Method Development for Solitary Bee Pesticide Risk Assessment Using Megachile Rotundata As a Surrogate Species

Method development for solitary bee pesticide risk assessment using Megachile rotundata as a surrogate species

Graham Ansell

A Thesis presented to The University of Guelph

In partial fulfilment of requirements for the degree of Master of Science in Environmental Sciences

Guelph, Ontario, Canada

ABSTRACT

METHOD DEVELOPMENT FOR SOLITARY BEE PESTICIDE RISK ASSESSMENT USING MEGACHILE ROTUNDATA AS A SURROGATE SPECIES

Graham Robert Ansell Advisor: University of Guelph, 2019 Dr. Cynthia Scott-Dupree

Solitary bee diversity and abundance declines worldwide are likely driven partly by pesticides. However, solitary bees may not be adequately protected under the current pesticide risk assessment paradigm that focuses primarily on the honey bee (Apis mellifera). My thesis focused on the development of toxicity testing methods for solitary bee pesticide risk assessment methods using the alfalfa leafcutting bee (Megachile rotundata) as a surrogate species. I generated baseline toxicity data for the positive control insecticides dimethoate, permethrin, and imidacloprid for use in laboratory topical and chronic oral assays as well as semi-field trials. I compared the toxicity of a range of pesticides to M. rotundata to known values for A. mellifera in the lab and found both species to be similarly sensitive. I also addressed methodological gaps, including the incorporation of individual bee mass in topical laboratory assays, the design of effective chronic oral treatment containers, and semi-field experimental parameters.

iii

ACKNOWLEDGEMENTS

I would like to thank my advisor Dr. Cynthia Scott-Dupree for her guidance and support to keep my project afloat and on track. Thank you also to Drs. Paul Sibley, Chris Cutler, and

Angela Gradish for serving on my advisory committee. This experience would not have been the same without the contribution of your different approaches and skill sets.

Thank you as well to my funding sources Bayer CropScience via the Bayer CropScience

Chair in Sustainable Pest Management (C. Scott-Dupree) and the Mitacs-Elevate grant awarded to Dr. Andrew J. Frewin in partnership with Syngenta Canada. Additionally, I would like to thank Keith Ardiel, Naomi Jakel, and Genna Wright from Bayer CropScience for help with locating, planting, and maintaining the field site in 2018; and Stephen Gradish for providing the field site in 2017 and assisting with its set-up and maintenance.

Thank you to Angela Gradish and Andrew Frewin, the most amazing postdocs a lab could ask for. Your willingness to help design experiments, spend a lifetime in the field, and edit everything is an incredible resource for any grad student. Andrew, none of this project would have been possible without your guidance and preliminary research. Thanks also to the rest of the lab for all the support: Dillon Muldoon, Alex Stinson, Caitlin MacDonald, and Tara Celetti.

Special thanks to the research assistants Glen, Abbie, Kevin, Dan, and Dana for doing the dirty work and keeping up with the crazy pace.

Thanks to my parents Stephen and Karen for all the support and encouragement during my education. And finally, thank you Lisandra, my love, for keeping me grounded and putting up with my mental absence. I can’t wait to see you when I’m back in the real world.

TABLE OF CONTENTS

Abstract ...... ii Acknowledgements ...... iii Table of Contents ...... iv List of Abbreviations ...... vi List of Figures ...... vii List of Tables ...... ix 1 Literature review...... 1 1.1 Bee life histories and pesticide exposure ...... 1 1.2 Pesticide risk assessment process for bees ...... 2 1.3 Exposure of bees to pesticides ...... 5 1.3.1 Behavioural differences affecting exposure profiles ...... 5 1.3.2 Physiological differences affecting pesticide exposure ...... 7 1.4 Megachile rotundata as a surrogate for solitary bee risk assessment ...... 10 1.4.1 Pesticide risk assessment protocol development for solitary bees...... 10 1.4.2 Surrogate solitary bees for pesticide risk assessment ...... 12 1.4.3 Megachile rotundata biology, life history, and production ...... 13 1.4.4 Overview of Megachile rotundata methods applicable to risk assessment ...... 17 1.4.5 Positive controls for Megachile rotundata risk assessment ...... 19 1.5 Conclusions ...... 19 1.6 Objectives ...... 20 2 Tier I pesticide risk assessment method development and generation of baseline pesticide toxicity data for Megachile rotundata ...... 21 2.1 Introduction ...... 21 2.2 Methods ...... 25 2.2.1 Test insects ...... 25 2.2.2 Pesticide treatment ...... 26 2.2.3 Statistical analyses ...... 28 2.3 Results ...... 29 2.4 Discussion ...... 32 3 Development of chronic oral laboratory toxicity assessment methods for tier I risk assessment with Megachile rotundata ...... 42

3.1 Introduction ...... 42 3.2 Introduction ...... 45 3.2.1 Test insects ...... 45 3.2.2 Treatment containers ...... 46 3.2.3 Insecticide treatments...... 48 3.2.4 Statistical analyses ...... 50 3.3 Results ...... 51 3.4 Discussion ...... 56 4 Method development for semi-field toxicity trials for use in pesticide risk assessment with Megachile rotundata ...... 61 4.1 Introduction ...... 61 4.2 Methods ...... 64 4.2.1 Test insects ...... 64 4.2.2 Planting purple tansy and field preparation ...... 64 4.2.3 Semi-field enclosure design ...... 65 4.2.4 Positive control experiment ...... 65 4.2.5 Release rate experiment ...... 71 4.2.6 Statistical analyses ...... 75 4.3 Results ...... 76 4.3.1 Positive control experiment ...... 76 4.3.2 Release rate experiment ...... 80 4.4 Discussion ...... 90 5 Conclusions and recommendations ...... 99 5.1 Potential positive control insecticides for tier I and II pesticide risk assessment using Megachile rotundata...... 100 5.2 Comparison of toxicity of a range of pesticides between Megachile rotundata and Apis mellifera...... 101 5.3 Method development for tier I and II pesticide risk assessment using Megachile rotundata females...... 102 6 References ...... 106

LIST OF ABBREVIATIONS

Alfalfa leafcutting bee (ALB)

Environmental Protection Agency (EPA)

Exposure toxicity ratio (ETR)

Hazard quotient (HQ)

Honey bee (HB)

Insect growth regulator (IGR)

International Commission for Plant-Pollinator Relationships (ICPPR)

Median lethal dose (LD50)

No observed adverse effect concentration (NOAEC)

Organisation for Economic Co-operation and Development (OECD)

Pest Management Regulatory Agency (PMRA)

Risk assessment (RA)

Risk quotient (RQ)

Society of Environmental Toxicology and Chemistry (SETAC)

Surface area to volume ratio (SA:V)

Tarnished plant bug (TPB)

vii

LIST OF FIGURES

Figure 1.1. An adult female Megachile rotundata...... 14

Figure 1.2. Megachile rotundata pupal cells removed from a nest tube (top) and broken apart by cell and placed in order (bottom)...... 15

Figure 2.1. Post-treatment containers for topical LD50 experiments using Megachile rotundata. 27

Figure 2.2. Dose response curves for mortality 72 h after contact exposure of female Megachile rodundata to dimethoate, permethrin, and imidacloprid...... 34

Figure 3.1. i) A feeding container used to house Megachile rotundata adults in a greenhouse before and during treatment...... 47

Figure 3.2. Mean (± SE) percent mortality of adult female Megachile rotundata fed 20% sucrose solution in either a greenhouse (22 ± 5.2 ºC) or controlled growth chamber (22 ± 1.5 ºC) over 28 d...... 52

Figure 3.3. Mean (± SE) percent mortality of adult female Megachile rotundata exposed ad libidum to 20% sucrose solution treated with formulated imidacloprid at 0.01, 0.02, 0.05 mg/L for 14 d ...... 54

Figure 3.4. Mean (± SE) percent mortality of adult female Megachile rotundata exposed ad libidum to sucrose solution treated with formulated dimethoate at 0.04, 0.2, 1, and 5 mg/L for 10 d...... 55

Figure 4.1. A 3.3 x 3.3 m screened enclosure in a field of blooming Phacelia tanacetifolia (left) and a Megachile rotundata nesting box (right)...... 68

Figure 4.2. Plot layout in the positive control semi-field experiment ...... 70

Figure 4.3. Examples of wing wear visible on female Megachile rotundata forewings...... 73

Figure 4.4. Plot layout in the release rate semi-field experiment ...... 74

Figure 4.5. Activities performed by adult female Megachile rotundata on Phacelia tanacetifolia plots treated with either water (control) or dimethoate at 0.2 g a.i./ha or 2 g a.i./ha...... 78

viii

Figure 4.9. Potted buckwheat plants placed in a screened enclosure on plots of Phacelia tanacetifolia containing 10 adult female Megachile rotundata after 2 d (left) and 6 d (right). .... 96

LIST OF TABLES

Table 1.1. Proposed trigger values for Apis and non-Apis tier I risk assessment...... 11

Table 2.3. Topical toxicity of three insecticides 72 hr after application to Megachile rotundata (ALB) and honey bees (HB)...... 35

Table 2.4. Hazard quotients of imidacloprid, dimethoate, and permethrin to Megachile rotundata calculated using acute topical toxicity values and the highest application rate from a formulated product containing the insecticide...... 36

Table 3.1. Results of likelihood ratio test (LRT) of model fit and Wald test for the effects of the parameters treatment, maximum daily temperature, and days after release on mortality of Megachile rotundata females fed sugar solution treated in a greenhouse with dimethoate for 10 d or imidacloprid for 14 d...... 53

Table 4.1. Model parameters and results of Wald test for several behaviour metrics of Megachile rotundata females recorded during the positive control experiment...... 79

Table 4.2. Model parameter estimates, results of Wald test, and estimated model fit for two behaviour metrics of Megachile rotundata females recorded during the positive control experiment...... 82

1 Literature review

1.1 Bee life histories and pesticide exposure

To delineate fields of study and general life histories, bees can pragmatically be classified into two broad groups: Apis and non-Apis. The only species belonging to the genus Apis in North

America is the honey bee (HB, Apis mellifera Linnaeus 1758), a non-native managed species primarily used for crop pollination and the production of hive products, such as honey and wax

(Pettis 2014). As a eusocial species, HB live in colonies containing a single reproductively active queen and thousands of infertile workers (Pettis 2014). Honey bees historically have been used as the surrogate test species for pesticide risk assessments (RA) for bees due to their ease of acquisition, importance in pollinating commercial crops, and relatively high sensitivity to pesticides (Pettis 2014).

Non-Apis bees in North America can be broadly divided into three groups: solitary cavity-nesting bees, solitary ground-nesting bees, and social bumble bees (EFSA 2013, Sgolastra et al. 2018). Some species of solitary bees, such as the cavity-nesters Osmia bicornis (Linnaeus

1758) and Megachile rotundata (Fabricius 1787), and the ground-nester Nomia melanderi

(Cockerell 1906) are managed for commercial crop pollination in North America (Richards

1993, Michener 2007). Bumble bees (Bombus sp.) are also extensively managed for greenhouse tomato pollination (Richards 1993). The more than 4,000 other bee species in North America

(Michener 2007, Wilson and Messinger Carril 2016) pollinate wild plants and provide crop pollination services without direct management (Richards 1993, Belfrage et al. 2005, Loose et al.

2005, Smagghe and Calderone 2012).

Recent declines in non-Apis bee abundance and species richness have been partially attributed to agricultural pesticides (Goulson et al. 2015, Koh et al. 2015, Potts et al. 2015, Leach

2 and Drummond 2018). However, examples of a direct correlation between pesticide use and reduced bee abundance in the field are limited and controversial. Rundlöf et al. (2016) observed that exposure to clothianidin through seed-treated canola reduced Bombus terrestris (L. 1758) colony strength, O. bicornis nesting, and wild bee density, but did not affect HB colony strength.

Park et al. (2015) observed a negative correlation in wild bee species richness with increasing insecticide and fungicide use in apple orchards. Woodcock et al. (2017), however, observed both negative and positive effects of neonicotinoids on HB, O. bicornis, and B. terrestris populations, depending on the country in which the study was conducted. Such studies highlight the need for a deeper understanding of the effects of pesticides on non-Apis bees, and how they interact with other factors, such as habitat complexity (Williams and Kremen 2006, Woodcock et al. 2017), anthropogenic activities (Kim et al. 2006), and parasite pressure (Potts et al. 2015).

1.2 Pesticide risk assessment process for bees

The Pest Management Regulatory Agency (PMRA) in Canada and Environmental

Protection Agency (EPA) in the United States aim to limit harmful, non-target environmental effects of pesticides through the regulation of pesticide registration and application (OECD

1998b, a, 2007, EPA 2012, 2014). Pesticide RA using HB is the primary source of data in North

America when incorporating bee safety into regulations (EPA 2012). For the purposes of this review, HB RA will refer to those practices standardized by the Organisation for Economic Co- operation and Development (OECD) and implemented by the EPA and PMRA (OECD 1998a, b;

OECD 2007; EPA 2012, 2014). Essential to this process is estimating the probability of exposure to a hazard and the effects of that hazard, from which a risk quotient (RQ) can be determined

(EPA 2012). If the risk posed to HB under field scenarios is found to be too high, a pesticide can be banned, restricted, or called back for re-formulation (Alix et al. 2014). Results from pesticide

RA can be extrapolated to approximate if a pesticide will negatively affect the economic yield associated with HB pollination and assess the potential environmental risk to other bees and unmanaged pollinators (Thompson and Hunt 1999, Alix et al. 2014).

HB RA is a three-tiered system that progresses from laboratory work up to field trials as described in EPA (2012, 2014). In turn, the tiers progress from highly controlled, low cost, restricted experiments (lab trials), to difficult-to-control, expensive, but realistic experiments

(field trials). Progression occurs from lower to higher tiers when tier-specific trigger values are reached. In this manner, low-risk pesticides are identified in relatively inexpensive laboratory experiments (tier I) while only those pesticides that may pose a risk to bees are further tested in semi-field (tier II) or field (tier III) experiments.

A tier I risk assessment involves determining the acute contact and oral median lethal dose (LD50) for individual adults and larvae, and chronic oral exposure experiments if required

(EPA 2012). Colony-level effects are not examined in these tests. Assessments are made with active ingredients and/or formulated product depending on the expected use patterns and chemistry of the test product, and predicted exposure scenarios are taken into account.

Additional routes of exposure, such as contaminated wind-blown pollen, water contamination, or dust from seed planting, are not examined in this tier. Progression to higher tiers in the United

States and Canada is triggered when the risk quotient (RQ), the ratio of the estimates of exposure and effects at a certain point, exceeds the level of concern (LOC), which is usually an estimate of exposure that will result in 10% mortality.

Tier II trials (EPA 2012) expand the scope of assessment to the entire colony in confined conditions. They can be used to generate data that tier I trials cannot, such as the sub-lethal effects on adults, effects on larvae in a colony, and exposure via multiple routes. The bees are

4 allowed to forage on treated plants outdoors, but are confined to an enclosure to ensure continuous exposure to the treatment. Under these conditions the colony is exposed through several routes at once, including diet, contact with treated foliage, and inhalation in the case of volatilizing compounds. The exact methods used for any tier II experiment depend on the specific concerns about the product and how it may affect the bees, but generally, colonies are placed in enclosures with a bee-attractive plant. When the plants are flowering, they are treated with the test pesticide at recommended rate(s), an appropriate positive control, and a negative control (OECD 2007). Common assessment endpoints for the entire colony are brood development, colony strength, and flight activity. Tier II trials are less expensive and easier to replicate than field studies, although the constant exposure to a treated monoculture represents a worst-case scenario. The duration of the experiment is also limited because HB colonies decline in health after 2 wk in enclosures (EPA 2012). This time restriction prevents the assessment of overwintering ability or long-term monitoring. There are no set trigger values in tier II trials to determine if it is necessary to progress to tier III. Instead, it is decided on the basis of the individual trials, chemistries, and use patterns whether more data need to be gathered in a realistic scenario.

Tier III studies (EPA 2012) are used to assess uncertainties identified in tier I and II studies by exposing colonies to maximum label rate applications in an unrestricted, realistic field scenario. The design of a tier III study depends on the knowledge gaps being addressed, and the chemistry and proposed use patterns of the pesticide being studied (EPA 2012). Therefore, protocols for Tier III studies must be designed and reviewed individually based on the purpose of the study (EPA 2014). Assessment endpoints may include refined tier II metrics as well as disease levels and overwintering success. In some cases, non-Apis bees may be studied in

5 addition to or in substitute of HB in a relevant context (EPA 2014). Due to their size and duration, these studies are expensive, time-consuming, and subject to more variability (e.g., differences in weather patterns within and between years and environmental differences between field sites). When conducted correctly, tier III studies can provide the most realistic estimates of the effects of pesticide exposure to bees (EPA 2012).

There is concern over the protectiveness of the current pesticide RA framework for solitary bees (EFSA 2013, Vaughn et al. 2014, Knäbe et al. 2017, Sgolastra et al. 2018). It was previously accepted that conservative experiments with HB using worst-case exposure scenarios in tier I trials (EFSA 2013) would be sufficient to protect solitary bees in most cases. However, current research suggests that differences in HB and solitary bee sociality, life history, and physiology may be too disparate for current risk estimates to be consistently protective beyond tier I experiments (EFSA 2013, Fischer and Moriarty 2014b, Sgolastra et al. 2018). In light of these concerns, other avenues to assess pesticide risk to solitary bees are being explored.

1.3 Exposure of bees to pesticides

1.3.1 Behavioural differences affecting exposure profiles

In general, bees can be exposed to pesticides via bodily contact with sprays or dust, either directly or as residues on plants, or contaminated nesting materials (e.g., wax, soil); or via the consumption of contaminated pollen or nectar (Wisk et al. 2014). However, the relative importance of these routes of exposure can be different for solitary bees and HB due to differences in life history (Wisk et al. 2014, Sgolastra et al. 2018). For instance, the highly lipophilic wax in HB colonies has been found to contain a range of pesticide residues (Chauzat and Faucon 2007). Non-foraging individuals that never leave the hive (i.e., queens, nurse bees, and larvae) may be exposed to these low-level residues in the wax (EPA 2014). The effects of

6 contaminated wax on HB have not been well studied, and the wax may in fact be sequestering residues and subsequently reducing exposure. In contrast, many solitary bees construct their nests with materials collected in or near field margins that may contain higher residues than those found in HB wax (Johansen et al. 2014). Examples of solitary bee nesting materials include fresh leaf pieces collected by leafcutting bees in the genus Megachile (George and Rincker

1982), mud collected by bees in the genus Osmia (Bosch and Kemp 2001), and leaf pubescences collected by wool-carder bees in the genus Anthidium (Gibbs and Sheffield 2009). For these bees, eggs and larvae will be in direct contact with the contaminated nesting materials until eclosion, resulting in prolonged contact exposure to pesticide residues and a potentially higher dose than that received by HB from wax (EFSA 2013, Sgolastra et al. 2018).

The location and composition of nesting sites also dictate differences in pesticide exposure between HB and solitary bees. Cavity-nesting bee exposure can occur through contaminated plant matter such as reeds or dead wood (Wisk et al. 2014), and ground-nesting solitary bees can be exposed to pesticide residues in soil for their entire development (Kim et al.

2006, Sgolastra et al. 2018). Solitary bees may also nest directly in or adjacent to agricultural landscapes, increasing their probability of insecticide exposure (Kim et al. 2006, Rundlöf et al.

2016, Sgolastra et al. 2018).

Sociality and colony organization can also affect pesticide exposure. Most notably, HB queens are usually protected within their nests from direct exposure to pesticide sprays or foliar residues, while the non-reproductive worker class assumes all the exposure risks associated with foraging (Dobrynin and Colombo 2007, Biddinger and Rajotte 2015). Honey bee workers only forage for the last few weeks of their lives, and therefore they experience the greatest exposure to pesticide residues when they are already at the end of their lives (Pettis 2014, Vaughn et al.

2014). In contrast, solitary bee females solely perform all foraging and nest-building activities and are exposed to pesticides for their entire adult lives (Sgolastra et al. 2018). Therefore, solitary bee females will experience comparatively higher pesticide exposure than HB queens and workers (EFSA 2013, Pettis 2014). Additionally, HB workers number in the tens of thousands per colony, and HB colonies can endure occasional losses of large numbers of workers without a noticeable decline in queen productivity (EFSA 2013, Vaughn et al. 2014). In contrast, the loss of each solitary bee female results in the loss of her potential offspring and could contribute to significant population declines for solitary bees (Thompson and Hunt 1999,

Sgolastra et al. 2018).

1.3.2 Physiological differences affecting pesticide exposure

Body size can have a significant effect on the dose of pesticide that bees are exposed to

(Uhl et al. 2016, Sgolastra et al. 2018). Large bees like Bombus spp. have a relatively small

SA:V, while smaller bees have a larger SA:V. The result is a higher dose per gram of body mass received from surface pesticide residues for smaller solitary bees such as M. rotundata, Osmia spp., and N. melanderia, compared to larger bees such as HB or Bombus spp. (Dobrynin and

Colombo 2007, Wisk et al. 2014).

Several studies have demonstrated that the susceptibility of bees to insecticides depends on body size. For instance, Devillers et al. (2003) reported that the susceptibility of HB and non-

Apis bees was inversely proportional to body size. Uhl et al. (2016) also found that body mass was predictive of dimethoate toxicity, although it did not entirely explain differences in toxicity between different bee species. Arena and Sgolastra (2014) reviewed the susceptibility of 18 species of bees from 44 publications and calculated a sensitivity ratio comparing HB to solitary bees in the lab. While there were some species that appeared to be far more sensitive to

8 pesticides than HB, approximately 95% of the species were still within the 10-fold assessment factor currently used in HB RA to predict risk for other bees (EFSA 2013). This assessment factor attempts to extrapolate non-Apis endpoints from HB endpoints in the lab by using a conservative value 1/10 of that found to be protective of HB. While promising for the protectiveness of HB RA to solitary bees, these results are only for laboratory work and do not incorporate any of the other factors that may affect pesticide exposure for bees in the field. Arena and Sgolastra (2014) also suggest that two economically important solitary bee species (M. rotundata and O. lignaria) are > 10 times as sensitive than HB to certain pesticides. To reach a stronger conclusion about the role of body mass in the relative sensitivity of solitary bees to pesticides, more bee species will need to be tested and life history traits will need to be considered outside of the laboratory (Arena and Sgolastra 2014, Wisk et al. 2014).

Body mass is not the only predictor of a species’ sensitivity to pesticides. Susceptibility to pesticides also depends on a species’ metabolism, detoxification pathways, and tolerance to a given pesticide and will vary, independently of body size, depending upon the species and chemistries involved (Dobrynin and Colombo 2007, Heard et al. 2017). In some cases, smaller bees, such as M. rotundata and N. melanderi, have been found to be 10 times less sensitive than

HB to a range of pesticides after accounting for differences in body size (Mayer et al. 1990,

Helson et al. 1994). Conversely, O. lignaria has been shown to be far more susceptible to the fungicide captan than the similarly-sized HB when given the same topical dose (Ladurner et al.

2005b), while M. rotundata may be even more susceptible to ingested captan than O. lignaria

(Huntzinger et al. 2008b, Huntzinger et al. 2008a). This variability between individual species makes it difficult to develop robust models to predict susceptibility to pesticides for all bees.

To reduce the probability of pesticide exposure to HB, pesticide regulations often limit applications to times of the day, such as the morning and evening, when HB are not actively foraging (Thompson and Hunt 1999). However, some solitary bees forage outside of this window and may be exposed to sprays timed to avoid HB (Thompson and Hunt 1999). Unlike social bees, solitary bee adults may not return to their nest for days at a time, and the males of many species may never reside in a hive (Pitts-Singer and Cane 2011, Wisk et al. 2014). In these cases, adults that spend the night in an agricultural area have a much higher chance of being exposed to direct pesticide sprays in the evening and early morning.

Differences in foraging distance and subsequent exposure patterns between solitary bees and HB are also not addressed in the current pesticide RA process (Vaughn et al. 2014, Sgolastra et al. 2018). HB can forage up to 10 km from the hive and are likely to encounter treated and untreated plants while foraging. In contrast, solitary bees are limited to a much smaller foraging distance of approximately 100 m, and when they nest near an agricultural area, they are likely to encounter comparatively more treated plants (Zurbuchen et al. 2010, Vaughn et al. 2014).

Additionally, many solitary bee species, such as squash bees (Peponapis spp.) or the alfalfa leafcutting bee, are specialists and only forage on a limited number of floral species (Kim et al.

2006, Pitts-Singer and Cane 2011). This may restrict their exposure to treated plants even further.

Honey bees alone are unlikely to provide an accurate representation for pesticide RA when considering the wide range of solitary bee physiologies, life histories, and protection goals.

To address this deficit, there has been work to design RA protocols for select solitary bee species to provide more inclusive protection (EFSA 2013, Cabrera et al. 2016, Knäbe et al. 2017, OECD

2017c, Sgolastra et al. 2018).

1.4 Megachile rotundata as a surrogate for solitary bee risk assessment

1.4.1 Pesticide risk assessment protocol development for solitary bees

There are currently no validated RA protocols for solitary bees (EFSA 2013, Sgolastra et al. 2018). Data from HB RA are usually extrapolated to roughly account for biological differences between HB and solitary bees (EFSA 2013). In cases where solitary bees are tested directly for a RA, methods for HB are modified for solitary bees (Sgolastra et al. 2018). These modified methods are not rigorously tested and thus are prone to unforeseen errors and poor design (EFSA 2013, Knäbe et al. 2017). Both extrapolation of data and insufficiently tested modification of HB methods will lead to inaccuracies and be untenable for widespread use if solitary bees are to be included in the RA process. For this reason, protocols are being designed across the European Union and North America to produce standardized RA methods for solitary bees (EFSA 2013, Vaughn et al. 2014, OECD 2017b). Progress is being made to develop semi- field methods for Osmia spp. (Knäbe et al. 2017), and these protocols can be used as a scaffold for other solitary bee RA development. Preliminary trigger values for solitary bee tier I studies in the European Union are presented in Table 1.1.

Table 1.1. Proposed trigger values for Apis and non-Apis tier I risk assessment. Hazard quotient (HQ) = application rate (g a.i./ha)/median lethal dose (ng a.i./bee), exposure toxicity ratio (ETR) = exposure/toxicity (EFSA 2013).

Trigger value Apis mellifera Solitary bees Bombus spp. Acute oral toxicity ETR < 0.2 ETR < 0.04 ETR < 0.036 Acute contact toxicity HQ < 42–85 HQ < 8–16 HQ < 7–14 Chronic oral toxicity ETR < 0.03 ETR < 0.0054 ETR < 0.0048 Larval toxicity ETR < 0.2 ETR < 0.2 ETR < 0.2

1.4.2 Surrogate solitary bees for pesticide risk assessment

Many solitary bees possess some traits that make them well-suited to RA testing. The relatively small foraging range of most solitary bees ensures they will feed only from plants in close proximity, which increases the probability that the bees are feeding from the treated crop in a tier III trial (Zurbuchen et al. 2010, Vaughn et al. 2014). Further, solitary bees appear less stressed than HB when confined to enclosures and may exhibit a more natural response to pesticides during tier II tests (Vaughn et al. 2014). It is also easy to manipulate solitary bees, and mark and track each bee during a study, as most species are univoltine or bivoltine and thus will not produce a second generation of adults during the course of a trial (Sheffield et al. 2011).

Several solitary species in the family Megachilidae have been suggested for RA in North

America and Europe based on their commercial availability and success in temperate regions

(Vaughn et al. 2014), including the alfalfa leafcutting bee (Megachile rotundata) and the blue orchard bee (Osmia lignaria Say 1837) for North America; and the horned-face bee (Osmia cornifrons Radoszkowski 1887), the red mason bee (Osmia rufa L. 1758) and the European orchard bee (Osmia cornuta L. 1805) for Europe (Vaughn et al. 2014). Current efforts aim to gather more data to make an informed selection of a representative for RA in each region and to promote RA method development for solitary bees (Vaughn et al. 2014, Sgolastra et al. 2018).

The life cycle of megachilid bees is easy to control for laboratory use because of their obligatory winter diapause and sturdy brood cell construction (Fairey et al. 1987, Rinehart et al.

2016). Brood cells containing overwintered pre-pupae can be incubated as needed to produce a predictable amount of adult bees at almost any time of the year indoors (Fairey et al. 1987). The alfalfa leafcutting bee (ALB, Megachile rotundata) in particular is especially well suited for RA in North America.

1.4.3 Megachile rotundata biology, life history, and production

Native to Europe, ALB (Figure 1.1) was accidentally introduced into North America in

1940, and it is now used for commercial pollination of several crops in Canada and the United

States, as well as Argentina, Australia, Chile, and New Zealand (Michener 2007, Pitts-Singer and

Cane 2011). In 2009, ALB contributed to the production of $4–7 billion of alfalfa seed (and associated alfalfa hay) in the United States (Smagghe and Calderone 2012).

Juvenile ALB spend all five larval instars and pupal stages inside cells fashioned from leaf disks cut and put together by their mothers (Trostle and Torchio 1994). These cylindrical cells are built piece by piece inside a cavity of 5 mm in diameter, such as a hollow stem or beetle exit cavity in a tree (Richards 1984, Pitts-Singer and Cane 2011, Sheffield et al. 2011). Each nesting cavity can contain up to 10 cells arranged end-to-end (Figure 1.2). Eggs are laid on pollen and nectar masses collected by the mother and will be the only source of nutrients for the developing larvae (Trostle and Torchio 1994). After developing into pre-pupae, larvae enter obligate diapause for the coming winter (Trostle and Torchio 1994, Pitts-Singer and Cane 2011).

This diapause will last 10–11 months in the wild, depending on ambient temperatures (Kemp and

Bosch 2000). After overwintering as cocooned pre-pupae, ALB develop into pupae and emerge as adults after several weeks (Richards 1984, Pitts-Singer and Cane 2011).

Figure 1.1. An adult female Megachile rotundata.

6 5 4 3 2 1

Figure 1.2. Megachile rotundata pupal cells removed from a nest tube (top) and broken apart by cell and placed in order (bottom). The deepest cell in the nest (1) is on the right, and the cell closest to the nest tube entrance (6) is on the left.

As in many bee species, male ALB have no function associated with the production of offspring other than to mate with females (Pitts-Singer and Cane 2011). Males will wait for females by nesting cavities and on flowers, vigorously attempting to couple with both females and other males that fly by (Rossi et al. 2010). A female will always resist mating attempts and will only successfully mate with one male in her lifetime (Blanchetot 1992, Fischer and Moriarty

2014a). At high male densities, sexual harassment can slow the reproductive activities of females

(Rossi et al. 2010).

Females will nest gregariously (Kemp and Bosch 2000), and while they do not cooperate in nest creation, it is possible for more than one female to consecutively occupy a nest over a single season (McCorquodale and Owen 1994). Reductions in offspring production have been observed at high female densities, potentially because of competition over nesting sites (Mayer

1994). Other external factors such as low-quality forage (Horne 1995a), long foraging distances

(Peterson et al. 2006), and cool rearing temperature (Tepedino and Parker 1986) can contribute to reduced progeny production, size, and provisioning (Klostermeyer et al. 1973, Peterson and

Roitberg 2006). ALB is partially bivoltine in North America, meaning two generations may be produced in one season (Tepedino and Parker 1988). The second generation of adults consists of a higher proportion of females than the first generation, and the adults are physically smaller than those from the first generation (Richards 1984, Tepedino and Parker 1988). The exact factors underlying second-generation emergence are unknown, but it has been suggested that poor larval nutrition or increased temperatures may initiate early adult emergence (Tepedino and Parker

1988, Horne 1995a, Pitts-Singer and Cane 2011). Second-generation bees rarely have a chance to reproduce in North American climates and appear to be of no benefit to the population (Richards

1984). A high proportion of second-generation bees therefore reduces the number of reproductive individuals for the next season.

For commercial pollination, overwintered ALB pre-pupae are released in large quantities into alfalfa fields before bloom: this allows the bees time to mature and emerge with the start of bloom (Richards 1984). Sheltered nesting boxes are placed throughout the field prior to bee emergence (Richards 1984). These boxes are made of wood or polystyrene and contain thousands of evenly spaced 5 mm diameter tunnels (Fairey et al. 1984, Richards 1984, Pitts-

Singer and Cane 2011). At the end of the season the nest boxes are collected and the brood cells are removed and stored loosely (Richards 1984). The brood cells are then artificially overwintered at 6 oC for 7–9 months (Richards et al. 1987) after which they can be emerged under artificial conditions at 29–30 oC and 50% RH (Kemp and Bosch 2000, Rinehart et al.

2016). Emergence can be delayed with no additional mortality for many months, but brood cell storage for more than a year significantly reduces successful emergence of adults (Richards et al.

1987, Rinehart et al. 2016).

1.4.4 Overview of Megachile rotundata methods applicable to risk assessment

While there are currently no standardized methods for pesticide RA with ALB, there are experimental methods that can be incorporated with proposed Osmia and Bombus protocols to build a RA framework (Knäbe et al. 2017, OECD 2017c). These methods are varied and inconsistent between studies but provide a number of potential protocols to assess ALB susceptibility to pesticides. There are several methods to assess the contact toxicity of pesticides to ALB that could be incorporated into tier I RA, including application to individuals using a pipette (Helson et al. 1994), and exposure to residues via treated filter paper (Huntzinger et al.

2008b) or foliage (Mayer et al. 1997). There are few methods to expose ALB orally to pesticides,

18 and they usually involve allowing adults to feed ad libitum from treated sugar solutions

(Huntzinger et al. 2008b) or carbaryl bran baits (Gregory et al. 1992, Peach et al. 1995). Methods used to orally treat Osmia spp. with specific doses (Ladurner et al. 2003, Ladurner et al. 2005a) are difficult to replicate with ALB and will need to be modified if they are included in ALB RA development. Methods to orally dose larvae in the lab involve treatment of the pollen provision of completed brood cells (Huntzinger et al. 2008a, Gradish et al. 2012).

Methods for ALB that could be developed into tier II and III protocols follow a similar framework as RA for HB. However, many of the endpoints used to assess HB (especially those related to sociality) do not apply to solitary bees. There are few field study protocols with ALB that could serve as a framework for developing tier II or III RA methods. The goal of many of the studies that have been performed was maximizing ALB production in alfalfa fields and minimizing loss due to pesticides, and as such the focus was on reproductive output and larval survival (Torchio 1973, George and Rincker 1982, Tasei and Carre 1985, Pitts-Singer and

Barbour 2016). If these methods are to be reworked for use in RA, additional metrics may need to be incorporated to fully account for the goals of a RA.

When adult ALB behavioural (sub-lethal) endpoints are examined in semi-field enclosures, they are done so by observations that are not feasible in a regulatory context. These studies focus on behaviours such as foraging time and flight time for individual bees throughout the day (Alston et al. 2007, Artz and Pitts-Singer 2015) or nest creation of only a few bees

(Klostermeyer and Gerber 1969). These methods would be too time consuming if expanded to produce the robust sample sizes required for RA but may work as a basis to modify current

Osmia semi-field protocols (Knäbe et al. 2017) for use with ALB.

1.4.5 Positive controls for Megachile rotundata risk assessment

A complete RA framework requires positive controls for tiers I and II (EFSA 2013). A positive control is a pesticide that will produce a known response when applied that can be used to confirm exposure in an experiment. There is not enough data on any pesticide that could be used as a positive control for ALB. HB, Bombus, and Osmia RA development all incorporate dimethoate as a positive control for tier I and II studies (EPA 2012, OECD 2017c, Sgolastra et al.

2018), and so dimethoate may be a good choice as a positive control for ALB RA to maintain consistency across RA protocols. However, an LD50 for dimethoate applied to ALB has only been estimated in one laboratory study (Torchio 1973), and dimethoate has not been examined in a semi-field setting. Future testing is needed to determine a suitable positive control for ALB

RA.

1.5 Conclusions

It is clear that the differences in life history and biology between HB and solitary bees potentially limit current bee RA practices from adequately protecting solitary bees. RA protocols will therefore need to be created for solitary bees. To achieve this, a larger body of toxicity and life history data will need to be generated for several representative solitary bee species. Progress has been made for semi-field work with Osmia spp. (Knäbe et al. 2017), but more species should be examined to adequately represent solitary bees in North America (Vaughn et al. 2014). Future developments should focus on determining the association between solitary bee mass and pesticide susceptibility to account for the potential variety of doses received by individual solitary bees, developing a positive control to use in tier I and II trials, developing species- specific methods for solitary bee lab and field research, and quantifying solitary bee activity and resource consumption to model exposure more precisely (OECD 2017c, Sgolastra et al. 2018).

Basic toxicological data must also be produced to develop a baseline understanding of the representative species’ response to pesticides (Vaughn et al. 2014).

ALB is an optimal choice for a surrogate solitary bee species for RA in North America.

Its biology is generally well understood (Vaughn et al. 2014), it thrives in the North American climate, commercial production of ALB provides a nearly limitless supply of individuals all year round, and they are easy to manipulate and observe in the lab and field.

1.6 Objectives

The goal of my research was to contribute to method development for pesticide RA with

ALB. I had three major objectives:

1. Generate toxicity data for potential positive controls for tier I (lab) and tier II (semi-field)

pesticide risk assessment using ALB (chapters 2, 3, and 4),

2. Compare the toxicity of a range of pesticides to ALB and to HB (chapters 2 and 3), and

3. Improve the efficiency and precision of tier I (lab) and tier II (semi-field) methods

applicable to pesticide RA using ALB (chapter 2, 3, and 4).

2 Tier I pesticide risk assessment method development and generation of

baseline pesticide toxicity data for Megachile rotundata

2.1 Introduction

Many solitary bees, including some managed species, are specialist foragers that pollinate certain crops, such as alfalfa, sweet cherry, apple, and cucurbits, with higher efficiency than honey bees (HB, Apis mellifera Linnaeus) (Vaughn et al. 2014). Solitary bees are also important pollinators in natural ecosystems (Vaughn et al. 2014). Some solitary bee populations are in decline, and these declines have been attributed to a variety of factors, including habitat loss and pesticide use (Goulson et al. 2015, Koh et al. 2015, Potts et al. 2015, Leach and Drummond

2018).

Solitary bee population declines have raised concerns about our ability to predict the effect of pesticides on solitary bees. Current pesticide risk assessment (RA) practices in North

America and the European Union focus almost exclusively on HB and may not sufficiently protect solitary bees (EFSA 2013, EPA 2014). Pesticide RA fails to account for the characteristic differences in life history, physiology, and behaviour between solitary bees and HB that may lead to increased pesticide susceptibility or exposure in solitary bees (Sgolastra et al. 2018).

International efforts are underway to develop pesticide RA protocols that will focus on solitary bees and provide a more accurate representation of their pesticide susceptibility and routes of exposure (EFSA 2013, EPA 2014, Fischer and Moriarty 2014b, OECD 2017c, Sgolastra et al.

2018). As it is unfeasible to produce regulatory guidelines for every species of solitary bee, surrogate species will need to be selected for different regions (Vaughn et al. 2014). The alfalfa leafcutting bee (ALB, Megachile rotundata Fabricius 1787) has been suggested as a potential surrogate for North American solitary bees as its biology and behaviour are well understood, it is

22 easy to work with in the lab, and individuals are commercially available in large quantities (Pitts-

Singer and Cane 2011).

HB RA is a three-tiered system involving tier I laboratory, tier II semi-field, and tier III field trials (OECD 1998a, b; EPA 2016). Tier I HB RA is used as a screening tool to quickly assess pesticides that are unlikely to be toxic and, if certain trigger values are reached, potentially harmful products are moved to higher levels of testing (EFSA 2013, Frazier et al. 2014, EPA

2016).

A well-characterized positive control is integral to tier I pesticide RA: By producing a known toxic effect in the test organism, the positive control is used in parallel with the pesticide under examination to confirm exposure in the experiment (EPA 2012). Here I generated topical

LD50 data for adult ALB to contribute to the characterization of dimethoate, permethrin, and imidacloprid as positive controls for tier I trials with ALB. Dimethoate is an organophosphate insecticide used as a positive control in HB RA and suggested as a positive control for Osmia and Bombus RA (OECD 1998b, 2017b, Uhl et al. 2016). Using dimethoate in ALB RA would maintain continuity in positive controls across all bee pesticide RA. Permethrin is a commonly used pyrethroid insecticide that has high acute toxicity to ALB (Helson et al. 1994) that makes it an excellent candidate as a positive control. Imidacloprid is a neonicotinoid insecticide widely used in seed treatments and as a foliar spray that has systemic activity. EFSA (2012) suggests that a systemic insecticide be used as a positive control, and current data suggests that it is highly toxic to ALB (Scott-Dupree et al. 2009), making it a good candidate as well.

Another integral step to address before developing tier I toxicity assessment methods for

ALB is to generate baseline pesticide toxicity data with ALB. These data will address basic knowledge gaps, including how ALB respond to pesticide exposure compared to HB in the lab.

To this end, I generated acute topical toxicity data for a range of pesticides applied to ALB.

These data are integral to develop an understanding of the topical toxicity of pesticides to ALB, which is limited (Torchio 1973, Helson et al. 1994, Mayer et al. 1998, Mayer and Lunden 1999,

Scott-Dupree et al. 2009, Gradish et al. 2012).

To better understand the response of ALB to a variety of pesticides, I generated toxicity data for two other pesticides with different modes of action: chlorantraniliprole, an anthranilic diamide insecticide that has not been tested with ALB; and captan, a fungicide previously reported as being orally toxic to ALB (Huntzinger et al. 2008b). I then compared these values to those generated for dimethoate, permethrin, and imidacloprid to assess the comparative toxicities of these pesticides to ALB.

Differences in bee body mass influence the dose received (dose/g bee) from a fixed exposure to pesticide (Klostermeyer et al. 1973, Peterson and Roitberg 2006, Peterson et al.

2006). Given that the effective dose of a pesticide received by an individual is inversely proportional to its body mass. Specifically, it can be expected that larger individuals will receive a smaller dose of pesticide and are likely to be less susceptible than smaller individuals. For instance, Uhl et al. (2016) found body size to be a predictive, though not constant, factor for sensitivity to topically applied dimethoate and estimated that Nomioides minutissimus is 128 times more susceptible to dimethoate than Bombus impatiens while only being 69 times smaller on average. In the same study, Lassioglossum malachurum was found to be 25 times more susceptible than B. impatiens to topically applied dimethoate and is 15 times smaller on average.

Devillers et al. (2003) also reported the susceptibility of HB and non-Apis bees to pesticides was inversely proportional to body size across species.

Body mass also affects the susceptibility to pesticides between different-sized individuals of the same species. Therefore, in pesticide RA scenarios where all individuals in a treatment receive the same amount of active ingredient, a higher variance in body mass within a population will result in a higher variance in the response to the treatment. ALB displays a higher variation in body mass between adult individuals than HB (Helson et al. 1994). This larger mass variation of ALB adults may be due to differences in environmental conditions and provision nutrient quality experienced by individual ALB larvae (O'Neill et al. 2010, Pitts-Singer 2015), whereas

HB are continuously provisioned and raised in cells of a consistent size (Pettis 2014). If this variation in induvial mass is not captured (as it is not in current HB RA methods), it is likely to decrease the precision of toxicity estimates. Only one study has assessed the significance of individual mass on pesticide toxicity estimates for ALB (Helson et al. 1994), and they found no significant effect of individual body mass in the models for six topically applied insecticides applied to four bee species, including ALB and HB. Further research should be done to fully assess the effect of body mass on the precision and consistency of these methods to assess toxicity.

The objectives of my study were to:

1. Assess dimethoate, permethrin, and imidacloprid for use as positive controls in tier I

RA with female ALB,

2. Compare the toxicity of a range of pesticides between ALB and HB to further

understand the differences in susceptibility to pesticides between these two species,

and,

3. Determine if incorporating individual body mass in tier I pesticide RA methods for

ALB increases the precision estimates of toxicity (LD50, 95% confidence interval,

model fit), and if so, determine what is the most efficient way to incorporate body

mass.

2.2 Methods

2.2.1 Test insects

ALB pre-pupae were purchased from NorthStar Seed Ltd. (Manitoba, Canada) in

November the year before the bees were used, and stored at 8 oC. Bees were not used beyond 10 months of purchase to ensure high-quality adults (Richards et al. 1987). When required, pre- pupae were placed in a plastic container and stored in a temperature-controlled growth cabinet at

27–30 oC., 60% RH, and 12:12 h light and dark until adult emergence (3–4 wk later). Large holes were cut in the lid of each container and covered on the inside with fine wire mesh to allow for ventilation and prevent the bees from escaping. The holes were covered externally with fine polyester noseeum fabric netting (Skeeta, Florida, USA) to prevent parasitoid wasps from transferring between containers. Containers were checked every other day for parasitoids, and all visibly parasitized cells and adult parasitoids were removed. In the second year, containers were placed in bags made from breathable low tunnel greenhouse screen to further restrict parasitoid mobility. Upon emergence, female bees were stored in groups of 10 in treatment cups (Figure

2.1) and fed 20% sucrose solution in a growth cabinet at 25 oC, 50% RH, and 12:12 h light and dark for 3 d to acclimate to experimental conditions. Treatment cups were composed of two 500 mL clear plastic cups, a piece of dental wick soaking in a reservoir of 20% sucrose solution as a food source, and a bent piece of aluminum mesh to provide a surface for the bees to stand on.

The feeding method proposed by Torchio (1973), which consisted of an inverted sugar feeder on the top of the container, did not elicit feeding from any bees in preliminary experiments.

2.2.2 Pesticide treatment

Technical grade (90–100% purity) dimethoate, permethrin, imidacloprid, chlorantraniliprole, and captan were used in the experiments (MilliporeSigma, Ontario, Canada).

Technical grade pesticides were used because they are available in all countries, but specific formulated products may be unavailable in certain countries. All pesticide dilutions were made with acetone and performed fresh each day. Bees were anesthetized with CO2 for 40 s in groups of 10. While anesthetized, 1 µL of pesticide dissolved in acetone was applied to the dorsal thorax of each bee using a micropipette. Controls were treated with acetone only. The bees were treated in a randomized complete block design where blocks were separated across days. Treated bees were placed in new feeding containers stocked with 20% sucrose solution and returned to the growth chamber at 25 ºC, 50% RH, and 12:12 h light and dark. Feeding containers were placed randomly in the growth chamber. Mortality was recorded daily. Bees were considered dead if they did not respond to a gentle squeeze on the thorax with forceps. After 3 d, all bees were placed in the freezer for at least 1 d, washed in de-ionized water to remove sucrose residues, desiccated thoroughly in a drying oven at 47 ºC, and weighed. These methods were adapted from current HB tier I RA (OECD 1998b) to be more suitable for ALB, such as a different feeding container design and adjustments to the growth chamber temperature, humidity, and lighting.

B C

F D

Figure 2.1. Post-treatment containers for topical LD50 experiments using Megachile rotundata. A) screened lid, B) outer 500 mL clear cup containing sugar solution, C) inner 500 mL clear cup with hole in center for dental wick, D) sugar solution, E) aluminum mesh, and F) sterile dental wick feeder soaking in sugar solution.

2.2.3 Statistical analyses

Data were analyzed using two methods of incorporating body mass, as well as one method not incorporating body mass, to identify the assessment method that optimizes both model precision and time required for weighing. These three assessment methods are as follows:

1. Not incorporating mass into the analysis, the simplest and fastest approach;

2. Including the mean mass of all the bees in each treatment replicate of 10 bees, a more

time-consuming process that estimated some of the mass differences in the

population;

3. Incorporating the mass of each individual bee, a process that added at least several

hours of labour to each experiment but was expected to provide the best estimate of

individual mass.

The same data were used in all three methods to maintain consistency: the mean bee mass per group of 10 bees was incorporated as an explanatory variable in the second method, and the mass of every individual bee was included as an explanatory variable in the third method. For all methods each bee was considered an experimental unit as the bees were treated individually and kept in nearly identical post-treatment containers.

All significance values were tested at α = 0.05. Data were analyzed using a generalized linear model in R with the glm function with a binomial distribution and probit transformation (R

Core Team 2017). The doses were not transformed to the log scale for analyses. The significance of each explanatory variable was tested with a Wald test (Agresti 1990) using the anova function

(test = “Chisq”) in R (R Core Team 2017). Model fit was approximated by calculating the pseudo R2 using Equation 2.1 (Alain et al. 2009).

푁푢푙푙 푑푒푣𝑖푎푛푐푒−푟푒푠𝑖푑푢푎푙 푑푒푣𝑖푎푛푐푒 푝푠푒푢푑표 푅2 = 100 푥 [2.1] 푛푢푙푙 푑푒푣𝑖푎푛푐푒

Control mortality was corrected for using the Henderson-Tilton equation (Henderson and

Tilton 1955).

Hazard quotients (HQ) were calculated using the highest field rate (calculated using the application rate and the concentration of active ingredient in the product) of one formulated product of each pesticide using Equation 2.2 (EFSA 2013). The HQ is used in the European

Union to estimate the risk posed to bees in the field by pesticides, but there is not enough information about ALB exposure to these insecticides to calculate a risk quotient as would be done in North America (EPA 2012).

푎푝푝푙𝑖푐푎푡𝑖표푛 푟푎푡푒 (𝑔 푎.𝑖./ℎ푎) 퐻푄 = [2.2] 퐿퐷50

The LD50 values were compared with previous values calculated for HB with and without accounting for the mean mass of a HB adult worker and the mean mass of the ALB females in my study.

2.3 Results

LD50 values could not be generated for chlorantraniliprole and captan as mortality never exceeded 50% at the highest possible concentration of each pesticide that could be dissolved in acetone: 3,500 ng a.i./bee and 30,000 ng a.i./bee, respectively. Imidacloprid was more toxic than permethrin, which was slightly more toxic than dimethoate (Table 2.1). The slope of the dose response across all methods of mass incorporation was steepest for imidacloprid, followed by dimethoate, and then permethrin (Table 2.2, Figure 2.2). Nominal treatment concentrations, sample sizes, arithmetic mean mortality, and mean mass and standard deviations of each treatment group are presented in Table 2.1.

Table 2.1. Treatments, sample sizes, the arithmetic mean mortality, mean dry mass and standard deviation (SD) of female Megachile rotundata 72 hr after topical treatment with five pesticides dissolved in 1 µL acetone. Insecticide Treatments Sample size Mean Mean dry mass (ng a.i./bee) mortality (%) (SD) in mg dimethoate 0 59 10.1 15.2 (2.42) 10 62 9.9 16.3 (3.45) 50 63 25.3 16.6 (3.87) 100 62 87.9 13.8 (3.12) 125 48 95.4 14.0 (1.70) 150 54 100 12.6 (1.63) permethrin 0 60 16.7 15.1 (2.78) 10 60 35.0 13.3 (4.31) 30 60 21.7 16.4 (3.13) 50 59 49.1 14.7 (3.03) 70 60 83.3 11.9 (2.37) 90 60 86.7 13.1 (2.33) imidacloprid 0 70 10 15.1 (2.24) 0.5 69 23.2 12.9 (2.13) 10 60 35.0 12.6 (2.62) 20 70 71.4 12.9 (1.71) 30 70 84.3 13.3 (1.54) 40 70 90.0 12.2 (1.74) chlorantraniliprole 0 30 20 NA 3,500 40 2.5 NA captan 0 30 20 NA 5,000 30 36.7 NA

When the mean mass of individual adults for each species was not taken into account, imidacloprid and dimethoate were more toxic to ALB than HB (Table 2.3). However, all three pesticides were less toxic to ALB than HB when the mean mass of individual adults was accounted for (Table 2.3). The mean dry mass of all bees in these trials was 14.4 mg with a standard deviation of 3.06 mg.

All surviving bees treated with imidacloprid at all doses presented various degrees of rigid paralysis (Sharf 2008) for the duration of the experiment, indicating that imidacloprid affects bees for more than 72 h. Paralyzed bees were characterized as having their legs and abdomen uncurled and with their wings together behind the thorax in a posture similar to a resting bee, but being nearly unable to move. Paralyzed bees were observed both ventral side down and ventral side up, but only responding to forecep stimulation with varying degrees of twitching in the legs and antennae. The degree of paralysis was dose-dependent, where bees that were treated with higher doses were less able to move than bees treated with lower doses.

Whether these effects would result in mortality after 72 h was not assessed. Bees treated with dimethoate presented varying degrees of spastic movements in the first day, but these symptoms disappeared by the third day. Bees treated with permethrin did not present any toxic symptoms beyond mortality.

Both mean bee mass per group and individual bee mass significantly affected the models

(Table 2.2). The LD50 values calculated for each method of mass incorporation within each pesticide were not notably different, as in all cases the confidence values overlapped between estimates (Table 2.2). Models that did not incorporate mass generally fit equally well or better than both the models that incorporated mean treatment replicate mass and individual bee mass

(Table 2.2).

The HQ values for both dimethoate and imidacloprid surpassed the proposed trigger value for solitary bees (EFSA 2013), while the HQ for permethrin did not (Table 2.4).

2.4 Discussion

My results suggest that it may be acceptable to use HB as a surrogate species for ALB

(and potentially all solitary bees) in tier I trials if assessment factors are incorporated. The toxicity of all three insecticides in my study were within the 10-fold assessment factor (a conservative 1/10 the LD50 for HB) proposed in EFSA (2013) based on HB LD50 values from the literature. For a more exact comparison between ALB and HB these bioassays could be repeated with both species concurrently, such as in Helson et al. (1994). In that study, 3 bee species including ALB were protected within the 10-fold assessment factor for 6 pesticides when individual mass was not included Helson et al. (1994). My findings are also consistent with those of Uhl et al. (2016) who found that although smaller bees are generally more susceptible to dimethoate than larger bees on a dose/bee basis, 95% of species are within the 10-fold assessment factor proposed in EFSA (2013). Providing further support to the use of HB as a surrogate in tier I experiments, ALB appeared to be less susceptible to all three of the insecticides I tested than HB when accounting for the mean mass of each species, which is also reflective of Helson et al. (1994). Therefore, the threshold values that are protective to HB via topical exposure are likely also protective of ALB in a lab environment after incorporating the

10-fold assessment factor. If HB tier I pesticide RA is protective of other solitary bees as well, it may be prudent to forego development of tier I RA for solitary bees. Time and money could then be directed to tier II and III trials for solitary bees where the exposure profiles of HB and solitary bees are more variable.

Table 2.2. Topical toxicity of dimethoate, permethrin, and imidacloprid to Megachile rotundata females either not incorporating body mass, or using one of two approaches to incorporate body mass as a model parameter: mean mass of each treatment replicate of 10 bees, and mass of each individual. Wald test X2 statistics are provided for both the effect of treatment and the effect of mass, where a high test statistic and significant P value indicate a significant effect on the model. Model parameters were generated using probit transformed mortality and untransformed dose data.

2 2 2 2 2 Insecticide Method of mass LD50 (± 95% CI) Slope at Wald X Wald X P Wald X Wald X P Pseudo R

incorporation (ng a.i./bee) the LD50 treatment treatment mass mass dimethoate No mass 69.84 (7.97) 3.06 e-2 34.04 < 0.0001 NA NA 89.91 Mean mass of container 62.25 (11.07) 2.14 e-2 50.65 < 0.0001 38.30 0.003 87.79 Individual mass 61.74 (8.56) 2.85 e-2 256.03 < 0.0001 118.61 < 0.0001 74.11 permethrin No mass 53.49 (8.25) 2.91 e-2 83.54 < 0.0001 NA NA 63.54 Mean mass of container 49.83 (11.66) 2.06 e-2 62.18 < 0.0001 34.18 < 0.0001 74.00 Individual mass 44.89 (8.17) 2.66 e-2 272.08 < 0.0001 179.39 < 0.0001 45.84 imidacloprid No mass 17.36 (2.73) 6.82 e-2 118.83 < 0.0001 NA NA 74.40 Mean mass of container 15.99 (3.93) 5.79 e-2 55.01 < 0.0001 43.44 0.007 76.66 Individual mass 12.90 (3.05) 6.38 e-2 282.87 < 0.0001 225.37 < 0.0001 43.24

Figure 2.2. Dose response curves for mortality 72 h after contact exposure of female Megachile rodundata to dimethoate, permethrin, and imidacloprid. These models were generated without incorporating bee mass. LD50 values (ng a.i./bee) and 95% confidence intervals are represented by vertical black lines and grey rectangles, respectively. The standard error of predicted values is represented by the grey ribbon around the dose response curve.

Table 2.3. Topical toxicity of three insecticides 72 hr after application to Megachile rotundata (ALB) and honey bees (HB). LD50 values for ALB in this study are presented along with the relative toxicity to HB and previous estimates for ALB. LD50 values incorporating mean body mass for each species are estimated using the mean mass of an ALB adult in this study and the mean mass of a HB adult. Pesticide LD50 (± 95% CI) Toxicity ratio to ALB in Toxicity ratio to HB Toxicity ratio to HB (ng a.i./bee) other studies (per bee) (per bee) (per mean body mass) dimethoate 69.84 (7.97) 1 : 0.01a 1 : 3.18c 1 : 0.36c,e permethrin 53.49 (8.25) 1 : 0.15b 1 : 1.37d 1 : 0.15d,e imidacloprid 17.36 (2.73) - 1 : 4.53d 1 : 0.46d,e aTorchio (1973),bHelson et al. (1994), cOECD (1998b), dSanchez-Bayo and Goka (2014), e(EPA 2012).

Insecticide LD50 (± 95% CI) Highest application Hazard quotient (ng a.i./bee) rate (g a.i./ha) dimethoate 69.84 (7.97) 1104a 17.55 permethrin 53.49 (8.25) 70b 1.60 imidacloprid 17.36 (2.73) 48c 2.76 aLagon® 480 E, Loveland Canada Products Inc., Dorchester, Ontario, Canada, bAmbush® 500 EC, AMVAC Chemical Corporation, Newport Beach, California, USA, cAdmire® 240, Bayer CropScience Inc., Calgary, Alberta, Canada.

The similarity in response between ALB and HB in laboratory trials is not surprising.

Laboratory topical exposure experiments exclude the effects of many of the differences between

ALB and HB, most notably the differences in life history, behaviour, and sociality (EFSA 2013,

Sgolastra et al. 2018). These differences are more likely to result in differences in susceptibility and exposure profiles at higher tiers of RA, and as such, HB may still not be suitable as a surrogate species for tier II and II trials.

If tier I toxicity testing methods are developed for pesticide RA using ALB, a positive control will be required. The toxic effects on the test organism caused by a positive control must be clearly measureable and replicable with relatively low variability. There are a limited number of laboratory studies that have assessed the acute susceptibility of ALB to topical applications of dimethoate or permethrin in the lab (Torchio 1973, Helson et al. 1994) that could be used as a reference for their development as a positive control, and there are no studies examining imidacloprid. The studies that have been completed with dimethoate and permethrin estimate

LD50 values much lower than those I estimated in my study. For example, Torchio (1973) also exposed ALB females to topical applications of dimethoate but reported an LD50 that is 77-fold lower than the LD50 value I generated (Table 2.2). There are several possible reasons for this difference primarily that the method used to feed the bees post-treatment (a honey saturated wick on top of the post-treatment container) in the Torchio (1973) study did not work in my preliminary experiments. The bees never flew or climbed to the top of the container and never encountered the sugar solution. Therefore, it is possible that the bees tested in Torchio (1973) may have been unable to feed for the duration of the experiment. There was low control mortality in those experiments after 72 hr, but ALB can survive for several days without food.

Starvation would weaken the ALB and increase their susceptibility to pesticides. As well, it is

38 unclear whether formulated or technical product was used by Torchio (1973). A formulated product could increase the cuticular absorption of dimethoate and increase the relative toxicity of the product, resulting in a lower LD50 (Mesenage and Antoniou 2018).

Helson et al. (1994) found permethrin to be 7-fold more toxic to ALB under lab conditions than in my study. This difference may be due to methodological differences. Helson et al. (1994) used formulated permethrin and maintained the bees at 16 oC following treatment.

Permethrin has been shown to be more toxic at lower temperatures to other insects (Whiten and

Peterson 2016). Additionally, ALB are slowed and likely have difficulty feeding at 16 oC.

Temperature stress and an inability to feed may have increased ALB susceptibility to permethrin in Helson et al. (1994) and resulted in a lower LD50 than in my study.

The variation between the toxicity of dimethoate and permethrin to ALB in my study compared to previously reported values highlights the importance of using repeatable, validated methods to produce consistent toxicity estimates. I tried to replicate the methods of Torchio

(1973) without success, and the experimental conditions used in Helson et al. (1994) are not suited to ALB biology. Tier I topical toxicity tests with a more replicable design will provide a better understanding of the expected toxic effects of permethrin, dimethoate, and imidacloprid on adult female ALB.

Imidacloprid appears to be a poor choice as a positive control because the paralysis it induced for the duration of my experiment makes it difficult to confirm mortality. Additionally, imidacloprid will likely continue to have a negative effect on survival beyond the short duration of a tier I RA (usually 48 or 72 h). In contrast, the observable effects of dimethoate and permethrin ceased by the end of the experiment, indicating that all individuals had either died or recovered. The cessation of toxic effects within a short time is a desirable trait when designing

39 short-term trials as the onset and conclusion of effects are clear and occur within the test period.

Using dimethoate as a positive control for tier I ALB RA would also maintain continuity with other bee pesticide RA methods, all of which use dimethoate, (OECD 1998b, 2015, Knäbe et al.

2017). Ring testing will be required to develop any of these insecticides as positive controls for

ALB RA. Thus far, my research is the only assessment of the toxicity of permethrin or dimethoate to ALB for the purposes of tier I pesticide RA method development.

My results suggest that incorporating individual mass in a dose response model will not improve the precision of the LD50 or model fit for acute topical application of pesticides to ALB.

Similarly, Helson et al. (1994) found that incorporating the mass of individuals did not change the LD50 when assessing the effect of acute topical applications of six pesticides on four bee species. Although I found that mass had a significant effect on my models, there was no observable change in the toxicity estimates or model fit. I therefore conclude that incorporating individual mass when calculating LD50’s for the purposes of pesticide RA with ALB is not necessary. However, I do recommend reporting the mean body mass of test populations to account for mass variation between trials. Risk assessments may use data compiled from different experiments and sample populations, and each population will have a different mean individual mass. Different populations of ALB are reared across North America, and the mean body mass of the individuals in these populations can be affected by latitude (Pankiw et al.

2012), production protocols (Pitts-Singer and Cane 2011), and environmental conditions and crop quality (Rothschild 1979). Sampling population mean mass can also be affected simply by sampling error from the same original population. It is possible that variation in mean body mass between sample populations may significantly affect toxicity estimates generated for different populations. There are few estimates of the variation in individual body mass between or within

40 different ALB populations to develop an understanding of how individual populations are affected by geographical origin, management methods, and year of production. There are a few individual records of the sample population mass in a few studies, but these have not been formally analyzed (Helson et al. 1994, O’Neil et al. 2010, Rothschild 1979). We therefore cannot predict the degree to which variation in individual body mass among sample populations will affect the results of tier I toxicity trials. It would be prudent to report the mean body mass of the bees used in future experiments in case this value can be used to improve the consistency of toxicity estimates between sample populations.

The mean dry mass of all the bees used in this experiment was 14.4 with a standard deviation of 3.06 mg. Helson et al. (1994) presents one of the few other examples of the mean mass and standard deviation of a sample population of ALB: a live mean mass of 28.6 mg with a standard deviation of 5.7 mg. The mass variation in my study and Helson et al. (1994) (presented as the ratio between mean mass and standard deviation, as dry mass and live mass cannot be compared directly) are 0.21 and 0.20, respectively. These values are almost identical, and indicate that the mass variation in my study was similar to at least one other study. This value needs to be recorded in other studies to generate a larger body of data regarding mass variation in

ALB sample populations.

The trigger values proposed by EFSA (2013) are preliminary estimates based on other solitary bee species and are not designed for ALB in particular. These HQ trigger values are more precautionary than the RQ trigger values used in North America by the EPA (EPA 2012).

The HQ of dimethoate (17.55) calculated from my results exceeded the trigger value range for solitary bees (8–16) proposed by EFSA. The risk of field applications of dimethoate posed to solitary bees may need to be examined in more detail.

In conclusion, my results suggest that it may not be necessary to develop topical tier I pesticide RA using ALB, as the LD50 values that I generated were within a 10–fold assessment factor of previously generated HB LD50 values as suggested in EFSA (2013). If solitary bee tier I

RA is developed, I do not recommend incorporating individual bee mass into LD50 analyses. I do, however, recommend reporting the mean mass of sample populations to address potential differences between sample populations. Finally, this is the first estimate of the toxicity of dimethoate or permethrin with ALB. There will need to be a substantial effort put into repeated testing before dimethoate or permethrin can be fully characterized as a positive control for tier I pesticide RA.

3 Development of chronic oral laboratory toxicity assessment methods for

tier I risk assessment with Megachile rotundata

3.1 Introduction

Solitary bees comprise the majority of North American bee species (Michener 2007) and are important pollinators in natural and agro-ecosystems (Smagghe and Calderone 2012). For some crops (e.g.,alfalfa (Medicago sativa) and cucurbits (Cucurbita)) and uncultivated plants, solitary bees are more effective pollinators than honey bees (HB, Apis mellifera). There are reports of worldwide declines in bee abundance and diversity, which has been attributed to a variety of stressors including pesticide applications (Goulson et al. 2015, Koh et al. 2015, Potts et al. 2015, Leach and Drummond 2018).

Pesticide risk assessment (RA) is the regulatory process used to assess the risk of pesticides to non-target organisms, including bees. Current RA methods for bees are focused almost exclusively on HB and may not account for differences in sociality, life history, and physiology between HB and solitary bees (EFSA 2013). Without a thorough understanding of the risk of pesticides to solitary bees, there is no way to know if solitary bees are adequately protected from pesticides. The potentially inadequate representation of solitary bees in RA has driven international efforts to develop RA protocols that will better characterize the risk posed to solitary bees by pesticides (EFSA 2013, EPA 2014, Fischer and Moriarty 2014c, Sgolastra et al.

2018).

The pesticide RA process consists of three tiers of testing, where higher-tier tests may be used to further refine the understanding of the risks posed by a pesticide in progressively more field- realistic scenarios (EPA 2012). Tier I is used as a preliminary screening and primarily focuses on acute laboratory pesticide exposure to adults and predictive modelling, but other tests, such as

43 larval toxicity or adult chronic oral exposures, can be included as well. Chronic oral tests are usually performed when it is expected that bees will be exposed to the pesticide over long durations. In general, these trials involve providing bees with a treated food source ad libidum for 1 wk or more, during which lethal and/or sub-lethal effects are assessed (EPA 2016, OECD

2017a). A chronic oral test is used to determine the oral no observed adverse effect concentration

(NOAEC), which can then be compared to a risk quotient (RQ) to characterize the risk posed to bees through ingestion (EPA 2012). The oral toxicity, NOAEC, and exposure routes for nearly all pesticides are currently unquantified for solitary bees (EFSA 2012, Sgolastra et al. 2018).

It is not feasible to develop RA protocols for every species of solitary bee, and so surrogate species must be selected. These species should represent the biology, physiology, and life history of managed and unmanaged solitary bees as closely as possible while also being easy to work with (Pettis et al. 2014). The alfalfa leafcutting bee (ALB, Megachile rotundata) is a good candidate for North American solitary bee RA (Pitts-Singer and Cane 2011). Introduced to

North America in the 1940s, it is now well established and demonstrates similar life history traits and environmental requirements as native species (Pitts-Singer and Cane 2011, Sheffield et al.

2011 ). Additionally, it is available for purchase in large quantities for nearly the entire year

(Pitts-Singer and Cane 2011). Finally, compared to other solitary bees, ALB biology, life history, and pesticide toxicology are relatively well understood.

Although they are easy to rear and handle under laboratory conditions, ALB are notoriously difficult to feed known amounts of food in acute or chronic oral pesticide toxicity studies (Ladurner et al. 2003, Ladurner et al. 2005b, Frewin personal comm.1). Previous feeders designed for ALB have included the use of artificial flowers (Ladurner et al. 2003, Ladurner et al. 2005b), glass pipettes (Torchio and Youssef 1973), and inverted sugar feeders (Huntzinger et al. 2008b). There is no documentation of these feeding designs being implemented by other researchers, and I could not successfully feed ALB with these feeder designs. Therefore, one of the objectives of my study was to develop an experimental method for consistently feeding ALB known amounts of sugar solution that could be used to assess the oral toxicity of pesticides to adult ALB.

A positive control is used to confirm exposure in tier I RA trials (EFSA 2012, OECD

2017a). There are no data available on any insecticide to determine an effective positive control for tier I chronic oral pesticide RA with ALB. Therefore, chronic oral toxicity exposure data need to be generated using potential positive controls with ALB to determine a suitable positive control. Dimethoate is a positive control used in tier I and II HB RA (EPA 2012, OECD 2017a) as well as RA for other non-Apis bees (Knäbe et al. 2017, OECD 2017b,c). It is highly toxic to bees and causes a rapid onset of mortality. It is therefore an excellent candidate for use in tier I chronic oral RA with ALB. Imidacloprid is another potential candidate as it has well-established lethal and sub-lethal toxicity to bees, although its effects on ALB are still largely unknown

(Mommaerts et al. 2009, Scott-Dupree et al. 2009, Blacquière et al. 2012). The use of a systemic insecticide as a positive control has been suggested by EFSA (EFSA 2012), and imidacloprid would fill that role. However, there is a potential issue with using imidacloprid as a positive control that needs to be addressed. Imidacloprid causes rigid paralysis in insects (Sharf 2008,

1 Dr. Andrew Frewin, [email protected] 45

Pisa et al. 2014), and my previous work (Chapter 2) has confirmed that this paralysis may last several days in ALB after topical application. This paralysis may prevent the bees from consuming a consistent amount of treated food, thus producing a less consistent response to treatments, which will in turn reduce the value of imidacloprid as a positive control.

The overall goal of my study was to contribute to method development for estimating the toxicity of pesticides in tier I chronic oral RA with ALB adults as a surrogate solitary bee species. My objectives were to:

1. Develop a feeding apparatus that adult ALB will consistently feed from, allowing

them to be exposed orally to pesticides for an extended duration, and,

2. Generate data on the lethal effects of dimethoate and imidacloprid as potential

positive controls in tier I chronic oral experiments using adult female ALB.

3.2 Introduction

3.2.1 Test insects

Alfalfa leafcutting bee pre-pupae were obtained from NorthStar Seed Ltd. (Manitoba,

Canada) in November, 2017 and stored at 8 ºC. Approximately 4 wk before the start of the experiment, pre-pupae were stored at 27 ± 2 ºC, 60% RH and a 12:12 h light:dark cycle in plastic containers until emergence. Each container had a large square hole in the lid, covered on the underside with fine wire mesh, to allow airflow and prevent adults from escaping. On the outside of the lids, the holes were covered with fine polyester noseeum fabric netting (Skeeta, Florida,

United States) to limit the transfer of parasitoids between containers. Upon emergence, ALB adults were transferred to treatment containers (Figure 3.1) and placed in a greenhouse at 22 ±

5.2 ºC and 60 ± 16.9% RH to acclimate to treatment conditions for 3 d.

3.2.2 Treatment containers

Each treatment container (Figure 3.1) consisted of a 500-mL clear plastic deli cup with a clear plastic lid with a 5 x 5 cm hole covered in fine wire mesh. A bent piece of 5 x 10 cm wire mesh was placed inside at the bottom of the cup for the bees to stand on. A 1-cm diameter hole was cut in the bottom of each cup and covered with an artificial flower. The artificial flower was a 3-cm diameter disc of paper painted with yellow and blue fluorescent acrylic paint (DEMCO

EnCouleurs Inc.) in a radiating design that had a 1-cm diameter hole in the centre. The artificial flower was based on the design by Ladurner et al. (2005b), but instead of coloured tape, I used ultraviolet paint as described by Ladurner et al. (2003) and Heard et al. (2017) to act as colour cues for the bees. The hole in the flower was covered with fine wire mesh through which the bees could feed from a sucrose solution-soaked dental wick. The purpose of the mesh was to limit physical contact between the bees and the treated food to ensure that they were exposed to the treatments only via ingestion.

The 500-mL cup sat inside a 1-L plastic deli cup. A 20-mL glass scintillation vial containing sugar solution was placed in the bottom of the 1-L cup. A dental wick saturated with treatment solution was placed in the scintillation vial, and centred under the hole in the 500-mL cup. The dental wick was cut to length to sit in the scintillation vial and gently contact the bottom of the 500-mL cup.

I did not incorporate scent cues into the feeders as Ladurner et al. (2005b) reported that

ALB did not respond well to scents.

i ii A

D E

Figure 3.1. i) A feeding container used to house Megachile rotundata adults in a greenhouse before and during treatment. Containers consisted of a A) plastic lid with wire screening, B) 1 L clear plastic cup, C) 500 mL clear plastic cup with a small hole cut in the bottom and covered with wire mesh, D) piece of wire mesh, E) artificial flower, F) sugar water-soaked dental wick pressed against mesh in 500 mL cup G) 20% sucrose solution reservoir, and H) sugar solution; ii) Close-up view of artificial flower (i-E). Yellow and blue markings were painted with flourescent paint to stimulate feeding by the bees on the sucrose solution-saturdated dental wick through the mesh.

3.2.3 Insecticide treatments

The formulated insecticides dimethoate (Lagon® 480 E, Loveland Canada Products Inc.,

Dorchester, Ontario, Canada) and imidacloprid (Admire® 240, Bayer CropScience Inc., Calgary,

Alberta, Canada) were diluted in a 20% sucrose and water solution to reach the desired concentrations. All pesticide dilutions were made fresh each application day. Imidacloprid was applied at 0.001, 0.01, 0.02 and 0.05 mg/L. These concentrations were chosen to be higher and lower than 0.01 mg/L, a concentration stated as having sub-lethal effects on other species of bees after limited exposure (Blacquière et al. 2012, Pisa et al. 2014). Therefore, it was expected that these concentrations would cause mortality after days of constant exposure. Dimethoate was applied at 0.04, 0.2, 1, and 5 mg/L. These concentrations were serial dilutions ranging from 12.5 fold lower to 5 fold higher than those proposed by the OECD for use with HB chronic oral tier I trials (OECD 2017a). All insecticide concentrations were nominal and were not confirmed with further chemical analyses.

The insecticides were mixed in the morning and provided to the bees in the afternoon between 2:00 and 5:00 pm. Bees were exposed to the insecticide treatments using the feeding containers described in section 3.2.2 (Figure 3.1) and kept in a greenhouse at 22 ± 5.2 ºC and 60

± 16.9% RH with a temperature range over 80 d of 13.28–38.30 ºC. Groups of 10–13 bees were placed in each container and provided with untreated sucrose solution for 3 d to acclimate to experimental conditions. After this acclimation period, the few bees that were not feeding and had died were removed from the experiment. The 20% sucrose solution was replaced with the corresponding treated solution and the number of acclimated bees in each container was recorded. Control containers received fresh untreated 20% sucrose solution. The bees were then placed back in the greenhouse. The solution in each container was replaced every 3–4 d. To

49 measure solution consumption by each group of bees, feeders were weighed before being placed in and after being removed from each container. The mortality in each cup was assessed every time the feeders were weighed. A bee was considered dead if it did not respond to a gentle squeeze on the thorax with forceps.

Each treatment was replicated 4–5 times for a total of 40–52 bees per treatment. Each group of treatments also included a control container of bees fed untreated sugar solution, and an evaporation control consisting of a treatment container without bees. The evaporation control was intended to be used to calculate how much treatment solution was consumed by the bees in each container, but in some cases the evaporation controls lost more mass than the containers with bees.

The experiments continued until 100% mortality was reached in the treatments or 25% mean mortality was reached in the controls. Depending on the treatment, the experiments lasted between 10 and 14 d. Data were analyzed up until the day before 25% mortality was reached or exceeded.

Temperature and humidity were recorded throughout the entire experiment using an environmental probe (HOBO® Pro v2 ext temp/RH, U23-002, ONSET) placed under a plastic dish to protect it from direct sun exposure.

In addition to the treatment containers in the greenhouse, female ALB were placed in a growth chamber with fluorescent lighting at 25 ± 1.5 ºC, 60% RH, and a 12:12 h light:dark cycle in treatment containers as described above. These bees were used to compare ALB survival in the treatment containers under fluorescent light (growth chamber) and natural light (greenhouse).

The greenhouse consisted of one brick wall and three walls and roof composed of glass panels.

These panels did not have any UV-resistant film and were made of plain glass. I placed seven

50 containers of 10 bees in the growth chamber that mirrored seven of the control containers used in the pesticide treatments for dimethoate and imidacloprid in the greenhouse. The bees in the growth chamber were provided with untreated sugar solution, and weighed and assessed for mortality on the same schedule as their greenhouse counterparts.

3.2.4 Statistical analyses

All analyses were completed in R version 3.4.3 (R Core Team 2017). Control mortality was corrected with the Henderson-Tilton equation (Henderson and Tilton 1955). Models were analyzed with a log-likelihood logistic analysis penalized by the Jeffreys prior with the logistf package (Heinze and Ploner 2018), and all significance tests were performed at α = 0.05.

Variance was partitioned into the effect of dose and the effect of the maximum temperature each day (Equation 3.1), where β0 is the intercept and βx are the parameters for insecticide concentration, maximum temperature, and days after release. The comparison between the bees maintained in the growth chamber and greenhouse was made using the same equation with location (greenhouse or growth chamber) as the fixed effect and block as the random variable.

1 푠푢푟푣푖푣푎푙 = [3.1] 1+푒−(훽0+푥푐훽푐+푥푡훽푡+푥푑훽푑)

The significance of each explanatory variable was tested with a Wald test (Agresti 1990) using the anova function (test = “Chisq”) in R (R Core Team 2017). Model fit was approximated by comparing the models with explanatory variables to null models with no explanatory variables using the likelihood ratio test in the logistf package in R (Heinze and Ploner 2018). The interactive effect between the treatment and maximum temperature was analyzed but not included in the final models, because in all cases it was not significant and was removed from the model in the final analysis.

3.3 Results

The bees in the greenhouse experienced significantly lower mortality over the course of the experiment (P < 0.0001) than those in the growth chamber (Figure 3.2). Both imidacloprid and dimethoate had a significant negative effect on mortality (P < 0.0001), and imidacloprid was

48 times more toxic than dimethoate (Table 3.1). Additionally, the maximum temperature had a significant negative effect (P < 0.0001) on ALB mortality for dimethoate, but not imidacloprid

(P = 1.00) (Table 3.1). Unadjusted mean control mortality surpassed 25% between 10–15 d depending on treatment, although only the day prior to this point was analyzed and reported

(Figure 3.3, Figure 3.4).

The bees exposed to imidacloprid experienced varied degrees of paralysis intermittently throughout the experiment in conjunction with feeding on the treated sucrose solution. Paralysis was considered any visible effect between slowed movement to near-complete cessation of motor function where the only sign of life was small twitches of the tarsi and antennae. Highly affected bees were frequently observed on their backs and unable to right themselves. Less- affected individuals moved lethargically and were unable to fly, but were able to right themselves. The duration of paralysis experienced by individuals was not recorded.

(%)

Mortality

Days

Figure 3.2. Mean (± SE) percent mortality of adult female Megachile rotundata fed 20% sucrose solution in either a greenhouse (22 ± 5.2 ºC) or controlled growth chamber (22 ± 1.5 ºC) over 28 d.

Insecticide LRT test statistic Pr LRT Parameter Estimate (± 95% CI) Wald X2 Wald X2 P imidacloprid 1836.60 < 0.0001 Treatment (mg/kg) 125.00 (15.85) > 1,000 < 0.0001 Temperature (ºC) -0.04 (0.07) 38.43 1.00 Days after release (d) 0.35 (0.06) > 1,000 < 0.0001 dimethoate 1341.71 < 0.0001 Treatment (mg/kg) 2.27 (0.30) > 1,000 < 0.0001 Temperature (ºC) -0.18 (0.05) 63.87 < 0.0001 Days after release (d) 1.69 (0.24) > 1,000 < 0.0001

Mortality (%) Mortality

Days

Figure 3.3. Mean (± SE) percent mortality of adult female Megachile rotundata exposed ad libidum to 20% sucrose solution treated with formulated imidacloprid at 0.01, 0.02, 0.05 mg/L for 14 d.

Mortality (%) Mortality

Days

Figure 3.4. Mean (± SE) percent mortality of adult female Megachile rotundata exposed ad libidum to sucrose solution treated with formulated dimethoate at 0.04, 0.2, 1, and 5 mg/L for 10 d.

3.4 Discussion

In this study I have made the first assessment of the toxic effects of dimethoate on adult female ALB following chronic oral exposure. I tested a range of concentrations that can be used as a benchmark for future experiments to further characterize dimethoate as a positive control for tier I RA. Dimethoate was highly toxic to female ALB and did not appear to inhibit feeding or repel the bees, making it an excellent candidate for use as a positive control. Dimethoate is also suggested as a positive control for HB RA in tier I chronic oral studies (OECD 2017a). An oral concentration of 0.2–1 mg/L for dimethoate is suggested in these HB RA protocols to reach 50% mortality in 10 d (OECD 2017a). I recorded 50% mortality at the lower end of that range after 10 d, as 0.2 mg/L resulted in a mean mortality of 42% and 1 mg/L resulted in a mean mortality of

96%. As the concentration resulting in 50% mortality of ALB in my experiments was within the range estimated to result in 50% mortality of HB over the same duration, dimethoate appears to have similar oral toxicity to ALB and HB. The concentrations of dimethoate used in my study may therefore be useful for developing dimethoate as a positive control with ALB, as a positive control used in chronic tier I pesticide RA should be applied at a rate that will result in 50% mortality (OECD 2017a). Ring testing and improvements on the treatment container design for

ALB will be required to thoroughly define dimethoate as a positive control and assess whether the oral toxicity of dimethoate is similar between ALB and HB. Future work to identify a

NOAEC for dimethoate can be built upon my work as well: The lowest concentration used in my study (0.04 mg/L) may be a good place to start the characterization of a NOAEC, as it resulted in only a small effect on mortality after 10 d.

My study also represents the first assessment of the chronic oral effects of imidacloprid on adult female ALB mortality. I do not recommend imidacloprid as a positive control for

57 chronic oral tier I RA because the bees experienced paralysis during treatment, which interfered with their activity and ability to consume treated food. This observation is consistent with previous work I have done with imidacloprid and ALB (Chapter 2), as well as with descriptions of the effects of neonicotinoids on honey bees (Pisa et al. 2014) and acetylcholinesterase inhibitors on insects in general (Sharf 2008). Paralysis resulting from the consumption of treatment solution can be expected to reduce the consistency of ALB feeding. The duration of paralysis is likely to be dependent on the dose consumed, which means that bees receiving higher doses will feed less often than bees that receive lower doses. Therefore, exposure to imidacloprid through treated sugar solution will essentially produce a dose-dependent antifeedant effect in

ALB. Additionally, not all bees will consume the same amount of treatment solution at each feeding, resulting in differences in the dose received by individuals within a treatment before they are paralyzed. A positive control should produce results that are as consistent as possible to confirm proper exposure, and this may not be the case if the bees are paralyzed.

ALB fed consistently from the feeders used in my experiment. I observed the bees feeding on the artificial flowers, walking, and flying for 10–15 d depending on the treatment block. Mean control mortality usually increased above 25% after 2 wk of observations, which may hinder the applicability of this method for studies lasting longer than 2 wk. However, HB

RA chronic oral tests only last 10 d, so this time limit may not be an issue in a RA context

(OECD 2017a).

Bees kept in the growth chamber experience nearly 20% higher mortality compared to bees maintained in the greenhouse. It is possible that the bees in the growth cabinet had difficulty locating the feeder under fluorescent light. Tezuka and Maeta (1993) reported that ALB are unable to re-locate their nests in greenhouses covered with ultraviolet light-absorbing film.

Additionally, Ladurner et al. (2005b) reported that ALB did not feed from an artificial flower design similar to mine under fluorescent lights without UV paint. In contrast, Ladurner et al.

(2003) reported nearly 100% ALB feeding success under natural and fluorescent light (15W

Cool White Sylvania® fluorescent) using real flowers, each containing a plastic ampoule with a ring of UV paint applied to the tip. I had limited success feeding ALB by incorporating real flowers in my preliminary experiments. The individual flowers also completely wilted within days and would need to be replaced several times over the course of a chronic experiment.

Therefore, I recommend the use of an artificial flower design similar to the one I used to allow the bees to feed for extended periods on a feeder that does not degrade. I also recommend the use of UV paint and natural light as ALB appear to favour visual cues outside the spectrum visible to humans. Further study to isolate the mechanism by which ALB respond to UV paint under natural light would be beneficial to develop methods to consistently feed ALB in well-controlled conditions under fluorescent lighting.

In contrast to the control mortality in my study (roughly 25% after 14 d), the chronic oral feeding protocol used by Huntzinger et al. (2008b) resulted in < 20% control mortality after 21 d of observations. Those experiments were performed in a controlled environment at 29 oC with feeders consisting of an inverted test tube covered in cloth. However, when I performed preliminary experiments using these methods, ALB would not feed consistently. Combining aspects of the inverted cloth covered feeder with additional UV paint marked artificial flowers may improve feeding and survival of adult ALB in chronic oral treatment containers compared to my study. An acceptable level of control mortality for ALB over long periods of time in laboratory or greenhouse conditions has not been well established. Further repetition and method

59 refinement is required to consistently limit control mortality and understand what control mortality is acceptable under these conditions.

The mortality I observed in the control bees kept in the greenhouse was likely not caused by starvation because I observed the bees feeding on a regular basis during the experiment.

Instead, the temperature variation and direct sun the ALB were exposed to may have increased the stress experienced by the bees and increased mortality. ALB have been recorded surviving and reproducing well at 30 °C (Bailey et al. 1982, Richards 2012), but the temperatures in the greenhouse in my study sometimes exceeded 35 °C. These temperatures exceed the temperatures that ALB usually forage in (Bailey et al. 1982, Richards 2012), and may have negative effects on bee health. Therefore, mortality may be reduced by placing the treatment containers in an area with a lower temperature maximum and less direct sun exposure. These experiments could also be conducted under fluorescent light if the factors affecting ALB preference for UV paint can be recreated artificially. Further assessments could be made to determine if the UV paint is necessary under natural light, or if ALB simply respond better to visual cues under natural light.

When the higher survival and feeding response under natural light is better understood, feeding containers may then be kept under well-controlled conditions while still successfully feeding

ALB. Alternatively, a source of shade and refuge from the sun could be introduced into the feeding containers to allow the bees to move themselves in or out of the shade to regulate their own internal temperature.

The mass lost by the evaporation control feeders was in many cases greater than the mass lost by the associated feeding containers with bees, which prevented me from estimating the amount of insecticide solution consumed by each container of bees. This high mass loss may have been due to small inconsistencies in the length of exposed dental wick (the area between the

60 labels E and F in Figure 3.1) resulting in higher rates of evaporation for some containers.

Variation in direct sun exposure between containers may also have been a contributing factor, although the position of the feeding containers on the greenhouse bench was randomized at each feeding. Evaporation may be reduced by reducing the amount of exposed wick, feeding bees with an angled capillary tube entering the side of the container similar to the design suggested by

Torchio and Youssef (1973), or incorporating the inverted cloth covered feeder proposed by

Huntzinger et al. (2008b).

In conclusion, the treatment container I designed can be successfully used to feed ALB for 10 d more effectively than most other feeding methods proposed for ALB (Ladurner et al.

2003, Ladurner et al. 2005b), as I had limited success using the mehod proposed by Huntzinger et al. (2008b). Key to the success of this design is the use of natural light and UV paint to make an artificial flower attractive to ALB. Further improvements upon this design should focus on measuring the sugar solution consumed in each container and reducing the temperature extremes inside the feeding containers to limit control mortality. I have also assessed the lethal effects of chronic oral exposure to dimethoate and imidacloprid on ALB for the first time. In doing so, I have generated baseline data necessary for developing dimethoate as a positive control in tier I chronic oral assays with ALB, as well as identifying treatments from which a NOAEC for dimethoate may be calculated. The data I have collected also may serve as a baseline when developing future toxicity assessment methods for tier I chronic oral pesticide RA using ALB.

4 Method development for semi-field toxicity trials for use in pesticide

risk assessment with Megachile rotundata

4.1 Introduction

Solitary bees comprise the majority of bees in North America (Michener 2007) and are responsible for an estimated 25% of the value of pollinated crops (Smagghe and Calderone

2012). There are several solitary bee species managed for commercial pollination in North

America, including the alfalfa leafcutting bee (ALB, Megachile rotundata), the orchard mason bee (Osmia lignaria), and the alkali bee (Nomia melanderi). Some solitary bee species are in decline, likely due to a combination of factors including disease, competition with non-native pollinators, climate change, and increased pesticide and fertilizer use associated with agricultural intensification (Goulson et al. 2015, Koh et al. 2015, Potts et al. 2015, Leach and Drummond

2018). These factors are likely to work in combination, where one factor increases the severity of the other factors, and no single factor is entirely to blame.

One of the underlying reasons for negative effects of pesticide use on solitary bees may be that the pesticide risk assessment (RA) paradigm is not adequately protective of solitary bees.

The current RA paradigm focuses almost exclusively on honey bees (HB, Apis mellifera) and may not properly account for differences in pesticide toxicity and exposure due to the differences in sociality, life history, and physiology between HB and solitary bees.

Pesticide RA methods unique to solitary bees should be considered to properly account for the differences between solitary bees and HB and thus more adequately protect solitary bees.

As it is unfeasible to develop RA methods for every species of solitary bee, surrogate species need to be chosen. The surrogate species should be regionally representative of the biology and life history of solitary bee species, amenable to use in the lab and field, and available for large-

62 scale testing; and have a well-understood biology and life history. For these reasons, the alfalfa leafcutting bee (ALB, Megachile rotundata) has been suggested as a surrogate species for North

American RA (Vaughn et al. 2014). Introduced in the 1940s, ALB is used for commercial pollination of alfalfa and has established wild populations across Canada and the United States

(Pitts-Singer and Cane 2011, Sheffield et al. 2011). ALB is commercially available in large quantities for research, individuals are easy to rear and handle in laboratory and field conditions, and individuals can be accessed nearly all year round (Pitts-Singer and Cane 2011). Additionally, compared to other solitary bees, the biology and life history of ALB are generally well understood, although there are still many knowledge gaps that need to be addressed before RA methods for ALB can be finalized.

Solitary bee RA will likely be based on the tiered framework already developed for HB

RA (EFSA 2012), where tier I consists of laboratory trials examining the effects of pesticides on individual bees, tier II focuses on colonies in semi-field trials where the bees are restricted to foraging on a treated crop, and tier III is performed on colonies in a treated field with no constraints on bee mobility (EFSA 2013). Tier II trials are limited in scope for HB RA because

HB do not function well in enclosures and overwintering colony success cannot be measured

(EPA 2012). Tier II experiments with solitary bees will allow analysis of foraging, reproduction, and overwintering ability inside an enclosure (Vaughn et al. 2014). Tier II RA is currently being developed for Osmia species in the European Union and can further inform other solitary bee RA development (Knäbe et al. 2017). There are currently no standardized RA methods for ALB.

The next step in developing pesticide RA methods for ALB is to understand key factors that can help further method development and be incorporated into the final methods. One such factor is the development of a positive control. Identifying a positive control early in the tier II

RA method development process will allow confirmation of exposure during method development and eventually in standardized tier II RA (EPA 2012). Dimethoate has been used as a positive control in tier II HB RA (EPA 2012) and is suggested for Osmia RA (Knäbe et al.

2017) and Bombus RA (OECD 2017c).

Before tier II RA methods can be created, one or more surrogate crops that ALB can reproduce successfully on must be identified. There has been some research on purple tansy

(Phacelia tanacetifolia) as a potential surrogate crop for ALB tier II toxicity assessment for use in RA (Frewin et al. 2019). Purple tansy is a bee-attractive crop with a high floral density. It has previously been assessed for use in Bombus RA (Cabrera et al. 2016, Gradish et al. 2016) and

Osmia (Knäbe et al. 2017) and shows promise as a surrogate crop for ALB.

Determining the optimal release rate of bees in a field enclosure of purple tansy will provide the highest number of bees to study (and consequently the highest number of offspring produced) while not limiting reproductive potential. There is a limit to how many ALB can be released in a certain area, as individual cell production and larval survival can be negatively affected by high female densities (Pitts-Singer and Bosch 2010). Female reproduction can also be reduced by male harassment if the ratio of males to females is too high (greater than or equal to three males for every female) in an enclosure (Rossi et al. 2010). Therefore, a release rate needs to be identified to find the optimum sample size in each tent and to maximize the reproductive output ALB in each enclosure.

The goal of my study was to generate baseline data to further the development of ALB tier II pesticide RA. I accomplished this goal with three objectives:

1. Generate toxicity data to support the use of dimethoate as a positive control in tier II

pesticide RA for ALB,

2. Assess the suitability of purple tansy as a surrogate crop for tier II RA studies by

assessing the reproductive output of ALB on purple tansy, and

3. Compare the effects of three potential ALB female release rates on reproduction and

nest provisioning activity in field enclosures for tier II pesticide RA development.

4.2 Methods

4.2.1 Test insects

ALB pre-pupae were obtained from NorthStar Seed Ltd. (Manitoba, Canada) in

November 2017 and stored at 8 oC until use (7–8 months later). A month before the start of experiments, the pre-pupae were placed in plastic rearing containers each with a plastic lid with an opening covered in fine polyester noseeum fabric netting (Skeeta, Florida, United States) to limit the transfer of parasitoids between containers. Pre-pupae were stored at 27 ± 2 ºC and 60%

RH with a 12:12 h light:dark cycle until emergence. The bees were placed in pre-treatment feeding containers (Figure 3.1) in groups of 13, and then placed in a greenhouse at 22 ± 5.2 ºC and 60 ± 16.9% RH for 3 d.

4.2.2 Planting purple tansy and field preparation

Experiments were conducted at a field site near Cambridge, Ontario, Canada at

43.314323º N 80.249219º W in well-drained silt loam soil. Purple tansy seeds were purchased from West Coast Seeds (British Columbia, Canada) in May 2018. The seeds were planted on 17

May 2018 with a drill seeder to a depth of 1.25 cm at a seeding rate of 19 kg/ha. This seeding rate was lowered from the rate of 33 kg/ha used by Frewin et al (2019) and Gradish et al. (2016) in sandy soil. No pesticides other than the treatments were applied to the plots after seeding.

Flowers on the weeds in each plot were removed but the plants were otherwise left intact to provide ALB with leaves for nesting material.

4.2.3 Semi-field enclosure design

For each experiment, individual 3.5 x 3.5 m plots of purple tansy were covered by a 3.3 x

3.3 m Coleman® Instant Screen House (Figure 4.1). As there was some heterogeneity in bloom across the field, plots were chosen where the purple tansy stand was of similar growth stage and floral abundance to maximize homogeneity across treatments. Each corner and all four guy-wires on each enclosure were secured to the ground with 20-cm ABS arrowhead tent pegs, and each side of the enclosure skirt was secured with two evenly spaced 30 x 2.5 cm slats of wood, each held in place with a 20-cm nail. These additions prevented the enclosures from blowing away in high winds and reduced the chance of bees escaping under the bottom. Enclosure entrances faced east. A nest box (Figure 4.1) was placed in each enclosure and attached to the top of a 1 m long stake. Each nest box consisted of a 20 x 20 cm section of Styrofoam® leafcutter nest block

(NorthStar Seeds Ltd., Manitoba, Canada) inside of a five-sided wooden box. The nest boxes each contained approximately 225 nesting cavities lined with unbleached craft paper facing out the open side of the box. The stakes were placed approximately 1 m in and to the right of the entrance of each enclosure. The entrance of the nest boxes faced the enclosure entrance, east, to catch the early morning sun. After erecting the enclosures, sweep nets were used to remove as many insects from the enclosures as possible

An environmental monitor (HOBO® Pro v2 ext temp/RH, U23-002, ONSET) was placed in one enclosure in each experiment to record temperature and humidity throughout the experiment.

4.2.4 Positive control experiment

In the morning prior to releasing the bees into the enclosures, each female was given a unique identification mark using a Sharpie® Poster-Paint water-based paint. The 3.5 x 3.5 m

66 plots for this experiment were organized in a randomized complete block design and were at least 20 m away from any other plot to avoid spray drift (Figure 4.4). Minor adjustments to the placement of the plots were made to increase the homogeneity of the purple tansy growth stage

(approximately 80% bloom) and visual floral density.

It has been shown that ALB prefer leaves with a high surface area, such as buckwheat leaves, for nest building, and reject small leaves (Horne 1995b). In anticipation of the potential for purple tansy, a small-leaved plant, to be a poor source of leaf material, a single 30 cm diameter pot of 5–8 mature buckwheat plants with their flowers removed was placed in the centre of each enclosure before the dimethoate was applied.

Dimethoate (Lagon® 480 E; Loveland Canada Products Inc., Ontario, Canada) was diluted in water to reach the desired nominal application rate of 2 g a.i./ha and 0.2 g a.i./ha.

These application rates were selected to produce a sub-lethal response in ALB. Application rates of 20 and 200 g a.i./ha were tested in similar preliminary trials in buckwheat (Fagopyrum esculentum) and resulted in 100% mortality of female ALB several days after application as well as high mortality in a second cohort of bees introduced a week after application (Frewin personal comm.1). Each plot was sprayed at approximately 80% bloom with the corresponding treatment

® using a CO2 backpack sprayer and handheld boom with Teejet 8002VS nozzles (Teejet

Technologies, Springfield, IL) at 19:00 on 9 July 2018. Control plots were sprayed with water only. Each treatment was replicated four times for a total of 12 treated plots. The following day, an enclosure and nest box were placed on each plot as described in section 4.2.3, and groups of

10 female and 15 male ALB adults were released directly into each enclosure 19 h after spraying.

1 Dr. Andrew Frewin, [email protected] 67

When the enclosures were put on the plots, a 1 m2 section in each of three plots spread diagonally across the field (Figure 4.2) was destructively sampled, and the number of broad- leafed weeds and purple tansy plants was counted. The number of purple tansy plants was used to estimate if any variation in plant density was present across the field.

Bees were observed three times a day on Monday, Wednesday, and Friday between the hours of 9:00 and 13:00 beginning on 11 July. The observation day was moved or skipped if there was rain as ALB activity is reduced when the sun is not out. The observer would enter the enclosure taking care not to let any bees escape, record the time, and then count the number of active bees twice by making a single visual sweep from one side of the enclosure to the other.

Bees were considered active if they were flying at plant canopy height or foraging. The observer would then watch the nest box for 15 min and record the identity of each bee that performed an activity at the nest. A female bee was recorded if it was provisioning a cell (carrying a leaf piece or pollen on its scopa), or repeatedly entering and exiting one or more cells >10 times. This latter activity was considered a sign of the bee needing to reorient itself with its nest (Guédot et al.

2013) and will be referred to herein as “erratic” behaviour. Nectar provisioning of cells was not recorded as it was impossible to differentiate between a bee returning with nectar or a bee performing some other activity.

Figure 4.1. A 3.3 x 3.3 m screened enclosure in a field of blooming Phacelia tanacetifolia (left) and a Megachile rotundata nesting box (right).

After the 15 min observation, the number of active ALB was counted twice more for a total of four measurements of bee activity. The observer would then leave the enclosure after making sure there were no bees on their clothing. The light conditions during each observation were recorded as cloudy (no direct sun), partly-cloudy (a cloud passing over the enclosure at least once), and sunny (direct sun for the duration of the observation).

The number of female ALB residing in the nest boxes at night was recorded every

Sunday and Wednesday after the sun went down by shining a flashlight in each tube of each nest box. The presence of a female in a nest was determined by observing their distinctive abdominal morphology. More than one individual female was never observed in a nest tube. This method was adapted from Pitts-Singer (2015) and was used to estimate the number of females still alive in the enclosures over time.

On the morning of 27 July, 17 d after adults were released, as many ALB as possible were removed from each enclosure with a sweep net. The adult bees were put in a cooler with ice and then placed in a freezer at -18 ºC later in the day. Nest boxes were removed from each enclosure, placed individually in sealed plastic bags, and returned to the lab. ALB were removed with forceps from the nest boxes and placed in a freezer at -18 ºC. Nest boxes were then put into bags made of breathable low-tunnel greenhouse cover fabric, which ALB adults were unable to chew through, and placed in the greenhouse at 21 ± 5 oC and 60 ± 17% RH. Nest boxes were observed twice a week for the emergence of second-generation bees, which were collected, placed in the freezer, and eventually dried and weighed. Second generation bees do not overwinter and emerge during the same season in which they are laid (Richards 1984).

Block 1

Block 2

Block 3

Block 4

Figure 4.2. Plot layout in the positive control semi-field experiment. White squares are control plots, grey squares are plots treated with 0.2 g a.i./ha dimethoate, black squares are plots treated with 2 g a.i./ha dimethoate, and crosshatched plots are plots sampled for Phacelia tanacetifolia density. Plots were grouped into four horizontal blocks in a randomized complete block design.

On September 10, 2018, a week after second generation bees stopped emerging, each brood cell was removed from the nest box, and its mass and position in the nesting tube were recorded (Figure 4.3). A cell at the back of the nesting tube (laid first) would be given a position rank of 1, and each subsequent cell would be given a position rank of 1 higher until the cell closest to the opening of the next tube was reached (Figure 4.3). The cells were then placed in the refrigerator at 8 oC to simulate overwintering conditions.

The adult ALB collected from the enclosures (n = 72) were dried, weighed, and sexed.

Wing wear measurements were taken from each adult: These measurements can be used to estimate the age of ALB in the wild through gradual wing wear (O’Neill et al. 2015) and thus can be used to compare the relative activity levels of ALB of a known age. Differences in wing wear between treatments over a uniform duration may then be measured as a sub-lethal effect of pesticide exposure. Wing wear measurements were taken following the methods of O’Neill et al.

(2015) (Figure 4.3).

4.2.5 Release rate experiment

Enclosures were erected in the same manner described in section 4.2.2 on 12 July 2018.

The bees were introduced during the morning of 16 July 2018. Due to patchy purple tansy growth and no need for enclosure separation as in the dimethoate-treated plots, plot locations were chosen that had visually similar amounts of bloom, regardless of how close together they were. Potted buckwheat was not placed in these enclosures; instead, any naturally occurring broad-leafed weeds were left in the plots for the ALB to use as a supplementary leaf source. The flowers were removed from the weeds to prevent ALB from foraging on plants other than purple tansy.

The three treatments consisted of the introduction of 10, 15, and 20 female ALB. These female release rates replicated those tested by (Frewin et al. 2019) on buckwheat and alfalfa.

Treatments were assigned to plots in a complete randomized design with no blocks (Figure 4.4).

Each enclosure also received 15 males. This consistent release rate of males was meant to provide enough males in the tents to have all females mated, as some males may die over the course of the experiment. The ratio between males and females was therefore 1.5, 1, and 0.75 and never reached the ration of 3 males per female where male harassment can affect female reproduction (Rossi et al. 2010). Further work will need to be performed to assess an optimal ratio of males to females. Bees were not individually marked for this experiment.

Alfalfa leafcutting bees were observed twice per day for 17 d using the protocols described in section 4.2.4. Data on the light conditions, mean number of active bees per enclosure, number of provisioning trips per enclosure, and number of bees behaving erratically per enclosure were recorded. Nest boxes were removed on 3 August, and brood cells were separated, weighed, and placed in overwintering conditions on 18 September using the methods described in section 4.2.4.

Plant density was measured across the field in the same manner as section 4.2.4 except that randomly selected 0.5 m2 quadrats were sampled to reduce labour time.

On 4 August, 10 s collections with a sweep net were carried out in each enclosure to measure the density of tarnished plant bug (TPB, Lygus lineolaris Palisot). Late-season infestations of TPB in the enclosures may have reduced the amount of forage available for the

ALB. As the TPB infestation was unexpected and their numbers were not noticeable in the positive control experiment, this metric was not recorded for that experiment.

Rank 0 1 2 3 4 5

Figure 4.3. Examples of wing wear visible on female Megachile rotundata forewings. Rank refers to the wing wear rank calculated using the method proposed by O’Neill et al. (2015): The apical margin of each forewing was rated from 0–5, where 0 = completely intact apical margin, 1 = 1–2 incisions on the apical margin, 2 = 3–10 incisions on the margin, 3 = > 10 incisions but some apical margin intact, 4 = no apical margin remaining, but incisions <0.5 the width of the distal submarginal cell, 5 = same as 4 but with incisions 0.5 > < 1.0 the width of the distal submarginal cell, 6 = same as 5 but with incisions > the width of the distal submarginal cell.

Figure 4.4. Plot layout in the release rate semi-field experiment. White squares represent 10 female and 10 male Megachile rotundata released per screened enclosure, grey squares represent 15 female and 10 male M. rotundata released per enclosure, black squares represent 20 female and 10 male M. rotundata released per enclosure, and crosshatched plots are plots sampled for Phacelia tanacetifolia density.

4.2.6 Statistical analyses

All analyses were completed in R version 3.4.3 (R Core Team 2017), and all tests were conducted using a significance threshold of alpha = 0.05. The effect of treatment on mean ALB activity in the enclosures during each observation period (measured as the number of bees foraging or flying at canopy height during one visual sweep of the enclosure), the total number of provisioning trips (pollen and leaf trips combined) in each observation period, and the total number of occurrences of erratic behaviour (exiting and entering one or more nesting tubes >10 times) in each observation period were all analyzed separately with linear mixed effects models using the lme function with the maximum likelihood method in the nlme package (Pinheiro et al.

2017). For these analyses, variance was partitioned into the fixed effects of the interaction between treatment (either rate of dimethoate or ALB release rate) and days after release, and the interaction between weather and time of day, while the random effects were partitioned by enclosure and observer.

Mean brood cell mass was also analyzed with a linear mixed effects model using the lme function and the maximum likelihood method (Pinheiro et al. 2017). Variance was partitioned into the fixed effects of the interactions between treatment (dimethoate rate or release rate), and the random effect of position in the nest tube. The total number of brood cells produced in each enclosure, wing wear of adult female ALB at the end of the experiment, and the mass and number of second generation bees of both sexes were analyzed with a linear model using the base function lm in R (R Core Team 2017) with treatment as a fixed effect.

The mean number of bees observed each night was analyzed with a generalized linear mixed-effects model using the glmer function with a Poisson distribution (Bates et al. 2015).

Variance was partitioned into the interaction between the fixed effects of treatment and day, and the random effect of enclosure.

The significance of each explanatory variable was tested with a Wald test (Agresti 1990) using the anova function (test = “Chisq”) in R (R Core Team 2017).

Data for both experiments were analyzed with the same statistical methods, except that the random effect of enclosure placement was analyzed as blocks for the positive control experiment, and as a continuous position variable for the release rate experiment because the release rate experiment could not be arranged in a complete randomized block design.

Additionally, every analysis of the release rate experiment included the number of TPB in the enclosure as a fixed effect. Model fit was estimated using the pseudo R2 value generated with the r.squared GLM function in the MuMIn package in R (Barton 2018).

4.3 Results

4.3.1 Positive control experiment

Bees were observed foraging from the purple tansy flowers. All females returning to the nest with pollen (n = 366) were carrying purple tansy pollen (purple tansy pollen is dark purple and easily distinguished from the pollen of other plants). Nearly 50% of the bees were recovered at the end of the experiment.

Treatment did not have a significant effect on any metric (Table 4.1). The number of provisioning trips during each observation period (Figure 4.5, Table 4.1) increased between 9:00 and 13:00 (P < 0.0001) and was significantly affected by the light conditions (P = 0.0006), where sunny weather was associated with higher provisioning activity than cloudy weather. The variability in the total number of provision trips in each enclosure was not well captured by this model (pseudo R2 = 0.26; Figure 4.5A).

The mean activity in each enclosure during each observation period (Figure 4.5, Table

4.1) decreased over the course of the experiment (P < 0.0001), increased over the course of each day (from 9:00 to 13:00) (P = 0.0002), and was positively influenced by the mean daily temperature (P < 0.0001) (Figure 4.5B, Table 4.1). The model predicted roughly half of the variability in the activity in each enclosure during the observation periods (pseudo R2 = 0.58).

Counts of erratic behaviour during each observation period decreased over the course of the experiment (P < 0.0001) (Figure 4.5C, Table 4.1). The model did not predict the variability of the total number of erratic bees during each enclosure observation (pseudo R2 = 0.25).

The number of ALB females residing inside the nest boxes during the night decreased throughout the experiment (P < 0.0001) but was not significantly affected by treatment (P =

0.688) (Figure 4.6, Table 4.2). The model captured around half of the variation in the number of female ALB spending the night in nest boxes (pseudo R2 = 0.58).

The mass of brood cells produced was not significantly affected by treatment (P = 0.762).

Brood cell mass was positively correlated with position in the nest tube: Cells closer to the back of the nest tube (produced first by the female) were significantly larger than those closer to the front (produced last by the female) (P = 0.019) when each nest tube was included as a random variable (Figure 4.6, Table 4.2). This model did not capture much of the variation between individual cell mass (pseudo R2 = 0.19).

A) B) C)

Figure 4.5. Activities performed by adult female Megachile rotundata on Phacelia tanacetifolia plots treated with either water (control) or dimethoate at 0.2 g a.i./ha or 2 g a.i./ha. A) The mean number of provisioning trips recorded each day, B) the mean number of active bees observed each day, and C) the mean number of bees behaving erratically (entering several nesting tubes > 10 times) each day. The solid lines represent the trend for the control, the dotted lines represent the trend for dimethoate applied at 0.2 g a.i./ha, and the dashed lines represent the trend for dimethoate applied at 2 g a.i./ha. Bars represent the standard error of all observations on each day for each treatment.

Table 4.1. Model parameters and results of Wald test for several behaviour metrics of Megachile rotundata females recorded during the positive control experiment. Parameter estimates represent the magnitude and direction of the effect of each parameter. Megachile rotundata were maintained on plots of Phacelia tanacetifolia that had been treated with water (control), 0.2 g a.i./ha dimethoate, or 2 g a.i./ha dimethoate. Response variable Pseudo R2 Parameter Estimate (95% CI) Wald X2 Wald Pr X2 total number of provisioning 0.26 Treatment (g a.i./ha) 0.03 (0.63) 0.01 0.924 trips recorded for each Days after release (d) -0.01 (0.06) 0.78 0.379 marked bee during each Time of day (h) 0.04 (0.17) 20.46 < 0.0001 observation Daily mean temperature (ºC) -0.04 (0.28) 0.08 0.776 Light conditions (partly cloudy) 0.11 (1.45) 7.79 0.0006 Light conditions (sunny) 1.62 (1.46) 7.79 0.0006 mean number of active bees 0.58 Treatment (g a.i./ha) 0.26 (0.92) 0.93 0.357 during each observation Days after release (d) -0.22 (0.08) 130.60 < 0.0001 Time of day (h) 0.21 (0.19) 14.29 0.0002 Daily mean temperature (ºC) 0.55 (0.33) 38.78 < 0.0001 Light conditions (partly cloudy) 0.31 (1.76) 2.22 0.112 Light conditions (sunny) 0.66 (1.65) 2.22 0.112 total number of bees 0.25 Treatment (g a.i./ha) 0.05 (0.18) 0.25 0.627 behaving erratically during Days after release (d) -0.13 (0.04) 59.76 < 0.0001 each observation Time of day (h) 0.02 (0.09) 0.12 0.729 Daily mean temperature (ºC) 0.10 (0.15) 2.51 0.115 Light conditions (partly cloudy) 0.73 (0.74) 1.90 0.153 Light conditions (sunny) 0.58 (0.69) 1.90 0.153

There was no significant effect of treatment on the number of cells produced (P = 0.630), second generation emergence of males (P = 0.815) or females (P = 0.574), mass of second generation males (P = 0.334) or females (P = 0.814), or adult female ALB wing wear (P =

0.727) (Table 4.3).

The variation in the number of purple tansy plants was <15% across the field in the three sampling plots (863, 802, and 705), and so the number of purple tansy plants was not included as a covariate in the analyses.

4.3.2 Release rate experiment

Higher release rates resulted in significantly higher wing wear of ALB adult females (P =

0.030) (Table 4.4). Release rates also had a significant positive effect on the mass of individual brood cells produced by female ALB (P = 0.023) but did not have a significant effect on any other metric (Table 4.4). There was no significant effect of TPB abundance or light conditions for any metric (Table 4.6).

The total number of provisioning trips during each observation period decreased over the course of the experiment (P < 0.0001) and was positively affected by the mean daily temperature

(P < 0.0001). The model accounted for around a third of the variation in the number of provisioning trips (pseudo R2 = 0.32).

The mean ALB activity during each observation period was negatively affected by the duration of the experiment (P < 0.0001) and positively affected by the daily mean temperature (P

< 0.0001). The model did not account for the variability of ALB activity in the enclosures very well (pseudo R2 = 0.27).

A) B)

Figure 4.6. A) The mean number of Megachile rotundata females residing the nest boxes at night for each treatment and B) the mean mass of individual M. rotundata brood cells produced grouped by their treatment and position in the nest tube. Cells closest to the back of the nest box were labeled 1. Megachile rotundata were maintained on plots of Phacelia tanacetifolia that had been treated with water (control), 0.2 g a.i./ha dimethoate, or 2 g a.i./ha dimethoate. The solid lines represent the trend for the control, the dotted lines represent the trend for dimethoate applied at 0.2 g a.i./ha, and the dashed lines represent the trend for dimethoate applied at 2 g

a.i./ha. Bars represent the standard error of all observations on each day for each treatment.

Table 4.2. Model parameter estimates, results of Wald test, and estimated model fit for two behaviour metrics of Megachile rotundata females recorded during the positive control experiment. The parameter estimates represent the magnitude and direction of the effect of each parameter on the model. Megachile rotundata were maintained on plots of Phacelia tanacetifolia that had been treated with water (control), 0.2 g a.i./ha dimethoate, or 2 g a.i./ha dimethoate. Response variable Pseudo R2 Parameter Estimate (95% Wald X2 Wald Pr X2 CI) total number of females per 0.58 Treatment (g a.i./ha) 0.16 (1.35) 0.17 0.689 nest box at night per Days after release (d) -0.22 (0.11) 43.00 <0.0001 observation period mass (mg) of individual M. 0.19 Treatment (g a.i./ha) 0.77 (4.80) 0.10 0.762 rotundata brood cells Cell depth -1.72 (1.41) 5.63 0.019 produced

Table 4.3. The arithmetic means of the total number of brood cells produced by female Megachile rotundata, mean mass of those cells, total number of second generation adults that emerged, mean mass of the second-generation adults bees, and the percent distribution of wing wear ratings of adult females that were collected in each treatment replicate at the end of the experiment. Bees were kept in screened enclosures in groups of 10 females (F) and 15 males (M) for 17 d on plots of Phacelia tanacetifolia that were sprayed with either water (control) or dimethoate at 0.2 g a.i./ha or 2 g a.i./ha.

Treatment Total number of Mean cell mass Total number of Mean mass (mg) of Adult female wing cells produced (±SE) (mg) (±SE) second-gen bees (±SE) second-gen adults (±SE) wear* (%) control 30.7 (5.24) 71.0 (5.61) F 3.00 (1.00) F 11.84 (1.56) 4 – 63.64 M 4.50 (0.87) M 10.43 (0.59) 3 – 36.36 0.2 g a.i./ha 27.0 (4.73) 71.3 (1.96) F 1.50 (0.50) F 13.86 (0.90) 4 – 66.67 M 4.50 (1.50) M 9.91 (0.55) 3 – 33.33 2 g a.i./ha 32.8 (8.84) 67.9 (1.86) F 1.75 (0.75) F 12.64 (0.53) 4 – 58.33 M 5.00 (3.00) M 9.37 (0.55) 3 – 41.67 *Wing wear measurements were determined using the methods proposed by O’Neill et al. (2015).

The total counts of erratic behaviour during each observation period decreased over the course of the experiment (P < 0.0001) and decreased with increasing temperature (P = 0.020).

The variability in erratic behaviour was not captured well by this model (pseudo R2 = 0.20).

The number of ALB adults present in the nest box at night decreased over the course of the experiment (P = 0.0001) but was not significantly affected by treatment (P = 0.321) or TPB density within treatment plots (P = 0.151) (Figure 4.8, Table 4.6). The variation in the number of females residing in the nest boxes was captured well by this simple model (pseudo R2 = 0.90).

The mass of individual brood cells was significantly correlated with their position in the nest tube (P < 0.001): Cells closer to the back of the nest tube were larger than those near the entrance (Figure 4.8, Table 4.6). Individual cell mass was also positively affected by treatment as stated above. The model accounted for approximately a third of the variation in cell weight

(pseudo R2 = 0.35).

Release rate did not have a significant effect on the number of cells produced (P = 0.123), the number of second generation females (P = 0.658) or males (P = 0.890) that emerged, or the mass of second generation males (P = 0.071) or females (P = 0.679) (Table 4.4).

The number of purple tansy plants was consistent across the three sampling plots (402,

403, and 396) and so was not used as a predictive variable.

Table 4.4. The arithmetic means and SE for the total number of brood cells produced by Megachile rotundata females, mean mass of those cells, total number of second-generation bees (M = male, F = Female) that emerged and did not diapause, mean mass of those second-generation bees, and the percent distribution of wing wear ratings of the adult females that were collected at the end of the experiment. The ALB were kept for 17 d in screened enclosure on Phacelia tanacetifolia and were released into these enclosures at a rate of 15 males and one of three different female release rates (10, 15, or 20).

Release Total number of cells Mean cell mass Total number of second- Mean mass (mg) of Adult female rate produced (±SE) (mg) (±SE) gen bees (±SE) second-gen adults (±SE) wing wear* % 10 F 10.4 (2.42) 67.7 (3.09) F 2.00 (7.07) F 14.1 (1.02) 4 – 52.94 M 0.40 (0.25) M 13.9 (2.90) 3 – 41.18 2 – 5.88 15 F 14.8 (3.02) 78.3 (3.35) F 2.25 (1.11) F 11.7 (0.65) 4 – 47.83 M 2.40 (1.50) M 12.0 (0.62) 3 – 52.17 20 F 18.4 (4.46) 77.1 (2.43) F 1.60 (0.40) F 13.4 (1.08) 4 – 84.21 M 0.60 (0.60) M 8.40 (0.46) 3 – 15.79 *Wing wear measurements were determined using the methods proposed by O’Neill et al. (2015).

A) B) C)

Figure 4.7. Activities performed by Megachile rotundata adult females over the course of 17 d in screened enclosures on plots of Phacelia tanacetifolia at three different release rates: 15 males with either 10 females, 15 females, or 20 females. Measurements were taken twice a day three times a week for A) mean number of provisioning trips (pollen and leaves) each day for each treatment, B) mean number of active bees during each observation each day for each treatment, and C) mean number of bees behaving erratically (entering > 10 nest tubes without leaving to forage) during each observation each day for each treatment. The solid, dotted, and dashed lines represent the trend for the release rate of 10, 15, and 20 females per enclosure, respectively. Bars represent the standard error of all observations for all enclosures in a treatment each day.

Table 4.5. Model parameters and results of Wald tests for several behaviour metrics of Megachile rotundata females recorded during the release rate experiment. The fit of the entire model was estimated with the pseudo R2. A high test statistic and low P value for a Wald test indicates a significant effect of the parameter on the model. Megachile rotundata were released at a rate of 10 males and either 10, 15, or 20 females in screened enclosures in a field of Phacelia tanacetifolia for 2 weeks.

Response variable Pseudo R2 Parameter Estimate (95% CI) Wald X2 Wald Pr X2 total number of provisioning 0.32 Treatment (females/enclosure) 0.05 (0.13) 0.53 0.481 trips recorded for each Days after release (d) -0.11 (0.06) 18.69 < 0.0001 marked bee during each Time of day (h) 0.03 (0.05) 0.41 0.522 observation Daily mean temperature (ºC) 0.60 (0.28) 19.38 < 0.0001 Light conditions (partly cloudy) 0.58 (1.05) 0.70 0.499 Light conditions (sunny) 0.68 (1.20) 0.70 0.499 Lygus lineolaris density 6.00 e-4 (3.40 e-3) 0.12 0.736 mean number of active bees 0.27 Treatment (females/enclosure) 0.02 (0.12) 0.09 0.764 observed during each Days after release (d) -0.10 (0.06) 16.90 < 0.0001 observation Time of day (h) 0.02 (0.05) 0.09 0.760 Daily mean temperature (ºC) 0.70 (0.30) 21.94 < 0.0001 Light conditions (partly cloudy) 0.89 (1.14) 2.16 0.120 Light conditions (sunny) 1.37 (1.30) 2.16 0.0120 Lygus lineolaris density 6.66 e-4 (3.05 e-3) 0.18 0.676 total number of bees 0.19 Treatment (females/enclosure) 1.73 e-3 (0.04) 0.04 0.848 behaving erratically during Days after release (d) -0.06 (0.02) 25.68 < 0.0001 each observation Time of day (h) -2.62 e-3 (0.02) 2.00 e-5 0.996 Daily mean temperature (ºC) -0.14 (0.12) 5.54 0.020 Light conditions (partly cloudy) 0.10 (0.43) 0.18 0.834 Light conditions (sunny) 0.06 (0.49) 0.18 0.834 Lygus lineolaris density 6.88 e-4 (9.19e-4) 2.15 0.168

A) B) C)

Figure 4.8. A) The mean number of Megachile rotundata females residing in nest boxes overnight over the course of the experiment for each treatment, B) mean mass of individual M. rotundata brood cells produced in each treatment grouped by their position in the nest tube (brood cells closest to the back of the nest box were labeled 1), and C) wing wear of adult females retrieved at the end of the experiment. Megachile rotundata were maintained in screened enclosures in a field of Phacelia tanacetifolia for 17 d with a release rate 15 males per enclosure and a treatment of either 10, 15, or 20 females per enclosure. The solid, dotted, and dashed lines in A and B represent the trend for the release rate of 10, 15, and 20 females per enclosure, respectively. Error bars represent the standard error of bees or brood cells sampled within each group.

Table 4.6. Model parameter estimates, model fit, and results of Wald test for several behaviour metrics of Megachile rotundata females recorded during the release rate experiment. The fit of the entire model was estimated using the pseudo R2. Megachile rotundata were released at a rate of 10 males and either 10, 15, or 20 females in screened enclosures in a field of Phacelia tanacetifolia for 2 weeks.

Response variable Pseudo R2 Parameter Estimate (95% CI) Wald X2 Wald Pr X2 number of female M. 0.90 treatment 0.28 (0.35) 1.06 0.229 rotundata residing in (females/enclosure) each nest box at night days after release (d) -0.16 (0.08) 17.45 0.0001 over the course of the Lygus lineolaris density 6.99 e-3 (8.92 e-3) 2.36 0.151 experiment mass (mg) of 0.35 treatment 1.08 (0.91) 5.33 0.023 individual M. (females/enclosure) rotundata brood cells cell depth -8.22 (2.88) 31.73 < 0.0001 produced Lygus lineolaris density 0.02 (0.02) 2.26 0.136

4.4 Discussion

In my study, dimethoate did not affect ALB behaviour, reproduction, or development at

0.2 or 2 g a.i./ha. Therefore, higher rates of dimethoate will need to be tested to determine the rate(s) that can be used as a positive control. In a similar study, ALB were confined to buckwheat plots that had been treated with dimethoate at 20 g a.i/ha or 200 g a.i./ha, and all ALB died within a few days of dimethoate application (Frewin, personal comm.1). In that study it was found that buckwheat was a low-quality crop for ALB feeding and reproduction, so the high mortality from dimethoate could not be statistically separated from the negative effects of being restricted to buckwheat. As well, differences in plant morphology between purple tansy and buckwheat could have played a role in the high mortality experienced previously. However, the rapid onset of mortality at both application rates indicated that dimethoate would be toxic at 20 g a.i./ha and 200 g a.i./ha to ALB regardless of crop type as the bees died well before starvation could be expected to affect them, or before they had much time to interact with the crop. I tested lower rates to avoid similar levels of mortality, but 2 and 0.2 g a.i./ha appear to be too low to cause obvious sub-lethal effects. Therefore, I suggest further experimentation with a rate between 2 and 20 g a.i./ha to determine an optimal application rate of dimethoate.

Tier II RA methods for Osmia suggest that dimethoate be applied to the surrogate crop

(Brassica napus or purple tansy) a rate of 75 g a.i./ha while bees are in flight to produce sub- lethal effects (Knäbe et al. 2017). Both of these recommendations will result in Osmia being exposed to more dimethoate ALB than in preliminary trials (Frewin, personal comm.1) that reached 100% female mortality in a few days (20 g a.i./ha and ALB introduced 19 hr after spray). Osmia may therefore be less sensitive than ALB to dimethoate if they only present sub- lethal effects at that rate and application method. If upon further testing ALB is found to be more

1 Dr. Andrew Frewin, [email protected] 91 sensitive than Osmia to a range of pesticides, ALB may be a more conservative surrogate for

North American solitary bee pesticide RA. The lack of effect of dimethoate in my study is unlikely to be the result of ALB not being exposed to dimethoate on purple tansy, however no residue analyses were performed to confirm the presence of dimethoate as funds were not available. After spray application, ALB would be exposed to dimethoate through contact with purple tansy pollen, as all pollen trips recorded in the positive control experiment were from purple tansy, as noted by its distinctive dark purple colour. Further, as the flowers of all other plants in the enclosures were removed throughout the experiment, and ALB cannot survive for 2 wk without a sugar source, the ALB must have consumed nectar from purple tansy flowers. Even flowers that opened after application would have been contaminated as dimethoate has noticeable systemic activity (Van Scoy et al. 2016). Treated leaves (weeds and buckwheat) used for nest construction were another source of dimethoate exposure for ALB. I recommend that future experiments include residue analyses on the flowers, leaves, and adult females after the experiment to confirm exposure to dimethoate.

I observed a high occurrence of erratic behaviour at the beginning of both of my experiments (Figure 4.5, Figure 4.7). Erratic behaviour likely represents the female looking for a new nest or being unable to find its current nest using olfactory (Guédot et al. 2013) or visual

(Tezuka and Maeta 1993) cues. Although not observed in my experiment, pesticide exposure may have a similar negative effect on ALB nest recognition as on HB navigation, recognition of olfactory cues, and behaviour (Belzunces and Tchamitchian 2012). Disruption of these faculties by pesticide intoxication may thus cause an increase in erratic behaviour. I did not observe an effect of dimethoate on erratic behaviour in my experiment, but dimethoate did not appear to

92 affect the bees by any metric at these rates. Therefore, erratic behaviour may still prove to be a useful metric for future assessment of pesticide intoxication with higher dimethoate rates.

Release rate did not have a significant effect on the number of brood cells produced in each enclosure; however, there did appear to be a trend for higher release rates to result in higher brood cell production. Further analyses with higher release rates or higher sample sizes may find a significant association between release rate and brood cell production. Frewin et al. (2019) recommended a release rate higher than 15 female ALB per enclosure using identical experimental methods. High ALB densities (>9.5 bees/m2) have been shown to limit reproductive success and increase larval mortality (Pitts-Singer and Bosch 2010), and a high proportion of females to available nesting sites can also reduce the number of progeny (Mayer

1994). The highest release rate in my experiment (20 female ALB per enclosure) resulted in a density of 1.6 bees/m2, so it appears to be possible to further increase the release rate without issue. In contrast, tier II methods for Osmia suggests a female density of 0.5–1 bees/m2 (Knäbe et al. 2017). ALB may be more cost-effective than Osmia for tier II RA as the relatively high densities that can be achieved with ALB will help to increase the available sample size.

Higher release rates resulted in higher female ALB wing wear and heavier brood cells produced in each enclosure over the course of the experiment, although the effects on both metrics were small (0.03 wing wear unit rank/additional female and 0.001 g cell mass/additional female, respectively). These results suggest that as female release rate increases, each individual bee produces larger, better provisioned cells, and flies more as a result (indicated by higher wing wear). Pitts-Singer and Bosch (2010) reported that ALB cell production is reduced at high release rates (the equivalent of 116 females per enclosure in my experiment). This indicates that there as an optimal release rate for reproduction below 116, but above 20 females per tent.

Sex ratio changed according to release rate (1.5, 1, 0.75 males per female for the release rates of 10, 15, and 20 females, respectively) and may have had some effect on the outcomes.

However, these rates still remained below 3 males per female, the rate previously recorded as having a negative effect on female reproduction (Rossi et al. 2010). I recommend that future experiments with higher female release rates use proportional male release rates. These release rates will be higher than the ones in my study and should allow all the females to be successfully mated even if the male population is reduced (males are likely to be more susceptible to dimethoate based on their smaller body size, and may experience mortality at doses that are sub- lethal to females).

ALB appeared to reproduce well on purple tansy in my study, which adds further support to its use as a surrogate crop for tier II RA as suggested by Frewin et al. (2019) and Artz and

Pitts-Singer (2015). Poorly provisioned juvenile ALB are more likely to emerge before diapause

(second generation bees) and are detrimental to population growth (Fischman et al. 2017). The emergence of second-generation bees may then be used as an indicator of provision quality.

There are few studies that focus on ALB reproduction on purple tansy, but approximately 20% of larvae produced on alfalfa have been shown to emerge as second- generation bees (Fischman et al. 2017) and in some cases the number of second-generation bees can be as high as 50% (Pitts-

Singer and Cane 2011). In both cases the release rate was not recorded and the data were collected in an open field. The second-generation emergence in my experiments was slightly lower, with the controls of the positive control experiment producing approximately 17% second-generation bees and the highest release rate of 20 females producing approximately 10% second-generation bees. Therefore, the lower emergence of second generation bees in my study compared to other crops is indicative of the quality of purple tansy as a surrogate crop.

I did not observe females returning to their nest with purple tansy leaves, possibly because purple tansy leaves are too narrow for use by ALB (Horne 1995b). However, I found that the pots of buckwheat in each positive control experimental plot were almost entirely defoliated by the ALB after only 5 d (Figure 4.9). Buckwheat leaves have high surface area and low toughness (resistance of the leaf to tearing), providing optimal leaf composition for ALB to use in nest construction (Horne 1995a, b). Introducing a new pot of buckwheat with flowers removed every few days could provide the female ALB in a purple tansy-dominated enclosure a surplus of leaf material for cell production in future experiments. Using potted plants would prevent the need to seed multiple plant species in the field at the same time and, along with the implementation of weed control, allow experimenters to keep the amount of available nesting material consistent across all plots.

Female ALB in the positive control experiment, including the control treatments, produced roughly three times as many brood cells per female as those in the release rate experiment despite both experiments occurring concurrently in the same field. The higher cell production in the positive control experiment may have been due to limited availability of leaves for nest building in the release rate experiment, as I was unable to place pots of buckwheat in that experiment. It may also have been due to the heavy TPB infestation in the release rate experiment. While there was no effect of TPB abundance on any metric within the release rate experiment, the TPB population may have been far larger in the release rate experiment than the positive control experiment. Unfortunately, the number of TPB in the enclosures in the positive control experiment was not recorded, because TPB numbers were too low at the time to be noticed as a potential problem. It is possible that in the week after the positive control experiment was completed, the TPB population in the field increased to levels that directly

95 affected flower availability to ALB, either through damage to the flowers or direct competition with the ALB. In addition to reduced leaf availability, flower damage caused by TPB would have put restrictions on the ability of ALB females to provision their nests. There is no damage threshold for TPB on purple tansy or safe method to control TPB populations in the presence of

ALB. I recommend developing stronger control methods for TPB that will not harm ALB, as well as developing a damage threshold for when TPB populations will impact ALB nest production.

I recommend continuing to record temperature throughout the day, time of observation, and light conditions for future tier II studies using ALB. Female ALB foraging is dependent on temperature and light intensity, and some nest building activities have been associated with specific times of day (Klostermeyer and Gerber 1969). Daily maximum temperature and time of observation captured significant variation in most of the models for both of my experiments.

These data are also easy to collect and add little cost to an experiment while providing valuable statistical information.

Figure 4.9. Potted buckwheat plants placed in a screened enclosure on plots of Phacelia tanacetifolia containing 10 adult female Megachile rotundata after 2 d (left) and 6 d (right).

My study is the first to address what I refer to as erratic behaviour in ALB as a measure of sub-lethal effects of pesticides. Osmia tier II RA (Knäbe et al. 2017) observations include recording the number of bees that enter the nest during the observation period and do not distinguish between foraging trips (nectar or pollen), mud collection trips for cell construction, or behaviour equivalent to erratic behaviour. As future method development for ALB RA may be based off of Osmia RA, I advise against pooling all nest entrance events for ALB research (i.e., provisioning trips and erratic behaviour) because erratic behaviour (entering nests >10 times without leaving to collect more provisions) could greatly inflate the counts of actual nest provisioning activities. Therefore, I recommend continuing to divide this metric into three behaviours when possible: clear provisioning trips of pollen or leaves, and erratic behaviour at the nest box. The need for such nuanced methodological considerations is another indication that

RA methods for one species cannot be directly applied to another, even within the same family.

I recommend the use of wing wear ratings as proposed by O’Neill et al. (2015) in tier II pesticide RA with ALB. On the basis of wing wear, I was able to detect a difference in activity in the release rate experiment, and wing wear may be useful for detecting changes in activity due to pesticide exposure when rates are higher than in my positive control experiment. Wing wear data are easy to collect: A newly trained individual was able to process 200 specimens in 3.5 h for my study.

The enclosures used in my experiments could be used in future tier II RA method development for ALB. Their small size allows for a high number of replicates in a field; and they are relatively inexpensive (~$200 CAD), commercially available, and easy to set up and take down over the course of one field season (EPA 2012, Knäbe et al. 2017). Other ALB researchers have used similarly sized enclosures (Pitts-Singer and Bosch 2010, Fischman et al. 2017, Frewin

98 et al. 2019) with success. The only issue is that over the course of 2–3 seasons of use, some of the enclosures were irreparably damaged by heavy wind. The durability of these enclosures may need to be considered in future method development.

The ALB nest boxes in my experiment were placed at crop canopy height which was uncomfortably low for most observers. Future studies could put the nest boxes at a standing height to reduce strain on the observer but should ensure that the nest boxes still receive direct sunlight.

I recommend a lower seeding rate than the one used in my experiment (19 kg/ha) as the purple tansy density was very high (808 ± 62.26 plants/m2) and many plants were choked out and unable to bloom. This rate was based on Gradish et al. (2016) and Frewin et al. (2019) who both seeded at 33 kg/ha to ensure a strong stand in sandy soil. Although rates as low as 6.3 kg/ha have been suggested by the United States Department of Agriculture when used as a cover crop

(Kilian 2016), I would suggest 8–13 kg/ha to ensure a maximum number of flowers.

In conclusion, I recommend that further ALB tier II RA method development use similar methods as in my study. My study is the first examination of dimethoate as a positive control in

ALB tier II RA on purple tansy. Based on my results and previous unpublished work (Frewin, personal comm.1), application rates between 2 and 20 g a.i./ha should be tested in the future. I recommend further testing with a release rate of female ALB above 1.6 bees/m2 or 20 females per enclosure, the highest rate in this experiment. My results are similar to Frewin et al. (2019) who suggested release rates higher than 15 females per enclosure. Higher release rates will increase the number of bees available for experimentation and maximize sample size Purple tansy was a successful surrogate crop for ALB tier II RA and resulted in even higher ALB reproduction when potted buckwheat was introduced to the tents. It is clear that the use of a

1 Dr. Andrew Frewin, [email protected] 99 sweep net at the beginning of the experiment was not an adequate control for all pests, although it was able to remove nearly all of the large pollinators and predators. Future tier II RA method development for ALB will need to incorporate species-specific metrics and the design features that I have discussed. My studies have provided important baseline data for release rate, positive control rates, and semi-field methods that will be important for future tier II pesticide RA method development using ALB.

5 Conclusions and recommendations

Current pesticide risk assessment (RA) protocols primarily focus on honey bees (HB,

Apis mellifera), and may not adequately protect solitary bees in the field (EPA 2013, Sgolastra et al. 2018). Therefore, solitary bee RA methods should be developed that will more fully account for the differences between solitary bee and HB life history, biology, and sociality that affect their pesticide exposure profiles (Sgolastra et al. 2018). The overall goal of my research was to contribute to the development of pesticide RA methods using the alfalfa leafcutting bee (ALB,

Megachile rotundata) as a surrogate species. To accomplish this, I focused on three objectives across three chapters:

1. Generate toxicity data for potential positive controls for lab (tier I) topical and

chronic oral trials as well as semi-field (tier II) trials,

2. Generate toxicity data for a range of pesticides in tier I topical and chronic oral

trials and compare these results with HB values, and

3. Improve and optimize methods for tier I) and tier II pesticide RA trials using

ALB.

100

5.1 Potential positive control insecticides for tier I and II pesticide risk assessment using Megachile rotundata.

I examined the toxicity of potential positive controls for tier I and II RA using ALB females. For tier I acute topical assessments, I assessed dimethoate, permethrin, and imidacloprid. I found that dimethoate and permethrin were both acutely toxic when applied topically and produced rapid mortality in ALB, while imidacloprid was acutely toxic but resulted in paralysis of varying degrees that lasted for longer than 72 h. The purpose of a positive control is to demonstrate exposure within an experimental design, and this is best achieved with a rapid, unambiguous response.

I further assessed the toxicity of dimethoate and imidacloprid at several concentrations for use in tier I chronic oral exposure trials over the course of several weeks. I found similar results as with the acute topical applications. Both insecticides were toxic to ALB, but exposure to imidacloprid resulted in extended paralysis of the bees. Individuals were observed feeding once, becoming paralyzed for several hours or days, then feeding again. This type of inconsistent feeding response produced by imidacloprid is likely to reduce its reliability as a positive control.

Additionally, it may be difficult to determine if bees affected by imidacloprid are dead or alive, which may add another source of error if imidacloprid is used.

I assessed the toxicity of dimethoate as a positive control for tier II trials in purple tansy at an application rate of 2 and 0.2 g a.i./ha. There was no observable effect of dimethoate on

ALB. In contrast, previous experiments of similar design on buckwheat saw 100% mortality of

ALB within a few days of dimethoate application at rates of 20 and 200 g a.i./ha, and high mortality in a second cohort of bees that was introduced a week later (Frewin personal comm.1).

These results suggest that dimethoate is acutely toxic to ALB at low doses and may be an

1 Dr. Andrew Frewin, [email protected] 101 effective positive control if the threshold for lethal and sub-lethal effects can be found. I recommend further analysis of the toxic effects of dimethoate at rates between 2 and 20 g/ha with ALB on purple tansy.

I recommend the use of dimethoate in tier I and II, and permethrin in tier I, pesticide RA using female ALB as a surrogate species. I have shown dimethoate to produce a consistent response in tier I trials with ALB, and to be potentially useful in tier II trials when further range finding tests have been completed. Dimethoate is also suggested for use as a positive control in pesticide RA for HB (EPA 2012), Bombus (OECD 2017c), and Osmia (Knäbe et al. 2017).

Further work to evaluate dimethoate as a positive control for ALB will allow for consistency across the RA protocols for different bees.

5.2 Comparison of toxicity of a range of pesticides between Megachile rotundata and Apis mellifera.

In addition to assessing the toxicity of potential positive controls for pesticide RA methods with ALB, I have generated data on a number of other pesticides. My study is the first assessment of acute topical applications of imidacloprid, chlorantraniliprole, and captan on ALB adult females, and the second such assessment of permethrin and dimethoate (Torchio 1973,

Helson et al. 1994). I have also presented the first assessment of chronic oral toxicity of dimethoate and imidacloprid to ALB. Further work to develop a baseline understanding of the response of ALB to a range of common pesticides, and the comparison of these values with those of HB, will be vital to further development of ALB pesticide RA.

It may not be necessary to use acute topical tests for solitary bee RA at all. In current RA practices, a value of 1/10 the HB LD50 is used to encompass the acute sensitivity range of wild bee species (Uhl et al. 2016). All topical LD50 values generated in my study, as well as in Helson

102 et al. (1994) who tested four bee species including HB and ALB, were within this 10-fold assessment factor. In a meta-analysis performed by Uhl et al. (2016) on nine bee species (not including ALB), the 10-fold assessment factor was protective of 95% of species. Therefore, performing acute topical tests on solitary bees may not provide any additional protection, as a product would in almost all cases be considered safe for both solitary bees and HB. The key differences in life history, biology, and sociality between solitary bees and HB are more likely to result in differences in exposure profiles in a tier II or III scenario.

The oral toxicity of dimethoate to ALB that I calculated is similar to that proposed by the

OECD for HB (OECD 2017a). I found 58% mean mortality in female ALB exposed to 0.2 mg/L dimethoate treated sucrose solution for 10 d. This falls within the range (0.2–1 mg a.i./L) proposed to result in 50% mortality of HB within 10 d in the OECD guidelines for tier I chronic laboratory oral exposure. Further assessment of the toxicity of dimethoate and other pesticides to

ALB should be performed to see if ALB are similarly or less susceptible than HB to chronic oral pesticide exposure. If this is the case, as it appears to be in topical tier I assays, then chronic oral tier I trials may also not be necessary in the solitary bee RA process.

5.3 Method development for tier I and II pesticide risk assessment using

Megachile rotundata females.

I do not recommend incorporating the mass of individual bees in tier I acute topical dose response analyses with ALB. Neither the model fit nor precision of the LD50 estimate was improved when mass was incorporated and so the additional time requirement of several hr to each experiment could not be justified. I do recommend recording the mean mass of each population tested (this is easily calculated by weighing the entire population at once) as differences in mean mass between sample populations may significantly change toxicity

103 estimates: ALB adult mass can be affected by the regional, nutritional, and methodological differences between sample populations (O'Neill et al. 2010, Pitts-Singer 2015). The possible variation in sample population mass has not been directly studied in ALB.

The feeding container I designed (Figure 3.1) for use in tier I chronic oral exposure trials can be used to feed ALB sugar solution for several wk. The use of UV paint (Ladurner et al.

2003, Heard et al. 2017) to attract ALB to the feeders appeared to elicit a stronger feeding response in the greenhouse when compared to the growth chamber. The stronger feeding response in the greenhouse may be due in part to the wider spectrum of light in the greenhouse.

The use of UV paint and a way to replicate the feeding response under natural light should be included in future chronic oral tests with ALB. Control mortality in my study was higher than the control mortality in Huntzinger et al. (2008b) and may be reduced by limiting temperature spikes above 30 ºC and providing the bees with shade.

I have furthered method development for tier II semi-field RA in several ways. First, I have provided further support that purple tansy (Phacelia tanacetifolia) provides high quality forage for ALB (Frewin et al. 2019). ALB in my experiments produced a smaller proportion of second-generation bees (non-diapausing adults that are unable to reproduce before fall) than

ALB foraging from alfalfa (Medicago sativa), buckwheat (Fagopyrum esculentum) (Frewin et al.

2019), or free-range forage (Fischman et al. 2017). I recommend adding a 30 cm diameter pot of deflowered buckwheat to the enclosures as buckwheat leaves are preferred by ALB for nest construction (Horne 1995b). The bees in my experiment that were provided with buckwheat produced threefold the number of cells as those that were only provided with purple tansy and weeds. I also recommend drill-seeding purple tansy to 1.25 cm at a rate less than 19 kg a.i./ha, as suggested by Kilian (2016), to prevent the purple tansy from being choked out.

104

I recommend further investigation into release rates of ALB above 20 females per enclosure (1.83 bees/m2) in tier II semi-field trials. This was the highest rate that I tested, and it resulted in the highest number of larval cells produced. A rate of >9.5 females/m2 has been shown to reduce ALB reproduction and larval survival in enclosures on alfalfa (Pitts-Singer and

Bosch 2010), but there is potential for an increase in release rate from my experiment without reaching that threshold. A higher release rate of females in each enclosure would allow a higher sample size in each enclosure and provide more statistical power for analyses. Alternatively, higher release rates would allow a study of similar sample size to occupy less space and require less resources than a study with a lower release rate. Higher release rates of females would need to be paired with a proportionally higher number of available nesting cavities and available males (Pitts-Singer and Bosch 2010, Frewin et al. 2019). However, males should be released at a proportion fewer than 3 males per female as higher proportions of males can reduce female productivity by attempted mating (Rossi et al. 2010). The number of males was consistent in my release rate experiments across treatments. This may have affected the results of my experiment as the ratio of males to females changed with higher female release rates. However, I released at most 1.5 males per female which is below the threshold observed by Rossi et al. (2010) to produce negative effects on reproduction. The optimal proportion of males to females in semi- field enclosures should be examined in more depth.

I recommend measuring “erratic” behaviour, defined as an individual female entering a nest box >10 times without leaving to collect resources, as a distinct behaviour from provisioning trips. Erratic behaviour is not mentioned as a distinct behaviour in tier II pesticide

RA with Osmia (Knäbe et al. 2017) or in other ALB studies focusing on RA (Fauria et al. 2004,

Frewin et al. 2019), although it was observed regularly in all enclosures in my study and

105 preliminary studies. Not accounting for erratic behaviour as a distinct activity may inflate the occurrence of other behaviours. Rapid entering and exiting of a nest cavity scored as nectar gathering trips, for example, would significantly inflate the number of nectar trips recorded.

Erratic behaviour may also be an indication of disruptions in chemosensory or visual homing ability (Tezuka and Maeta 1993, Fauria et al. 2004, Guédot et al. 2013) resulting from pesticide exposure. Therefore, it may be a useful metric to measure pesticide toxicity. In my study there was no effect of dimethoate on any metric, so I cannot conclude that erratic behaviour is a useful estimate of toxicity. The effects of pesticides on ALB, and dimethoate in particular, resulting in erratic behaviour should be examined and characterized more fully for future tier II method development. Either erratic behaviour can be included as a measurement of sub-lethal intoxication or it could be listed as a distinct behaviour to not record to avoid inflation of other measurements. This is the first discussion of erratic behaviour as a distinct and predictable series of actions for use in pesticide RA using ALB females.

The baseline data presented in my thesis along with the novel tier I and II methods I have designed will further the development of methods to assess the toxicity of pesticides to ALB as part of a broader pesticide RA process for solitary bees.

106

6 References

Agresti, A. 1990. Categorical data analysis, 3 ed. John Wiley & Sons, Inc., New York, NY.

Alain, F. Z., E. N. Ieno, N. J. Walker, A. S. Anatoly, and G. M. Smith. 2009. Mixed effects models and extensions in ecology with R. Springer Science + Buisness Media, New York, NY. doi: 10.1007/978-0-387-87458-6.

Alix, A., T. Steeger, C. Brittain, D. Fischer, R. Johnson, T. Moriarty, E. Johansen, F. Streissel, R. Fischer, M. Miles, C. Lee-Steere, M. Vaughn, B. Vaissière, G. Maynard, K. Kasina, R. C. F. Nocelli, C. Scott-Dupree, M. Coulson, A. Dinter, and M. Fry. 2014. Chapter 10: overview of a proposed ecological risk assessment process for honey bees (Apis mellifera) and non-Apis bees, pp. 121–148. In D. Fischer and T. Moriarty (eds.), Pesticide risk assessment for pollinators. John Wiley & Sons, Inc., Oxford, United Kingdom.

Alston, D. G., V. J. Tepedino, B. A. Bradley, T. R. Toler, T. L. Griswold, and S. M. Messinger. 2007. Effects of the insecticide phosmet on solitary bee foraging and nesting in orchards of Capitol Reef National Park, Utah. Environ. Entomol. 36: 811–816. doi: 10.1603/0046- 225x(2007)36[811:eotipo]2.0.co;2

Arena, M., and F. Sgolastra. 2014. A meta-analysis comparing the sensitivity of bees to pesticides. Ecotoxicology. 23: 324–334. doi: 10.1007/s10646-014-1190-1

Artz, D. R., and T. L. Pitts-Singer. 2015. Effects of fungicide and adjuvant sprays on nesting behavior in two managed solitary bees, Osmia lignaria and Megachile rotundata. PLoS ONE. 10: e0135688. doi:10.1371/journal.pone.0135688

Bailey, W. G., H. Lehrer, and P. F. Mills. 1982. Humidity and the pollination activity of Megachile rotundata. Environ. Entomol.11: 1063–1066.

Barton, K. 2018. MuMIn: multi-model inference. R package version 1.42.1. https://CRAN.R- project.org/package=MuMIn.

Bates, D., M. Maechler, B. Bolker, and S. Walker. 2015. Fitting linear mixed-effects models using lme4. J. Stat. Softw. 67: 1–48. doi:10.18637/jss.v067.i01

Belfrage, K., J. Björklund, and L. Salomonsson. 2005. The effects of farm size and organic farming on diversity of birds, pollinators, and plants in a Swedish landscape. Ambio. 34: 582–588.

Belzunces, L. P., and J. B. Tchamitchian. 2012. Neural effects of insecticides in the honey bee. Apidologie. 43: 348–370. doi: 10.1007/s13592-012-0134-0

Biddinger, D. J., and E. G. Rajotte. 2015. Integrated pest and pollinator management—adding a new dimension to an accepted paradigm. Curr. Opin. Insect. Sci. 10: 204–209. doi: 10.1016/j.cois.2015.05.012.

Blacquière, T., G. Smagghe, C. A. M. van Gestel, and V. Mommaerts. 2012. Neonicotinoids in bees: a review on concentrations, side-effects and risk assessment. Ecotoxicology. 21: 973–992. doi: 10.1007/s10646-012-0863-x.

107

Blanchetot, A. 1992. DNA fingerprinting analysis in the solitary bee Megachile rotundata: variability and nest mate genetic relationships. Genome. 35: 681–688.

Bosch, J., and W. P. Kemp. 2001. How to manage the blue orchard bee. Sustainable Agriculture Network, Beltsville, MD.

Cabrera, A. R., M. T. Almanza, G. C. Cutler, D. L. Fischer, S. Hinarejos, G. Lewis, D. Nigro, A. Olmstead, J. Overmyer, D. A. Potter, N. E. Raine, C. Stanley-Stahr, H. Thompson, and J. van der Steen. 2016. Initial recommendations for higher-tier risk assessment protocols for bumble bees, Bombus spp. (Hymenoptera: Apidae). Integr. Environ. Assess. Manag.12: 222–229. doi: 10.1002/ieam.1675.

Chauzat, M.-P., and J.-P. Faucon. 2007. Pesticide residues in beeswax samples collected from honey bee colonies (Apis mellifera L.) in France. Pest. Manag. Sci. 63: 1100–1106. 10.1002/ps.1451.

Devillers, J., A. Decourtye, H. Budzinski, M. H. Pham-Delègue, S. Cluzeau, and G. Maurin. 2003. Comparative toxicity and hazards of pesticides to Apis and non-Apis bees. A chemometrical study. SAR. QSAR. Environ. Res. 14: 389–403. doi: 10.1080/10629360310001623980.

Dobrynin, N. D., and M. Colombo. 2007. Factors affecting pesticides hazard to different kinds of pollinators. Boll. Zool. Agrar. Bachic. 39: 243–253. doi: 10.1080/0005772x.1983.11097927

(EFSA) European Food Safety Authority. 2012. Scientific opinion on the science behind the development of a risk assessment of plant protection products on bees (Apis mellifera, Bombus spp., and solitary bees). EFSA. Journal. 10: 275. doi: 10.2903/j .efsa.2012.2668.

(EFSA) European Food Safety Authority. 2013. Guidance on the risk assessment of plant protection products on bees (Apis mellifera, Bombus spp. and solitary bees). EFSA. J. 11: 268. doi: 10.2903/j.efsa.2013.3295.

(EPA) Environmental Protection Agency. 2012. White paper in support of the proposed risk assessment process for bees. Office of Chemical Safety and Pollution Prevention, Office of Pesticide Programs, Environmental Assessment Division, Washington, D.C.

(EPA) Environmental Protection Agency. 2014. Guidance for assessing pesticide risk to bees. Office of Pesticide Programs, United States Environmental Protection Agency, Washington, D.C.

(EPA) Environmental Protection Agency. 2016. Guidance on exposure and effects testing for assessing risks to bees. Office of Pesticide Programs, U.S. Environmental Protection Agency. Washington, D.C.

Fairey, D. T., L. P. Lefkovich, and D. L. Nelson. 1987. Viability of prepupae of the leafcutting bee Megachile rotundata (Fab.) during extended storage. Can. J. Zool. 65: 1853–1856. doi: 10.1139/z87-281

Fairey, D. T., L. P. Lefkovich, J. A. C. Lieverse, and B. Siemens. 1984. Materials for leafcutting bee (Megachile rotundata F.) shelters in north west Canada. J. Appl. Entomol. 106: 119–122. doi: 10.1111/j.1439-0418.1988.tb00574.x

Fauria, K., R. Campan, and A. Grimal. 2004. Visual marks learned by the solitary bee Megachile rotundata for localizing its nest. Anim. Behav. 67: 523–530. doi: 10.1016/j.anbehav.2003.06.002

108

Fischer, D., and T. Moriarty. 2014a. Appendix 1: elements for a chronic adult oral toxicity study, pp. 179-180. In D. Fischer and T. Moriarty (eds.), Pesticide risk assessment for pollinators. John Wiley & Sons, Inc., Oxford, United Kingdom.

Fischer, D., and T. Moriarty. 2014b. Pesticide risk assessment for pollinators, First ed. Society of Environmental Toxicology and Chemistry (SETAC). John Wiley & Sons, Inc., Oxford, United Kingdom.

Fischer, D., and T. Moriarty. 2014c. Chapter 1: introduction, pp. 1-2. In D. Fischer and T. Moriarty (eds.), Pesticide risk assessment for pollinators. John Wiley & Sons, Inc., Oxford, United Kingdom. doi: 10.1002/9781118852408.ch1

Fischman, B. J., T. L. Pitts-Singer, and G. E. Robinson. 2017. Nutritional regulation of phenotypic plasticity in a solitary bee (Hymenoptera: Megachilidae). Environ. Entomol.46: 1070–1079. doi: 10.1093/ee/nvx119

Frazier, J., J. Pflugfleder, P. Aupinel, A. Decourtye, J. Ellis, C. Scott-Dupree, Z. Huang, H. Thompson, P. Bachman, A. Dinter, M. Vaughn, B. E. Vaissière, G. Maynard, M. Kasina, E. Johansen, C. Brittain, M. Coulson, and R. C. F. Nocelli. 2014. Chapter 8: assessing effects through laboratory toxicity testing, pp. 75-94 In D. Fischer and T. Moriarty (eds.), Pesticide risk assessment for pollinators. John Wiley & Sons, Inc., Oxford, United Kingdom. doi: 10.1002/9781118852408.ch8

Frewin, A. J., A. E. Gradish, G. R. Ansell, and C. D. Scott-Dupree. 2019. Potential surrogate plants for use in semi-field pesticide risk assessment with Megachile rotundata. PeerJ. 6: 1–17. doi: 10.7717/PeerJ..6278

George, D. A., and C. M. Rincker. 1982. Residues of commercially used insecticides in the environment of Megachile rotundata. J. Econ. Entomol. 75: 319–323. doi: 10.1093/jee/75.2.319

Gibbs, J., and C. S. Sheffield. 2009. Rapid range expansion of the wool-carder bee, Anthidium manicatum (Linnaeus) (Hymenoptera: Megachilidae), in North America. J. Kansas. Entomol. Soc. 82: 21–29. doi: 10.2317/jkes805.27.1

Goulson, D., E. Nicholls, C. Botias, and E. L. Rotheray. 2015. Bee declines driven by combined stress from parasites, pesticides, and lack of flowers. Science. 347: 1255957–1255957. 10.1126/science.1255957.

Gradish, A. E., C. D. Scott-Dupree, and G. C. Cutler. 2012. Susceptibility of Megachile rotundata to insecticides used in wild blueberry production in Atlantic Canada. J. Pest. Sci. 85: 133–140. doi: 10.1007/s10340-011-0391-0

Gradish, A. E., G. C. Cutler, A. J. Frewin, and C. D. Scott-Dupree. 2016. Comparison of buckwheat, red clover, and purple tansy as potential surrogate plants for use in semi-field pesticide risk assessments with Bombus impatiens. PeerJ. 4: e2228. doi: 10.7717/PeerJ. 2228

Gregory, D. A., D. L. Johnson, B. H. Thompson, and K. W. Richards. 1992. Laboratory evaluation of the effects of carbaryl and chlorpyrifos bran baits and sprays used in grasshopper control, on alfalfa leafcutting bees (Megachile rotundata [F.]). J. Agr. Entomol. 9: 109–115.

109

Guédot, C., J. S. Buckner, M. M. Hagen, J. Bosch, W. P. Kemp, and T. L. Pitts-Singer. 2013. Nest marking behavior and chemical composition of olfactory cues involved in nest recognition in Megachile rotundata. Environ. Entomol.42: 779–789. doi: 10.1603/en13015

Heard, M. S., J. Baas, J.-L. Dorne, E. Lahive, A. G. Robinson, A. Rortais, D. J. Spurgeon, C. Svendsen, and H. Hesketh. 2017. Comparative toxicity of pesticides and environmental contaminants in bees: are honey bees a useful proxy for wild bee species? Sci. Total. Environ.578: 357–365. doi: 10.1016/j.scitotenv.2016.10.180

Heinze, G., and M. Ploner. 2018. Logistf: Firth's bias-reduced logistic regression. R package version 1.23. https://CRAN.R-project.org/package=logistf

Helson, B. V., K. N. Barber, and P. D. Kingsbury. 1994. Laboratory toxicology of six forestry insecticides to four species of bee (Hymenoptera: Apoidea). Arch. Environ. Con. Tox. 27: 107– 114. doi: 10.1007/bf00203895

Henderson, C. F., and E. W. Tilton. 1955. Tests with acaricides against the brown wheat mite. J. Econ. Entomol. 48: 157–161. doi: 10.1093/jee/48.2.157

Horne, M. 1995a. Pollen preference and its relationship to nesting success of Megachile rotundata (Hymenoptera: Megachilidae). Ann. Entomol. Soc. Am.88: 862–867. doi: 10.1093/aesa/88.6.862

Horne, M. 1995b. Leaf area and toughness: effects of nesting material preferences of Megachile rotundata (Hymenoptera: Megachilidae). Ann. Entomol. Soc. Am.88: 868–875. doi: 10.1093/aesa/88.6.868

Huntzinger, C. I., R. R. James, and W. P. Kemp. 2008a. Laboratory bioassays to evaluate fungicides for chalkbrook control in larvae of the alfalfa leafcutting bee (Hymenoptera: Megachilidae). J. Econ. Entomol. 101: 660–667. doi: 10.1603/0022-0493(2008)101[660:lbteff]2.0.co;2

Huntzinger, C. I., R. R. James, J. Bosch, and W. P. Kemp. 2008b. Fungicide tests on adult alfalfa leafcutting bees (Hymenoptera: Megachilidae). J. Econ. Entomol. 101: 1088–1094. doi: 10.1603/0022-0493(2008)101[1088:ftoaal]2.0.co;2

Johansen, E., M. Fry, and T. Moriarty. 2014. Chapter 13: risk mitigation and performance criteria, pp. 165–172. In D. Fischer and T. Moriarty (eds.), Pesticide risk assessment for pollinators. John Wiley & Sons, Inc., Oxford, United Kingdom. doi: 10.1002/9781118852408.ch13

Kemp, W. P., and J. Bosch. 2000. Development and emergence of the alfalfa pollinator Megachile rotundata (Hymenoptera: Megachilidae). Ann. Entomol. Soc. Am. 93: 904–911.

Kilian, R. 2016. Plant materials technical note No. MT-113. In Natural Resources Conservation Service Service [ed.]. United States Department of Agriculture, Bridger, MT.

Kim, J., N. Williams, and C. Kremen. 2006. Effects of cultivation and proximity to natural habitat on ground-nesting native bees in California sunflower fields. J. Kansas. Entomol. Soc. 79: 309–320. doi: 10.2317/0507.11.1

Klostermeyer, E. C., and H. S. Gerber. 1969. Nesting behavior of Megachile rotundata (Hymenopera: Megachilidae) monitored with an event recorder. Ann. Entomol. Soc. Am.62: 1321–1325. doi: 10.1093/aesa/62.6.1321

110

Klostermeyer, E. C., J. M. Stephen, and W. B. Rasmussen. 1973. Sex and weight of Megachile rotundata (Hymenoptera: Megachilidae) progeny associated with provision weights. J. Kansas. Entomol. Soc. 46: 536–548. www.jstor.org/stable/25082604

Knäbe, S., L., Franke, J. Fricke, O. Klein, A. Alscher, C. Classen, N. Exeler, M. Frommberger, J. S. Guerola, S. Kimmel, J. Lückmann, C. Molitor, B. Peters, S. Hecht-Rost, C. Shneider. 2017. Summary of an ICPPR non-Apis workshop–subgroup higher TIER (bumble beees and solitary bees) with recommendations for an semifield experimental design. doi: 10.13140/RG.2.2.30425.72805

Koh, I., E. V. Lonsdorf, N. M. Williams, C. Brittain, R. Isaacs, J. Gibbs, and T. H. Ricketts. 2015. Modeling the status, trends, and impacts of wild bee abundance in the United States. PNAS. 113: 140–145. doi: 10.1073/pnas.1517685113

Ladurner, E., J. Bosch, S. Maini, and W. P. Kemp. 2003. A method to feed individual bees (Hymenoptera: Apiformes) known amounts of pesticides. Apidologie. 34: 597–602. doi: 10.1051/apido:2003055

Ladurner, E., J. Bosch, W. P. Kemp, and S. Maini. 2005a. Evaluation of a standard artificial flower design to feed individual bees known amounts of pesticides. Apidologie. 36: 379–387. doi: 10.1051/apido:2005025

Ladurner, E., J. Bosch, W. P. Kemp, and S. Maini. 2005b. Assessing delayed and acute toxicity of five formulated fungicides to Osmia lignaria Say and Apis mellifera. Apidologie. 36: 449–460. doi: 10.1051/apido:2005032

Leach, M. E., and F. Drummond. 2018. A review of native wild bee nutritional health. International J. Ecol. 2018: 1–10. doi: 10.1155/2018/9607246

Loose, J. L., F. A. Drummond, C. Stubbs, S. Woods, and S. Hoffmann. 2005. Conservation and management of native bees in cranberry. Maine Agricultural and Forest Experiment Station Technical Bulletin 191. http://feedthebees.org/wp-content/uploads/0201/02/Conservation-and- Management-of-Native-Bees-in-Cranberry.pdf

Mayer, D. F. 1994. Effects of male to female ratio and number of females per nesting tunnel on sex ratio and number of progeny of the alfalfa leafcutter bee Megachile rotundata (Hymenoptera: Megachilidae). J. Entomol. Soc. B.C. 91: 43–46.

Mayer, D. F., E. R. Miliczky, and J. D. Lunden. 1990. Esfenvalerate insecticide and domesticated pollinators. Bee. Sci. 1: 33–39.

Mayer, D. F., J. D. Lunden, and G. Kovacs. 1997. Susceptibility of four bee species (Hymenoptera: Apoidea) to field weathered insecticide residues. J. Entomol. Soc. B.C. 94: 27–30.

McCorquodale, D. B., and R. E. Owen. 1994. Laying sequence, diploid males, and nest usurpation in the leafcutter bee, Megachile rotundata (Hymenoptera: Megachilidae). J. Insect. Behav. 7: 731– 738. doi: 10.1007/bf01997441

Mesenage, R., and M. N. Antoniou. 2018. Ignoring adjuvent toxicity falsifies the safety profile of commercial pesticides. Front. Public. Health. 5: 1–8. doi: 10.3389/fpubh.2017.00361

111

Michener, C. D. 2007. The bees of the world, vol. 2. John Hopkins University Press, Baltimore, MD.

Mommaerts, V., S. Reynders, J. Boulet, L. Besard, G. Sterk, and G. Smagghe. 2009. Risk assessment for side-effects of neonicotinoids against bumblebees with and without impairing foraging behavior. Ecotoxicology. 19: 207–215. doi: 10.1007/s10646-009-0406-2

O'Neill, K. M., A. M. Pearce, R. P. O'Neill, and R. S. Miller. 2010. Offspring size and sex ratio variation in a feral population of alfalfa leafcutting bees (Hymenoptera: Megachilidae). Ann. Entomol. Soc. Am.103: 775–784. doi: 10.1603/an09183

O’Neill, K. M., C. M. Delphia, and T. L. Pitts-Singer. 2015. Seasonal trends in the condition of nesting females of a solitary bee: wing wear, lipid content, and oocyte size. PeerJ. 3: e930. doi: 10.7717/PeerJ..930

(OECD) Organisation for Economic Co-operation and Development. 1998a. OECD guidelines for the testing of chemicals, Test No. 213: Honeybees, acute oral toxicity test. In OECD guidelines for the testing of chemicals section 2: effects on biotic systems. OECD Publishing, Paris, France. doi: 10.1787/9789264070165-en

(OECD) Organisation for Economic Co-operation and Development. 1998b. OECD guidelines for the testing of chemicals, Test No. 214: Honeybees, acute contact toxicity test. In OECD guidelines for the testing of chemicals section 2: effects on biotic systems. OECD Publishing, Paris, France. doi: 10.1787/9789264070189-en

(OECD) Organisation for Economic Co-operation and Development. 2007. Guidance document on the honey bee (Apis mellifera L.) brood test under semi-field conditions, Test No. 75. In OECD series on testing and assessment. OECD Publishing, Paris, France. doi: 10.1787/9789264085510- en

(OECD) Organisation for Economic Co-operation and Development. 2017a. OECD guidelines for the testing of chemicals, Test No. 245: honey bee (Apis mellifera L.), chronic oral toxicity test (10- day feeding). In OECD guidelines for the testing of chemicals section 2: effects on biotic systems. OECD Publishing, Paris, France. doi: 10.1787/9789264284081-en

(OECD) Organisation for Economic Co-operation and Development. 2017b. OECD guideline for the testing of chemicals: Test No. 246: Bumblebee, acute contact toxicity test. In OECD Guidelines for the Testing of Chemicals, Section 2. OECD Publishing, Paris, France doi: 10.1787/9789264284104-en

(OECD) Organisation for Economic Co-operation and Development. 2017c. OECD guidelines for the testing of chemicals, Test No. 247: bumblebee, acute oral toxicity test. In OECD series on testing and assessment. OECD Publishing, Paris, France. doi: 10.1787/9789264284128-en

Pankiw, P., J. A. C. Lieverse, and B. Siemens. 2012. The relationship between latitude and the emergence of alfalfa leafcutter bees, Megachile rotundata (Hymenoptera: Megachilidae). Can. Ent. 112: 555–558. doi: 10.4039/Ent112555-6

Park, M. G., E. J. Blitzer, J. Gibbs, J. E. Losey, and B. N. Danforth. 2015. Negative effects of pesticides on wild bee communities can be buffered by landscape context. P. R. Soc. B. 282: 20150299. doi: 10.1098/rspb.2015.0299

112

Peach, M. L., D. G. Alston, and V. J. Tepedino. 1995. Sublethal effects of carbaryl bran bait on nesting performance, parental investment, and offspring size and sex ratio of the alfalfa leafcutting bee (Hymenoptera: Megachilidae). Environ. Entomol. 24: 34–39. doi: 10.1093/ee/24.1.34

Peterson, J. H., and B. D. Roitberg. 2006. Impact of resource levels on sex ratio and resource allocation in the solitary bee, Megachile rotundata. Environ. Entomol. 35: 1404–1410. doi: 10.1093/ee/35.5.1404

Peterson, J. H., B. D. Roitberg, and J. H. Peterson. 2006. Impacts of flight distance on sex ratio and resource allocation to offspring in the leafcutter bee, Megachile rotundata. Behav. Ecol. Sociobiol. 59: 589–596. doi: 10.1007/s00265-005-0085-9

Pettis, J. 2014. Chapter 2: overview of the honey bee, pp. 3–4 In D. Fischer and T. Moriarty (eds.), Pesticide risk assessment for pollinators. John Wiley & Sons, Inc., Oxford, United Kingdom. doi: 10.1002/9781118852408.ch2

Pettis, J., I. Tornier, M. Clook, K. Wallner, B. Vaissiere, T. Stadler, W. Hou, H. Maynard, R. Becker, M. Coulson, P. Jourdan, M. Vaughn, R. C. F. Nocelli, C. Scott-Dupree, E. Johansen, C. Brittain, A. Dinter, and M. Kasina. 2014. Chapter 9: assessing effects through semi-field and field toxicity testing, pp. 295–119 In D. Fischer and T. Moriarty (eds.), Pesticide risk assessment for pollinators. John Wiley & Sons, Inc., Oxford, United Kingdom. doi: 10.1002/9781118852408.ch9

Pinheiro, B., S. DebRoy, D. Sarkar, and R Core Team. 2017. nlme: linear and nonlinear mixed effects models. R package version 3.1-131. URL: https://CRAN.R-project.org/package=nlme

Pisa, L. W., V. Amaral-Rogers, L. P. Belzunces, J. M. Bonmatin, C. A. Downs, D. Goulson, D. P. Kreutzweiser, C. Krupke, M. Liess, M. McField, C. A. Morrissey, D. A. Noome, J. Settele, N. Simon-Delso, J. D. Stark, J. P. Van der Sluijs, H. Van Dyck, and M. Wiemers. 2014. Effects of neonicotinoids and fipronil on non-target invertebrates. Environ. Sci. Pollut. R. 22: 68– 102. doi: 10.1007/s11356-014-3471-x

Pitts-Singer, T. L. 2015. Resource effects on solitary bee reproduction in a managed crop pollination system. Environ. Entomol. 44: 1125–1138. doi: 10.1093/ee/nvv088

Pitts-Singer, T. L., and R. R. James. 2005. Emergence success and sex ratio of commerical alfalfa leafcutting bees from the United States and Canada. J. Econ. Entomol. 98: 1785–1790. doi: 10.1093/jee/98.6.1785

Pitts-Singer, T. L., and J. Bosch. 2010. Nest establishment, pollination efficiency, and reproductive success of Megachile rotundata (Hymenoptera: Megachilidae) in relation to resource availability in field enclosures. Environ. Entomol. 39: 149–158. doi:10.1603/en09077

Pitts-Singer, T. L., and J. H. Cane. 2011. The alfalfa leafcutting bee, Megachile rotundata: the world's most intensively managed solitary bee. Annu. Rev. Entomol. 56: 221–237. doi: 10.1146/annurev- ento-120709-144836 Pitts-Singer, T. L., and J. D. Barbour. 2016. Effects of residual novaluron on reproduction in alfalfa leafcutting bees, Megachile rotundata F. (Megachilidae). Pest. Manag. Sci. 73: 153–159. doi: 10.1002/ps.4356

113

Potts, S., K. Biesmeijer, R. Bommarco, T. Breeze, L. Carvalheiro, M. Franzén, J. González-Varo, A. Holzschuh, A. Kleijn, B. Kunin, T. Lecoq, O. Lundin, D. Michez, P. Neumann, A. Nieto, L. Penev, P. Rasmont, O. Ratamäki, V. Riedinger, S. Robert, M. Rundlöf, J. Scheper, P. Sørensen, I. Steffan-Dewenter, P. Stoev, M. Vilà, and O. Schweiger. 2015. Status and trends of European pollinators: Key findings from the STEP project. Pensoft Publishers, Sofia, Bulgaria.

R Core Team. 2017. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.

Richards, K. W. 1984. Alfalfa leafcutter bee management in Western Canada, 1st ed. Agriculture Canada Publication. 1485/E. doi: 10.1080/0005772x.1987.11098930

Richards, K. W. 1993. Non-Apis bees as crop pollinators. Revue Suisse de Zoologie 100: 807–822. doi: 10.5962/bhl.part.79885

Richards, K. W. 2012. Effect of environment and equipment on productivity of alfalfa leafcutter bees (Hymenoptera: Megachilidae) in southern Alberta, Canada. Can. Entomol. 128: 47–56. doi: 10.4039/Ent12847-1

Richards, K. W., G. H. Whitfield, and G. B. Schaalje. 1987. Effects of temperature and duration of winter storage on survival and period of emergence for the alfalfa leafcutter bee (Hymenoptera: Megachilidae). J. Kansas. Entomol. Soc. 60: 70–76.

Rinehart, J. P., G. D. Yocum, W. P. Kemp, and J. H. Bowsher. 2016. Optimizing fluctuating thermal regime storage of developing Megachile rotundata (Hymenoptera: Megachilidae). J. Econ. Entomol. 109: 993–1000. doi: 10.1093/jee/tow026

Rossi, B. H., P. Nonacs, and T. L. Pitts-Singer. 2010. Sexual harassment by males reduces female fecundity in the alfalfa leafcutting bee, Megachile rotundata. Anim. Behav. 79: 165–171. doi: 10.1016/j.anbehav.2009.10.023

Rothschild, M. 1979. Factors influencing size and sex ratios in Megachile rotundata (Hymenoptera: Megachilidae). J. Kansas. Entomol. Soc. 52: 392–401.

Rundlöf, M., G. K. S. Andersson, R. Bommarco, I. Fries, V. Hederstrom, L. Herbertsson, O. Jonsson, B. K. Klatt, T. R. Pederson, J. Yourstone, and H. G. Smith. 2016. Seed coating with a neonicotinoid insecticide negatively affects wild bees. Nature. 521: 77–80. doi: 10.1038/nature14420

Sanchez-Bayo, F., and K. Goka. 2014. Pesticide residues and bees – a risk assessment. PLoS ONE. 9: 1–16. doi: 10.1371/journal.pone.0094482

Scott-Dupree, C. D., L. Conroy, and C. R. Harris. 2009. Impact of currently used or potentially useful insecticides for canola agroecosystems on Bombus impatiens (Hymenoptera: Apidae), Megachile rotundata (Hymenoptera: Megachilidae), and Osmia lignaria (Hymenoptera: Megachilidae). J. Econ. Entomol. 102: 177–182. doi: 10.1603/029.102.0125

Sgolastra, F., S. Hinarejos, T. L. Pitts-Singer, N. K. Boyle, T. Joseph, J. Lūckmann, N. E. Raine, R. Singh, N. M. Williams, and J. Bosch. 2018. Pesticide exposure assessment paradigm for solitary bees. Environ. Entomol.. 20: 1–14 doi: 10.1093/ee/nvy105

114

Sharf, M. E. 2008. Neurological effects of insecticides and the insect nervous system. In J. L. Capinera [ed.], Encyclopedia of entomology. Springer, Dordrecht, Netherlands.

Sheffield, C. S., C. Ratti, L. Packer, and T. L. Griswold. 2011. Leafcutter and mason bees of the genus Megachile Latrielle (Hymenoptera: Megachilidae) in Canada and Alaska. Can. J. Arthropod. Identif. 18: 1–107. doi: 10.3752/cjai.2011.18

Smagghe, G., and N. W. Calderone. 2012. Insect pollinated crops, insect pollinators and US agriculture: trend analysis of aggregate data for the period 1992–2009. PLoS ONE. 7: e37235. doi: 10.1371/journal.pone.0037235

Tasei, J. N., and S. Carre. 1985. Effects du traitement de luzerne en fleurs (Medicago sativa L.) avec de la deltamethrine et de la phosalone sur l'abeille solitaire: Megachile rotundata F. (Hym., Megachilidae). Acta. Oecol-Oec. Appl. 6: 165–173.

Tepedino, V. J., and F. D. Parker. 1986. Effect of rearing temperature on mortality, second-generation emergence, and size of adult in Megachile rotundata (Hymenoptera: Megachilidae). J. Econ. Entomol. 79: 974–977. doi: 10.1093/jee/79.4.974

Tepedino, V. J., and F. D. Parker. 1988. Alternation of sex ratio in a partially bivoltine bee, Megachile rotundata (Hymenoptera: Megachilidae). Ann. Entomol. Soc. Am. 81: 467–476. doi: 10.1093/aesa/81.3.467

Tezuka, T., and Y. Maeta. 1993. Effect of UVA film on extrandial activities of three species of bees. Japanese J. Appl. Entomol. Z. 37: 175–180. doi: 10.1303/jjaez.37.175

Thompson, H. M., and L. V. Hunt. 1999. Extrapolation from honeybees to bumblebees in pesticide risk assessment. Ecotoxicology. 8: 147–466. doi: 10.1023/A:10264440

Torchio, P. F. 1973. Relative toxicity of insecticides to the honey bee, alkali bee and alfalfa leafcutting bee (Hymenoptera: Apidae, Halictidae, Megachilidae). J. Kansas. Entomol. Soc. 46: 446–453.

Torchio, P. F., and N. N. Youssef. 1973. A method of feeding bees measured amounts of insecticides in solution. Can. Entomol. 105: 1011–1014. doi: 10.4039/Ent1051011-7

Trostle, G., and P. F. Torchio. 1994. Comprative nesting behaviour and immature development of Megachile rotundata (Fabricius) and Megachile apicalis Spinola (Hymenoptera: Megachilidae). J. Kansas. Entomol. Soc. 67: 53–72.

Uhl, P., L. A. Franke, C. Rehberg, C. Wollmann, P. Stahlschmidt, L. Jeker, and C. A. Brühl. 2016. Interspecific sensitivity of bees towards dimethoate and implications for environmental risk assessment. Sci. Rep - UK. 6. doi: 10.1038/srep34439

Van Scoy, A., A. Pennell, X. Zhang. 2016. Environmental fate and toxicology of dimethoate. Rev. Environ. Contam. Toxicol. 237: 53-70. doi: 10.1007/978-3-319-23573-8_3.

Vaughn, M., B. E. Vassière, G. Maynard, M. Kasina, R. C. F. Nocelli, C. Scott-Dupree, E. Johansen, C. Brittain, M. Coulson, and A. Dinter. 2014. Chapter 3: overview of non-Apis bees, pp. 15-18 In D. Fischer and T. Moriarty (eds.), Pesticide risk assessment for pollinators. John Wiley & Sons, Inc., Oxford, United Kingdom. doi: 10.1002/9781118852408.ch3

115

Whiten, S. R., and R. K. D. Peterson. 2016. The influence of ambient temperature on the susceptibility of Aedes aegypti (Diptera: Culicidae) to the pyrethroid insecticide permethrin. J. Med. Entomol. 53: 139–143. doi: 10.1093/jme/tjv159

Williams, N. M., and C. Kremen. 2006. Resource distributions among habitats determine solitary bee offspring production in a mosaic landscape. Ecol. Appl. 17: 910–921. doi: 10.1890/06-0269

Wilson, J. S., and O. J. Messinger Carril. 2016. The bees in your backyard: a guide to North America's bees. Princeton University Press, Princeton, NJ.

Wisk, J. D., J. Pistorius, M. Beevers, R. Bireley, Z. Browning, M. P. Chauzat, A. Nikolakis, J. Overmyer, R. Rose, R. Sebastien, B. E. Vaissière, G. Maynard, M. Kasina, R. C. F. Nocelli, C. Scott-Dupree, E. Johansen, C. Brittain, M. Coulson, A. Dinter, and M. Vaughn. 2014. Chapter 7: assessing exposure of pesticides to bees, pp. 45–74. In D. Fischer and T. Moriarty (eds.), Pesticide risk assessment for pollinators. John Wiley & Sons, Inc., Oxford, United Kingdom. doi: 10.1002/9781118852408.ch7

Woodcock, B. A., J. M. Bullock, R. F. Shore, M. S. Heard, M. G. Pereira, J. Redhead, L. Ridding, H. Dean, D. Sleep, P. Henrys, J. Peyton, S. Hulmes, L. Hulmes, M. Sárospataki, C. Saure, M. Edwards, E. Genersch, S. Knäbe, and R. F. Pywell. 2017. Country-specific effects of neonicotinoid pesticides on honey bees and wild bees. Science. 356: 1393–1395. doi: 10.1126/science.aaa1190

Zurbuchen, A., L. Landert, J. Klaiber, A. Müller, S. Hein, and S. Dorn. 2010. Maximum foraging ranges in solitary bees: only few individuals have the capability to cover long foraging distances. Biol. Conserv. 143: 669–676. doi: 10.1016/j.biocon.2009.12.003