An Abstract of the Thesis Of

AN ABSTRACT OF THE THESIS OF

Katherine Anne Arstingstall for the degree of Master of Science in Wildlife Science presented on June 5, 2020.

Title: Using DNA Metabarcoding to Study Interactions Between Native Bees and Plants: Technical Development and Applications.

Abstract approved:

______Sandra J. DeBano

Abstract

Pollination is an essential ecosystem service that sustains functioning ecosystems and aids in food production. In response to recent, widespread declines of managed and native bee populations, many land managers have shown interest in developing conservation and restoration plans for enhancing native bee habitat. However, there is a lack of data on which plant species serve as important food sources for native bees. Traditional methods for describing plant-bee interactions (e.g., bee foraging observations, microscopy) are time consuming, require specialized expertise, and often result in low taxonomic resolution. DNA metabarcoding of bee-collected pollen has been introduced as a more effective tool for describing the relationship between bees and flowering plants. However, there are still some concerns with this relatively new technique that need to be examined and resolved before pollinator researchers can be confident in the results that it produces. My thesis examines the strengths and some limitations of using DNA metabarcoding of bee pollen to describe plant-bee interactions. Chapter 1 provides an introduction to the topic of native bees and their ecological and societal importance. I identify current knowledge gaps, and I introduce the topic of DNA metabarcoding, its many uses, and its potential limitations. This chapter provides background knowledge and introduces the questions that are addressed in the following chapters.

Chapter 2 addresses some unresolved questions that exist in the field of pollen metabarcoding, including which type of sequence count removal threshold is most appropriate for studying plant-pollinator interactions, the quantitative abilities of pollen metabarcoding, the ability of pollen metabarcoding to detect rare flower visits by bees, and the potential role of environmental contamination in mischaracterizing bee foraging behavior. We collected pollen from five plant species, created pollen mixtures in the laboratory that varied in species richness and evenness, and used metabarcoding of the ITS2 region to identify the plant species in the mixtures. We analyzed the sequencing data using two different sequence count removal threshold protocols: one liberal and one conservative. We were able to correctly identify all plant species in the mixtures, confirming the qualitative abilities of ITS2 pollen metabarcoding. When using the liberal threshold, six additional plant species that were not used to create the pollen mixtures were detected in the single-species samples. When using the conservative threshold, no additional species were detected in the single-species mixtures, but some species used to create the pollen mixtures were not detected above the sequence count removal threshold in mixtures 2-5, resulting in false negatives. We compared the proportion of pollen by mass to the proportion of sequencing reads produced for each plant species in the mixtures. Regardless of the threshold used, the proportion of pollen and sequencing reads were not significantly related for two of the four mixtures, and certain species were consistently over and underrepresented. We examined each plant species separately, and the proportion of pollen and sequencing reads was significantly and positively related, but proportions of the over and underrepresented species varied strongly from a one-to- one relationship. In Chapter 3, we examine the strengths and limitations of using pollen metabarcoding to study plant-native bee interactions in three different habitat types. We sampled 403 native bees from three different habitat types in eastern Oregon. We documented foraging observations for each bee, and we used DNA metabarcoding of the ITS2 region and rbcL gene to identify the plant species present in each bee’s pollen load. We compared plant-pollinator networks created from bee foraging observations with networks created from plant species assignments obtained using DNA metabarcoding to

determine whether these data are consistent or if DNA metabarcoding provides additional information on bee foraging behavior. We also compared plant species assignments produced by DNA metabarcoding when using a larger, “regional” reference database to those produced using a site specific, “local” reference database. Plant-pollinator networks produced using data derived from DNA metabarcoding had significantly higher connectance, linkage density, and bee generality and significantly lower specialization when compared to networks based on bee foraging observations. Approximately 15% more plant species were assigned when using the regional database than when using the local database. Using a local reference database reduced the possibility of erroneous taxonomic assignments. Chapter 4 provides a summary of the key findings from Chapters 2 and 3. I make some suggestions for researchers that wish to use DNA metabarcoding to study interactions between bees and plants and identify some areas for future research. Ultimately, the results of this thesis show that DNA metabarcoding of bee pollen is a promising technique which provides additional information on bee foraging behavior that cannot easily be discovered using bee foraging observations or microscopy. Although ITS2 sequence reads cannot be used to determine the amount of a plant species in a pollen sample, ITS2 metabarcoding provides accurate species identification and does not appear to overestimate the number of plant species on which bees forage when using our more conservative sequence count removal threshold. The results obtained from studies such as these can be used to create bee species- and region-specific plant lists that can inform restoration and conservation plans to increase native bee habitat quantity and quality in any location.

Using DNA Metabarcoding to Study Interactions Between Native Bees and Plants: Technical Development and Applications

by Katherine Anne Arstingstall

A THESIS

submitted to

Oregon State University

in partial fulfillment of the requirements for the degree of

Master of Science

Presented June 5, 2020 Commencement June 2020

Master of Science thesis of Katherine Anne Arstingstall presented on June 5, 2020

APPROVED:

Major Professor, representing Wildlife Science

Head of the Department of Fisheries and Wildlife

Dean of the Graduate School

I understand that my thesis will become part of the permanent collection of Oregon State University libraries. My signature below authorizes release of my thesis to any reader upon request.

Katherine Anne Arstingstall, Author

ACKNOWLEDGEMENTS

I would like to thank many individuals and organizations involved with this project. Without their support, the completion of this thesis would not have been possible.

First and foremost, I would like to thank my major advisor and mentor, Dr. Sandy DeBano for her constant encouragement, guidance, support and positivity. I am so grateful to have had the opportunity to work with someone so knowledgeable, patient, and hard-working, and I will always cherish my time as her graduate student. I would also like to thank my committee members, Dr. Ken Frost and Dr. Dave Wooster for their advice, their suggestions during field and lab work, and their help during the writing process. A special thanks to Ken for funding my graduate research assistantship and for always allowing me to use laboratory space and supplies.

Next, I would like to thank the organizations and grants that helped fund this project: the USDA Forest Service, Pacific Northwest Research Station, the Northwest Potato Research Consortium, the Oregon Department of Forestry, Oregon State University’s Agricultural Research Fund, the Oregon Zoo Foundation’s Future for Wildlife Pacific Northwest Fund, the Oren Pollak Memorial Student Research Grant for Grassland Science, the Mazamas Research Grants Program, and the Western SARE Graduate Student Grant. I’d also like to thank the United States Forest Service, The Nature Conservancy and Threemile Canyon Farms for access to study sites and allowing us to sample bees.

Several researchers at The Nature Conservancy and the United States Forest Service Pacific Northwest Research Station helped with planning and field work for this project. I would like to thank Mary M. Rowland for her help with study design, edits and suggestions during the writing process, and her encouragement throughout the years. A special thanks to Heidi Schmalz, Josh Averett, and Cali Thomas for their help identifying plant species. I’d also like to thank Skyler Burrows for identifying the hundreds of bees that were collected for this study.

I would like to thank Xiaoping Li for constructing and implementing the bioinformatics pipeline used in this project, his responsiveness and his willingness to

help. Thank you to Victoria Skillman for her helpfulness and assistance in the laboratory. I would like to thank our field and laboratory technicians: James McKnight, Coltyn Kidd, and Marisa McCaskey. This project would not have been possible without all their work in the field and in the laboratory. It was a pleasure to work with such inquisitive, kind, and hard-working individuals. I would especially like to thank my lab-mate, Scott Mitchell, for all of his help in the field and in the laboratory as well as his friendship and support throughout the years.

Finally, I would like to thank my family and friends for always believing in me and supporting my decision to move across the country to achieve my goals. I want to especially thank my parents for their constant encouragement, love and support. I could not have done this without them.

CONTRIBUTION OF AUTHORS

Sandra J. DeBano contributed to all parts of this thesis including study design, field wok, data analysis, and writing. Ken Frost and David Wooster contributed laboratory space, equipment and materials, and provided edits and suggestions for all chapters. Xiaoping Li constructed and implemented the bioinformatics pipeline used in this project and wrote the bioinformatics methods in Chapters 2 and 3. Mary M. Rowland contributed to study design and provided edits and suggestions for Chapter 3. Skyler Burrows identified all bee specimens and provided edits for Chapter 3.

TABLE OF CONTENTS

Page Chapter 1: Introduction……………………………………………………………………1 Literature Cited……………………………………………………………………6 Chapter 2: Testing Quantitative Capabilities of DNA Metabarcoding for Use in Bee Foraging Studies Using Pollen Mixtures of Known Concentrations…………………….11

Abstract…...……………………………………………………………………...11 Introduction………………………………………………………………………12 Methods…………………………………………………………………………..17 Flower Collection………………………………………………………...17 Isolating Pollen from Flowers……………………………………………17 Preparation of Pollen Mixtures..…………………………………………18 DNA Extraction and PCR………………………………………………..18 Bioinformatics- Read Quality Filtering and Denoising………………….19 Bioinformatics- Reference Database Construction………………………20 Bioinformatics- Training Classifier and Taxonomy Classification……...20 Sequence Count Removal Thresholds………………………………..….20 Data Analysis…………………………………………………………….21 Results……………………………………………………………………………21 Detection Using a Liberal Threshold………………………….…………21 Detection Using a Conservative Threshold………………...……………22 Quantification: Comparing Proportion of Pollen by Mass to Proportion of Sequencing Reads……………………………………………………..22

Discussion………………………………………………………………………..22 Literature Cited…………………………………………………………………..28 Figures and Tables……………………………………………………………….33 Chapter 3: Capabilities and Limitations of Using DNA Metabarcoding to Study Plant- Pollinator Interactions……………………………………………………………………43

Abstract…………………………………………………………………………..43 Introduction………………………………………………………………………44

TABLE OF CONTENTS (Continued) Page Methods…………………………………………………………………………..47 Study Sites……………………………………………………………….47 Bee Sampling…………………………………………………………….48 Pollen Isolation…………………………………………………………..48 DNA Extraction and PCR………………………………………………..49 Bioinformatics- Read Quality Filtering and Denoising………………….50 Bioinformatics- Reference Database Construction………………………50 Bioinformatics- Training Classifier……………………………………...51 Bioinformatics- Taxonomy Classification……………………………….51 Sequence Count Removal Threshold…………………………………….52 Network Analyses………………………………………………………..52 Results……………………………………………………………………………53 Comparing Plant-Pollinator Networks Created from Bee Foraging Observations and Plant Species Assignments Obtained Using DNA Metabarcoding…………………………………………………………...53

Comparing Plant Species Assignments Obtained Using MB-RDB and MB-LDB…………………………………………………………………54

Inconsistencies Among Bee Foraging Observations and DNA Metabarcoding Data……………………………………………………...55

Discussion………………………………………………………………………..56 Literature Cited…………………………………………………………………..62 Figures and Tables……………………………………………………………….68 Chapter 4: Conclusion……………………………………………………………………79 Literature Cited…………………………………………………………………..83 Bibliography……………………………………………………………………………..84 Appendix 1: Additional Figures and Tables for Chapter 2………………………………93 Appendix 2: Additional Tables for Chapter 3……………………………………………97

LIST OF FIGURES

Figure Page 2.1: Proportion of pollen and sequencing reads for each species in mixtures (conservative threshold) …………………………….....…………………………………………...…..33

2.2: Proportion of pollen by mass vs. proportion of sequencing reads (conservative threshold) ……………………………………………………………...………………...34

3.1: Workflow of QIIME2 pipeline……………………………………………………...68 3.2: Plant-pollinator networks for Threemile Canyon Farms……………………………69 3.3: Plant-pollinator networks for the USFS Starkey Experimental Forest and Range….70 3.4: Plant-pollinator networks for TNC’s Zumwalt Prairie Preserve……………………71 3.5: Bar plots of statistically significant network parameters……….…………………...72 3.6: Average number of plant species identified per pollen load………………………..73

LIST OF TABLES

Table Page 2.1: Proportion of pollen and sequencing reads for plant species in mixtures and single- species samples (liberal threshold) …...…………………………………………………35

2.2: Original and adjusted number of sequencing reads (conservative threshold) ….…..37 2.3: Proportion of pollen and sequencing reads for plant species in mixtures and single- species samples (conservative threshold) ………………….……………………………39

2.4: Summary of linear regression for plant species in mixtures 1-5……………………41 2.5: Summary of linear regressions for each plant species across all mixtures………….42 3.1: Number of plant taxonomic units assigned for each detected method...……………74 3.2: Top 10 plant species identified for each detection method at each location………..75 3.3: Percent of bee foraging observations that are consistent with DNA metabarcoding data…………………………………………………………………………………….…77

3.4: Mismatches of regional database metabarcoding…….……………………………..78

LIST OF APPENDIX FIGURES

Figure Page

A1.1: Proportion of pollen and sequencing reads for species in mixtures (liberal threshold) ………………………………………………………………………...……...94

A1.2: Proportion of pollen by mass vs. proportion of sequencing reads (liberal threshold)…………………………………………….…………………………………..95

LIST OF APPENDIX TABLES

Table Page A1.1: Additional plant species in single-species samples (liberal threshold)…………....96 A2.1: Complete list of sampled bee species……………………………………………..98 A2.2: Complete list of plant species identified using all three detection methods……..100 A2.3: Legends for Threemile Canyon Farms plant-pollinator networks……………….106 A2.4: Legends for the USFS Starkey Experimental Forest and Range plant-pollinator networks……………………………………………………………………...…………108

A2.5: Legends for TNC’s Zumwalt Prairie Preserve plant-pollinator networks……….113

Chapter 1: Introduction

Pollinators play an important role in both natural and agricultural ecosystems, sustaining flowering plant communities and aiding in food production (Buchmann & Nabhan 1996; Losey & Vaughan 2006; Ollerton,et al., 2011). It is estimated that about 90% of all flowering plant species are pollinated by insects, with bees serving as the primary pollinators (Richards, 1997; Buchmann & Nabhan, 1996; Ollerton et al., 2012). Most crop pollination is provided by the European honeybee (Apis mellifera), but wild bees also play a significant role, with their pollination services estimated at $150 billion globally (Gallai et al., 2009). Declines in wild bee populations have been reported worldwide due to many inter-connected factors including land-use change, increased pesticide application, and reduced resource diversity (Potts et al., 2010; Cameron et al., 2011; Martins et al., 2013; Scheper et al., 2014). One way to reverse native bee population declines is to increase the quantity and quality of native bee habitat by planting flowers that are important food sources for native bees (Durant & Otto, 2019). Identifying these food sources requires a greater understanding of the relationship between bees and flowering plants. Traditional methods for describing bee-flower interactions include direct observation of foraging behavior and indirect observation by collecting pollen loads from bees and using microscopy to identify plant species based on morphological characteristics of pollen grains (Erdtman, 1943). Direct observation is time consuming and only reveals a snapshot of a bee’s foraging behavior because the observer is limited by their ability to follow a foraging bee from flower to flower. Microscopy is also time consuming, requires specialized expertise, and often results in low taxonomic resolution (Rahl, 2008; Cornman et al., 2015). Furthermore, only a subsample of a bee’s pollen load is typically analyzed using microscopy, which can result in a lack of detection of pollen from plant species present in low abundance (Von Der Ohe et al., 2004; Whittington et al., 2004; Sajwani et al., 2014). DNA metabarcoding of pollen collected from foraging bees is a promising tool for describing plant-pollinator networks (Cornman et al., 2015; Keller et al., 2015;

Kraaijeveld et al., 2015; Richardson et al., 2015a; Bell et al., 2017; Smart et al., 2017). DNA metabarcoding combines species identification using DNA barcoding with next generation sequencing technology. In pollen DNA metabarcoding methods, DNA is extracted from pollen samples containing one or more plant species and barcode regions with low intraspecific and high interspecific variation are amplified and sequenced using next generation sequencing. The species composition of a pollen sample is then determined based on the presence of DNA sequencing “barcodes” in the pollen samples and their similarity to sequences of known reference plant species (Hebert et al., 2003; Sickel et al., 2015). DNA metabarcoding is used in many different systems to study a wide range of organisms. For example, metabarcoding methods can be used to monitor aquatic biodiversity, conduct diet assessments, and characterize bacterial and fungal community diversity (Kartzinel et al., 2015; Valentini et al., 2016; Djemiel et al., 2017). Most early bee pollen metabarcoding studies focused on honey bees, collecting composite pollen samples from many individuals in pollen traps attached to the entrance of hives. Pollen metabarcoding techniques could then be used on the pollen collected by an entire hive to determine the floral composition of honey or to describe the overall hive’s foraging habits by identifying the plant species in the pollen loads (Hawkins et al., 2015; Richardson et al., 2015a,b; De Vere et al., 2017). Several studies have compared the number of plant taxa detected using DNA metabarcoding to the number of plant taxa identified using microscopic methods, finding that pollen metabarcoding consistently detects an equal or greater number of plant taxa than microscopic pollen identification (Keller et al., 2015; Richardson et al., 2015a; Smart et al., 2017). Bell et al. (2017) introduced the idea that pollen metabarcoding could be used on individual pollen loads collected from bees and built a proof-of-concept plant-pollinator network using plant species assignments obtained from metabarcoding of 38 individual bee pollen loads using two commonly examined barcode regions (ITS2 and rbcL). Since its first use, researchers have questioned whether sequencing reads can be used as a proxy for the quantity of pollen from each plant species in a mixed pollen sample. To examine this question, researchers compared the proportion of sequencing

reads produced for each plant species using DNA metabarcoding to the proportion of pollen grains identified via microscopy (Keller et al., 2015; Kraaijeveled et al., 2015; Richardson et al., 2015a; Bell et al., 2019; Richardson et al., 2019). The results of these studies have been inconsistent: while some found a positive relationship between the proportion of sequencing reads and the proportion of pollen grains for a plant species, others found no reliable association, leaving the question of a quantitative relationship unresolved. Despite the continued use of direct observation to describe bee foraging behavior and floral preferences (Roof et al., 2018; Brunet et al., 2019; Estravis Barcala et al., 2019), few studies have compared plant-pollinator networks based on bee foraging observations to those based on plant species assignments obtained from DNA metabarcoding of bee pollen loads. Those that have either did not solely focus on bees (Pornon et al. 2016; Galliot et al., 2017) or lacked taxonomic resolution of the plants and pollinators involved (Potter et al., 2019). Examining the consistency, or lack of consistency, between plant-pollinator networks derived from bee foraging observations and DNA metabarcoding of bee pollen loads is important for determining whether observation alone can sufficiently identify major food sources for bees, or whether they tend to oversimplify plant-pollinator networks. The few studies that have compared direct observations of pollinator foraging behavior with pollen metabarcoding data show that plant-pollinator networks created using DNA metabarcoding data are more complex, containing more plant taxa and interactions than those based on direct observation (Pornon et al. 2016; Potter et al., 2019, Arstingstall et al., in revision). Although this may be an accurate depiction of foraging behavior, the tendency of metabarcoding networks to be more complex could be explained by other mechanisms. For example, pollen is moved around the environment by wind and other insects (Janzen, 1983; Whitehead, 1983; Latta et al., 1998), so a flower could have pollen from several other plant species residing on it before a bee visits it. Thus, a bee could inadvertently pick up pollen from multiple plant species in one flower visit. Yet no studies have isolated pollen collected from flowers of single species in natural settings and sequenced their pollen to determine if additional plant species are

4 detected. Such data would shed light on whether plant species assignments obtained using DNA metabarcoding of bee pollen tend to overestimate the complexity of plant- pollinator networks, and if so, allow for estimating the proportion of interactions that can be attributed to bee foraging versus those that may be attributed to movement of pollen by wind and other insects. Another unresolved question in the field of pollen metabarcoding is whether we can detect rare flower visits by examining a bee’s pollen load, or if those species might not be detected because they fall below sequence count removal thresholds. Some contamination is inevitable in pollen metabarcoding, but we can control for contamination in the laboratory by using negative controls at the DNA isolation and amplification stages, which contain sterile water in place of genetic material. We can then use the number of sequencing reads found in each negative control to create a “sequence count removal threshold,” and any taxonomic assignment whose read count falls below this threshold can be considered contamination or other “background noise” and removed from further analysis (Bell et al., 2017, 2019; Macgregor et al., 2019). There is currently no standardized protocol for creating and using a sequence count removal threshold in pollen metabarcoding studies, despite the fact that different threshold protocols can yield different interpretations of the same results. Richardson et al. (2019) discarded “taxonomic groups represented by <0.01% of the data”. In contrast, Bell et al. (2019) used the maximum number of sequencing reads in a negative control as a threshold, and set any taxonomic assignment with fewer reads than this threshold to zero. One type of threshold removes entire taxonomic groups, while the other removes specific bee-plant interactions, likely yielding different results and different interpretations of pollinator behavior. Identifying a standardized threshold protocol would eliminate this problem and allow for comparison of pollinator foraging behavior (e.g., plant-pollinator networks) across multiple pollen metabarcoding studies. However, there is concern that this may also eliminate rare interactions occurring at low frequencies (Bell et al., 2017). Bell et al. (2019) created pollen mixtures ranging in rarity (5-100%) to test the qualitative abilities of pollen metabarcoding, but no study has sequenced a mixture

containing plant species present in amounts similar to the amount of pollen collected by a bee in a single flower visit. Another challenge associated with pollen metabarcoding is the issue of correctly identifying plant taxa to the species-level (Keller et al., 2015; Kraaijeveld et al., 2015; Richardson et al., 2015a). While DNA metabarcoding consistently identifies a greater number of taxa compared to traditional methods (i.e., direct observation, microscopy), species assignments obtained through DNA metabarcoding can be prone to sequencing and database errors, resulting in erroneous taxonomic assignments (Cornman et al., 2015; Keller et al., 2015; Smart et al., 2017). The accuracy of plant species assignments is highly dependent on the reference database; only the plant species that are included in the reference database can be assigned to DNA sequences. Researchers often include all plant DNA sequences in a database for a specific barcode marker (Cornman et al., 2015; Keller et al., 2015; Sickel et al., 2015; Galliot et al., 2017) or from a geographic location consistent with their area of study, which is often as large as an entire state or country (Richardson et al., 2015b; De Vere et al., 2017; Potter et al., 2019). This ensures that all possible forage resources are included, but it can also increase the potential for false detections because species and genera that share a significant portion of their DNA can be difficult to differentiate (Gao et al., 2010). Currently no work has been conducted to examine the potential benefits of using reference databases that contain only the plant species that are known to occur in the location of interest (i.e., a “local” reference database) versus a reference database that includes all plant species known to occur in a larger region (i.e., a “regional” reference database). Chapter 2 addresses several unanswered questions in the field of pollen metabarcoding. Here, we sequence the DNA of five types of laboratory-prepared pollen mixtures of known concentrations and five single-species pollen samples and compare the results after using two different types of sequence count removal thresholds that are commonly used in pollen metabarcoding studies. We use the sequencing results for the pollen mixtures to determine whether the proportion of sequencing reads for a plant species can be used to determine the proportion of that plant species in a mixed pollen sample (e.g., a bee-collected pollen load). Next, we used sequencing data from three

6 replicates of a highly skewed laboratory-prepared pollen mixture with plant species present in quantities similar to the amount of pollen collected by a bee in a single flower visit to determine whether pollen DNA metabarcoding can consistently detect plant species in such small amounts above the two types of sequence count removal thresholds. Finally, we used the sequence data from the five single-species pollen samples to examine the potential role of environmental contamination in overestimating the number of species on which an individual bee forages. Chapter 3 focuses on practical applications of pollen metabarcoding techniques, identifying strengths and limitations of using this method to study individual native bee- plant interactions. We sampled 403 native bees from three locations in eastern Oregon, recorded foraging observations for each bee, and sequenced their pollen loads individually. We used this data to determine whether plant-pollinator networks based on bee foraging observations are consistent with networks based on plant species assignments obtained using pollen metabarcoding or if bee foraging observations tend to oversimplify bee foraging behavior. We also compared plant species assignments produced by DNA metabarcoding when using a larger, “regional” reference database to those produced using a site specific, “local” reference database to determine whether there are benefits to using a more site-specific, local reference database. In the last chapter, I provide a summary of the key findings of my thesis, make suggestions for other researchers that wish to use pollen DNA metabarcoding to study plant-bee interactions or identify important food sources for bees, and identify several areas of future research based on the subjects presented here.

Literature Cited Arstingstall, K. A., DeBano, S. J., Li, X., Wooster, D. E., Rowland, M. M., Burrows, S., Frost, K. (In Revision). Capabilities and limitations of using DNA metabarcoding to study plant-pollinator interactions. Molecular Ecology. Bell, K. L., Fowler, J., Burgess, K. S., Dobbs, E. K., Gruenewald, D., Lawley, B., … Brosi, B. J. (2017). Applying pollen DNA metabarcoding to the study of plant– pollinator interactions. Applications in Plant Sciences, 5(6), 1600124. https://doi.org/10.3732/apps.1600124 Bell, K. L., Burgess, K. S., Botsch, J. C., Dobbs, E. K., Read, T. D., & Brosi, B. J. (2019). Quantitative and qualitative assessment of pollen DNA metabarcoding

using constructed species mixtures. Molecular Ecology, 28(2), 431–455. https://doi.org/10.1111/mec.14840 Brunet, J., Zhao, Y., & Clayton, M. K. (2019). Linking the foraging behavior of three bee species to pollen dispersal and gene flow. PLOS ONE, 14(2), e0212561. https://doi.org/10.1371/journal.pone.0212561 Buchmann, S. L., & Nabhan, G. P. (1996). The Forgotten Pollinators. Island Press. Cameron, S. A., Lozier, J. D., Strange, J. P., Koch, J. B., Cordes, N., Solter, L. F., & Griswold, T. L. (2011). Patterns of widespread decline in North American bumble bees. Proceedings of the National Academy of Sciences, 108(2), 662–667. https://doi.org/10.1073/pnas.1014743108 Cornman, R. S., Otto, C. R. V., Iwanowicz, D., & Pettis, J. S. (2015). Taxonomic Characterization of Honey Bee (Apis mellifera) Pollen Foraging Based on Non- Overlapping Paired-End Sequencing of Nuclear Ribosomal Loci. PLOS ONE, 10(12), e0145365. https://doi.org/10.1371/journal.pone.0145365 de Vere, N., Jones, L. E., Gilmore, T., Moscrop, J., Lowe, A., Smith, D., … Ford, C. R. (2017). Using DNA metabarcoding to investigate honey bee foraging reveals limited flower use despite high floral availability. Scientific Reports, 7(1), 42838. https://doi.org/10.1038/srep42838 Djemiel, C., Grec, S., & Hawkins, S. (2017). Characterization of Bacterial and Fungal Community Dynamics by High-Throughput Sequencing (HTS) Metabarcoding during Flax Dew-Retting. Frontiers in Microbiology, 8. https://doi.org/10.3389/fmicb.2017.02052 Durant, J. L., & Otto, C. R. V. (2019). Feeling the sting? Addressing land-use changes can mitigate bee declines. Land Use Policy, 87, 104005. https://doi.org/10.1016/j.landusepol.2019.05.024 Erdtman, G. (1943). An Introduction to Pollen Analysis. chronica Botanica, waltham. Mass., uSa. Estravis Barcala, M. C., Palottini, F., & Farina, W. M. (2019). Honey bee and native solitary bee foraging behavior in a crop with dimorphic parental lines. PLOS ONE, 14(10), e0223865. https://doi.org/10.1371/journal.pone.0223865 Gallai, N., Salles, J.-M., Settele, J., & Vaissière, B. E. (2009). Economic valuation of the vulnerability of world agriculture confronted with pollinator decline. Ecological Economics, 68(3), 810–821. https://doi.org/10.1016/j.ecolecon.2008.06.014 Galliot, J.-N., Brunel, D., Bérard, A., Chauveau, A., Blanchetête, A., Lanore, L., & Farruggia, A. (2017). Investigating a flower-insect forager network in a mountain grassland community using pollen DNA barcoding. Journal of Insect Conservation, 21(5–6), 827–837. https://doi.org/10.1007/s10841-017-0022-z Gao, T., Yao, H., Song, J., Zhu, Y., Liu, C., & Chen, S. (2010). Evaluating the feasibility of using candidate DNA barcodes in discriminating species of the large Asteraceae family. BMC Evolutionary Biology, 10(1), 324. https://doi.org/10.1186/1471-2148-10-324 Hawkins, J., Vere, N. de, Griffith, A., Ford, C. R., Allainguillaume, J., Hegarty, M. J., … Adams-Groom, B. (2015). Using DNA Metabarcoding to Identify the Floral Composition of Honey: A New Tool for Investigating Honey Bee Foraging

Preferences. PLOS ONE, 10(8), e0134735. https://doi.org/10.1371/journal.pone.0134735 Hebert, P. D. N., Cywinska, A., Ball, S. L., & deWaard, J. R. (2003). Biological identifications through DNA barcodes. Proceedings of the Royal Society of London. Series B: Biological Sciences, 270(1512), 313–321. https://doi.org/10.1098/rspb.2002.2218 Janzen, D. H. (1983). Seed and pollen dispersal by animals: Convergence in the ecology of contamination and sloppy harvest. Biological Journal of the Linnean Society, 20(1), 103–113. https://doi.org/10.1111/j.1095-8312.1983.tb01592.x Kartzinel, T. R., Chen, P. A., Coverdale, T. C., Erickson, D. L., Kress, W. J., Kuzmina, M. L., Rubenstein, D. I., Wang, W., & Pringle, R. M. (2015). DNA metabarcoding illuminates dietary niche partitioning by African large herbivores. Proceedings of the National Academy of Sciences of the United States of America, 112(26), 8019–8024. Keller, A., Danner, N., Grimmer, G., Ankenbrand, M., Ohe, K. von der, Ohe, W. von der, … Steffan‐Dewenter, I. (2015). Evaluating multiplexed next-generation sequencing as a method in palynology for mixed pollen samples. Plant Biology, 17(2), 558–566. https://doi.org/10.1111/plb.12251 Kraaijeveld, K., Weger, L. A. de, García, M. V., Buermans, H., Frank, J., Hiemstra, P. S., & Dunnen, J. T. den. (2015). Efficient and sensitive identification and quantification of airborne pollen using next-generation DNA sequencing. Molecular Ecology Resources, 15(1), 8–16. https://doi.org/10.1111/1755- 0998.12288 Latta, R. G., Linhart, Y. B., Fleck, D., & Elliot, M. (1998). Direct and indirect estimates of seed versus pollen movement within a population of ponderosa pine. Evolution, 52(1), 61–67. https://doi.org/10.1111/j.1558-5646.1998.tb05138.x Losey, J. E., & Vaughan, M. (2006). The economic value of ecological services provided by insects. BioScience, 56(4), 311. https://doi.org/10.1641/0006- 3568(2006)56[311:TEVOES]2.0.CO;2 Macgregor, C. J., Kitson, J. J. N., Fox, R., Hahn, C., Lunt, D. H., Pocock, M. J. O., & Evans, D. M. (2019). Construction, validation, and application of nocturnal pollen transport networks in an agro-ecosystem: A comparison using light microscopy and DNA metabarcoding. Ecological Entomology, 44(1), 17–29. https://doi.org/10.1111/een.12674 Martins, A. C., Gonçalves, R. B., & Melo, G. A. R. (2013). Changes in wild bee fauna of a grassland in Brazil reveal negative effects associated with growing urbanization during the last 40 years. Zoologia (Curitiba), 30(2), 157–176. https://doi.org/10.1590/S1984-46702013000200006 Ollerton, J., Winfree, R., & Tarrant, S. (2011). How many flowering plants are pollinated by animals? Oikos, 120(3), 321–326. https://doi.org/10.1111/j.1600- 0706.2010.18644.x Ollerton, J., Price, V., Armbruster, W. S., Memmott, J., Watts, S., Waser, N. M., Totland, Ø., Goulson, D., Alarcón, R., Stout, J. C., & Tarrant, S. (2012). Overplaying the role of honey bees as pollinators: A comment on Aebi and Neumann (2011).

Trends in Ecology & Evolution, 27(3), 141–142. https://doi.org/10.1016/j.tree.2011.12.001 Pornon, A., Escaravage, N., Burrus, M., Holota, H., Khimoun, A., Mariette, J., Pellizzari, C., Iribar, A., Etienne, R., Taberlet, P., Vidal, M., Winterton, P., Zinger, L., & Andalo, C. (2016). Using metabarcoding to reveal and quantify plant-pollinator interactions. Scientific Reports, 6(1), 27282. https://doi.org/10.1038/srep27282 Potter, C., Vere, N. de, Jones, L. E., Ford, C. R., Hegarty, M. J., Hodder, K. H., … Franklin, E. L. (2019). Pollen metabarcoding reveals broad and species-specific resource use by urban bees. PeerJ, 7, e5999. https://doi.org/10.7717/peerj.5999 Potts, S. G., Biesmeijer, J. C., Kremen, C., Neumann, P., Schweiger, O., & Kunin, W. E. (2010). Global pollinator declines: Trends, impacts and drivers. Trends in Ecology & Evolution, 25(6), 345–353. https://doi.org/10.1016/j.tree.2010.01.007 Rahl, M. (2008). Microscopic Identification and Purity Determination of Pollen Grains. In M. G. Jones & P. Lympany (Eds.), Allergy Methods and Protocols (pp. 263– 269). https://doi.org/10.1007/978-1-59745-366-0_22 Richards, A. J. (1997). Plant Breeding Systems. Garland Science. Richardson, R. T., Lin, C.-H., Sponsler, D. B., Quijia, J. O., Goodell, K., & Johnson, R. M. (2015a). Application of ITS2 metabarcoding to determine the provenance of pollen collected by honey bees in an agroecosystem. Applications in Plant Sciences, 3(1), 1400066. https://doi.org/10.3732/apps.1400066 Richardson, R. T., Lin, C.-H., Quijia, J. O., Riusech, N. S., Goodell, K., & Johnson, R. M. (2015b). Rank-based characterization of pollen assemblages collected by honey bees using a multi-locus metabarcoding approach. Applications in Plant Sciences, 3(11), 1500043. https://doi.org/10.3732/apps.1500043 Richardson, R. T., Curtis, H. R., Matcham, E. G., Lin, C.-H., Suresh, S., Sponsler, D. B., Hearon, L. E., & Johnson, R. M. (2019). Quantitative multi-locus metabarcoding and waggle dance interpretation reveal honey bee spring foraging patterns in Midwest agroecosystems. Molecular Ecology, 28(3), 686–697. https://doi.org/10.1111/mec.14975 Roof, S. M., DeBano, S., Rowland, M. M., & Burrows, S. (2018). Associations between blooming plants and their bee visitors in a riparian ecosystem in eastern Oregon. Northwest Science, 92(2), 119. https://doi.org/10.3955/046.092.0205 Sajwani, A., Farooq, S. A., & Bryant, V. M. (2014). Studies of bee foraging plants and analysis of pollen pellets from hives in Oman. Palynology, 38(2), 207–223. https://doi.org/10.1080/01916122.2013.871652 Scheper, J., Reemer, M., Kats, R. van, Ozinga, W. A., Linden, G. T. J. van der, Schaminée, J. H. J., Siepel, H., & Kleijn, D. (2014). Museum specimens reveal loss of pollen host plants as key factor driving wild bee decline in The Netherlands. Proceedings of the National Academy of Sciences, 111(49), 17552– 17557. https://doi.org/10.1073/pnas.1412973111 Sickel, W., Ankenbrand, M. J., Grimmer, G., Holzschuh, A., Härtel, S., Lanzen, J., … Keller, A. (2015). Increased efficiency in identifying mixed pollen samples by meta-barcoding with a dual-indexing approach. BMC Ecology, 15(1), 20. https://doi.org/10.1186/s12898-015-0051-y

Smart, M. D., Cornman, R. S., Iwanowicz, D. D., McDermott-Kubeczko, M., Pettis, J. S., Spivak, M. S., & Otto, C. R. V. (2017). A comparison of honey bee-collected pollen from working agricultural lands using light microscopy and ITS metabarcoding. Environmental Entomology, 46(1), 38–49. https://doi.org/10.1093/ee/nvw159 Valentini, A., Taberlet, P., Miaud, C., Civade, R., Herder, J., Thomsen, P. F., Bellemain, E., Besnard, A., Coissac, E., Boyer, F., Gaboriaud, C., Jean, P., Poulet, N., Roset, N., Copp, G. H., Geniez, P., Pont, D., Argillier, C., Baudoin, J.-M., … Dejean, T. (2016). Next-generation monitoring of aquatic biodiversity using environmental DNA metabarcoding. Molecular Ecology, 25(4), 929–942. https://doi.org/10.1111/mec.13428 Von Der Ohe, W., Persano Oddo, L., Piana, M. L., Morlot, M., & Martin, P. (2004). Harmonized methods of melissopalynology. Apidologie, 35(Suppl. 1), S18–S25. https://doi.org/10.1051/apido:2004050 Whitehead, D. R. (1983). Wind pollination: some ecological and evolutionary perspectives. Pollination Biology, 97(08). Whittington, R., Winston, M. L., Tucker, C., & Parachnowitsch, A. L. (2004). Plant- species identity of pollen collected by bumblebees placed in greenhouses for tomato pollination. Canadian Journal of Plant Science, 84(2), 599–602. https://doi.org/10.4141/P02-192

Chapter 2: Testing Quantitative Capabilities of DNA Metabarcoding for Use in Bee Foraging Studies Using Pollen Mixtures of Known Concentrations

Katherine A. Arstingstall, Sandra J. DeBano, Xiaoping Li, David E. Wooster, Mary M. Rowland, Skyler Burrows, Kenneth Frost

Abstract Many bee populations worldwide are experiencing significant declines, and one approach used to reverse these declines is to increase bee habitat quality and quantity by planting flowers that are important food sources for bees. Although multiple methods can be used to identify bee-preferred floral resources, recently, metabarcoding of bee pollen has been used to successfully identify which plants bees are foraging on. However, several questions about this method are unresolved, including the type of contamination threshold to use, the extent to which these data are quantitative, the ability of the technique to detect rare flower visits by bees, and the potential role of environmental contamination in mischaracterizing bee foraging behavior. To address these questions, we collected pollen from five plant species, created pollen mixtures in the laboratory that varied in species richness and evenness, and used DNA metabarcoding of the ITS2 region to identify plant species in the mixtures. We analyzed sequencing data using two types of contamination thresholds: a liberal threshold, in which plant taxa making up <0.01% of total sequencing reads were removed, and a more conservative threshold, in which any taxonomic assignment whose sequence read count was less than a negative control-based threshold was set to zero. We compared the proportion of pollen by mass to the proportion of sequencing reads for each plant species in the mixtures after using both types of thresholds. In both cases, we found that the number of sequencing reads produced by ITS2 metabarcoding cannot be interpreted quantitatively as a measure of the relative proportion of a species in a mixed pollen load. The relationship between the proportion of pollen and sequencing reads was not statistically significant for two of the four mixtures analyzed, and certain species were consistently over and underrepresented. When each individual species was examined separately, the relationship between the

proportion of pollen and sequencing reads was positive and statistically significant, but proportions for two species varied strongly from a one-to-one relationship. When using the liberal threshold, all plant species in the mixtures were detected above the threshold. However, additional plant species were detected in the single-species samples. When using the conservative threshold, no additional plant species were detected in the single- species samples, but some species used to make pollen loads were not detected above the threshold in three of the five mixture types, resulting in false negatives. The results of this study suggest that sequence reads from ITS2 metabarcoding are not good proxies for relative abundance of pollen in a mixed sample, and that the type of sequence count removal threshold used can have a major effect on conclusions drawn from studies that use metabarcoding of bee pollen to study plant-pollinator interactions.

Introduction Bee populations worldwide are currently experiencing significant declines (Goulson et al., 2015; Cameron & Sadd, 2020), and one way to reverse these declines is to increase the quantity or quality of their habitat by planting flowers that are important food sources for bees (Durant & Otto, 2019). However, determining which plant species are significant food sources for bees can be challenging. Traditional methods for describing bee-flower interactions (e.g., observation, microscopy) are time consuming, can require specialized expertise, and often lack taxonomic resolution (Free, 1970; Rahl, 2008; Cornman et al., 2015). However, the recent use of DNA metabarcoding on bee pollen shows potential to be a more effective method for identifying food sources for bees than traditional methods, resulting in higher taxonomic resolution and revealing a more detailed record of bee foraging behavior (Keller et al., 2015; Richardson et al., 2015a,b; Pornon et al., 2016; Smart et al., 2017; Potter et al., 2019, Arstingstall et al., in review). However, a number of questions remain about pollen metabarcoding analyses and potential limitations of metabarcoding of pollen to establish floral resource preferences of bees. For example, there is currently no standardized sequence count removal threshold protocol for analyzing pollen metabarcoding data. Different thresholds are used in many

pollen metabarcoding studies, despite the fact that the type of threshold used may affect how results of pollinator studies are interpreted (Pornon et al., 2016; Bell et al., 2019; Macgregor et al., 2019; Potter et al., 2019; Richardson et al., 2019). There is also continuing disagreement over whether sequence read data can be used to estimate the abundance of a particular plant species in a mixed pollen sample (Keller et al., 2015; Kraaijeveld et al., 2015; Pornon et al., 2016; Richardson et al., 2015a,b; Smart et al., 2017; Bell et al., 2019). Another question relates to the ability of metabarcoding techniques to detect rare flower visits by bees. Finally, the extent of environmental contamination of flower pollen, and how that may complicate conclusions about bee foraging behavior is not clear. While many gene regions can be used in metabarcoding studies, ITS2 has been identified as a successful universal DNA barcode for plant identification (Chen et al., 2010). ITS2 sequences are short and thus easy to amplify using one pair of universal primers; the ITS2 gene region mutates frequently, allowing it to differentiate between closely related plant species; and past work has shown up to 92.7% species identification accuracy using the ITS2 region (Chen et al., 2010; Song et al., 2012). Other studies have found that, with the use of a complete reference database, pollen DNA metabarcoding of the ITS2 gene region provides reliable presence/absence data on which floral resources bees are using (Richardson et al., 2015a,b; Smart et al., 2017; Bell et al., 2019; Arstingstall et al., In Review). The ITS2 barcoding region, which is a nuclear ribosomal marker, can also be used with chloroplast markers (e.g., trnL, rbcL, matK, psbA-trnH) to increase resolution and reduce amplification bias (Chen et al., 2010; Kraaijeveld et al., 2014; Richardson et al., 2015b; Bell et al., 2019). The issue of whether sequence read data can be used to quantify each plant species in a mixed pollen load is unresolved. Studies that use the same barcode markers continue to find contradictory results. For example, when comparing proportions of sequencing reads obtained using DNA metabarcoding of the ITS2 gene region to proportions of pollen grains counted using microscopy for plant species in a pollen load, Keller et al. (2015) found similar relative abundances, while Richardson et al. (2015a) found no association. Pornon et al (2016) found a positive relationship between the

amount of isolated DNA from each plant species and the number of corresponding ITS1 and trnL sequencing reads. Bell et al. (2019) used the ITS2 marker coupled with a chloroplast marker (rbcL) in a multiplex polymerase chain reaction (PCR) and analyzed the sequencing reads for each barcode marker separately. They found that although a statistically significant correlation existed between the proportion of ITS2 and rbcL sequence reads for each species and the proportion of pollen grains in the mixtures, only a small amount of variance was explained by the proportion of pollen grains. Richardson et al. (2019) used the ITS2 marker with three chloroplast markers (trnL, rbcL, and trnH), running separate PCR reactions for each barcode marker and removing any taxonomic assignments discovered using only one barcode marker. They found that the relationship between the microscopy pollen counts and the median number of sequence reads produced for each plant family was statistically significant. When using pollen metabarcoding to identify plant species in a mixed species sample, studies have shown that certain plant species can be over or underrepresented by the number of sequencing reads when compared to the proportion of pollen grains in a sample (Richardson et al., 2015a; Smart et al., 2017; Bell et al., 2019). Misrepresentation might occur for several reasons including variable gene copy number (Álvarez et al., 2003; Kembel et al., 2012), isolation bias (Brooks et al., 2015), and amplification bias (Bell et al., 2017; Pawluczyk et al., 2015). Brooks et al. (2015) found that different combinations of DNA extraction kits and number of PCR cycles resulted in dramatically different proportions of sequencing reads per taxa. Additionally, sequence variation at the barcode priming site can affect the amplification efficiency of a plant species, resulting in false negatives (Pawluczyk et al., 2015; Pompanon et al., 2012). Nevertheless, studies have found quantitative relationships and propose that sequencing reads are a good proxy for the relative abundance of a species in a pollen load (Keller et al., 2015; Kraaijeveld et al., 2015; Pornon et al., 2016). Resolving the issue of whether DNA metabarcoding of bee pollen produces quantitative results is necessary for correctly interpreting plant- pollinator relationships derived using the approach. Another issue that must be addressed when using metabarcoding to describe bee foraging behavior relates to contamination. Although field contamination (e.g., pollen

moving around in the environment via wind and other insects) is difficult to control for, we can control for contamination in the laboratory using extraction blanks and PCR negatives, which contain sterile water in place of genetic material. These negative controls quantify the number of sequencing reads in a sample that can be attributed to laboratory contamination alone. The number of sequencing reads found in negative controls can be used to create a sequence count removal threshold, and any taxonomic assignment whose read count falls below this threshold can be considered contamination or other “background noise” and removed from further analysis (Bell et al., 2017, 2019; Macgregor et al., 2019). However, there is concern that this method may also eliminate rare plant-bee interactions occurring at low frequencies (Bell et al., 2017). Creating laboratory mixtures with plant species present in amounts small enough to represent a single plant-bee interaction (~1,100 pollen grains or 0.3 mg of pollen [Harder, 1990; Goodell & Thomson, 1996]) could help determine whether rare plant-bee interactions can be detected using DNA metabarcoding of bee pollen or if the number of sequencing reads produced by the pollen collected during these rare interactions might fall below the sequence count removal threshold. The ability of DNA metabarcoding to detect rare plant-bee interactions is dependent on the type of threshold that is used. There is currently no standardized protocol for creating and using a sequence count removal threshold in pollen metabarcoding studies, despite the fact that different threshold protocols can potentially yield different interpretations of the same results. Richardson et al. (2019) discarded “taxonomic groups represented by <0.01% of the data,” whereas, Bell et al. (2019) used the maximum number of sequencing reads in a negative control as a threshold, and set any taxonomic assignment with fewer reads than this threshold to zero. One type of threshold removes entire taxonomic groups, while the other removes specific bee-plant interactions, likely yielding different results and different interpretations of pollinator behavior. Identifying a standardized threshold protocol would eliminate this problem and allow for comparison of pollinator foraging behavior (e.g., plant-pollinator networks) across multiple pollen metabarcoding studies.

Another unresolved issue relates to how well plant-pollinator networks produced from plant species assignments obtained using metabarcoding data represent bee foraging behavior and floral preferences. Several studies have found that plant-pollinator networks based on metabarcoding data are more complex, including more plant taxa and interactions than networks created using bee foraging observations (Pornon et al., 2016; Potter et al., 2019). Although this may be an accurate depiction of bee foraging behavior, the tendency of metabarcoding networks to be more complex than those based on foraging observations alone could be explained by other mechanisms. For example, pollen is moved around the environment by wind and other insects (Janzen, 1983; Whitehead, 1983; Latta et al., 1998), so a flower could have pollen from several other plant species residing on it before a bee visits it. Thus, a bee could inadvertently pick up pollen from multiple plant species in one flower visit. Yet no studies have isolated pollen collected from flowers of single species in natural settings and sequenced their pollen to determine if additional plant species are detected. Such data would shed light on whether plant species assignments obtained using DNA metabarcoding of bee pollen tend to overestimate the complexity of plant-pollinator networks, and if so, allow for estimating the proportion of interactions that can be attributed to bee foraging versus those that may be attributed to movement of pollen by wind and other insects. Here, we collected pollen by hand from five plant species to create single-species pollen samples and five types of mixtures of known concentrations, varying in species richness (3-5 species) and evenness (uniform to highly skewed). We used DNA metabarcoding of the ITS2 gene region to identify the plant species in the samples, and analyzed the sequencing data using both a liberal and a conservative sequence count removal threshold. The specific objectives of this study were to determine whether 1) the two types of threshold yield different results, 2) the proportions of plant species in a pollen mixture based on mass correspond with proportions of plant species in a pollen mixture based on number of sequencing reads, 3) rare plant-bee interactions can be detected using DNA metabarcoding, and 4) additional plant species are detected in single-species pollen samples.

Methods Flower Collection Flowers were collected in July 2017 from the United States Forest Service (USFS) Starkey Experimental Forest and Range (Starkey) and from Hermiston, both located in eastern Oregon. Slender cinqefoil (Potentilla gracilis), Oregon checkerbloom (Sidalcea oregana), and mountain goldenbanner (Thermopsis montana) flowers were collected at Starkey (45.2332°N, 118.5511°W). Starkey is a long-term research site in Union County (elevation 1,130-1,500 m) that was established in 1940. Sampling took place at sites along Meadow Creek, a major tributary of the Upper Grande Ronde River that flows through Starkey. Hairy vetch (Vicia villosa) and Scotch thistle (Onopordum acanthium) flowers were collected in Hermiston (45.8169°N, 119.2846°W). Flowers from each plant species were collected by hand and stored in separate plastic zip lock bags. Each bag contained flowers sampled from multiple plants of the same species. Flowers were stored at -20°C until further processing.

Isolating Pollen from Flowers All tools used during pollen isolation were placed in 10% bleach for at least one min prior to use and in between each bout of pollen isolation to reduce contamination. Stamens were removed from each flower using forceps and placed in a glass vial filled with sterilized water. All stamens from each plant species were placed in the same vial. Vials were shaken vigorously for one min to detach pollen from anthers. The solution was poured through a fine mesh sieve into a vacuum filtration system. The pollen was collected on a 5 µm mixed cellulose ester filter. The filters were placed on a watch glass and examined under a dissecting microscope. If any small insects, anthers, or other debris were found on the filter, they were removed using bleached forceps. The filters were then placed in a drying oven at 65°C for approximately two hours. Once dry, filters were placed in individual 50 mL centrifuge tubes, and 10 mL of acetone was added to each tube. The centrifuge tubes were vortexed until the filters were completely dissolved, and the solution was mixed thoroughly. The tubes were then centrifuged for 3 min at 2,000 rpm, the supernatant was discarded, 1 mL of acetic acid was added to each tube, and the

pellet was re-suspended. The solution was transferred to a 1.5 mL microcentrifuge tube. The mass of the 1.5 mL microcentrifuge tube was recorded before transferring the solution in order to obtain a final pollen mass at the end of the process. The solution was centrifuged for 30 s at 12,000 rpm, and the supernatant was discarded. Two washes were performed with 1 mL of sterilized water for 30 s at 12,500 rpm. This was followed by two washes with 1 mL of ethanol for 30 s at 12,500 rpm and 3 min at 13,500 rpm. The supernatant was discarded, and the pollen pellet was dried for 30 min. The mass of each 1.5 mL screw cap microcentrifuge tube containing pollen was recorded, and the mass of the empty microcentrifuge tube was subtracted from this value to give the mass of the dry pollen pellet.

Preparation of Pollen Mixtures Pollen mixtures were created from stock pollen samples that were isolated from five plant species (Table 2.1). All tools were placed in 10% bleach for at least one min before being used on different plant species. Appropriate amounts of pollen were transferred from each stock sample into 1.5 mL screw cap microcentrifuge tubes to create five single-species samples with three replicates each and five mixtures with three replicates each, representing a range of species richness (3-5 species) and evenness (Table 2.1). A total of 30 pollen samples were prepared.

DNA Extraction and PCR DNA extractions were conducted during fall of 2017. We used the Macherey- Nagel Nucleospin Food kit (Macherey-Nagel, Bethlehem, Pennsylvania, USA), following the “isolation of genomic DNA from honey or pollen” supplementary protocol. We added 1 mm glass beads to 1.5 mL screw cap microcentrifuge tubes, and the samples were homogenized in a Mini-BeadBeater-24 (Biospec Products, Bartlesville, Oklahoma, USA). Negative controls (i.e., sterilized water in place of pollen) were included with each round of DNA extraction. Polymerase chain reactions (PCR) were conducted in January and February of 2019. Following Sickel et al. (2015), we used the dual-indexing pollen DNA

metabarcoding strategy and the second internal transcribed spacer (ITS2) primers ITS S2F and ITS4R. We added Illumina overhang adapter sequences to each primer (Klindworth et al., 2013). Each PCR reaction contained 10 uL of 5X Green GoTaq® Reaction Buffer, 0.3 uL of 10 mM dNTPs, 1 uL of each primer, 0.2 uL of GoTaq® DNA Polymerase, 35.5 uL of water, and 2 uL of the template DNA with a total volume of 50 uL per reaction. The PCR cycles began with an initial heat activation period for 3 min at 95°C, which was followed by 35 cycles of 30 s at 95°C, 30 s at 55°C, and 1 min at 72°C. We included a final extension of 10 min at 72°C, and the samples were held at 10°C until further processing. We included negative controls in each round of PCR, using water in place of template DNA. A 5 uL sample of each PCR product was electrophoresed in a 2% agarose gel, stained with GelRedTM (Biotium Inc, Fremont, CA), and visualized under UV light to confirm the presence of the appropriately sized amplicons. We purified the PCR products using the Promega Wizard SV Gel and PCR Clean-Up System. We quantified the DNA for each sample using a Nanodrop 2000 Spectrophotometer (Thermo Scientific). Samples were diluted with sterilized water in 96-well plates. At the Center for Genome Research and Biocomputing at Oregon State University, the samples were indexed, pooled, and sequenced in a standard flow cell v3 300bp paired-end full service run of the Illumina MiSeq instrument.

Bioinformatics- Read Quality Filtering and Denoising A total of 4,530,819 raw reads were retrieved across 69 samples for ITS2 (data from 39 of these samples were used for a separate study). We used the open-source QIIME2 pipeline (version 2019-04) (Bolyen et al., 2019) to analyze species composition of pollen loads. We used the QIIME2 DADA2 plugin (Callahan et al., 2016) to filter read quality and denoise reads (--p-trunc-len-f 255, --p-trunc-len-r 203). Noisy reads and sequence reads with a median sequencing quality score below 30 were disregarded. We joined paired-end reads to create contigs, and we removed duplicate sequences and sequences with chimera. This resulted in a feature table containing optimized sequences, amplicon sequence variants (ASV), and their abundance, which was used for additional

analyses (e.g., taxonomic classification). Approximately 2,608,088 (56%) paired-end reads were retained for ITS2 after filtering and denoising.

Bioinformatics- Reference Database Construction A list of plant species known to occur at Starkey has been developed over multiple years by botanists. Reference sequences of the ITS2 region for the plants included in this list were extracted from publicly available data of Sickel et al. (2015) (https://www.biozentrum.uni-wuerzburg.de/dna-analytics/molecular-biodiversity- group/downloads/). We used a python package: NCBI-Companion (https://github.com/lixiaopi1985/NCBI_Companion) to query the NCBI nucleotide database to obtain sequences of species that were not present in the data of Sickel et al. (2015). ITSx (Bengtsson-Palme et al. 2013) software was used to detect and extract ITS2 sequences. The database contained 77% of the total species on the Starkey plant list. ITS2 sequences from Onopordum acanthium and Vicia villosa were manually added to the local Starkey database.

Bioinformatics- Training Classifier and Taxonomy Classification We used the Naïve Bayes algorithm provided by the QIIME2 feature-classifier plugin (Bokulich et al. 2018) to train a classifier for taxonomic assignment based on the ITS2 database. The default settings were applied (i.e., kmer length = 7, confidence threshold = 0.7) with high accuracy and low recall. We used the classify-sklearn tool of the QIIME2 feature-classifier plugin (Bokulich et al. 2018) to import the reference sequences and then assign taxonomy to each ASV with default settings.

Sequence Count Removal Thresholds We used two types of sequence count removal thresholds to analyze the sequencing data: a liberal threshold and a conservative threshold. To create the liberal threshold, we found the total number of sequencing reads and the sum of sequencing reads for each plant taxa. Any taxonomic group whose sequence count was less than

0.01% of the total number of sequencing reads was removed before further analysis (following Richardson et al. 2019). To create the conservative threshold, we found the average and the standard deviation of the number of sequencing reads in the negative control samples. We added 1.645 standard deviations to the average and used this number as our sequence count removal threshold, accounting for 95% of possible “background noise” detected in the pollen samples (Armbruster & Pry, 2008). We set the sequence count to zero for any taxonomic assignment whose read count fell below the threshold.

Data Analysis We examined the relationship between the proportion of pollen by mass and the proportion of sequencing reads produced from each plant species in four of the mixtures using a linear regression. We did not conduct a regression for the uniform mixture (mixture 3) because we lacked variation in the proportion of pollen by mass for the sample. We also conducted regressions of proportion of pollen mass and reads for each individual species across all mixtures. All analyses were performed in R, version 4.0.0 (R Core Team, 2020).

Results Detection Using a Liberal Threshold All plant species included in the laboratory-prepared mixtures were correctly identified using plant species assignments obtained from DNA metabarcoding and were detected above the sequence count removal threshold (Figure 2.1, Table 2.1). Six additional plant species that were not used to create the laboratory-prepared mixtures were detected above the sequence count removal threshold in the mixtures and single- species samples (Table A1.1). Trace amounts of Onopordum acanthium and Thermopsis montana were detected in all single-species samples, and sequencing reads assigned to Potentilla gracilis made up an average of 1.6% of the sequencing reads in Sidalcea oregana single-species samples (Table 2.1). All “rare” plant species in mixture 5 were detected above the sequence count removal threshold (Table 2.1).

Detection Using a Conservative Threshold All plant species included in the laboratory-prepared mixtures were correctly identified using plant species assignments obtained from DNA metabarcoding, but some were not consistently detected above the sequence count removal threshold, resulting in false negatives (Table 2.2). Onopordum acanthium and T. montana were always detected above the sequence count removal threshold (Figure 2.1, Table 2.2). S. oregana was not detected above the sequence count removal threshold in mixtures 2-5 (Table 2.2). P. gracilis and V. villosa were not detected above the sequence count removal threshold for one replicate of mixture 5 (Table 2.2). No additional species were detected in the single- species pollen mixtures, but trace amounts of P. gracilis were detected above the sequence count removal threshold in two S. oregana single-species sample replicates (Table 2.3).

Quantification: Comparing Proportion of Pollen by Mass to Proportion of Sequencing Reads For mixtures 1 and 5, there were statistically significant correlations between the proportion of sequencing reads and the mass of pollen from the plant species in those mixtures (Table 2.4). However, no statistically significant relationships were observed for mixtures 2 and 4 (Table 2.4). When each species was examined separately, there was a statistically significant relationship between the proportion of pollen by mass and the proportion of sequencing reads across all mixtures for each species (Table 2.5). However, T. montana was consistently overrepresented by the number of sequencing reads in mixtures 2 - 5 and S. oregana was consistently underrepresented by the number of sequencing reads in mixtures 1 – 5 (Figure 2.2).

Discussion Each of the five plant species used to create the laboratory-prepared mixtures was correctly identified using plant species assignments obtained from ITS2 metabarcoding data (Table 2.1). However, when using the conservative threshold, certain species were

not detected above the sequence count removal threshold in mixtures 2-5, resulting in several false negatives (Table 2.2). Bell et al. (2019), who used a similar threshold approach, found comparable results: plant species assignments from ITS2 pollen metabarcoding were accurate, but certain plant species were consistently underrepresented, often resulting in false negatives. Although past work suggests that nuclear ribosomal sequencing reads can be interpreted quantitatively as a measure of the amount of pollen in a sample (Keller et al., 2015; Pornon et al., 2016), Bell et al. (2019) and Richardson et al. (2019) found that sequencing results from ITS2 pollen DNA metabarcoding could not be interpreted quantitatively. The results of this study further validate these conclusions. There were not statistically significant relationships between mass of pollen and number of ITS2 sequencing reads for every mixture regardless of the type of threshold used (Table 2.4). Several studies using chloroplast barcode markers have found positive relationships between proportion of pollen grains and proportion of sequencing reads for plant species in mixed pollen samples (Kraaijeveld et al., 2015; Richardson et al., 2015b; Richardson et al., 2019). The PCR reactions in this study were performed as a multiplex, including ITS2 and rbcL primer sets. However, only 2% of the raw sequencing reads were rbcL sequences, from which meaningful data could not be derived. It is unclear why the ITS2 region was so heavily favored in our multiplex PCR reaction. Bell et al. (2019) was successful when multiplexing ITS2 and rbcL and, when examining the sequencing reads from each barcode marker separately, found a statistically significant relationship between the number of sequencing reads and the proportion of pollen from each species in the laboratory prepared mixtures. However, for both barcode markers, only a small amount of variance in sequence reads was explained by the proportion of pollen grains, leading them to conclude that the number of sequence reads could not be used to determine the amount of pollen of a particular plant species in a pollen sample, regardless of whether nuclear or chloroplast barcode markers were used. Richardson et al. (2019) also used ITS2 and rbcL barcode markers. They conducted separate PCR reactions for each primer set and found that rbcL metabarcoding results were strongly correlated with microscopy results. Future research could be conducted to determine whether rbcL genes

24 are amplified more efficiently in separate PCR reactions as opposed to a multiplex reaction and whether there is a difference in the relationship between proportion of sequencing reads and proportion of pollen in a sample using either method. Certain plant species were consistently over and underrepresented in the laboratory-created pollen mixtures. The proportion of sequencing reads produced for T. montana was on average 6.4 and 6.5 times greater than the proportion of pollen by mass after using the liberal and conservative thresholds respectively (Table 2.1, Table 2.3). When using the liberal threshold, the proportion of sequencing reads produced for S. oregana was on average 33.7 times lower than the proportion of pollen by mass (Table 2.1). When using the conservative threshold, S. oregana was only detected above the sequence count removal threshold in one fifth of the mixture samples. In these samples, the proportion of sequencing reads produced was on average 3.3 times lower than the proportion of pollen by mass (Table 2.3). There are several reasons why the number of sequencing reads produced for a species may not be proportional to the mass of the pollen: DNA isolation bias (Brooks et al., 2015), amplification bias (Pompanon et al., 2012; Pawluczyk et al., 2015), and differences in gene copy number (Kembel et al., 2012; Song et al., 2012). It is unlikely that DNA isolation bias occurred during this study because we used the same DNA extraction kit for all samples. However, amplification bias and gene copy number could have affected the number of sequencing reads obtained for T. montana and S. oregana. ITS2 is a multi-copy gene region, with high variation in copy number among different plant species. T. montana may have a greater number of ITS2 gene regions than the other species used to create mixtures in this study, resulting in overrepresentation of the species in DNA metabarcoding results. The underrepresentation of S. oregana may have been caused by a reduced number of ITS2 gene regions relative to the other plant species used in the mixtures or by reduced amplification efficiency, which can result in false negatives when sequence variation occurs at the priming site for certain species (Pawluczyk et al., 2015). False negatives can also occur from PCR inhibition. PCR inhibitors are a diverse group of substances representing a wide range of physical properties and mechanisms that inhibit the polymerase chain reaction. Plant PCR inhibitors (e.g., pectin, polyphenols,

polysaccharides, xylan) can often be removed through dilution of the isolated nucleic acids, which results in the dilution of the PCR inhibitors (Schrader et al., 2012). We encountered PCR inhibition when amplifying DNA from P. gracilis pollen. The mixtures containing pollen from P. gracilis did not produce visible bands when gel electrophoresed. We performed a series of dilutions on the mixtures, some of which produced visible bands, and we chose the dilution ratio that resulted in the PCR product which produced the brightest band. The inhibition likely occurred because of the high concentration of DNA from individual plant species in our samples, which is less likely in field collected bee pollen loads. However, researchers should be aware of PCR inhibitors and the steps they can take to remove them or decrease their effects. When researchers use DNA metabarcoding to determine which plant species are in a pollen sample, the reference library to which they compare the DNA sequences in their samples may contain hundreds or even thousands of plant species. Attempts to derive quantitative information from sequence reads would require determining which plant species may be over or underrepresented, and every combination of species would need to be tested because a species could react differently when in combination with different plant species (Bell et al., 2019; Pornon et al., 2016), a task that would be practically impossible. Several researchers have proposed the use of correction factors to derive quantitative information from DNA metabarcoding sequence reads (Richardson et al., 2015a; Thomas et al., 2016). However, our results and work by others suggest that a correction factor would not be effective. The proportions of plant species in a mixed pollen sample are affected by negative correlation bias and thus cannot be examined independently (Gloor et al., 2017). Although sequencing reads produced by pollen metabarcoding cannot be used to quantify the amount of a plant species in a mixed pollen sample, sequencing technologies are rapidly advancing and this type of quantification may be possible in the near future (Kane et al., 2012; Taberlet et al., 2012; Robert et al., 2013; Tang et al., 2014). When using DNA metabarcoding to identify plant species in a pollen mixture, a certain amount of contamination is inevitable. Contamination can occur at any stage of the DNA metabarcoding process, so extra precautions need to be taken in the field and

laboratory (e.g., using clean nets to catch individual bees, bleaching workspaces and tools) to reduce contamination. Sequencing negative controls from each stage in the laboratory process (e.g., DNA extraction, PCR amplification), which contain water in place of genetic material, help us determine the number of sequencing reads that can be attributed to laboratory contamination alone. Using the number of sequencing reads detected in the negative controls, we can define a sequence count removal threshold, which allows us to remove potential contamination and other “background noise” from further analysis. However, this approach could lead to a trade off when using DNA metabarcoding to study plant-bee interactions: while metabarcoding can detect a greater number of plant species than traditional methods, plant species that occur in lower quantities in a pollen load (e.g., from rare flower visits by bees) may fall below the sequence count removal threshold and be removed from analysis. The results presented here suggest that the detection of rare bee-flower interactions is dependent on the type of threshold used. Our highly skewed mixture contained 8.3 mg (81%) of pollen from O. acanthium, 1 mg (10%) of pollen from S. oregana, and 0.3 mg (3%) of pollen from each of P. gracilis, T. montana, and V. villosa. Pollen grain size ranges between 15 and 200 µm (Pacini et al., 2015). The average size of a maize pollen grain is 80 µm, making it a relatively average sized pollen grain (Porter et al., 1981). Porter et al., (1981) and Miller (1982) estimated that one mg of maize pollen contains about 2,000 to 4,000 grains. Therefore, 0.3 mg of pollen should contain somewhere between 600 and 1,200 pollen grains. Previous work indicates that bumble bees collect at least 1,100 pollen grains per flower visit (Harder, 1990; Goodell & Thomson, 1996). Therefore, 0.3 mg should be a good estimate of the lower limit of pollen collected by a bumble bee in a single (i.e., rare) visit. When using the liberal threshold, which removes any plant taxa whose total number of sequencing reads falls below 0.01% of total reads, all “rare” plant species were detected above the threshold (Table 2.1). However, when using the conservative threshold, S. oregana was not detected above the sequence count removal threshold in any of the highly skewed mixture replicates (Table 2.2). Furthermore, P. gracilis and V. villosa were not detected above the sequence count removal threshold in one of the

highly skewed mixture replicates. Therefore, it appears that we can consistently detect “rare” bee-flower visits when using the liberal threshold, but not when using the conservative threshold. However, using the conservative threshold may not present a serious limitation when studying plant-pollinator networks because single-bee flower visits are not necessarily ecologically significant. While many precautions can be taken in the laboratory to avoid and control for contamination issues, contamination is more difficult to control in the field. Wind and other insects can move pollen around the environment before sampling occurs. Therefore, pollen from other plant species could be present on a flower before a bee visits it, resulting in the bee picking up pollen from multiple plant species in one visit. If this is the case, we would expect to see multiple plant species that were not part of our artificial mixtures in our single-species samples. When using the liberal threshold, six additional plant species (Achillea millefolium, Hypericum perforatum, Lupinus sulphureous, Solidago missouriensis, Symphyotrichum spathulatum, and Trifolium repens) were detected in the single-species samples, while no additional plant species were detected in the single-species mixtures when using the conservative threshold. These results suggest that the use of the liberal threshold for analyzing pollen metabarcoding data may overestimate resource use by bees by including contamination and background noise (e.g., wind-pollinated plant species, pollen brought to flower by other insects). In this study, certain plant species were consistently over and underrepresented by the number of ITS2 sequence reads produced for the laboratory-created pollen mixtures, suggesting that DNA metabarcoding of the ITS2 gene region cannot be used to estimate the abundance of a plant species in a mixed pollen sample. However, more research is needed to confirm the quantitative abilities of sequencing reads produced using chloroplast barcode markers. Plant species assignments obtained using DNA metabarcoding of the ITS2 gene region were accurate, confirming the reliability of presence/absence data obtained using this method. We were able to consistently detect plant species present in amounts representing single bee-flower visits when using the liberal threshold, but several additional plant species that were not used to create the laboratory-prepared mixtures were also detected in the single-species samples. When we

used the conservative threshold, some false negatives occurred, but we did not detect any additional plant species in the single-species samples. Based on the information presented here, the conservative threshold seems to be most appropriate for the study of plant- pollinator networks. When describing plant-pollinator networks, it would be better to lose a few rare flower visits that may not be ecologically relevant, than to include potential field contamination and other background noise that could significantly skew results.

Literature Cited Álvarez, I., & Wendel, J. F. (2003). Ribosomal ITS sequences and plant phylogenetic inference. Molecular Phylogenetics and Evolution, 29(3), 417–434. https://doi.org/10.1016/S1055-7903(03)00208-2 Armbruster, D. A., & Pry, T. (2008). Limit of blank, limit of detection and limit of quantitation. The Clinical Biochemist Reviews, 29(Suppl 1), S49–S52. Arstingstall, K. A., DeBano, S. J., Li, X., Wooster, D. E., Rowland, M. M., Burrows, S., Frost, K. (In Revision). Capabilities and limitations of using DNA metabarcoding to study plant-pollinator interactions. Molecular Ecology Bell, K. L., Fowler, J., Burgess, K. S., Dobbs, E. K., Gruenewald, D., Lawley, B., Morozumi, C., & Brosi, B. J. (2017). Applying pollen DNA metabarcoding to the study of plant–pollinator interactions. Applications in Plant Sciences, 5(6), 1600124. https://doi.org/10.3732/apps.1600124 Bell, K. L., Burgess, K. S., Botsch, J. C., Dobbs, E. K., Read, T. D., & Brosi, B. J. (2019). Quantitative and qualitative assessment of pollen DNA metabarcoding using constructed species mixtures. Molecular Ecology, 28(2), 431–455. https://doi.org/10.1111/mec.14840 Bengtsson‐Palme, J., Ryberg, M., Hartmann, M., Branco, S., Wang, Z., Godhe, A., Wit, P. D., Sánchez‐García, M., Ebersberger, I., Sousa, F. de, Amend, A., Jumpponen, A., Unterseher, M., Kristiansson, E., Abarenkov, K., Bertrand, Y. J. K., Sanli, K., Eriksson, K. M., Vik, U., … Nilsson, R. H. (2013). Improved software detection and extraction of ITS1 and ITS2 from ribosomal ITS sequences of fungi and other eukaryotes for analysis of environmental sequencing data. Methods in Ecology and Evolution, 4(10), 914–919. https://doi.org/10.1111/2041-210X.12073 Bokulich, N. A., Kaehler, B. D., Rideout, J. R., Dillon, M., Bolyen, E., Knight, R., Huttley, G. A., & Gregory Caporaso, J. (2018). Optimizing taxonomic classification of marker-gene amplicon sequences with QIIME 2’s q2-feature- classifier plugin. Microbiome, 6(1), 90. https://doi.org/10.1186/s40168-018-0470- z Bolyen, E., Rideout, J. R., Dillon, M. R., Bokulich, N. A., Abnet, C. C., Al-Ghalith, G. A., Alexander, H., Alm, E. J., Arumugam, M., Asnicar, F., Bai, Y., Bisanz, J. E., Bittinger, K., Brejnrod, A., Brislawn, C. J., Brown, C. T., Callahan, B. J., Caraballo-Rodríguez, A. M., Chase, J., … Caporaso, J. G. (2019). Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2.

Nature Biotechnology, 37(8), 852–857. https://doi.org/10.1038/s41587-019-0209- 9 Brooks, J. P., Edwards, D. J., Harwich, M. D., Rivera, M. C., Fettweis, J. M., Serrano, M. G., Reris, R. A., Sheth, N. U., Huang, B., Girerd, P., Vaginal Microbiome Consortium (additional members), Strauss, J. F., Jefferson, K. K., & Buck, G. A. (2015). The truth about metagenomics: Quantifying and counteracting bias in 16S rRNA studies. BMC Microbiology, 15(1), 66. https://doi.org/10.1186/s12866-015- 0351-6 Callahan, B. J., McMurdie, P. J., Rosen, M. J., Han, A. W., Johnson, A. J. A., & Holmes, S. P. (2016). DADA2: High-resolution sample inference from Illumina amplicon data. Nature Methods, 13(7), 581–583. https://doi.org/10.1038/nmeth.3869 Cameron, S. A., & Sadd, B. M. (2020). Global trends in bumble bee health. Annual Review of Entomology, 65(1), 209–232. https://doi.org/10.1146/annurev-ento- 011118-111847 Chen, S., Yao, H., Han, J., Liu, C., Song, J., Shi, L., Zhu, Y., Ma, X., Gao, T., Pang, X., Luo, K., Li, Y., Li, X., Jia, X., Lin, Y., & Leon, C. (2010). Validation of the ITS2 region as a novel DNA barcode for identifying medicinal plant species. PLOS ONE, 5(1), e8613. https://doi.org/10.1371/journal.pone.0008613 Cornman, R. S., Otto, C. R. V., Iwanowicz, D., & Pettis, J. S. (2015). Taxonomic characterization of honey bee (Apis mellifera) pollen foraging based on non- overlapping paired-end sequencing of nuclear ribosomal loci. PLOS ONE, 10(12), e0145365. https://doi.org/10.1371/journal.pone.0145365 Durant, J. L., & Otto, C. R. V. (2019). Feeling the sting? Addressing land-use changes can mitigate bee declines. Land Use Policy, 87, 104005. https://doi.org/10.1016/j.landusepol.2019.05.024 Free, J. B. (1970). The flower constancy of bumblebees. Journal of Animal Ecology, 39(2), 395–402. JSTOR. https://doi.org/10.2307/2978 Gloor, G. B., Macklaim, J. M., Pawlowsky-Glahn, V., & Egozcue, J. J. (2017). Microbiome datasets are compositional: And this is not optional. Frontiers in Microbiology, 8. https://doi.org/10.3389/fmicb.2017.02224 Goodell, K., & Thomson, J. D. (1996, June). Comparisons of pollen removal and deposition by honey bees and bumblebees visiting apple. In VII International Symposium on Pollination 437 (pp. 103-108). Goulson, D., Nicholls, E., Botías, C., & Rotheray, E. L. (2015). Bee declines driven by combined stress from parasites, pesticides, and lack of flowers. Science, 347(6229). https://doi.org/10.1126/science.1255957 Harder, L. D. (1990). Pollen removal by bumble bees and its implications for pollen dispersal. Ecology, 71(3), 1110–1125. JSTOR. https://doi.org/10.2307/1937379 Janzen, D. H. (1983). Seed and pollen dispersal by animals: Convergence in the ecology of contamination and sloppy harvest. Biological Journal of the Linnean Society, 20(1), 103–113. https://doi.org/10.1111/j.1095-8312.1983.tb01592.x Kane, N., Sveinsson, S., Dempewolf, H., Yang, J. Y., Zhang, D., Engels, J. M. M., & Cronk, Q. (2012). Ultra-barcoding in cacao (Theobroma spp.; Malvaceae) using

whole chloroplast genomes and nuclear ribosomal DNA. American Journal of Botany, 99(2), 320–329. https://doi.org/10.3732/ajb.1100570 Keller, A., Danner, N., Grimmer, G., Ankenbrand, M., Ohe, K. von der, Ohe, W. von der, Rost, S., Härtel, S., & Steffan‐Dewenter, I. (2015). Evaluating multiplexed next- generation sequencing as a method in palynology for mixed pollen samples. Plant Biology, 17(2), 558–566. https://doi.org/10.1111/plb.12251 Kembel, S. W., Wu, M., Eisen, J. A., & Green, J. L. (2012). Incorporating 16S gene copy number information improves estimates of microbial diversity and abundance. PLOS Computational Biology, 8(10). https://doi.org/10.1371/journal.pcbi.1002743 Kraaijeveld, K., Weger, L. A. de, García, M. V., Buermans, H., Frank, J., Hiemstra, P. S., & Dunnen, J. T. den. (2015). Efficient and sensitive identification and quantification of airborne pollen using next-generation DNA sequencing. Molecular Ecology Resources, 15(1), 8–16. https://doi.org/10.1111/1755- 0998.12288 Latta, R. G., Linhart, Y. B., Fleck, D., & Elliot, M. (1998). Direct and indirect estimates of seed versus pollen movement within a population of ponderosa pine. Evolution, 52(1), 61–67. https://doi.org/10.1111/j.1558-5646.1998.tb05138.x Macgregor, C. J., Kitson, J. J. N., Fox, R., Hahn, C., Lunt, D. H., Pocock, M. J. O., & Evans, D. M. (2019). Construction, validation, and application of nocturnal pollen transport networks in an agro-ecosystem: A comparison using light microscopy and DNA metabarcoding. Ecological Entomology, 44(1), 17–29. https://doi.org/10.1111/een.12674 Miller, P. D. (1982). Maize pollen: collection and enzymology. In Sheridan WF (ed) Maize for Biological Research. University Press, University of North Dakota, Grand Forks, pp 279-293 Pacini, E. (2015). Pollination. In Reference Module in Earth Systems and Environmental Sciences. Elsevier. https://doi.org/10.1016/B978-0-12-409548-9.09315-5 Pawluczyk, M., Weiss, J., Links, M. G., Egaña Aranguren, M., Wilkinson, M. D., & Egea-Cortines, M. (2015). Quantitative evaluation of bias in PCR amplification and next-generation sequencing derived from metabarcoding samples. Analytical and Bioanalytical Chemistry, 407(7), 1841–1848. https://doi.org/10.1007/s00216- 014-8435-y Pompanon, F., Deagle, B. E., Symondson, W. O. C., Brown, D. S., Jarman, S. N., & Taberlet, P. (2012). Who is eating what: Diet assessment using next generation sequencing. Molecular Ecology, 21(8), 1931–1950. https://doi.org/10.1111/j.1365-294X.2011.05403.x Pornon, A., Escaravage, N., Burrus, M., Holota, H., Khimoun, A., Mariette, J., Pellizzari, C., Iribar, A., Etienne, R., Taberlet, P., Vidal, M., Winterton, P., Zinger, L., & Andalo, C. (2016). Using metabarcoding to reveal and quantify plant-pollinator interactions. Scientific Reports, 6(1), 27282. https://doi.org/10.1038/srep27282 Porter, E. K. (1981). Origins and genetic nonvariability of the proteins which diffuse from maize pollen. Environmental Health Perspectives, 37, 53-59.

Potter, C., Vere, N. de, Jones, L. E., Ford, C. R., Hegarty, M. J., Hodder, K. H., Diaz, A., & Franklin, E. L. (2019). Pollen metabarcoding reveals broad and species-specific resource use by urban bees. PeerJ, 7, e5999. https://doi.org/10.7717/peerj.5999 Rahl, M. (2008). Microscopic identification and purity determination of pollen grains. In M. G. Jones & P. Lympany (Eds.), Allergy Methods and Protocols (pp. 263–269). Humana Press. https://doi.org/10.1007/978-1-59745-366-0_22 Richardson, R. T., Lin, C.-H., Sponsler, D. B., Quijia, J. O., Goodell, K., & Johnson, R. M. (2015a). Application of ITS2 metabarcoding to determine the provenance of pollen collected by honey bees in an agroecosystem. Applications in Plant Sciences, 3(1), 1400066. https://doi.org/10.3732/apps.1400066 Richardson, R. T., Lin, C.-H., Quijia, J. O., Riusech, N. S., Goodell, K., & Johnson, R. M. (2015b). Rank-based characterization of pollen assemblages collected by honey bees using a multi-locus metabarcoding approach. Applications in Plant Sciences, 3(11), 1500043. https://doi.org/10.3732/apps.1500043 Richardson, R. T., Curtis, H. R., Matcham, E. G., Lin, C.-H., Suresh, S., Sponsler, D. B., Hearon, L. E., & Johnson, R. M. (2019). Quantitative multi-locus metabarcoding and waggle dance interpretation reveal honey bee spring foraging patterns in Midwest agroecosystems. Molecular Ecology, 28(3), 686–697. https://doi.org/10.1111/mec.14975 Roberts, R. J., Carneiro, M. O., & Schatz, M. C. (2013). The advantages of SMRT sequencing. Genome Biology, 14(6), 405. https://doi.org/10.1186/gb-2013-14-6- 405 Schrader, C., Schielke, A., Ellerbroek, L., & Johne, R. (2012). PCR inhibitors— Occurrence, properties and removal. Journal of Applied Microbiology, 113(5), 1014–1026. https://doi.org/10.1111/j.1365-2672.2012.05384.x Sickel, W., Ankenbrand, M. J., Grimmer, G., Holzschuh, A., Härtel, S., Lanzen, J., Steffan-Dewenter, I., & Keller, A. (2015). Increased efficiency in identifying mixed pollen samples by meta-barcoding with a dual-indexing approach. BMC Ecology, 15(1), 20. https://doi.org/10.1186/s12898-015-0051-y Smart, M. D., Cornman, R. S., Iwanowicz, D. D., McDermott-Kubeczko, M., Pettis, J. S., Spivak, M. S., & Otto, C. R. V. (2017). A comparison of honey bee-collected pollen from working agricultural lands using light microscopy and ITS metabarcoding. Environmental Entomology, 46(1), 38–49. https://doi.org/10.1093/ee/nvw159 Song, J., Shi, L., Li, D., Sun, Y., Niu, Y., Chen, Z., Luo, H., Pang, X., Sun, Z., Liu, C., Lv, A., Deng, Y., Larson-Rabin, Z., Wilkinson, M., & Chen, S. (2012). Extensive pyrosequencing reveals frequent intra-genomic variations of internal transcribed spacer regions of nuclear ribosomal DNA. PLOS ONE, 7(8). https://doi.org/10.1371/journal.pone.0043971 Taberlet, P., Coissac, E., Pompanon, F., Brochmann, C., & Willerslev, E. (2012). Towards next-generation biodiversity assessment using DNA metabarcoding. Molecular Ecology, 21(8), 2045–2050. https://doi.org/10.1111/j.1365- 294X.2012.05470.x

Tang, M., Tan, M., Meng, G., Yang, S., Su, X., Liu, S., Song, W., Li, Y., Wu, Q., Zhang, A., & Zhou, X. (2014). Multiplex sequencing of pooled mitochondrial genomes— A crucial step toward biodiversity analysis using mito-metagenomics. Nucleic Acids Research, 42(22), e166–e166. https://doi.org/10.1093/nar/gku917 Thomas, A. C., Deagle, B. E., Eveson, J. P., Harsch, C. H., & Trites, A. W. (2016). Quantitative DNA metabarcoding: Improved estimates of species proportional biomass using correction factors derived from control material. Molecular Ecology Resources, 16(3), 714–726. https://doi.org/10.1111/1755-0998.12490 Villanueva-Gutiérrez, R., & Roubik, D. W. (2016). More than protein? Bee–flower interactions and effects of disturbance regimes revealed by rare pollen in bee nests. Arthropod-Plant Interactions, 10(1), 9–20. https://doi.org/10.1007/s11829- 015-9413-9 Wang, X.-C., Liu, C., Huang, L., Bengtsson‐Palme, J., Chen, H., Zhang, J.-H., Cai, D., & Li, J.-Q. (2015). ITS1: A DNA barcode better than ITS2 in eukaryotes? Molecular Ecology Resources, 15(3), 573–586. https://doi.org/10.1111/1755- 0998.12325 Whitehead, D. R. (1983). Wind pollination: some ecological and evolutionary perspectives. Pollination biology, 97(08).

Figures and Tables

Figure 2.1: Distinct pollen types in pollen mixtures representing the proportion of each plant species’ contribution to a mixed pollen sample based on mass (i.e., “Pollen”) and the proportion of sequencing reads produced for each plant species for 3 replicates obtained using DNA metabarcoding (i.e., “SR 1-3”) when using the conservative threshold. See Figure A1.1 for liberal threshold results.

Figure 2.2: Proportion of pollen of each plant species in mixtures vs. proportion of sequence reads using the conservative threshold. Light gray line represents expected one- to-one relationship. See Figure A1.2 for liberal threshold results.

Table 2.1: Proportion by mass and proportion of sequencing reads for each plant species in mixtures and single-species samples using the liberal threshold. See Table A1.1 for more information on additional species. OA: Onopordum acanthium; SO: Sidalcea oregana; PG: Potentilla gracilis; TM: Thermopsis montana; VV: Vicia villosa.

Proportion by Mass Proportion of Sequencing Reads OA SO PG TM VV OA SO PG TM VV Additional Species Mixture 1-1 0.47 0.28 0.25 0 0 0.5775 0.0664 0.3516 0.0003 0 0.0042 Mixture 1-2 0.47 0.28 0.25 0 0 0.5715 0.0815 0.3416 0.0006 0 0.0048 Mixture 1-3 0.47 0.28 0.25 0 0 0.5373 0.1197 0.3362 0.0009 0 0.0059 Mixture 2-1 0.4 0.2 0.2 0.2 0 0.0694 0.005 0.0565 0.8679 0 0.0011 Mixture 2-2 0.4 0.2 0.2 0.2 0 0.0891 0.0068 0.0567 0.8466 0 0.0009 Mixture 2-3 0.4 0.2 0.2 0.2 0 0.0709 0.0061 0.0538 0.8678 0 0.0014 Mixture 3-1 0.2 0.2 0.2 0.2 0.2 0.06 0.0057 0.031 0.7824 0.1198 0.001 Mixture 3-2 0.2 0.2 0.2 0.2 0.2 0.0335 0.0028 0.0229 0.8734 0.0662 0.0011 Mixture 3-3 0.2 0.2 0.2 0.2 0.2 0.0399 0.0068 0.0245 0.8685 0.0594 0.0009 Mixture 4-1 0.6 0.1 0.1 0.1 0.1 0.2111 0.0013 0.0527 0.6124 0.1214 0.0012 Mixture 4-2 0.6 0.1 0.1 0.1 0.1 0.2616 0.0014 0.0448 0.5836 0.1083 0.0003 Mixture 4-3 0.6 0.1 0.1 0.1 0.1 0.3102 0.0018 0.0359 0.5685 0.082 0.0016 Mixture 5-1 0.81 0.1 0.03 0.03 0.03 0.6727 0.003 0.0179 0.2836 0.0225 0.0004 Mixture 5-2 0.81 0.1 0.03 0.03 0.03 0.6924 0.0073 0.0129 0.2643 0.022 0.0011 Mixture 5-3 0.81 0.1 0.03 0.03 0.03 0.5302 0.0075 0.0061 0.4454 0.0098 0.0011 OA-1 1 0 0 0 0 0.9995 0 0 0.0005 0 0 OA-2 1 0 0 0 0 0.9988 0 0 0.0004 0 0.0008 OA-3 1 0 0 0 0 0.9983 0 0 0.0006 0.0003 0.0008 SO-1 0 1 0 0 0 0.0006 0.9683 0.0156 0.0015 0 0.014 SO-2 0 1 0 0 0 0.0021 0.9593 0.0203 0.0038 0 0.0144 SO-3 0 1 0 0 0 0.0005 0.9678 0.0132 0.0031 0.0003 0.0151 PG-1 0 0 1 0 0 0.0004 0.0006 0.9948 0.0008 0 0.0034 PG-2 0 0 1 0 0 0.0003 0 0.9943 0.001 0.0001 0.0042

PG-3 0 0 1 0 0 0.0003 0 0.9945 0.0005 0 0.0047 TM-1 0 0 0 1 0 0.0002 0 0 0.9989 0 0.0009 TM-2 0 0 0 1 0 0.0007 0.0003 0.0003 0.9987 0 0 TM-3 0 0 0 1 0 0.0005 0 0 0.9989 0 0.0006 VV-1 0 0 0 0 1 0.0003 0 0 0.0015 0.9972 0.001 VV-2 0 0 0 0 1 0.0003 0.0003 0 0.0007 0.9986 0.0001 VV-3 0 0 0 0 1 0.0004 0 0 0.0004 0.9989 0.0003

Table 2.2: Number of sequencing reads detected for each species before and after applying the conservative threshold. Values shaded in pink represent false negatives. OA: Onopordum acanthium; SO: Sidalcea oregana; PG: Potentilla gracilis; TM: Thermopsis montana; VV: Vicia villosa.

Number of sequencing reads Adjusted number of sequencing reads OA SO PG TM VV OA SO PG TM VV Mixture 1-1 25247 2904 15372 11 0 25247 2904 15372 0 0 Mixture 1-2 20162 2874 12052 20 0 20162 2874 12052 0 0 Mixture 1-3 31356 6983 19619 53 0 31356 6983 19619 0 0 Mixture 2-1 3528 256 2870 44093 0 3528 0 2870 44093 0 Mixture 2-2 4660 354 2968 44294 0 4660 0 2968 44294 0 Mixture 2-3 3629 310 2753 44387 0 3629 0 2753 44387 0 Mixture 3-1 2615 250 1352 34081 5219 2615 0 1352 34081 5219 Mixture 3-2 1575 130 1077 41003 3108 1575 0 1077 41003 3108 Mixture 3-3 1527 259 938 33251 2274 1527 0 938 33251 2274 Mixture 4-1 10119 60 2524 29357 5818 10119 0 2524 29357 5818 Mixture 4-2 9349 51 1600 20857 3872 9349 0 1600 20857 3872 Mixture 4-3 10822 64 1251 19836 2861 10822 0 1251 19836 2861 Mixture 5-1 21290 94 565 8976 712 21290 0 565 8976 712 Mixture 5-2 22933 243 426 8755 728 22933 0 426 8755 728 Mixture 5-3 13456 190 154 11304 248 13456 0 0 11304 0 OA-1 40563 0 0 19 0 40563 0 0 0 0 OA-2 35895 0 0 14 0 35895 0 0 0 0 OA-3 38116 0 0 21 13 38116 0 0 0 0 SO-1 15 25330 408 38 0 0 25330 408 0 0 SO-2 47 20982 444 84 0 0 20982 444 0 0 SO-3 9 18178 247 59 5 0 18178 0 0 0 PG-1 12 18 30314 25 0 0 0 30314 0 0 PG-2 11 0 31458 33 4 0 0 31458 0 0

PG-3 15 0 45021 21 0 0 0 45021 0 0 TM-1 10 0 0 40925 0 0 0 0 40925 0 TM-2 25 10 10 33754 0 0 0 0 33754 0 TM-3 13 0 0 28010 0 0 0 0 28010 0 VV-1 7 0 0 43 27775 0 0 0 0 27775 VV-2 12 0 0 28 38379 0 0 0 0 38379 VV-3 14 0 0 12 31814 0 0 0 0 31814

Table 2.3: Proportion by mass and proportion of sequencing reads for each plant species in mixtures and single-species samples using the conservative threshold. Values shaded in pink represent false negatives. OA: Onopordum acanthium; SO: Sidalcea oregana; PG: Potentilla gracilis; TM: Thermopsis montana; VV: Vicia villosa.

Proportion by Mass Proportion of Sequencing Reads OA SO PG TM VV OA SO PG TM VV Mixture 1-1 0.47 0.28 0.25 0 0 0.5801 0.0667 0.3532 0 0 Mixture 1-2 0.47 0.28 0.25 0 0 0.5746 0.0819 0.3435 0 0 Mixture 1-3 0.47 0.28 0.25 0 0 0.541 0.1205 0.3385 0 0 Mixture 2-1 0.4 0.2 0.2 0.2 0 0.0699 0 0.0568 0.8733 0 Mixture 2-2 0.4 0.2 0.2 0.2 0 0.0898 0 0.0572 0.8531 0 Mixture 2-3 0.4 0.2 0.2 0.2 0 0.0715 0 0.0542 0.8743 0 Mixture 3-1 0.2 0.2 0.2 0.2 0.2 0.0604 0 0.0312 0.7877 0.1206 Mixture 3-2 0.2 0.2 0.2 0.2 0.2 0.0337 0 0.023 0.8768 0.0665 Mixture 3-3 0.2 0.2 0.2 0.2 0.2 0.0402 0 0.0247 0.8753 0.0599 Mixture 4-1 0.6 0.1 0.1 0.1 0.1 0.2116 0 0.0528 0.6139 0.1217 Mixture 4-2 0.6 0.1 0.1 0.1 0.1 0.262 0 0.0448 0.5846 0.1085 Mixture 4-3 0.6 0.1 0.1 0.1 0.1 0.3112 0 0.036 0.5705 0.0823 Mixture 5-1 0.81 0.1 0.03 0.03 0.03 0.675 0 0.0179 0.2846 0.0226 Mixture 5-2 0.81 0.1 0.03 0.03 0.03 0.6983 0 0.013 0.2666 0.0222 Mixture 5-3 0.81 0.1 0.03 0.03 0.03 0.5435 0 0 0.4565 0 OA-1 1 0 0 0 0 1 0 0 0 0 OA-2 1 0 0 0 0 1 0 0 0 0 OA-3 1 0 0 0 0 1 0 0 0 0 SO-1 0 1 0 0 0 0 0.9841 0.0159 0 0 SO-2 0 1 0 0 0 0 0.9793 0.0207 0 0 SO-3 0 1 0 0 0 0 1 0 0 0 PG-1 0 0 1 0 0 0 0 1 0 0 PG-2 0 0 1 0 0 0 0 1 0 0

PG-3 0 0 1 0 0 0 0 1 0 0 TM-1 0 0 0 1 0 0 0 0 1 0 TM-2 0 0 0 1 0 0 0 0 1 0 TM-3 0 0 0 1 0 0 0 0 1 0 VV-1 0 0 0 0 1 0 0 0 0 1 VV-2 0 0 0 0 1 0 0 0 0 1 VV-3 0 0 0 0 1 0 0 0 0 1

Table 2.4: Summary of linear regression examining the relationship between proportion of pollen by mass and proportion of sequencing reads for the plant species in laboratory- prepared mixtures after using the liberal and conservative thresholds. A linear regression was not conducted for mixture 3 because no variation existed in the proportion of pollen by mass for the sample.

R Square p-value Liberal Conservative Liberal Conservative Mixture 1 0.79 0.79 <0.0001 <0.0001 Mixture 2 0.01 0.01 0.78 0.78 Mixture 3 N/A N/A N/A N/A Mixture 4 0.02 0.02 0.61 0.61 Mixture 5 0.69 0.69 <0.001 <0.001

Table 2.5: Summary of linear regression examining the relationship between the proportion of pollen by mass and proportion of sequencing reads for each plant species across all mixtures when using the liberal and conservative thresholds.

R square p-value

Liberal Conservative Liberal Conservative

Onopordum acanthium 0.59 0.59 <0.001 <0.001

Sidalcea oregana 0.58 0.56 0.001 0.001

Potentilla gracilis 0.47 0.47 0.005 0.005

Thermopsis montana 0.93 0.92 <0.0001 <0.0001

Vicia villosa 0.62 0.62 <0.001 <0.001

Chapter 3: Capabilities and Limitations of Using DNA Metabarcoding to Study Plant-Pollinator Interactions

Katherine A. Arstingstall, Sandra J. DeBano, Xiaoping Li, David E. Wooster, Mary M. Rowland, Skyler Burrows, and Kenneth Frost

Abstract Many pollinator populations are experiencing severe declines, emphasizing the need for a better understanding of the complex relationship between bees and flowering plants. Using DNA metabarcoding to describe plant pollinator interactions eliminates some challenges associated with traditional methods for doing so and has the potential to reveal a more detailed record of bee foraging behavior. Here we use DNA metabarcoding of the ITS2 region and rbcL gene to identify plant species present in pollen loads of 403 bees from three habitat types in eastern Oregon. Our specific objectives were to 1) determine whether plant species in pollen loads identified using DNA metabarcoding data are consistent with plant species identified by direct observations of bee foraging and 2) compare plant species assignments produced by DNA metabarcoding when using a “regional” reference database to those produced using a “local” database. Of plant species that bees were observed visiting, 34-98% were detected using DNA metabarcoding. Plant-pollinator networks produced using data derived from DNA metabarcoding had significantly higher connectance, linkage density and bee generality and significantly lower specialization when compared to networks based on bee foraging observations. Approximately 15% more plant species were assigned when using the regional database than when using the local database. Using a reference database that only included sequence information for plant species present in the study area reduced the possibility of erroneous taxonomic assignments. Here, we examine some strengths and limitations of using DNA metabarcoding to identify plant species present in bee pollen loads and provide guidance for future research.

Introduction Pollination is a crucial ecosystem service that aids in food production and sustains ecosystem functioning (Buchmann & Nabhan 1996; Losey & Vaughan 2006; Ollerton et al., 2011). With rapid and widespread declines of bee populations, there is a clear and urgent need for a greater understanding of the complex interactions occurring between bees and the plant species they use (Winfree et al., 2009; Potts et al., 2010; Koh et al., 2016). Network analyses and diagrams are important conservation tools often used to visualize and quantify the strength of plant-pollinator interactions and to identify major food sources of the pollinator species involved (Kearns & Inouye, 1993; Bosch et al., 2009; Cusser & Goodell 2013). Traditional methods for documenting bee-flower relationships include direct observation of foraging behavior and indirect examination by collecting pollen loads from bees and using microscopy to separate and identify plant species based on morphological characteristics of pollen grains (Erdtman, 1943). There are challenges and limitations associated with each of these methods. Direct observation is time consuming and only reveals a snapshot of bee foraging behavior. The difficulty of following a foraging bee from one flower to another often limits the observer to viewing one or two flower visits per bee and has led to overestimates of floral fidelity by individual bees (Free, 1970). Plant species identification by microscopy of pollen is time-consuming, requires specialized expertise, and can result in low taxonomic resolution (Rahl, 2008; Cornman et al., 2015). In addition, typically only a subsample of an insect’s pollen load is analyzed using this technique, which can result in a lack of detection of pollen from plant species present in the load in low abundance (Von Der Ohe et al., 2004; Whittington et al., 2004; Sajwani et al., 2014). DNA metabarcoding of pollen collected from foraging bees is a promising tool for identifying plant species that are present in pollen loads and investigating plant- pollinator networks (Cornman et al., 2015; Keller et al., 2015; Kraaijeveld et al., 2015; Richardson et al., 2015a; Bell et al., 2017; Smart et al., 2017). DNA metabarcoding combines species identification using DNA barcoding with next generation sequencing technology. In pollen DNA metabarcoding methods, DNA is extracted from pollen

samples containing one or more plant species and barcode regions with low intraspecific and high interspecific variation are amplified and sequenced using next generation sequencing. The species composition of a pollen sample is then determined based on the presence of DNA sequencing “barcodes” in the pollen samples and their similarity to sequences of known reference plant species. (Hebert et al., 2003; Sickel et al., 2015). Although traditional methods are still used more frequently to describe plant- pollinator networks than DNA metabarcoding (e.g., Wood et al., 2015; Morandin & Kremen, 2013; Roof et al., 2018; Dibble et al., 2020a,b), several studies have compared the taxonomic resolution of DNA metabarcoding with that of microscopy and found that DNA metabarcoding resulted in equal or higher taxa richness (Keller et al., 2015; Richardson et al., 2015a,b; Smart et al., 2017). Yet very few studies have compared observations of bee foraging with DNA metabarcoding data, despite the fact that studies involving bee foraging observation are becoming more common because of the rising popularity of conducting citizen science (Domroese & Johnson, 2017). The few studies that have compared these approaches either did not examine bees exclusively (Pornon et al., 2016; Galliot et al., 2017) or lacked taxonomic resolution of plants and pollinators (Potter et al., 2019). Examining the consistency, or lack of consistency, between plant- pollinator networks derived from bee foraging observations and DNA metabarcoding of bee pollen loads is important for determining whether observations alone can sufficiently identify major food sources used by bees, or whether they tend to oversimplify plant- pollinator networks. There are challenges associated with pollen DNA metabarcoding, including the issue of correctly identifying plant taxa to the species-level (Keller et al., 2015; Kraaijeveld et al., 2015; Richardson et al., 2015a). While DNA metabarcoding increases the number of taxa assigned at the species level compared to traditional methods, species assignments obtained through DNA metabarcoding can be prone to sequencing and database errors, resulting in erroneous species identifications (Cornman et al., 2015; Keller et al., 2015; Smart et al., 2017). The accuracy of plant species assignments is also highly dependent on the reference database. Reference databases are constructed by acquiring DNA sequences of known plant species from curated databases or the National

Center for Biotechnology Information (NCBI) nucleotide database (Sickel et al., 2015). Researchers often include all plant DNA sequences in a database for a specific barcode marker (Cornman et al., 2015; Keller et al.; 2015; Sickel et al., 2015; Galliot et al., 2017) or from a geographic location consistent with their area of study, which is often as large as an entire state or country (Richardson et al., 2015b; De Vere et al., 2017; Potter et al., 2019). This ensures that all possible plant species used as forage resources are included in the database, but it can also increase the potential for false detections associated with misidentification. Species and genera of certain plant families (e.g., Asteraceae) share a significant portion of their DNA (sometimes more than 99%), making differentiation of these species difficult (Gao et al., 2010). Currently no work has been conducted to document the potential benefits of using reference databases that are limited to plant species that occur in the location of interest (i.e., a “local” reference database) versus a reference database that includes all plant species known to occur in a larger region (i.e., a “regional” reference database). One potential benefit of using a local reference database is reducing misidentification of plant species that can occur when using a regional reference database. This benefit may outweigh the extra time and effort required to conduct plant surveys of an area before using plant species assignments derived from DNA metabarcoding of pollen to study the plant-pollinator interactions occurring there. Here, we examine the accuracy of plant species assignments using DNA metabarcoding of pollen in the context of a study aimed at examining native bee-plant associations in eastern Oregon. The specific objectives of this study were to 1) determine whether plant-pollinator networks based on plant species identifications from DNA metabarcoding of pollen are consistent with plant-pollinator networks based on visual observations of bee foraging behavior, and 2) compare plant species assignments obtained using a reference database that includes sequence data of plants present in the region (i.e., a regional database) to species assignments obtained using a reference database that only includes sequence data for plants present at a specific location (i.e., a local database).

Methods Study Sites The field component of this study was conducted in June, July and August of 2018 at three locations in eastern Oregon, US: Threemile Canyon Farms (45.7513° N, 119.9376° W), the United States Forest Service (USFS) Starkey Experimental Forest and Range (45.2332° N, 118.5511° W), and The Nature Conservancy’s (TNC) Zumwalt Prairie Preserve (45.5559° N, 116.9587° W). These locations were chosen with the goal of examining plant-bee associations in different land use types (i.e., agroecosystem, riparian forest, and grassland, respectively). The three study sites occur within 350 km of each other. Threemile Canyon Farms (Threemile) is a 37,636 ha industrial, center pivot- irrigated farm located in Morrow County (elevation 100-300 m). Threemile has 9,308 ha in conservation area, 15,985 ha of irrigated conventionally managed cropland, and 6,178 ha of irrigated organically managed cropland. Uncultivated habitat includes the conservation area, which consists of arid grassland and shrub-steppe, as well as field margins, which are dominated by non-native vegetation. The Starkey Experimental Forest and Range (Starkey) is located in Union County (elevation 1,130-1,500 m); this long-term research site established in 1940 is grazed by cattle under a standard USFS grazing permit, and also supports herds of deer (Odocoileus spp.) and elk (Cervus canadensis) (Rowland et al., 1997). Sampling at Starkey occurred on Meadow Creek, a major tributary of the upper Grande Ronde River that flows through Starkey. Dominant riparian shrubs include willow (Salix spp.), black hawthorn (Crataegus douglasii Lindl.), thinleaf alder (Alnus incana [L.] Moench ssp. tenuifolia [Nutt.] Breitung), black cottonwood (Populus balsamifera L. ssp. trichocarpa [Torr. & A. Gray ex Hook.] Brayshaw), and common snowberry (Symphoricarpos albus [L.] S.F. Blake). Scattered ponderosa pine (Pinus ponderosa Lawson & C. Lawson), Douglas-fir (Pseudotsuga menziesii [Mirb.] Franco), and western larch (Larix occidentalis Nutt.) are also found in the riparian corridor (Averett, Endress, Rowland, Naylor, & Wisdom, 2017). Herbaceous vegetation includes >200 species of forbs (Roof et al., 2018).

TNC’s Zumwalt Prairie Preserve (Zumwalt) is a 13,269 ha remnant bunchgrass prairie in Wallowa County (elevation 1,100-1,700 m). The Zumwalt Prairie has been used as summer pasture for horse, sheep, and cattle for over 100 years, but the majority of the area remains dominated by native plant species including Idaho fescue (Festuca idahoensis Elmer), prairie Junegrass (Koeleria macrantha [Ledeb.] Schult.), and bluebunch wheatgrass (Pseudoroegneria spicata [Pursh] Á. Löve) (Kennedy et al., 2009) and a rich forb community (>112 species of forbs) (Kimoto et al., 2012).

Bee Sampling Bees were sampled during peak foraging hours (0900-1800) once a month from each site during three bouts (13-28 June, 3-25 July, 6-31 August). Each bee was caught directly in a glass vial, if possible, or with an insect net and placed in an individual glass vial. All vials and nets were placed in 10% bleach for at least one minute and dried prior to sampling. Nets were replaced with clean, bleached nets after each use to prevent pollen contamination among samples. No killing agent was used in the vials, and bees were frozen at the end of each field day. Vials were labeled with the time, date, location, and flower species that the bee was foraging on when it was collected. The flower species noted for each bee was considered the bee foraging observation for that specimen. Each bee was given a unique sample ID so that it could later be associated with its specific pollen load. Bees were pinned, labeled, sexed and identified to the lowest taxonomic level possible, usually species, using the methods described in Kuhlman and Burrows (2017).

Pollen Isolation Bees were washed in the glass vial in which they were collected in the field. Sterilized water was added to the vial, and the vial was shaken vigorously until all visible pollen was removed from the bee. The pollen solution was pipetted from the vial and transferred to a 50 mL centrifuge tube. If pollen was still visible on the bee or the walls of the vial, it was rinsed with additional sterilized water, shaken vigorously, and the solution was pipetted from the vial and transferred to the same 50 mL centrifuge tube. The bees

were then stored in a freezer until they could be pinned and identified. Each tube containing an individual bee’s pollen load was centrifuged at 2,000 rpm for 2 min, and the supernatant was discarded. The pollen pellet was resuspended in 1 mL of water, and the solution was transferred to a 1.5 mL screw cap microcentrifuge tube. The screw cap tubes were centrifuged at 12,500 rpm for 30 s and the supernatant was discarded. The pollen pellet was washed in 1 mL of 100% ethanol, and the supernatant was discarded. The pollen pellet was then dried for 30 min in an Eppendorf vacufuge. Dried pollen pellets were stored at -20°C until further processing.

DNA Extraction and PCR DNA extractions took place during October and November 2018 using the Macherey-Nagel Nucleospin Food kit (Macherey-Nagel, Bethlehem, Pennsylvania, USA). We followed the “isolation of genomic DNA from honey or pollen” supplementary protocol. We added 1 mm glass beads to each 1.5 mL screw cap microcentrifuge tube and homogenized the samples in a Mini-BeadBeater-24 (Biospec Products Bartlesville, Oklahoma, USA). We included negative controls with each round of DNA extraction using sterilized water instead of pollen. We used the dual-indexing pollen DNA metabarcoding strategy of Sickel et al. (2015). We used the second internal transcribed spacer (ITS2) primers ITS S2F and ITS4R used by Sickel et al. (2015) and existing universal ribulose-1,5-bisphosphate carboxylase/oxygenase large subunit (rbcL) primers, rbcL2 (Palmieri et al., 2009) and rbcLaR (Kress & Erickson, 2007). Illumina overhang adapter sequences were added to each of the primers (Klindworth et al., 2013). The polymerase chain reaction (PCR) was conducted as a multiplex, targeting both sequence regions for amplification in one reaction. The PCR reactions contained 10 μL of 5X Green GoTaq® Reaction Buffer, 0.3 uL of 10 mM dNTPs, 1 uL of each primer, 0.2 uL of GoTaq® DNA Polymerase, 33.5 uL of water, and 2 uL of the template DNA for a total volume of 50 uL per reaction. PCR cycles included an initial period of heat activation for 3 min at 95°C. This was followed by 35 cycles of 30 s at 95°C, 30 s at 55°C, and 1 min at 72°C. There was a final extension of 10 min at 72°C and then the samples were held at 10°C until further processing. A

negative control was included in each round of PCR, where 2 uL of water were used instead of template DNA. A 5 uL sample of each PCR product was electrophoresed in a 2% agarose gel, stained with GelRedTM (Biotium Inc, Fremont, CA), and visualized under UV light to confirm the presence of the appropriately-sized amplicons. The PCR products were purified using the Promega Wizard SV Gel and PCR Clean-Up System. DNA was quantified for each sample using a Nanodrop 2000 Spectrophotometer (Thermo Scientific), and the samples were diluted with sterilized water in 96-well plates. The samples were indexed, pooled, and sequenced at the Center for Genome Research and Biocomputing at Oregon State University in a standard flow cell v3 300bp paired-end full service run of the Illumina MiSeq instrument.

Bioinformatics- Read Quality Filtering and Denoising A total of 41,152,294 and 735,102 raw reads were retrieved across 558 samples for ITS2 and rbcL, respectively. The open-source QIIME2 pipeline (version 2019-04) (Bolyen et al. 2019) was used to analyze species composition of pollen loads (Figure 3.1). Initially, the QIIME2 DADA2 plugin (Callahan et al. 2016) was used to filter read quality and denoise reads. Sequence reads with a median sequencing quality score below 22 and noisy reads were filtered out. Contigs were created by joining paired-end reads, and duplicate sequences and sequences with chimera were removed. The resulting feature table containing optimized sequences and amplicon sequence variants (ASV) was used for additional analyses (e.g., taxonomic classification). After filtering and denoising, approximately 19,105,591 and 452,283 paired-end reads were retained for ITS2 and rbcL, respectively.

Bioinformatics- Reference Database Construction Lists of all plant species known to occur at the three study sites have been developed over multiple years by botanists conducting field work at each site. A regional plant list was developed that included the unique species of each study site. Reference sequences of the ITS2 region and rbcL gene were extracted from publicly available data of Sickel et al. (2015) and Bell et al. (2017) (ITS2 download:

https://www.biozentrum.uni-wuerzburg.de/dna-analytics/molecular-biodiversity- group/downloads/ and rbcL download: https://figshare.com/collections/rbcL_reference_library/3466311 ). A python package, NCBI-Companion (https://github.com/lixiaopi1985/NCBI_Companion), was used to query the NCBI nucleotide database to obtain sequences of species that were not present in the data of Sickel et al. (2015) and Bell et al. (2017). ITSx (Bengtsson-Palme et al. 2013) and MetaCurator software (Richardson et al., 2020) were used to detect and extract ITS2 and rbcL regions in the NCBI queried reference sequences, respectively. The final “regional” ITS2 and rbcL database was then split into three smaller “local” databases for ITS2 and rbcL that contained only species known to occur at each individual site. From the master plant list with all unique plant species, the regional database contained 79% and 85% of the total species for ITS2 and rbcL respectively. The local Threemile, Starkey, and Zumwalt ITS2 databases contained 63%, 77%, and 82% of the total species at each study site respectively. The local Threemile, Starkey, and Zumwalt rbcL databases contained 57%, 68%, and 80% of the total species at each study site respectively.

Bioinformatics- Training Classifier The Naïve Bayes algorithm provided by the QIIME2 feature-classifier plugin (Bokulich et al. 2018) was used to train a classifier for taxonomic assignment based on the ITS2 and rbcL databases. Initially, the default settings were applied (i.e., kmer length = 7, confidence threshold = 0.7) with high accuracy and low recall. In one case (Zumwalt rbcL paired-end), taxonomic assignment of the classifier resulted in most of the samples being unassigned, so the kmer length was increased and confidence threshold reduced (high recall kmer = 32; confidence threshold = 0.6) to increase taxonomic assignments.

Bioinformatics- Taxonomy Classification The classify-sklearn tool of the QIIME2 feature-classifier plugin (Bokulich et al. 2018) was used to assign taxonomy to each ASV. The feature table outputted by DADA2 was also filtered to ensure that the abundance of the feature was larger than 1000 and

could be found in more than four samples. We then applied the QIIME2 taxa plugin to visualize the relative abundance in each sample in barplots. In addition to analyzing the regional data, we separated the raw reads for ITS2 and rbcL by study sites. For each site, the reads were subjected to the pipeline above (DADA2) to produce an ASV table and representative sequences with adjusted truncating parameters to maintain appropriate overlapping length for reads pairing (for ITS: Zumwalt –p-trunc-len-f 288 –p-trunc-len-r 237; Starkey –p-trunc-len-f 299 –p-trunc-len-r 244; Threemile –p-trunc-len-f 299 –p-trunc-len-r 243; for rbcL paired end: Zumwalt –p- trunc-len-f 299 –p-trunc-len-r 243; Starkey –p-trunc-len-f 299 –p-trunc-len-r 268; Threemile –p-trunc-len-f 299 –p-trunc-len-r 243). Abundance of the features was set at 1,000 and could be found in no less than four samples. Corresponding local databases were used to assign taxonomy annotation to the features for each study site.

Sequence Count Removal Threshold Using the average and standard deviation of the number of sequencing reads in all negative control samples, we added 1.645 standard deviations to the average and used this number as our sequence count removal threshold (Armbruster & Pry, 2008). This allowed us to account for 95% of possible “background noise” detected in the pollen samples. We set the sequence count to zero for any taxonomic assignment whose read count fell below the sequence count removal threshold.

Network Analyses Bipartite plant-pollinator networks were created, and network statistics were calculated using the bipartite package in R (Dormann, Gruber, & Fruend, 2008). Plant- pollinator networks were based on bee foraging observations and plant species assignments derived from DNA metabarcoding data using reference libraries of regional (MB-RDB) and local (MB-LDB) plant species. For each location, data from all sampling periods were combined into one network for each detection method (i.e., bee foraging observations, MB-RDB, and MB-LDB). Analysis of variance (ANOVA) was used to compare four parameters of pollinator networks: connectance (the realized proportion of

possible links), H2’ (a frequency based index that increases with greater ecological specialization), linkage density (the mean number of links per species within the network), and bee generality (the mean number of plant species visited by each pollinator species within the network) (Gresty et al., 2018; Delmas et al. 2019; Macgregor et al., 2019). Network parameter values were scaled to the highest value for a given parameter at a given site prior to analysis to account for differences in location. Tukey’s honest significant difference test (HSD) was used to compare means among detection methods. Analyses were conducted using R, version 4.0.0 (R Core Team, 2020).

Results Comparing Plant-Pollinator Networks Created from Bee Foraging Observations and Plant Species Assignments Obtained Using DNA Metabarcoding

Each sampling location was characterized by a distinctive group of bees interacting with varied plant communities. At Threemile Canyon Farms, the 139 bees used in the network analyses included 14 genera and 22 species. Anthophora curta was the most abundant species, making up 24% of bees sampled, and was observed visiting four plant species. Other common bee species included Agapostemon femoratus, Melissodes bimatris, and Melissodes pallidisignatus (Table A2.1). Pollen loads of bees sampled from Threemile contained an average of 1.6 plant species using MB-RDB and an average of 2.1 plant species using MB-LDB. At Starkey, the 168 bees used in the network analyses included 12 genera and 39 species. Bombus bifarius was the most commonly sampled species, making up 21% of bees sampled, and was observed visiting 12 plant species. Other common species included Bombus flavifrons, Halictus ligatus, and Bombus californicus (Table A2.1). Pollen loads of bees sampled from Starkey contained an average of 2.0 and 2.4 plant species identified using MB-RDB and MB-LDB respectively. At Zumwalt, the 96 bees used in the network analyses included 9 genera and 35 species. Bombus californicus and Bombus centralis were the most commonly sampled species, making up 17% and 13% of bees sampled respectively (Table A2.1). B. californicus was observed visiting 8 plant species, and B. centralis was observed visiting

6 plant species. Pollen loads of bees sampled from Zumwalt contained an average of 1.7 plant species identified using MB-RDB and MB-LDB. When using taxonomic assignments from ITS2 and rbcL combined, MB-RDB and MB-LDB detected and assigned more plant species than bee foraging observations at all locations (Table 3.1). Plant-pollinator networks created from MB-RDB and MB-LDB were more complex than those created from bee foraging observations (Figures 3.2, 3.3, & 3.4). Mean connectance, linkage density, bee generality and H2’ of networks produced by the three detection methods were statistically different. When compared to networks based on bee foraging observations and networks produced from MB-RDB, networks produced from MB-LDB had significantly higher connectance (Figure 3.5a; F = 14.6, p = 0.005). Networks produced from MB-RDB and MB-LDB had significantly higher linkage density (Figure 3.5b; F = 15.5, p = 0.004), and bee generality (Figure 3.5c; F = 14.6, p = 0.005) than networks based on bee foraging observations. Specialization (H2’) was significantly higher for networks produced from bee foraging observations compared to those produced by MB-RDB and MB-LDB (Figure 3.5d; F = 16.3, p = 0.004). The mean number of plant species per pollen load identified through observation was significantly lower than the mean number of plant species per pollen load identified by MB-RDB and MB-LDB (Figure 3.6; F = 16.3, p = 0.004).

Comparing Plant Species Assignments Obtained Using MB-RDB and MB-LDB Of the ten plant species detected in the greatest number of pollen loads using each detection method (Table 3.2), MB-RDB and MB-LDB assigned 60%, 50% and 90% of the same species at Threemile, Starkey, and Zumwalt respectively. At all locations, MB- RDB and MB-LDB assigned 78.6% of plant families and 61.2% of plant genera that bees were visiting when sampled≥ (Table 3.3). At Threemile and≥ Starkey, we found that using the local database to assign taxonomy resulted in a higher percentage of detections of the plant species that bees were visiting when sampled relative to the regional database. At the Zumwalt sites, lower percentages of the plant species that bees were visiting when sampled were detected using MB-LDB relative to the other two sites (Table 3.3).

In this study, we define a regional mismatch as a plant species that was identified using MB-RDB but is not known to occur in the local area of interest. We determined that 20%, 10%, and 9% of the plant species identified using MB-RDB at Threemile, Starkey, and Zumwalt respectively were regional mismatches (Table 3.4). The use of the local database eliminated the potential for regional mismatches.

Inconsistencies Among Bee Foraging Observations and DNA Metabarcoding Data At Threemile, three of the nine plant species that bees were observed visiting made up 81% of observations: diffuse knapweed (Centaurea diffusa Lam.) (41%), hoary tansyaster (Machaeranthera canescens [Pursh] A. Gray ssp. canescens var. canescens) (24%), and yellow star-thistle (Centaurea solstitialis L.) (16%) (Table 3.2). While both metabarcoding approaches detected yellow star-thistle in a large proportion of pollen loads, diffuse knapweed was not detected by MB-RDB, and hoary tansyaster was not detected by MB-LDB (Table A2.2). Of the 26 plant species that bees were observed visiting at Starkey, only three made up 51% of observations: slender cinquefoil (Potentilla gracilis Douglas ex Hook.) (22%), Missouri goldenrod (Solidago missouriensis Nutt.) (20%), and mountain monardella (Monardella odoratissima Benth.) (9%) (Table 3.2). Slender cinquefoil was not detected above the sequence count removal threshold using MB-RDB. Mountain monardella was detected with the rbcL gene using both MB-RDB and MB-LDB. However, it was not included in the ten plant species that were most commonly detected in the pollen samples for MB-RDB or MB-LDB (Table 3.2). Furthermore, Missouri goldenrod was not detected using MB-RDB (Table A2.2). At Starkey, seven plant species that bees were observed visiting were not identified using plant species assignments derived from DNA metabarcoding data. Of these, only one species (largeflower triteleia [Triteleia grandiflora Lindl.]) was missing ITS2 sequence data from the reference library and therefore pollen sequences could not possibly be assigned based on the ITS2 region. Of the 23 plant species that bees were observed visiting at Zumwalt, four made up 40.6% of behavioral observations: silky lupine (Lupinus sericeus Pursh) (10.4%), whitestem frasera (Frasera albicaulis Douglas ex Griseb.) (10.4%), shaggy fleabane

(Erigeron pumilus Nutt.) (10.4%), and slender cinquefoil (9.4%) (Table 3.2). While whitestem frasera was identified by both the ITS2 region and rbcL gene using the regional and local reference libraries, slender cinquefoil, shaggy fleabane, and silky lupine were not detected using either MB-RDB or MB-LDB (Table A2.2). At Zumwalt, 14 plant species that bees were observed visiting were not detected by DNA metabarcoding. Of these, three species were missing ITS2 and rbcL sequence data from the reference library, making species assignment impossible (Table A2.2).

Discussion Bees contribute to global food security and are vital to sustaining functioning ecosystems and biodiversity. Considering recent population declines of many bee populations, a better understanding of bee-flower relationships is needed. This study illustrates some of the potential strengths and highlights some limitations of using plant species assignments derived from DNA metabarcoding data to study plant-pollinator interactions. We found that plant-pollinator networks created using plant species identifications derived from DNA metabarcoding methods were more complex, with more plant taxa, greater connectance, and greater linkage density, than those created using bee foraging observations (Table 3.1; Figures 3.5a,b). Other studies have found similar results (Pornon et al. 2016; Potter et al., 2019). We found that many of the plant species detected in pollen loads with DNA metabarcoding but with no observed bee visitation were small in stature or rare at the study sites, (e.g., slender phlox [Microsteris gracilis (Hook.) Greene], black medick [Medicago lupulina L.], smallflower woodland-star [Lithophragma parviflorum (Hook.) Nutt. Ex Torr. & A. Gray]) suggesting that DNA metabarcoding may help to document interactions not easily observed in the field. Plant species that were very abundant in the environment but were not observed being visited by bees were also detected using plant species assignments from DNA metabarcoding. Yellow rabbitbrush (Chrysothamnus viscidiflorus [Hook.] Nutt.) was commonly observed growing at Threemile, but it was never observed being visited by bees. Due to its sheer abundance in the environment at Threemile, we suspected that it played a larger role in plant-pollinator networks than the

observation data would indicate. Yellow rabbitbrush was detected in 49 pollen loads (35% of Threemile pollen loads) using MB-LDB, which suggests that yellow rabbitbrush may be an important food source for the native bees at our agricultural sites. However, the common detection of rabbitbrush despite the lack of foraging observations could be due to the movement of rabbitbrush pollen via wind or other insects onto flowers that bees were foraging on. Plant-pollinator networks created from bee foraging observations had low bee generality values and relatively high H2’ values (Figures 3.5c,d). Many bee species in these networks appear to be specialists, visiting only one plant species (Figures 3.2, 3.3, & 3.4). However, in networks created from plant species identified using DNA metabarcoding of pollen loads, these same bee species appear to have a more generalist diet, visiting multiple plant species (Figures 3.2, 3.3, & 3.4), resulting in higher bee generality values and lower H2’ values (Figures 3.5c,d). Networks created from plant species assignments derived from DNA metabarcoding data also had significantly higher values of connectance and linkage density (Figures 3.5a,b), which is believed to be correlated with greater resilience to species loss (Dunne et al., 2002; Montoya et al., 2006). Our results suggest that bees are foraging on a wider variety of resources than indicated by bee foraging observations alone (Figure 3.5). For the 73 bee species sampled in this study, we identified an average of 1.8 plant species per pollen load using MB- RDB and an average of 2.1 plant species per pollen load using MB-LDB (Figure 3.6). Recent DNA metabarcoding work by Potter et al. (2019) found similar patterns for Bombus terrestris and bees of the Halictidae family. Pollen loads collected from foraging B. terrestris and halictid bees had an average of 4.2 and 4.1 plant genera respectively in their pollen loads. Larson et al. (2018) also found that bees from certain families (i.e., Andrenidae, Apidae, Colletidae, and Megachilidae) are most likely to carry mixed pollen loads. If the diet of many bees is broader than bee foraging observations suggest, their resilience in the face of environmental perturbation and resulting alteration of plant communities may be greater than generally appreciated. Pollinator species that have a generalist diet may be less vulnerable to extinction or local extirpation because they are

able to shift their diet to more readily available forage resources. This could be especially important as climate change causes phenological mismatches between bees and plants (Kudo & Ida, 2013). It is reasonable to assume that the plant species on which a bee was foraging when sampled would be detected in its pollen load using plant species assignments from DNA metabarcoding data. However, there are several reasons why this may not occur. We sampled bees that were observed coming in contact with plant reproductive parts, but we did not determine whether pollen was collected from the flower before the bee was sampled. Therefore, the bee may not have had time to collect pollen from the observed plant species before it was sampled. Furthermore, bees forage for pollen and nectar for energy and to provision their young (Ginsberg, 1983; Goulson, 1999). Some of the bees in this study may have been foraging for nectar at the time of sampling and may not have had any pollen from the flower on them when collected. Male bees and cleptoparasitic bees also do not actively collect pollen from the flowers that they visit. However, male bees passively collect small amounts of pollen on their bodies and can play a substantial role in pollination in certain situations (Ostevik et al., 2010). Inconsistencies between bee foraging observations and species assignments derived from DNA metabarcoding data may also be due to plant misidentification in the field. Certain plant species are difficult to differentiate due to similar morphology and other features. One example is longleaf fleabane (Erigeron corymbosus Nutt.) and western mountain aster (Symphyotrichum spathulatum [Lind.] G.L. Nesom). Bees were commonly observed visiting longleaf fleabane, but this species was never detected using DNA metabarcoding data. However, western mountain aster was commonly detected in the pollen loads using both MB-RDB and MB-LDB, and it was detected in every pollen sample on which the bee was collected on longleaf fleabane. Sequence data for both the ITS2 region and the rbcL gene of longleaf fleabane were included in the reference library, and western mountain aster and longleaf fleabane share less than 96% of their ITS2 and rbcL DNA sequences. Therefore, it is unlikely that this was an erroneous species assignment, and it is possible that we misidentified western mountain aster as longleaf fleabane, causing the mismatch in our data.

Only plant species that are included in the reference library could be assigned to sequence data from the pollen loads. This could explain the low percentage of species identifications at Zumwalt (Table 3.3). DNA sequences for at least one of the two barcode markers used in this study (ITS2 and rbcL) were missing from our reference library for 35% of the plant species that bees were observed visiting at Zumwalt (Table A2.2). Some of these plant species with missing sequence data (e.g., palouse thistle [Cirsium brevifolium Nutt.], sticky purple geranium [Geranium viscosissimum Fisch. & C.A. Mey. Ex C.A. Mey.]) appear to be important for bees based on our foraging observations (Table 3.2). If some key pollen sources for native bees are unable to be assigned using DNA metabarcoding because of missing sequence data, this could present problems for land managers that are managing for specific plant and bee species. This highlights the need to continue accumulating sequence information for plant species that have no sequence data or with sequence data that are not available to the public, especially commonly occurring and ecologically important species. Increasing the amount of publicly available sequence data will increase the accuracy of taxonomic assignments and allow for more plant species assignments in future studies. In this study, we used multiple barcode markers for taxonomic identification (i.e., ITS2 and rbcL), and we found that the number of plant taxonomic units identified using both barcode markers (ITS2 + rbcL) was higher than that using ITS2 or rbcL alone 83% of the time (Table 3.1). ITS2 sequence data for certain plant species (e.g., mountain monardella, Nootka rose [Rosa nutkana C. Presl]) were not available to add to our reference library, so these species could not be identified using the ITS2 region. However, we did have the rbcL gene sequence data for these plant species in our reference library, resulting in identification of these species by rbcL alone. The NCBI nucleotide database is far from complete, and until we have complete reference libraries, using multiple loci allows us to obtain additional plant species assignments. We can also be more confident in plant species that were detected using both loci (e.g, whitestem frasera, common St. Johnswort, thinleaved owl’s-clover [Orthocarpus tenuifolius (Pursh) Benth.]) because the probability of a false-positive identification occurring at both gene

regions is smaller than the probability of one occurring at just one gene region for a single plant species (Richardson et al., 2015b). Here we used a multiplex approach in which both primer sets were added to the same PCR reaction. The majority of plant species assignments were made using the ITS2 region, with 98% of the raw reads being ITS2 sequences. Only 2% of the raw reads were rbcL sequences, suggesting that amplification bias was occurring in favor of ITS2. rbcL primer sets typically require an annealing temperature of 50°C (Bruni et al., 2012; Hawkins et al., 2015; Potter et al., 2019), while the annealing temperature of ITS2 primer sets can range from 50 – 60°C (Bálint et al., 2014; Richardson et al., 2015a). In our multiplex PCR reaction, we used an annealing temperature of 55°C, which may have favored amplification of the ITS2 region. Future research is needed to compare amplification of the rbcL barcode marker in a separate PCR reaction to amplification of the rbcL barcode marker when it is multiplexed with another barcode marker (e.g., ITS2) to determine whether it performs better in a separate PCR reaction. Plant-pollinator networks created using MB-RDB included about 15% more taxa than those created using MB-LDB (Table 3.1). One obvious reason that these networks included more plant taxa is because results obtained using the regional database include “regional mismatches” at each location (Table 3.4). We defined regional mismatches as the assignment of a plant species using the regional plant database that is not known to occur in the specific location where the bee was sampled. For example, high mountain cinquefoil (Potentilla flabellifolia Hook. Ex Torr. & A. Gray), a plant species that has only been documented to occur at Zumwalt, was detected in 37 pollen loads at Starkey using the regional database. High mountain cinquefoil has never been documented to occur at Starkey. In contrast, slender cinquefoil is a very common forb at Starkey (Roof et al., 2018), and when we use the local reference database, which does not include high mountain cinquefoil, slender cinquefoil is detected in the same 37 pollen loads. High mountain cinquefoil and slender cinquefoil have 97.4% and 99.8% sequence similarity of the ITS2 region and rbcL gene respectively. Therefore, we can assume that when high mountain cinquefoil is included in the reference library, slender cinquefoil is incorrectly assigned as high mountain cinquefoil to DNA sequences. Similar erroneous species

assignments may be occurring with spotted knapweed (Centaurea stoebe L.) and diffuse knapweed at Threemile as well as Pacific lupine (Lupinus lepidus Douglas ex Lindl.) and silky lupine at Zumwalt. These erroneous species assignments could be especially troubling for land managers who wish to use plant species assignments obtained from DNA metabarcoding of pollen to inform pollinator habitat creation or restoration decisions. Problems distinguishing between plant species in the same genus is a long- standing challenge for metabarcoding studies (Gao et al., 2010). The degree to which this issue poses a problem for investigations of plant-pollinator interactions depends on the type of questions being addressed. Incorrect plant species assignments or assignments simply made at the “operational taxonomic unit” level may not pose a serious problem for certain studies, for example, basic ecological studies examining questions related to diversity of resources and/or attributes of pollinator networks. However, in other circumstances, exact identification of plant species is required to address the question of interest, such as those related to habitat restoration for pollinators or the use of non-native plants by pollinators. As an example, yellow star-thistle is an invasive, non-native aster that was detected in pollen loads of bees collected at Zumwalt (Table A2.2). However, this plant species is relatively rare on the prairie, and it is quite likely that the pollen identified as yellow star-thistle is another aster that shares a high proportion of genetic material at the two barcode markers we examined. Currently, pollen DNA metabarcoding approaches may not be able to address research questions that require a high degree of certainty about species identification. However, sequencing technologies and associated applications are advancing rapidly, and future technologies will likely be able to overcome some of these current limitations (Kane et al., 2012; Taberlet et al., 2012; Roberts et al., 2013; Tang et al., 2014). Using a local reference database reduced the potential for regional mismatches that can occur when using a regional reference database. However, we believe that both sets of plant species assignments obtained from DNA metabarcoding data (i.e., MB-RDB and MB-LDB) provide a better representation of plant species use by native bees at these locations than do bee foraging observations despite some erroneous species assignments

that we believe occurred. Here, we had the advantage of having well-described study sites, but that is often not the case. Those that wish to use plant species assignments derived from DNA metabarcoding of pollen to study the plant-pollinator interactions in locations that are not well described should err on the side of caution, and the chosen method of plant species identification will depend on the types of questions being asked. Past work has shown that for questions that can be answered by family or genus level identification, DNA metabarcoding of pollen works very well (Lucas et al., 2018; Potter et al., 2019). However, if plant species identification is important, extra steps may need to be taken (e.g., using additional barcode markers, plant surveys, observations, microscopy) to confirm that plant species identifications based on DNA metabarcoding data are accurate.

Literature Cited Aird, D., Ross, M. G., Chen, W.-S., Danielsson, M., Fennell, T., Russ, C., … Gnirke, A. (2011). Analyzing and minimizing PCR amplification bias in Illumina sequencing libraries. Genome Biology, 12(2), R18. https://doi.org/10.1186/gb-2011-12-2-r18 Armbruster, D. A., & Pry, T. (2008). Limit of blank, limit of detection and limit of quantitation. The Clinical Biochemist Reviews, 29(Suppl 1), S49–S52. Averett, J. P., Endress, B. A., Rowland, M. M., Naylor, B. J., & Wisdom, M. J. (2017). Wild ungulate herbivory suppresses deciduous woody plant establishment following salmonid stream restoration. Forest Ecology and Management, 391, 135–144. https://doi.org/10.1016/j.foreco.2017.02.017 Bálint, M., Schmidt, P.-A., Sharma, R., Thines, M., & Schmitt, I. (2014). An Illumina metabarcoding pipeline for fungi. Ecology and Evolution, 4(13), 2642–2653. https://doi.org/10.1002/ece3.1107 Bell, K. L., Fowler, J., Burgess, K. S., Dobbs, E. K., Gruenewald, D., Lawley, B., … Brosi, B. J. (2017). Applying pollen DNA metabarcoding to the study of plant– pollinator interactions. Applications in Plant Sciences, 5(6), 1600124. https://doi.org/10.3732/apps.1600124 Bengtsson‐Palme, J., Ryberg, M., Hartmann, M., Branco, S., Wang, Z., Godhe, A., … Nilsson, R. H. (2013). Improved software detection and extraction of ITS1 and ITS2 from ribosomal ITS sequences of fungi and other eukaryotes for analysis of environmental sequencing data. Methods in Ecology and Evolution, 4(10), 914– 919. https://doi.org/10.1111/2041-210X.12073 Bokulich, N. A., Kaehler, B. D., Rideout, J. R., Dillon, M., Bolyen, E., Knight, R., … Gregory Caporaso, J. (2018). Optimizing taxonomic classification of marker-gene amplicon sequences with QIIME 2’s q2-feature-classifier plugin. Microbiome, 6(1), 90. https://doi.org/10.1186/s40168-018-0470-z

Bolyen, E., Rideout, J. R., Dillon, M. R., Bokulich, N. A., Abnet, C. C., Al-Ghalith, G. A., … Caporaso, J. G. (2019). Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2. Nature Biotechnology, 37(8), 852–857. https://doi.org/10.1038/s41587-019-0209-9 Bosch, J., González, A. M. M., Rodrigo, A., & Navarro, D. (2009). Plant–pollinator networks: Adding the pollinator’s perspective. Ecology Letters, 12(5), 409–419. https://doi.org/10.1111/j.1461-0248.2009.01296.x Bruni, I., De Mattia, F., Martellos, S., Galimberti, A., Savadori, P., Casiraghi, M., … Labra, M. (2012). DNA barcoding as an effective tool in improving a digital plant identification system: A case study for the area of Mt. Valerio, Trieste (NE Italy). PLOS ONE, 7(9), e43256. https://doi.org/10.1371/journal.pone.0043256 Buchmann, S. L., & Nabhan, G. P. (2012). The Forgotten Pollinators. Island Press. Callahan, B. J., McMurdie, P. J., Rosen, M. J., Han, A. W., Johnson, A. J. A., & Holmes, S. P. (2016). DADA2: High-resolution sample inference from Illumina amplicon data. Nature Methods, 13(7), 581–583. https://doi.org/10.1038/nmeth.3869 Cornman, R. S., Otto, C. R. V., Iwanowicz, D., & Pettis, J. S. (2015). Taxonomic characterization of honey bee (Apis mellifera) pollen foraging based on non- overlapping paired-end sequencing of nuclear ribosomal loci. PLOS ONE, 10(12), e0145365. https://doi.org/10.1371/journal.pone.0145365 Cusser, S., & Goodell, K. (2013). Diversity and distribution of floral resources influence the restoration of plant–pollinator networks on a reclaimed strip mine. Restoration Ecology, 21(6), 713–721. https://doi.org/10.1111/rec.12003 de Vere, N., Jones, L. E., Gilmore, T., Moscrop, J., Lowe, A., Smith, D., … Ford, C. R. (2017). Using DNA metabarcoding to investigate honey bee foraging reveals limited flower use despite high floral availability. Scientific Reports, 7(1), 42838. https://doi.org/10.1038/srep42838 Delmas, E., Besson, M., Brice, M.-H., Burkle, L. A., Riva, G. V. D., Fortin, M.-J., … Poisot, T. (2019). Analysing ecological networks of species interactions. Biological Reviews, 94(1), 16–36. https://doi.org/10.1111/brv.12433 Dibble, A. C., Drummond, F. A., & Stack, L. B. (2020a). Bee visitation on flowers in maine, united states, reveals the relative attractiveness of plants through space and time: part I. Environmental Entomology. https://doi.org/10.1093/ee/nvaa028 Dibble, A. C., Drummond, F. A., & Stack, L. B. (2020b). Plant origin and other attributes impact bee forage patterns in a common garden study in maine, united states: part II. Environmental Entomology. https://doi.org/10.1093/ee/nvaa029 Domroese, M. C., & Johnson, E. A. (2017). Why watch bees? Motivations of citizen science volunteers in the Great Pollinator Project. Biological Conservation, 208, 40–47. https://doi.org/10.1016/j.biocon.2016.08.020 Dormann, C. F., Gruber, B., & Fründ, J. (2008). The bipartite package. Version 0.73. R Project for Statistical Computing, Vienna, Austria. Dunne, J. A., Williams, R. J., & Martinez, N. D. (2002). Network structure and biodiversity loss in food webs: Robustness increases with connectance. Ecology Letters, 5(4), 558–567. https://doi.org/10.1046/j.1461-0248.2002.00354.x Erdtman, G. (1943). An Introduction to Pollen Analysis. chronica Botanica, waltham. Mass., uSa.

Free, J. B. (1970). The flower constancy of bumblebees. Journal of Animal Ecology, 39(2), 395–402. JSTOR. https://doi.org/10.2307/2978 Galliot, J.-N., Brunel, D., Bérard, A., Chauveau, A., Blanchetête, A., Lanore, L., & Farruggia, A. (2017). Investigating a flower-insect forager network in a mountain grassland community using pollen DNA barcoding. Journal of Insect Conservation, 21(5–6), 827–837. https://doi.org/10.1007/s10841-017-0022-z Gao, T., Yao, H., Song, J., Zhu, Y., Liu, C., & Chen, S. (2010). Evaluating the feasibility of using candidate DNA barcodes in discriminating species of the large Asteraceae family. BMC Evolutionary Biology, 10(1), 324. https://doi.org/10.1186/1471-2148-10-324 Ginsberg, H. S. (1983). Foraging ecology of bees in an old field. Ecology, 64(1), 165– 175. https://doi.org/10.2307/1937338 Goulson, D. (1999). Foraging strategies of insects for gathering nectar and pollen, and implications for plant ecology and evolution. Perspectives in Plant Ecology, Evolution and Systematics, 2(2), 185–209. https://doi.org/10.1078/1433-8319- 00070 Gresty, C. E. A., Clare, E., Devey, D. S., Cowan, R. S., Csiba, L., Malakasi, P., … Willis, K. J. (2018). Flower preferences and pollen transport networks for cavity-nesting solitary bees: Implications for the design of agri-environment schemes. Ecology and Evolution, 8(15), 7574–7587. https://doi.org/10.1002/ece3.4234 Hawkins, J., Vere, N. de, Griffith, A., Ford, C. R., Allainguillaume, J., Hegarty, M. J., … Adams-Groom, B. (2015). Using DNA metabarcoding to identify the floral composition of honey: a new tool for investigating honey bee foraging preferences. PLOS ONE, 10(8), e0134735. https://doi.org/10.1371/journal.pone.0134735 Hebert, P. D. N., Cywinska, A., Ball, S. L., & deWaard, J. R. (2003). Biological identifications through DNA barcodes. Proceedings of the Royal Society of London. Series B: Biological Sciences, 270(1512), 313–321. https://doi.org/10.1098/rspb.2002.2218 Janzen, D. H. (1983). Seed and pollen dispersal by animals: Convergence in the ecology of contamination and sloppy harvest. Biological Journal of the Linnean Society, 20(1), 103–113. https://doi.org/10.1111/j.1095-8312.1983.tb01592.x Kane, N., Sveinsson, S., Dempewolf, H., Yang, J. Y., Zhang, D., Engels, J. M. M., & Cronk, Q. (2012). Ultra-barcoding in cacao (Theobroma spp.; Malvaceae) using whole chloroplast genomes and nuclear ribosomal DNA. American Journal of Botany, 99(2), 320–329. https://doi.org/10.3732/ajb.1100570 Kearns, C. A., & Inouye, D. W. (1993). Techniques for pollination biologists. Techniques for Pollination Biologists. Retrieved from https://www.cabdirect.org/cabdirect/abstract/19940200882 Keller, A., Danner, N., Grimmer, G., Ankenbrand, M., Ohe, K. von der, Ohe, W. von der, … Steffan‐Dewenter, I. (2015). Evaluating multiplexed next-generation sequencing as a method in palynology for mixed pollen samples. Plant Biology, 17(2), 558–566. https://doi.org/10.1111/plb.12251 Kennedy, P. L., DeBano, S. J., Bartuszevige, A. M., & Lueders, A. S. (2009). Effects of native and non-native grassland plant communities on breeding passerine birds:

implications for restoration of northwest bunchgrass prairie. Restoration Ecology, 17(4), 515–525. https://doi.org/10.1111/j.1526-100X.2008.00402.x Kimoto, C., DeBano, S. J., Thorp, R. W., Rao, S., & Stephen, W. P. (2012). Investigating temporal patterns of a native bee community in a remnant north american bunchgrass prairie using blue vane traps. Journal of Insect Science, 12(108), 1– 23. https://doi.org/10.1673/031.012.10801 Klindworth, A., Pruesse, E., Schweer, T., Peplies, J., Quast, C., Horn, M., & Glöckner, F. O. (2013). Evaluation of general 16S ribosomal RNA gene PCR primers for classical and next-generation sequencing-based diversity studies. Nucleic Acids Research, 41(1), e1–e1. https://doi.org/10.1093/nar/gks808 Kraaijeveld, K., Weger, L. A. de, García, M. V., Buermans, H., Frank, J., Hiemstra, P. S., & Dunnen, J. T. den. (2015). Efficient and sensitive identification and quantification of airborne pollen using next-generation DNA sequencing. Molecular Ecology Resources, 15(1), 8–16. https://doi.org/10.1111/1755- 0998.12288 Kudo, G., & Ida, T. Y. (2013). Early onset of spring increases the phenological mismatch between plants and pollinators. Ecology, 94(10), 2311–2320. https://doi.org/10.1890/12-2003.1 Kuhlman, M., & Burrows, S. (2017). Checklist of bees (Apoidea) from a private conservation property in west-central Montana. Biodiversity Data Journal, 5. https://doi.org/10.3897/BDJ.5.e11506 Koh, I., Lonsdorf, E. V., Williams, N. M., Brittain, C., Isaacs, R., Gibbs, J., & Ricketts, T. H. (2016). Modeling the status, trends, and impacts of wild bee abundance in the United States. Proceedings of the National Academy of Sciences, 113(1), 140– 145. https://doi.org/10.1073/pnas.1517685113 Larson, D. L., Larson, J. L., & Buhl, D. A. (2018). Conserving all the pollinators: variation in probability of pollen transport among insect taxa. Natural Areas Journal, 38(5), 393. https://doi.org/10.3375/043.038.0508 Latta, R. G., Linhart, Y. B., Fleck, D., & Elliot, M. (1998). Direct and indirect estimates of seed versus pollen movement within a population of ponderosa pine. Evolution, 52(1), 61–67. https://doi.org/10.1111/j.1558-5646.1998.tb05138.x Losey, J. E., & Vaughan, M. (2006). The economic value of ecological services provided by insects. BioScience, 56(4), 311. https://doi.org/10.1641/0006- 3568(2006)56[311:TEVOES]2.0.CO;2 Lucas, A., Bodger, O., Brosi, B. J., Ford, C. R., Forman, D. W., Greig, C., … de Vere, N. (2018). Floral resource partitioning by individuals within generalised hoverfly pollination networks revealed by DNA metabarcoding. Scientific Reports, 8(1), 5133. https://doi.org/10.1038/s41598-018-23103-0 Macgregor, C. J., Kitson, J. J. N., Fox, R., Hahn, C., Lunt, D. H., Pocock, M. J. O., & Evans, D. M. (2019). Construction, validation, and application of nocturnal pollen transport networks in an agro-ecosystem: A comparison using light microscopy and DNA metabarcoding. Ecological Entomology, 44(1), 17–29. https://doi.org/10.1111/een.12674 Montoya, J. M., Pimm, S. L., & Solé, R. V. (2006). Ecological networks and their fragility. Nature, 442(7100), 259–264. https://doi.org/10.1038/nature04927

Morandin, L. A., & Kremen, C. (2013). Hedgerow restoration promotes pollinator populations and exports native bees to adjacent fields. Ecological Applications, 23(4), 829–839. https://doi.org/10.1890/12-1051.1 Ollerton, J., Winfree, R., & Tarrant, S. (2011). How many flowering plants are pollinated by animals? Oikos, 120(3), 321–326. https://doi.org/10.1111/j.1600- 0706.2010.18644.x Ostevik, K., Manson, J., & Thompson, J. (2010). Pollination potential of male bumble bees (Bombus impatiens): Movement patterns and pollen-transfer efficiency. Journal of Pollination Ecology, 2, 21–26. https://doi.org/10.26786/1920- 7603(2010)3 Pornon, A., Escaravage, N., Burrus, M., Holota, H., Khimoun, A., Mariette, J., Pellizzari, C., Iribar, A., Etienne, R., Taberlet, P., Vidal, M., Winterton, P., Zinger, L., & Andalo, C. (2016). Using metabarcoding to reveal and quantify plant-pollinator interactions. Scientific Reports, 6(1), 27282. https://doi.org/10.1038/srep27282 Potter, C., Vere, N. de, Jones, L. E., Ford, C. R., Hegarty, M. J., Hodder, K. H., … Franklin, E. L. (2019). Pollen metabarcoding reveals broad and species-specific resource use by urban bees. PeerJ, 7, e5999. https://doi.org/10.7717/peerj.5999 Potts, S. G., Biesmeijer, J. C., Kremen, C., Neumann, P., Schweiger, O., & Kunin, W. E. (2010). Global pollinator declines: Trends, impacts and drivers. Trends in Ecology & Evolution, 25(6), 345–353. https://doi.org/10.1016/j.tree.2010.01.007 Rahl, M. (2008). Microscopic identification and purity determination of pollen grains. In M. G. Jones & P. Lympany (Eds.), Allergy Methods and Protocols (pp. 263–269). https://doi.org/10.1007/978-1-59745-366-0_22 Richardson, R. T., Lin, C.-H., Sponsler, D. B., Quijia, J. O., Goodell, K., & Johnson, R. M. (2015a). Application of ITS2 metabarcoding to determine the provenance of pollen collected by honey bees in an agroecosystem. Applications in Plant Sciences, 3(1), 1400066. https://doi.org/10.3732/apps.1400066 Richardson, R. T., Lin, C.-H., Quijia, J. O., Riusech, N. S., Goodell, K., & Johnson, R. M. (2015b). Rank-based characterization of pollen assemblages collected by honey bees using a multi-locus metabarcoding approach. Applications in Plant Sciences, 3(11), 1500043. https://doi.org/10.3732/apps.1500043 Richardson, R. T., Sponsler, D. B., McMinn-Sauder, H., & Johnson, R. M. (2020). MetaCurator: A hidden Markov model-based toolkit for extracting and curating sequences from taxonomically-informative genetic markers. Methods in Ecology and Evolution 11(1), 181-186. https://doi.org/10.1101/672782 Roberts, R. J., Carneiro, M. O., & Schatz, M. C. (2013). The advantages of SMRT sequencing. Genome Biology, 14(6), 405. https://doi.org/10.1186/gb-2013-14-6- 405 Roof, S. M., DeBano, S., Rowland, M. M., & Burrows, S. (2018). Associations between blooming plants and their bee visitors in a riparian ecosystem in eastern Oregon. Northwest Science, 92(2), 119. https://doi.org/10.3955/046.092.0205 Rowland, M. M., Bryant, L. D., Johnson, B. K., Noyes, J. H., Wisdom, M. J., & Thomas, J. W. (1997). The Starkey project: history, facilities, and data collection methods for ungulate research. Gen. Tech. Rep. PNW-GTR-396. Portland, OR: US

Department of Agriculture, Forest Service, Pacific Northwest Research Station. 62 p, 396. Sajwani, A., Farooq, S. A., & Bryant, V. M. (2014). Studies of bee foraging plants and analysis of pollen pellets from hives in Oman. Palynology, 38(2), 207–223. https://doi.org/10.1080/01916122.2013.871652 Sickel, W., Ankenbrand, M. J., Grimmer, G., Holzschuh, A., Härtel, S., Lanzen, J., … Keller, A. (2015). Increased efficiency in identifying mixed pollen samples by meta-barcoding with a dual-indexing approach. BMC Ecology, 15(1), 20. https://doi.org/10.1186/s12898-015-0051-y Smart, M. D., Cornman, R. S., Iwanowicz, D. D., McDermott-Kubeczko, M., Pettis, J. S., Spivak, M. S., & Otto, C. R. V. (2017). A comparison of honey bee-collected pollen from working agricultural lands using light microscopy and ITS metabarcoding. Environmental Entomology, 46(1), 38–49. https://doi.org/10.1093/ee/nvw159 Taberlet, P., Coissac, E., Pompanon, F., Brochmann, C., & Willerslev, E. (2012). Towards next-generation biodiversity assessment using DNA metabarcoding. Molecular Ecology, 21(8), 2045–2050. https://doi.org/10.1111/j.1365- 294X.2012.05470.x Tang, M., Tan, M., Meng, G., Yang, S., Su, X., Liu, S., Song, W., Li, Y., Wu, Q., Zhang, A., & Zhou, X. (2014). Multiplex sequencing of pooled mitochondrial genomes— A crucial step toward biodiversity analysis using mito-metagenomics. Nucleic Acids Research, 42(22), e166–e166. https://doi.org/10.1093/nar/gku917 Von Der Ohe, W., Persano Oddo, L., Piana, M. L., Morlot, M., & Martin, P. (2004). Harmonized methods of melissopalynology. Apidologie, 35(Suppl. 1), S18–S25. https://doi.org/10.1051/apido:2004050 Whitehead, D. R. (1983). Wind pollination: some ecological and evolutionary perspectives. Pollination biology, 97(08). Whittington, R., Winston, M. L., Tucker, C., & Parachnowitsch, A. L. (2004). Plant- species identity of pollen collected by bumblebees placed in greenhouses for tomato pollination. Canadian Journal of Plant Science, 84(2), 599–602. https://doi.org/10.4141/P02-192 Winfree, R., Aguilar, R., Vázquez, D. P., LeBuhn, G., & Aizen, M. A. (2009). A meta- analysis of bees’ responses to anthropogenic disturbance. Ecology, 90(8), 2068– 2076. https://doi.org/10.1890/08-1245.1 Wood, T. J., Holland, J. M., & Goulson, D. (2015). Pollinator-friendly management does not increase the diversity of farmland bees and wasps. Biological Conservation, 187, 120–126. https://doi.org/10.1016/j.biocon.2015.04.022

Figures and Tables

Figure 3.1: Workflow of the QIIME2 pipeline for bee pollen composition analysis.

Figure 3.2: Plant-pollinator networks at Threemile Canyon Farms created from a) bee foraging observations, b) DNA metabarcoding data using a reference library of regional plants, and c) DNA metabarcoding data using a reference library of local plants only. In each network, the top row represents plant species, and the bottom row represents bee species. Thickness of the lines represents the frequency of the interactions. Complete species listings for bees and plants can be found in Table A2.2.

Figure 3.3: Plant-pollinator networks at Starkey Experimental Forest and Range created from a) bee foraging observations, b) DNA metabarcoding data using a reference library of regional plants, and c) DNA metabarcoding data using a reference library of local plants only. In each network, the top row represents plant species, and the bottom row represents bee species. Thickness of the lines represents the frequency of the interactions. Complete species listings for bees and plants can be found in Table A2.4.

Figure 3.4: Plant-pollinator networks at Zumwalt created from a) bee foraging observations, b) DNA metabarcoding data using a reference library of regional plants, and c) DNA metabarcoding data using a reference library of local plants only. In each network, the top row represents plant species, and the bottom row represents bee species. Thickness of the lines represents the frequency of the interactions. Complete species listings for bees and plants can be found in Table A2.5.

Figure 3.5: Bar plots of a) connectance, b) linkage density, c) bee generality, and d) H2’ for DNA metabarcoding using a local reference database (MB-LDB), DNA metabarcoding using a regional reference database (MB-RDB), and bee foraging observations. Different letters denote averages (of values that were scaled to the highest value for a given parameter at a given site) that differ significantly according to a post hoc Tukey’s HSD test. Error bars are ± SE.

Figure 3.6: Number of plant species detected in a single pollen load for each detection method. MB-LDB, DNA metabarcoding using a local reference database; MB-RDB, DNA metabarcoding using a regional reference database.

Table 3.1: Number of plant taxonomic units assigned by each detection method for all pollen loads collected at a location. BFO: Bee foraging observations; MB-RDB: DNA metabarcoding using a regional reference database; MB-LDB: DNA metabarcoding using a local reference database.

BFO MB-RDB MB-LDB ITS2 rbcL ITS2 + rbcL ITS2 rbcL ITS2 + rbcL Threemile Families 4 6 4 7 5 3 6 Genera 8 12 3 13 8 4 11 Species 9 14 3 15 10 4 13 Starkey Families 10 17 7 18 14 7 14 Genera 21 42 12 45 34 10 38 Species 26 46 10 49 38 10 44 Zumwalt Families 13 15 7 16 15 6 16 Genera 21 33 6 34 30 6 31 Species 23 34 4 34 29 4 29

Table 3.2: Top 10 plant species identified in plant-pollinator networks at Threemile Canyon Farms (Threemile), the USFS Starkey Experimental Forest and Range (Starkey), and The Nature Conservancy’s Zumwalt Prairie Preserve (Zumwalt) using behavioral observations and DNA metabarcoding with a regional (MB-RDB) and local (MB-LDB) reference library. Numbers in parentheses following location names indicate total pollen loads sampled from that location. Numbers in parentheses following plant species names indicate number of pollen loads in which that plant species was identified. A “‡” indicates a plant species that was identified using all three detection methods. A “†” indicates a plant species that was identified using two detection methods.

Observations MB-RDB MB-LDB Threemile (139) Centaurea diffusa (57) † Centaurea stoebe (66) Centaurea solstitialis (84) ‡ Machaeranthera canescens (33) † Centaurea solstitialis (59) ‡ Centaurea diffusa (64) † Centaurea solstitialis (22) ‡ Machaeranthera canescens (41) † Chrysothamnus viscidiflorus (49) Gutierrezia sarothrae (7) ‡ Sisymbrium altissimum (12) ‡ Gutierrezia sarothrae (48) ‡ Onopordum acanthium (7) ‡ Gutierrezia sarothrae (11) ‡ Sisymbrium altissimum (12) ‡ Sisymbrium altissimum (7) ‡ Onopordum acanthium (8) ‡ Onopordum acanthium (8) ‡ Psoralidium lanceolatum (4) † Salsola kali (8) ‡ Salsola kali (8) ‡ Achillea millefolium (1) ‡ Sisymbrium loeselii (8) † Chenopodium leptophyllum (7) Salsola kali (1) ‡ Kali turgidum (7) Plectritis macrocera (5) N/A Psoralidium lanceolatum (5) † Sisymbrium loeselii (3) † Starkey (168) Potentilla gracilis (37) † Potentilla flabellifolia (37) Potentilla gracilis (37) † Solidago missouriensis (34) † Achillea millefolium (33) ‡ Achillea millefolium (33) ‡

Monardella odoratissima (15) ‡ Solidago canadensis (30) Solidago missouriensis (30) † Cirsium vulgare (13) ‡ Symphyotrichum spathulatum (30) ‡ Symphyotrichum spathulatum (30) ‡ Penstemon procerus (13) † Lupinus lepidus (26) Potentilla recta (29) Erigeron corymbosus (10) Hypericum perforatum (26) ‡ Lupinus sulphureus (26) † Potentilla arguta (9) Perideridia gairdneri (18) † Hypericum perforatum (26) ‡ Thermopsis montana (6) ‡ Senecio integerrimus (18) Perideridia gairdneri (18) † Achillea millefolium (4) ‡ Trifolium hybridum (15) Senecio serra (18) † Hypericum perforatum (4) ‡ Cirsium vulgare (14) ‡ Cirsium vulgare (17) ‡ Zumwalt (96) Lupinus sericeus (10) Lupinus lepidus (42) † Lupinus lepidus (42) † Frasera albicaulis (10) ‡ Frasera albicaulis (21) ‡ Frasera albicaulis (21) ‡ Erigeron pumilus (10) Potentilla flabellifolia (13) † Symphyotrichum spathulatum (20) † Potentilla gracilis (9) Achillea millefolium (11) ‡ Potentilla flabellifolia (14) † Arnica sororia (6) ‡ Symphyotrichum spathulatum (10) † Achillea millefolium (11) ‡ Grindelia nana (6) † Arnica sororia (9) ‡ Arnica sororia (9) ‡ Orthocarpus tenuifolius (6) ‡ Grindelia nana (7) † Orthocarpus tenuifolius (6) ‡ Cirsium brevifolium (5) Orthocarpus tenuifolius (6) ‡ Cirsium foliosum (5) † Lupinus leucophyllus (4) Cirsium foliosum (4) † Delphinium nuttallianum (4) ‡ Clarkia pulchella (3) Perideridia gairdneri (4) † Perideridia gairdneri (4) †

Table 3.3: Percent of samples in which metabarcoding detected the plant species from which the corresponding bee was collected. MB-RDB: DNA metabarcoding using a regional reference database; MB-LDB: DNA metabarcoding using a local reference database.

Threemile Starkey Zumwalt MB-RDB MB-LDB MB-RDB MB-LDB MB-RDB MB-LDB Family 97.8% 97.1% 92.9% 93.5% 79.6% 78.6% Genus 95.7% 69.8% 73.2% 78.6% 69.4% 61.2% Species 54.7% 63.3% 30.4% 70.8% 38.8% 33.7%

Table 3.4: Mismatches of regional database metabarcoding. Mismatches for the regional database are species that were identified with DNA metabarcoding that are not known to occur in the area. See Table A2.2 for more detail. A “†” indicates a plant species assigned using the ITS2 gene region, a “‡” indicates a plant species assigned using the rbcL gene region, and a “§” indicates a plant species assigned using both ITS2 and rbcL gene regions.

Location Regional Mismatches Threemile Centaurea stoebe † Kali turgidum ‡ Rhaponticum repens † Starkey Erythranthe guttata † Micranthes oregana † Phacelia heterophylla † Potentilla flabellifolia § Trifolium hybridum † Zumwalt Hieracium scouleri † Phacelia hastata † Trifolium pratense †

Chapter 4: Conclusion Native bee populations are declining worldwide, and reversing these declines requires a greater understanding of the complex relationship between bees and flowering plants. Although traditional methods for describing plant-bee interactions (e.g., bee foraging observations and microscopy) are still frequently used today, DNA metabarcoding of bee-collected pollen provides researchers with a more detailed record of bee foraging behavior. This method is ultimately less time intensive and costly than microscopy, and results in greater taxonomic resolution than traditional methods. In Chapter 2, we addressed several unresolved questions about pollen metabarcoding techniques including whether 1) different sequence count removal threshold protocols result in different interpretations of pollen metabarcoding data, 2) DNA metabarcoding results can be interpreted quantitatively, 3) DNA metabarcoding can detect rare plant-bee interactions, and 4) DNA metabarcoding data might overestimate the number of plant species on which bees forage. We determined that, regardless of the threshold type used, the number of sequencing reads produced for a plant species cannot be used to quantify the amount of that species’ pollen in a mixed species sample. Certain plant species were consistently over and underrepresented by the number of sequencing reads produced by DNA metabarcoding, causing the relationship between the number of sequencing reads and the amount of pollen to vary strongly from a one-to-one relationship. Although we cannot determine the amount of pollen from each plant species in a bee’s pollen load using these metabarcoding techniques, we can still estimate the proportion of pollen loads in which a plant species is detected. This can be used along with plant survey data (e.g., presence and abundance during sampling periods) to get an idea of which plant species are preferred by bees. We were able to correctly identify all plant species used to create the laboratory- prepared mixtures in Chapter 2. However, after using the conservative threshold to remove potential contamination, some species were not detected above the threshold in mixtures 2-5, resulting in false negatives. All “rare” plant species in mixture 5 were detected when using the liberal threshold, but not the conservative threshold, suggesting that single plant-bee interactions are not consistently detected when using a more

conservative sequence count removal threshold. However, this may not be a serious limitation for pollinator networks studies because single flower visits that may be difficult to detect are most likely not ecologically relevant. When using the liberal threshold, six plant species that were not used to create the laboratory-prepared pollen mixtures were detected in the single-species samples, while no additional species were detected in the single-species samples when using the conservative threshold. This suggests that the use of a more liberal threshold could introduce false positive plant identifications (due to sequencing background noise, contamination, or wind dispersed pollen) that arguably should not be included in plant-pollinator network analyses, and plant-pollinator networks created using the conservative threshold likely do not overestimate the number of plant species on which bees forage because of contamination issues. In Chapter 3, we sequenced pollen loads collected from 403 native bees individually. We compared plant species assignments obtained using a regional reference database containing all plant species known to occur in the larger region to those obtained using a local reference database containing only the plant species known to occur in the location in which the bees were sampled. We found that using the regional database overestimated the number of plants visited by about 15% due to regional mismatches and other erroneous taxonomic assignments, suggesting that a local database may provide more accurate plant species assignments. We also compared bee foraging observations to plant species assignments obtained from DNA metabarcoding data. We found that DNA metabarcoding identified more plant species than bee foraging observations at all three locations, and the mean number of flower visits detected per bee was significantly greater using metabarcoding techniques. Plant-pollinator networks created using plant species assignments obtained from DNA metabarcoding data had significantly higher linkage density and bee generality and significantly lower H2’ (specialization) compared to plant-pollinator networks based on bee foraging observations. MB-LDB networks also had significantly higher connectance than MB-RDB and bee foraging observation networks. There were some inconsistencies among bee foraging observations and plant species assignments

obtained using DNA metabarcoding. These inconsistencies could be explained by bee behavior as described in Chapter 3 (e.g., not having enough time to collect pollen before being sampled; collecting nectar instead of pollen), but a number of them are likely due to bioinformatic deficiencies (e.g., incomplete reference database; erroneous taxonomic assignments). Erroneous taxonomic assignments are more difficult, and potentially impossible, to control, but incomplete reference databases can be addressed relatively easily. Researchers who wish to use the methods presented in Chapter 3 to study plant- bee interactions should be cautious and take several pre-sequencing steps to ensure the most accurate results. Vegetation surveys should be conducted at the time of bee sampling, not only at bee sampling locations, but in the surrounding area (up to 10 km), in order to create a comprehensive local reference library that will include all potential forage resources. If researchers have prior knowledge of the blooming plants in the area, a search for DNA sequences of these species on GenBank or in other curated databases will allow them to create a list of plant species for which no DNA sequence data is available. Then, field samples of these plant species should be collected, sequenced and added to the reference library manually. If no prior knowledge of the vegetation at the study site is available, field samples should be taken for all plant species identified at the study site in case they need to be sequenced and added to the reference library. If this is not feasible, and a larger, regional reference database is used (e.g., all angiosperms in the state of Oregon), taxonomic assignments should be limited to the family or genus level because it is likely that erroneous taxonomic assignments will occur at the species level. Although these initial steps may be time intensive, a thorough examination of the vegetation in the immediate sampling area and the surrounding area is the best way to ensure all possible forage resources are included in the reference database. Pollen DNA metabarcoding has repeatedly been shown to identify more plant species and plant-bee interactions than bee foraging observations and microscopic pollen analysis. The time spent preparing an accurate and complete reference library for metabarcoding at a specific study location would be far more productive and informative than the equivalent time spent using traditional methods to describe bee-flower interactions. Bee foraging

observations are severely constrained by the observer’s ability to follow a bee from flower to flower, which would require the fast and accurate identification of multiple plant species in the field as well as the successful sampling of the observed bee. Microscopy is limited by the fact that only a subsample of a bee’s pollen load is examined and the researcher’s ability to correctly identify pollen grains to genus or species, a difficult task that is typically only accomplished by experts. Although there are some drawbacks to using pollen DNA metabarcoding, the positives far outweigh the negatives, and many of the issues identified in Chapter 3 can be resolved by taking the initial steps to create a comprehensive, site-specific reference database. The results presented in this thesis open several areas for future research. There is some evidence suggesting that the number of sequencing reads produced using chloroplast barcode markers are a good proxy for the amount of pollen in a sample (Kraaijeveld et al., 2014; Richardson et al., 2019). Unfortunately, we did not gain any useful rbcL sequence data from our multiplex reaction in Chapter 2. Future studies could create pollen mixtures like ours, sequence samples using rbcL barcode markers separately, and examine the relationship between number of sequence reads and amount of pollen for each plant species. A significant, one-to-one relationship would add to a growing body of evidence that chloroplast barcode markers may be used to quantify the amount of each plant species in a bee’s pollen load. Another area for future research would be to use the local database metabarcoding results presented here to examine landscape effects (e.g., disturbed vs. natural) on bee communities. For example, we might expect more generalist foraging behavior from bee species in the disturbed, agricultural sites and more specialized foraging behavior from bee species in our natural sites due to greater plant species diversity in natural habitats, allowing for greater specialization. These data could also be separated by month to determine whether there are a higher proportion of rare, specialist bee species in June and more common, generalist bee species in August. Plant species diversity is typically greater early in the season, allowing for more opportunities for specialization. Finally, plant survey data (e.g., blooming stem presence and abundance) conducted during each period of bee sampling could be used with local database

metabarcoding results to estimate visitation rates for each plant species. Plant species with a high visitation rate would appear to be preferred by bees. These plant species as well as plant species that are detected in a large proportion of bee pollen loads (which may not necessarily have a high visitation rate due to their abundance in the environment) could be used to create seed mixes and inform land management decisions involving restoration of native bee habitat. The techniques presented here could be used in any part of the world, in any habitat type to create region specific plant lists for bees.

Literature Cited Richardson, R. T., Curtis, H. R., Matcham, E. G., Lin, C.-H., Suresh, S., Sponsler, D. B., Hearon, L. E., & Johnson, R. M. (2019). Quantitative multi-locus metabarcoding and waggle dance interpretation reveal honey bee spring foraging patterns in Midwest agroecosystems. Molecular Ecology, 28(3), 686–697. https://doi.org/10.1111/mec.14975 Kraaijeveld, K., Weger, L. A. de, García, M. V., Buermans, H., Frank, J., Hiemstra, P. S., & Dunnen, J. T. den. (2015). Efficient and sensitive identification and quantification of airborne pollen using next-generation DNA sequencing. Molecular Ecology Resources, 15(1), 8–16. https://doi.org/10.1111/1755- 0998.12288

Bibliography Aird, D., Ross, M. G., Chen, W.-S., Danielsson, M., Fennell, T., Russ, C., … Gnirke, A. (2011). Analyzing and minimizing PCR amplification bias in Illumina sequencing libraries. Genome Biology, 12(2), R18. https://doi.org/10.1186/gb-2011-12-2-r18 Álvarez, I., & Wendel, J. F. (2003). Ribosomal ITS sequences and plant phylogenetic inference. Molecular Phylogenetics and Evolution, 29(3), 417–434. https://doi.org/10.1016/S1055-7903(03)00208-2 Armbruster, D. A., & Pry, T. (2008). Limit of blank, limit of detection and limit of quantitation. The Clinical Biochemist Reviews, 29(Suppl 1), S49–S52. Arstingstall, K. A., DeBano, S. J., Li, X., Wooster, D. E., Rowland, M. M., Burrows, S., Frost, K. (In Revision). Capabilities and limitations of using DNA metabarcoding to study plant-pollinator interactions. Molecular Ecology Averett, J. P., Endress, B. A., Rowland, M. M., Naylor, B. J., & Wisdom, M. J. (2017). Wild ungulate herbivory suppresses deciduous woody plant establishment following salmonid stream restoration. Forest Ecology and Management, 391, 135–144. https://doi.org/10.1016/j.foreco.2017.02.017 Bálint, M., Schmidt, P.-A., Sharma, R., Thines, M., & Schmitt, I. (2014). An Illumina metabarcoding pipeline for fungi. Ecology and Evolution, 4(13), 2642–2653. https://doi.org/10.1002/ece3.1107 Bell, K. L., Fowler, J., Burgess, K. S., Dobbs, E. K., Gruenewald, D., Lawley, B., … Brosi, B. J. (2017). Applying pollen DNA metabarcoding to the study of plant– pollinator interactions. Applications in Plant Sciences, 5(6), 1600124. https://doi.org/10.3732/apps.1600124 Bell, K. L., Burgess, K. S., Botsch, J. C., Dobbs, E. K., Read, T. D., & Brosi, B. J. (2019). Quantitative and qualitative assessment of pollen DNA metabarcoding using constructed species mixtures. Molecular Ecology, 28(2), 431–455. https://doi.org/10.1111/mec.14840 Bengtsson‐Palme, J., Ryberg, M., Hartmann, M., Branco, S., Wang, Z., Godhe, A., … Nilsson, R. H. (2013). Improved software detection and extraction of ITS1 and ITS2 from ribosomal ITS sequences of fungi and other eukaryotes for analysis of environmental sequencing data. Methods in Ecology and Evolution, 4(10), 914– 919. https://doi.org/10.1111/2041-210X.12073 Bokulich, N. A., Kaehler, B. D., Rideout, J. R., Dillon, M., Bolyen, E., Knight, R., … Gregory Caporaso, J. (2018). Optimizing taxonomic classification of marker-gene amplicon sequences with QIIME 2’s q2-feature-classifier plugin. Microbiome, 6(1), 90. https://doi.org/10.1186/s40168-018-0470-z Bolyen, E., Rideout, J. R., Dillon, M. R., Bokulich, N. A., Abnet, C. C., Al-Ghalith, G. A., … Caporaso, J. G. (2019). Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2. Nature Biotechnology, 37(8), 852–857. https://doi.org/10.1038/s41587-019-0209-9 Bosch, J., González, A. M. M., Rodrigo, A., & Navarro, D. (2009). Plant–pollinator networks: Adding the pollinator’s perspective. Ecology Letters, 12(5), 409–419. https://doi.org/10.1111/j.1461-0248.2009.01296.x Brooks, J. P., Edwards, D. J., Harwich, M. D., Rivera, M. C., Fettweis, J. M., Serrano, M. G., Reris, R. A., Sheth, N. U., Huang, B., Girerd, P., Vaginal Microbiome

Consortium (additional members), Strauss, J. F., Jefferson, K. K., & Buck, G. A. (2015). The truth about metagenomics: Quantifying and counteracting bias in 16S rRNA studies. BMC Microbiology, 15(1), 66. https://doi.org/10.1186/s12866-015- 0351-6 Bruni, I., De Mattia, F., Martellos, S., Galimberti, A., Savadori, P., Casiraghi, M., … Labra, M. (2012). DNA barcoding as an effective tool in improving a digital plant identification system: a case study for the area of Mt. Valerio, Trieste (NE Italy). PLOS ONE, 7(9), e43256. https://doi.org/10.1371/journal.pone.0043256 Buchmann, S. L., & Nabhan, G. P. (1996). The Forgotten Pollinators. Island Press. Callahan, B. J., McMurdie, P. J., Rosen, M. J., Han, A. W., Johnson, A. J. A., & Holmes, S. P. (2016). DADA2: High-resolution sample inference from Illumina amplicon data. Nature Methods, 13(7), 581–583. https://doi.org/10.1038/nmeth.3869 Cameron, S. A., Lozier, J. D., Strange, J. P., Koch, J. B., Cordes, N., Solter, L. F., & Griswold, T. L. (2011). Patterns of widespread decline in North American bumble bees. Proceedings of the National Academy of Sciences, 108(2), 662–667. https://doi.org/10.1073/pnas.1014743108 Cameron, S. A., & Sadd, B. M. (2020). Global trends in bumble bee health. Annual Review of Entomology, 65(1), 209–232. https://doi.org/10.1146/annurev-ento- 011118-111847 Chen, S., Yao, H., Han, J., Liu, C., Song, J., Shi, L., Zhu, Y., Ma, X., Gao, T., Pang, X., Luo, K., Li, Y., Li, X., Jia, X., Lin, Y., & Leon, C. (2010). Validation of the ITS2 region as a novel dna barcode for identifying medicinal plant species. PLOS ONE, 5(1), e8613. https://doi.org/10.1371/journal.pone.0008613 Cornman, R. S., Otto, C. R. V., Iwanowicz, D., & Pettis, J. S. (2015). Taxonomic characterization of honey bee (Apis mellifera) pollen foraging based on non- overlapping paired-end sequencing of nuclear ribosomal loci. PLOS ONE, 10(12), e0145365. https://doi.org/10.1371/journal.pone.0145365 Cusser, S., & Goodell, K. (2013). Diversity and distribution of floral resources influence the restoration of plant–pollinator networks on a reclaimed strip mine. Restoration Ecology, 21(6), 713–721. https://doi.org/10.1111/rec.12003 de Vere, N., Jones, L. E., Gilmore, T., Moscrop, J., Lowe, A., Smith, D., … Ford, C. R. (2017). Using DNA metabarcoding to investigate honey bee foraging reveals limited flower use despite high floral availability. Scientific Reports, 7(1), 42838. https://doi.org/10.1038/srep42838 Delmas, E., Besson, M., Brice, M.-H., Burkle, L. A., Riva, G. V. D., Fortin, M.-J., … Poisot, T. (2019). Analysing ecological networks of species interactions. Biological Reviews, 94(1), 16–36. https://doi.org/10.1111/brv.12433 Dibble, A. C., Drummond, F. A., & Stack, L. B. (2020a). Bee visitation on flowers in maine, united states, reveals the relative attractiveness of plants through space and time: part I. Environmental Entomology. https://doi.org/10.1093/ee/nvaa028 Dibble, A. C., Drummond, F. A., & Stack, L. B. (2020b). Plant origin and other attributes impact bee forage patterns in a common garden study in maine, united states; part II. Environmental Entomology. https://doi.org/10.1093/ee/nvaa029 Djemiel, C., Grec, S., & Hawkins, S. (2017). Characterization of bacterial and fungal community dynamics by high-throughput sequencing (HTS) metabarcoding

during flax dew-retting. Frontiers in Microbiology, 8. https://doi.org/10.3389/fmicb.2017.02052 Domroese, M. C., & Johnson, E. A. (2017). Why watch bees? Motivations of citizen science volunteers in the Great Pollinator Project. Biological Conservation, 208, 40–47. https://doi.org/10.1016/j.biocon.2016.08.020 Dormann, C. F., Gruber, B., & Fründ, J. (2008). The bipartite package. Version 0.73. R Project for Statistical Computing, Vienna, Austria. Dunne, J. A., Williams, R. J., & Martinez, N. D. (2002). Network structure and biodiversity loss in food webs: Robustness increases with connectance. Ecology Letters, 5(4), 558–567. https://doi.org/10.1046/j.1461-0248.2002.00354.x Durant, J. L., & Otto, C. R. V. (2019). Feeling the sting? Addressing land-use changes can mitigate bee declines. Land Use Policy, 87, 104005. https://doi.org/10.1016/j.landusepol.2019.05.024 Erdtman, G. (1943). An Introduction to Pollen Analysis. chronica Botanica, waltham. Mass., uSa. Free, J. B. (1970). The flower constancy of bumblebees. Journal of Animal Ecology, 39(2), 395–402. JSTOR. https://doi.org/10.2307/2978 Gallai, N., Salles, J.-M., Settele, J., & Vaissière, B. E. (2009). Economic valuation of the vulnerability of world agriculture confronted with pollinator decline. Ecological Economics, 68(3), 810–821. https://doi.org/10.1016/j.ecolecon.2008.06.014 Galliot, J.-N., Brunel, D., Bérard, A., Chauveau, A., Blanchetête, A., Lanore, L., & Farruggia, A. (2017). Investigating a flower-insect forager network in a mountain grassland community using pollen DNA barcoding. Journal of Insect Conservation, 21(5–6), 827–837. https://doi.org/10.1007/s10841-017-0022-z Gao, T., Yao, H., Song, J., Zhu, Y., Liu, C., & Chen, S. (2010). Evaluating the feasibility of using candidate DNA barcodes in discriminating species of the large Asteraceae family. BMC Evolutionary Biology, 10(1), 324. https://doi.org/10.1186/1471-2148-10-324 Ginsberg, H. S. (1983). Foraging ecology of bees in an old field. Ecology, 64(1), 165– 175. https://doi.org/10.2307/1937338 Gloor, G. B., Macklaim, J. M., Pawlowsky-Glahn, V., & Egozcue, J. J. (2017). Microbiome datasets are compositional: And this is not optional. Frontiers in Microbiology, 8. https://doi.org/10.3389/fmicb.2017.02224 Goodell, K., & Thomson, J. D. (1996, June). Comparisons of pollen removal and deposition by honey bees and bumblebees visiting apple. In VII International Symposium on Pollination 437 (pp. 103-108). Goulson, D. (1999). Foraging strategies of insects for gathering nectar and pollen, and implications for plant ecology and evolution. Perspectives in Plant Ecology, Evolution and Systematics, 2(2), 185–209. https://doi.org/10.1078/1433-8319- 00070 Goulson, D., Nicholls, E., Botías, C., & Rotheray, E. L. (2015). Bee declines driven by combined stress from parasites, pesticides, and lack of flowers. Science, 347(6229). https://doi.org/10.1126/science.1255957 Gresty, C. E. A., Clare, E., Devey, D. S., Cowan, R. S., Csiba, L., Malakasi, P., … Willis, K. J. (2018). Flower preferences and pollen transport networks for cavity-nesting

solitary bees: Implications for the design of agri-environment schemes. Ecology and Evolution, 8(15), 7574–7587. https://doi.org/10.1002/ece3.4234 Harder, L. D. (1990). pollen removal by bumble bees and its implications for pollen dispersal. Ecology, 71(3), 1110–1125. JSTOR. https://doi.org/10.2307/1937379 Hawkins, J., Vere, N. de, Griffith, A., Ford, C. R., Allainguillaume, J., Hegarty, M. J., … Adams-Groom, B. (2015). Using DNA metabarcoding to identify the floral composition of honey: a new tool for investigating honey bee foraging preferences. PLOS ONE, 10(8), e0134735. https://doi.org/10.1371/journal.pone.0134735 Hebert, P. D. N., Cywinska, A., Ball, S. L., & deWaard, J. R. (2003). Biological identifications through DNA barcodes. Proceedings of the Royal Society of London. Series B: Biological Sciences, 270(1512), 313–321. https://doi.org/10.1098/rspb.2002.2218 Janzen, D. H. (1983). Seed and pollen dispersal by animals: Convergence in the ecology of contamination and sloppy harvest. Biological Journal of the Linnean Society, 20(1), 103–113. https://doi.org/10.1111/j.1095-8312.1983.tb01592.x Kane, N., Sveinsson, S., Dempewolf, H., Yang, J. Y., Zhang, D., Engels, J. M. M., & Cronk, Q. (2012). Ultra-barcoding in cacao (Theobroma spp.; Malvaceae) using whole chloroplast genomes and nuclear ribosomal DNA. American Journal of Botany, 99(2), 320–329. https://doi.org/10.3732/ajb.1100570 Kartzinel, T. R., Chen, P. A., Coverdale, T. C., Erickson, D. L., Kress, W. J., Kuzmina, M. L., Rubenstein, D. I., Wang, W., & Pringle, R. M. (2015). DNA metabarcoding illuminates dietary niche partitioning by African large herbivores. Proceedings of the National Academy of Sciences of the United States of America, 112(26), 8019–8024. JSTOR. Kearns, C. A., & Inouye, D. W. (1993). Techniques for pollination biologists. Techniques for Pollination Biologists. Retrieved from https://www.cabdirect.org/cabdirect/abstract/19940200882 Keller, A., Danner, N., Grimmer, G., Ankenbrand, M., Ohe, K. von der, Ohe, W. von der, … Steffan‐Dewenter, I. (2015). Evaluating multiplexed next-generation sequencing as a method in palynology for mixed pollen samples. Plant Biology, 17(2), 558–566. https://doi.org/10.1111/plb.12251 Kembel, S. W., Wu, M., Eisen, J. A., & Green, J. L. (2012). Incorporating 16S gene copy number information improves estimates of microbial diversity and abundance. PLOS Computational Biology, 8(10). https://doi.org/10.1371/journal.pcbi.1002743 Kennedy, P. L., DeBano, S. J., Bartuszevige, A. M., & Lueders, A. S. (2009). Effects of native and non-native grassland plant communities on breeding passerine birds: implications for restoration of northwest bunchgrass prairie. Restoration Ecology, 17(4), 515–525. https://doi.org/10.1111/j.1526-100X.2008.00402.x Kimoto, C., DeBano, S. J., Thorp, R. W., Rao, S., & Stephen, W. P. (2012). Investigating temporal patterns of a native bee community in a remnant north american bunchgrass prairie using blue vane traps. Journal of Insect Science, 12(108), 1– 23. https://doi.org/10.1673/031.012.10801

Klindworth, A., Pruesse, E., Schweer, T., Peplies, J., Quast, C., Horn, M., & Glöckner, F. O. (2013). Evaluation of general 16S ribosomal RNA gene PCR primers for classical and next-generation sequencing-based diversity studies. Nucleic Acids Research, 41(1), e1–e1. https://doi.org/10.1093/nar/gks808 Kraaijeveld, K., Weger, L. A. de, García, M. V., Buermans, H., Frank, J., Hiemstra, P. S., & Dunnen, J. T. den. (2015). Efficient and sensitive identification and quantification of airborne pollen using next-generation DNA sequencing. Molecular Ecology Resources, 15(1), 8–16. https://doi.org/10.1111/1755- 0998.12288 Kudo, G., & Ida, T. Y. (2013). Early onset of spring increases the phenological mismatch between plants and pollinators. Ecology, 94(10), 2311–2320. https://doi.org/10.1890/12-2003.1 Kuhlman, M., & Burrows, S. (2017). Checklist of bees (Apoidea) from a private conservation property in west-central Montana. Biodiversity Data Journal, 5. https://doi.org/10.3897/BDJ.5.e11506 Koh, I., Lonsdorf, E. V., Williams, N. M., Brittain, C., Isaacs, R., Gibbs, J., & Ricketts, T. H. (2016). Modeling the status, trends, and impacts of wild bee abundance in the United States. Proceedings of the National Academy of Sciences, 113(1), 140– 145. https://doi.org/10.1073/pnas.1517685113 Larson, D. L., Larson, J. L., & Buhl, D. A. (2018). Conserving all the pollinators: variation in probability of pollen transport among insect taxa. Natural Areas Journal, 38(5), 393. https://doi.org/10.3375/043.038.0508 Latta, R. G., Linhart, Y. B., Fleck, D., & Elliot, M. (1998). Direct and indirect estimates of seed versus pollen movement within a population of ponderosa pine. Evolution, 52(1), 61–67. https://doi.org/10.1111/j.1558-5646.1998.tb05138.x Losey, J. E., & Vaughan, M. (2006). the economic value of ecological services provided by insects. BioScience, 56(4), 311. https://doi.org/10.1641/0006- 3568(2006)56[311:TEVOES]2.0.CO;2 Lucas, A., Bodger, O., Brosi, B. J., Ford, C. R., Forman, D. W., Greig, C., … de Vere, N. (2018). Floral resource partitioning by individuals within generalised hoverfly pollination networks revealed by DNA metabarcoding. Scientific Reports, 8(1), 5133. https://doi.org/10.1038/s41598-018-23103-0 Macgregor, C. J., Kitson, J. J. N., Fox, R., Hahn, C., Lunt, D. H., Pocock, M. J. O., & Evans, D. M. (2019). Construction, validation, and application of nocturnal pollen transport networks in an agro-ecosystem: A comparison using light microscopy and DNA metabarcoding. Ecological Entomology, 44(1), 17–29. https://doi.org/10.1111/een.12674 Martins, A. C., Gonçalves, R. B., & Melo, G. A. R. (2013). Changes in wild bee fauna of a grassland in Brazil reveal negative effects associated with growing urbanization during the last 40 years. Zoologia (Curitiba), 30(2), 157–176. https://doi.org/10.1590/S1984-46702013000200006 Miller, P. D. (1982). Maize pollen: collection and enzymology. In Sheridan WF (ed) Maize for Biological Research. University Press, University of North Dakota, Grand Forks, pp 279-293

Montoya, J. M., Pimm, S. L., & Solé, R. V. (2006). Ecological networks and their fragility. Nature, 442(7100), 259–264. https://doi.org/10.1038/nature04927 Morandin, L. A., & Kremen, C. (2013). Hedgerow restoration promotes pollinator populations and exports native bees to adjacent fields. Ecological Applications, 23(4), 829–839. https://doi.org/10.1890/12-1051.1 Ollerton, J., Winfree, R., & Tarrant, S. (2011). How many flowering plants are pollinated by animals? Oikos, 120(3), 321–326. https://doi.org/10.1111/j.1600- 0706.2010.18644.x Ollerton, J., Price, V., Armbruster, W. S., Memmott, J., Watts, S., Waser, N. M., Totland, Ø., Goulson, D., Alarcón, R., Stout, J. C., & Tarrant, S. (2012). Overplaying the role of honey bees as pollinators: A comment on Aebi and Neumann (2011). Trends in Ecology & Evolution, 27(3), 141–142. https://doi.org/10.1016/j.tree.2011.12.001 Ostevik, K., Manson, J., & Thompson, J. (2010). Pollination potential of male bumble bees (Bombus impatiens): Movement patterns and pollen-transfer efficiency. Journal of Pollination Ecology, 2, 21–26. https://doi.org/10.26786/1920- 7603(2010)3 Pacini, E. (2015). Pollination. In Reference Module in Earth Systems and Environmental Sciences. Elsevier. https://doi.org/10.1016/B978-0-12-409548-9.09315-5 Pawluczyk, M., Weiss, J., Links, M. G., Egaña Aranguren, M., Wilkinson, M. D., & Egea-Cortines, M. (2015). Quantitative evaluation of bias in PCR amplification and next-generation sequencing derived from metabarcoding samples. Analytical and Bioanalytical Chemistry, 407(7), 1841–1848. https://doi.org/10.1007/s00216- 014-8435-y Pompanon, F., Deagle, B. E., Symondson, W. O. C., Brown, D. S., Jarman, S. N., & Taberlet, P. (2012). Who is eating what: Diet assessment using next generation sequencing. Molecular Ecology, 21(8), 1931–1950. https://doi.org/10.1111/j.1365-294X.2011.05403.x Pornon, A., Escaravage, N., Burrus, M., Holota, H., Khimoun, A., Mariette, J., Pellizzari, C., Iribar, A., Etienne, R., Taberlet, P., Vidal, M., Winterton, P., Zinger, L., & Andalo, C. (2016). Using metabarcoding to reveal and quantify plant-pollinator interactions. Scientific Reports, 6(1), 27282. https://doi.org/10.1038/srep27282 Porter, E. K. (1981). Origins and genetic nonvariability of the proteins which diffuse from maize pollen. Environmental Health Perspectives, 37, 53-59. Potter, C., Vere, N. de, Jones, L. E., Ford, C. R., Hegarty, M. J., Hodder, K. H., … Franklin, E. L. (2019). Pollen metabarcoding reveals broad and species-specific resource use by urban bees. PeerJ, 7, e5999. https://doi.org/10.7717/peerj.5999 Potts, S. G., Biesmeijer, J. C., Kremen, C., Neumann, P., Schweiger, O., & Kunin, W. E. (2010). Global pollinator declines: Trends, impacts and drivers. Trends in Ecology & Evolution, 25(6), 345–353. https://doi.org/10.1016/j.tree.2010.01.007 Rahl, M. (2008). Microscopic identification and purity determination of pollen grains. In M. G. Jones & P. Lympany (Eds.), Allergy Methods and Protocols (pp. 263–269). https://doi.org/10.1007/978-1-59745-366-0_22 Richards, A. J. (1997). Plant Breeding Systems. Garland Science.

Richardson, R. T., Lin, C.-H., Sponsler, D. B., Quijia, J. O., Goodell, K., & Johnson, R. M. (2015a). Application of ITS2 metabarcoding to determine the provenance of pollen collected by honey bees in an agroecosystem. Applications in Plant Sciences, 3(1), 1400066. https://doi.org/10.3732/apps.1400066 Richardson, R. T., Lin, C.-H., Quijia, J. O., Riusech, N. S., Goodell, K., & Johnson, R. M. (2015b). Rank-based characterization of pollen assemblages collected by honey bees using a multi-locus metabarcoding approach. Applications in Plant Sciences, 3(11), 1500043. https://doi.org/10.3732/apps.1500043 Richardson, R. T., Curtis, H. R., Matcham, E. G., Lin, C.-H., Suresh, S., Sponsler, D. B., Hearon, L. E., & Johnson, R. M. (2019). Quantitative multi-locus metabarcoding and waggle dance interpretation reveal honey bee spring foraging patterns in Midwest agroecosystems. Molecular Ecology, 28(3), 686–697. https://doi.org/10.1111/mec.14975 Richardson, R. T., Sponsler, D. B., McMinn‐Sauder, H., & Johnson, R. M. (2020). MetaCurator: A hidden Markov model-based toolkit for extracting and curating sequences from taxonomically-informative genetic markers. Methods in Ecology and Evolution, 11(1), 181–186. https://doi.org/10.1111/2041-210X.13314 Roberts, R. J., Carneiro, M. O., & Schatz, M. C. (2013). The advantages of SMRT sequencing. Genome Biology, 14(6), 405. https://doi.org/10.1186/gb-2013-14-6- 405 Roof, S. M., DeBano, S., Rowland, M. M., & Burrows, S. (2018). Associations between blooming plants and their bee visitors in a riparian ecosystem in eastern Oregon. Northwest Science, 92(2), 119. https://doi.org/10.3955/046.092.0205 Rowland, M. M., Bryant, L. D., Johnson, B. K., Noyes, J. H., Wisdom, M. J., & Thomas, J. W. (1997). The Starkey project: history, facilities, and data collection methods for ungulate research. Gen. Tech. Rep. PNW-GTR-396. Portland, OR: US Department of Agriculture, Forest Service, Pacific Northwest Research Station. 62 p, 396. Sajwani, A., Farooq, S. A., & Bryant, V. M. (2014). Studies of bee foraging plants and analysis of pollen pellets from hives in Oman. Palynology, 38(2), 207–223. https://doi.org/10.1080/01916122.2013.871652 Scheper, J., Reemer, M., Kats, R. van, Ozinga, W. A., Linden, G. T. J. van der, Schaminée, J. H. J., Siepel, H., & Kleijn, D. (2014). Museum specimens reveal loss of pollen host plants as key factor driving wild bee decline in The Netherlands. Proceedings of the National Academy of Sciences, 111(49), 17552– 17557. https://doi.org/10.1073/pnas.1412973111 Schrader, C., Schielke, A., Ellerbroek, L., & Johne, R. (2012). PCR inhibitors— Occurrence, properties and removal. Journal of Applied Microbiology, 113(5), 1014–1026. https://doi.org/10.1111/j.1365-2672.2012.05384.x Sickel, W., Ankenbrand, M. J., Grimmer, G., Holzschuh, A., Härtel, S., Lanzen, J., … Keller, A. (2015). Increased efficiency in identifying mixed pollen samples by meta-barcoding with a dual-indexing approach. BMC Ecology, 15(1), 20. https://doi.org/10.1186/s12898-015-0051-y Smart, M. D., Cornman, R. S., Iwanowicz, D. D., McDermott-Kubeczko, M., Pettis, J. S., Spivak, M. S., & Otto, C. R. V. (2017). A comparison of honey bee-collected

pollen from working agricultural lands using light microscopy and ITS metabarcoding. Environmental Entomology, 46(1), 38–49. https://doi.org/10.1093/ee/nvw159 Song, J., Shi, L., Li, D., Sun, Y., Niu, Y., Chen, Z., Luo, H., Pang, X., Sun, Z., Liu, C., Lv, A., Deng, Y., Larson-Rabin, Z., Wilkinson, M., & Chen, S. (2012). Extensive pyrosequencing reveals frequent intra-genomic variations of internal transcribed spacer regions of nuclear ribosomal DNA. PLOS ONE, 7(8). https://doi.org/10.1371/journal.pone.0043971 Taberlet, P., Coissac, E., Pompanon, F., Brochmann, C., & Willerslev, E. (2012). Towards next-generation biodiversity assessment using DNA metabarcoding. Molecular Ecology, 21(8), 2045–2050. https://doi.org/10.1111/j.1365- 294X.2012.05470.x Tang, M., Tan, M., Meng, G., Yang, S., Su, X., Liu, S., Song, W., Li, Y., Wu, Q., Zhang, A., & Zhou, X. (2014). Multiplex sequencing of pooled mitochondrial genomes— A crucial step toward biodiversity analysis using mito-metagenomics. Nucleic Acids Research, 42(22), e166–e166. https://doi.org/10.1093/nar/gku917 Thomas, A. C., Deagle, B. E., Eveson, J. P., Harsch, C. H., & Trites, A. W. (2016). Quantitative DNA metabarcoding: Improved estimates of species proportional biomass using correction factors derived from control material. Molecular Ecology Resources, 16(3), 714–726. https://doi.org/10.1111/1755-0998.12490 Valentini, A., Taberlet, P., Miaud, C., Civade, R., Herder, J., Thomsen, P. F., Bellemain, E., Besnard, A., Coissac, E., Boyer, F., Gaboriaud, C., Jean, P., Poulet, N., Roset, N., Copp, G. H., Geniez, P., Pont, D., Argillier, C., Baudoin, J.-M., … Dejean, T. (2016). Next-generation monitoring of aquatic biodiversity using environmental DNA metabarcoding. Molecular Ecology, 25(4), 929–942. https://doi.org/10.1111/mec.13428 Villanueva-Gutiérrez, R., & Roubik, D. W. (2016). More than protein? Bee–flower interactions and effects of disturbance regimes revealed by rare pollen in bee nests. Arthropod-Plant Interactions, 10(1), 9–20. https://doi.org/10.1007/s11829- 015-9413-9 Von Der Ohe, W., Persano Oddo, L., Piana, M. L., Morlot, M., & Martin, P. (2004). Harmonized methods of melissopalynology. Apidologie, 35(Suppl. 1), S18–S25. https://doi.org/10.1051/apido:2004050 Wang, X.-C., Liu, C., Huang, L., Bengtsson‐Palme, J., Chen, H., Zhang, J.-H., Cai, D., & Li, J.-Q. (2015). ITS1: A DNA barcode better than ITS2 in eukaryotes? Molecular Ecology Resources, 15(3), 573–586. https://doi.org/10.1111/1755- 0998.12325 Whitehead, D. R. (1983). Wind pollination: some ecological and evolutionary perspectives. Pollination biology, 97(08). Whittington, R., Winston, M. L., Tucker, C., & Parachnowitsch, A. L. (2004). Plant- species identity of pollen collected by bumblebees placed in greenhouses for tomato pollination. Canadian Journal of Plant Science, 84(2), 599–602. https://doi.org/10.4141/P02-192

Winfree, R., Aguilar, R., Vázquez, D. P., LeBuhn, G., & Aizen, M. A. (2009). A meta- analysis of bees’ responses to anthropogenic disturbance. Ecology, 90(8), 2068– 2076. https://doi.org/10.1890/08-1245.1 Wood, T. J., Holland, J. M., & Goulson, D. (2015). Pollinator-friendly management does not increase the diversity of farmland bees and wasps. Biological Conservation, 187, 120–126. https://doi.org/10.1016/j.biocon.2015.04.022

Appendix 1: Additional Figures and Tables for Chapter 2 This appendix includes two figures and one table that are supplementary materials to Chapter 2.

Figure A1.1: Distinct pollen types in pollen mixtures representing the proportion of each plant species’ contribution to a mixed pollen sample based on mass (i.e., “Pollen”) and the proportion of sequencing reads produced for each plant species for 3 replicates obtained using DNA metabarcoding (i.e., “SR 1-3”) when using the liberal threshold.

Figure A1.2: Proportion of pollen of each plant species in mixtures vs. proportion of sequence reads using the liberal threshold. Light gray line represents expected one-to-one relationship.

Table A1.1: Proportions of sequencing reads for additional plant species detected in single-species samples using the liberal threshold. Number of sequencing reads are in parentheses.

Single-Species Achillea Hypericum Lupinus Solidago Symphyotrichum Trifolium Sample millefolium perforatum sulphureus missouriensis spathulatum repens OA-1 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) OA-2 0 (0) 0.0003 (9) 0 (0) 0 (0) 0.0006 (20) 0 (0) OA-3 0 (0) 0.0001 (5) 0 (0) 0 (0) 0.0006 (24) 0 (0) SO-1 0.0068 (177) 0.0007 (18) 0.002 (52) 0 (0) 0.0017 (45) 0.0029 (75) SO-2 0.006 (131) 0 (0) 0.002 (43) 0.0007 (15) 0.0027 (60) 0.003 (66) SO-3 0.0069 (130) 0 (0) 0.0027 (51) 0 (0) 0.0026 (49) 0.0029 (54) PG-1 0.0021 (63) 0.0004 (12) 0.0005 (16) 0 (0) 0.0004 (12) 0 (0) PG-2 0.003 (94) 0.0003 (9) 0.0003 (10) 0.0006 (19) 0 (0) 0 (0) PG-3 0.0035 (159) 0.0003 (13) 0.0002 (11) 0.0005 (24) 0.0002 (7) 0 (0) TM-1 0 (0) 0 (0) 0 (0) 0 (0) 0.0009 (35) 0 (0) TM-2 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) TM-3 0 (0) 0.0003 (9) 0 (0) 0.0003 (9) 0 (0) 0 (0) VV-1 0 (0) 0 (0) 0.0004 (10) 0 (0) 0.0006 (17) 0 (0) VV-2 0 (0) 0.0001 (5) 0 (0) 0 (0) 0 (0) 0 (0) VV-3 0 (0) 0 (0) 0 (0) 0.0003 (8) 0 (0) 0 (0)

Appendix 2: Additional Tables for Chapter 3 This appendix includes two tables that are supplementary materials to Chapter 3.

Table A2.1: Number of individuals of each taxon observed visiting flowers at Threemile Canyon Farms (Threemile), USFS Starkey Experimental Forest and Range (Starkey), and The Nature Conservancy’s Zumwalt Prairie Preserve (Zumwalt), northeastern Oregon, USA. Subgenus names are in parentheses.

Bee Species (Subgenus) Threemile Starkey Zumwalt Total Agapostemon (Agapostemon) femoratus 15 0 0 15 Agapostemon (Agapostemon) texanus 4 0 0 4 Andrena (Euandrena) astragali 0 0 1 1 Andrena (Leucandrena) barbilabris 0 1 0 1 Andrena (Trachandrena) cyanophila 0 3 0 3 Andrena (Micrandrena) melanochroa 0 7 3 10 Andrena (Euandrena) nigrocaerulea 0 0 2 2 Andrena (Melandrena) nivalis 0 0 1 1 Andrena (Simandrena) pallidifovea 0 0 5 5 Andrena (Plastandrena) prunorum 2 0 2 4 Andrena (Tysandrena) sp. A 0 1 0 1 Andrena (Euandrena) sp. 1 0 0 2 2 Andrena (Melandrena) transnigra 0 0 1 1 Andrena (Andrena) vicinoides 0 0 2 2 Anthophora (Melea) bomboides 0 1 0 1 Anthophora (Micranthophora) curta 34 0 0 34 Anthophora (Lophanthophora) pacifica 0 0 1 1 Anthophora (Mystacanthophora) urbana 3 1 0 4 Bombus (Subterraneobombus) appositus 0 4 3 7 Bombus (Pyrobombus) bifarius 0 35 2 37 Bombus (Fervidobombus) californicus 0 10 16 26 Bombus (Pyrobombus) centralis 0 5 12 17 Bombus (Psithyrus) fernaldae 0 2 0 2 Bombus (Fervidobombus) fervidus 0 0 1 1 Bombus (Pyrobombus) flavifrons 0 15 1 16 Bombus (Separatobombus) griseocollis 4 0 0 4 Bombus (Pyrobombus) huntii 1 0 3 4 Bombus (Psithyrus) insularis 0 2 1 3 Bombus (Pyrobombus) mixtus 0 4 1 5 Bombus (Bombias) nevadensis 3 0 3 6 Bombus (Cullumanobombus) rufocinctus 0 0 5 5 Bombus (Pyrobombus) vosnesenskii 0 5 1 6 Coelioxys (Boreocoelioxys) rufitarsis 1 2 0 3 Colletes (Hyalinus) phaceliae 2 0 0 2 Diadasia nigrifrons 0 0 1 1 Halictus (Nealictus) farinosus 1 4 0 5

Halictus (Odontalictus) ligatus 0 13 6 19 Halictus (Protohalictus) rubicundus 0 1 4 5 Halictus (Seladonia) tripartitus 0 1 1 2 Hylaeus (Hylaeus) verticalis 0 1 0 1 Lasioglossum (Dialictus) albipenne 0 0 1 1 Lasioglossum (Evylaeus) cooleyi 0 1 0 1 Lasioglossum (Dialictus) incompletum 0 2 2 4 Lasioglossum (Dialictus) laevissimum 0 3 0 3 Lasioglossum (Dialictus) nevadense 0 3 0 3 Lasioglossum (Lasioglossum) olympiae 0 7 0 7 Lasioglossum (Evylaeus) ovaliceps 0 0 1 1 Lasioglossum (Dialictus) sedi 0 1 0 1 Lasioglossum (Hemihalictus) sp. 0 2 0 2 Lasioglossum (Dialictus) sp. 8 0 0 8 Lasioglossum (Hemihalictus) sp. 2 0 10 0 10 Lasioglossum (Lasioglossum) titusi 0 0 3 3 Megachile (Eutricharaea) apicalis 8 0 0 8 Megachile (Megachiloides) nevadensis 5 0 0 5 Megachile (Litomegachile) onobrychidis 8 0 2 10 Megachile (Xanthosarus) perihirta 0 3 2 5 Melissodes (Eumelissodes) bimatris 12 0 0 12 Melissodes (Callimelissodes) lupinus 1 0 1 2 Melissodes (Eumelissodes) lutulentus 5 0 0 5 Melissodes (Eumelissodes) microstictus 0 1 0 1 Melissodes (Eumelissodes) pallidisignatus 11 0 0 11 Melissodes (Heliomelissodes) rivalis 0 0 2 2 Melissodes (Eumelissodes) subagilis 3 0 0 3 Osmia (Melanosmia) atrocyanea 0 1 0 1 Osmia (Melanosmia) brevis 0 2 0 2 Osmia (Melanosmia) bucephala 0 2 0 2 Osmia (Acanthosmioides) integra 0 1 0 1 Osmia (Melanosmia) juxta 0 2 0 2 Osmia (Centrosmia) raritatis 0 1 0 1 Osmia sp. 2 0 1 0 1 Panurginus torchio 0 6 1 7 Perdita (Pygoperdita) wyomingensis 0 1 0 1 Sphecodes sp. 1 0 0 1 Svastra (Epimelissodes) obliqua 3 0 0 3 Triepeolus paenepectoralis 3 0 0 3 Xeromelecta (Melectomorpha) californica 1 0 0 1

100

Table A2.2: Plant species identified at Threemile Canyon Farms (Threemile), USFS Starkey Experimental Forest and Range (Starkey), and The Nature Conservancy’s Zumwalt Prairie Preserve (Zumwalt) using bee foraging observations (BFO), DNA metabarcoding using a regional reference database (MB-RDB), and DNA metabarcoding using a local reference database (MB- LDB). A “†” indicates a plant species that was detected using the ITS2 gene region, a “‡” indicates a plant species that was detected using the rbcL gene region, a “§” indicates a plant species that was identified using both ITS2 and rbcL gene regions, and a “¶” indicates a plant species that was identified using behavioral observations. Gray shading indicates a regional mismatch (i.e., a plant species assigned using the regional reference database that is not known to occur in the area). A “✔” indicates a plant species that was included in our ITS2 and/or rbcL reference library and a “✖” indicates a plant species that was missing from our ITS2 and/or rbcL reference library.

Threemile Starkey Zumwalt GenBank Plant Species BFO MB- MB- BFO MB- MB- BFO MB- MB- ITS2 rbcL RDB LDB RDB LDB RDB LDB Achillea millefolium ¶ † † ¶ † † ¶ † † ✔ ✔ Aconitum columbianum † † ✔ ✔ Agastache urticifolia † † † ✔ ✔ Arenaria congesta ¶ ✖ ✖ Arnica sororia ¶ † † ✔ ✔ Allium acuminatum † † ✔ ✔ Allium validum † † ✔ ✔ Antennaria microphylla † ✔ ✔ Arceuthobium † ✔ ✔ campylopodum Blepharipappus scaber † † † ✔ ✔ Buglossoides arvensis † † ✔ ✔ Calochortus eurycarpus ¶ ✖ ✔ Castilleja lutescens ¶ ✔ ✖ Castilleja miniata † ✔ ✔

101

Centaurea diffusa ¶ † ✔ ✔ Centaurea solstitialis ¶ § § † † ✔ ✔ Centaurea stoebe † † ✔ ✔ Chenopodium leptophyllum ‡ ✔ ✔ Chrysothamnus viscidiflorus ‡ ✔ ✔ Cirsium arvense ¶ † † ✔ ✔ Cirsium brevifolium ¶ ✖ ✖ Cirsium foliosum † § ✔ ✔ Cirsium vulgare ¶ † § ✔ ✔ Clarkia pulchella ¶ ✔ ✔ Collomia linearis ¶ ✔ ✔ Cornus sericea ‡ ✔ ✔ Crepis acuminata † ✔ ✔ Cryptantha torreyana † † ✔ ✔ Delphinium nuttallianum ¶ † † ✔ ✔ Draba verna ‡ ✔ ✔ Drymocallis glandulosa § ✔ ✔ Erigeron corymbosus ¶ ✔ ✔ Erigeron pumilus ¶ ✔ ✔ Erigeron speciosus ¶ † † ✔ ✔ Eriogonum heracleoides ¶ ¶ ✔ ✔ Eriogonum umbellatum † ✔ ✔ Eriophyllum lanatum † † ✔ ✔ Erythranthe guttata † ✔ ✔ Frasera albicaulis ¶ § § ✔ ✔ Gentiana affinis † † ✔ ✔

102

Geranium pusillum † † ✔ ✔ Geranium viscosissimum ¶ ✖ ✔ Geum macrophyllum † ✔ ✔ Geum triflorum † † ✔ ✔ Grindelia nana ¶ † ¶ † ✔ ✔ Gutierrezia sarothrae ¶ † † ✔ ✔ Helianthus annuus † † ✔ ✔ Heracleum maximum † † ✔ ✔ Heuchera cylindrica † ✔ ✔ Hieracium albiflorum † ✔ ✔ Hieracium cynoglossoides ¶ ✖ ✔ Hieracium scouleri † † ✔ ✔ Holodiscus discolor † ✔ ✔ Horkelia fusca † ✔ ✔ Hydrophyllum capitatum † ✔ ✔ Hypericum perforatum ¶ § § ✔ ✔ Kali turgidum ‡ ✔ ✔ Lactuca serriola † † † † ✔ ✔ Lithophragma parviflorum † ✔ ✔ Lupinus caudatus ¶ ✖ ✖ Lupinus leucophyllus ¶ ✔ ✔ Lupinus sericeus ¶ ✔ ✖ Lupinus lepidus † † † ✔ ✔ Lupinus sulphureus ¶ † ✔ ✔ Machaeranthera canescens ¶ † ✔ ✔ Medicago lupulina † ✔ ✔

103

Medicago sativa † † † † † ✔ ✔ Melilotus officinalis † ✔ ✔ Mentha arvensis ¶ † † ✔ ✔ Micranthes oregana † ✔ ✔ Microseris nutans † ✔ ✔ Microsteris gracilis ‡ ✔ ✔ Monardella odoratissima ¶ ‡ ‡ ✖ ✔ Nothocalais troximoides † ✔ ✔ Onopordum acanthium ¶ † † ✔ ✔ Orthocarpus tenuifolius ¶ § § ✔ ✔ Pedicularis groenlandica † † ✔ ✔ Penstemon confertus ‡ ✔ ✔ Penstemon procerus ¶ † ✔ ✔ Penstemon richardsonii † ✔ ✔ Perideridia gairdneri † † † † ✔ ✔ Phacelia hastata † † ✔ ✔ Phacelia heterophylla † † † ✔ ✔ Philadelphus lewisii † † ✔ ✔ Physocarpus malvaceus † † ✔ ✔ Plectris macrocera † ✔ ✔ Polemonium occidentale ¶ † † ✔ ✔ Potentilla arguta ¶ ✔ ✔ Potentilla gracilis ¶ † ¶ ✔ ✔ Potentilla flabellifolia § § § ✔ ✔ Potentilla recta ‡ ✔ ✔ Psoralidium lanceolatum ¶ † § ✔ ✔

104

Rhaponticum repens † ✔ ✔ Rosa gymnocarpa † ✔ ✔ Rosa nutkana ‡ ✔ ✔ Rosa woodsii † ✔ ✔ Salsola kali ¶ † † ✔ ✔ Sedum stenopetalum † ✔ ✔ Senecio integerrimus § ✔ ✔ Senecio serra ¶ † ✔ ✖ Sidalcea oregana ¶ † † ¶ † † ✔ ✔ Sisymbrium altissimum ¶ † † ¶ † † ✔ ✔ Sisymbrium loeselii † † ✔ ✔ Solanum dulcamara † ✔ ✔ Solidago canadensis † † † ✔ ✔ Solidago missouriensis ¶ † ¶ ✔ ✔ Spiraea betulifolia ‡ ‡ ✔ ✔ Symphyotrichum ¶ † † † † ✔ ✔ spathulatum Thermopsis montana ¶ § † ✔ ✔ Toxicoscordion venenosum ¶ † ✔ ✔ Tragopogon dubius ¶ † † ✔ ✔ Tribulus terrestris † ✔ ✔ Trifolium eriocephalum † ✔ ✔ Trifolium hybridum † ✔ ✔ Trifolium pratense ¶ § § † ✔ ✔ Trifolium repens † † ✔ ✔ Trifolium wormskioldii ¶ ✔ ✔

105

Triteleia grandiflora ¶ ✖ ✔ Vicia americana § § ✔ ✔ Vicia cracca ¶ ✔ ✔ Vicia villosa § † ✔ ✔

106

Table A2.3: Legends for Threemile Canyon Farms plant-pollinator networks created from behavioral observations, DNA metabarcoding using a regional reference database (MB-RDB), and DNA metabarcoding using a local reference database (MB-LDB) (Figure 3.2).

Network ID Bee Species 1 Agapostemon (Agapostemon) femoratus 2 Agapostemon (Agapostemon) texanus 3 Andrena (Plastandrena) prunorum 4 Anthophora (Micranthophora) curta 5 Anthophora (Mystacanthophora) urbana 6 Bombus (Separatobombus) griseocollis 7 Bombus (Pyrobombus) huntii 8 Bombus (Bombias) nevadensis 9 Coelioxys (Boreocoelioxys) rufitarsis 10 Colletes (Hyalinus) phaceliae 11 Halictus (Nealictus) farinosus 12 Lasioglossum (Dialictus) sp. 13 Megachile (Eutricharaea) apicalis 14 Megachile (Megachiloides) nevadensis 15 Megachile (Litomegachile) onobrychidis 16 Melissodes (Eumelissodes) bimatris 17 Melissodes (Callimelissodes) lupinus 18 Melissodes (Eumelissodes) lutulentus 19 Melissodes (Eumelissodes) pallidisignatus 20 Melissodes (Eumelissodes) subagilis 21 Sphecodes sp. 22 Svastra (Epimelissodes) obliqua 23 Triepeolus paenepectoralis 24 Xeromelecta (Melectomorpha) californica

Behavioral Observations (Figure 3.2a) Network ID Plant Taxa X1 Achillea millefolium X2 Centaurea diffusa X3 Centaurea solstitialis X4 Dieteria canescens X5 Gutierrezia sarothrae X6 Onopordum acanthium X7 Psoralidium lanceolatum X8 Salsola kali X9 Sisymbrium altissimum

107

MB-RDB (Figure 3.2b) Network ID Plant Taxa X1 Dieteria canescens X2 Rhaponticum repens X3 Achillea millefolium X4 Centaurea solstitialis X5 Centaurea stoebe X6 Gutierrezia sarothrae X7 Onopordum acanthium X8 Solanum dulcamara X9 Brassicaceae X10 Sisymbrium altissimum X11 Sisymbrium loeselii X12 Medicago sativa X13 Psoralidium lanceolatum X14 Tribulus terrestris X15 Salsola kali X16 Kali turgidum MB-LDB (Figure 3.2c) Network ID Plant Taxa X1 Achillea millefolium X2 Centaurea spp. X3 Centaurea diffusa X4 Centaurea solstitialis X5 Gutierrezia sarothrae X6 Onopordum acanthium X7 Plectritis macrocera X8 Sisymbrium altissimum X9 Sisymbrium loeselii X10 Medicago sativa X11 Salsola kali X12 Chrysothamnus viscidiflorus X13 Draba verna X14 Chenopodium leptophyllum

108

Table A2.4: Legends for the USFS Starkey Experimental Forest and Range plant- pollinator networks created from behavioral observations, DNA metabarcoding using a regional reference database (MB-RDB), and DNA metabarcoding using a local reference database (MB-LDB) (Figure 3.3).

Network ID Bee Species 1 Andrena (Leucandrena) barbilabris 2 Andrena (Trachandrena) cyanophila 3 Andrena (Micrandrena) melanochroa 4 Andrena (Thysandrena) sp. A 5 Anthophora (Melea) bomboides 6 Anthophora (Mystacanthophora) urbana 7 Bombus (Subterraneobombus) appositus 8 Bombus (Pyrobombus) bifarius 9 Bombus (Fervidobombus) californicus 10 Bombus (Pyrobombus) centralis 11 Bombus (Psithyrus) fernaldae 12 Bombus (Pyrobombus) flavifrons 13 Bombus (Psithyrus) insularis 14 Bombus (Pyrobombus) mixtus 15 Bombus (Pyrobombus) vosnesenskii 16 Coelioxys (Boreocoelioxys) rufitarsis 17 Halictus (Nealictus) farinosus 18 Halictus (Odontalictus) ligatus 19 Halictus (Protohalictus) rubicundus 20 Halictus (Seladonia) tripartitus 21 Hylaeus (Hylaeus) verticalis 22 Lasioglossum (Evylaeus) cooleyi 23 Lasioglossum (Dialictus) incompletum 24 Lasioglossum (Dialictus) laevissimum 25 Lasioglossum (Dialictus) nevadense 26 Lasioglossum (Lasioglossum) olympiae 27 Lasioglossum (Dialictus) sedi 28 Lasioglossum (Dialictus) sp. 29 Lasioglossum (Hemihalictus) sp. 2 30 Megachile (Xanthosarus) perihirta 31 Melissodes (Eumelissodes) microstictus 32 Osmia (Melanosmia) atrocyanea 33 Osmia (Melanosmia) brevis 34 Osmia (Melanosmia) bucephala 35 Osmia (Acanthosmioides) integra 36 Osmia (Melanosmia) juxta

109

37 Osmia (Centrosmia) raritatis 38 Osmia sp. 2 39 Panurginus sp. A 40 Perdita (Pygoperdita) wyomingensis

Behavioral Observations (Figure 3.3a) Network ID Plant Taxa X1 Achillea millefolium X2 Cirsium arvense X3 Cirsium vulgare X4 Collomia linearis X5 Erigeron corymbosus X6 Erigeron speciosus X7 Eriogonum heracleoides X8 Grindelia nana X9 Hypericum perforatum X10 Lupinus sulphureus X11 Mentha arvensis X12 Monardella odoratissima X13 Penstemon procerus X14 Polemonium occidentale X15 Potentilla arguta X16 Potentilla gracilis X17 Senecio serra X18 Sidalcea oregana X19 Solidago missouriensis X20 Symphyotrichum spathulatum X21 Thermopsis montana X22 Tragopogon dubius X23 Trifolium pratense X24 Trifolium spp. X25 Trifolium wormskioldii X26 Triteleia grandiflora X27 Vicia cracca MB-RDB (Figure 3.3b) Network ID Plant Taxa X1 Hieracium scouleri X2 Erythranthe guttata X3 Drymocallis glandulosa X4 Geum macrophyllum X5 Rosa gymnocarpa X6 Sedum stenopetalum X7 Micranthes oregana

110

X8 Allium acuminatum X9 Allium validum X10 Heracleum maximum X11 Perideridia gairdneri X12 Asteraceae X13 Achillea millefolium X14 Blepharipappus scaber X15 Centaurea stoebe X16 Cirsium arvense X17 Cirsium vulgare X18 Erigeron speciosus X19 Eriophyllum lanatum X20 Grindelia nana X21 Lactuca serriola X22 Microseris nutans X23 Senecio integerrimus X24 Solidago canadensis X25 Symphyotrichum spathulatum X26 Tragopogon dubius X27 Phacelia heterophylla X28 Polemonium occidentale X29 Lamiaceae X30 Mentha arvensis X31 Pedicularis groenlandica X32 Penstemon spp. X33 Penstemon richardsonii X34 Lupinus lepidus X35 Medicago sativa X36 Melilotus officinalis X37 Thermopsis montana X38 Trifolium spp. X39 Trifolium hybridum X40 Trifolium pratense X41 Trifolium repens X42 Vicia americana X43 Hypericum perforatum X44 Sidalcea oregana X45 Holodiscus discolor X46 Physocarpus malvaceus X47 Potentilla flabellifolia X48 Aconitum columbianum X49 Arceuthobium campylopodum X50 Heuchera cylindrica X51 Monardella odoratissima

111

X52 Spiraea betulifolia X53 Cornus sericea MB-LDB (Figure 3.3c) Network ID Plant Taxa X1 Allium acuminatum X2 Allium validum X3 Heracleum maximum X4 Perideridia gairdneri X5 Antennaria microphylla X6 Cirsium arvense X7 Achillea millefolium X8 Cirsium vulgare X9 Erigeron speciosus X10 Eriophyllum lanatum X11 Hieracium albiflorum X12 Lactuca serriola X13 Nothocalais troximoides X14 Senecio serra X15 Solidago missouriensis X16 Symphyotrichum spathulatum X17 Tragopogon dubius X18 Hydrophyllum capitatum X19 Phacelia hastata X20 Polemonium occidentale X21 Lamiaceae X22 Agastache urticifolia X23 Mentha arvensis X24 Pedicularis groenlandica X25 Penstemon procerus X26 Lupinus sulphureus X27 Medicago lupulina X28 Thermopsis montana X29 Trifolium spp. X30 Trifolium eriocephalum X31 Trifolium pratense X32 Trifolium repens X33 Vicia americana X34 Hypericum perforatum X35 Sidalcea oregana X36 Horkelia fusca X37 Physocarpus malvaceus X38 Potentilla spp. X39 Potentilla gracilis X40 Aconitum columbianum

112

X41 Lithophragma parviflorum X42 Microsteris gracilis X43 Monardella odoratissima X44 Penstemon confertus X45 Spiraea betulifolia X46 Rosa nutkana X47 Potentilla recta

113

Table A2.5: Legends for TNC’s Zumwalt Prairie Preserve plant-pollinator networks created from behavioral observations, DNA metabarcoding using a regional reference database (MB-RDB), and DNA metabarcoding using a local reference database (MB- LDB) (Figure 3.4).

Network ID Bee Species 1 Andrena (Euandrena) astragali 2 Andrena (Micrandrena) melanochroa 3 Andrena (Euandrena) nigrocaerulea 4 Andrena (Melandrena) nivalis 5 Andrena (Simandrena) pallidifovea 6 Andrena (Plastandrena) prunorum 7 Andrena (Melandrena) transnigra 8 Andrena (Euandrena) sp. 1 9 Andrena (Andrena) vicinoides 10 Anthophora (Lophanthophora) pacifica 11 Bombus (Subterraneobombus) appositus 12 Bombus (Pyrobombus) bifarius 13 Bombus (Fervidobombus) californicus 14 Bombus (Pyrobombus) centralis 15 Bombus (Fervisobombus) fervidus 16 Bombus (Pyrobombus) flavifrons 17 Bombus (Pyrobombus) huntii 18 Bombus (Psithyrus) insularis 19 Bombus (Pyrobombus) mixtus 20 Bombus (Bombias) nevadensis 21 Bombus (Cullmanobombus) rufocinctus 22 Bombus (Pyrobombus) vosnesenskii 23 Diadasia nigrifrons 24 Halictus (Odontalictus) ligatus 25 Halictus (Protohalictus) rubicundus 26 Halictus (Seladonia) tripartitus 27 Lasioglossum (Dialictus) albipenne 28 Lasioglossum (Dialictus) incompletum 29 Lasioglossum (Evylaeus) ovaliceps 30 Lasioglossum (Lasioglossum) titusi 31 Megachile (Litomegachile) onobrychidis 32 Megachile (Xanthosarus) perihirta 33 Melissodes (Callimelissodes) lupinus 34 Melissodes (Heliomelissodes) rivalis 35 Panurginus sp. A

114

Behavioral Observations (Figure 3.4a) Network ID Plant Taxa X1 Achillea millefolium X2 Arenaria congesta X3 Arnica sororia X4 Calochortus eurycarpus X5 Castilleja lutescens X6 Cirsium brevifolium X7 Clarkia pulchella X8 Delphinium nuttallianum X9 Erigeron pumilus X10 Eriogonum heracleoides X11 Frasera albicaulis X12 Geranium viscosissimum X13 Grindelia nana X14 Hieracium cynoglossoides X15 Lupinus caudatus X16 Lupinus leucophyllus X17 Lupinus sericeus X18 Lupinus spp. X19 Orthocarpus tenuifolius X20 Potentilla gracilis X21 Sidalcea oregana X22 Sisymbrium altissimum X23 Solidago missouriensis X24 Toxicoscordion venenosum MB-RDB (Figure 3.4b) Network ID Plant Taxa X1 Hieracium scouleri X2 Frasera albicaulis X3 Toxicoscordion venenosum X4 Perideridia gairdneri X5 Achillea millefolium X6 Arnica sororia X7 Blepharipappus scaber X8 Centaurea solstitialis X9 Cirsium foliosum X10 Grindelia nana X11 Helianthus annuus X12 Lactuca serriola X13 Solidago canadensis X14 Symphyotrichum spp. X15 Symphyotrichum spathulatum X16 Buglossoides arvensis

115

X17 Cryptantha torreyana X18 Phacelia spp. X19 Phacelia hastata X20 Phacelia heterophylla X21 Philadelphus lewisii X22 Gentiana affinis X23 Castilleja miniata X24 Agastache urticifolia X25 Orthocarpus tenuifolius X26 Sisymbrium altissimum X27 Lupinus lepidus X28 Medicago sativa X29 Trifolium pratense X30 Geranium pusillum X31 Vicia villosa X32 Sidalcea oregana X33 Geum triflorum X34 Potentilla flabellifolia X35 Rosa woodsii X36 Delphinium nuttallianum X37 Calochortus spp. MB-LDB (Figure 3.4c) Network ID Plant Taxa X1 Perideridia gairdneri X2 Achillea millefolium X3 Arnica sororia X4 Blepharipappus scaber X5 Centaurea solstitialis X6 Cirsium foliosum X7 Crepis acuminata X8 Helianthus annuus X9 Lactuca serriola X10 Solidago canadensis X11 Symphyotrichum spathulatum X12 Buglossoides arvensis X13 Cryptantha torreyana X14 Phacelia heterophylla X15 Philadelphus lewisii X16 Frasera albicaulis X17 Gentiana affinis X18 Sisymbrium altissimum X19 Orthocarpus tenuifolius X20 Agastache urticifolia X21 Lupinus lepidus

116

X22 Medicago sativa X23 Trifolium spp. X24 Vicia villosa X25 Geranium pusillum X26 Sidalcea oregana X27 Potentilla flabellifolia X28 Geum triflorum X29 Eriogonum umbellatum X30 Delphinium nuttallianum X31 Calochortus spp.