment × Date F6,45 = 1.05, P =

0.41; Treatment F2,15 = 1.63, P = 0.23; Figure 5-2b). Aphid abundance was generally low, with < 20% of stalks infested in any treatment over the course of the experiment. Pod borer damage was also low in this experiment (< 3.5% in all treatments), and did not differ by treatment (Welch’s ANOVA F2,6.3 = 0.244, P = 0.79; data not shown). In contrast, insecticide treatments exerted significant effects on thrips abundance over time

(Treatment × Date F6,45 = 5.22, P < 0.001). The largest differences among treatments

162 occurred on the third sample date, when thrips abundance was reduced by roughly half in quinalphos-treated plots compared to neem-treated and untreated plots (Figure 5-2).

Thrips abundances on the other sampling dates were similar across treatments (Figure

5-2a).

In Experiment 2, as in the first experiment, the percentage of inflorescences infested by bean aphids was not significantly influenced by insecticide regime (F5,12 =

1.87, P = 0.17), although numerically thiamethoxam, quinalphos, and neem tended to reduce aphid infestation (Figure 5-3b). Aphid infestation was again generally low, with less than 10% of inflorescences over the season infested with aphids. Pod borer damage was also low (< 6% in all treatments, Figure 5-3a), but nonetheless was influenced by treatment (F5,12 = 6.4, P = 0.004). Plots treated with spinosad alone or with alternating spinosad and neem had lower pod borer damage than plots treated with quinalphos, thiamethoxam, or no insecticide (Figure 5-3a). Insecticide treatments strongly influenced thrips abundance (F5,12 = 20.7, P < 0.001). The number of thrips per blossom over the season was significantly lower in quinalphos-treated plots and thiamethoxam-treated plots than the untreated, neem-treated, and spinosad-treated plots (Figure 5-3). Plots treated with alternating spinosad and neem had intermediate thrips abundance (Figure

5-3c).

Insecticide effects on flower production and yield

In the first experiment, insecticide treatments had a strong effect on flowering phenology (Figure 5-2b; Treatment × Date F8,60 = 160, P < 0.0001; compound symmetry model with heterogeneous variance by date). Starting roughly one month after beginning

163 insecticide treatments, plots treated with quinalphos had ~75 inflorescences per plot while the neem-treated or untreated plots had virtually none (Figure 5-2b). Several weeks later, the pattern reversed as the neem-treated and untreated plots experienced a second flush of flowers, but the quinalphos-treated plots did not (Figure 5-2b). Insecticide treatments also significantly affected lablab bean yield (Welch’s ANOVA F2,9.7 = 28.4, P

< 0.001). Bonferroni-corrected, pairwise comparisons showed that untreated plots had the lowest yield [4.4 ± 0.2 kg (mean ± SE)], neem-treated plots had intermediate yield (5.4 ±

0.2 kg), and quinalphos-treated plots had the highest yield (6.3 ± 0.1 kg).

In Experiment 2, insecticide treatments again influenced lablab bean flower production (F5,12 = 4.0, P = 0.02; Figure 5-3d). Mean inflorescences over the season were significantly higher in thiamethoxam-treated plots than in the untreated control, with intermediate values in the rest of the treatments (Figure 5-3d). Flowering activity generally followed a pattern opposite that of seasonal thrips abundance (Figure 5-3d).

Insecticide treatments did not significantly affect post-treatment yield (F5,12 = 1.3, P =

0.33), although the numerical trend was for higher yields in plots treated with thiamethoxam, quinalphos, and alternating spinosad/neem (Figure 5-3e).

Relationship between pest abundance and lablab bean yield

An exploratory regression model based on the data from Experiment 2 suggested that lablab bean yield decreased linearly with thrips abundance (t = -4.1, P = 0.001;

Figure 5-4a) and pod-borer damage (t = -4.8, P < 0.001; Figure 5-4b), with marginal evidence that aphids reduced yield (t = -2.0, P = 0.07). The overall model was highly significant and the three pests together explained over half the variability in lablab bean

164

2 yield in the experiment (F3,14 = 10.1, P < 0.001, adjusted R = 61.7%). The fitted slope for thrips abundances [-0.091± 0.022 (estimate ± SE)] suggested that high densities of thrips

(25 thrips/blossom, the higher end of what we observed) corresponded to ~20% decrease in lablab bean yield. Interestingly, the fitted slope for pod-borer damage (-0.29 ± 0.06) suggested that pod borers caused damage beyond infesting pods; for every 1% of pod damage, lablab bean yield declined by 5%. This may be because M. vitrata, the main pod-boring species at our site, can damage buds and flowers as well as pods (Sharma

1998).

Characterizing the thrips community in lablab bean

DNA extraction and amplification

We performed DNA extractions on 138 thrips specimens, yielding 57 successful

PCR reactions with high-quality sequence data (48 from Gazipur, 9 from Jessore). The low success rate of PCR reactions relative to the number of DNA extractions may have been related to degradation of samples in transit from Bangladesh to Taiwan, when we had to remove storage ethanol for > 48 h for air travel. An amplicon ranging from 563-

683 bp was produced from each sample by PCR using the coxI gene specific primers dgLCO-1490 and dgHCO-2198. Sequence alignment and editing resulted in a consensus sequence of 537 bp across all thrips samples. The sequences have been submitted to

BOLD and NCBI GenBank and the accessions numbers are given in Table 5-4.

165

Correspondence between morphological and molecular identifications

Using taxonomic keys for morphological identification, we placed adult specimens of Thripidae into five species. All but one Phlaeothripidae specimen were placed in the genus Haplothrips; the remaining specimens could be identified only to family level.

For four of the species in Thripidae (Frankliniella intonsa, Thrips hawaiiensis,

Thrips palmi, and Megalurothrips usitatus), there was good agreement between morphological identification, sequences from reference specimens, and BIN numbers assigned by BOLD (Figure 5-5). However for one species, Megalurothrips distalis, the various approaches to identification yielded contradictory findings. In the neighbor- joining tree, the specimens which we morphologically identified as M. distalis formed a well-supported clade, but this clade included reference specimens of both M. distalis and

M. peculiaris (Figure 5-5). In a follow-up neighbor-joining tree that combined our

Megalurothrips specimens with all publicly available records from this genus and its corresponding BINs, we found again that M. usitatus formed an unambiguous clade, but that M. distalis and M. peculiaris were indistinguishable (Figure 5-6). Therefore, we referred to specimens in this clade as M. distalis/peculiaris.

Of the ten species in Phlaeothripidae previously recorded in Bangladesh

(ThripsWiki 2016), five lacked reference specimens in the BOLD database (Ethirothrips obscurus, Bamboosiella exastis, Ecacanthothrips tibialis, Gynaikothrips bengalensis, and

Podothrips lucasseni), and so could not be included for comparison in our neighbor- joining tree. The tree together with BIN numbers suggested that our specimens were mainly in one clade that is likely a species of Haplothrips, agreeing with our

166 morphological identifications (Figure 5-5). Notably, this clade appeared distinct from the reference specimens for the three Haplothrips species previously recorded in Bangladesh

(Figure 5-5). We ran all of our Phlaeothripidae through the BOLD identification engine

(Public Record Barcode Database) to determine if they matched any other species for which molecular data are available, but they did not. On the basis of the neighbor-joining tree together with our morphological identifications we classified our specimens in

Phlaeothripidae as Haplothrips sp. 1 or 2, and Phlaeothripidae sp. 1.

All of the larval specimens unambiguously clustered with a clade of adult specimens (Figure 5-5). Larvae belonged to M. usitatus, M. distalis/peculiaris, or

Haplothrips sp. 2. Notably, we did not find larvae of F. intonsa or Thrips spp., suggesting that these taxa may not complete their life cycles in lablab bean.

Genetic distance

Results from the genetic distance analysis supported the groupings that we identified through the combination of morphological and molecular identification, but also highlighted the importance of the sampling frame. Considering only our specimens, maximum genetic distances within BINs ranged from 0.19% to 1.30%, while nearest neighbors were all > 3% (Table 5-3). When we examined genetic distances including all

BIN-associated sequences in BOLD, distances within BINs predictably increased, ranging from 0.52% to 3.30% (Table 5-3). Nearest neighbors were again > 3% (Table

5-3). Importantly, when each BIN was considered individually, there was still evidence of a ‘barcode gap’ (Meyer & Paulay 2005), a difference between intra-BIN and inter-BIN genetic distances.

167

One of our sequences (Haplothrips sp. 1) was sufficiently distinct from other records in BOLD to be assigned to a new BIN (Table 5-3). All other sequences were grouped with at least two existing records, with most matches originating from India,

Pakistan, and China, although the cosmopolitan F. intonsa grouped with records from

Europe and North America as well as Asia (Table 5-3). Our single specimen of Thrips palmi, which is associated with multiple BINs, grouped with records from India and

Pakistan (Table 5-3).

Conclusions about the thrips community

Matching our initial morphotaxa with our eventual identification, the thrips community in lablab bean was dominated by Megalurothrips spp., which made up 77% of adults and 95% of larvae destructively sampled from lablab bean inflorescences.

Haplothrips spp. contributed 11% of adults and the remaining 5% of larvae. F. intonsa contributed 9% and Thrips spp. 2% to the adult collections.

Discussion

Our study had two main goals: i) to test biopesticides as a replacement for conventional synthetic insecticides in lablab bean IPM programs, and ii) to characterize flower-inhabiting thrips as a potential emerging pest in this crop. Generally the biopesticides we tested did not reliably control insect pests or increase yield, although low aphid and pod borer densities made it challenging to test this hypothesis. An important exception was that spinosad controlled pod borers more effectively than

168 conventional insecticides. Concerning our second goal, morphological identification combined with DNA barcoding revealed that flower-inhabiting thrips in lablab bean were dominated by Megalurothrips usitatus and M. distalis/peculiaris. High densities of thrips were associated with reduced flowering and yield, suggesting that IPM programs in this crop may need to incorporate thrips control.

The failure of biopesticides to reliably increase yield may have been related to the realized pest community in our experiments. In particular, low aphid abundance precluded a strong conclusion concerning the efficacy of neem in this system. Future studies could test neem at sites with higher aphid abundance, perhaps trialing a variety of formulations, which appear to vary somewhat in efficacy against this pest (Dalwadi et al.

2008; Das et al. 2008). Spinosad did show good efficacy against pod-borers, consistent with both its general potency against lepidopteran larvae (Thompson et al. 2000), and recent studies showing good control of Maruca spp. in particular (Sreekanth et al. 2015;

Umbarkar & Parsana 2014; Yadav & Singh 2014). Interestingly, spinosad suppressed pod borers similarly regardless of whether it was applied weekly or every other week; less frequent application might ameliorate the higher cost of spinosad compared to some conventional insecticides (Sreekanth et al. 2015). On the basis of acute toxicity, spinosad also has very good selectivity for insects versus mammals (Thompson et al. 2000), and so may help reduce farm-worker health problems associated with insecticide use (Dasgupta

2003; Dasgupta et al. 2005).

Details on morphological characterization of thrips species infesting lablab bean or other vegetable and grain legumes are scanty in Bangladesh. Using morphological identification combined with DNA barcoding, we were able to identify the major thrips

169 species at our site as M. usitatus and M. distalis/peculiaris, laying the groundwork for future studies on the damage potential of these pests. Our finding of larvae of the two

Megalurothrips species and one Haplothrips species demonstrate that at least these three species can complete their development in lablab bean. The occurrence at lower numbers of adults of F. intonsa and T. palmi are concerning, as these species are known vectors of damaging tospoviruses in legumes (Pappu et al. 2009). Most relevant to this study, T. palmi can transmit Groundnut bud necrosis virus, which has been documented in lablab bean in India (Jain et al. 2007; Pappu et al. 2009).

While Megalurothrips sjostedti is a major pest of cowpea in Africa (Tamò et al.

1993), there is little literature on the damage potential of Megalurothrips spp. in Asian legumes. In lablab bean, determining the influence of flower-inhabiting thrips on yield is complicated by two factors. First, this crop naturally aborts many of its flowers without producing pods (Ayyangar & Nambiar 1935), and second, thrips are able to pollinate lablab bean flowers (Velayudhan et al. 1985), potentially increasing yield, although lablab bean is generally regarded as self-pollinating (Ayyangar & Nambiar 1935). In our experiments, insecticides that reduced thrips abundance tended to increase yield, and a preliminary regression based on our second field experiment found a linear decrease in yield with increasing seasonal thrips abundance. These findings suggest that thrips may indeed have a strong influence on pod production in lablab bean, and emphasize the need for a careful investigation of the interaction between thrips and lablab bean flower development and yield. Future studies should include the influence of larval feeding on bud and flower development, since this appears to be the mechanism largely responsible for damage by M. sjostedti in cowpea (Tamò et al. 1993).

170

Previous DNA barcoding studies, including studies on thrips, have suggested that named species are in fact complexes of cryptic species (Hajibabaei et al. 2006; Iftikhar et al. 2016). We found the opposite for one of the taxa studied here: thrips that we identified as M. distalis failed to reliably segregate from M. peculiaris based on molecular data.

Interestingly, these two species also failed to be distinguished based on a multivariate analysis of morphological characters (Palmer 1987). M. peculiaris has a sensorium base similar to Odontothrips, and has only been recorded from India and Bangladesh (Mirab-

Balou et al. 2013; ThripsWiki 2016). This species was not reported in the three previous studies we found on thrips in L. purpureus, all from India (Dalwadi et al. 2008; Maisnam et al. 2012; Velayudhan et al. 1985). In contrast, all three studies recorded M. distalis on

L. purpureus, and this species has been recorded as a major thrips species infesting about

50 legume species and is widely distributed across Asia, Africa, Australia, the Americas and Europe (Plantwise Knowledge Bank 2016). Based on our results, we cannot say whether M. peculiaris and M. distalis are really one species, or if one is simply a young species that is not yet distinct from the other at the coxI locus (Meyer & Paulay 2005).

Resolving this question will likely require the use of multiple molecular loci combined with morphological traits.

Our experience with DNA barcoding highlights both the challenges and the promise of this approach to species identification. While DNA barcoding allowed us to match unknown larval specimens to adult specimens and reinforced some of our adult identifications, it was of limited use in identifying unknown adult specimens. A complete and well-verified reference library is a prerequisite for molecular identification (Collins

& Cruickshank 2013; Meyer & Paulay 2005), and unfortunately the public reference

171 library on BOLD is still quite incomplete, at least for the focal taxa of this study (e.g. only half of the Phlaeothripidae species known to occur in Bangladesh are represented).

More positively, BOLD is growing quickly in its regional and taxonomic coverage. The

BIN numbers associated with our unknown specimens provide a mechanism for these records to be matched with future records. More broadly, taxonomic expertise is a major barrier to pest management programs, especially in developing countries. There is a shortage of taxonomic expertise in Bangladesh, a country of >150 million people and a major agricultural producer of many crops including grain and vegetable legumes. Our case study supports the contention that DNA barcoding can be a valuable tool under these constraints (Floyd et al. 2010; Miller 2007), although to reach its full potential it will require a major investment in reference libraries.

In conclusion, we found that spinosad has value against pod-borer pests in lablab bean production but that neem requires further testing under heavy aphid pest pressure.

Flower-inhabiting thrips in this crop were dominated by M. usitatus and M. peculiaris/distalis, and at high densities they appeared to reduce lablab bean flowering with consequent reductions to yield. We suggest that future work investigate the influence of Megalurothrips larvae on lablab bean bud and flower development, and explore potential control tactics for this emerging pest.

Acknowledgements

We thank Mei-ying Lin (AVRDC) for logistical help and members of the BARI

IPM laboratory and field crew for help maintaining field experiments and collecting data.

This work was supported by the U.S. Agency for International Development (USAID)

172 under the U.S. Borlaug Fellows in Global Food Security Program at Purdue University

(subaward no. 8000059691 to M. Douglas and J. Tooker), and AVRDC’s project entitled

"Improving Incomes, Nutrition and Health in Bangladesh through Potatoes,

Sweetpotatoes and Vegetables” also funded by USAID, through the International Potato

Centre (CIP).

173

Tables

Table 5-1. Insecticide formulations and rates applied to lablab bean (Lablab purpureus) in field experiments in Bangladesh.

Active ingredient Formulation Distributor Rate MaviMNPVa 1 X 109 occlusion S. Taiwan University of Science 1.0 g/L bodies/g & Technology Bt sbsp. Kurstaki E-911 60% w/w FWUSOW Industry Co., Ltd. 1.0 g/L Azadirachtin Bioneem Plus 1EC Ispahani Biotech Ltd. 1.0 ml/L Quinalphos Corolux 25 EC Corbet International Ltd. 2.0 ml/L Spinosad Tracer 45 SC Auto Cropcare Ltd. 0.4 ml/L Thiamethoxam Actara 25 WG Syngenta Bangladesh Ltd. 0.2 g/L a This is not a commercial insecticide; it is produced through mass rearing and isolating MaviMNPV through a joint project of AVRDC and the S. Taiwan University of Science & Technology.

174

Table 5-2. Reference specimens used in the molecular identification of thrips specimens

GenBank BOLD sequence Family Species Country BIN Reference accession number Phlaeothripidae Ananthakrishnana Pakistan KP871449 MATHR454-12 ACA2783 Iftikhar et al. 2016 euphorbiae Dyothrips pallescens N/A KC513147 GBMIN12475-13 AAY4258 Buckman et al. 2013 Haplothrips ganglbaueri Pakistan KP871455 MATHR059-10 ACF1370 Iftikhar et al. 2016 Haplothrips gowdeyi Pakistan KP871389 MATHR050-10 AAN5798 Iftikhar et al. 2016 Haplothrips tenuipennis Pakistan KP871477 MATHR051-10 AAN4488 Iftikhar et al. 2016 Thripidae Frankliniella intonsa Japan AB587606 GBMIN39469-13 AAF6737 N/A Japan AB277214 GBMH2773-07 N/A Inoue & Sakurai 2007 Megalurothrips distalis China HQ540407 GBMIN30301-13 AAN6623 N/A Megalurothrips peculiaris Pakistan HQ991655 MATHR027-10 AAN6623 Iftikhar et al. 2016 Megalurothrips usitatus Pakistan KP845774 MATHR015-10 AAM8053 Iftikhar et al. 2016 Thrips hawaiiensis Pakistan KP871187 MATHR315-11 AAZ8516 Iftikhar et al. 2016 Thrips palmi Pakistan JF839936 MATHR178-10 AAE7913 Iftikhar et al. 2016 Pakistan HQ991644 MATHR013-10 AAN2747 Iftikhar et al. 2016 Canada KT708311 RRINV150-15 AAD4600 Telfer et al. 2015

175

Table 5-3. Pairwise genetic distances (p-distances) for thrips taxa collected from lablab bean (Lablab purpureus) in Bangladesh. Genetic distances are shown for only the specimens from this study (“This study”), as well as for the specimens from this study combined with all other records for each BIN in the Barcode of Life Data Systems (BOLD).

Genetic distance (%) Taxon BIN Records N Mean Max NNa Other countries with BIN records Haplothrips sp. 1 ACZ7039 This study 1 - - 3.2 - BOLD 1 - - 3.4 Haplothrips sp. 2b ACY8730 This study 7 0.11 0.37 3.7 India BOLD 8 0.17 0.52 3.4 Phlaeothripidae sp. 1 ACI5935 This study 1 - - 18.3 India BOLD 3 1.61 2.49 17.0 Frankliniella intonsa AAF6737 This study 14 0.03 0.19 16.8 China, India, Japan, Norway, Russia, BOLD 44 0.71 3.30 12.7 Serbia, United States Megalurothrips AAN6623 This study 9 0.11 0.37 8.2 China, Pakistan distalis/peculiaris BOLD 49 0.33 1.10 8.4 Megalurothrips usitatusc AAM8053 This study 22 0.63 1.30 8.2 Australia, China, French Polynesia, India, BOLD 56 0.59 1.65 8.4 Indonesia, Taiwan Thrips hawaiiensis AAZ8516 This study 1 - - 14.7 China, India, Pakistan BOLD 13 0.50 2.75 4.3 Thrips palmi AAN2747 This study 1 - - 17.1 India, Pakistan BOLD 112 0.20 0.97 5.7 a NN = distance to nearest neighbor b One specimen in this BIN on BOLD was identified as Haplothrips ceylonicus c Two specimens in this BIN on BOLD were identified as Pseudodendrothrips bhatti

176

Table 5-4. Collection and curation information for thrips specimens collected from lablab bean (L. purpureus) in Bangladesh and their associated coxI sequences.

Life stage Date BOLD GenBank Classified as BIN Location /sex collected Sequence ID Accession no. Phlaeothripidae sp. 1 ACI5935 AF Jessore 2/4/15 LLBT048-16 KX233565 Haplothrips sp. 1 ACZ7039 AF Gazipur 1/14/15 LLBT002-16 KX233529 Haplothrips sp. 2 ACY8730 AF Gazipur 1/14/15 LLBT001-16 KX233530 Haplothrips sp. 2 ACY8730 AF Gazipur 1/14/15 LLBT049-16 KX233528 Haplothrips sp. 2 ACY8730 AF Gazipur 1/14/15 LLBT050-16 KX233526 Haplothrips sp. 2 ACY8730 L Gazipur 1/14/15 LLBT051-16 KX233564 Haplothrips sp. 2 ACY8730 AF Gazipur 1/14/15 LLBT054-16 KX233527 Haplothrips sp. 2 ACY8730 L Gazipur 1/14/15 LLBT061-16 KX233523 Haplothrips sp. 2 ACY8730 AF Gazipur 1/14/15 LLBT067-16 KX233524 Frankliniella intonsa AAF6737 AF Gazipur 2/24/15 LLBT047-16 KX233516 Frankliniella intonsa AAF6737 AF Gazipur 2/24/15 LLBT010-16 KX233514 Frankliniella intonsa AAF6737 AM Gazipur 2/24/15 LLBT013-16 KX233515 Frankliniella intonsa AAF6737 AF Gazipur 2/24/15 LLBT014-16 KX233510 Frankliniella intonsa AAF6737 AF Gazipur 3/9/15 LLBT024-16 KX233521 Frankliniella intonsa AAF6737 AF Gazipur 3/9/15 LLBT025-16 KX233522 Frankliniella intonsa AAF6737 AF Gazipur 3/9/15 LLBT029-16 KX233511 Frankliniella intonsa AAF6737 AF Gazipur 3/9/15 LLBT030-16 KX233513 Frankliniella intonsa AAF6737 AF Jessore 2/4/15 LLBT042-16 KX233509 Frankliniella intonsa AAF6737 AF Jessore 2/4/15 LLBT043-16 KX233520 Frankliniella intonsa AAF6737 AF Gazipur 1/14/15 LLBT044-16 KX233519 Frankliniella intonsa AAF6737 AF Gazipur 1/14/15 LLBT045-16 KX233518 Frankliniella intonsa AAF6737 AF Gazipur 2/24/15 LLBT046-16 KX233517 Frankliniella intonsa AAF6737 AF Gazipur 2/24/15 LLBT056-16 KX233512 Megalurothrips distalis/peculiaris AAN6623 AF Gazipur 3/4/15 LLBT031-16 KX233541 Megalurothrips distalis/peculiaris AAN6623 AF Jessore 2/4/15 LLBT039-16 KX233540 Megalurothrips distalis/peculiaris AAN6623 L Jessore 2/4/15 LLBT040-16 KX233539

177

Life stage Date BOLD GenBank Classified as BIN Location /sex collected Sequence ID Accession no. Megalurothrips distalis/peculiaris AAN6623 L Gazipur 1/14/15 LLBT058-16 KX233538 Megalurothrips distalis/peculiaris AAN6623 L Gazipur 1/14/15 LLBT059-16 KX233537 Megalurothrips distalis/peculiaris AAN6623 L Gazipur 1/14/15 LLBT060-16 KX233536 Megalurothrips distalis/peculiaris AAN6623 AM Gazipur 1/14/15 LLBT062-16 KX233535 Megalurothrips distalis/peculiaris AAN6623 AM Gazipur 1/14/15 LLBT066-16 KX233534 Megalurothrips distalis/peculiaris AAN6623 AF Gazipur 12/17/14 LLBT068-16 KX233533 Megalurothrips sp. AAM8053 AF Gazipur 2/24/15 LLBT009-16 KX233532 Megalurothrips sp. AAM8053 AF Gazipur 2/24/15 LLBT016-16 KX233531 Megalurothrips usitatus AAM8053 AF Gazipur 2/24/15 LLBT003-16 KX233554 Megalurothrips usitatus AAM8053 AF Gazipur 1/14/15 LLBT004-16 KX233553 Megalurothrips usitatus AAM8053 AF Gazipur 12/17/14 LLBT005-16 KX233552 Megalurothrips usitatus AAM8053 L Gazipur 3/4/15 LLBT006-16 KX233551 Megalurothrips usitatus AAM8053 L Gazipur 3/4/15 LLBT007-16 KX233563 Megalurothrips usitatus AAM8053 AF Gazipur 2/24/15 LLBT008-16 KX233562 Megalurothrips usitatus AAM8053 AF Gazipur 2/24/15 LLBT011-16 KX233561 Megalurothrips usitatus AAM8053 AF Gazipur 2/24/15 LLBT012-16 KX233560 Megalurothrips usitatus AAM8053 L Gazipur 2/24/15 LLBT015-16 KX233558 Megalurothrips usitatus AAM8053 AF Gazipur 3/9/15 LLBT021-16 KX233550 Megalurothrips usitatus AAM8053 AF Gazipur 3/9/15 LLBT022-16 KX233549 Megalurothrips usitatus AAM8053 AF Gazipur 3/9/15 LLBT026-16 KX233545 Megalurothrips usitatus AAM8053 L Gazipur 3/9/15 LLBT027-16 KX233544 Megalurothrips usitatus AAM8053 AF Gazipur 3/9/15 LLBT028-16 KX233543 Megalurothrips usitatus AAM8053 L Gazipur 3/4/15 LLBT032-16 KX233556 Megalurothrips usitatus AAM8053 L Gazipur 3/4/15 LLBT033-16 KX233546 Megalurothrips usitatus AAM8053 L Gazipur 3/4/15 LLBT034-16 KX233547 Megalurothrips usitatus AAM8053 AF Jessore 2/4/15 LLBT035-16 KX233548 Megalurothrips usitatus AAM8053 AF Jessore 2/4/15 LLBT041-16 KX233559 Megalurothrips usitatus AAM8053 L Gazipur 1/14/15 LLBT053-16 KX233555

178

Life stage Date BOLD GenBank Classified as BIN Location /sex collected Sequence ID Accession no. Megalurothrips usitatus AAM8053 AF Gazipur 1/14/15 LLBT055-16 KX233542 Thrips palmi AAN2747 AF Jessore 2/4/15 LLBT038-16 KX233569 Thrips hawaiiensis AAZ8516 AF Jessore 2/4/15 LLBT037-16 KX233568

179

Figures

Figure 5-1. Adult and larval bean thrips (Megalurothrips spp.) infesting the flowers (A), pods (B), and leaves (C) of lablab bean in a field experiment in Gazipur, Bangladesh

180

15 A 100 B Neem OP Untreated 75 10

50

5 25 Inflorescences/plot Thrips/inflorescence

0 0 12/16 12/23 12/30 1/6 1/21 1/28 2/12 2/25 3/4

Figure 5-2. Thrips abundance (A) and flowering activity (B) in response to insecticide treatments in Experiment 1 (mean ± standard error). OP = organophosphate.

181

10.0 A A 7.5

5.0

A 2.5 A AB B B Caterpillar damage (% pods) 0.0 Untreated Neem Spin/Neem Spin THX OP

12 B

9

6

3

Aphid infestation (% inflorescences) Aphid infestation Untreated Neem Spin/Neem Spin THX OP

25 C A A A

20 AB

15 B

10 C

5 Thrips infestation (adults/blossom) Thrips infestation Untreated Neem Spin/Neem Spin THX OP

40 D A AB 30

AB 20 B AB AB

Mean inflorescences 10

Untreated Neem Spin/Neem Spin THX OP

4.5 E

4.0

3.5

3.0

Marketable yield (kg) Marketable 2.5

Untreated Neem Spin/Neem Spin THX OP

182

Figure 5-3. Pod-borer damage (A), aphid infestation (B), thrips abundance (C), mean inflorescences (D), and marketable yield (E) after insecticide treatments in Experiment 2. Aphid infestation and pod damage were logit-transformed for analysis, but untransformed means are shown here. Means (± standard error) reflect seasonal totals (yield, pod-borer damage), seasonal means (inflorescences), or a cumulative measure of seasonal abundance (aphids, thrips; see Materials & Methods for details). Based on Tukey’s HSD test, letters above means indicate significant differences at α = 0.05.

183

A. B. 5 5 ● ● ● ● ● ● ● 4 4 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● 3 ● 3 ● ● ● ● ●● ● ● ● 2 2

1 1 Lablab bean yield (kg) Lablab bean yield (kg) Lablab ●

0 0

5 10 15 20 25 0 2 4 6 8 10 Thrips per blossom Pod−borer damage (%)

Figure 5-4. Partial regression plots for relationships between pest abundance and lablab bean yield. The fitted model was used to generate the slopes (dotted lines) and their 95% confidence bands (shown in gray). Lablab bean yield was negatively related to thrips abundance (A) and pod- borer damage (B), but only marginally related to aphid abundance (not shown).

184

LLBT026-16|Megalurothrips_usitatus LLBT006-16|Thripidae_larva LLBT027-16|Thripidae_larva LLBT016-16|Megalurothrips_sp LLBT041-16|Megalurothrips_usitatus LLBT034-16|Thripidae_larva LLBT015-16|Thripidae_larva LLBT004-16|Megalurothrips_usitatus LLBT055-16|Megalurothrips_usitatus LLBT012-16|Megalurothrips_usitatus 99 LLBT022-16|Megalurothrips_usitatus LLBT033-16|Thripidae_larva M. usitatus (AAM8053) LLBT028-16|Megalurothrips_usitatus 100 LLBT053-16|Thripidae_larva LLBT008-16|Megalurothrips_usitatus LLBT035-16|Megalurothrips_usitatus LLBT009-16|Megalurothrips_distalis LLBT005-16|Megalurothrips_usitatus 100 LLBT007-16|Thripidae_larva LLBT011-16|Megalurothrips_usitatus LLBT021-16|Megalurothrips_usitatus 86 LLBT003-16|Megalurothrips_usitatus MATHR015-10|Megalurothrips_usitatus|AAM8053 LLBT032-16|Thripidae_larva 100 LLBT066-16|Megalurothrips_distalis LLBT062-16|Megalurothrips_distalis 71 LLBT040-16|Thripidae_larva LLBT031-16|Megalurothrips_sp. LLBT039-16|Megalurothrips_distalis LLBT060-16|Thripidae_larva M. distalis/peculiaris (AAN6623) LLBT059-16|Thripidae_larva LLBT058-16|Thripidae_larva LLBT068-16|Megalurothrips_distalis 81 MATHR027-10|Megalurothrips_peculiaris|AAN6623 GBMIN30301-13|Megalurothrips_distalis|AAN6623 100 100 MATHR013-10|Thrips_palmi|AAN2747 T. palmi (AAN2747) 86 LLBT038-16|Thrips_palmi 77 MATHR178-10|Thrips_palmi|AAE7913 RRINV150-15|Thrips_palmi|AAD4600 100 74 MATHR315-11|Thrips_hawaiiensis|AAZ8516 LLBT037-16|Thrips_hawaiiensis T. hawaiiensis (AAZ8516) LLBT030-16|Frankliniella_intonsa 100 GBMIN39469-13|Frankliniella_intonsa|AAF6737 GBMH2773-07|Frankliniella_intonsa|n/a LLBT025-16|Frankliniella_intonsa LLBT014-16|Frankliniella_intonsa LLBT046-16|Frankliniella_intonsa LLBT044-16|Frankliniella_intonsa LLBT043-16|Frankliniella_intonsa F. intonsa (AAF6737) LLBT042-16|Frankliniella_intonsa LLBT010-16|Frankliniella_intonsa LLBT056-16|Frankliniella_intonsa LLBT013-16|Frankliniella_intonsa LLBT029-16|Frankliniella_intonsa LLBT047-16|Frankliniella_intonsa LLBT045-16|Frankliniella_intonsa LLBT024-16|Frankliniella_intonsa 100 LLBT048-16|Phlaeothripidae Phlaeothripidae sp. 1 (ACI935) 96 GBMIN12475-13|Dyothrips_pallescens|AAY4258 87 MATHR050-10|Haplothrips_gowdeyi|AAN5798 MATHR059-10|Haplothrips_ganglbaueri|ACF1370 95 MATHR454-12|Ananthakrishnana_euphorbiae|ACA2783 MATHR051-10|Haplothrips_tenuipennis|AAN4488 99 LLBT002-16|Haplothrips_sp. Haplothrips sp. 1 (ACZ7039) 100 LLBT001-16|Haplothrips_sp. LLBT054-16|Haplothrips_sp. LLBT061-16|Phlaeothripidae_larva LLBT067-16|Haplothrips_sp. Haplothrips sp. 2 (ACY8730) LLBT049-16|Haplothrips_sp. LLBT051-16|Phlaeothripidae_larva LLBT050-16|Haplothrips_sp.

0.02

185

Figure 5-5. Neighbor-joining tree based on the 5’ region of the mitochondrial cytochrome oxidase I gene of thrips samples collected from lablab bean plants in Bangladesh. Each specimen is labeled with its sequence number in the Barcode of Life Data System (BOLD). Reference specimens are in gray. Larval samples are in purple. One sample for which the morphological and molecular identifications conflict is in orange. Clades are labeled with Barcode Index Numbers (BINs), which have been proposed to be molecular taxonomic units similar to species. The tree is based on p-distances and was built with node support assessed with 1000 bootstrap iterations in Mega (version 6.06). Bootstrap values less than 70% are not shown.

186

GMBCM2513-15|Thysanoptera|Bangladesh LLBT027-16|Thripidae_larva LLBT026-16|Megalurothrips_usitatus LLBT055-16|Megalurothrips_usitatus LLBT012-16|Megalurothrips_usitatus LLBT033-16|Thripidae_larva LLBT022-16|Megalurothrips_usitatus LLBT028-16|Megalurothrips_usitatus LLBT006-16|Thripidae_larva LLBT016-16|Megalurothrips_sp LLBT004-16|Megalurothrips_usitatus MATHR084-10|Megalurothrips_usitatus|Pakistan MATHR099-10|Megalurothrips_usitatus|Pakistan LLBT041-16|Megalurothrips_usitatus LLBT034-16|Thripidae_larva LLBT015-16|Thripidae_larva GBMHT252-13|Megalurothrips_usitatus|India GBMHT253-13|Megalurothrips_usitatus|India 88 GBMHT254-13|Megalurothrips_usitatus|India RFTHY005-10|Thysanoptera|COI-5P|Australia GBMHT169-13|Megalurothrips|Indonesia SYC7868-14|Thysanoptera|French_Polynesia SYC7143-14|Thysanoptera|French_Polynesia SYC7728-14|Thysanoptera|French_Polynesia SYC7867-14|Thysanoptera|French_Polynesia 99 MATHR141-10|Megalurothrips_usitatus|Pakistan LLBT008-16|Megalurothrips_usitatus LLBT053-16|Thripidae_larva MATHR015-10|Megalurothrips_usitatus|Pakistan 79 LLBT003-16|Megalurothrips_usitatus LLBT021-16|Megalurothrips_usitatus LLBT032-16|Thripidae_larva GBMHT094-10|Pseudodendrothrips_bhattii|India LLBT011-16|Megalurothrips_usitatus LLBT007-16|Thripidae_larva LLBT005-16|Megalurothrips_usitatus GMBCM2507-15|Thysanoptera|Bangladesh LLBT009-16|Megalurothrips_distalis MATHR114-10|Megalurothrips_usitatus|Pakistan LLBT035-16|Megalurothrips_usitatus 99 LLBT066-16|Megalurothrips_distalis MATHR108-10|Megalurothrips_peculiaris|Pakistan GMBCA5993-15|Thysanoptera|Bangladesh GMBCB3700-15|Thysanoptera|Bangladesh MATHR112-10|Megalurothrips_peculiaris|Pakistan LLBT062-16|Megalurothrips_distalis GBMIN30300-13|Megalurothrips_distalis|China GBMIN30379-13|Megalurothrips_distalis|China GBMIN30301-13|Megalurothrips_distalis|China GBMIN30382-13|Megalurothrips_distalis|China GBMIN39690-13|Megalurothrips_sp._2_YF-2010|Unknown GBMIN30302-13|Megalurothrips_distalis|China GBMIN30380-13|Megalurothrips_distalis|China GBMIN30381-13|Megalurothrips_distalis|China LLBT040-16|Thripidae_larva LLBT031-16|Megalurothrips_sp. GMBCA5992-15|Thysanoptera|Bangladesh MATHR101-10|Megalurothrips_peculiaris|Pakistan LLBT039-16|Megalurothrips_distalis LLBT060-16|Thripidae_larva LLBT059-16|Thripidae_larva LLBT058-16|Thripidae_larva LLBT068-16|Megalurothrips_distalis GBMIN39682-13|Megalurothrips_sp._1_YF-2010|Unknown MATHR106-10|Megalurothrips_peculiaris|Pakistan MATHR102-10|Megalurothrips_peculiaris|Pakistan MATHR035-10|Megalurothrips_peculiaris|Pakistan MATHR034-10|Megalurothrips_peculiaris|Pakistan MATHR100-10|Megalurothrips_peculiaris|Pakistan MATHR105-10|Megalurothrips_peculiaris|Pakistan GBA8578-12|Megalurothrips_distalis|China MATHR027-10|Megalurothrips_peculiaris|Pakistan GBMHT481-15|Megalurothrips_distalis|Kenya 99 GMKMI576-15|Thysanoptera|Kenya

0.02 Figure 5-6. Neighbor-joining tree based on the 5’ region of the mitochondrial cytochrome oxidase I gene of Megalurothrips and related samples (BINs AAM8053, AAN6623, ACS4755) from the present study and the Barcode of Life Database (BOLD). Each specimen is labeled with its BOLD sequence number; the specimens from the current study start with “LLBT”. Colors correspond to morphological identifications. The tree is based on p-distances and was built with node support assessed with 1000 bootstrap iterations in Mega (version 6.06). Bootstrap values less than 70% are not shown.

187

Conclusions

The major findings of this dissertation are as follows:

• The use of seed-applied neonicotinoids in field crops rose rapidly over the past

two decades, and these products now constitute a major portion of overall

insecticide use in field crops.

• Neonicotinoid seed treatments are often used as ‘insurance’ against

unmeasured pest populations, rather than as part of an IPM framework driven

by pest monitoring or prediction.

• Neonicotinoid seed treatments in corn and soybeans travel from treated plants

to various soil invertebrates, in concentrations that can poison insect predators

that eat them.

• Under some conditions, seed-applied neonicotinoids can disrupt biological

control of slugs, enhancing slug populations and their damage to plants.

• In no-till field crop production in Pennsylvania, neonicotinoid seed treatments

do not reliably increase yield, and can decrease yield by fostering slugs.

• In lablab bean production in Bangladesh, more research is necessary to assess

whether and how biopesticides might contribute to an overall IPM strategy.

Spinosad is effective against pod-boring caterpillars (Maruca vitrata) and

could displace broad-spectrum insecticides for this pest.

• Flower-inhabiting thrips in lablab bean are dominated by Megalurothrips spp.,

and will likely need to be addressed in an IPM program for this crop.

188 Future research directions

Throughout the studies in this dissertation I have often treated ‘potential slug predators’ as a group because of the limited evidence concerning which predatory taxa prey on slugs in North America. More broadly, the diets of many predatory arthropods in agricultural systems are incompletely known, or inferred from laboratory feeding trials that bear little resemblance to field conditions. A typical crop field in Pennsylvania can host more than 45 species of ground beetles (Leslie et al. 2010), further compounding the challenge of untangling trophic relationships. The advent of molecular methods to elucidate the diets of field-collected predators by identifying prey DNA or proteins in the gut represents a major advance in researchers’ ability to determine the diet-composition of predatory species (Sheppard & Harwood 2005). Future studies could use molecular methods to describe predator-prey networks in agricultural systems and how management choices (such as insecticide use) alter these networks, analogous to the way researchers have examined the influence of various anthropogenic factors on plant-pollinator networks (e.g. Lopezaraiza-Mikel et al. 2007). A better understanding of predator diet composition would also aid the rational design of conservation biological control programs.

In my meta-analysis, I found that seed-applied neonicotinoids decreased abundance of predatory arthropods by an average of 16%, but the real-world significance of this reduction is uncertain for several reasons. First, while there are some aggregate estimates of the value of biological control over large scales (e.g. Losey & Vaughan

2006), there are few if any tools to estimate the value of biological control in particular crops at a field-scale relevant to farm management (Naranjo et al. 2015). The underlying

189 functions that would be necessary to build such a tool (relationships between predator abundance, pest abundance, and crop yield) are not available in most cases. Describing these relationships, their underlying variability, and their generalizability would be a fruitful area for future research. Second, a 16% reduction at the field level is likely to have different long-term consequences in landscapes with varying composition. The influence of neonicotinoids on long-term population trends of natural enemies is likely to be more pronounced in a state like Iowa, where corn and soybeans make up more than

60% of the landscape, than in a state like Pennsylvania, where these crops account for less than 7% and semi-natural areas make up much of the rest (USDA NASS 2016).

More generally, in the U.S. we lack long-term datasets with standardized methodology that describe population trends for natural enemies. Such datasets, built through standardized sampling across a network of research sites throughout major crop production regions, would provide a reliable indicator of whether our crop management practices, in aggregate, are conserving natural enemies of crop pests.

Understanding the long-term consequences of insecticide use in agricultural systems should explore not only the influence of insecticides on predator populations, but also the way these products alter selection pressure on pests and their predators. I have established that at least slugs, earthworms, and black cutworms can contain biologically meaningful concentrations of neonicotinoids from seed applications and it seems unlikely that these are the only soil-associated organisms in which neonicotinoids can be found. It stands to reason that various prey species will vary in the insecticide concentrations they contain, based on differences in how and where they encounter insecticide residues in the environment and how quickly they excrete these residues. In turn, consistent patterns in

190 the concentration of insecticides in various types of prey could select for populations or communities of predators that avoid prey with high insecticide concentrations, or even shift from prey to non-prey foods (e.g. weed seeds). Variation in insecticide concentrations within prey populations also raises the possibility that prey could be selected to repurpose insecticide residues as a defense, similar to the way many herbivorous insects in nature have evolved to tolerate and sequester plant defenses. This line of reasoning is admittedly speculative, but in my opinion deserves to be explored.

In my three site-years of field experiments, the influence of seed-applied neonicotinoids on crop yield was variable. In soybeans, these insecticides were associated with a 5% decrease in yield, while in two years of corn experiments they did not significantly affect yield. These results are broadly consistent with field studies in other locations, which have found the yield response to seed-applied neonicotinoids to be generally small on average, and variable across sites and years. It would be valuable to move beyond documenting this variability, and striving to understand its sources to enable better prediction of the characteristics of fields that would benefit from neonicotinoid seed treatment. A good example of this approach is a study on corn production in Virginia, which found that fall populations of white grubs could predict the likelihood of yield benefit from neonicotinoid seed treatment the following season

(Jordan et al. 2012). Similar efforts in each production region, focused on the most relevant pest groups, could be used to create farmer-friendly tools to predict neonicotinoid value (or cost). Such tools could reduce the ubiquity of neonicotinoid seed treatments and the concomitant risk of non-target effects.

191

Appendix A

Supplemental details on methods used in experiments on the effects of neonicotinoid seed treatments on a soybean-slug-beetle food chain

Laboratory experiment

Soybean-slug and slug-ground beetle bioassays

In the soybean-slug bioassays, we lined the containers (16-oz Reynolds Del-Pak

®) with 2.5 cm of sifted, moist potting soil (Premier ® Pro-Mix ® BX). The four soybean seeds were evenly spaced. During daily observations, slug status was scored as alive or dead, and seedlings were scored as undamaged, damaged, or killed.

In the slug-ground beetle bioassays, the sides of the containers were coated with

Fluon (BioQuip Products, Inc., Rancho Dominguez, CA) to keep slugs in the arena where beetles could attack them (Symondson 1993). Because these lab experiments depended on slugs generated in the soybean-slug study, they were similarly blocked into three trials. We observed interactions in containers under low light conditions for the first 3.5 hours after beetles were introduced, recording the status of slugs and beetles every 15 to

30 minutes. During daily observations, we classified slugs as alive or dead, and inspected dead slugs under a microscope to confirm predation. We also recorded beetle mortality and took qualitative notes on beetle behavior that could be associated with poisoning

(altered gait, partial paralysis, trembling, etc.).

192 Field experiment

Study site and crop management

For both treatments, we used soybean variety HS31A03 (GROWMARK, Inc.,

Bloomington, IL), which had been inoculated with Bradyrhizobium japonicum (Optimize

®, Novozymes Bio Ag Inc., Brookfield, WI). Experimental plots were established in a field that had been farmed using no-till practices for seven years, and had not received insecticide applications (including seed treatments) for at least one year. The planter did not use pressurized air to move seeds and so was unlikely to expel insecticide-ladened dust (as in Krupke et al. 2012). The plot layout is below, with gray squares indicating plots planted with thiamethoxam and fungicide-treated seed, and white squares indicating plots planted with untreated seed.

Invertebrate activity-density

Each pitfall trap comprised a 16-oz plastic container (Reynolds Del Pak ®) sunk into the ground, with an identical container nested inside the first so that it could be emptied without disturbing the surrounding soil. A white plastic plate supported by nails served as a trap cover and the killing agent was 50% propylene glycol with unscented dish soap (a few drops/gallon) to break surface tension. The first sample occurred roughly three weeks after planting and sampling continued monthly for the rest of the growing

193 season (sample dates: June 9th, July 13th, August 21st, September 24th). Upon collection, we strained samples (1-mm mesh) and stored them in 80% ethanol. We identified slugs to genus (some characters cannot be assessed on preserved specimens; Chichester & Getz

1973; McDonnell et al. 2009) and most predators to family or order (Triplehorn &

Johnson 2005; Ubick et al. 2005).

Predation

We pinned caterpillars to a small piece of modeling clay that we buried so that the caterpillar rested on the soil surface, excluding vertebrates with a cylindrical, hardware cloth cage (9.5-cm tall, 11.5-cm diameter, mesh size: 1.3 cm) topped with a plastic lid.

On each sample date, we assessed both diurnal and nocturnal predator activity: one 12-h sample started at 8:30, and the other at 20:30. After the first 12 h, we replaced any attacked, missing, or compromised caterpillars. We recorded caterpillars as whole and alive, partially eaten, or missing. Dead caterpillars showing no sign of predation were excluded from analyses.

Insecticide analyses

In the laboratory, we assigned juvenile gray garden slugs (0.14 ± 0.07 g) singly to microcosms planted with low-rate or high-rate thiamethoxam-treated soybeans (n =

48/treatment) and allowed them to feed for one week as previously described. After being re-weighed, we randomly selected a subset of slugs from each treatment to feed to

C.tricolor (n = 20-23), while we selected others for insecticide testing (n = 7-11). We left

194 beetles overnight to feed upon slugs and in the morning, we collected all beetles from each treatment and assessed them for symptoms of toxicity. Those that had fed upon slug tissue were pooled into a single insecticide sample (n = 18-19 beetles/sample). For each plant sample, we randomly selected two containers and harvested the above-ground portion of the seedlings for analysis (n = 7-8 seedlings/sample). We thoroughly mixed the soil from those same randomly selected containers and took a roughly 45-ml subsample of soil for analysis.

In the field, when soybeans were at the cotyledon stage, we generated each soil and soybean sample by pooling material from four random locations in each plot. At each location, we cut two soybean seedlings at the soil surface, for a total of eight seedlings per plot. After harvesting the seedlings, we centered a golf-cup cutter over the row and took a soil core ~10 cm deep and 10.8 cm in diameter. We combined the soil from the four locations into a bin where we broke up the cores and mixed them thoroughly, then took a roughly 50-ml subsample and put aside half to measure soil moisture. In the course of our soil sampling at the cotyledon stage, we also found and collected several earthworms. We collected slugs for insecticide samples when soybeans were at the cotyledon and one-leaf stages. While slugs were active (21:15-22:45), we collected ≥ 15 slugs per plot directly from soybean plants, collecting slugs from at least four distinct locations within each plot. To see whether insecticide residues persisted in slugs over the season, we also collected slugs shortly before harvest (4/plot, from refuge traps because soybeans had senesced). Throughout our field sampling, we used dedicated sampling equipment for each treatment to avoid cross-contamination.

195 We did not collect ground beetles or other predators for neonicotinoid testing for several reasons. For one thing the species contributing to slug predation in North America are not well described, so it would have been difficult to choose focal taxa. Also, to achieve adequate sensitivity we would have needed to collect a large number of predators, which ruled out hand-collection. We could have collected predators using pitfall traps, but this would complicate interpretation because insecticide exposure could make beetles more or less likely to be trapped. We therefore limited our field sampling to soil, plants, and slugs and relied on the insect toxicology literature to interpret concentrations of neonicotinoids in slugs relative to their risk to predators.

Limits of detection (LOD) for our analysis by LC/MS-MS ranged from 1 to 19 ppb for the primary neonicotinoids thiamethoxam, clothianidin (also a thiamethoxam metabolite), and imidacloprid, and 63 to 937 (most ≤ 188 ppb) for other neonicotinoid metabolites. Differences in LOD arise from differences in sample mass and in the sensitivity of the technique for different analytes.

196

Appendix B

Supplemental results from experiments on the effects of neonicotinoid seed treatments on a soybean-slug-beetle food chain

To test the generality of our results in the soybean-slug bioassay, we conducted a similar experiment with smaller juvenile slugs.

Methods

Bioassays were similar to those described in the main text (see Soybean-slug bioassays), except that the slug treatment comprised four smaller juvenile slugs (0.045 ±

0.015 [SD] g) per container rather than a single, medium-sized juvenile slug. The experiment was again a four by two factorial design with the four types of seed treatments crossed with presence or absence of slugs (n = 8/seed-slug treatment). Data collection was as described for the medium-slug experiment.

Analysis

Slug survival and mass change (as % of starting mass) did not meet parametric assumptions, so we assessed the influence of seed treatment on these outcomes by

Kruskal-Wallis analysis of variance (ANOVA). To examine the influence of seed treatments on patterns of slug damage to plants over time, we performed mixed-effects repeated measures analyses with an AR1 model for the covariance.

197 Results

Slugs readily attacked soybean seedlings grown from each of the four seed treatments, with no significant differences in the number of seedlings damaged over time

(Seed treatment: F3,28 = 0.99, P = 0.41; Day*Seed treatment: F18,168= 0.85, P = 0.64).

Survival of small slugs was > 90% in each treatment, and statistically similar across all treatments (Kruskal-Wallis H = 6.46, d.f. = 3, P = 0.09). Although the P-value for this test was marginal, there was no discernible pattern in slug survival across treatments

(F+H > F > U > F+L). Slug body mass increased by 35% on average over the study, with no significant differences in slug mass change by seed treatment (Kruskal-Wallis H =

0.89: d.f. = 3, P = 0.83). Overall, seed treatment did not influence slug herbivory, survival, growth, or behavior.

198

Appendix C

Supplemental results from a meta-analysis of the effect of seed-applied neonicotinoids on natural enemies

Table C-1. Studies included in our meta-analysis of neonicotinoid seed treatment effects on natural enemies (full references at end).

Early Peer Functional Arth. N Plot size Crop Study abbrev. A.I.(s) Alt. insect. Habitat samp. review group(s) class plots (m2) (%) Barley SoteloCardona2010 THX M, Pr, Pa F A, I 25 16 112 Canola Echegaray2009 IMI OP, OC, O, Pr S I 20 3 - 4 100 - 752 PYR Cotton/ Krauter2001 IMI Pr F A, I 0 3 110000 Sorghum Maize Ahmad20052006 X CLO PYR M, O, Pr F, S A, I 0 - 33 4 232 Albajes2003 X IMI M, Pr F, S A, I 20 4 7000 AlDeeb2003 X IMI, CLO OP, PYR M, O, Pr F, S A, I 0 - 33 4 23 - 28 Babendreier2015 X CLO BIO, PYR M, Pr S A, I 0 7 - 61 1 Bhatti2005 X IMI Bt, PYR M, O, Pr, Pa F, S A, Ch, I 20 - 33 4 335 delaPoza2005 X IMI Bt M, Pr F, S A, Ch, I 0 - 20 3 - 4 1833 - 5500 Farinos2008 Harmon2006 CLO M, O, Pr, Pa F, S A, Ch, I 66 - 100 4 729 - 1080 Soybean Douglas2015 X THX M, O, Pr, Pa S A, I 25 6 1080 Hallett2014 X IMI, THX Pr, Pa F A, I 20 - 25 3 111 HeidelBaker2012 THX PYR Pr, Pa F I 0 - 10 4 1003 Carter2013 Ohnesorg2009 X IMI, THX PYR, TRI M, Pr, Pa F A, I 14 6 150 Seagraves2012 X IMI, THX F A, I 13 4 41

199

Early Peer Functional Arth. N Plot size Crop Study abbrev. A.I.(s) Alt. insect. Habitat samp. review group(s) class plots (m2) (%) Soybean Spigler2013 THX Pr F I 65 4 409 Tinsley2012 X THX Pr F I 0 4 70 Sugar beet Baker2002 IMI, CLO, THX PYR M S A, I 33 4 108 Sunflower Charlet et al. 2007 X THX CARB Pa F I 0 4 24 Wheat Schmidt et al. 2002 IMI M S I 0 4 10000 Key to abbreviations: A.I. = active ingredient, Alt. insect. = alternative insecticide compared to neonicotinoid seed treatment, Early samp. (%) = the percentage of samples taken during the first 40 days of crop growth, IMI = imidacloprid, CLO = clothianidin, THX = thiamethoxam, BIO = biopesticide, Bt = transgenic Bt crop, CARB = carbamate, OC = organochlorines, OP = organophosphate, PYR = pyrethroid, M = mixed, O = omnivore, Pr = predator, Pa = parasitoid, F = foliar/aboveground, S = soil/belowground, I = insect, A = arachnid, Ch = chilopod

200

study Ahmad20052006 Albajes2003 300 AlDeeb2003 Babendreier2015 Baker2002 Bhatti2005 Charlet2007 delaPoza2005Farinos2008 Douglas2015 200 Echegaray2009 Hallett2014 Harmon2006 HeidelBaker2012Carter2013 Krauter2001 Ohnesorg2009 100 Schmidt2002 Seagraves2012 SoteloCardona2010 Spigler2013 Tinsley2012

0

−5.0 −2.5 0.0 2.5 Effect size (d)

Figure C-1. Weighted histogram for the effect of seed-applied neonicotinoids on natural enemies (relative to no-insecticide controls), color-coded by study (n = 607 observations from 56 site-years and 20 studies).

201

100

study Ahmad20052006 AlDeeb2003 Babendreier2015 Baker2002 Bhatti2005 Echegaray2009 50 Hallett2014 Ohnesorg2009

0

−4 0 4 8 Effect size (d)

Figure C-2. Weighted histogram for the effect of seed-applied neonicotinoids on natural enemies (relative to pyrethroid controls), color-coded by study (n = 384 observations from 15 site-years and 8 studies).

202

vs. No insecticide Thrips: THX

Slugs: THX

Aphids: THX

IMI

Aphids: THX vs. PYR vs. Foliar insecticide

IMI vs. PYR

THX vs. PYM

IMI vs. PYM

IMI vs. Foliar IMI

THX vs. Foliar IMI

−150 −100 −50 0 50 100 150 200 Change in predator−prey ratio (%)

Figure C-3. Change in predator-prey ratio (PP) as a result of seed-applied neonicotinoids, relative to control plots treated with either no insecticide or a foliar insecticide (calculated as 100% X (PPNeonic - PPControl)/PPControl)). Each point represents a treatment comparison within a given study; negative values indicate that predator-prey ratios were lower in the neonicotinoid-treated plots versus controls, while positive values indicate the opposite. Points with the same color were derived from the same study. PYR = pyrethroid, PYM = pymetrozine, IMI = imidacloprid, and THX = thiamethoxam.

203

Works cited

Afifi M, Lee E, Lukens L, and Swanton C. 2014. Thiamethoxam as a seed treatment alters the

physiological response of maize (Zea mays) seedlings to neighbouring weeds. Pest

Management Science 71:505-514.

Agency for Toxic Substances & Disease Registry. 2002. Aldrin and Dieldrin - ToxFAQs.

Available at http://www.atsdr.cdc.gov/toxfaqs/tfacts1.pdf.

Ahmad A, Wilde GE, Whitworth RJ, and Zolnerowich G. 2006. Effect of corn hybrids expressing

the coleopteran-specific Cry3Bb1 protein for corn rootworm control on aboveground

insect predators. Journal of Economic Entomology 99:1085-1095.

Ahmad A, Wilde GE, and Zhu KY. 2005. Detectability of coleopteran-specific Cry3Bb1 protein

in soil and its effect on nontarget surface and below-ground arthropods. Environmental

Entomology 34:385-394.

Ahmed KU, Rahnam MM, Alam MZ, and Ahmed SU. 2004. Methods of pest control and direct

yield loss assessment of country bean (Dolichos lablab) at farmers' field condition: a

survey finding. Pakistan Journal of Biological Sciences 7:287-291.

Al-Deeb MA, and Wilde GE. 2003. Effect of Bt corn expressing the Cry3Bb1 toxin for corn

rootworm control on aboveground nontarget arthropods. Environmental Entomology

32:1164-1170.

Al-Deeb MA, Wilde GE, Blair JM, and Todd TC. 2003. Effect of Bt corn for corn rootworm

control on nontarget soil microarthropods and nematodes. Environmental Entomology

32:859-865.

Albajes R, Lopez C, and Pons X. 2003. Predatory fauna in cornfields and response to

imidacloprid seed treatment. Journal of Economic Entomology 96:1805-1813.

204 Ananthakrishnan TN. 1993. Bionomics of thrips. Annual Review of Entomology 38:71-92.

Atikullah SM, Hossain ABME, Miah GU, Rahman MM, and Khan SA. 2013. Homestead

vegetables in the south-central coastal region of Bangladesh: species richness, utilization

and conservation. 6th Annual International ESP Conference. Bali, Indonesia. p 18.

Ayre K. 2001. Effect of predator size and temperature on the predation of Deroceras reticulatum

(Müller) (Mollusca) by carabid beetles. Journal of Applied Entomology-Zeitschrift Fur

Angewandte Entomologie 125:389-395. 10.1046/j.1439-0418.2001.00568.x

Ayyangar GNR, and Nambiar KKK. 1935. Studies in Dolichos lablab (Roxb.) and (L.)- the

Indian field and garden bean. Proceedings of the Indian Academy of Sciences - Section B

1:857-867.

Babendreier D, Jeanneret P, Pilz C, and Toepfer S. 2015. Non-target effects of insecticides,

entomopathogenic fungi and nematodes applied against western corn rootworm larvae in

maize. Journal of Applied Entomology 139:457-467.

Bahlai CA, vander Werf W, O'Neal M, Hemerik L, and Landis DA. 2015. Shifts in dynamic

regime of an invasive lady beetle are linked to the invasion and insecticidal management

of its prey. Ecological Applications 25:1807-1818.

Bahlai CA, Xue Y, McCreary CM, Schaafsma AW, and Hallett RH. 2010. Choosing organic

pesticides over synthetic pesticides may not effectively mitigate environmental risk in

soybeans. PLoS ONE 5:e11250. 10.1371/journal.pone.0011250

Baker P, Haylock LA, Gamer BH, Sands RJN, and Dewar AM. 2002. The effects of insecticide

seed treatments on beneficial invertebrates in sugar beet. Farnham: British Crop

Protection Council.

Barbosa P, Saunders JA, Kemper J, Trumbule R, Olechno J, and Martinat P. 1986. Plant

allelochemicals and insect parasitoid effects of nicotine on Cotesia congregata (Say)

205 (Hymenoptera: Braconidae) and Hyposoter annulipes (Cresson) (Hymenoptera:

Ichneumonidae). Journal of Chemical Ecology 12:1319-1328.

Barker GM. 2004. Natural Enemies of Terrestrial Molluscs. Wallingford: CABI Publishing.

Bayer CropScience. 2010. Poncho ® 600 Insecticide Label. Monheim am Rhein, Germany.

Beck's Hybrids. 2014. Seed treatments. Available at

http://www.beckshybrids.com/Products/Seed-Treatments.

Benbrook CM. 2009. The magnitude and impacts of the biotech and organic seed price premium.

Washington, DC: The Organic Center.

Benbrook CM. 2012. Impacts of genetically engineered crops on pesticide use in the US - the first

sixteen years. Environmental Sciences Europe 24:24.

Bengtsson J, Ahnström J, and Weibull A-C. 2005. The effects of organic agriculture on

biodiversity and abundance: a meta-analysis. Journal of Applied Ecology 42:261-269.

Bhatti MA, Duan J, Head G, Jiang CJ, McKee MJ, Nickson TE, Pilcher CL, and Pilcher CD.

2005a. Field evaluation of the impact of corn rootworm (Coleoptera: Chrysomelidae)-

protected Bt corn on ground-dwelling invertebrates. Environmental Entomology 34:1325-

1335.

Bhatti MA, Duan J, Head GP, Jiang CJ, McKee MJ, Nickson TE, Pilcher CL, and Pilcher CD.

2005b. Field evaluation of the impact of corn rootworm (Coleoptera : Chrysomelidae)-

protected Bt corn on foliage-dwelling arthropods. Environmental Entomology 34:1336-

1345.

Bilde T, and Toft S. 1998. Quantifying food limitation of arthropod predators in the field.

Oecologia 115:54-58.

Blacquière T, Smagghe G, van Gestel CAM, and Mommaerts V. 2012. Neonicotinoids in bees: a

review on concentrations, side effects and risk assessment. Ecotoxicology 21:973-992.

206 Boethel DJ. 2004. Integrated management of soybean insects. In: Boerma HR, and Specht JE,

eds. Soybeans: improvement, production, and uses. Madison: American Society of

Agronomy/Crop Science Society of America/Soil Science Society of America, 853-881.

Bonmatin JM, Giorio C, Girolami V, Goulson D, Kreutzweiser DP, Krupke C, Liess M, Long E,

Marzaro M, Mitchell EA, Noome DA, Simon-Delso N, and Tapparo A. 2015.

Environmental fate and exposure; neonicotinoids and fipronil. Environmental Science

and Pollution Research 22:35-67.

Bradshaw JD, and Rice ME. 2003. Bean leaf beetles: a current and historical perspective.

Integrated Crop Management Newsletter: Iowa State University. p 20-21.

Bradshaw JD, Rice ME, and Hill JH. 2008. Evaluation of management strategies for bean leaf

beetles (Coleoptera : Chrysomelidae) and bean pod mottle virus (Comoviridae) in

soybean. Journal of Economic Entomology 101:1211-1227.

Brassard DW, and Yusuf I. 2002. Impact analysis of the seed treatment uses of lindane on wheat,

barley, oats, rye, corn, sorghum, and canola. Washington, D.C.: Environmental

Protection Agency, Biological and Economic Analysis Division.

Brust GE, Stinner BR, and McCartney DA. 1985. Tillage and soil insecticide effects on predator-

black cutworm (Lepidoptera: Noctuidiae) interactions in corn agroecosystems. Journal of

Economic Entomology 78:1389-1392.

Buckman RS, Mound LA, and Whiting MF. 2013. Phylogeny of thrips (Insecta: Thysanoptera)

based on five molecular loci. Systematic Entomology 38:123-133.

Budge GE, Garthwaite D, Crowe A, Boatman ND, Delaplane KS, Brown MA, Thygesen HH, and

Pietravalle S. 2015. Evidence for pollinator cost and farming benefits of neonicotinoid

seed coatings on oilseed rape. Scientific Reports 5:12574.

207 Byers RA, and Calvin DD. 1994. Economic injury levels to field corn from slug

(Stylommatophora: Agrolimacidae) feeding. Journal of Economic Entomology 87:1345-

1350.

Carson R. 1962. Silient Spring. Boston, MA: Houghton Mifflin Company.

Carter ME. 2013. Functional response and the effects of insecticidal seed treatment on the

soybean aphid parasitoid, Binodoxys communis.M.S. Thesis. University of Minnesota.

Chagnon M, Kreutzweiser D, Mitchell EA, Morrissey CA, Noome DA, and Van der Sluijs JP.

2015. Risks of large-scale use of systemic insecticides to ecosystem functioning and

services. Environmental Science and Pollution Research 22:119-134.

Charlet LD, Aiken RM, Meyer RF, and Gebre-Amlak A. 2007. Impact of combining planting

date and chemical control to reduce larval densities of stem-infesting pests of sunflower

in the central plains. Journal of Economic Entomology 100:1248-1257.

Chichester LF, and Getz LL. 1973. The terrestrial slugs of northeastern North America. Sterkiana

51:11-42.

Cloyd RA, and Bethke JA. 2011. Impact of neonicotinoid insecticides on natural enemies in

greenhouse and interiorscape environments. Pest Management Science 67:3-9.

Codling G, Al Naggar Y, Giesy JP, and Robertson AJ. 2016. Concentrations of neonicotinoid

insecticides in honey, pollen and honey bees (Apis mellifera L.) in central Saskatchewan,

Canada. Chemosphere 144:2321-2328.

Collins RA, and Cruickshank RH. 2013. The seven deadly sins of DNA barcoding. Molecular

Ecology Resources 13:969-975.

Cox WJ, and Cherney JH. 2011. Location, variety, and seeding rate interactions with soybean

seed-applied insecticide/fungicides. Agronomy Journal 103:1366-1371.

Cox WJ, Cherney JH, and Shields E. 2007a. Clothianidin seed treatments inconsistently affect

corn forage yield when following soybean. Agronomy Journal 99:543-548.

208 Cox WJ, Shields E, Cherney DJR, and Cherney JH. 2007b. Seed-applied insecticides

inconsistently affect corn forage in continuous corn. Agronomy Journal 99:1640-1644.

Croft BA, and Brown AWA. 1975. Responses of arthropod natural enemies to insecticides.

Annual Review of Entomology 20:285-335.

Croft BA, and Nelson EE. 1972. An index to predict efficient interactions of Typhlodromus

occidentalis in control of Tetranychus mcdanieli in Southern California apple trees.

Journal of Economic Entomology 65:310-312.

CropLife Foundation. 2013. The role of seed treatment in modern U.S. crop production: a review

of benefits. Washington, DC.

Cumming G. 2012. Understanding the new statistics: Effect sizes, confidence intervals, and meta-

analysis. New York, NY: Taylor & Francis Group.

Dalwadi MM, Korat DM, and Tank BD. 2008. Bio-efficacy of some botanical insecticides

against major insect pests of Indian bean, Lablab purpureus L. Karnataka Journal of

Agricultural Sciences 21:295-296.

Das BC, Sarker PK, and Rahman MM. 2008. Aphidicidal activity of some indigenous plant

extracts against bean aphid Aphis craccivora Koch (Homoptera: Aphididae). Journal of

Pest Science 81:153-159.

Dasgupta S. 2003. Bangladesh Pesticide Use (BPU) Ref. BGD_2003_BPU_v01_M. The World

Bank. Available at http://www.microdata.worldbank.org.

Dasgupta S, Meisner C, and Huq M. 2005. Health effects and pesticide perception as

determinants of pesticide use: evidence from Bangladesh. Washington, D.C.: World Bank

Policy Research Working Paper 3776. de la Poza M, Pons X, Farinós GP, López C, Ortego F, Eizaguirre M, Castanera P, and Albajes R.

2005. Impact of farm-scale Bt maize on abundance of predatory arthropods in Spain.

Crop Protection 24:677-684.

209 Donaldson JR, Myers SW, and Gratton C. 2007. Density-dependent responses of soybean aphid

(Aphis glycines Matsumura) populations to generalist predators in mid to late season

soybean fields. Biological Control 43:111-118.

Douglas MR. 2012. The influence of farming practices on slugs and their predators in reduced-

tillage field crops in Pennsylvania. M.S. Thesis. The Pennsylvania State University.

Douglas MR, Rohr JR, and Tooker JF. 2015. Neonicotinoid insecticide travels through a soil food

chain, disrupting biological control of non-target pests and decreasing soya bean yield.

Journal of Applied Ecology 52:250-260.

Douglas MR, and Tooker JF. 2012. Slug (Mollusca: Agriolimacidae, Arionidae) ecology and

management in no-till field crops, with an emphasis on the mid-Atlantic region. Journal

of Integrated Pest Management 3:1-9.

Douglas MR, and Tooker JF. 2015. Large-scale deployment of seed treatments has driven rapid

increase in use of neonicotinoid insecticides and preemptive pest management in US field

crops. Environmental Science & Technology 49:5088-5097.

Dow AgroSciences. 2014. Mycogen ® Seeds, Grain Corn. Available at

http://mycogen.com/Grain_Corn/Information/SitePages/Traits.aspx.

Echegaray FJ. 2009. Environmental effects of insecticides on non target predator and parasitoid

insects. M.S. Thesis. Royal Roads University.

Elbert A, Haas M, Springer B, Thielert W, and Nauen R. 2008. Applied aspects of neonicotinoid

uses in crop protection. Pest Management Science 64:1099-1105.

Eskelson MJ, Chapman EG, Archbold DD, Obrycki JJ, and Harwood JD. 2011. Molecular

identification of predation by carabid beetles on exotic and native slugs in a strawberry

agroecosystem. Biological Control 56:245-253.

Esker PD, and Conley SP. 2012. Probability of yield response and breaking even for soybean

seed treatments. Crop science 52:351-359.

210 European Commision. 2013. Commission implementing regulations (EU). 485/2013. Official

Journal of the European Union. p 12-26.

Farinós GP, de la Poza M, Hernández-Crespo P, Ortego F, and Castañera P. 2008. Diversity and

seasonal phenology of aboveground arthropods in conventional and transgenic maize

crops in Central Spain. Biological Control 44:362-371.

Fausti SW, McDonald TM, Lundgren JG, Li J, Keating AR, and Catangui M. 2012. Insecticide

use and crop selection in regions with high GM adoption rates. Renewable Agriculture

and Food Systems 27:295-304.

Fernandez-Cornejo J, Greene C, Jans S, Osteen CD, and Padgitt M. 2000. Pest management

practices. In: Anderson W, Magleby R, and Heimlich R, editors. Agricultural Resources

and Environmental Indicators: U. S. Department of Agriculture ERS. p 1-48.

Fleischer SJ, Orzolek MD, De Mackiewicz D, and Otjen L. 1998. Imidacloprid effects on

Acalymma vittata (Coleoptera : Chrysomelidae) and bacterial wilt in cantaloupe. Journal

of Economic Entomology 91:940-949.

Floyd R, Lima J, deWaard J, Humble L, and Hanner R. 2010. Common goals: policy implications

of DNA barcoding as a protocol for identification of arthropod pests. Biological

Invasions 12:2947-2954.

Fregin S, Haase M, Olsson U, and Alstrom P. 2012. Pitfalls in comparisons of genetic distances:

a case study of the avian family Acrocephalidae. Molecular Phylogenetics and Evolution

62:319-328.

Furlan L, and Kreutzweiser D. 2015. Alternatives to neonicotinoid insecticides for pest control:

case studies in agriculture and forestry. Environmental Science and Pollution Research

22:135-147.

211 Gaspar AP, Marburger DA, Mourtzinis S, and Conley SP. 2014. Soybean seed yield response to

multiple seed treatment components across diverse environments. Agronomy Journal

106:1955-1962.

Geiger F, Bengtsson J, Berendse F, Weisser WW, Emmerson M, Morales MB, Ceryngier P, Liira

J, Tscharntke T, Winqvist C, Eggers S, Bommarco R, Pärt T, Bretagnolle V, Plantegenest

M, Clement LW, Dennis C, Palmer C, Oñate JJ, Guerrero I, Hawro V, Aavik T, Thies C,

Flohre A, Hänke S, Fischer C, Goedhart PW, and Inchausti P. 2010. Persistent negative

effects of pesticides on biodiversity and biological control potential on European

farmland. Basic and Applied Ecology 11:97-105.

Giesler LJ, Ghabrial SA, Hunt TE, and Hill JH. 2002. Bean pod mottle virus: A threat to US

soybean production. Plant Disease 86:1280-1289.

Glare T, Caradus J, Gelernter W, Jackson T, Keyhani N, Köhl J, Marrone P, Morin L, and

Stewart A. 2012. Have biopesticides come of age? Trends in Biotechnology 30:250-258.

Glogoza P. 2005. Corn insects of North Dakota affecting planting decisions. Fargo, ND: North

Dakota State University Extension Service.

Glogoza P, McMullen M, Zollinger R, Thostenson A, DeJong T, Meyer W, Schauer N, and Olson

J. 2002. Pesticide use and pest management practices for major crops in North Dakota

2000.

Godfray HC, Blacquière T, Field LM, Hails RS, Petrokofsky G, Potts SG, Raine NE, Vanbergen

AJ, and McLean AR. 2014. A restatement of the natural science evidence base

concerning neonicotinoid insecticides and insect pollinators. Proceedings of the Royal

Society of London Series B, Biological Sciences 281:20140558.

Godfray HC, Blacquiere T, Field LM, Hails RS, Potts SG, Raine NE, Vanbergen AJ, and McLean

AR. 2015. A restatement of recent advances in the natural science evidence base

212 concerning neonicotinoid insecticides and insect pollinators. Proceedings of the Royal

Society of London Series B, Biological Sciences 282:20151821.

Goulson D. 2013. An overview of the environmental risks posed by neonicotinoid insecticides.

Journal of Applied Ecology 50:977-987.

Government of Ontario. 2015. Ontario Regulation 139/15 made under the Pesticides Act. The

Ontario Gazette: Ministry of Government Services. p 1402-1442.

Gray ME. 2011. Relevance of traditional integrated pest management (IPM) strategies for

commercial corn producers in a transgenic agroecosystem: A bygone era? Journal of

Agricultural and Food Chemistry 59:5852-5858.

Gray ME, Sappington TW, Miller NJ, Moeser J, and Bohn MO. 2009. Adaptation and

invasiveness of Western corn rootworm: intensifying research on a worsening pest.

Annual Review of Entomology 54:303-321.

Hadi BA, Bradshaw JD, Rice ME, and Hill JH. 2012. Bean leaf beetle (Coleoptera:

Chrysomelidae) and bean pod mottle virus in soybean: biology, ecology, and

management. Journal of Integrated Pest Management 3:1-7.

Hajibabaei M, Janzen DH, Burns JM, Hallwachs W, and Hebert PD. 2006. DNA barcodes

distinguish species of tropical Lepidoptera. Proceedings of the National Academy of

Sciences of the United States of America 103:968-971.

Hallett RH, Bahlai CA, Xue Y, and Schaafsma AW. 2014. Incorporating natural enemy units into

a dynamic action threshold for the soybean aphid, Aphis glycines (Homoptera:

Aphididae). Pest Management Science 70:879-888.

Hallmann CA, Foppen RP, van Turnhout CA, de Kroon H, and Jongejans E. 2014. Declines in

insectivorous birds are associated with high neonicotinoid concentrations. Nature

511:341-343.

213 Hammond RB. 2000. Simulation of moderate levels of slug injury to soybean. Crop Protection

19:113-120.

Hammond RB, and Byers RA. 2002. Agriolimacidae and Arionidae as pests in conservation-

tillage soybean and maize cropping in North America. In: Barker GM, ed. Molluscs as

crop pests, 301-314.

Harmon HE. 2006. Non-target effects of clothianidin seed-treated corn on the invertebrate

community. M.S. Thesis. Towson University.

Hebert PD, Ratnasingham S, and deWaard JR. 2003. Barcoding animal life: cytochrome c

oxidase subunit 1 divergences among closely related species. Proceedings of the Royal

Society of London Series B, Biological Sciences 270 Suppl 1:S96-99.

Heidel-Baker TT. 2012. Compatibility of soybean aphid integrated pest management strategies.

Ph.D. Dissertation. University of Minnesota.

Henry M, Béguin M, Requier F, Rollin O, Odoux J-F, Aupinel P, Aptel J, Tchamitchian S, and

Decourtye A. 2012. A common pesticide decreases foraging success and survival in

honey bees. Science 336:348-350.

Herbert A, and Kennedy G. 2015. New survey shows high level and widespread resistance of

thrips to neonicotinoid insecticides. Available at http://blogs.ext.vt.edu/ag-pest-

advisory/files/2015/02/NeonicThripsResistance.pdf.

Higgins JPT, Thompson SG, Deeks JJ, and Altman DG. 2003. Measuring inconsistency in meta-

analyses. British Medical Journal 327:557-560.

Hladik ML, Kolpin DW, and Kuivila KM. 2014. Widespread occurrence of neonicotinoid

insecticides in streams in a high corn and soybean producing region, USA.

Environmental Pollution 193:189-196.

Hodgson EW, Kemis M, and Geisinger B. 2012. Assessment of Iowa soybean growers for insect

pest management practices. Journal of Extension 50:4RIB6.

214 Hopwood J, Hoffman Black S, Vaughan M, and Lee-Mader E. 2013. Beyond the birds and the

bees: Effects of neonicotinoid insecticides on agriculturally important beneficial

invertebrates. Portland, OR: The Xerces Society for Invertebrate Conservation. p 32.

Hoque M, Mollik SR, Nazimuddin M, Alam SN, Khorsheduzzaman M, Jasmine HS, Sultana NA,

Rahman M, Ahmed B, Rajotte E, and Luther G. 2002. Survey of bean pod borer (Maruca

testulalis), whitefly and aphids and their natural enemies on country bean (Lablab

purpureus). Tenth Annual Report of the Integrated Pest Management Collaborative

Research Support Program (IPM CRSP). p 123-124.

Horowitz J, Ebel R, and Ueda K. 2010. "No-till" farming is a growing practice. Economic

Research Information Bulletin No 70. Washington, DC: U. S. Department of Agriculture,

ERS.

Hull LA, and Beers EE. 1985. Ecological selectivity: modifying chemical control practices to

preserve natural enemies. In: Hoy M, editor. Symposium on Biological Control in

Agricultural Integrated Pest Management Systems. Citrus Research and Education

Center, University of Florida: Academic Press. p 103-122.

Hurley T, and Mitchell PD. 2014a. The value of neonicotinoids in North American agriculture:

Methods and assumptions for estimating the impact of neonicotinoid insecticides on pest

management practices and costs for U. S. corn, soybean, wheat, cotton and sorghum

farmers. AgInformatics LLC.

Hurley T, and Mitchell PD. 2014b. The value of neonicotinoids in North American agriculture:

value of insect pest management to U.S. and Canadian corn, soybean, and canola farmers.

AgInformatics LLC.

Iftikhar R, Ashfaq M, Rasool A, and Hebert PD. 2016. DNA barcode analysis of thrips

(Thysanoptera) diversity in Pakistan reveals cryptic species complexes. PLoS ONE

11:e0146014.

215 Inoue T, and Sakurai T. 2007. The phylogeny of thrips (Thysanoptera: Thripidae) based on partial

sequences of cytochrome oxidase I, 28S ribosomal DNA and elongation factor-1⍺ and

the association with vector competence of tospoviruses. Applied Entomology and

Zoology 42:71-81.

Islam MS, Rahman MM, and Hossain T. 2010. Physico-morphological variation in hyacinth bean

[Lablab purpureus (L.) Sweet]. Bangladesh Journal of Agricultural Research 35:431-

438.

Jain RK, Bag S, Umamaheswaran K, and Mandal B. 2007. Natural infection by tospoviruses of

cucurbitaceous and fabaceous vegetable crops in India. Journal of Phytopathology

155:22-25.

Jeschke P, and Nauen R. 2008. Neonicotinoids - from zero to hero in insecticide chemistry. Pest

Management Science 64:1084-1098.

Jeschke P, Nauen R, Schindler M, and Elbert A. 2011. Overview of the status and global strategy

for neonicotinoids. Journal of Agricultural and Food Chemistry 59:2897-2908.

Johnson KD, O'Neal ME, Bradshaw JD, and Rice ME. 2008. Is preventative, concurrent

management of the soybean aphid (Hemiptera : aphididae) and bean leaf beetle

(Coleoptera : chrysomelidae) possible? Journal of Economic Entomology 101:801-809.

Johnson KD, O'Neal ME, Ragsdale DW, Difonzo CD, Swinton SM, Dixon PM, Potter BD,

Hodgson EW, and Costamagna AC. 2009. Probability of cost-effective management of

soybean aphid (Hemiptera: Aphididae) in North America. Journal of Economic

Entomology 102:2101-2108.

Johnson MW, and Tabashnik BE. 1999. Enhanced biological control through pesticide

selectivity. In: Bellows TS, and Fisher TW, eds. Handbook of Biological Control. San

Diego, CA: Academic Press, 297-317.

216 Jones DR. 2005. Plant viruses transmitted by thrips. European Journal of Plant Pathology

113:119-157.

Jordan TA, Youngman RR, Laub CL, Tiwari S, Kuhar TP, Balderson TK, Moore DM, and Saphir

M. 2012. Fall soil sampling method for predicting spring infestation of white grubs

(Coleoptera: Scarabaeidae) in corn and the benefits of clothianidin seed treatment in

Virginia. Crop Protection 39:57-62.

Kabir KH, Baksh ME, Roup FMA, Karim MA, and Ahmed A. 1996. Insecticide usage pattern on

vegetables at farmers' level of Jessore region in Bangladesh: a survey finding.

Bangladesh Journal of Agricultural Research 21:241-254.

Kadirvel P, Srinivasan R, Hsu YC, Su FC, and De La Pena R. 2013. Application of cytochrome

oxidase I sequences for phylogenetic analysis and identification of thrips species

occurring on vegetable crops. Journal of Economic Entomology 106:408-418.

Kamel A. 2010. Refined methodology for the determination of neonicotinoid pesticides and their

metabolites in honey bees and bee products by liquid chromatography-tandem mass

spectrometry (LC-MS/MS). Journal of Agricultural and Food Chemistry 58:5926-5931.

Klümper W, and Qaim M. 2014. A meta-analysis of the impacts of genetically modified crops.

PLoS ONE 9:e111629.

Kogan M. 1998. Integrated pest management: Historical perspectives and contemporary

developments. Annual Review of Entomology 43:243-270.

Koh I, Lonsdorf EV, Williams NM, Brittain C, Isaacs R, Gibbs J, and Ricketts TH. 2016.

Modeling the status, trends, and impacts of wild bee abundance in the United States.

Proceedings of the National Academy of Sciences of the United States of America

113:140-145.

Korbie DJ, and Mattick JS. 2008. Touchdown PCR for increased specificity and sensitivity in

PCR amplification. Nature Protocols 3:1452-1456.

217 Koricheva J, Gurevitch J, and Mengersen K. 2013. Handbook of Meta-analysis in Ecology and

Evolution. Princeton, NJ: Princeton University Press.

Krauter PC, Sansone CG, and Heinz KM. 2001. Assessment of Gaucho (R) seed treatment effects

on beneficial insect abundance in sorghum. Southwestern Entomologist 26:143-146.

Krupke CH, Hunt GJ, Eitzer BD, Andino G, and Given K. 2012. Multiple routes of pesticide

exposure for honey bees living near agricultural fields. PLoS ONE 7:e29268.

Kullik SA, Sears MK, and Schaafsma AW. 2011. Sublethal effects of Cry 1F Bt corn and

clothianidin on black cutworm (Lepidoptera: Noctuidae) larval development. Journal of

Economic Entomology 104:484-493.

Kumar P, Pandit SS, Steppuhn A, and Baldwin IT. 2014. Natural history-driven, plant-mediated

RNAi-based study reveals CYP6B46's role in a nicotine-mediated antipredator herbivore

defense. Proceedings of the National Academy of Sciences of the United States of

America 111:1245-1252.

Kutner MH, Nachstheim CJ, Neter J, and Li W. 2005. Applied Linear Statistical Models, 5th

Edition. Boston, MA: McGraw-Hill.

Landis DA, Gardiner MM, van der Werf W, and Swinton SM. 2008. Increasing corn for biofuel

production reduces biocontrol services in agricultural landscapes. Proceedings of the

National Academy of Sciences of the United States of America 105:20552-20557.

Lange WH. 1959. Seed treatment as a method of insect control. Annual Review of Entomology

4:363-388.

Laurent FM, and Rathahao E. 2003. Distribution of C-14 imidacloprid in sunflowers (Helianthus

annuus L.) following seed treatment. Journal of Agricultural and Food Chemistry

51:8005-8010.

Laurino D, Manino A, Patetta A, and Porporato M. 2013. Toxicity of neonicotinoid insecticides

on different honey bee genotypes. Bulletin of Insectology 66:119-126.

218 Lee S-T, Srinivasan R, Wu Y-J, and Talekar NS. 2007. Occurrence and characterization of a

nucleopolyhedrovirus from Maruca vitrata (Lepidoptera, Pyralidae) isolated in Taiwan.

Biocontrol 52:801-819.

Lenth RV. 2016. Least-squares means: The R package lsmeans. Journal of Statistical Software

69:1-33.

Leslie TW, Biddinger DJ, Rohr JR, and Fleischer SJ. 2010. Conventional and seed-based insect

management strategies similarly influence nontarget coleopteran communities in maize

Environmental Entomology 39:2045-2055.

Lilly JH. 1956. Soil insects and their control. Annual Review of Entomology 1:203-222.

Limay-Rios V, Forero LG, Xue Y, Smith J, Baute T, and Schaafsma A. 2016. Neonicotinoid

insecticide residues in soil dust and associated parent soil in fields with a history of seed

treatment use on crops in southwestern Ontario. Environmental Toxicology and

Chemistry 35:303-310.

Littell RC, Milliken GA, Stroup WW, Wolfinger RD, and Schabenberger O. 2006. Analysis of

repeated measures data. SAS for Mixed Models, Second Edition. Cary, NC: SAS

Institute, Inc., 159-203.

Lopezaraiza-Mikel ME, Hayes RB, Whalley MR, and Memmott J. 2007. The impact of an alien

plant on a native pollinator network: an experimental approach. Ecology Letters 10:539-

550.

Losey J, and Vaughan M. 2006. The economic value of ecological services provided by insects.

Bioscience Journal 56:311-323.

Lucas E, Giroux S, Demougeot S, Duchesne RM, and Coderre D. 2004. Compatibility of a

natural enemy, Coleomegilla maculata lengi (Col., Coccinellidae) and four insecticides

used against the Colorado potato beetle (Col., Chrysomelidae). Journal of Applied

Entomology 128:233-239.

219 Lundgren JG. 2009. Plant-incorporated pest resistance and natural enemies. In: Lundgren JG, ed.

Relationships of Natural Enemies and Non-Prey Foods: Springer Science + Business

Media B.V., 309-331.

Lundgren JG, Shaw JT, Zaborski ER, and Eastman CE. 2006. The influence of organic transition

systems on beneficial ground-dwelling arthropods and predation of insects and weed

seeds. Renewable Agriculture and Food Systems 21:227-237.

Lundgren JG, and Wiedenmann RN. 2002. Coleopteran-specific Cry3Bb toxin from transgenic

corn pollen does not affect the fitness of a nontarget species, Coleomegilla maculata

DeGeer (Coleoptera : Coccinellidae). Environmental Entomology 31:1213-1218.

Lundin O, Rundlof M, Smith HG, Fries I, and Bommarco R. 2015. Neonicotinoid insecticides

and their impacts on bees: a systematic review of research approaches and identification

of knowledge gaps. PLoS ONE 10:e0136928.

Maass BL, Knox MR, Venkatesha SC, Angessa TT, Ramme S, and Pengelly BC. 2010. Lablab

purpureus - a crop lost for Africa? Tropical Plant Biology 3:123-135.

Macharia I, Backhouse D, Skilton R, Ateka E, Wu SB, Njahira M, Maina S, and Harvey J. 2015.

Diversity of thrips species and vectors of tomato spotted wilt virus in production systems

in Kenya Journal of Economic Entomology 108:20-28.

Magalhaes LC, Hunt TE, and Siegfried BD. 2009. Efficacy of neonicotinoid seed treatments to

reduce soybean aphid populations under field and controlled conditions in Nebraska.

Journal of Economic Entomology 102:187-195.

Main AR, Headley JV, Peru KM, Michel NL, and Cessna AJ. 2014. Widespread use and frequent

detection of neonicotinoid insecticides in wetlands of Canada's prairie pothole region.

PLoS ONE 9:e92821.

220 Maisnam S, Dwijamani Singh O, and Varatharajan R. 2012. Diversity and diagnostics of

Thysanoptera inhabiting leguminous plants with a note on life cycle of Megalurothrips

peculiaris Bagnall. Indian Journal of Entomology 74:274-280.

Marvier M, McCreedy C, Regetz J, and Kareiva P. 2007. A meta-analysis of effects of Bt cotton

and maize on nontarget invertebrates. Science 316:1475-1477.

McCarville MT, O'Neal ME, Potter BD, Tilmon KJ, Cullen EM, McCornack BP, Tooker JF, and

Prischmann-Voldseth DA. 2014. One gene versus two: a regional study on the efficacy of

single gene versus pyramided resistance for soybean aphid management. Journal of

Economic Entomology 107:1680-1687.

McCornack BP, and Ragsdale DW. 2006. Efficacy of thiamethoxam to suppress soybean aphid

populations in Minnesota soybean. Crop Management. 10.1094/CM-2006-0915-01-RS

McDonnell RJ, Paine TD, and Gormally MJ. 2009. Slugs: A Guide to the Invasive and Native

Fauna of California. Oakland: Diision of Agriculture and Natural Resources, University

of California.

Meng X, Zhang Y, Guo B, Sun H, Liu C, and Liu Z. 2015. Identification of key amino acid

differences contributing to neonicotinoid sensitivity between two nAChR alpha subunits

from Pardosa pseudoannulata. Neuroscience Letters 584:123-128.

Meyer CP. 2003. Molecular systematics of cowries (Gastropoda: Cypraeidae) and diversification

patterns in the tropics. Biological Journal of the Linnean Society 79:401-459.

Meyer CP, and Paulay G. 2005. DNA barcoding: error rates based on comprehensive sampling.

PLoS Biology 3:e422.

Miller SE. 2007. DNA barcoding and the renaissance of taxonomy. Proceedings of the National

Academy of Sciences of the United States of America 104:4775-4776.

Minnesota Department of Agriculture. 2014. Minnesota pesticide sales information. Available at

http://www.mda.state.mn.us/chemicals/pesticides/useandsales.aspx

221 Mirab-Balou M, Minaei K, and Chen XX. 2013. An illustrated key to the genera of Thripinae

(Thysanoptera, Thripidae) from Iran. ZooKeys 317:27-52.

Mitchell PD. 2014. A meta-analysis approach to estimating the yield effects of neonicotinoids.

The value of neonicotinoids in North American agriculture: AgInformatics, LLC.

Mizell RF, III, and Sconyers MC. 1992. Toxicity of imidacloprid to selected arthropod predators

in the laboratory. Florida Entomologist 75:277-280.

Moritz G, Mound LA, Morris DC, and Goldarazena A. 2004. Pest thrips of the world - an

identification system using molecular and microscopial methods. University of

Queensland: CBIT.

Mortiz G, O'Donnell C, and Parelia M. 2009. Pest thrips of North America associated with

domestic and imported crops. The University of Queensland: QAAFI Biological

Information Technology (QBIT).

Moser SE, and Obrycki JJ. 2009. Non-target effects of neonicotinoid seed treatments; mortality of

coccinellid larvae related to zoophytophagy. Biological Control 51:487-492.

Mound L, and Sartiami D. 2013. Identification of the terebrantian thrips (Insecta, Thysanoptera)

associated with cultivated plants in Java, Indonesia. ZooKeys 306:1-21.

Mound LA. 2013. Homologies and host-plant specificity: recurrent problems in the study of

thrips. Florida Entomologist 96:318-322.

Mullin CA, Saunders MC, Leslie TW, Biddinger DJ, and Fleischer SJ. 2005. Toxic and

behavioral effects to Carabidae of seed treatments used on Cry3Bbl- and Cry1Ab/c-

protected corn. Environmental Entomology 34:1626-1636.

Musser FR, Catchot AL, Davis JA, Herbert DA, Leonard BR, Lorenz GM, Reed T, Reisig DD,

and Stewart SD. 2012. Soybean insect losses in the Southern U. S. Midsouth

Entomologist 5:11-22.

222 Myers C, and Hill E. 2014. Benefits of Neonicotinoid Seed Treatments to Soybean Production.

Washington, D.C.: U.S. Environmental Protection Agency.

Naranjo S. 2005. Long-term assessment of the effects of transgenic Bt cotton on the function of

the natural enemy community. Environmental Entomology 34:1211-1223.

Naranjo S. 2009. Impacts of Bt crops on non-target invertebrates and insecticide use patterns.

CAB Reviews: Perspectives in Agriculture, Veterinary Science, Nutrition and Natural

Resources 4:1-23.

Naranjo SE, Ellsworth PC, and Frisvold GB. 2015. Economic value of biological control in

integrated pest management of managed plant systems. Annual Review of Entomology

60:621-645.

Newman JA, Bergelson J, and Grafen A. 1997. Blocking factors and hypothesis tests in ecology:

Is your statistics text wrong? Ecology 78:1312-1320.

NOAA. 2016. Climate Data Online. Available at http://www.ncdc.noaa.gov/cdo-

web/search?datasetid=GHCNDMS.

Norris RF, Caswell-Chen EP, and Kogan M. 2002. Concepts in Integrated Pest Management.

Upper Saddle River, NJ: Prentice Hall.

O'Neal ME, and Hogsdon E. 2014. Personal communication to M. Douglas and J. Tooker. Iowa

State University, Department of Entomology.

Obopile M, Hammond RB, and Thomison PR. 2013. Interaction among planting dates, transgenic

maize, seed treatment, corn rootworm damage and grain yield. Journal of Applied

Entomology 137:45-55. 10.1111/j.1439-0418.2012.01716.x

Ohnesorg WJ, Johnson KD, and O'Neal ME. 2009. Impact of reduced-risk insecticides on

soybean aphid and associated natural enemies. Journal of Economic Entomology

102:1816-1826.

223 Olson K, Badibanga T, and Difonzo CD. 2008. Farmers' awareness and use of IPM for soybean

aphid control: report of survey results for the 2004, 2005, 2006, and 2007 crop years.

Department of Applied Economics, University of Minneosota.

Ontario Ministry of Agriculture Food and Rural Affairs. 2014. Do you need to use insecticide-

treated corn and soybean seed? Available at

http://www.omafra.gov.on.ca/english/crops/reduceneonics.html.

Ooi PAC. 1982. Attempts at forecasting rice planthopper populations in Malaysia. Entomophaga

27 (Special Issue):89-98.

Osteen CD, and Fernandez-Cornejo J. 2013. Economic and policy issues of US agricultural

pesticide use trends. Pest Management Science 69:1001-1025.

Palmer J. 1987. Megalurothrips in the flowers of tropical legumes: a morphometric study. In:

Holman J, Pelikan J, Dixon AFG, and Weisman L, eds. Population structure, genetics,

and taxonomy of aphids and Thysanoptera. The Hague, NL: SPB Publishing, 480-495.

Pappu HR, Jones RA, and Jain RK. 2009. Global status of tospovirus epidemics in diverse

cropping systems: successes achieved and challenges ahead. Virus Research 141:219-

236.

Pedersen P. 2014. Personal communication to J. F. Tooker. Seedcare Technology Manager for

Syngenta.

Pinheiro J, Bates D, DebRoy S, and Sarkar D. 2013. nlme: Linear and Nonlinear Mixed Effects

Models. R package version 3.1-113.

Pioneer Hi-Bred International Inc. 2012. Environmental report for the determination of

nonregulated status for insect-resistant and herbicide-tolerant 4114 maize. Submitted to

USDA-APHIS.

Pioneer Hi-Bred International Inc. 2014a. Pioneer premium seed treatment - corn. Available at

https://www.pioneer.com/home/site/us/products/corn/seed-treatments.

224 Pioneer Hi-Bred International Inc. 2014b. Pioneer premium seed treatment - soybean. Available

at https://www.pioneer.com/home/site/us/products/soybean/seed-treatments.

Pisa LW, Amaral-Rogers V, Belzunces LP, Bonmatin JM, Downs CA, Goulson D, Kreutzweiser

DP, Krupke C, Liess M, McField M, Morrissey CA, Noome DA, Settele J, Simon-Delso

N, Stark JD, Van der Sluijs JP, Van Dyck H, and Wiemers M. 2015. Effects of

neonicotinoids and fipronil on non-target invertebrates. Environmental Science and

Pollution Research 22:68-102.

Plantwise Knowledge Bank. 2016. Megalurothrips distalis. Available at

http://www.plantwise.org/KnowledgeBank/Datasheet.aspx?dsid=52629.

Pons X, and Albajes R. 2002. Control of maize pests with imidacloprid seed dressing treatment in

Catalonia (NE Iberian Peninsula) under traditional crop conditions. Crop Protection

21:943-950.

Prasifka JR, Hellmich RL, Dively GP, Higgins LS, Dixon PM, and Duan JJ. 2008. Selection of

nontarget arthropod taxa for field research on transgenic insecticidal crops: using

empirical data and statistical power. Environmental Entomology 37:1-10.

R Core Team. 2014. R: A Language Environment for Statistical Computing. Vienna: R

Foundation for Statistical Computing.

R Core Team. 2015. R: A Language and Environment for Statistical Computing. Vienna: R

Foundation for Statistical Computing.

Ragsdale DW, Landis DA, Brodeur J, Heimpel GE, and Desneux N. 2011. Ecology and

management of the soybean aphid in North America. Annual Review of Entomology

56:375-399. 10.1146/annurev-ento-120709-144755

Ratnasingham S, and Hebert PD. 2007. BOLD: The barcode of life data system

(http://www.barcodinglife.org). Molecular Ecology Notes 7:355-364.

225 Ratnasingham S, and Hebert PD. 2013. A DNA-based registry for all animal species: the barcode

index number (BIN) system. PLoS ONE 8:e66213.

Rebijith KB, Asokan R, Krishna V, Ranjitha HH, Kumar NKK, and Ramamurthy VV. 2014.

DNA barcoding and elucidation of cryptic diversity in thrips (Thysanoptera). Florida

Entomologist 97:1328-1347.

Reisig DD, Herbert DA, and Malone S. 2012. Impact of neonicotinoid seed treatments on thrips

(Thysanoptera: Thripidae) and soybean yield in Virginia and North Carolina. Journal of

Economic Entomology 105:884-889.

Rothenbücher J, and Schaefer M. 2006. Submersion tolerance in floodplain arthropod

communities. Basic and Applied Ecology 7:398-408.

Schmidt T, Schmuck R, and Maus C. 2002. Data variability in carabid field studies (Coleoptera:

Carabidae) and how to deal with small-scaled inhomogeneities of environmental

conditions. Bulletin OILB/SROP 25:37-42.

Seagraves MP, and Lundgren JG. 2012. Effects of neonicotinoid seed treatments on soybean

aphid and its natural enemies. Journal of Pest Science 85:125-132.

Settle WH, Ariawan H, Astuti ET, Cahyana W, Hakim AL, Hindayana D, Lestari A, and

Pajarnigsih. 1996. Managing tropical rice pests through conservation of generalist natural

enemies and alternative prey. Ecology 77:1975-1988.

Sharma HC. 1998. Bionomics, host plant resistance, and management of the legume pod borer,

Maruca vitrata - a review. Crop Protection 17:373-386.

Sheppard SK, and Harwood JD. 2005. Advances in molecular ecology: tracking trophic links

through predator-prey food-webs. Functional Ecology 19:751-762.

Shipp JL, and Zariffa N. 1991. Spatial patterns of and sampling methods for western flower thrips

(Thysanoptera: Thripidae) on greenhouse sweet pepper. Canadian Entomologist 123:989-

1000.

226 Simmons CL, Pedigo LP, and Rice ME. 1998. Evaluation of seven sampling techniques for

wireworms (Coleoptera : Elateridae). Environmental Entomology 27:1062-1068.

Simms LC, Ester A, and Wilson MJ. 2006. Control of slug damage to oilseed rape and wheat

with imidacloprid seed dressings in laboratory and field experiments. Crop Protection

25:549-555.

Simon-Delso N, Amaral-Rogers V, Belzunces LP, Bonmatin JM, Chagnon M, Downs C, Furlan

L, Gibbons DW, Giorio C, Girolami V, Goulson D, Kreutzweiser DP, Krupke CH, Liess

M, Long E, McField M, Mineau P, Mitchell EAD, Morrissey CA, Noome DA, Pisa L,

Settele J, Stark JD, Tapparo A, Van Dyck H, Van Praagh J, Van der Sluijs JP, Whitehorn

PR, and Wiemers M. 2015. Systemic insecticides (neonicotinoids and fipronil): trends,

uses, mode of action and metabolites. Environmental Science and Pollution Research

22:5-34.

Smith GJ. 1987. Pesticide use and toxicity in relation to wildlife: organophosphorous and

carbamate compounds. Resource Publication: U. S. Department of the Interior, Fish and

Wildlife Service.

Smith JF, Catchot AL, Musser FR, and Gore J. 2013. Effects of aldicarb and neonicotinoid seed

treatments on twospotted spider mite on cotton. Journal of Economic Entomology

106:807-815.

Sotelo-Cardona PA. 2010. Interactions among biological control, cultural control and barley

resistance to the Russian wheat aphid, Diuraphis noxia (Kurdjumov), in Colorado,

Kansas and Nebraska. Ph.D. Dissertation. Kansas State University.

South A. 1992. Terrestrial Slugs: Biology, Ecology, and Control. London: Chapman & Hall.

Sparks TC. 2013. Insecticide discovery: an evaluation and analysis. Pesticide Biochemistry and

Physiology 107:8-17.

227 Spigler MI. 2013. Effects of thiamethoxam seed treatments on nutritive sources available to Orius

insidiosus say in Indiana soybean agroecosystems. M.S. Thesis. Purdue University.

Sreekanth M, Lakshmi M, and Rao Y. 2015. Efficacy and economics of certain new generation

novel insecticides against legume pod borer, Maruca vitrata (Geyer) on pigeonpea

(Cajanus cajan L.). Journal of Applied Biology and Biotechnology 3:7-10.

Sreekanth M, and Seshamahalakshmi M. 2012. Studies on relative toxicity of biopesticides to

Helicoverpa armigera (Hubner) and Maruca vitrata (Geyer) on pigeonpea (Cajanus

cajan L.). Journal of Biopesticides 5:191-195.

Srinivasan R. 2008. Susceptibility of legume pod borer (LPB), Maruca vitrata to delta-

endotoxins of Bacillus thuringiensis (Bt) in Taiwan. Journal of Invertebrate Pathology

97:79-81.

Srinivasan R, Yule S, Chang JC, Malini P, Lin MY, Hsu YC, and Schafleitner R. 2012. Towards

developing a sustainable management strategy for legume pod borer, Maruca vitrata on

yard-long bean in Southeast Asia. SEAVEG2012 Regional Symposium. p 76-82.

Srivathsan A, and Meier R. 2012. On the inappropriate use of Kimura-2-parameter (K2P)

divergences in the DNA-barcoding literature. Cladistics 28:190-194.

Stamm MD, Heng-Moss TM, Baxendale FP, Siegfried BD, Blankenship EE, and Nauen R. 2015.

Uptake and translocation of imidacloprid, clothianidin and flupyradifurone in seed-

treated soybeans. Pest Management Science 72:1099-1109.

Steffey KL, and Gray ME. 2000. Should we expect more from wireworms, white grubs, grape

colaspis, et al. in the future? . Proceedings of the Ilinois Crop Protection Technology

Conference. University of Illinois at Urbana-Champaign.

Stern VM, Smith RF, van den Bosch R, and Hagen KS. 1959. The Integrated Control concept.

Hilgardia 29:81-101.

228 Stinner BR, and House GJ. 1990. Arthropods and other invertebrates in conservation-tillage

agriculture. Annual Review of Entomology 35:299-318.

Stokstad E, and Grullón G. 2013. Infographic: Pesticide Planet. Science 341:730-731.

Sur R, and Stork A. 2003. Uptake, translocation and metabolism of imidacloprid in plants.

Bulletin of Insectology 56:35-40.

Symondson WOC. 2004. Coleoptera (Carabidae, Staphylinidae, Lampyridae, Drilidae and

Silphidae) as predators of terrestrial gastropods. In: Barker GM, ed. Natural Enemies of

Terrestrial Molluscs, 37-84.

Symondson WOC, Glen DM, Erickson ML, Liddell JE, and Langdon CJ. 2000. Do earthworms

help to sustain the slug predator Pterostichus melanarius (Coleoptera : Carabidae) within

crops? Investigations using monoclonal antibodies. Molecular Ecology 9:1279-1292.

Symondson WOC, Sunderland KD, and Greenstone MH. 2002. Can generalist predators be

effective biocontrol agents? Annual Review of Entomology 47:561-594.

Syngenta. 2016. Cruiser ® 5FS Insecticide Label. Basel, Switzerland.

Szczepaniec A, Creary SF, Laskowski KL, Nyrop JP, and Raupp MJ. 2011. Neonicotinoid

insecticide imidacloprid causes outbreaks of spider mites on elm trees in urban

landscapes. PLoS ONE 6:e20018.

Szczepaniec A, Raupp MJ, Parker RD, Kerns D, and Eubanks MD. 2013. Neonicotinoid

Insecticides Alter Induced Defenses and Increase Susceptibility to Spider Mites in

Distantly Related Crop Plants. PLoS ONE 8. 10.1371/journal.pone.0062620

Tamò M, Baumbartner J, Delucchi V, and Herren HR. 1993. Assessment of key factors

responsible for the pest status of the bean flower thrips Megalurothrips sjostedti

(Thysanoptera: Thripidae) in West Africa. Bulletin of Entomological Research 83:251-

258.

229 Tamura K, Stecher G, Peterson D, Filipski A, and Kumar S. 2013. MEGA6: Molecular

Evolutionary Genetics Analysis Version 6.0. Molecular Biology and Evolution 30:2725-

2729.

Tanaka K, Endo S, and Kazano H. 2000. Toxicity of insecticides to predators of rice

planthoppers: spiders, the mirid bug and the dryinid wasp. Applied Entomology and

Zoology 35:177-187.

Telfer AC, Young MR, Quinn J, Perez K, Sobel CN, Sones JE, Levesque-Beaudin V, Derbyshire

R, Fernandez-Triana J, Rougerie R, Thevanayagam A, Boskovic A, Borisenko AV, Cadel

A, Brown A, Pages A, Castillo AH, Nicolai A, Glenn Mockford BM, Bukowski B,

Wilson B, Trojahn B, Lacroix CA, Brimblecombe C, Hay C, Ho C, Steinke C, Warne

CP, Garrido Cortes C, Engelking D, Wright D, Lijtmaer DA, Gascoigne D, Hernandez

Martich D, Morningstar D, Neumann D, Steinke D, Marco DeBruin DD, Dobias D, Sears

E, Richard E, Damstra E, Zakharov EV, Laberge F, Collins GE, Blagoev GA, Grainge G,

Ansell G, Meredith G, Hogg I, McKeown J, Topan J, Bracey J, Guenther J, Sills-Gilligan

J, Addesi J, Persi J, Layton KK, D'Souza K, Dorji K, Grundy K, Nghidinwa K,

Ronnenberg K, Lee KM, Xie L, Lu L, Penev L, Gonzalez M, Rosati ME, Kekkonen M,

Kuzmina M, Iskandar M, Mutanen M, Fatahi M, Pentinsaari M, Bauman M, Nikolova N,

Ivanova NV, Jones N, Weerasuriya N, Monkhouse N, Lavinia PD, Jannetta P, Hanisch

PE, McMullin RT, Ojeda Flores R, Mouttet R, Vender R, Labbee RN, Forsyth R, Lauder

R, Dickson R, Kroft R, Miller SE, MacDonald S, Panthi S, Pedersen S, Sobek-Swant S,

Naik S, Lipinskaya T, Eagalle T, Decaens T, Kosuth T, Braukmann T, Woodcock T,

Roslin T, Zammit T, Campbell V, Dinca V, Peneva V, Hebert PD, and deWaard JR.

2015. Biodiversity inventories in high gear: DNA barcoding facilitates a rapid biotic

survey of a temperate nature reserve. Biodiversity Data Journal:e6313.

230 Thelin GP, and Stone WW. 2013. Estimation of annual agricultural pesticide use for counties of

the conterminous United States, 1992-2009. USGS Scientific Investigations Report. p 1-

54.

Therneau T. 2014. A package for survival analysis in S. R package version 2.37-7.

Thiele H-U. 1977. Carabid Beetles in their Environments. Berlin: Springer-Verlag.

Thompson GD, Dutton R, and Sparks TC. 2000. Spinosad - a case study: an example from a

natural products discovery programme. Pest Management Science 56:696-702.

Thorpe KW, and Barbosa P. 1986. Effects of consumption of high and low nicotine tobacco by

Manduca sexta (Lepidoptera: Sphingidae) on survival of gregarious endoparasitoid

Cotesia congregata (Hymenoptera: Braconidae) Journal of Chemical Ecology 12:1329-

1337.

ThripsWiki. 2016. ThripsWiki - providing information on the World's thrips. Available at

http://thrips.info/wiki/Main_Page (accessed 12 April 2016).

Tinsley NA, Steffey KL, Estes RE, Heeren JR, Gray ME, and Diers BW. 2012. Field-level effects

of preventative management tactics on soybean aphids (Aphis glycines Matsumura) and

their predators. Journal of Applied Entomology 135:361-371.

Tomizawa M, and Casida JE. 2005. Neonicotinoid insecticide toxicology: mechanisms of

selective action. Annual Review of Pharmacology and Toxicology 45:247-268.

Torres JB, and Ruberson JR. 2004. Toxicity of thiamethoxam and imidacloprid to Podisus

nigrispinus (Dallas) (Heteroptera: Pentatomidae) nymphs associated to aphid and

whitefly control in cotton. Neotropical Entomology 33:99-106.

Triplehorn CA, and Johnson NF. 2005. Borror and Delong's Introduction to the Study of Insects,

7th edition. Belmont: Thomson Brooks/Cole.

Tyler-Julian K, Funderburk J, and Mound L. 2014. Megalurothrips distalis (Thysanoptera:

Thripidae) in the flowers of kudzu in Florida. Florida Entomologist 97:835-840.

231 U. K. DEFRA. 2010. A review of slug control in winter wheat and oilseed rape - PS2803.

Available at http://sciencesearch.defra.gov.uk (accessed 03 June 2014).

U. S. Department of Agriculture ERS. 2014a. Commodity costs and returns. Available at

http://www.ers.usda.gov/data-products/commodity-costs-and-returns.aspx.

U. S. Department of Agriculture ERS. 2014b. Recent trends in GE adoption. Available at

http://www.ers.usda.gov/data-products/adoption-of-genetically-engineered-crops-

in-the-us/recent-trends-in-ge-adoption.aspx - .VC1c1hCITq4.

U. S. Department of Agriculture NASS. 2014a. Agricultural Chemical Use Program. Available at

http://www.nass.usda.gov/Surveys/Guide_to_NASS_Surveys/Chemical_Use

U. S. Department of Agriculture NASS. 2014b. Agricultural prices. Available at

http://usda.mannlib.cornell.edu/MannUsda/viewDocumentInfo.do?documentID=1

002

U. S. Department of Agriculture NASS. 2014c. Quick Stats 2.0. Available at

http://quickstats.nass.usda.gov.

U. S. Department of Agriculture National Institute of Food and Agriculture. 2013. National Road

Map for Integrated Pest Management. Available at

http://www.csrees.usda.gov/nea/pest/pdfs/ipm_roadmap.pdf.

U. S. Environmental Protection Agency. 1998. Imidacloprid; pesticide tolerances for emergency

exemptions. Federal Register 63:66438-66447.

U. S. Environmental Protection Agency. 2003. Imidacloprid; pesticide tolerances for emergency

exemptions. Federal Register 68:61624-61634.

U. S. Environmental Protection Agency. 2014a. Acetamiprid Registration Review. Docket EPA-

HQ-OPP-2012-0329. . Available at http://www.regulations.gov/ -

!docketDetail;D=EPA-HQ-OPP-2012-0329 (accessed October 15 2014).

232 U. S. Environmental Protection Agency. 2014b. Clothianidin Registration Review. Docket EPA-

HQ-OPP-2011-0865. Available at http://www.regulations.gov/ -

!docketDetail;D=EPA-HQ-OPP-2011-0865 (accessed October 15 2014).

U. S. Environmental Protection Agency. 2014c. Dinotefuran Registration Review. Docket EPA-

HQ-OPP-2011-0920. . Available at http://www.regulations.gov/ -

!docketDetail;D=EPA-HQ-OPP-2011-0920 (accessed October 15 2014).

U. S. Environmental Protection Agency. 2014d. Imidacloprid Registration Review. Docket EPA-

HQ-OPP-2008-0844. Available at http://www.regulations.gov/ -

!docketDetail;D=EPA-HQ-OPP-2008-0844 (accessed October 15 2014).

U. S. Environmental Protection Agency. 2014e. Thiacloprid Registration Review. Docket EPA-

HQ-OPP-2012-0218. . Available at http://www.regulations.gov/ -

!docketDetail;D=EPA-HQ-OPP-2012-0218 (accessed October 15 2014).

U. S. Environmental Protection Agency. 2014f. Thiamethoxam Registration Review. Docket

EPA-HQ-OPP-2011-0581. . Available at http://www.regulations.gov/ -

!docketDetail;D=EPA-HQ-OPP-2011-0581 (accessed October 15 2014).

U. S. Environmental Protection Agency. 2015. EPA actions to protect pollinators. Available at

http://www2.epa.gov/pollinator-protection/epa-actions-protect-pollinators.

U. S. Geological Survey. 2014. Pesticide national synthesis project, annual pesticide use maps:

1992-2011. Available at https://water.usgs.gov/nawqa/pnsp/usage/maps/index.php

(accessed 31 July 2014).

Ubick D, Paquin P, Cushing PE, and Roth V. 2005. Spiders of North America: an identification

manual. New Hampshire: American Arachnological Society.

233 Umbarkar PS, and Parsana GJ. 2014. Field efficacy of different insecticides against spotted pod

borer, Maruca vitrata (Geyer) infesting greengram. Journal of Industrial Pollution

Control 30:227-230.

USDA NASS. 2014. Tillage practices released. Available at

https://www.nass.usda.gov/Statistics_by_State/Pennsylvania/Publications/Survey

_Results/tillage_practices14.pdf (accessed May 8 2016).

USDA NASS. 2016. CropScape - Cropland Data Layer. Available at

https://nassgeodata.gmu.edu/CropScape/. van der Sluijs JP, Amaral-Rogers V, Belzunces LP, van Lexmond MFIJB, Bonmatin JM,

Chagnon M, Downs CA, Furlan L, Gibbons DW, Giorio C, Girolami V, Goulson D,

Kreutzweiser DP, Krupke C, Liess M, Long E, McField M, Mineau P, Mitchell EAD,

Morrissey CA, Noome DA, Pisa L, Settele J, Simon-Delso N, Stark JD, Tapparo A, Van

Dyck H, Van Praagh J, Whitehorn PR, and Wiemers M. 2015. Conclusions of the

Worldwide Integrated Assessment on the risks of neonicotinoids and fipronil to

biodiversity and ecosystem functioning. Environmental Science and Pollution Research

22:148-154.

Van Dijk TC, Van Staalduinen MA, and Van der Sluijs JP. 2013. Macro-invertebrate decline in

surface water polluted with imidacloprid. PLoS ONE 8:e62374.

Velayudhan K, Gopinathan K, and Bakthavatsalam N. 1985. Pollination potential, population

dynamics and dispersal of thrips species (Thysanoptera: Insecta) infesting flowers of

Dolichos lablab L. (Fabaceae). Proceedings of the Indian National Science Academy, B

51:574-580.

Viechtbauer W. 2007a. Accounting for heterogeneity via random-effects models and moderator

analyses in meta-analysis. Zeitschrift für Psychologie / Journal of Psychology 215:104-

121.

234 Viechtbauer W. 2007b. Hypothesis tests for population heterogeneity in meta-analysis. British

Journal of Mathematical and Statistical Psychology 60:29-60.

Viechtbauer W. 2010. Conducting meta-analyses in R with the metafor package. Journal of

Statistical Software 36:1-48.

Wang HY, Yang Y, Su JY, Shen JL, Gao CF, and Zhu YC. 2008. Assessment of the impact of

insecticides on Anagrus nilaparvatae (Pang et Wang) (Hymenoptera : Mymanidae), an

egg parasitoid of the rice planthopper, Nilaparvata lugens (Hemiptera : Delphacidae).

Crop Protection 27:514-522.

Whitehorn PR, O'Connor S, Wackers FL, and Goulson D. 2012. Neonicotinoid pesticide reduces

bumble bee colony growth and queen production. Science 336:351-352.

Wilde G, Roozeboom K, Ahmad A, Claassen M, Gordon B, Heer W, Maddux L, Martin V, Evans

P, Kofoid K, Long J, Schlegel A, and Witt M. 2007. Seed treatment effects on early-

season pests of corn and on corn growth and yield in the absence of insect pests. Journal

of Agricultural and Urban Entomology 24:177-193.

Wilde G, Roozeboom K, Claassen M, Janssen K, and Witt M. 2004. Seed treatment for control of

early-season pests of corn and its effect on yield. Journal of Agricultural and Urban

Entomology 21:75-85.

Williams MR. 2014. Cotton crop loss data. Mississippi State University Cooperative Extension

Service.

Wolfenbarger LL, Naranjo SE, Lundgren JG, Bitzer RJ, and Watrud LS. 2008. Bt crop effects on

functional guilds of non-target arthropods: a meta-analysis. PLoS ONE 3:e2118.

Yadav NK, and Singh PS. 2014. Bio-efficacy of chemical Insecticides against Spotted Pod Borer,

Maruca testulalis (Geyer) on Cowpea. International Journal of Agriculture, Environment

and Biotechnology 7:187.

235 Zhang W, and Swinton SM. 2012. Optimal control of soybean aphid in the presence of natural

enemies and the implied value of their ecosystem services. Journal of Environmental

Management 96:7-16.

Zollinger R, Dexter A, Dahl G, Fitterer S, McMullen M, Waldhaus E, Glogoza P, and

Ignaszewaski K. 1996. Pesticide use and pest management practices in North Dakota.

North Dakota State University.

Zollinger R, Glogoza P, McMullen M, Bradley C, Dexter A, Knopf D, Wilson E, DeJong T, and

Meyer W. 2006. Pesticide use and pest management practices in North Dakota. North

Dakota State University.

Zollinger R, Markell S, Knodel JJ, Gray J, Jantzi D, Hagemeister K, and Kilpatrick P. 2012.

Pesticide use and pest management practices in North Dakota. North Dakota State

University.

Zollinger R, McMullen M, Dahl G, Dexter A, Nalewaja J, Hamlin W, and Beckler D. 1992.

Pesticide use and pest management practices in North Dakota. North Dakota State

University.

Zollinger R, McMullen M, Knodel JJ, Gray J, Jantzi D, Kimmet G, Hagemeister K, and Schmitt

C. 2009. Pesticide use and pest management practices in North Dakota. North Dakota

State University.

VITA

Margaret R. Douglas

EDUCATION Ph.D. Entomology & International Agriculture & Development (2016) Department of Entomology, Pennsylvania State University. Advised by Dr. John F. Tooker. M.S. Entomology (2012) Department of Entomology, Pennsylvania State University. Advised by Dr. John F. Tooker. B.A. Biology (2004) Oberlin College. Minor in Philosophy. Advised by Dr. Mary Garvin.

PEER-REVIEWED PUBLICATIONS Douglas, M. and J. Tooker. 2015. Large-scale deployment of seed treatments has driven rapid increase in use of neonicotinoid insecticides and preemptive pest management in U.S. field crops. Environmental Science & Technology 49(8): 5088-5097. Douglas, M., J. Rohr, and J. Tooker. 2015. Neonicotinoid insecticide travels through a soil food chain, disrupting biological control of non-target pests and decreasing soya bean yield. Journal of Applied Ecology 52(1): 250-260. [Editor’s Choice/Cover] Shipanski, M., S. Bailey, M. Barbercheck, M. Douglas, D. Finney, K. Haider, J. Kaye, A. Kemanian, D. Mortensen, M. Ryan, J. Tooker, and C. White. 2014. A conceptual framework for evaluating multifunctionality of cover crops in agroecosystems. Agricultural Systems 125: 12-22. Wimp, G., S. Murphy, D. Lewis, M. Douglas, R. Ambikapathi, L. Van Tull, C. Gratton, and R. Denno. 2013. Predator hunting mode influences patterns of prey use from grazing and detrital food webs. Oecologia 171(2): 505-515. Douglas, M. and J. Tooker. 2012. Slug (Mollusca: Agriolimacidae, Arionidae) ecology and management in no-till field crops, with an emphasis on the mid-Atlantic region. Journal of Integrated Pest Management 3(1): C1-C9. Bentley, T., M. Douglas, I. Grettenberger, E. Lastro, C. Sidhu, and J. Smith. 2012. Student Debate: Global climate change will have substantial long-term negative effects on arthropod diversity: Pro Position. American Entomologist 58(2): 99 - 100. Allainguillaume, J., et al. 2011. Permanent genetic resources added to Molecular Ecology resources database 1 August 2010 – September 2010. Molecular Ecology Resources 11: 219-222.

GRANTS, AWARDS, AND FELLOWSHIPS Penn State University Ralph O. Mumma Award for Outstanding Achievement (2015) USAID U.S. Borlaug Fellowship in Global Food Security (2014) Penn State University Evans Family Award for Graduate Student Extension Achievement (2013) Pennsylvania Department of Agriculture Grant (2012), with J. F. Tooker USDA NE Sustainable Agriculture Research & Education Graduate Student Grant (2012) Entomology Society of America (Plant-Insect Section) M.S. Student Award (2012) International Org. of Biological Control (Nearctic Section) M.S. Student Award (2012) Sigma Xi Grant in Aid (2010)