Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Intraspecific trait variation structures species-interaction networks : a case study with the host plant… Barbour, Matthew A. 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2016_november_barbour_matthew.pdf [ 5.44MB ]
Metadata
JSON: 24-1.0314221.json
JSON-LD: 24-1.0314221-ld.json
RDF/XML (Pretty): 24-1.0314221-rdf.xml
RDF/JSON: 24-1.0314221-rdf.json
Turtle: 24-1.0314221-turtle.txt
N-Triples: 24-1.0314221-rdf-ntriples.txt
Original Record: 24-1.0314221-source.json
Full Text
24-1.0314221-fulltext.txt
Citation
24-1.0314221.ris

Full Text

INTRASPECIFIC TRAIT VARIATION STRUCTURES SPECIES-INTERACTION NETWORKS: A CASE STUDY WITH THE HOST PLANT SALIX HOOKERIANA  by  Matthew A. Barbour  M.Sc. in Ecology, San Diego State University, 2011  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Zoology)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  September 2016  © Matthew A. Barbour, 2016   ii Abstract The strength and organization of species interactions determine the structure and dynamics of ecological communities. However, interactions do not occur between species per se – interactions occur between individuals. A rising challenge for ecology is to understand how intraspecific trait variation scales up to affect community structure and dynamics. My dissertation research has taken an empirical approach to addressing this challenge. Specifically, I conducted common garden experiments with the host plant Salix hookeriana in multiple environments to study the contributions of host-plant genetic and environmental variation to functional trait variation and, in turn, the structure of ecological interactions with its associated community. In Chapter 2, I show that Salix hookeriana exhibited genetic variation in its resistance to a diverse community of herbivorous arthropods. Rather than there being a single key trait that explained herbivore responses, I found that a range of plant-growth and leaf quality traits mediated the responses of different herbivore species and feeding guilds. In Chapter 3, I show that the effects of willow genetic variation on a community of insect herbivores cascades up to affect interactions between these herbivores and their insect parasitoids. Direct and indirect genetic effects resulted in distinct compositions of multi-trophic interactions associated with each host-plant genotype. When I simulated the additive effects of genetic variation on food-web complexity, I observed a strong positive relationship, indicating that intraspecific genetic and phenotypic variation can play a key role in structuring ecological networks, which may in turn affect community persistence. In Chapter 4, I show that host plant genotypic effects on the structure of foliar arthropod and ecotomycorrhizal communities were strong, despite variability in the biotic (ant-aphid interactions) and abiotic (wind exposure) environment of a coastal dune ecosystem. Taken together, my dissertation suggests that heritable trait variation can play a key   iii role in shaping the structure of species-interaction networks. By linking the ecological consequences of heritable trait variation to network structure, my dissertation paves the way for empirical and theoretical studies of the interplay between ecological and evolutionary processes in affecting the maintenance of biodiversity.    iv Preface A version of Chapter 2 was published in the journal, Functional Ecology under the following citation: Matthew A. Barbour, Rodriguez-Cabal, M. A., Wu, E. T., Julkunen-Tiitto, R., Ritland, C.  E., Miscampbell, A. E., Jules, E. S. and Crutsinger, G. M. (2015), Multiple plant traits shape the genetic basis of herbivore community assembly. Functional Ecology, 29: 995–1006. doi:10.1111/1365-2435.12409our. I designed the study as well as collected and analyzed all of the plant trait data, except for the quantification of leaf secondary metabolites (credit: R. Julkunen-Tiitto). I also collected and analyzed all of the arthropod community data. I wrote 95% of the manuscript with all co-authors contributing minor edits (5%). M. Rodriguez-Cabal and G. Crutsinger contributed more than the other co-authors to the editing process. E. Wu, E. Jules, and G. M. Crutsinger provided resources for and planted the common garden used in this study. C.E. Ritland and A.E. Miscampbell genotyped the plants in the common garden.   A version of Chapter 3 was published in the journal, Proceedings of the National Academy of Sciences of the USA under the following citation: Matthew A. Barbour, Fortuna, M. A., Bascompte, J., Nicholson, J.R., Julkunen-Tiitto, R., Jules, E. S. and Crutsinger, G. M. (2015), Genetic specificity of a plant-insect food web: Implications for linking genetic variation to network complexity. Proceedings of the National Academy of Sciences of the USA, 113: 2128–2133. doi:10.1073/pnas.1513633113. I designed the study and collected all of the insect galls from the field. J.R. Nicholson collected data on gall-parasitoid interactions in the lab. I analyzed all of the data and wrote 95% of the manuscript with all co-authors contributing minor edits (5%). M. Fortuna and J. Bascompte provided feedback on the conceptual framework of the manuscript and contributed more to the writing than other co-authors (except for G. Crutsinger). R. Julkunen-Tiitto, E. Jules, and G. Crutsinger contributed in the same ways as in Chapter 2 (see previous paragraph).  Chapter 4 has not been published yet. I designed both experiments and collected all of the field data, except for quantifying the root-associated ectomycorrhiza and bacteria (credit: Sonya Erlandson). I analyzed all of the data and wrote 95% of the manuscript with my committee members contributing minor edits (5%).     v Table of Contents  Abstract .......................................................................................................................................... ii Preface ........................................................................................................................................... iv Table of Contents ...........................................................................................................................v List of Tables ................................................................................................................................ xi List of Figures .............................................................................................................................. xii Acknowledgements .................................................................................................................... xiii Dedication .....................................................................................................................................xv Chapter 1: Introduction ................................................................................................................1 1.1 Background ........................................................................................................................ 1 1.2 General Approach and Study System ................................................................................ 4 Chapter 2: Multiple plant traits shape the genetic basis of herbivore community assembly .6 2.1 Introduction ........................................................................................................................ 6 2.2 Materials and Methods ....................................................................................................... 8 2.2.1 Study system ............................................................................................................... 8 2.2.2 Common garden .......................................................................................................... 9 2.2.3 Molecular methods.................................................................................................... 10 2.2.4 How do herbivore communities respond to host-plant genotype?............................ 10 2.2.4.1 Sampling ............................................................................................................ 10 2.2.4.2 Analyses ............................................................................................................. 11 2.2.5 How heritable are different host-plant traits? ........................................................... 12 2.2.5.1 Leaf quality ........................................................................................................ 12   vi 2.2.5.2 Architecture........................................................................................................ 14 2.2.5.3 Analyses ............................................................................................................. 15 2.2.6 Which plant traits account for herbivore community responses to host-plant genotype? .............................................................................................................................. 16 2.3 Results .............................................................................................................................. 18 2.3.1 How do herbivore communities respond to host-plant genotype?............................ 18 2.3.1.1 Community-level ............................................................................................... 18 2.3.1.2 Feeding guilds .................................................................................................... 21 2.3.2 How heritable are different host-plant traits? ........................................................... 24 2.3.2.1 Leaf quality ........................................................................................................ 24 2.3.2.2 Architecture........................................................................................................ 24 2.3.3 Which plant traits account for herbivore community responses to host-plant genotype? .............................................................................................................................. 26 2.3.3.1 Community-level ............................................................................................... 26 2.3.3.2 Feeding guilds .................................................................................................... 29 2.4 Discussion ........................................................................................................................ 29 2.4.1 How do herbivore communities respond to host-plant genotype?............................ 30 2.4.2 How heritable are different host-plant traits? ........................................................... 31 2.4.3 Which plant traits account for herbivore community responses to host-plant genotype? .............................................................................................................................. 32 2.4.4 Conclusions ............................................................................................................... 34 Chapter 3: Genetic specificity of a plant-insect food web: implications for linking genetic variation to network complexity .................................................................................................36   vii 3.1 Introduction ...................................................................................................................... 36 3.2 Materials and Methods ..................................................................................................... 41 3.2.1 Common garden experiment and plant traits ............................................................ 41 3.2.2 Quantifying the genetic specificity of the plant-insect food web ............................. 42 3.2.3 Simulating the additive effects of genetic variation on network complexity ........... 45 3.3 Results and Discussion .................................................................................................... 47 3.3.1 Quantifying the genetic specificity of the plant-insect food web ............................. 47 3.3.2 Simulating the additive effects of genetic variation on network complexity ........... 53 3.3.3 Conclusions ............................................................................................................... 56 Chapter 4: Host-plant genetic and environmental variation structure above and belowground communities in a coastal dune ecosystem ...........................................................57 4.1 Introduction ...................................................................................................................... 57 4.2 Materials and Methods ..................................................................................................... 60 4.2.1 Study site ................................................................................................................... 60 4.2.2 Experimental design.................................................................................................. 61 4.2.2.1 Ant-aphid experiment ........................................................................................ 62 4.2.2.2 Wind experiment ................................................................................................ 63 4.2.3 What is the relative importance of willow genotype vs. the biotic and abiotic environment in structuring associated communities? ........................................................... 64 4.2.4 Do willow genetic and environmental variation have different effects on above and belowground communities? .................................................................................................. 64 4.2.5 What are the potential mechanisms by which willow genetic and environmental variation affects community responses? ............................................................................... 66   viii 4.2.5.1 Plant traits .......................................................................................................... 66 4.2.5.2 Soil characteristics ............................................................................................. 67 4.2.6 Statistical analyses .................................................................................................... 69 4.2.6.1 Community responses ........................................................................................ 69 4.2.6.2 Plant traits .......................................................................................................... 70 4.2.6.3 Soil characteristics ............................................................................................. 71 4.2.6.4 Direct and indirect effects .................................................................................. 71 4.3 Results .............................................................................................................................. 73 4.3.1 What is the relative importance of willow genotype vs. the biotic and abiotic environment in structuring associated communities? ........................................................... 73 4.3.1.1 Ant-aphid experiment ........................................................................................ 73 4.3.1.2 Wind experiment ................................................................................................ 78 4.3.2 Do willow genetic and environmental variation have different effects on above and belowground communities? .................................................................................................. 82 4.3.3 What are the potential mechanisms by which willow genetic and environmental variation affects community responses? ............................................................................... 83 4.3.3.1 Ant-aphid experiment ........................................................................................ 83 4.3.3.2 Wind experiment ................................................................................................ 88 4.4 Discussion ........................................................................................................................ 93 4.4.1 What is the relative importance of willow genotype vs. the biotic and abiotic environment in structuring associated communities? ........................................................... 93 4.4.2 Do above and belowground communities differ in their responses to willow genetic and environmental variation? ................................................................................................ 94   ix 4.4.3 What are the potential mechanisms by which willow genetic and environmental variation affects community responses? ............................................................................... 95 4.4.3.1 Ant-aphid experiment ........................................................................................ 95 4.4.3.2 Wind experiment ................................................................................................ 96 4.4.4 Conclusions ............................................................................................................... 97 Chapter 5: Conclusion .................................................................................................................99 5.1 Ecological Consequences of Genetic Variation Within Populations ............................... 99 5.2 Ecological Consequences of Genetic Variation Between Populations .......................... 102 5.3 Eco-Evolutionary Feedbacks in Complex Food Webs .................................................. 103 5.4 Conclusion ..................................................................................................................... 104 References ...................................................................................................................................105 Appendices ..................................................................................................................................123 Appendix A ............................................................................................................................. 123 A.1 Microsatellite analysis............................................................................................... 123 A.2 Non-tannin phenolic compounds .............................................................................. 124 A.3 Diagram of minimum convex hull ............................................................................ 125 A.4 Heatmap of phenotypic trait correlations .................................................................. 126 A.5 Principal component analysis of leaf phenolic compounds ...................................... 127 A.6 Ordination of herbivore community response .......................................................... 128 Appendix B ............................................................................................................................. 130 B.1 Statistical models testing the genetic specificity of the plant-insect food web. ........ 130 B.2 Statistical models explaining insect food web responses .......................................... 132 B.3 Generalized linear models of leaf gall parasitism ..................................................... 133   x B.4 Relatedness and functional-trait diversity of willow genotypes ............................... 133 B.5 Sampling interactions in gall-parasitoid network...................................................... 134 B.6 Calculating quantitative-weighted linkage density (food-web complexity) ............. 135 B.7 Asymptotic vs. non-asymptotic models .................................................................... 135 B.8 Results for simulations of sampling effort and genetic variation.............................. 136 B.9 Assessing the accuracy of the asymptotic model ...................................................... 138 B.10 Structural equation model of food-web complexity ................................................ 139 Appendix C ............................................................................................................................. 142 C.1 Abundance responses of key arthropod guilds in ant-aphid experiment .................. 142 C.2 Abundance responses of key arthropod guilds in wind experiment.......................... 143 C.3 Principal components analysis of aboveground plant traits ...................................... 144 C.4 Principal components analysis of soil properties in wind experiment ...................... 145    xi List of Tables Table 2.1 Correlations among key herbivore species ................................................................... 20 Table 2.2 Correlations among herbivore feeding guilds ............................................................... 23 Table 2.3 Herbivore responses to plant traits................................................................................ 27 Table 4.1 Summary of statistical models in ant-aphid experiment ............................................... 75 Table 4.2 Summary of statistical models in wind experiment ...................................................... 79 Table 4.3 Redundancy analyses of above and belowground communities .................................. 93    xii List of Figures Figure 2.1 Herbivore community responses ................................................................................. 19 Figure 2.2 Herbivore feeding guild responses .............................................................................. 22 Figure 2.3 Heritable trait variation ................................................................................................ 25 Figure 3.1 Genetic specificity of trophic interactions ................................................................... 39 Figure 3.2 Conceptual model of how genetic variation affects food-web complexity ................. 40 Figure 3.3 Response of gall community ....................................................................................... 48 Figure 3.4 Response of gall-parasitoid interaction network ......................................................... 51 Figure 3.5 Mechanisms affecting leaf gall parasitism .................................................................. 52 Figure 3.6 Simulated effect of genetic variation on food-web complexity .................................. 54 Figure 4.1 Arthropod community responses in ant-aphid experiment ......................................... 76 Figure 4.2 Arthropod community responses in wind experiment................................................. 81 Figure 4.3 Compositional responses of above and belowground communities ............................ 83 Figure 4.4 Aphid, ant, and plant trait responses ........................................................................... 84 Figure 4.5 Mechanisms of community assembly in ant-aphid experiment .................................. 87 Figure 4.6 Soil characteristics and plant trait responses ............................................................... 89 Figure 4.7 Mechanisms of community assembly in wind experiment ......................................... 92        xiii Acknowledgements First and foremost, I’d like to thank my immediate family and best friends for their love and support before and during my PhD. Most of them don’t really know what I’m doing, but they continue to give me their support. For scientific inspiration, I would like to thank the late Bob Paine. His 1966 and 1969 papers in The American Naturalist inspired my interest in food webs. Kevin McCann’s 2012 book got me interested in theoretical aspects of food webs and shaped much of my thinking on this topic. The Department of Zoology at University of British Columbia (UBC), was a fantastic place to do my PhD and there are numerous faculty, graduate students, and staff to thank. My supervisor Greg Crutsinger taught me how to be a successful scientist and made me laugh a lot in the process. Both the times that I have agreed and disagreed with Greg have been fundamental in shaping my development as an ecologist. My committee members, Diane Srivastava and Mary O’Connor, have contributed substantially to my thinking as an ecologist as their approach to ecology is very different from Greg’s and from each other. Sally Otto’s teaching and patience has enabled me to develop basic skills in mathematical modeling (none of which I used in my dissertation chapters, but I will in the future!). With many graduate students and post docs I have had lots of laughs and stimulating conversations about ecology, but I’d especially like to thank Andrew MacDonald, Angelica Gonzalez, Seth Rudman, Mariano Rodriguez-Cabal, Bill harrower, Kat Anderson, and Norah Brown. For improving the quality of my dissertation research, I am grateful to many people for their time and effort. Eric Jules provided me with lab space and made me feel welcome during my summers of fieldwork at Humboldt State University. Eric Nelson, Andrea Pickart, and the staff of Humboldt Bay National Wildlife Refuge (U.S. Fish and Wildlife Service) facilitated experimental logistics and gave me permission to conduct field experiments at Lanphere Dunes. Elizabeth Wu and numerous   xiv undergraduates from Humboldt State University assisted with the establishment and maintenance of the main common garden at Humboldt Bay National Wildlife Refuge. Nathan and Mandy Ask provided me with a nice place to live during my fieldwork in Humboldt County. George Argus, John Bair, and Peter Haggard assisted with plant and arthropod identifications. Lindsay Mackas-Burns, Brendan Locke, Joshua Nicholson, Chris Greyson-Gaito, Arezoo Sootodeh, Sinikka Sorsa, Melissa DeSiervo, Jenell Jackson and Ruthie Espanol assisted with lab and fieldwork. Riitta Julkunen-Tiitto and Jordi Bascompte took a chance on collaborating with a first year PhD student and provided me with leaf chemistry data (Riitta) as well as a broader perspective on the structure and dynamics of ecological networks (Jordi). Greg Crutsinger, Mariano Cabal-Rodriguez, Jordi Bascompte, Miguel Fortuna, Marc Johnson, Cris Hochwender, Robyn Zerebecki, Chris Greyson-Gaito, Diane Srivastava, Mary O’Connor, Dominic Gravel and 4 anonymous reviewers commented on earlier drafts of the dissertation. For financial support, I thank UBC for awarding me a Biodiversity Research Integrative Training & Education (BRITE) Graduate Student Fellowship (Sep. 2011 – Aug. 2013), James Robert Thompson Fellowship (Sep. 2013 – Aug. 2014), and a Four-Year Fellowship (Sep. 2013 – Aug. 2016). I would also like to thank UBC for awarding me a VPRI Graduate Student Travel Fund. This award enabled me to conduct a 6-week lab rotation in Dr. Jordi Bascompte’s lab in Seville, Spain, which proved instrumental in helping me conceptually identify the links between intraspecific genetic variation and the structure of ecological networks. My field research was supported by grants to my advisor, Greg Crutsinger, from the Miller Institute for Basic Research in Science and a National Science Engineering & Research Council (NSERC) Discovery grant. Grants from my advisor as well as a Zoology Graduate Student Travel Award (from UBC) enabled me to attend numerous international conferences that have contributed substantially to my professional development.   xv Dedication I dedicate my dissertation to my friend, Rusty Ligon, and my fiancé, Ruthie Espanol. Rusty helped me realize that I could actually make a living studying nature and he was my role model as I pursued graduate school in ecology. Ruthie’s empathy, in addition to so many other wonderful things about her, helped me realize that pursuing my fascination with the natural world and trying to figure out how it works was the right career path for me. Without their support and guidance, I may have ended up as an investment banker.    1 Chapter 1: Introduction 1.1 Background A core goal of ecology is to understand the processes that shape the structure of species rich communities. The strength and organization of species interactions define the structure of ecological communities, which in turn is a key determinant of the population dynamics of species as well as the flow of energy through ecosystems (Paine 1966, 1969; Dunne 2006; McCann 2012; Bascompte & Jordano 2014). Network theory provides a powerful approach for quantifying community structure and making predictions about how the gain or loss of species will affect community dynamics (Dunne et al. 2002; Stouffer & Bascompte 2011; Rohr et al. 2014). For example, network theory has shown that several structural characteristics promote the persistence of multi-trophic communities (food webs), including weak interaction strengths (McCann 2000; McCann 2012), modularity (distinct groups of interconnected species, Stouffer & Bascompte 2011), and greater complexity (number of interactions per species, MacArthur 1955; Dunne et al. 2002; Kondoh 2003). Therefore, the structure of ecological communities can have fundamental consequences for the maintenance of biodiversity.  Network theory typically models ecological interactions as occurring between species; however, species do not interact per se – interactions occur between individuals. A species-level simplification is useful when populations consist of homogenous sets of individuals, all of which interact equally with individuals of different species the same way. Yet, we know from the field of evolutionary ecology that most populations are not homogenous and that intraspecific differences dictate the strength and direction of natural selection (Endler 1986). Network theory, however, has virtually ignored the role of intraspecific trait variation in shaping ecological   2 interactions in species rich communities (Ings et al. 2008; Rudolf & Lafferty 2011). Given the mounting empirical evidence that intraspecific variation can affect interspecific interactions (Murdoch et al. 2003; Post et al. 2008; Clark 2010) there is a clear need to understand if/when intraspecific trait variation scales up to affect the structure of ecological networks and, in turn, community dynamics (Bolnick et al. 2011; Moya-Laraño 2011).    Genetic variation within species is a key driver of intraspecific trait variation, which in turn determines interactions with associated species (Antonovics 1992; Whitham 2003). Common garden experiments are a powerful approach for quantifying genetic variation in functional traits (Lynch & Walsh 1998) as well as the genetic specificity of species interactions (Fritz 1995; Bailey et al. 2006; Whitham et al. 2012). A key finding that has emerged from common garden studies over the past several decades is that genetic variation within a variety of taxa can have cascading effects on the diversity and composition of associated communities across multiple trophic levels (Maddox & Root 1987, 1990; Fritz & Price 1988; Johnson 2008; Bassar et al. 2011; Whitham et al. 2012; Rudman et al. 2015). This prior work forms a clear expectation that intraspecific genetic variation is capable of scaling up to affect the strength and organization of multi-trophic interactions, and thus community dynamics, although this has yet to be explicitly tested.   In addition to genetics, the biotic and abiotic environment can play a fundamental role in the expression of an individual’s traits (e.g. phenotypic plasticity, Scheiner 1993). Phenotypic plasticity is considered to be an important driver of species interactions (Agrawal 2001; Fordyce 2006), although its relative importance compared to an individual’s genotype in determining   3 functional trait variation and, in turn, the structure of ecological communities is still unclear. Answering this question has important implications for determining if/when community ecologists should spend the extra effort to study the community-level effects of genetic variation (Chase & Knight 2003; Hersch-Green et al. 2011). Moreover, teasing apart genotypic and environmental effects will gives us a more detailed understanding of the mechanisms by which intraspecific trait variation shapes community structure.   If we want to understand the role that intraspecific differences play in structuring species assemblages, we have to identify the key traits mediating ecological interactions. Identifying these key traits can be challenging. Traits are often highly correlated, making it difficult to determine which trait (or suite of traits) associated community members are cueing in on (Agrawal & Fishbein 2006). Quantifying the heritability of functional traits, their covariation, as well as how these traits affect species interactions is crucial though for predicting the direction and tempo of eco-evolutionary dynamics (Moya-Laraño et al. 2012). Knowing the heritability and phenotypic plasticity of functional traits can also enable predictions about whether evolutionary processes, such as natural selection, are capable of affecting ecological dynamics. For example, the concentration of condensed tannins in leaves of narrowleaf cottonwood (Populus angustifolia) is a highly heritable trait that also has a strong influence on the diversity and composition of foliar arthropods and soil microbial communities (Whitham et al. 2006). Consequently, strong directional selection on leaf tannin content could rapidly alter the structure of cottonwood’s associated above and belowground communities. Given the growing awareness that evolutionary processes can affect ecological dynamics (Yoshida et al. 2003; Fussmann et al. 2007; Schoener 2011; Agrawal et al. 2012), identifying how genetic and environmental variation   4 affects intraspecific trait variation and species interactions will give insight to the dynamic interplay between ecological and evolutionary processes in structuring ecological communities (Johnson & Stinchcombe 2008; Hughes et al. 2008).   1.2 General Approach and Study System For my dissertation, I used a series of common garden experiments and detailed trait screens to study how genetic variation within a dominant species of host plant interacts with the environment to affect functional trait variation and, in turn, ecological interactions with the diverse assemblage of species it interacts with. The bulk of my dissertation consists of three chapters. In Chapter 2, I quantify the heritability of a diverse array of plant functional traits (40 traits in total) and determine the relative importance of these traits in determining the structure of the host plant’s associated community of herbivorous arthropods. In Chapter 3, I examine the cascading effects of heritable variation in plant traits on the structure of an associated insect food web of herbivorous galling insects and their parasitoids. In Chapter 4, I examine how natural variation in the biotic (ant-aphid mutualisms) and abiotic (wind exposure) environment of a coastal dune ecosystem affects the community-level responses of foliar arthropods and root-associated ectomycorrhiza and microbes to host-plant genetic variation.   For these experiments, I used the host plant Salix hookeriana (coastal willow). Salix hookeriana is a deciduous shrub (< 8 m) that occurs along the Pacific coast ranging from northern California to Alaska. This willow species grows primarily in meadows, floodplains, and coastal dunes, and is generally restricted to less than 100 m elevation (Argus 2013). As with other willows, S. hookeriana is dioecious and reproduces both sexually (wind and insect pollination; e.g., Sacchi   5 & Price 1988) and asexually through vegetative growth (Argus 2013). The genus Salix has been a model system for examining the role of host-plant genetics in shaping interactions with its associated community (e.g., Fritz & Price 1988; Hochwender & Fritz 2004) for a number of reasons. First, willows support a diverse community of arthropods that include many different feeding guilds, such as gallers, leaf miners, leaf chewers, xylem and phloem feeders (Roche & Fritz 1997; Sipura 1999). Second, there can be considerable genetic variation within willow populations (Brunsfeld, Soltis & Soltis 1991), with different genotypes displaying extensive phenotypic differences in morphology (Fritz & Price 1988) and phenolic chemistry (Nichols-Orians, Fritz & Clausen 1993). Finally, preference and performance of individual herbivore species has already been linked to some willow traits in other species (Matsuki & MacLean Jr. 1994; Björkman, Dalin & Ahrné 2008; Boeckler, Gershenzon & Unsicker 2011), which provides an informative background for S. hookeriana.   6 Chapter 2: Multiple plant traits shape the genetic basis of herbivore community assembly  2.1 Introduction For over two decades, researchers studying plant-herbivore interactions have been interested in how host-plant genetic variation affects associated arthropod communities. Early work by Fritz and Price (1988) with willow (Salix lasiolepis) and Maddox and Root (1987, 1990) with goldenrod (Solidago altissima) demonstrated that different plant genotypes can host unique combinations of herbivore species. Since then, greenhouse experiments, common garden studies, and field observations from a variety of host-plant systems have provided further evidence that plant genetic variation is an important driver of herbivore community assembly (reviewed in Whitham et al. 2012). Nevertheless, the specific traits mediating herbivore responses to different host-plant genotypes remains unclear, as most studies neglect to screen plant phenotypes in sufficient detail (Hughes et al. 2008; Hersch-Green, Turley & Johnson 2011). Consequently, we are lacking a mechanistic understanding of the role host-plant genetic variation plays in the assembly of herbivore communities for most systems.   Identifying the specific host-plant characteristics that shape herbivore community composition can be a complex task. A single herbivore species is often correlated with multiple plant traits (Agrawal 2005; Agrawal & Fishbein 2006) and different herbivore species within a community may exhibit divergent responses to the same traits (Agrawal 2004, 2005; Agrawal & Fishbein 2006). For example, studies of common milkweed (Asclepias syriaca) have shown that latex and   7 trichomes negatively affect chewing herbivores, whereas these same traits are either ineffective (latex) or positively (trichomes) associated with sap-sucking insects (Agrawal 2004, 2005; Agrawal & Fishbein 2006). Furthermore, traits other than those related to leaf quality (e.g., secondary metabolites, trichomes, leaf C:N) are often overlooked in plant-herbivore studies, but warrant further consideration. For example, aspects of plant architecture (e.g., biomass, height, branching complexity) can vary within host-plant species and also have strong effects on insect herbivores, particularly in woody plants (Carmona, Lajeunesse & Johnson 2011; Crutsinger et al. 2014). Therefore, we need studies that screen plant traits in detail at both the leaf and whole-plant level to understand the mechanisms underlying herbivore community responses (Hughes et al. 2008; Hersch-Green et al. 2011).   Identifying the plant phenotypes mediating herbivore responses is also a critical step toward a mechanistic understanding of plant-herbivore eco-evolutionary dynamics (Hersch-Green et al. 2011). For example, we know that temporal changes in the genetic composition of host-plant populations can directly affect the abundance of associated consumer species (Agrawal et al. 2013). Yet, predicting these consequences will require us to move beyond simply identifying plant genotype-herbivore associations to research that characterizes: (1) herbivore responses to host-plant traits, and (2) the magnitude of variation and heritability of these plant phenotypes (Geber & Griffen 2003). From there, we can build both a mechanistic understanding of the genetic basis to herbivore community assembly, as well as make predictions about the cascading effects of host-plant evolution on the species that feed upon them. To date, such a comprehensive examination is lacking for the majority of host-plant study systems (but see Agrawal 2005 and Johnson et al. 2009).   8 In this study, we used a large common garden experiment to examine arthropod herbivore community responses to genetic and phenotypic variation in the dominant host-plant species, Salix hookeriana. Specifically, we sought to address three questions: (1) How do herbivore communities respond to host-plant genotype? (2) How heritable are different host-plant traits? (3) Which plant traits account for herbivore community responses to host-plant genotype?  2.2 Materials and Methods 2.2.1 Study system Salix hookeriana (coastal willow) is a deciduous shrub (< 8 m) that occurs along the Pacific coast ranging from northern California to Alaska. This willow species grows primarily in meadows, floodplains, and coastal dunes, and is generally restricted to less than 100 m elevation (Argus 2013). As with other willows, S. hookeriana is dioecious and reproduces both sexually (wind and insect pollination; e.g., Sacchi & Price 1988) and asexually through vegetative growth (Argus 2013).   The genus Salix has been a model system for examining the role of host-plant genetics in shaping plant-herbivore interactions (e.g., Fritz & Price 1988; Hochwender & Fritz 2004) for a number of reasons. First, willows support a diverse community of arthropods that include many different feeding guilds, such as gallers, leaf miners, leaf chewers, xylem and phloem feeders (Roche & Fritz 1997; Sipura 1999). Second, there can be considerable genetic variation within willow populations (Brunsfeld, Soltis & Soltis 1991), with different genotypes displaying extensive phenotypic differences in morphology (Fritz & Price 1988) and phenolic chemistry (Nichols-Orians, Fritz & Clausen 1993). Finally, preference and performance of individual herbivore   9 species has already been linked to some willow traits in other species (Matsuki & MacLean Jr. 1994; Björkman, Dalin & Ahrné 2008; Boeckler, Gershenzon & Unsicker 2011), which provides an informative background for S. hookeriana.  2.2.2 Common garden In February 2009, we established a common garden experiment consisting of clones from 27 different individuals of S. hookeriana (‘willow’ hereafter) at Humboldt Bay National Wildlife Refuge (HBNWR) (40°40'53"N, 124°12'4"W) near Loleta, California, USA. We haphazardly chose willow individuals (13 males, 14 females) from a single population growing locally around Humboldt Bay in both riparian areas (23 of 27) and dune swales (4 of 27) and subsequently genotyped them using microsatellite markers (see Molecular Methods below). We propagated clonal replicates of each individual using 25 cm cuttings that had been soaked in water for two weeks and planted directly into the ground in two hectares of a former cattle pasture at HBNWR. We planted cuttings in a completely randomized design with 25 replicates per willow individual (27 individuals × 25 replicates = 675 willows total), and cuttings spaced 3 m apart in a 45 m × 135 m grid. Each cutting was surrounded by a 1 × 1 m square of heavy-duty weed cloth to prevent vegetation growth in the immediate area. A 2.5 m tall fence was built around the experiment to exclude deer and cattle. Willows in our garden began flowering in February and reached their peak growth in late July to early August. During this study, willows had reached 2-3 m in height.    10 2.2.3 Molecular methods To confirm that willow individuals were genetically unique, we genotyped each individual using two microsatellite loci, SB80 and SB194 (Barker et al. 2003). Polymerase chain reaction (PCR) amplifications were performed in 10 µl reaction volumes containing 5 ng DNA, 1 pmol each of forward and reverse primers, 0.5 pmol M13 IRD-labeled primer, 200 µM dNTP (New England Biolabs, Ipswitch, MA, USA), 1 X Paq5000 PCR buffer (Agilent Technologies Canada Inc., Toronto, ON), 1 U Paq5000 DNA polymerase (Agilent Technologies Canada Inc.) and 2.0 mM MgSO4. Cycling conditions were 94 °C/2 min, 35 cycles of 94 °C/40 s, 54 °C/1 min, 72 °C/1 min, 72°C/10 min. The PCR products were analyzed on a LiCor 4200 automatic sequencer using 5.5% polyacrylamide gels (KBplus, LiCor Biotechnology) and scored using RFLPscan (LiCor Biotechnology, Lincoln, NE) due to their tetraploidy. Of the 27 individuals collected, 26 were found to be genetically unique and were used in this study (13 males, 13 females; Appendix A.1).    2.2.4 How do herbivore communities respond to host-plant genotype? 2.2.4.1 Sampling In July 2011, we used two techniques to sample the herbivore community on about five randomly chosen individuals of each of the 26 genotypes (n = 132, range = 4 – 7 for each genotype). For mobile herbivores, we vacuumed the entire crown of each willow using a modified leaf blower/vacuum (Craftsman 25 cc 2-cycle; Sears Holding Corporation, Hoffman Estates, Illinois, USA) with a fine insect net attached. We brought samples to the lab immediately where we counted each individual and identified them to species or morphospecies under a dissecting scope. For sedentary herbivores, we visually surveyed the entire shrub for   11 different species of galls (leaf and stem) and leaf mines. All herbivores were further assigned to one of the following feeding guilds: gallers, leaf miners, leaf chewers, xylem feeders and phloem feeders. To score damage from leaf chewers, we haphazardly selected five shoots per plant. Starting with the first fully expanded leaf on each shoot, we visually assigned damage scores to every other leaf for six leaves. We scored each leaf to one of 11 damage categories based on percent leaf area removed (0, 1-5, 5-10, 10-20, 20-30…90-100%). The same observer (MAB) scored all damage to maintain consistency across samples. We averaged damage scores for each shoot and then for all six shoots to obtain a single estimate of percent leaf area removed per replicate willow. 2.2.4.2 Analyses To examine how the herbivore community responded to willow genotype, we used separate one-way ANOVAs (with “stats” package in R; R Core Team 2013) to test for differences in the following responses: total richness, abundance, rarefied richness (using individual-based rarefaction; Gotelli & Colwell 2001), evenness (1E = exp(Shannon entropy)/richness; Tuomisto 2012), and percent leaf area removed (PLAR). Total richness, abundance, and PLAR were log-transformed and evenness was logit-transformed prior to analysis to improve normality and reduce heteroscedasticity. We also used separate generalized linear models (GLMs) to test for differences in abundance of several herbivore species and feeding guilds among genotypes (with “MASS” package in R). GLMs were appropriate because they account for response variables with non-normal distributions and heteroscedasticity that were not improved by transformations (O’Hara & Kotze 2010). To test for differences in community composition among willow genotypes, we normalized our community data (site-by-species matrix) using the chord transformation (sum of squared species relative abundances equal to one for each sample;   12 Legendre & Gallagher 2001) and conducted a redundancy analysis (RDA, 1000 permutations; with “vegan” package in R). RDA is analogous to an ANOVA on pairwise community dissimilarity values. Lastly, we calculated Pearson’s r (with “psych” package in R) to determine whether individual species and feeding guilds exhibited correlated responses among willow shrubs (phenotypic correlations, n = 131) and genotypes (genetic correlations, n = 26). Phenotypic correlations were estimated using the abundance (or damage for PLAR) of each species or feeding guild observed on each shrub, whereas genetic correlations were estimated from the mean abundance (or damage for PLAR) of each species or feeding guild found on each genotype.   2.2.5 How heritable are different host-plant traits? We measured 40 different plant traits that have been linked to herbivore preference and performance on willows and other host-plant species (Lawton 1983; Matsuki & MacLean Jr. 1994; Cornelissen et al. 2003; Björkman et al. 2008; Barbehenn & Constabel 2011; Boeckler et al. 2011). These traits were grouped into two larger categories, including leaf quality (36 traits) and plant architecture (four traits).  2.2.5.1 Leaf quality Phenolics are among the most abundant secondary metabolites in leaves of species within the family Salicaceae (Palo 1984) and have been shown to influence the preference and performance of several species of leaf chewing beetles (Family: Chrysomelidae) and sawflies (Family: Tenthridinidae) that specialize on willows (e.g., Tahvanainen, Julkunen-Tiitto & Kettunen 1985; Roininen & Tahvanainen 1989). We measured seven different types of phenolic compounds:   13 condensed tannins (two types), salicylates (eight types), phenolic acids (eight types), flavones (seven types), flavonols (three types), flavanones (eriodictyol 7-glycoside), and flavanonols (two types). To measure phenolics, we collected two fully expanded and undamaged leaves from about five shrubs of each genotype (n = 140, range = 4-7) in early August of 2012. Leaves were stored in paper coin envelopes and allowed to air-dry at room temperature until they could be analyzed (Julkunen-Tiitto & Sorsa 2001). Leaf samples were then ground dry and extracted with 100% methanol prior to high-performance liquid chromatography (HPLC)(Agilent, Series 1100, Agilent Technologies, Waldbronn, Germany) analysis of salicylates, phenolic acids, and flavonoids (Nybakken & Julkunen-Tiitto 2013). We identified phenolic metabolites by comparing their retention times and UV spectrum to standards (Appendix A.2). After HPLC runs, we quantified condensed tannin content using re-dissolved methanol extracts (soluble condensed tannins) and dried extraction residue (insoluble condensed tannins)(Nybakken & Julkunen-Tiitto 2013). Methodological details for leaf phenolic processing and extraction are given in Nybakken and Julkunen-Tiitto (2013).    In addition to our extensive characterization of the phenolic profiles of different genotypes, we measured other putatively important traits that could shape leaf quality for herbivores, including specific leaf area (SLA), water content, trichome density, percent carbon (C) and nitrogen (N), and C:N. For SLA, water content, and trichome density, we excised a single fully expanded and undamaged leaf from an average of five replicates of each genotype in July 2012 (n = 137, range = 4-7). We placed leaf samples into separate plastic bags within a cooler and immediately brought them back to the lab. We then weighed leaves to obtain fresh mass (g), digitally scanned them to measure leaf area (mm2) using ImageJ (Abràmoff, Magalhães & Ram 2004), and oven-  14 dried them at 60° C for 72 hours to obtain dry weight (g)(Cornelissen et al. 2003). We calculated SLA as leaf area/dry weight (Cornelissen et al. 2003). Leaf water content was calculated as the (fresh weight - dry weight)/dry weight (Munns & PrometheusWiki Contributors 2010). To measure trichome density, we counted the number of trichomes along an 11 mm x 1 mm transect in the center of the leaf, half-way between the leaf edge and the mid-vein, under a dissecting scope. To measure percent C and N, we collected ten fully expanded and undamaged leaves from the outer crown of an average of five replicates of each willow genotype in July 2010 (n = 130, range = 4-6). Leaves were air-dried and ground to a fine powder using a ball mill (SPEX, SamplePrep Mixer/Mill 8000D, Metuchen, New Jersey, USA). Subsamples of each material were then analyzed for percent C and N on an elemental analyzer (NC 2500 Carlo-Erba, Milan, Italy) using acetanilide (10.36% N and 71.09% C) as a reference standard. Shrubs sampled for percent C and N did not correspond with the same replicates sampled for other plant traits; therefore, we used the mean values for each genotype for calculating phenotypic correlations with other plant traits (further details in Analyses below) and for use in multiple regression analyses (further details in Which plant traits account for herbivore responses?).  2.2.5.2 Architecture Plant architectural traits included plant size, plant height, foliage density, and fractal dimension (an index of architectural complexity). We measured architectural traits by setting up a white tarp (5.5 m by 7.6 m) as a backdrop behind an average of five replicates per genotype (n = 132, range = 4 – 7) in late July 2011. We then took a photo on a tripod with a standard focal length (no zoom) from a standardized position (4 m distance, facing SW direction). Using ImageJ, we first removed shadows created by the foliage and then converted photos to black-and-white   15 images. We estimated plant height as the vertical height of the shrub in each image and plant size as the total two-dimensional area (m2) covered by the shrub in each image using a known scale. We calculated foliage density using plant size divided by the minimum convex hull area of the plant. The minimum convex hull represents a connected series of straight segments convexly enclosing all of the foreground pixels in our plant images (Appendix A.3). To calculate fractal dimension, we used the box counting method incorporated in the FracLac plugin for ImageJ. Fractal dimension is an index of complexity that measures how detail in a pattern changes with the scale of measurement. This architectural trait is also known to display heritable variation among Populus hybrids (Bailey et al. 2004) and can influence the abundance and size-distribution of arthropods on plants (Morse et al. 1985). 2.2.5.3 Analyses We used separate restricted maximum likelihood (REML) models to test for differences in plant traits among willow genotypes (with “nlme” package in R). We specified plant genotype as a random effect in all models and evaluated its significance using a likelihood ratio test. We did not include plant sex in our model because exploratory analyses showed that it was only weakly associated with a couple of salicylate compounds that were unimportant in affecting herbivore responses. Traits were transformed as needed to improve normality and reduce heteroscedasticity. To calculate the broad-sense heritability of plant traits, we used the equation: H2 = VG / VP, where VG is the total genotypic variance among clones, and VP is the total phenotypic variance, calculated as the sum of the residual and genetic variance (Lynch & Walsh 1998). Broad-sense heritability values range between 0-1, where values close to zero indicate low heritability (i.e., the trait is strongly influenced by the environment), and values close to 1 indicate high heritability (i.e., the trait is strongly controlled by underlying genetic variation). We   16 also calculated phenotypic (range of n = 115-140 shrubs) correlations (Pearson’s r) between all plant traits. We explored phenotypic trait correlations (Appendix A.4) to determine how to mitigate the effects of plant trait multicollinearity on multiple regression analysis (further details in Which plant traits account for herbivore community responses to host-plant genotype?).    2.2.6 Which plant traits account for herbivore community responses to host-plant genotype? We used multiple regression analyses to identify the host-plant traits that best accounted for herbivore community responses; however, we first had to mitigate the effects of multicollinearity. We used three different methods to reduce multicollinearity. For leaf phenolic chemistry, we conducted separate principle components analysis (PCA, with “labdsv” package in R) on the following groups of highly correlated compounds (Appendices A.4 and A.5): salicylates/condensed tannins, phenolic acids, flavones/flavonols (flavonoids), and flavanones/flavanonols (miscellaneous flavonoids). Performing separate PCAs allowed us to interpret the relationships between different classes of phenolic compounds and the herbivore community. Prior to PCA, we first transformed phenolics as necessary to linearize correlated relationships and then standardized each trait (mean = 0, SD = 1) to give them each equal weight in the analysis. We used scree plots and tables of variable loadings to select representative principal components (Appendix A.5). When certain pairs of traits were highly correlated with each other (0.4 < |r| < 0.8), we used the residuals from a linear regression of the two traits as a new predictor variable that was no longer correlated with the other trait (Graham 2003). These   17 trait pairs included: plant size and height (r = 0.59, P < 0.001), plant size and foliage density (r = 0.47, P < 0.001), as well as SLA and water content (r = 0.60, P < 0.001). In two cases, pairs of traits scaled closely with one another (|r| > 0.80), so we retained the trait that had a more intuitive ecological interpretation and discarded the other. Therefore, we retained plant size instead of fractal dimension (r = 0.85, P < 0.001), and kept C:N instead of N content (r = -0.97, P < 0.001). The three different methods we used to reduce multicollinearity resulted in 12 predictor variables. Leaf quality traits included: salicylate/tannin PC1, phenolic acid PC1-2, flavonoid PC1-2, miscellaneous flavonoids PC1 (Appendix S5), water content, SLA residuals, and C:N. Plant architectural traits included plant size, height residuals, and foliage density residuals. Finally, we conducted variance inflation factor (VIF) analysis on these 12 predictor variables (with “car” package in R) to calculate how much of the variance of an estimated regression coefficient is increased due to collinearity. All VIF values were < 1.8, indicating that multicollinearity had only a minor influence on our subsequent multiple regression analyses (Dormann et al. 2013).  Using this subset of predictor variables, we used multiple regression with forward model selection to identify the key traits accounting for herbivore responses. We restricted these analyses to herbivore responses that varied significantly among willow genotypes (P < 0.05). We used the forward model selection approach advocated by Blanchet et al. (2008), which prevents inclusion of spurious variables (i.e., inflated Type 1 error) and overestimation of explained variance (i.e., R2). This method first tested whether the full model, which included all 12 predictor variables, was significant (P < 0.05). We then proceeded with forward model selection using two stopping criteria: (1) P < 0.05 for including a variable in the model, (2) the adjusted R2   18 calculated on the full model. Whenever forward selection identified a variable that brought one or the other criterion over the fixed threshold, the variable was rejected, and the procedure stopped. To assess the relative importance of each predictor variable in the final model, we calculated the change in explanatory variance when a variable was removed (R2, with “rockchalk” package in R). After identifying a final model, we then used sequential sum-of-squares (i.e., Type 1 SS) to test whether including genotype as a factor still had a significant effect. If it did, this indicated that we either did not identify all of the relevant plant traits or that our study failed to capture some other important interaction (e.g., competition or predation) mediated by plant genotype.  2.3 Results 2.3.1 How do herbivore communities respond to host-plant genotype? 2.3.1.1 Community-level Total herbivore abundance (F25,105 = 1.64, P = 0.044; Fig. 2.1A) and community composition (F25,105 = 1.62, P = 0.001; Fig. 2.1C) exhibited strong responses to willow genotype, whereas herbivore richness (F25,105 = 1.33, P = 0.162; Fig. 2.1B), rarefied richness (F25,105 = 1.11, P = 0.348), and evenness did not (F25,105 = 1.40, P = 0.123). Herbivore abundance varied 3.5-fold among clones, ranging from an average of 24 to 84 individuals between the most disparate genotypes (Fig. 2.1A). Willow genotype explained 27.3% of the variance in community composition (Fig. 2.1C), with differences driven primarily by two leaf miners (weevil Tachyerges salicis, moth Caloptilia sp.), two leaf gallers (midge Iteomyia salicisverruca; mite Aculops tetanothrix), and a xylem feeding leaf hopper (Cicadellidae nymph sp. 1)(Appendix A.6). Of these species, the two leaf miners (T. salicis, 225,105 = 80.62, P < 0.001; Caloptilia sp.,   19 225,105 = 56.62, P < 0.001) and two gallers (I. salicisverruca, 225,105 = 63.62, P < 0.001; A. tetanothrix, 225,105 = 54.73, P = 0.001) varied between 3.7 and 10-fold in their abundance among willow genotypes, whereas Cicadellidae nymph sp.1 exhibited only a marginally significant response (225,105 = 35.96, P = 0.072). While Caloptilia and I. salicisverruca exhibited a positive phenotypic (i.e., shrub level) correlation (r = 0.20, P = 0.020; Table 2.1), no species pairs displayed correlated responses among the different willow genotypes (Table 2.1).  Figure 2.1 Herbivore community responses  Herbivore community responses to 26 different genotypes of Salix hookeriana growing in a common garden. Community-level variables included: (A) total richness, (B) total abundance, and (C) an ordination of community composition based on Euclidean distances of chord-transformed community data in which each axis represents the percent variance explained by the corresponding axis from redundancy analysis (RDA). For (A) and (B), genotypes are ordered based on total herbivore abundance, with circles and error bars representing means and SEs, respectively. For (C), the position of each letter corresponds to the centroid for −0.6 −0.4 −0.2 0.0 0.2 0.4 0.6−0.6−0.4−0.20.00.20.4CABDEFGHIJKLMNOPQRSTUVWXY Z8101214161820No. of species20406080100XK L RASDHGVBUQECNOY I J Z FMP TWS. hookeriana  genotypeNo. of individualsRDA 1 (7.4%)RDA 2 (5.1%)(A)(B)(C)  20 each genotype and the ellipses represent the SE of the centroid’s position. The ellipses of 5 of 26 genotypes are highlighted to illustrate the differences in herbivore community composition along these axes.     Table 2.1 Correlations among key herbivore species  Tachyergesa Caloptiliab Iteomyiaa Aculopsa Tachyerges salicisa 1 0.04 -0.08 -0.11 Caloptilia sp.b 0.15 1 0.20 0.00 Iteomyia salicisverrucaa -0.30 -0.11 1 0.09 Aculops tetanothrixa -0.07 0.01 0.11 1  Pearson correlations (r) of dominant herbivore species occuring on Salix hookeriana. Italicized values below the diagonal represent genetic correlations (n = 26), while values above the diagonal are phenotypic correlations (n = 131). Notes: alog(x+1) transformed; bsquare-root transformed. Statistically significant correlations (P < 0.05) are indicated in boldface type.   21  2.3.1.2 Feeding guilds As with total herbivores and individual species, the abundance of most herbivore feeding guilds varied by several fold among willow genotypes (Fig. 2.2A-F). For example, leaf chewer (225,105 = 41.62, P = 0.020; Fig. 2.2A) and phloem feeder (225,105 = 41.16, P = 0.022; Fig. 2.2C) abundance varied 9.8- and 14-fold among genotypes, respectively, but xylem feeders displayed only a marginally significant response (225,105 = 36.39, P = 0.066; Fig. 2.2B). Percent leaf area removed (PLAR) also varied 6.3-fold, ranging from an average of 4.7% to 29.6% leaf area removed among genotypes (F25,105 = 2.80, P < 0.001; Fig. 2.2D). Finally, galler and leaf miner abundance differed by 13.1-fold (225,105 = 44.06, P = 0.011; Fig. 2E) and 3.7-fold (225,105 = 82.06, P < 0.001; Fig. 2.2F) among genotypes, respectively.   Several of these guild-level responses exhibited significant phenotypic correlations. For example, leaf chewers were positively correlated with both phloem feeders (r = 0.26, P = 0.003) and leaf miners (r = 0.31, P < 0.001; Table 2.2). Phloem feeders were also positively correlated with leaf miners (r = 0.19, P = 0.031; Table 2.2). In contrast, gallers were negatively correlated with PLAR (r = -0.20, P = 0.021; Table 2.2). Despite the handful of significant phenotypic correlations, no guilds were genetically correlated (Table 2.2).    22 Figure 2.2 Herbivore feeding guild responses Herbivore feeding guild and percent leaf area removed (PLAR) responses to 26 different genotypes of Salix hookeriana: (A) leaf chewers, (B) xylem feeders, (C) phloem feeders, (D) percent leaf area removed, (E) gallers, and (F) leaf miners. Circles and error bars represent means and SEs.         23  Table 2.2 Correlations among herbivore feeding guilds  PLARa L. chewerb P. feederb  Gallerb L. minera PLARa 1 0.08 0.15 -0.20 -0.01 Leaf chewerb 0.24 1 0.26 0.11 0.31 Phloem feederb 0.37 0.30 1 -0.03 0.19 Gallerb -0.35 -0.06 0.03 1 0.15 Leaf minera 0.04 0.20 0.22 -0.07 1 Pearson correlations (r) of herbivore guild abundances and percent leaf area removed (PLAR). Italicized values below the diagonal represent genetic correlations (n = 26), while values above the diagonal are phenotypic correlations (n = 131). Notes: alog-transformed; blog(x+1) transformed. Statistically significant correlations (P < 0.05) are indicated in boldface type.   24  2.3.2 How heritable are different host-plant traits? 2.3.2.1 Leaf quality Leaf quality traits displayed a remarkable amount of variation among willow genotypes and were highly heritable (mean H2 = 0.72; Fig. 2.3A-C; Appendix A.7). For example, genotypes varied 3.3- and 88.5-fold in total condensed tannins and total salicylates, with broad-sense heritability values of 0.61 (21 = 67.75, P < 0.001) and 0.68 (21 = 92.40, P < 0.001; Fig. 2.3A), respectively. Similarly, willow genotypes varied 8.3- and 2.3-fold in concentration of total phenolic acids (H2 = 0.88, 21 = 202.92, P < 0.001; Fig. 2.3B) and total flavones (H2 = 0.70, 21 = 92.94, P < 0.001; Fig. 2.3C), both of which exhibited high degrees of heritability. Leaf trichome density was also both highly variable (25.9-fold among genotypes) and heritable (H2 = 0.62, 21 = 77.58, P < 0.001; Fig. 2.3F).  Although leaf C:N only varied 1.7-fold among genotypes it was highly heritable (H2 = 0.61, 21 = 64.03, P < 0.001; Fig. 2.3D). In contrast to the other leaf quality traits, both SLA (21 = 5.19, P = 0.023, Fig. 2.3E) and leaf water content (21 = 14.41, P < 0.001) varied 1.4-fold among genotypes and displayed relatively low heritability values of 0.15 and 0.27, respectively.   2.3.2.2 Architecture Variability and heritability of plant architecture was low relative to most leaf quality traits (mean H2 = 0.27; Fig. 3G-I; Appendix A.7). Plant size varied 2.4-fold among genotypes with a corresponding heritability of 0.15 (21 = 5.41, P = 0.020; Fig. 3G). In comparison, foliage density (H2 = 0.38, 21 = 25.65, P < 0.001; Fig. 2.3H) and plant height (H2 = 0.38, 21 = 23.77,   25 P < 0.001; Fig. 2.3I) varied 1.6 and 1.7-fold among genotypes, but were more than twice as heritable as plant size.  Figure 2.3 Heritable trait variation  Variation in plant traits among 26 genotypes of Salix hookeriana. Leaf quality traits (A-F): (A) total salicylates, (B) total phenolic acids, and (C) total flavones, (D) carbon:nitrogen, (E) specific leaf area (SLA), and (F) trichome density. Plant architecture traits (G-I): (G) plant size, (H) foliage density, and (I) plant height. Circles and error bars represent means and SEs, respectively.     26 2.3.3 Which plant traits account for herbivore community responses to host-plant genotype? 2.3.3.1 Community-level Leaf phenolic chemistry and plant size tended to be the best predictors of herbivore community responses (Table 2.3). For example, total herbivore abundance was best explained by variation in plant size, trichome density, and flavonoid PC2 (Table 2.3). Specifically, larger plants with fewer trichomes and negative loadings on flavonoid PC2 hosted more herbivore individuals. Herbivore community composition was influenced by plant size and a different set of leaf phenolics (phenolic acid PC1-2 and miscellaneous flavonoids PC1), but these traits did not fully explain the effect of willow genotype (Table 2.3).  Plant traits corresponding to individual herbivore species responses did not always match the traits that explained overall community composition (Table 2.3). For example, the leaf mining weevil, T. salicis, was more abundant on large shrubs with dense foliage, positive loadings on phenolic acid PC1, but low leaf water content (Table 2.3). The leaf mining moth, Caloptilia sp., had higher abundances on larger shrubs with negative loadings on flavonoid PC1 (Table 2.3). The leaf galling midge, I. salicisverruca, did not vary with phenolic chemistry, but had higher abundances on larger shrubs with higher leaf C:N (Table 2.3). In contrast, the leaf galling mite, A. tetanothrix, was more abundant on plants with negative loadings on phenolic acid PC2 (Table 2.3). Despite finding several significant herbivore-trait associations, genotype was maintained as a significant predictor of all herbivore species in our trait analyses, with the exception of A. tetanothrix.    27 Table 2.3 Herbivore responses to plant traits Response Variable coef (± SE) P ∆R2 total R2 Genotype P Herbivore abundancea Plant sizea 0.53 ± 0.09 <0.001 0.208 0.292 0.074  Flavonoids PC2 -0.11 ± 0.03 0.001 0.078     Trichome densityb -0.09 ± 0.04 0.018 0.037     Community compositionc Phenolic acids PC1     0.032 0.017 0.113 0.008  Phenolic acids PC2   0.001 0.034    Plant sizea    0.001 0.023    Miscellaneous flavonoids PC1  0.019 0.019     Flavonoids PC2     0.044 0.016     Tachyerges salicisb Phenolic acid PC1 0.11 ± 0.03 0.001 0.078 0.212 0.049  Foliage density residuals 2.64 ± 1.09 0.017 0.043    Plant sizea 0.36 ± 0.16 0.023 0.039     Water content -0.60 ± 0.27 0.031 0.035     Caloptilia sp.d Plant sizea 0.70 ± 0.19 <0.001 0.137 0.212 0.037  Flavonoids PC1 -0.12 ± 0.04 0.001 0.081   Iteomyia salicisverrucab Plant sizea 0.48 ± 0.19 0.013 0.051 0.116 <0.001   C:N 0.03 ± 0.01 0.02 0.044     Aculops tetanothrixb Phenolic acid PC2 -0.18 ± 0.04 <0.001 0.142 0.142 0.304 Leaf chewer abundanceb Plant sizea 0.79 ± 0.16 <0.001 0.177 0.177 0.119 Phloem feeder abundanceb Plant sizea 0.68 ± 0.18 <0.001 0.105 0.201 0.633  SLA residuals 0.16 ± 0.06 0.012 0.048    Phenolic acids PC1 -0.08 ± 0.04 0.031 0.035     Height residuals 0.58 ± 0.28 0.040 0.032     PLARa Full trait model     0.128   0.164   Galler abundanceb Full trait model     0.209   0.147              28 Response Variable coef (± SE) P ∆R2 total R2 Genotype P  Leaf miner abundanceb Plant sizea 0.41 ± 0.09 <0.001 0.120 0.334 0.052  Foliage density residuals 2.34 ± 0.7 0.001 0.070    Phenolic acids PC2 -0.10 ± 0.03 0.001 0.075    Phenolic acids PC1 0.05 ± 0.02 0.011 0.041     Height residuals 0.37 ± 0.16 0.026 0.031      Results from multiple regression and redundancy analyses, after forward model selection, of herbivore responses to plant traits of Salix hookeriana growing in a common garden. Additionally, we tested whether willow genotype continued to be a significant predictor of herbivore responses after accounting for the variation explained by plant traits (Genotype P). Notes: alog-transformed; blog(x+1) transformed;  credundancy analysis on chord-transformed herbivore community data (site-by-species matrix) with significance evaluated after 1000 permutations of the data; dsquare-root transformed.    29 2.3.3.2 Feeding guilds As with most herbivore community responses, feeding guilds were principally linked with leaf phenolic chemistry and plant architecture (Table 2.3). For example, leaf chewers were more abundant on larger plants (Table 2.3). Phloem feeders were also more abundant on larger plants, but also responded positively to taller plants and those that had high SLA and negative loadings on phenolic acid PC1 (Table 2.3). The full trait models for PLAR and galler abundance were not significant, suggesting that unmeasured plant traits may be underlying their response to different willow genotypes. In contrast, phenolic acid PC1 and PC2, as well as all three architecture traits explained 33.4% of the variance in leaf miner abundance. Specifically, plants with greater architectural complexity (larger, taller and more dense foliage), positive loadings on phenolic acid PC1, and negative loadings on phenolic acid PC2 hosted more leaf miners (Table 2.3).  2.4 Discussion Our results demonstrate that host-plant genetic variation is a key factor shaping S. hookeriana’s associated arthropod herbivore community in a willow population in northern California. Total herbivore abundance, community composition, as well as individual species and feeding guilds exhibited strong responses to different willow genotypes. These differences corresponded with extensive phenotypic variation in leaf quality and plant architecture; however, there was no single trait that explained herbivore community responses. Rather, there was a range of host-plant traits that were associated with different herbivore species and feeding guilds.     30 2.4.1 How do herbivore communities respond to host-plant genotype? Our study highlights that a genetic basis to arthropod herbivore community composition occurs via differential responses among guilds and species, a result that has been demonstrated in a variety of host-plant systems (Whitham et al. 2012). Although species- and guild-level abundances varied among S. hookeriana genotypes, none of these responses were correlated among genotypes. Similarly, Roche and Fritz (1997) with Salix sericea found little evidence for genetically correlated responses among the 12 species of galling, leaf mining, and leaf folding herbivores they examined. Work in other host-plant systems have observed strong genetic correlations among herbivore species (Maddox & Root 1990), as well as correlations that vary from year-to-year (Johnson & Agrawal 2007). The absence of genetic correlations in our study could be indicative of different herbivore species responding to different suites of plant traits. Alternatively, the magnitude of correlated responses measured in our study may have been dampened by naturally occurring competitive interactions among herbivores, or predation, or both (Leimu & Koricheva 2006). Either way, this lack of genetic correlation suggests that selection for resistance traits imposed by these herbivores on S. hookeriana, and possibly other Salix sp., are independent of one another. Given that herbivore communities are often highly heterogeneous in space and time (Lewinsohn, Novotny & Basset 2005), species turnover in S. hookeriana’s diverse herbivore assemblage could result in highly variable selection pressures on many different plant traits. This explanation may contribute to why Salix sp. often exhibit considerable genetic and phenotypic variation within natural populations (Fritz & Price 1988; Brunsfeld et al. 1991; Nichols-Orians et al. 1993).    31 2.4.2 How heritable are different host-plant traits? Salix hookeriana genotypes varied in all traits that we measured; however, the magnitude of variation among genotypes was much greater for leaf phenolics and trichome density compared to other leaf quality traits or plant architecture. Moreover, leaf quality traits had 2.7-fold higher broad-sense heritability values (mean H2 = 0.72) compared to plant architectural traits (mean H2 = 0.27) in S. hookeriana, a pattern primarily driven by leaf phenolic chemistry. While broad-sense heritability values tend to overestimate the capacity for evolution, these relative differences in plant trait heritability may be quite general. For example, a meta-analysis by Geber and Griffen (2003) found that the mean heritability of plant secondary chemistry was more than two times greater than the heritability of plant morphology, phenology, and vegetative performance traits. This pattern may have important implications for community genetics research in plant-herbivore systems, especially when there is considerable plant phenotypic variation. For example, traits under weaker genetic control (i.e., low heritability) will be strongly influenced by the environment in which a host-plant is growing. If herbivores cue in on weakly heritable traits (e.g., plant size), predicting community responses will be difficult without explicitly incorporating environmental variation. Of the plant-herbivore genotype-by-environment (G × E) studies that have been done (e.g., Garibaldi, Kitzberger & Chaneton 2011; Silfver et al. 2014), much of this work has examined a limited number of traits (similar to the purely genetic studies), environments, and spatial scales (Tack, Johnson & Roslin 2011). Consequently, integrating detailed trait screenings within G × E studies should be a priority for future research.    32 2.4.3 Which plant traits account for herbivore community responses to host-plant genotype? Recently, the primacy of plant secondary metabolites in mediating host-plant resistance to arthropod herbivores has been questioned (Carmona et al. 2011). In concordance, we found that plant size tended to explain nearly twice the variation (mean R2 = 0.105) in herbivore responses compared to any single axis of phenolic variation (mean R2 = 0.057). This result is consistent with a recent meta-analysis demonstrating a positive relationship between the architecture of woody plants and arthropod herbivores (Carmona et al. 2011). This result also corresponds with predictions from the plant vigor hypothesis (Price 1991), which states that herbivores prefer either larger modules (e.g., shoots, leaves) within plants or larger plant individuals instead of smaller ones, due to increased resource availability. There are many other potential explanations though. For example, plant size was positively correlated with plant height, architectural complexity (fractal dimension), and foliage density. Therefore, larger plants are likely more apparent to herbivores (Castagneyrol et al. 2012), provide habitat heterogeneity that decreases predator foraging efficiency (Kareiva & Sahakian 1990), and buffer microclimate conditions (Raghu, Drew & Clarke 2004). Partitioning these causal mechanisms will require future manipulative experiments that hold plant size constant while varying other architectural traits. Nevertheless, we did find independent and positive relationships between plant height or foliage density and the abundance of certain herbivore species (Tachyerges salicis) and feeding guilds (leaf miners and phloem feeders), suggesting that plant apparency and habitat heterogeneity may contribute to genetic variation in host-plant susceptibility.    33 Another possible explanation for the relatively weak role of leaf phenolic chemistry compared to plant architecture is that the dominant herbivores in our study all specialize on members of the genus Salix. A priori, we would expect that specialist herbivores would be less sensitive to variation in the most abundant secondary metabolites, since they would have had to evolve some degree of physiological tolerance to these chemicals. In line with this, we found little correspondence between herbivore responses and salicylates and condensed tannins, which were the most abundant secondary metabolites in our leaf samples. Instead, we found that when phenolic compounds did show a relationship with herbivores, they were the less abundant ones (e.g., flavonoids and phenolic acids). Consequently, our results suggest that screening a range of secondary metabolites, above and beyond the most abundant compounds, is necessary for understanding herbivore community assembly.  Despite our detailed characterization of willow phenotypes, the traits we measured did not fully explain the effect of willow genotype on the herbivore community (Table 2.3), suggesting that other unmeasured traits could be relevant. For example, we did not measure shoot length, which has been identified as one of the best predictors of abundance for a few species of willow galling sawflies, presumably because more vigorous growing shoots have higher resource availability (Price 1991). We also measured traits at the peak of the growing season, thereby neglecting potential differences in phenology among willow genotypes—an important suite of traits in other systems (Johnson & Agrawal 2005). In addition to unmeasured plant traits, our study focused on community composition, thereby neglecting the diverse competitive and predatory interactions that are likely going on throughout this community. For example, Fritz (1990) showed that the strength of competition between gall-inducing sawflies varied among genotypes of Salix   34 lasiolepis. The size and toughness of stem galls induced by the galling sawfly, Euura lasiolepis, on S. lasiolepis is determined by plant genotype, which in turn, affects parasitoid attack rates (Craig, Itami & Price 1990). Thus, a comprehensive understanding of host-plant genetic effects on herbivore communities may also require incorporating interactions among species both within and between trophic levels.  2.4.4 Conclusions Our research provides several insights into the trait-based mechanisms mediating herbivore community responses to host-plant genetic variation. First, there is emerging evidence from our study and others that plant secondary chemistry tends to be more heritable than plant architectural traits (Geber & Griffin 2003). However, in woody plant systems, plant architecture appears to be a dominant and predictable driver of herbivore community responses, relative to the more idiosyncratic effects of plant secondary chemistry (Carmona et al. 2011). Since environmental variation is more likely to shape variation in plant architecture (because it is less heritable), it will be particularly important for future work in woody plant systems to explicitly incorporate plant responses across variable environments to understand herbivore community assembly. Next, herbivore responses were not correlated among genotypes, likely because individual herbivore species and feeding guilds are cueing in on different suites of plant traits. These uncorrelated responses also imply separate genetic control of resistance to these species and the lack of potential for multispecies selection on the same resistance traits (Fritz and Simms, 1992). Thus, studies should consider a range of traits and partition herbivore species and guild responses to host-plant genotypes. Finally, the direct effects of plant traits on herbivores are likely not the only pathways by which genetic variation structures herbivore communities.   35 Incorporating similar detailed comparisons of competitive or predatory interactions will be an important step for building a mechanistic understanding of the genetic basis to community assembly and the eco-evolutionary dynamics between plants and herbivores.   36 Chapter 3: Genetic specificity of a plant-insect food web: implications for linking genetic variation to network complexity 3.1 Introduction Network theory has provided both a conceptual and quantitative approach for mapping interactions between species and making predictions about how the gain or loss of species will affect the structure and dynamics of ecological networks (Dunne et al. 2002; Stouffer & Bascompte 2011; Rohr, Saavedra & Bascompte 2014). Representing a network at the species-level, however, makes the implicit assumption that each species consists of a homogenous population of individuals, all of which interact equally with individuals of different species. Yet, most populations are heterogeneous mixtures of individuals that vary in their phenotypes and there is growing evidence that this intraspecific variation is an important factor governing the assembly of ecological communities (Clark 2010; Bolnick et al. 2011; Violle et al. 2012). Consequently, there is a clear need to account for the role of intraspecific variation in structuring ecological networks (Poisot, Stouffer & Gravel 2015).  Genetic variation is a key driver of intraspecific variation and many studies have now demonstrated direct and indirect genetic effects on species interactions (Fritz 1995; Bailey et al. 2006; Abdala‐ Roberts & Mooney 2013) and the composition of communities across multiple trophic levels (Fritz & Price 1988; Maddox & Root 1990; Post et al. 2008; Harmon et al. 2009). This prior work forms a clear expectation that intraspecific genetic variation is capable of scaling up to affect the structure of an ecological network. In particular, we expect that network structure will be affected by genetic variation through at least two different mechanisms. For a food web   37 (network of trophic interactions), genetic variation in the quality of a basal resource may alter the (i) abundances or (ii) phenotypes of consumer species or both (Bukovinszky et al. 2008). These direct genetic effects on consumers may then have cascading effects on the strength of trophic interactions between consumers and their predators (Bukovinszky et al. 2008), resulting in distinct compositions of trophic interactions associated with different genotypes of the basal resource (Fig. 3.1). If such genetic specificity in the composition of trophic interactions occurs, then theory predicts that increasing genetic variation will result in more interactions per species  (Bolnick et al. 2011; Moya-Laraño 2011), and therefore greater food-web complexity (Fig. 3.2). Moreover, greater complexity may in turn affect food web dynamics, as more complex food webs are predicted to be more robust to species extinctions (MacArthur 1955; Dunne et al. 2002). However, whether genetic variation is capable of scaling up to affect food-web complexity is currently unclear.  In this study, we quantify the genetic specificity of trophic interactions and use these data to simulate the additive effects of genetic variation on food-web complexity. To do this, we used a common garden experiment of a host plant (26 genotypes of coastal willow, Salix hookeriana) and its associated food web of insect galls and parasitoids (Fig. 3.1). We focused on this plant-insect food web for three reasons. First, we have demonstrated in previous work that S. hookeriana (hereafter, willow) displays heritable variation in traits associated with leaf quality (36 traits, mean H2 = 0.72) and plant architecture (4 traits, mean H2 = 0.27), some of which are also associated with resistance to its community of galling herbivores (Barbour et al. 2015). Second, the unique biology of galling insects makes them ideal for building quantitative food webs. In particular, galls provide a refuge for larva from attack by most generalist predators   38 (Hawkins, Cornell, & Hochberg 1997); therefore, galls and their natural enemies often form a distinct subset of the larger food web associated with host-plants. In our system, all of the natural enemies are insect parasitoids that complete their development within the gall after parasitizing larva, making it easy to identify and quantify all of the trophic interactions within this food web. Third, the biology of galls is also ideal for identifying the mechanisms mediating trophic interactions. In particular, gall size is a key trait that affects the ability of parasitoids to successfully oviposit through the gall wall and into the larva within the gall (i.e. larger galls provide a refuge from parasitism: Abrahamson & Weis 1997). Moreover, gall size is determined, in part, by the genotype of the plant (Abrahamson & Weis 1997), so we have a clear mechanism by which genetic variation can affect the strength of trophic interactions. Taken together, our study seeks to examine how intraspecific genetic variation influences the structure of ecological networks. In doing so, our study takes a crucial step toward a more predictive understanding of how the gain or loss of genetic variation will affect the dynamics of ecological networks.     39 Figure 3.1 Genetic specificity of trophic interactions  The species comprising the food web in this study include a host plant (coastal willow, Salix hookeriana), four herbivorous galling insects, and six insect parasitoids (species details in Materials & Methods). The plant-insect food web consists of 16 trophic interactions (4 willow-gall and 12 gall-parasitoid) aggregated from all plant individuals sampled in this common garden experiment, whereas each genotype subweb represents the trophic interactions aggregated from all plant individuals of the corresponding genotype. We depicted three genotype subwebs (of 26) to illustrate the differences in trophic interactions associated with each willow genotype. The width of each grey segment is proportional to the number of individuals associated with each trophic interaction. Note that we scaled the width of trophic interactions to be comparable among genotype   40 subwebs, but not between subwebs and the aggregated food web, in order to emphasize the differences among subwebs.   Figure 3.2 Conceptual model of how genetic variation affects food-web complexity  Conceptual model of how increasing genetic variation (number of shades of green circles) results in greater food-web complexity (number of interactions per species). If different genotypes of a basal resource are associated with distinct compositions of trophic interactions (i.e. genetic specificity of trophic interactions), then increasing genetic variation in the resource will result in a more complex food web because of the increase in the number of interactions per species at all three trophic levels. Colors correspond to different trophic levels (green = basal resource, blue = primary consumer, orange = secondary consumer), while different shapes within each trophic level correspond to different species.     41 3.2 Materials and Methods  3.2.1 Common garden experiment and plant traits  To isolate the effects of coastal willow (S. hookeriana Barratt ex Hooker) genetic variation on the plant-insect food web, we used a common garden experiment consisting of 26 different willow genotypes (13 males; 13 females), located at Humboldt Bay National Wildlife Refuge (HBNWR) (40°40'53"N, 124°12'4"W) near Loleta, California, USA. Willow genotypes were collected from a single population of willows growing around Humboldt Bay. While relatedness among these genotypes is unknown, their phenotypes in multivariate trait space are quite distinct from each other (details in supplementary information), suggesting that we can treat them as independent from one another. This common garden was planted in February 2009 with 25 clonal replicates (i.e. stem cuttings) of each willow genotype in a completely randomized design in two hectares of a former cattle pasture at HBNWR. Willows in our garden begin flowering in February and reach their peak growth in early August. During this study, willows had reached 2 - 4 m in height. Further details on the genotyping and planting of the common garden are available in Barbour et al. (2015).  To identify the plant traits that may be determining resistance to galling insects, we measured 40 different traits associated with leaf quality (36 traits) and plant architecture (4 traits). Each of these 40 traits exhibited significant, broad-sense heritable variation (mean leaf quality H2 = 0.72; mean architecture H2 = 0.27; range of H2 for all traits = 0.15 - 0.97). For further details on how these willow traits were sampled and quantified, see methods in Barbour et al. (2015). We then reduced these 40 traits into 13 composite traits that had a negligible degree of multicollinearity using either principle components analysis (PCA), sequential regression (residuals of one trait   42 after accounting for correlation between two traits), or removing one trait from a pair of highly correlated traits (details on methods in Barbour et al. 2015). The final set of leaf quality traits included salicylates/tannins PC1, flavones/flavonols PC1-2, phenolic acids PC1-2, flavanones/flavanonols PC1 (Table S3 of Barbour et al. 2015), carbon-to-nitrogen ratio (C:N), water content, specific leaf area (residuals from water content), and trichome density. The final set of plant architecture traits included plant size, plant height (residuals from plant size), and foliage density (residuals from plant size).  3.2.2 Quantifying the genetic specificity of the plant-insect food web  To build a quantitative food web for each willow genotype, we collected galls from about 5 randomly chosen replicates of each genotype in September 2012 (N = 145 willows, range = 4 - 9 replicates per genotype). For each replicate willow, we collected all galls occurring on one randomly selected basal branch. We restricted our gall collections to those induced by midges in the insect family Cecidomyiidae (4 species). These species included a leaf gall (Iteomyia salicisverruca), bud gall (Rabdophaga salicisbrassicoides), apical-stem gall (unknown midge species), and mid-stem gall (Rabdophaga salicisbattatus). To quantify the abundance of gall-parasitoid interactions, we placed collected galls into 30 mL plastic transport vials (loosely capped at the end), which we maintained at room temperature in the lab for four months. We then opened galls under a dissecting scope and determined whether the gall survived or was parasitized, and if parasitized, the identity of the parasitoid species. In total, we identified five species of hymenopteran parasitoids, including Platygaster sp. (Family: Platygastridae), Mesopolobus sp. (Family: Pteromalidae), Torymus sp. (Family: Torymidae), Tetrastichus sp. (Family: Eulophidae), and an unknown species of Mymaridae (hereafter, Mymarid sp. A), as   43 well as one predatory midge (Lestodiplosis sp., Family: Cecidomyiidae). This predatory midge is functionally similar to the other parasitoids so we collectively referred to this natural enemy community (6 species) as parasitoids for brevity. All together, we documented 12 unique gall-parasitoid interactions (Fig. 3.1), which appears to represent the vast majority of interactions in the gall-parasitoid network (details in Appendix B.5). We omitted from analyses those galls for which we could not reliably determine the cause of mortality. We quantified gall abundance by counting the number of surviving and parasitized larva for each gall species collected from each branch. For gall size, we measured galls to the nearest 0.01 mm at their maximum diameter (perpendicular to the direction of plant tissue growth).  To quantify the genetic specificity of trophic interactions with galling insects, we tested for differences in gall sizes, abundances, and community composition among willow genotypes. For gall size, we analyzed separate linear models with willow genotype as the predictor variable and average gall size as the response variable, but we weighted the analysis by the number of galls used to calculate average gall size. We weighted the analysis because we expected that averages based on more galls reflect a more accurate estimate of the average size of galls found on a willow individual. For gall abundances, we analyzed multivariate generalized linear models (multivariate GLMs, error distribution = negative binomial, link function = log) with willow genotype as the predictor variable and a matrix of gall abundances as the response variable. For gall community composition, we used permutational MANOVA (PERMANOVA) with willow genotype as the predictor variable and a matrix of Bray-Curtis dissimilarities in gall abundances as the response variable. To identify the plant traits mediating resistance to galling insects, we used the same analyses as for gall sizes (weighted linear models) and abundances (multivariate   44 GLMs) except that our predictor variable was now a matrix of willow traits. To select a final model of willow traits, we sequentially removed traits based on Aikaike information criteria (AIC) to identify a nested set of candidate statistical models. We then used likelihood ratio tests to identify the statistical model of willow traits that best predicted gall abundances or gall sizes.   To quantify the genetic specificity of the network of gall-parasitoid interactions, we tested for differences in the abundance, composition, and strength of gall-parasitoid interactions among willow genotypes. For the abundance and composition of gall-parasitoid interactions, we used the same analytical approach as we did to test for differences in gall abundances and community composition. For these analyses though, we had a matrix of the abundance (multivariate GLMs) or dissimilarity (PERMANOVA) of unique gall-parasitoid interactions as the response variable. To identify the mechanisms determining the abundance of gall-parasitoid interactions, we again used multivariate GLMs except that our predictor variable was now a matrix of gall abundances and gall sizes. We then used the same approach as we did to identify the willow traits that best predicted gall abundances (i.e. AIC and likelihood ratio tests), to identify which gall sizes and abundances best predicted the abundance of gall-parasitoid interactions. For the strength of gall-parasitoid interactions, we used separate GLMs (error distribution = binomial, link function = logit) with willow genotype as the predictor variable and the proportion of galls parasitized as our response variable for each gall species. If we detected an effect of willow genotype on total parasitism rates, then we analyzed separate GLMs for each parasitoid species to determine which parasitoids were driving total parasitism rates. Finally, we again used AIC and likelihood ratio tests to examine whether parasitism rates were due to gall abundance, gall size, or their interaction.   45  3.2.3 Simulating the additive effects of genetic variation on network complexity For our index of complexity, we chose to use quantitative-weighted linkage density, LDq, which is based on Shannon diversity and is the average of the effective number of prey and predatory interactions for a given species, weighted by their energetic importance (details on how LDq  was calculated are available in Appendix B.6 and in Bersier, Banašek-Richter & Cattin 2002; Banaaek-Richter et al. 2009). LDq  (hereafter, food-web complexity) is less sensitive to variation in sample size compared to other measures of food-web complexity  (Banaaek-Richter et al. 2009), making it an appropriate measure of complexity for our study.   To examine whether genetic variation increases food-web complexity, we designed a resampling procedure to estimate the complexity of the plant-insect food web at different levels of genetic variation (range = 1 to 25 genotype mixtures) from our empirical data. We omitted 1 of the 26 genotypes from this analysis (Genotype U) because we did not find any galls on the branches we sampled. Our resampling procedure consisted of the following two steps. (i) Generate quantitative matrices: In order to ensure willow genotypes had equal sampling effort, we randomly sampled 4 individual willows of each genotype (without replacement) and their corresponding trophic interactions (willow-gall and gall-parasitoid). Next, we calculated the total abundance of each trophic interaction associated with each genotype, resulting in a quantitative matrix of 25 genotypes (rows) and 16 unique trophic interactions (columns, 4 willow-gall and 12 gall-parasitoid). (ii) Sampling genetic variation: with this matrix, we randomly sampled 1 to 25 genotypes (without replacement), 200 times each, and calculated the total abundance of each trophic interaction associated with each level of genetic variation. We removed redundant   46 combinations of genotypes that were generated by our random sampling. We then calculated food-web complexity for each sample, and then calculated the average complexity for each level of genetic variation. Finally, we repeated this sampling procedure on 50 different matrices to quantify the variability in our estimates of average food-web complexity. This resampling procedure is analogous to methods used in experimental studies (e.g. Crutsinger et al. 2006; Johnson et al. 2006) to estimate the expected additive effects of genetic variation on arthropod diversity.  One constraint of our experimental design and resampling procedure is that estimates of complexity from mixtures with more genotypes are based off more plants (e.g. 1-genotype 4-plant mixtures vs. 25-genotype 100-plant mixtures). This would not be a problem if, for example, we had measures of trophic interactions on 25 replicate plants of each willow genotype, because we could directly compare 1-genotype 25-plant mixtures with 25-genotype 25-plant mixtures. Therefore, it is important to account for the increase in food-web complexity that may come from simply sampling more plants. We estimated this sampling effect by first using our resampling procedure to generate 1,000 estimates of average complexity for 1-genotype mixtures based on progressively higher levels of sampling effort (1 – 4 plants). We then used an asymptotic model (Colwell & Coddington 1994) to predict the average complexity of food webs in 1-genotype 100-plant mixtures to use as a baseline for estimating the additive effects of genetic variation (dashed line in Fig. 3.6). Details of the asymptotic model and our evaluation of alternative models are given in Appendices B.7-9.     47 To examine the pathways by which genetic variation influences food-web complexity, we built a piecewise structural equation model (details given in Appendix B.10) using data from one of the 50 replicates of our resampling procedure. We observed the same qualitative results when we explored other replicates, so we only report the quantitative results from the first replicate.  All statistical analyses were conducted in R version 3.1.2 (R Core Team 2014).   3.3 Results and Discussion 3.3.1 Quantifying the genetic specificity of the plant-insect food web In concordance with previous work in this system (Barbour et al. 2015), we observed clear differences in the abundance of 3 of the 4 galling insects among willow genotypes (multivariate GLM, χ225,119 = 202.40, P = 0.001; Appendix B.1). Specifically, we found that the average abundance of leaf, bud, and apical-stem galls varied 10-, 8-, and 1.4-fold among willow genotypes, respectively (Fig. 3.3A-C). This variation resulted in 69% dissimilarity in the average composition of galls among willow genotypes (F22,89 = 1.96, P = 0.001). Moreover, we found that the average diameter of leaf galls varied 2-fold among willow genotypes (Fig. 3.3D). This observed genetic specificity in the abundance and phenotypes of insect herbivores corroborates decades of work in other plant-gall (Fritz & Price 1988; Abrahamson & Weis 1997; Bailey et al. 2006) and plant-herbivore systems (Maddox & Root 1990; Whitham et al. 2012).    48 Figure 3.3 Response of gall community  Direct effects of willow (Salix hookeriana) genetic variation on its associated community of galling insects. Among the 26 willow genotypes we surveyed in our common garden experiment, we found that: (A) average abundance of leaf galls varied 10-fold (GLM, χ225,119 = 74.60, P = 0.001); (B) average abundance of bud galls varied 8-fold (GLM, χ225,119 = 55.02, P = 0.006); (C) average abundance of apical-stem galls varied 1.4-fold (GLM, χ225,119 = 44.47, P = 0.042); and (D) average diameter of leaf galls varied 2-fold (weighted linear model, F23,57 = 2.17, P = 0.009). Plots (A – C) display the median (bar within box), 25th to 75th percentiles (IQR, box edges), 1.5 × IQR (whiskers), and outliers (points) for gall abundances found on each willow genotype. For   49 plot (D), each circle corresponds to the average gall diameter associated with an individual willow and the size of the circle is scaled according to the number of galls used to calculate the weighted average for each willow genotype (diamond). Colors correspond to different gall species (orange = leaf gall, blue = bud gall, grey = apical-stem gall). For all plots, we ordered willow genotypes based on average leaf gall abundance (low to high).   Importantly though, our extensive screening of willow phenotypes (Materials and Methods) enabled us to identify traits that may be mediating the genetic specificity of trophic interactions with galling insects. In particular, we found that leaf C:N, certain leaf secondary metabolites (flavanones/flavanonols PC1), and plant size were associated with changes in the abundance of galling insects (multivariate GLM, χ23,104 = 28.44, P = 0.004; Appendix B.2), whereas leaf gall diameter was associated with variation in a different suite of leaf secondary metabolites (salicylates/tannins PC1 and flavones/flavonols PC1)(weighted linear model, F2,59 = 8.27, P < 0.001; Appendix B.2). These results highlight that accounting for intraspecific variation in multiple plant traits is important for predicting antagonistic interactions between plants and insect herbivores (Barbour et al. 2015), and should therefore be incorporated into mechanistic models of food-web structure.   We found that the effects of willow genetic variation extended beyond pairwise interactions with herbivores (Fritz & Price 1988; Maddox & Root 1990; Whitham et al. 2012) and simple tri-trophic interactions (Fritz 1995; Abrahamson & Weis 1997; Bailey et al. 2006; Abdala‐ Roberts & Mooney 2013) to determine the assembly of the network of gall-parasitoid interactions (multivariate GLM, χ225,119 = 357.10, P = 0.001; Appendix B.1). In particular, we found that the frequency of parasitism from three parasitoids (Platygaster sp., Mesopolobus sp., and Torymus   50 sp.) on leaf galls varied 270%, 30%, and 40% among willow genotypes, respectively (Fig. 3.4A-C). This variation resulted in 78% dissimilarity in the average composition of gall-parasitoid interactions among willow genotypes (F12,45 = 1.57, P = 0.007). Furthermore, we found that the probability of a gall being parasitized also depended on willow genotype (Appendix B.1), a pattern that was particularly strong for leaf galls (Fig. 3.4D).   The genetic specificity of the network of gall-parasitoid interactions was determined by variation in both the abundance and size of galling insects. Specifically, we found that the abundance of 67% (8 of 12) of the gall-parasitoid interactions increased with the abundance of their associated galls, and that leaf gall size affected trophic interactions with both leaf and bud galls (multivariate GLM, χ24,76 = 179.80, P = 0.001; Appendix B.2). In terms of interaction strength, we found that the odds of a leaf gall being parasitized decreased by 25% with every 1 mm increase in leaf gall diameter (GLM, χ21,79 = 22.28, P < 0.001). Nevertheless, the strength of trophic interactions with individual parasitoid species depended on both leaf gall size and abundance (Fig. 3.5A-B; Appendix B.3), suggesting that natural selection has the potential to shape food-web structure. For example, if there were selection on willows for increased resistance to leaf galls through smaller galls and lower gall abundances, then we would expect to see more parasitism overall and a shift in dominance from Platygaster to Mesopolobus, since Mesopolobus had its highest attack rates on small galls at low abundances (Fig. 3.5A). While our results are limited to examining the effects of standing genetic variation on a tri-trophic food web over a single season, there is ample evidence from other studies that natural selection can play an important role in shaping consumer-resource dynamics (Yoshida et al. 2003; Agrawal et al. 2012). Understanding how evolutionary processes affect the structure and dynamics of   51 ecological networks, and vice versa (Melián et al. 2011; Moya-Laraño et al. 2012), is likely a fruitful topic for future research.  Figure 3.4 Response of gall-parasitoid interaction network  Indirect effects of willow (Salix hookeriana) genetic variation on its associated network of gall-parasitoid interactions. Among the 26 willow genotypes we surveyed in our common garden experiment, we found that: (A) leaf gall parasitism by Platygaster sp. varied 270% (GLM, χ225,119 = 79.51, P = 0.001); (B) leaf gall parasitism by Mesopolobus sp. varied 30% (GLM, χ225,119 = 50.00, P = 0.009); (C) leaf gall parasitism by Torymus sp. varied 40% (GLM, χ225,119 = 60.11, P = 0.001); and (D) the proportion of leaf galls parasitized   52 varied between 0.0 and 1.0 (GLM, χ223,58 = 75.79, P < 0.001). Plots (A – C) display the median (bar within box), 25th to 75th percentiles (IQR, box edges), 1.5 × IQR (whiskers), and outliers (points) for the abundance of gall-parasitoid interactions associated with each willow genotype. For plot (D), each circle corresponds to the proportion of galls parasitized on each replicate willow and the size of the circle is scaled according to the number of galls used to calculate the weighted average for each willow genotype (diamond). Colors correspond to different gall-parasitoid interactions. As with Fig. 3.3, we ordered willow genotypes based on average leaf gall abundance (low to high).  Figure 3.5 Mechanisms affecting leaf gall parasitism    53 Variation in the size and abundance of leaf galls on willows is associated with changes in the strength and composition of gall-parasitoid interactions. (A – B) In general, the proportion of leaf galls parasitized by both Platygaster (blue, solid line) and Mesopolobus (green, short-dashed line) decreases as gall size increases, while Torymus (orange, long-dashed line) exhibits the opposite pattern. On willows with small leaf galls though (< 8 mm), Mesopolobus had the highest attack rate at low gall abundances (1 – 4 leaf galls per branch, N = 46 per parasitoid species), whereas Platygaster was the dominant parasitoid at high gall abundances (5 – 22 leaf galls per branch, N = 35 per parasitoid species). Lines correspond to slopes estimated from generalized linear models (GLMs). Points were jittered slightly to avoid overlapping values.   3.3.2 Simulating the additive effects of genetic variation on network complexity To examine this, we used our empirical data to simulate how the complexity of the plant-insect food web would change across different levels of willow genetic variation (Materials and Methods). After accounting for sampling effort (dashed line, Fig. 3.6), our simulations suggest that food-web complexity would increase by 20% with increasing genetic variation (Fig. 3.6). This positive relationship was primarily due to an increased likelihood of sampling genotypes with complementary trophic interactions, as we found that willow genotypes differed by 73% in the average composition of their trophic interactions (inset Fig. 3.6). To more precisely understand the relationship between genetic variation, the addition of complementary interactions, and food-web complexity, we used a structural equation model (Materials and Methods). We found that increasing genetic variation resulted in a more diverse community of galls and a more generalized network of gall-parasitoid interactions, albeit through two main pathways (Appendix B.10). On the one hand, increasing genetic variation resulted in higher gall species richness, which had a positive direct effect on food-web complexity (standardized path   54 effect = 0.21). On the other hand, increasing genetic variation resulted in higher gall abundances, which indirectly increased complexity by increasing the effective number of parasitoid species per gall (standardized path effect = 0.26). Other pathways had comparatively small and idiosyncratic effects on food-web complexity (Appendix B.10).  Figure 3.6 Simulated effect of genetic variation on food-web complexity  Simulations of our empirical data indicate that increasing willow (Salix hookeriana) genetic variation results in a more complex plant-insect food web due to complementarity in trophic interactions. Specifically, we found that the average complexity (LDq, quantitative-weighted linkage density) of the plant-insect food web increased by 20% over the range of genetic variation (number of genotypes) in the experimental population of willows. Grey circles correspond to the average food-web complexity estimates for each replicate simulation (N = 50 for each level of genetic variation), whereas blue circles correspond to the overall average   55 complexity of food webs at each level of genetic variation. Black circles correspond to the average complexity of 1-genotype mixtures at 4 different levels of sampling effort (i.e. number of plants sampled), and the dashed line represents the predicted increase in complexity of 1-genotype mixtures with greater sampling effort. The inset shows how the average composition of trophic interactions (willow-gall and gall-parasitoid) differed by 73% among willow genotypes (PERMANOVA on Bray-Curtis dissimilarities, F22,89 = 1.90, P = 0.001), suggesting an important role of complementarity in determining food-web complexity. In this ordination plot, black letters and grey ovals correspond to the centroid and standard error of the centroid, respectively, for the composition of trophic interactions found on each willow genotype. Centroids and their standard errors were calculated from a constrained analysis of principal coordinates (CAP) on Bray-Curtis dissimilarities.  An important limitation of our simulation and experimental design is that we were unable to estimate the extent to which food-web complexity is influenced by non-additive effects of genetic variation. Non-additive effects may arise in a variety of ways (e.g. competition and facilitation, associational resistance/susceptibility, source-sink dynamics), and prior work has shown that host-plant genetic variation can have positive (Crutsinger et al. 2006), neutral  (Johnson, Lajeunesse & Agrawal 2006), or negative (McArt, Cook-Patton & Thaler 2012) non-additive effects on the diversity of upper trophic levels. Future experiments are needed that explicitly manipulate levels of genetic variation and test for the presence and magnitude of non-additive effects on food-web structure. It is worth noting though that our qualitative conclusion, namely that genetic variation likely increases food-web complexity, will still hold unless negative, non-additive effects are equal or greater in magnitude compared to the additive effect we observed.      56 3.3.3 Conclusions Our results suggest that the gain or loss of genetic variation within a key species may fundamentally alter food-web complexity and therefore the persistence of food webs. There are two main conclusions from our work. First, intraspecific variation in multiple traits is an important driver of network structure; therefore, mechanistic models of food-web structure should incorporate such variability within species (Poisot et al. 2015), as this can enhance the accuracy of these models in predicting trophic interactions (Woodward et al. 2010). Given that plants, insect herbivores, and their parasitoids comprise over half of all known species of metazoans (Price 1980; Strong, Lawton & Southwood 1984), accounting for intraspecific variation in a wide range of functional traits should be a priority for future food web models  (Henri & van Veen 2011). Second, understanding the direct and indirect effects of genetic variation on trophic interactions is essential for predicting how evolutionary processes will affect the structure and persistence of food webs over time. Indeed, our simulations suggest that the loss of genetic variation will result in less complex food webs. Moreover, genetic variation provides the raw material for evolution by natural selection; therefore, losing genetic variation in key species may hinder the adaptive capacity of both the species and the food web under future environmental change (Jump, Marchant & Peñuelas 2009; Carroll et al. 2014). At this point though, we are currently lacking a theoretical and empirical understanding of how genetic variation scales up to affect the dynamics of food webs. Given that the current rate of population extinction is orders of magnitude higher than the rate of species extinction (Hughes, Daily & Ehrlich 1997), our study highlights the pressing need for research examining how the loss of genetic variation within and among populations will affect food webs and the ecosystem services they provide  (Luck, Daily & Ehrlich 2003; Schindler et al. 2010).    57 Chapter 4: Host-plant genetic and environmental variation structure above and belowground communities in a coastal dune ecosystem 4.1 Introduction Intraspecific genetic variation is a key driver of trait variation within host plants, which in turn can have cascading effects on associated species and entire communities of organisms (Fritz & Price 1988; Maddox & Root 1990; Antonovics 1992; Lamit et al. 2015). For example, genetic variation in the leaf chemistry of cottonwoods (Whitham et al. 2006) and in the plant architecture of coyote bush (Crutsinger et al. 2014) structures diverse assemblages of species, from foliar arthropods aboveground to soil microbes below. While community-level consequences of genetic variation (commonly referred to as ‘community genetics’, sensu Antonovics 1992) have been documented in a variety of host-plant taxa (Whitham et al. 2012), evidence comes primarily from common garden experiments that minimize environmental variation. These controlled environments likely limit effects of the biotic and abiotic environment on the expression of host plant traits (Gratani 2014) as well as the diversity and composition of species assemblages (MacArthur 1972; Gaston 2000). Therefore, the importance of host-plant genetic variation in structuring ecological communities in naturally varying environments remains unclear (Hersch-Green, Turley & Johnson 2011; Tack, Johnson & Roslin 2012; Crutsinger 2016).   Key challenges for advancing studies of community genetics beyond the common garden are to: (i) identify important environmental factors that structure communities associated with host plants; and (ii) distinguish whether environmental effects are independent (E) or modified by host-plant genotype (G x E). For example, a series of experiments in common milkweed   58 (Asclepias syriaca) have shown that a diversity of biotic factors, such as light competition from neighboring plants (Agrawal & Zandt 2003), caterpillar herbivory (Abdala‐ Roberts, Agrawal & Mooney 2012), and aphid-tending ants (Mooney & Agrawal 2008; Abdala‐ Roberts et al. 2012) can act independently or interact with milkweed genotype to shape its associated community of foliar arthropods. These experiments demonstrated that the community genetics of milkweed are contingent on the biotic environment. Similarly, abiotic factors, such as soil nutrient availability, can act independently or modify the effects of host-plant genotype on herbivore assemblages (Orians & Fritz 1998) and tri-trophic interactions (Rossi & Stiling 1998; Abdala-Roberts & Mooney 2013). Still, we are lacking explicit comparisons of host-plant genotypic effects across natural gradients in both biotic and abiotic factors, so their relative importance in structuring communities is unclear.   Although host plants provide essential resources for a diverse array of taxa both above- and belowground, the majority of community genetics studies have focused on aboveground assemblages (Whitham et al. 2012). Studies that have simultaneously examined above- and belowground communities have found variable results, with host-plant genetic effects on aboveground communities being stronger (Crutsinger et al. 2008; Bailey et al. 2009) or comparable and coupled (Crutsinger et al. 2014) with those belowground. Above- and belowground linkages can have important consequences for both plant fitness (Whitham et al. 2006) and terrestrial ecosystem processes (Wardle et al. 2004). In addition, feedbacks between above- and belowground assemblages may depend strongly on the biotic and/or abiotic environment (Wardle et al. 2004). Consequently, a rising challenge for community genetics is to   59 understand the linkages between above- and belowground communities (Crutsinger et al. 2014; Lamit et al. 2015) and whether these linkages are modified by environmental variation.  Host-plant traits determine the quantity and quality of resources for the diverse organisms that colonize them; therefore, measuring functional trait responses of host-plant genotypes to different environments can give insight to mechanisms of community assembly in genotype-by-environment studies. Phenotypic traits can vary in their plasticity (change in trait expression of a genotype in response to the environment: Scheiner 1993) and may even be plastic in response to one environmental gradient but not another (Scheiner & Goodnight 1984, Garbutt & Bazzaz 1987, Scheiner 1993). In addition, multiple plant traits can be important in structuring associated communities on host plants (Agrawal 2004, 2005; Agrawal & Fishbein 2006; Barbour et al. 2015, 2016). Simultaneous measurements of multiple functional traits and community-level patterns in genotype-by-environment studies can distinguish the effects of heritable trait variation (proportion of variance in a trait explained by genotype: Lynch & Walsh 1998), phenotypic plasticity, and direct environmental effects on species assemblages.   Here, we use common garden experiments to examine how host-plant genotypic variation as well as the biotic and abiotic environment structure communities associated with the willow Salix hookeriana in a coastal dune ecosystem. Prior work in this system has shown that willow genotypes host distinct arthropod communities and that multiple plant traits are important in determining community assembly (Barbour et al. 2015; 2016). Importantly, these traits varied substantially in their degree of heritability (plant growth, mean H2 = 0.26; leaf quality, mean H2 = 0.72), suggesting that the environment may influence them in different ways. We sought to   60 answer the following questions: (1) what is the relative importance of willow genotype vs. the biotic and abiotic environment in structuring associated communities? (2) Do willow genetic and environmental variation have different effects on above and belowground communities? (3) What are the potential mechanisms by which willow genetic and environmental variation affects community responses?   4.2 Materials and Methods 4.2.1 Study site We conducted this research at Lanphere Dunes (40° 53’29.85”N, 124° 8’49.06”W), a pristine coastal dune ecosystem managed by US Fish and Wildlife service in Humboldt County, California, U.S.A. Coastal willow (Salix hookeriana ex Barratt ex Hooker) naturally occurs in nearshore dune swales – seasonal freshwater wetlands that form in depressions between dune ridges (Pickart 2009). Aside from coastal willow (hereafter willow), the dominant vegetation in these swales consists of beach pine (Pinus contorta ssp. Contorta) and slough sedge (Carex obnupta).    During preliminary surveys, we qualitatively identified two important sources of environmental variation for willows in the dunes – one abiotic (wind exposure) and one biotic (the presence of ant-aphid mutualisms). Willows growing in wind-exposed habitats often exhibit reduced growth, especially at the their leading edge, appearing to be “swept back” by the wind. We hypothesized that wind exposure may influence associated communities through three non-mutually exclusive mechanisms: (i) reduced plant size due to wind pruning; (ii) reduced soil moisture due to increased evaporation; and (iii) direct inhibition of ovipositing female arthropods. We also   61 observed that the aphid Aphis farinosa was an abundant herbivore at Lanphere Dunes. Aphis farinosa is usually found at the tips of new shoot growth where it feeds on willow phloem. As with many other aphid species, A. farinosa excretes carbohydrate-rich honeydew while feeding, which attracts ants that tend the aphids and feed on the honeydew. This ant-aphid interaction is often mutualistic, because the ants will defend aphids from predatory arthropods and also eat other herbivores that may be competing with the aphids (Floate & Whitham 1994; Mooney & Agrawal 2008). The ant species we observed most frequently tending A. farinosa was the western thatching ant, Formica obscuripes. Western thatching ants create distinct dome-shaped mounds from nearby plant-material and are known to reduce herbivory from leaf chewing arthropods on S. hookeriana at our study site (Crutsinger & Sanders 2005), presumably by deterring ovipositing females or predating young larva. This work suggests that the presence of aphids and the proximity to ant mounds influences associated communities through three non-mutually exclusive mechanisms: (i) increased abundance of aphid-tending ants, which could deter other arthropods; (ii) attraction of predators or deterrence of other herbivores, by aphids; (iii) alteration of plant-growth or leaf quality traits by aphids.   4.2.2 Experimental design Prior to bud burst in February 2012, we took shoot cuttings (40 cm length & ~0.5 cm diameter) from one to two replicates of 10 different willow genotypes from a pool of 26 locally collected willow genotypes planted in a large common garden experiment. Details about the establishment of this common garden are given in Barbour et al. (2015). These 10 genotypes displayed substantial variation in both plant-growth and leaf-quality traits (Barbour et al. 2015). Shoot cuttings were soaked in water overnight and then planted in a mixture of 80% perlite, 20% peat   62 moss (dolomite lime added to balance pH) inside “Cone-tainers” (Stuewe & Sons, Inc.). We grew cuttings under ambient weather conditions outside the greenhouse at Humboldt State University until we transplanted willows into multiple common gardens at Lanphere Dunes.  4.2.2.1 Ant-aphid experiment To examine how the presence of aphids, proximity to ant mounds, and willow genotype affected associated communities, we established common gardens around 5 different ant mounds (treated as blocks) in late May 2012. Within each block, we randomly planted 20 cuttings (2 replicates of each of 10 genotypes) with 0.5 m spacing in plots that were at a distance of 1, 6, and 12 meters from the edge of the ant mound, for a total of 60 cuttings per ant mound (300 cuttings for entire experiment). Within each plot, we randomly assigned the aphid treatment (aphid presence vs. absence) to one of the two replicates for each genotype. On May 22, we collected aphids (Aphis farinosa) from a single willow patch at Lanphere Dunes and placed 5 adult apterate aphids on the tips of willow cuttings in the aphid treatment using a moist paintbrush. We bagged aphids onto the apical shoots of cuttings using organza bags to promote aphid establishment on plants. Similarly, we placed organza bags on all control plants. On May 27, we checked aphid treatments to ensure there were 5 adult aphids and removed bags from all cuttings. If necessary, we added aphids to these treatments until there were 5 adults and we removed any aphid nymphs that were produced since initial establishment. We checked plants for aphids on June 6, June 13, June 24, July 4, July 14, and July 20, 2012. If plants in the aphid treatment had less than 5 apterate aphids, we noted their abundance and added aphids until there were at least 5 individuals. The ant-aphid experiment was restricted to the summer of 2012, because in the   63 summer of 2013 there was high willow mortality induced by drought and A. farinosa was too low in abundance on naturally occurring willows to allow us to repeat the experiment.  4.2.2.2 Wind experiment To examine how wind exposure and willow genotype affected associated communities in the coastal dunes, we planted 200 willow cuttings in a split-plot experimental design in late May of 2012. At each of 10 different willow patches (treated as blocks), we established an ‘exposed’ and a ‘unexposed’ common garden with exposed gardens facing prevailing winds during the growing season. Each garden consisted of one replicate cutting of each of 10 genotypes randomly planted in 2 m by 0.5 m grid with 0.5 m spacing between plants. The center of exposed and unexposed gardens within each block were the same distance (2 m) from the edge of the willow patch to control for insect accessibility. To estimate the difference in wind conditions experienced by exposed vs. unexposed plants, we went out on a representative windy afternoon in September 2012. A nearby weather station estimated wind speeds of 22 km/h during this period (Arcata, CA). We used a hand-held anemometer (Kestrel 1000) to measure wind speed at a height of 37 cm aboveground (approximate height of tallest plants in the garden in 2012) in each plot of our experiment. For each block, we randomly selected the order in which exposed and unexposed plots were measured and took maximum wind speed measurements over a 30 s period. We found that willows growing in wind-exposed plots experienced up to 3.7-fold higher wind speeds compared to unexposed plots (F1,9 = 187.32, P < 0.001), suggesting that the location of our plots were effective manipulations of wind exposure.    64 4.2.3 What is the relative importance of willow genotype vs. the biotic and abiotic environment in structuring associated communities?  To address this question, we visually surveyed plants for arthropods to determine the abundances of different (morpho)species. For the ant-aphid experiment, we surveyed arthropods on 5 different occasions between early June and late July 2012. For the wind experiment, we surveyed arthropods once at the end of July 2012 and then once a month in May, June, and July of 2013. So that individuals were not counted twice between sampling dates, we took the maximum abundance for each arthropod (morpho)species from each plant across all sampling dates within each year. This approach provides a conservative estimate of the total number of individuals of each (morpho)species that occurred on individual plants through the summer. Given the relatively low abundances of individual morphospecies, we grouped arthropods at the Family-level for insects and at the Order-level for all other arthropods prior to analyzing community composition (details in Statistical Analyses section below).   4.2.4 Do willow genetic and environmental variation have different effects on above and belowground communities? To address this question, we dug up the willows from the wind experiment to sample ectomychorriza and bacteria communities associated with willow roots in late July of 2013. We did not sample belowground communities of plants in the ant-aphid experiment due to the high mortality of plants in 2013. To sample these belowground communities, we removed willows with the surrounding soil intact to preserve root systems, separated shoots and roots, then brushed soil off root systems and stored roots in separate plastic bags. Within 6 hours of excavation, root systems were stored at 4°C. To process roots, we gently rinsed them in tap water   65 until free of visible soil. In order to randomly select roots for molecular analysis, second order roots were cut up into 2 cm lengths, spread out on a grid, and then – using a random number generator – a total of 30 cm of root length was picked from numbered grid cells. These random root subsamples were flash frozen in liquid N, and kept at -80°C until DNA extraction. To extract DNA, flash frozen root samples were physically disrupted with 2 beads per 2 mL tube (3.0 mm Yttria stabilized Zirconidea Grinding Media) for 30 seconds at 1500 strokes per minute (SPEX SamplePrep 200 geno/grinder). Total DNA was extracted from frozen root samples using MoBio PowerSoil 96 sample DNA extraction kits following the manufacture’s instructions.  To sequence and identify ectomycorrhiza and bacteria OTUs, we used custom barcode primer sets ITS1f/ITS4 and 515f/806r (Caporaso et al. 2012) to PCR amplify the fungal ITS1, 5.8S, and ITS2 region of ribosomal DNA and the V4 region of bacterial 16S ribosomal DNA from total root DNA extractions. Product quality was assessed by gel electrophoresis. PCR products were cleaned with magnetic beads, quantified with Qubit fluorometric kit, and all samples were pooled at a bacteria:fungal concentration ratio of 2:1. Pooled amplicon libraries were sequenced as single-index (the reverse barcode was uniquely indexed) 300 base pair reads at Stanford Functional Genomics Facility on one lane of an Illumina MiSeq. Reads were quality controlled by trimming low quality bases and sequenced adaptors and removing reads with average error rates greater than 0.25 using UPARSE (Edgar 2013). Only high quality, paired forward and reverse reads were used for OTU clustering at 97% identity and then checked for chimeras against the GOLD 16s rRNA database (Reddy et al. 2015) and UNITE fungal ITS database ver6_97_13.05.2014 (Kõljalg et al.) with UPARSE. Taxonomy was assigned using the RDP Classifer (Wang et al. 2007) and UNITE (ver6_97_13.05.2014) in QIIME (Caporaso et al.   66 2010). We then normalized datasets and discarded some OTUs and samples based on the following conditions: OTUs with no known taxonomy (any OTU that did not blast to at least Kingdom Fungi, Bacteria or Archaea); root samples with fewer than 6000 fungal reads; and mitochondrial and chloroplast OTUs with samples with less than 9000 bacterial reads.  4.2.5 What are the potential mechanisms by which willow genetic and environmental variation affects community responses? 4.2.5.1 Plant traits Prior work in this study system has demonstrated that variation in both plant growth and leaf quality traits affect the likelihood of willows being colonized by foliar arthropods (Barbour et al. 2015). To quantify plant-growth traits, we measured plant height, the number of shoots produced, and average shoot length in late July of each year (end of growing season) for both experiments. We quantified plant height as the distance (mm) from the ground to the tip of the tallest shoot. We quantified average shoot length by measuring every shoot on each plant to the nearest millimeter and calculating the average shoot length for each plant. We also measured several traits that could shape leaf quality for herbivores, including water content, trichome density, specific leaf area (SLA), percentage carbon (C) and nitrogen (N), and C:N. To measure these traits, we excised fully expanded and undamaged leaves from plants in late July of each year, stored leaf samples with a moist paper towel in separate plastic bags within a cooler and immediately brought them back to the laboratory. We then weighed leaves to obtain fresh mass (g), digitally scanned them to measure leaf area (mm2) using ImageJ (Abrámoff, Magalhães, and Ram 2004), and oven-dried them at 60 °C for 72 h to obtain dry weight (g) (Cornelissen et al. 2003). We calculated SLA as 𝑙𝑒𝑎𝑓 𝑎𝑟𝑒𝑎𝑑𝑟𝑦 𝑚𝑎𝑠𝑠 (Cornelissen et al. 2003). We calculated leaf water   67 content as the (𝑓𝑟𝑒𝑠ℎ 𝑚𝑎𝑠𝑠−𝑑𝑟𝑦 𝑚𝑎𝑠𝑠)𝑑𝑟𝑦 𝑚𝑎𝑠𝑠 (Munns & PrometheusWiki Contributors 2010). To measure trichome density, we counted the number of trichomes along an 11 mm by 1 mm transect in the center of the leaf, halfway between the leaf edge and the mid-vein, under a dissecting scope. To measure percentage C and N, we ground oven-dried leaves to a fine powder using a ball mill (Mixer/Mill 8000D, SPEX SamplePrep; Metuchen, NJ, USA). Subsamples of each material were then analyzed for percentage C and N on an elemental analyzer (ECS 4010; Costech Analytical Technologies, Valencia, California, USA) using atropine (4.84% N and 70.56% C) as a reference standard. For root-associated communities, we hypothesized that variation in root C:N may affect community assembly. A small subsample of roots from each plant was oven-dried, crushed with a razor blade and approximately 4 mg were flash combusted on a Carlo-Erba 1500 elemental analyzer to measure percentage C and N.  4.2.5.2 Soil characteristics   Soil nutrients, total organic matter, and moisture may all influence plant traits and the assembly of ectomychorizzal and bacterial communities on plant roots (Erlandson et al. 2015). Moreover, we expected that wind exposure would affect these soil characteristics (Lortie & Cushman 2007); therefore, we measured soil nutrients, percent organic matter, and moisture within each plot of the wind experiment (one exposed and one unexposed plot per block).   To estimate soil nutrient uptake by willows, we installed Plant Root Simulator (PRS) Probes (Western Ag Innovations, Saskatchewan, Canada) at three randomly selected locations within each plot for 11 days in September 2012. PRS Probes estimate nutrient supply rates to roots by   68 continuously adsorbing charged ionic elements over the burial period. For our study, we estimated potential root uptake of NO3+, NH4-, Ca, Mg, K, P, Fe, Mn, Cu, Zn, B, S, Pb, Al, and Cd. From this nutrient data, we calculated total N as NO3+ + NH4-, and then used principal components analysis to condense these nutrients into a single axis (nutrients PC1) that explained 34% of the variation. Nutrients PC1 described the negative correlation between nitrogen compounds (NO3+, NH4-) and the rest of the ionic elements, with positive values indicating high supply rates of all ionic elements except for the nitrogen compounds. To measure percent organic matter content (%OM), we used a trowel to collect soil (depth = 0 – 15 cm) adjacent to the randomly positioned PRS probes in September 2012. Soils were transported back to the lab in plastic bags, sieved into fragments less than 2 mm, randomly subsampled using a soil splitter, and dried at 105 °C for 72 hours. We then weighed a subsample of the oven dried soil into an oven dried crucible and placed the crucible and soil into a furnace to be combusted at 375 °C for 16 hours.  We then weighed the combusted samples, placed them in a desiccator for 20 minutes, and weighed them again. We calculated percent organic matter as %𝑂𝑀 = 𝑂𝑣𝑒𝑛 𝑑𝑟𝑦 𝑚𝑎𝑠𝑠 (𝑔) – 𝐶𝑜𝑚𝑏𝑢𝑠𝑡𝑒𝑑 𝑀𝑎𝑠𝑠 (𝑔)𝑂𝑣𝑒𝑛 𝐷𝑟𝑦 𝑀𝑎𝑠𝑠 (𝑔)× 100. To measure soil moisture (volumetric water content, m3/m3), we used a 5TE soil sensor coupled to an EM50 Digital/Analog Data Logger (Decagon Devices, Pullman, Washington, USA). In September 2012, while PRS probes were in the ground, we measured soil moisture at a depth of 5 cm in three random locations within each plot on three different days between 1100 – 1500 hours. We repeated this same sampling scheme in early July 2013. Plot levels measurements of soil moisture were highly correlated between years (Pearson’s r = 0.93, t18 = 10.91, P < 0.001), so we averaged these soil moisture estimates to determine a single soil moisture value per plot.   69  4.2.6 Statistical analyses 4.2.6.1 Community responses To examine how willow genotype, the environment, and their interaction influenced richness, abundance, and rarefied richness of aboveground arthropods as well as ectomycorrhiza and root bacteria, we used separate generalized linear mixed-effect models (GLMMs) (Bolker et al. 2009). For the ant-aphid experiment, we omitted A. farinosa and F. obscuripes from our calculations of arthropod community properties because we expected our treatments to manipulate their abundances. We specified block (ant mound) and plots nested within block (the 3 different distances from ant mound) as random effects. We specified willow genotype, aphid treatment, distance from ant mound, and their 3-way interaction as fixed effects in the model. For the wind experiment, we specified block (willow patch) and plots nested within block (the 2 wind exposure treatments) as random effects. We specified willow genotype, wind treatment, sampling year, and their 3-way interaction as fixed effects in the model. Plant mortality in each experiment resulted in unbalanced designs, so we used Type II sum-of-squares to test the significance of fixed effects. For continuous responses (rarefied richness, normalized abundances of ectomycorrhiza and bacteria) we specified Gaussian error distributions in our models and tested the significance of fixed effects using F-tests with Kenward-Roger approximated degrees of freedom. For count responses (richness and arthropod abundances), we specified Poisson error distributions in our models and tested the significance of fixed effects using likelihood-ratio tests. If necessary, we accounted for overdispersion in these Poisson models by specifying an individual-level random effect.     70 To examine whether community composition depended on willow genotype, the environment, or their interaction, we applied a Hellinger transformation to our community data (square root of proportional abundance of species found on each willow;  (Legendre & Gallagher 2001) and conducted separate redundancy analyses (RDA, 1000 permutations on Euclidean distances) for the arthropod, ectomycorrhiza, and bacteria communities. A Hellinger transformation was appropriate because calculating Euclidean distances on raw abundance data can sometimes result in two communities that do not share the same species being more similar than two communities that do share the same species (Legendre & Gallagher 2001). We incorporated the same fixed effects structured as we used to analyze the univariate community responses for each experiment. To test the significance of each effect, we used Type II sum-of-squares and compared the observed community dissimilarities to the dissimilarities we would expect by random chance with a permutation test that controls for the blocked design of our experiment. To test the significance of treatments that varied at the plot-level (wind exposure and distance from ant mound), we first calculated the community’s centroid in multivariate space for each plot. We then included block as a covariate and ran the same permutation test as previously described. This ensured that our significance tests of treatments that varied at the plot-level were based on the appropriate residual degrees of freedom (wind exposure residual df = 9; distance from ant mound residual df = 4).  4.2.6.2 Plant traits  To analyze how willow genotype, the environment, and their interaction influenced willow phenotypes, we used separate GLMMs with the same structure described in the “Community responses” section. For the wind experiment, we lacked multiple years of data on leaf trichome   71 density (2012 only), SLA (2013 only), leaf C:N (2013 only), and root C:N (2013 only); therefore, we removed sampling year, and its interactions, from the fixed effects structures of these GLMMs.  4.2.6.3 Soil characteristics To examine the effect of wind exposure on soil characteristics (total N, nutrients PC1, %OM, and soil moisture), we used separate mixed effect models with wind treatment as a fixed effect and block (willow patch) as a random effect. Since all soil characteristics were continuous responses, we specified Gaussian error distributions in our models and tested the significance of fixed effects using F-tests with Kenward-Roger approximated degrees of freedom.  4.2.6.4 Direct and indirect effects To identify potential mechanisms by which willow genetic and environmental variation affects community responses, we used piecewise structural equation models (SEMs) (Lefcheck 2015). We only modeled potential mechanisms and community responses that were statistically significant in our prior analyses. For example, if we did not observe a G×E effect on a community response variable, then we did not model this interactive effect in the piecewise SEM.    An advantage of piecewise SEMs is that they are flexible, allowing users to account for correlated structure (i.e. random effects) in their experimental design. However, as with any technique that relies on multiple regression, structural equation models can give misleading results if there is collinearity among predictor variables. To mitigate the effects of collinearity,   72 we used principal components analysis (PCA) to condense aboveground willow phenotypes as well as soil properties into a small number of uncorrelated variables. For aboveground willow traits in the wind experiment, we analyzed separate PCAs for 2012 and 2013 since we did not always have data on the same traits in each year. At times, we lacked data for all traits on each plant or all soil properties measured in each plot. Therefore, we used a regularized iterative PCA algorithm to impute missing values (Josse et al. 2012). For each PCA, we retained principal components with eigenvalues greater than 1.   To calculate standardized coefficients in our piecewise SEM, we scaled all predictor and response variables to mean = 0 and SD = 1 prior to analyzing them with GLMMs (error distribution = Gaussian). For willow genotype, we specified the average effect for the 10 genotypes as the reference level (i.e. deviation contrasts) and calculated the standard deviation of the coefficients to determine its standardized coefficient. To evaluate the explanatory power of our separate GLMMs, we report marginal R2 (Nakagawa & Schielzeth 2013). Marginal R2 values do not adjust for the variance explained by our random effects; therefore, they give us a truer sense of the explanatory power of our models. To evaluate the fit of the full structural equation model, we used a test of directed separation (Shipley 2000). This test identifies missing paths in the model, calculates the P-value for each missing pathway, and then calculates a test statistic, Fisher’s C, using the following equation: C = -2∑ ln(𝑃𝑖)𝑘𝑖=1 , where 𝑃𝑖 is the P-value of the ith missing pathway and k is the total number of missing pathways. Fisher’s C can then be compared to a chi-square distribution with 2k degrees of freedom. If there are many missing pathways with low P-values, this will result in a lower P-value for the structural equation model. Therefore, a P-value < 0.05 indicates a poor fit for the structural equation model, whereas a P-value > 0.05   73 indicates a good fit. Note that if we have included the key plant traits as well as biotic and abiotic factors, then there should be no missing paths between willow genotype and our environmental treatments.   All analyses were conducted in R version 3.2.4 (R Core Team 2016).  4.3 Results 4.3.1 What is the relative importance of willow genotype vs. the biotic and abiotic environment in structuring associated communities? 4.3.1.1 Ant-aphid experiment Willow genotype and the biotic environment had independent and interactive effects on the arthropod community (Table 4.1). We found that arthropod richness varied from 1.2 to 3.2 species among genotypes (Fig. 4.1A), while arthropod abundance varied 4-fold among the different clones (Fig. 4.1B). The effect of willow genotype on arthropod richness was explained by correlated responses in arthropod abundance, as there was no difference in rarefied richness among genotypes (Table 4.1). Aphid treatment was the only factor that affected rarefied richness (Table 4.1), leading to a 16% decrease in rarefied richness when aphids were added to willows (Fig. 4.1D); however, this effect of aphid treatment did not translate into an effect on total richness Table 4.1). Willows in the aphid treatment also had 2-fold more arthropods, but only at the furthest distance from ant mounds (Eaphid×Eant, Table 4.1, Fig. 4.1C). Proximity to ant mounds did not influence any other aspect of the arthropod community (Table 4.1). Arthropod community composition was influenced by an interaction between willow genotype and the aphid treatment (Table 4.1, Fig. 4.1E). This G×Eaphid effect was primarily due to the differential   74 response of other aphids to a single willow genotype (Appendix C.1, Fig. 4.1F). If we remove this genotype from the analysis, we still find strong, but independent effects of willow genotype (F8,156 = 1.66, P = 0.007) and the addition of aphids (F1,156 = 2.93, P = 0.017) on community composition.      75  Table 4.1 Summary of statistical models in ant-aphid experiment  Responses Genotype (G) Eaphid Eant G×Eaphid G×Eant Eaphid×Eant G×Eaphid×Eant Foliar arthropods        Total richnessa 41.35(9) 0.45(1) 1.12(1) 7.66(9) 7.84(9) 1.15(1) 6.17(9) Total abundancea 34.86(9) 1.79(1) 1.42(1) 14.61(9) 9.00(9) 8.12(1) 9.18(9) Rarefied richnessb 0.83(9,138.8) 5.34(1,139.2) 0.46(1,8.2) 1.81(9,139.7) 0.94(9,139.7) 0.41(1,140.6) 0.70(9,139.4) Community compositionc 1.52(9,176) 2.04(1,176) 1.01(1,9) 1.45(9,157) 0.97(9,157) 0.69(1,157) 0.93(9,148) Ant-aphid interactions         A. farinosa abund.a 20.83(9) - 0.55(1) - 10.25(9) - - F. obscuripes abund.a 2.42(1*) 9.77(1) 1.68(1) 6.26(2*) - - - Plant traits        Heightb 15.83(9,204.2) 0.63(1,204.3) 0.31(1,9.1) 0.93(9,204.5) 0.98(9,204.4) 0.07(1,204.3) 1.62(9,204.7) Shoot countb 65.84(9) 2.76(1) 0.21(1) 12.11(9) 8.80(9) 4.20(1) 9.21(9) Shoot lengthb 7.27(9,204.2) 2.39(1,204.2) 0.10(1,9.1) 1.05(9,204.5) 0.70(9,204.3) 1.24(1,204.3) 0.56(9,204.6) Leaf trichome densitya 38.17(9) 0.44(1) 0.81(1) 23.17(8) 8.41(9) 0.84(1) - log(Leaf water content)b 1.33(9,69.6) 0.01(1,69.4) 1.02(1,7.1) 0.48(8,70.4) 0.79(9,69.5) 0.36(1,70.6) 1.02(7,72.0)  Summary of statistical models that analyze the effects of willow genotype, aphid treatment, and distance from ant mounds on the arthropod community, ant-aphid interactions, and plant traits. We report the test statistic and include the degrees of freedom for each test in parentheses. Font type denotes statistical significance (bold P < 0.05, italic P < 0.10, normal P > 0.10). Notes: aLikelihood-ratio test and degrees of freedom calculated using a generalized linear mixed-effect model (error distribution = Poisson, link function = log); bF-test and Kenward-Roger approximated degrees of freedom calculated using a linear mixed-effect model; cF-test calculated using redundancy analysis on Hellinger-transformed community data; *indicates that predictor was modeled as a random effect and its significance was determined using a likelihood ratio test.   76 Figure 4.1 Arthropod community responses in ant-aphid experiment  Responses of the arthropod community to genetic variation within the willow Salix hookeriana, the addition of the aphid Aphis farinosa, and proximity to mounds of the ant Formica obscuripes. We found that willow genotype influenced both the total richness (A) and abundance (B) of arthropods. Arthropod abundance was also influenced by the addition of aphids, but only at the furthest distances from ant mounds (C). The addition of aphids reduced the probability of encountering a different arthropod species (rarefied richness) by 16% across all treatments (D). We also found that the addition of aphids modified the effect of willow genotype on the composition of the arthropod community (E). This interaction between willow genotype and aphid treatment was solely due to the differential effect of genotype J on the abundance of non-A. farinosa aphids in the aphid treatment (F). Symbols and error bars correspond to the response variable’s mean ± 95% confidence interval. We calculated mean and confidence intervals based on the full models Table 2.1) using the ‘effects’ package in R. Black squares correspond to the effect of willow genotype after controlling for   77 other treatments, while grey diamonds and white circles represent the aphid treatment and control, respectively.    78 4.3.1.2 Wind experiment The abiotic environment and willow genotype had strong, but independent effects on the arthropod community (Table 4.2). In particular, willows growing in wind-exposed plots hosted 51% fewer species, 47% fewer individuals, and 60% fewer rarefied species compared to unexposed willows (Fig. 4.2A,C,E). In spite of the effects of wind exposure, willow genotype had a strong effect on both the richness (~3-fold differences, Fig. 4.2B) and abundance (~5-fold differences, Fig. 4.2D) of arthropods, but only a marginal effect on rarefied richness (Fig. 4.2F). Arthropod communities on willows had both more species and more individuals in the second year of the experiment compared to the first (Table 4.2); however, we also conducted more arthropod surveys for the wind experiment in 2013 vs. 2012. In terms of community composition, we observed strong effects of wind exposure by the end of experiment (Table 4.2, Fig 4.2A). These compositional differences were due to several key arthropod taxa (gall midges, leaf-mining moths, and spiders) being less abundant on wind-exposed willows, whereas leaf-tiering moths were insensitive to wind exposure (and therefore relatively more abundant; Appendix C.2). Although several arthropod taxa varied in abundance among willow genotypes (Appendix C.2), we did not detect an effect of genotype on community composition in either year of the experiment (Table 4.2).  79 Table 4.2 Summary of statistical models in wind experiment Responses Genotype (G) Ewind Eyear G×Ewind G×Eyear Ewind×Eyear G×Ewind×Eyear Foliar arthropods        Richnessa 28.01(9) 10.33(1) 13.55(1) 3.74(9) 9.85(9) 0.92(1) 7.04(9) Abundancea 25.25(9) 5.48(1) 6.72(1) 7.33(9) 8.22(9) 1.65(1) 11.85(9) Rarefied richnessb 1.96(9,71.1) 22.82(1,7.8) 1.13(1,82.7) 0.66(9,80.9) - 0.67(1,81.9) - Community composition2012 0.96(9,51) 1.26(1,7)  0.91(9,42)    Community composition2013 1.17(9,68) 5.70(1,9)  0.69(6,62)    Root-associated Mycorrhiza        Richness2013b 1.28(9,95.0) 1.01(1,8.8) - 1.23(9,95.8) - - - Abundance2013b 0.80(9,95.5) 0.36(1,8.7) - 1.03(9,96.4) - - - Rarefied richness2013b 0.87(9,95.1) 0.88(1,8.8) - 0.93(9,95.9) - - - Community composition2013 1.01(9,117) 1.18(1,9) - 0.87(9,108) - - - Root-associated Bacteria     - - - Richness2013b 1.35(9,100.8) 4.53(1,7.9) - 0.87(9,101.5) - - - Abundance2013b 1.39(9,102.3) 2.00(1,8.0) - 0.64(9,103.2) - - - Rarefied richness2013b 1.48(9,99.9) 6.03(1,7.8) - 1.35(9,100.5) - - - Community composition2013 0.93(9,120) 1.38(1,9) - 0.87(9,111) - - - Soil characteristics         Total Nb - 5.08(1,9) - - - - - Soil moistureb - 3.52(1,9) - - - - - Percent organic matterb - 0.68(1,8.4) - - - - - Nutrient PC1b - 1.31(1,9) - - - - -                                             80 Responses Genotype (G) Ewind Eyear G×Ewind G×Eyear Ewind×Eyear G×Ewind×Eyear Plant traits        Heightb 9.13(9,145.3) 29.10(1,9.0) 210.09(1,156.3) 0.71(9,147.9) 0.80(9,157.8) 16.69(1,158.4) 1.84(9,160.9) Shoot counta 47.42(9) 9.91(1) 5.68(1) 10.70(9) 18.26(9) 12.53(1) 5.76(9) Shoot lengthb 4.97(9,144.2) 10.44(1,9.0) 75.36(1,158.5) 0.84(9,146.9) 1.61(9,160.1) 0.05(1,160.7) 0.70(9,163.2)         Leaf water contentb 4.90(9,129.0) 0.97(1,8.7) 2.93(1,139.7) 0.47(9,132.0) 2.80(9,141.6) 2.03(1,141.5) 1.56(9,144.1) Leaf trichome density2012b 67.31(9) 0.02(1) - 10.45(9) - - - SLA2013b 4.21(9,122.5) 0.34(1,8.9) - 1.19(9,123.4) - - - Leaf C:N2013b 4.88(9,70.48) 1.54(1,7.8) - 1.31(9,71.6) - - - Root C:N2013b 0.85(9,107.0) 0.31(1,8.7) - 0.33(9,107.5) - - -  Summary of statistical models that analyze the effects of willow genotype and wind exposure on associated communities, soil characteristics, and plant traits. We report the test statistic and include the degrees of freedom for each test in parentheses. Font type denotes statistical significance (bold P < 0.05, italic P < 0.10, normal P > 0.10). Notes: aLikelihood-ratio test and degrees of freedom calculated using a generalized linear mixed-effect model (error distribution = Poisson, link function = log); bF-test and Kenward-Roger approximated degrees of freedom calculated using a linear mixed-effect model; cF-test calculated using redundancy analysis on Hellinger-transformed community data; *indicates that predictor was modeled as a random effect and its significance was determined using a likelihood ratio test.    81 Figure 4.2 Arthropod community responses in wind experiment  Arthropod community responses to wind exposure and genetic variation within the willow Salix hookeriana. We found that both wind exposure and willow genotype had strong, but independent effects on the arthropod community. Specifically, arthropod communities on wind-exposed willows had lower richness (A), abundance (C), and rarefied richness (E) compared to unexposed willows. Willow genotype had a strong effect on the richness (B) and abundance (D) of arthropods, but only a marginal effect on rarefied richness (F). Points and error bars correspond to the response variable’s mean ± 95% confidence interval. We calculated mean and confidence intervals based on the full models Table 4.2) using the ‘effects’ package in R.  82  4.3.2 Do willow genetic and environmental variation have different effects on above and belowground communities? Willow genotype and wind exposure had distinct effects on root-associated ectomycorrhizal and bacterial communities compared to foliar arthropods. Neither wind exposure nor willow genotype influenced the richness, abundance, or rarefied richness of ectomycorrhiza OTUs (Table 4.2). However, willow genotype explained 7% of the variation in the composition of the ectomycorrhizal community (Fig. 4.3B) with no detectable effect of wind-exposure (Table 4.2). In contrast to the ectomycorrhizal community, wind exposure slightly influenced multiple indices of the bacteria community (Table 4.2), but in the opposite direction of foliar arthropods. For example, the roots of wind-exposed plants tended to host more bacteria OTUs than unexposed plants (10% increase, Table 2). The effect of wind exposure on bacterial richness was likely a result of the significant increase in rarefied richness on wind-exposed plants (Table 4.2), but the effect size for rarefied richness was very small (wind-exposed mean = 0.9993, unexposed mean = 0.9992). While wind exposure did not affect the total abundance of bacteria OTUs (Table 4.2), it had a marginal effect on the composition of the bacteria community (Fig. 4.3C). There was no detectable effect of willow genotype on any aspect of the bacteria community.    83 Figure 4.3 Compositional responses of above and belowground communities  Community dissimilarity of foliar arthropods (A) as well as root-associated ectomycorrhiza (B) and bacteria (C) in response to wind exposure and genetic variation within the willow Salix hookeriana. Black text and grey ellipses correspond to the community centroid ± 95% confidence interval. Grey numbers denote blocks and each unique number is the community centroid for the plot within each block. Grey circles mark the location of individual willow communities in multivariate space. We calculated the locations of centroids ± 95% confidence interval and individual samples using redundancy analysis on Hellinger-transformed community data.  4.3.3 What are the potential mechanisms by which willow genetic and environmental variation affects community responses? 4.3.3.1 Ant-aphid experiment  We hypothesized that the effect of willow genetic variation and the biotic environment on arthropod communities would be mediated, in part, by variation in the abundance of A. farinosa and F. obscuripes. While distance from ant mounds had little effect on A. farinosa, willow genotype had a strong effect, with the average number of aphids ranging from 0.05 to 7 among the most disparate willow genotypes in the aphid treatment (Fig. 4.4A, Table 1). This strong effect of willow genotype on A. farinosa in the aphid treatment resulted in a G×Eaphid effect on the abundance of F. obscuripes (Table 4.1), with ant abundance varying from 0 to ~0.5  84 individuals (on average) among clones in the aphid treatment, whereas they were virtually absent in the absence of aphids (Fig. 4.4B). Proximity to ant mounds had no effect on the abundance of F. obscuripes (Table 4.1).  Figure 4.4 Aphid, ant, and plant trait responses  Variability in ant-aphid interactions and plant traits explained by genetic variation within the willow Salix hookeriana, the addition of the aphid Aphis farinosa, and distance from mounds of the ant Formica obscuripes. In the aphid treatment, we found that willow genotype influenced the abundance of the aphid Aphis farinosa (A). The effect of willow genotype on A. farinosa resulted in willow genotype determining the abundance of the ant Formica obscuripes, but only in the aphid treatment (B). Plant height was solely determined by willow genetic variation (C). In contrast, the aphid treatment modified the effect of willow  85 genotype on leaf trichome density (D). Symbols and error bars correspond to the response variable’s mean ± 95% confidence interval. We calculated mean and confidence intervals based on the full models Table 4.1) using the ‘effects’ package in R. Black squares correspond to the effect of willow genotype after controlling for other treatments, while grey diamonds and white circles represent the aphid treatment and control, respectively.   In addition to ant-aphid interactions, we hypothesized that the effect of willow genetic variation and the biotic environment on arthropod communities would be mediated by plant traits. We observed both direct and interactive effects of willow genotype and the biotic environment on plant traits (Table 4.1). All of the plant-growth traits we measured varied approximately 2-fold among the most disparate willow genotypes (Table 4.1, Fig. 4.4C). Willows did appear to produce 28% more shoots in the absence of aphids, but only at the furthest distance from ant mounds (Eaphid×Eant effect, Table 4.1). While there was little apparent effect of willow genotype and the biotic environment on leaf water content (Table 4.1), we found that the addition of aphids modified the effect of certain willow genotypes on leaf trichome density (G×Eaphid effect, Table 4.1). Specifically, two clones (S and T) produced ~4-fold more trichomes when aphids were absent, whereas genotype L produced 3-fold more trichomes when aphids were present (solid lines in Fig. 4.1F).   We used structural equation models (for richness, abundance, and rarefied richness) and redundancy analysis (for community composition) to tease apart the direct and indirect effects of willow genetic variation and the biotic environment on the arthropod community. For the traits we measured, we found that the indirect effect of willow genotype on arthropod richness and abundance was mediated primarily by plant trait PC1 (Fig. 4.5A,B). Plant height, shoot count,  86 and shoot length all had strong, positive loadings on trait PC1 (Appendix C.3), indicating that larger willows hosted more arthropod species and individuals. Arthropod abundance was also positively influenced by the addition of aphids, primarily because A. farinosa also attracted other ant species (Pearson’s r = 0.42, t282 = 7.74, P < 0.001; Fig. 4.5D) and these other ants were the second most abundant taxonomic group in the community. In contrast to total abundance, the addition of aphids negatively affected rarefied richness. This negative effect was due in part to aphid additions attracting more F. obscuripes, an active generalist predator that likely consumed or inhibited the colonization of other arthropods. In terms of composition, we found that the abundance of A. farinosa was the only factor (of the mechanisms we modeled) influencing the arthropod community. Specifically, higher abundance of A. farinosa resulted in an increase in the proportional abundance of other ant species in the community (Fig. 4.5D).  Despite our detailed analysis of potential mechanisms, our structural equation models revealed multiple missing paths (dotted lines, Fig. 4.5A,B,C), resulting in rather poor fits for most models (richness: C2 = 8.49, P = 0.014; abundance: C32 = 37.66, P = 0.226; rarefied richness: C4=11.10, P = 0.025). For example, after accounting for the traits we measured, willow genotype still had a strong effect on arthropod richness (Fig. 4.5A) and A. farinosa abundance (Fig. 4.5B), indicating that we failed to identify key pathways (likely unmeasured traits) by which genetic variation influenced these responses. Similarly, we failed to fully identify the Eaphid×Eant effect on arthropod abundance (Fig. 4.5B) as well as how the addition of aphids negatively affected rarefied richness (Fig. 4.5C). For our redundancy analysis of community composition, we found that A. farinosa abundance explained the effect of the aphid treatment (F1,173 = 0.90, P = 0.447), but we still failed to detect the effect of both willow genotype (F9,173 = 1.53, P = 0.014) and the  87 G×Eaphid effect (F9,164 = 1.71, P = 0.004), suggesting that we failed to measure important constitutive and inducible plant traits.   Figure 4.5 Mechanisms of community assembly in ant-aphid experiment  Statistical models of the processes mediating arthropod community assembly in the ant-aphid experiment. Piecewise structural equation models of arthropod richness (A), abundance (B), and rarefied richness (C). Colored and white boxes represent exogenous and endogenous variables, respectively. Solid, single-headed arrows correspond to modeled pathways between predictor and response variables, and may be either positive (black) or negative (red). Grey, double-headed arrows denote variables with no direct relationship and that we assumed to be driven by the same underlying factor. For clarity, we only plotted paths with standardized coefficients > 0.10. Numbers next to all arrows represent the standardized path coefficient, which also corresponds to the thickness of arrows. (B) Redundancy analysis illustrating the effect of plant traits (Trait PC1 & PC2) on arthropod community composition (Hellinger-transformed = square root of  88 proportional abundances of species found on each willow). Black and blue arrows correspond to plant traits and species, respectively, while grey dots represent the position of individual willow communities.  4.3.3.2 Wind experiment One of the mechanisms by which wind exposure could influence willow-associated communities is through accumulated effects on soil properties; however, we observed only modest effects of wind exposure on soil properties (Table 4.2). Specifically, soil in wind-exposed plots was marginally drier (Fig. 4.6A) with higher amounts of total Nitrogen (Fig. 4.6B) than in unexposed plots, but there was no clear difference in either percent organic matter or nutrient composition (Table 4.2).   As with the ant-aphid experiment, we hypothesized that the effects of wind exposure and willow genotype on associated communities would be mediated by plant traits. Interestingly, we found that plant-growth and leaf quality traits responded differently to wind exposure and willow genetic variation (Table 4.2). For example, wind exposure negatively affected all plant-growth traits (Table 4.1). Moreover, the negative effects of wind exposure were magnified by the end of the experiment for both plant height (Fig. 4.6C) and the number of shoots produced (Table 4.2). Still, willow genotype had a pronounced effect on all plant-growth traits, resulting in willows that varied over 2-fold in height (Fig. 4.6D), number of shoots, and shoot length among the most disparate genotypes. While the effect of willow genotype on shoot length changed by the end of the experiment (Table 4.2), this G×Eyear effect was relatively small (R2 = 0.05) compared to the effect of genotype alone (R2 = 0.13). In contrast to plant-growth traits, willow genotype was the primary factor in determining leaf traits across both years of the experiment (Table 4.2). The leaves of willow genotypes varied 46-fold in trichome density, 1.5-fold in SLA, and 1.6-fold in  89 C:N (Fig. 4.6E). We had data available on leaf water content for 2012 and 2013, and we found that the amount of variation explained by willow genotype depended on the sampling year (2012, R2 = 0.11; 2013, R2 = 0.16). Unlike aboveground plant traits, root C:N did not appear to be influenced by either wind exposure or willow genotype (Table 4.2, Fig. 4.6F).  Figure 4.6 Soil characteristics and plant trait responses   90 Variability in soil characteristics and plant traits explained by wind exposure and genetic variation within the willow Salix hookeriana. Wind exposure had marginal effects on both soil moisture (A) and Nitrogen availability (B). The negative effect of wind exposure on plant height was magnified in the second year of the experiment (C); however, plant height still varied ~2-fold among the most disparate willow genotypes (D). Willow genotype was a good predictor of leaf C:N (E), but a poor predictor of root C:N (F). Symbols and error bars correspond to the response variable’s mean ± 95% confidence interval. We calculated mean and confidence intervals based on the full models Table 4.2) using the ‘effects’ package in R.    In contrast to the ant-aphid experiment, our structural equation models provided good fits to our data (i.e., P > 0.05), indicating that we identified the key processes affecting the richness (Fig. 4.7A; C38 = 28.83, P = 0.858), abundance (C38 = 32.8, P = 0.709), and rarefied richness (C38 = 21.63, P = 0.985) of willow-associated communities.   Aboveground, we found that wind exposure had a direct, negative effect on arthropod richness (Fig. 4.7A), abundance (std. coef. = -0.08), and rarefied richness (std. coef. = -0.26). In addition, we found that both trait PC1 and PC2 mediated the indirect effects of wind exposure (negative) and willow genetic variation on the arthropod community (Fig. 4.7A). Trait PC1 had a strong, positive effect on arthropod richness (Fig. 4.7A), abundance (std. coef. = 0.28), and rarefied richness (std. coef = 0.37). Similar to the ant-aphid experiment, trait PC1 had strong, positive associations with plant height, shoot count, and shoot length (Appendix C.3), indicating that larger willows hosted more arthropod species. Trait PC2 had a smaller, but negative effect on arthropod richness (Fig. 4.7A), abundance (-0.15), and rarefied richness (-0.12). Trait PC2 had a strong positive correlation with leaf C:N, but strong negative correlations with leaf water content and SLA (Appendix C.3), indicating that willows with poorer quality leaf tissue hosted fewer  91 arthropod species. These qualitative patterns held for the richness, abundance, and rarefied richness of foliar arthropods in the first year of the experiment as well (C22 = 26.02, P = 0.251), except that trait PC2 was determined by different traits (Appendix C.3) and did not appear to affect any aspect of the arthropod community (richness, P = 0.657; abundance, P = 0.104; rarefied richness, P = 0.850). For community composition, we only analyzed the data from the second year of the experiment because this was the only year for which we detected a significant effect of wind exposure (Table 4.2). We found that the effects of wind exposure on community composition were primarily mediated by plant trait PC1. Positive values of trait PC1 (i.e. larger plants) had greater proportional abundance of gall midges, leaf-mining moths, and spiders, whereas leaf-tiering moths were insensitive to plant size (Fig. 4.7B).  Belowground, we found that different processes determined the structure of root-associated ectomycorrhiza and bacteria communities. For example, soil PC1, and to a lesser extent root C:N, negatively affected ectomycorrhiza richness (Fig. 4.7A), abundance (std. coefs: soil PC1 = -0.28; root C:N = -0.15), and rarefied richness (std. coefs: soil PC1 = -0.28; root C:N = -0.22). Soil PC1 had strong positive correlations with soil moisture and organic matter, but negative correlations with NO3- and NH4+, indicating that ectomycorrhiza communities were more diverse in drier environments with more available nitrogen. In contrast, soil PC2 was the primary factor in determining bacteria richness (Fig. 4.7A), abundance (std. coef. = 0.23), and rarefied richness (std. coef. = 0.28). Micronutrients such as Ca2+, Mg2+, and Cd2+ had strong positive loadings on soil PC2, indicating that bacteria richness was greater in environments with more of these    92 Figure 4.7 Mechanisms of community assembly in wind experiment  Statistical models of the processes mediating community assembly in the wind experiment. (A) Piecewise structural equation model of the richness of foliar arthropods as well as root-associated ectomycorrhiza and bacteria. Colored and white boxes represent exogenous and endogenous variables, respectively. Solid, single-headed arrows correspond to modeled pathways between predictor and response variables, and may be either positive (black) or negative (red). Grey, double-headed arrows denote variables with no direct relationship and that we assumed to be driven by the same underlying factor. For clarity, we only plotted paths with standardized coefficients > 0.10. Numbers next to all arrows represent the standardized path coefficient, which also corresponds to the thickness of arrows. (D) Redundancy analysis illustrating the effect of plant traits (Trait PC1 & PC2) on arthropod community composition (Hellinger-transformed = square root of proportional abundances of species found on each willow). Black and blue arrows correspond to plant traits and species, respectively, while grey dots represent the position of individual willow communities.   93 micronutrients. Although we detected clear effects of soil properties and root C:N on richness, abundance, and rarefied richness of root-associated communities, none of these characteristics were strong predictors of their compositions (Table 4.3). Indeed, although we detected a significant effect of willow genetic variation on ectomycorrhizal composition (Table 4.2), we failed to identify the process mediating the effect of willow genotype (F9,106 = 1.03, P = 0.002). Our failure to identify this process is not surprising though, given that we measured only one belowground plant trait (root C:N) and it was not strongly influenced by willow genotype (Table 4.2).   Table 4.3 Redundancy analyses of above and belowground communities Community composition Ewind Trait PC1 Trait PC2 Root C:N Soil PC1 Soil PC2 Arthropods 0.83(1,9) 12.05(1,76) 0.65(1,76) - - - Mycorrhiza  - - - 1.17(1,115) 1.89(1,8) 0.85(1,8) Bacteria - - - 1.31(1,116) 1.90(1,8) 0.81(1,8)  Redundancy analyses of foliar arthropods and root-associated ectomycorrhiza and bacteria. We report F-statistics and degrees of freedom in parenthesis. Font type denotes statistical significance (bold P < 0.05, italic P < 0.10, normal P > 0.10).   4.4 Discussion 4.4.1 What is the relative importance of willow genotype vs. the biotic and abiotic environment in structuring associated communities? The relative importance of host-plant genetic effects on associated arthropod, bacterial, and ectomycorrhizal communities depended on the type of environmental gradient. In the ant-aphid  94 experiment, willow genetic variation tended to have a stronger effect on arthropod community structure compared to aphid additions and proximity to ant mounds. Still, we did find that aphid additions modified the effect of willow genotype on the composition of the arthropod community, suggesting that accurately predicting community assembly requires an understanding of both factors. In the wind experiment, we found that wind exposure trumped willow genotype in the strength of its effect on foliar arthropods and root-associated bacteria; however, willow genotype was the only factor that influenced the composition of root-associated ectomycorrhiza. Moreover, despite the importance of wind exposure in shaping arthropod composition, willow genotype still had predictable effects on individual arthropod guilds. Taken together, our study suggests that host-plant genetic variation, aphid presence, and the amount of wind exposure can be equally important in structuring associated communities in coastal dunes ecosystems.  4.4.2 Do above and belowground communities differ in their responses to willow genetic and environmental variation? Although diverse assemblages of above and belowground taxa colonize host plants, there are no genotype-by-environment studies, to our knowledge, that have simultaneously measured the responses of above and belowground assemblages. We found that foliar arthropods, ectomycorrhiza, and root microbes all responded differently to willow genetic and environmental variation, suggesting that these communities are responding to different plant traits and environmental correlates of wind exposure. Similarly, Lamit et al. (2015) found that communities of foliar arthropods and ectomycorrhiza did not covary across genotypes of narrowleaf cottonwood (Populus angustifolia). The lack of covariation between these  95 communities is likely because these assemblages are responding to different plant traits, although we did not fully identify the root traits mediating ectomycorrhiza and bacteria colonization. While we have made substantial progress in the past decade understanding plant-arthropod interactions, it is time that community genetics research turns its attention belowground to understand the plant traits influencing these diverse assemblages. This will have the added benefit of understanding associations between above and belowground traits which will be important for predicting when we would expect linkages between above and belowground communities.  4.4.3 What are the potential mechanisms by which willow genetic and environmental variation affects community responses? 4.4.3.1 Ant-aphid experiment Our study supports an emerging generalization that host-plant genetic variation has strong bottom-up effects on ant-aphid interactions (Johnson 2008; Mooney & Agrawal 2008; Abdala‐Roberts et al. 2012). As with other studies, we found that ant abundances were strongly mediated by the effects of host-plant genetic variation on aphid densities (Johnson 2008; Mooney & Agrawal 2008). Johnson (2008) found that heritable variation in leaf water content and trichome density were important determinants of the densities of Aphis oestlundi on evening primrose (Oenothera biennis); however, leaf water content and trichome density were not associated with A. farinosa abundance in our study, suggesting the traits mediating plant-aphid interactions are species specific (Züst & Agrawal 2016).   96 In contrast to previous work (Johnson 2008; Mooney & Agrawal 2008), we found that the effects of host-plant genotype on the richness and abundance of other arthropods were mediated, in part, by genotypic differences in plant size rather than ant abundance. One possible reason for this discrepancy is that the absolute variation in F. obscuripes abundance was quite low in our experiment (max. genotype average = ~0.5 individuals) compared to other studies (max. genotype average = ~3 individuals), which was likely due to the rather low abundance of aphids we observed (max. genotype average = 7 individuals). While we failed to fully identify the mechanisms (likely unmeasured plant traits) explaining arthropod community responses in this experiment, our results do suggest that host-plant evolution could have strong effects on arthropod community structure, despite variability in this biotic factor.   4.4.3.2 Wind experiment Our study supports the notion that wind is a key environmental factor in structuring communities associated with host plants in coastal dune ecosystems (Miller and Weis 1999; Crutsinger et al. 2010, 2014). We found that the negative effects of wind exposure on arthropod richness and abundance resulted from a combination of direct effects on colonization as well as indirect effects mediated by wind pruning and, to a lesser extent, reductions in leaf quality. Similarly, Crutsinger et al. (2014) found that there were more arthropod species and individuals on prostrate vs. erect morphs of coyote bush (Baccharis piluaris), due to their low-lying growth form which enabled them to be more productive than erect morphs at their windy coastal dunes site.    97 Few studies have examined community-level responses to natural variability in the abiotic environment and host-plant genetic variation, making it difficult to draw other useful comparisons. The majority of genotype-by-abiotic environment studies to date have used fertilizers to manipulate soil nutrient availability (Abdala-Roberts & Mooney 2013, Orians & Fritz 1996, Ross & Stiling 1998), but it is unclear whether these manipulations reflect natural variation in soil nutrients. This may explain why the effects of variation in soil nutrients range from being independent and weak (Abdala-Roberts & Mooney 2013) to being strong modifiers (Orians & Fritz 1996) of host-plant genotype on arthropod communities. The only other genotype-by-abiotic environment study that we are aware of manipulated sun exposure by sea daisy (Borrichia frutescens, Ross & Stiling 1998) and, similar to our study, observed strong, independent effects of sun exposure on densities of the gall midge, Asphondylia borrichiae, presumably through changes in carbon-based secondary metabolites. If we are to make progress on understanding the relative importance of willow genotype vs. environment for associated communities, future experimental work should focus on manipulating natural variation in specific abiotic factors, or, at the very least, measuring variability in abiotic factors to begin to identify putative causal factors.  4.4.4 Conclusions Overall, our study reinforces the importance of host-plant genetic variation in shaping associated communities and extends this finding to natural biotic and abiotic gradients in coastal dune ecosystems. Our findings also suggest that predicting community responses to genetic and environmental variation is a complex task that may depend on historical processes that have shaped the genetic architecture for the populations of interest (i.e. sensitivity to specific  98 environmental factors). Still, the effects of willow genetic variation were clear at both the level of plant traits and the community structure of foliar arthropods and ectomycorrhiza. Importantly, this suggests that host-plant evolution could have a strong influence on the biodiversity of above and belowground communities. Future studies should work toward understanding how these diverse communities feedback to impose selection pressures on host plants as well as how host-plant traits mediate interactions between above and belowground communities. In doing so, we will be able to work toward a more synthetic understanding of the evolutionary ecology of host plants and their diverse associated communities.   99 Chapter 5: Conclusion To summarize, I found that heritable trait variation within the host plant Salix hookeriana can have cascading effects on consumer-resource interactions across multiple trophic levels (Chapter 2, 3, and 4). The effect of host-plant genetic variation was important despite natural variation in biotic (ant-aphid interactions) and abiotic (wind exposure) environmental factors (Chapter 4). When host-plant genotypes support distinct sets of trophic interactions, there is a clear expectation for increasing population-level genetic diversity to increase food-web complexity (Chapter 3). My dissertation informs us about how intraspecific variation in key species can affect the structure of species-rich communities. Moreover, by demonstrating the link between heritable trait variation and the structure of species-interaction networks, my research suggests that host-plant evolution can affect the dynamics of ecological communities. In the remainder of this discussion, I attempt to integrate these key results to propose future directions for work that will help us understand the consequences of genetic variation for the structure of ecological networks and the maintenance of biodiversity.   5.1 Ecological Consequences of Genetic Variation Within Populations There are a growing number of studies demonstrating that the effects of genetic diversity within species can extend beyond the population-level to affect associated community and ecosystem patterns (Crutsinger et al. 2006; Johnson et al. 2006; Hughes et al. 2008; Moreira & Mooney 2013; Whitlock 2014). In Chapter 3, I argued that future empirical studies should quantify the effects of genetic diversity on species interactions in order to identify the processes underlying patterns at the community- and ecosystem-level. In addition to this, I would argue that we are in need of theoretical and empirical work that examines if/when genetic diversity is capable of  100 buffering food webs (multi-trophic communities) from rapid environmental change (Hughes et al. 2008).   Similar to species diversity (Tilman et al. 2006; Haddad et al. 2011), intraspecific genetic diversity could buffer food webs from environmental change through a portfolio effect that maintains population-level abundance (Jump et al. 2009). For example, genetic diversity within a key resource species could maintain diversity of higher trophic levels if different genotypes exhibit complementary or independent dynamics (i.e. changes in abundance) in response to environmental change. There is empirical evidence for this type of portfolio effect below the species level. For example, Hughes and Stachowicz (2004) found that plots with more genotypes of the seagrass Zostera marina maintained higher productivity in the face of a disturbance caused by grazing geese, presumably due to certain genotypes being less palatable to geese. Since seagrass provides critical habitat for a diversity of marine species (Duffy et al. 2015; Huang et al. 2015), intraspecific genetic diversity could be important for sustaining biodiversity in this marine ecosystem. Similarly, local adaptation and genetic variation in life-history characteristics of sockeye salmon populations in Bristol Bay, Alaska results in asynchronous population dynamics. This asynchrony enables the aggregate of the populations to sustain its productivity in fluctuating marine and freshwater environments, thereby providing a stable, annual resource-pulse for consumers (Schindler et al. 2010; Ruff et al. 2011). In addition to the stable abundances produced by a portfolio effect, my results from Chapter 3 suggest that genetic diversity in resource quality could increase food-web complexity. In theory, complexity will enhance food-web resilience because predators will have more potential prey to optimally reallocate their foraging efforts after a disturbance (Kondoh 2003). Future work could test the  101 plausibility of this genetic diversity-complexity mechanism by: (1) testing whether plots with higher host-plant genetic diversity support more complex food webs; and (2) determining whether food webs supported by diverse host-plant polycultures are buffered from an environmental disturbance, such as elevated temperature. Although I have discussed this genetic diversity-complexity prediction in terms of resource diversity, genetic diversity in consumers could have a similar effect (Moya-Laraño 2012).  While prior studies have focused on genotypic richness, there are a large number of ways that populations can vary genetically. It is important that future work stay grounded in understanding how the four evolutionary processes (natural selection, genetic drift, gene flow, and mutations) affect genetic diversity within populations (Hughes et al. 2008). In a sense, we already know how these evolutionary processes will affect genetic diversity. Gene flow and mutations should increase genetic diversity within a population whereas genetic drift will decrease genetic diversity. Therefore, we might expect that processes such as gene flow and mutations in foundation or keystone species may increase food-web complexity, whereas genetic drift will decrease food-web complexity. These predictions are ignoring, however, the fact that gene flow and mutations may result in maladaptive phenotypes that cause decreases in population abundances, which, if present in key species, could decrease food-web complexity. The effects of natural selection on genetic diversity depends on the form of selection (directional, stabilizing, or divergent), but even stabilizing selection, although decreasing genetic variation, may have the indirect effect of selecting for more productive genotypes, which could favor food-web complexity. These nuances are important because they will give insight into our expectations for how evolutionary processes should affect ecological dynamics in complex communities. We are  102 in much need of experimental work that mimics these evolutionary processes to test these predictions and further develop theory.   It is also important to consider how genetic and phenotypic diversity in multiple species may affect food-web dynamics. One of the problems with studying the effects of genetic diversity beyond a single species is that the inherent increase in complexity can make the issue intractable. For mathematical models, this increase in complexity limits analytical assessments of models, which results in only a limited fraction of parameter space that can be examined through simulations. One way to constrain the inherent increase in model complexity is to empirically identify constraints on the parameter space of these models. This approach has been successfully used in dynamical models of ecological networks. For example, Jordi Bascompte and his colleagues have used this approach to identify how empirically observed network architectures (e.g. modularity and nestedness) influence the dynamics of diverse ecological communities (Stouffer & Bascompte 2011; Rohr et al. 2014). A similar approach could be applied to studying the food-web effects of genetic diversity within multiple species by quantifying how genetic diversity is distributed among species within food webs. Interestingly, such data are available (Leinonen et al. 2013); however, they have yet to be synthesized and incorporated into models of food-web dynamics. Given the complexity of the problem, developing theory through this approach could help guide appropriate experimental tests of realistic parameter space.  5.2 Ecological Consequences of Genetic Variation Between Populations My thesis focused on how within-population genetic variation can affect associated species interactions; however, between-population genetic variation could be a key factor in shaping  103 species interactions in species rich communities on large geographic scales. Indeed, there is a surge of interest in understanding how intraspecific trait variation shapes species-interaction networks change across broad geographic scales (Poisot et al. 2012, 2015). For example, there is mounting evidence that local adaptation in host plants (Crutsinger et al. 2014; Rudman et al. 2015) and predatory fish (Post et al. 2008; Bassar et al. 2010; Rudman et al. 2015) can affect patterns at the community- and ecosystem-level. It would be useful for future studies to explicitly quantify how species interactions change within these diverse aquatic and terrestrial communities. Indeed, changes in species-interaction networks due to local adaptation may explain why there are coevolutionary “hot spots” and “cold spots” (Gomulkiewicz et al. 2000;  Thompson 2005). Future empirical work should examine how local adaptation structures communities through reciprocal transplant experiments with multiple, interacting species (Nuismer & Gandon 2008).  5.3 Eco-Evolutionary Feedbacks in Complex Food Webs The results from my third chapter suggest the intraspecific genetic diversity can increase food-web complexity; however, it is interesting to consider how food-web complexity may feedback to affect genetic diversity. Food-web complexity is determined by the diversity (richness and evenness) of trophic interactions (Bersier et al. 2002; Banaaek-Richter et al. 2009). Surprisingly, there appears to be little theoretical and empirical work for how greater diversity of trophic interactions may affect the strength and direction of natural selection. One hypothesis is that interaction diversity may actually reduce the strength of natural selection, thereby preserving genetic and phenotypic diversity within a population. For example, if interacting species exert differential selection pressures on a focal species, then increasing diversity of trophic interactions  104 could end up muting the net directional effects of selection on the focal species. Therefore, food-web complexity could create heterogeneity in the biotic environment, which is thought to be an important mechanism in the maintenance of genetic variation (Levene 1953; Züst et al. 2012; Kerwin et al. 2015). Empiricists could test this hypothesis using experimental evolution with a focal species that is embedded in communities of varying complexity, although it would be necessary to first confirm that members of the community impose different selection pressures on the focal species.   5.4 Conclusion Evolutionary biologists have often focused on how heritable trait variation results in local adaptation and the origin of species. On the other hand, ecologists have sought to understand how interactions between species affect the structure and dynamics of ecological communities. My dissertation research has taken an important step toward bridging the fields of evolutionary biology and community ecology by empirically identifying how heritable variation in a suite of plant traits can affect species-interaction networks, even in the face of natural variability in the biotic and abiotic environment. My results have intriguing implications for how evolutionary processes will affect the structure of species rich food webs, although we are in need of much empirical work that explicitly examines the dynamic feedback between ecology and evolution that likely occurs in diverse species assemblages. In addition, the results of my dissertation provide a precedent for future theoretical studies of the eco-evolutionary dynamics of food webs. Going forward, it will be important that both theoreticians and empiricists collaborate to study the factors that will enhance the adaptive capacity of complex communities under future environmental change.   105 References Abràmoff, M.D., Magalhães, P.J. & Ram, S.J. (2004) Image Processing with ImageJ. Biophotonics International, 11, 36–42. Abdala‐Roberts, L., Agrawal, A. & Mooney, K. (2012) Ant–aphid interactions on Asclepias syriaca are mediated by plant genotype and caterpillar damage. Oikos, 1905–1913. Abdala‐Roberts, L. & Mooney, K. (2013) Environmental and plant genetic effects on tri‐trophic interactions. Oikos, 122, 1157–1166. Abrahamson, W.G. & Weis, A.E. (1997) Evolutionary Ecology across Three Trophic Levels: Goldenrods, Gallmakers, and Natural Enemies. Princeton University Press, Princeton. Agrawal, A.A. (2001) Phenotypic plasticity in the interactions and evolution of species. Science, 294, 321-326. Agrawal, A.A. & Zandt, P.A. (2003) Ecological play in the coevolutionary theatre: genetic and environmental determinants of attack by a specialist weevil on milkweed. Journal of Ecology, 91, 1049–1059. Agrawal, A.A. (2004) Plant defense and density dependence in the population growth of herbivores. The American Naturalist, 164, 113-120.  Agrawal, A.A. (2005) Natural selection on common milkweed (Asclepias syriaca) by a community of specialized insect herbivores. Evolutionary Ecology Research, 7, 651–667. Agrawal, A.A. & Fishbein, M. (2006) Plant defense syndromes. Ecology, 87, S132–S149. Agrawal, A.A., Hastings, A.P., Johnson, M.T.J., Maron, J.L. & Salminen, J.-P. (2012) Insect herbivores drive real-time ecological and evolutionary change in plant populations. Science, 338, 113–116.  106 Agrawal, A.A., Johnson, M.T.J., Hastings, A.P.,  Maron, J.L. (2013) A field-experiment demonstrating plant life-history evolution and its eco-evolutionary feedback to seed predator populations. The American Naturalist, 181, S35-S45. Argus, G.W. (2013) Salix, in Jepson Flora Project (eds.). URL http://ucjeps.berkeley.edu/cgi-bin/get_IJM.pl?tid=42833 [accessed 2 November 2013] Antonovics, J. (1992) Toward community genetics. Plant resistance to herbivores and pathogens: ecology, evolution, and genetics (eds. Fritz, R. & Simms, E.), pp. 426–449, University of Chicago Press, Chicago. Bailey, J.K., Bangert, R.K., Schweitzer, J.A., Trotter III, R.T., Shuster, S.M. & Whitham, T.G. (2004) Fractal geometry is heritable in trees. Evolution, 58, 2100–2102. Bailey, J.K., Wooley, S.C., Lindroth, R.L. & Whitham, T.G. (2006) Importance of species interactions to community heritability: a genetic basis to trophic‐level interactions. Ecology Letters, 9, 78–85. Bailey, J.K., Schweitzer, J.A., Úbeda, F., Koricheva, J., LeRoy, C.J., Madritch, M.D., Rehill, B.J., Bangert, R.K., Fischer, D.G., Allan, G.J. & Whitham, T.G. (2009) From genes to ecosystems: a synthesis of the effects of plant genetic factors across levels of organization. Philosophical Transactions of the Royal Society of London B: Biological Sciences, 364, 1607–1616. Banaaek-Richter, C., Bersier, L.-F., Cattin, M.-F., Baltensperger, R., Gabriel, J.-P., Merz, Y., Ulanowicz, R., Tavares, A.F., Williams, D.D., Ruiter, P.C. de, Winemiller, K.O. & Naisbit, R.E. (2009) Complexity in quantitative food webs. Ecology, 90, 1470–1477.  107 Barbour, M.A., Rodriguez-Cabal, M.A., Wu, E.T., Julkunen-Tiitto, R., Ritland, C.E., Miscampbell, A.E., Jules, E.S. & Crutsinger, G.M. (2015) Multiple plant traits shape the genetic basis of herbivore community assembly. Functional Ecology, 29, 995–1006. Barbour, M., Fortuna, M., Bascompte, J., Nicholson, J., Julkunen-Tiitto, R., Jules, E. & Crutsinger, G. (2016) Genetic specificity of a plant–insect food web: Implications for linking genetic variation to network complexity. Proceedings of the National Academy of Sciences of the USA, 2128–2133. Barbehenn, R. V. & Constabel, C.P. (2011) Tannins in plant-herbivore interactions. Phytochemistry, 72, 1551–1565. Barker, J.H.A., Pahlich, A., Trybush, S., Edwards, K.J. & Karp, A. (2003) Microsatellite markers for diverse Salix species. Molecular Ecology Notes, 3, 4–6. Bascompte, J. & Jordano. P. (2014) Mutualistic networks. Princeton University Press, Princeton. Bersier, L.-F., Banašek-Richter, C. & Cattin, M.-F. (2002) Quantitative descriptors of food-web matrices. Ecology, 83, 2394–2407. Björkman, C., Dalin, P. & Ahrné, K. (2008) Leaf trichome responses to herbivory in willows: induction, relaxation and costs. The New Phytologist, 179, 176–84. Blanchet, F.G., Legendre, P. & Borcard, D. (2008) Forward selection of explanatory variables. Ecology, 89, 2623–2632. Boeckler, G.A., Gershenzon, J. & Unsicker, S.B. (2011) Phenolic glycosides of the Salicaceae and their role as anti-herbivore defenses. Phytochemistry, 72, 1497–1509. Bolker BM (2008) Ecological Models and Data in R. Princeton University Press, Princeton, New Jersey.  108 Bolker, B., Brooks, M., Clark, C., Geange, S., Poulsen, J., Stevens, H. & White, J.-S. (2009) Generalized linear mixed models: a practical guide for ecology and evolution. Trends in Ecology & Evolution, 24, 127–135. Bolnick, D.I., Amarasekare, P., Araújo, M.S., Bürger, R., Levine, J.M., Novak, M., Rudolf, V.H.W., Schreiber, S.J., Urban, M.C. & Vasseur, D.A. (2011) Why intraspecific trait variation matters in community ecology. Trends in Ecology and Evolution, 26, 183–192. Brunsfeld, S.J., Soltis, D.E. & Soltis, P.S. (1991) Patterns of genetic variation in Salix section Longifoliae (Salicaceae). American Journal of Botany, 78, 855–869. Bukovinszky, T., Veen, F.J.F. van, Jongema, Y. & Dicke, M. (2008) Direct and indirect effects of resource quality on food web structure. Science, 319, 804–807. Caporaso, G., Kuczynski, J., Stombaugh, J., Bittinger, K., Bushman, F., Costello, E., Fierer, N., Peña, A., Goodrich, J., Gordon, J., Huttley, G., Kelley, S., Knights, D., Koenig, J., Ley, R., Lozupone, C., McDonald, D., Muegge, B., Pirrung, M., Reeder, J., Sevinsky, J., Turnbaugh, P., Walters, W., Widmann, J., Yatsunenko, T., Zaneveld, J. & Knight, R. (2010) QIIME allows analysis of high-throughput community sequencing data. Nature Methods, 7, 335–336. Caporaso, G., Lauber, C., Walters, W., Berg-Lyons, D., Huntley, J., Fierer, N., Owens, S., Betley, J., Fraser, L., Bauer, M., Gormley, N., Gilbert, J., Smith, G. & Knight, R. (2012) Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. The ISME Journal, 6, 1621–1624. Carmona, D., Lajeunesse, M.J. & Johnson, M.T.J. (2011) Plant traits that predict resistance to herbivores. Functional Ecology, 25, 358–367. Carroll, S.P., Jørgensen, P.S., Kinnison, M.T., Bergstrom, C.T., Denison, R.F., Gluckman, P., Smith, T.B., Strauss, S.Y. & Tabashnik, B.E. (2014) Applying evolutionary biology to address  109 global challenges. Science, 346, 1245993. Castagneyrol, B., Lagache, L., Giffard, B., Kremer, A. & Jactel, H. (2012) Genetic diversity increases insect herbivory on oak saplings. PloS ONE, 7, e44247. Chase, J.M. & Knight, T.M. (2003) Community genetics: toward a synthesis. Ecology, 84, 580–582. Clark, J.S. (2010) Individuals and the variation needed for high species diversity in forest trees. Science, 327, 1129–1132. Colwell, R.K. & Coddington, J. (1994) Estimating terrestrial biodiversity through extrapolation. Philosophical Transactions of the Royal Society of London B: Biological Sciences, 345, 101–118. Cornelissen, J.H.C., Lavorel, S., Garnier, E., Díaz, S., Buchmann, N., Gurvich, D.E., Reich, P.B., Steege, H. ter, Morgan, H.D., Van Der Heijden, M.G.A., Pausas, J.G. & Poorter, H. (2003) A handbook of protocols for standardised and easy measurement of plant functional traits worldwide. Australian Journal of Botany, 51, 335–380. Craig, T.P., Itami, J.K. & Price, P.W. (1990) The window of vulnerability of a shoot-galling sawfly to attack by a parasitoid. Ecology, 71, 1471–1482. Crutsinger, G.M. & Sanders, N.J. (2005) Aphid-tending ants affect secondary users in leaf shelters and rates of herbivory on Salix hookeriana in a coastal dune habitat. American Midland Naturalist, 154, 296–304. Crutsinger, G.M., Collins, M.D., Fordyce, J.A., Gompert, Z., Nice, C.C. & Sanders, N.J. (2006) Plant genotypic diversity predicts community structure and governs an ecosystem process. Science, 313, 966–968. Crutsinger, G.M., Reynolds, W., Classen, A.T. & Sanders, N.J. (2008) Disparate effects of plant  110 genotypic diversity on foliage and litter arthropod communities. Oecologia, 158, 65-75. Crutsinger, G.M., Rodriguez-Cabal, M.A., Roddy, A.B., Peay, K.B., Bastow, J.L., Kidder, A.G., Dawson, T.E., Fine, P.V., Rudgers, J.A. (2014) Genetic variation within a dominant shrub structures green and brown community assemblages. Ecology, 95, 387-98.  Crutsinger, G., Rudman, S., Rodriguez‐ Cabal, M., McKown, A., Sato, T., MacDonald, A., Heavyside, J., Geraldes, A., Hart, E., LeRoy, C. & El‐ Sabaawi, R. (2014) Testing a ‘genes‐ to‐ecosystems’ approach to understanding aquatic–terrestrial linkages. Molecular Ecology, 23, 5888–5903. Crutsinger, G. (2016) A community genetics perspective: opportunities for the coming decade. New Phytologist, 210, 65–70. Dormann, C.F., Elith, J., Bacher, S., Buchmann, C., Carl, G., Carré, G., Marquéz, J.R.G., Gruber, B., Lafourcade, B., Leitão, P.J., Münkemüller, T., McClean, C., Osborne, P.E., Reineking, B., Schröder, B., Skidmore, A.K., Zurell, D. & Lautenbach, S. (2013) Collinearity: a review of methods to deal with it and a simulation study evaluating their performance. Ecography, 36, 027–046. Duffy, E., Reynolds, P., Boström, C., Coyer, J., Cusson, M., Donadi, S., Douglass, J., Eklöf, J., Engelen, A., Eriksson, B., Fredriksen, S., Gamfeldt, L., Gustafsson, C., Hoarau, G., Hori, M., Hovel, K., Iken, K., Lefcheck, J., Moksnes, P., Nakaoka, M., O’Connor, M., Olsen, J., Richardson, P., Ruesink, J., Sotka, E., Thormar, J., Whalen, M. & Stachowicz, J. (2015) Biodiversity mediates top–down control in eelgrass ecosystems: a global comparative‐experimental approach. Ecology Letters, 18, 696-705. Dunne, J., Williams, R. & Martinez, N. (2002) Network structure and biodiversity loss in food webs: robustness increases with connectance. Ecology Letters, 5, 558-567.  111 Dunne, J. (2006) The network structure of food webs. Ecological networks: Linking structure to dynamics in food webs (eds. Dunne, J. & Pascual, M.), pp. 27-86, Oxford University Press, New York. Edgar, R. (2013) UPARSE: highly accurate OTU sequences from microbial amplicon reads. Nature Methods, 10, 996–998. Endler, J.A. (1986) Natural Selection in the Wild. Princeton University Press, Princeton. Erlandson, S., Savage, J., Cavender-Bares, J. & Peay, K. (2015) Soil moisture and chemistry influence diversity of ectomycorrhizal fungal communities associating with willow along an hydrologic gradient. FEMS microbiology ecology, 92, fiv148. Floate, K. & Whitham, T. (1994) Aphid-ant interaction reduces chrysomelid herbivory in a cottonwood hybrid zone. Oecologia, 97, 215–221. Fordyce, J.A. (2006) The evolutionary consequences of ecological interactions mediated through phenotypic plasticity. The Journal of Experimental Biology, 209, 2377-2383. Fritz, R.S. & Price, P.W. (1988) Genetic variation among plants and insect community structure: willows and sawflies. Ecology, 69, 845–856. Fritz, R.S. (1990) Variable competition between insect herbivores on genetically variable host plants. Ecology, 71, 1208–1211. Fritz, R.S. & Simms, (1992) Plant resistance: ecology, evolution and genetics. University of Chicago Press, Chicago.  Fritz, R.S. (1995) Direct and indirect effects of plant genetic variation on enemy impact. Ecol Entomol, 20, 18–26. Fussmann, G., Loreau, M. & Abrams, P. (2007) Eco-evolutionary dynamics of communities and ecosystems. Functional Ecology, 21, 465–477.  112 Garbutt, K. & Bazzaz, FA. (1987) Population niche structure: Differential response of Abutilon theophrasti progeny to resource gradients. Oecologia, 72, 291–296. Garibaldi, L. A., Kitzberger, T. & Chaneton, E.J. (2011) Environmental and genetic control of insect abundance and herbivory along a forest elevational gradient. Oecologia, 167, 117–29. Gaston. (2000) Global patterns in biodiversity. Nature, 405, 220–227. Geber, M.A. & Griffen, L.R. (2003) Inheritance and natural selection on functional traits. International Journal of Plant Sciences, 164, S21–S42. Gomulkiewicz, R., Thompson, J., Holt, R., Nuismer, S. & Hochberg, M. (2000) Hot Spots, Cold Spots, and the Geographic Mosaic Theory of Coevolution. The American Naturalist, 156, 156–174. Gotelli, N.J. & Colwell, R.K. (2001) Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecology Letters, 4, 379–391. Gotelli NJ, Colwell RK (2010) Estimating species richness. Oxford University Press, Oxford. Gratani, L. (2014) Plant Phenotypic Plasticity in Response to Environmental Factors. Advances in Botany, 2014, 1–17. Haddad, N., Crutsinger, G., Gross, K., Haarstad, J. & Tilman, D. (2011) Plant diversity and the stability of foodwebs. Ecology Letters, 14, 42–46. Harmon, L.J., Matthews, B., Roches, S. Des, Chase, J.M., Shurin, J.B. & Schluter, D. (2009) Evolutionary diversification in stickleback affects ecosystem functioning. Nature, 458, 1167–1170. Henri, D.C. & Veen, F.J.F. van. (2011) Body size, life history and the structure of host–parasitoid networks. Advances in Ecological Research, 45, 135–180.  113 Hersch-Green, E.I., Turley, N.E. & Johnson, M.T.J. (2011) Community genetics: what have we accomplished and where should we be going? Philosophical Transactions of the Royal Socieity B: Biological Sciences, 366, 1453–1460. Hochwender, C.G. & Fritz, R.S. (2004) Plant genetic differences influence herbivore community structure: evidence from a hybrid willow system. Oecologia, 138, 547–557. Huang, A., Essak, M. & O’Connor, M. (2015) Top–down control by great blue herons Ardea herodias regulates seagrass‐ associated epifauna. Oikos, 124, 1492–1501. Hughes, A.R. & Stachowicz, J.J. (2004) Genetic diversity enhances the resistance of a seagrass ecosystem to disturbance. Proceedings of the National Academy of Sciences of the USA, 101, 8998–9002. Hughes, A.R., Inouye, B.D., Johnson, M.T.J., Underwood, N. & Vellend, M. (2008) Ecological consequences of genetic diversity. Ecology Letters, 11, 609–623. Hughes, J.B., Daily, G.C. & Ehrlich, P.R. (1997) Population diversity: its extent and extinction. Science, 278, 689–692. Ings, T.C., Montoya, J.M., Bascompte, J., Blüthgen, N., Brown, L., Dormann, C.F., Edwards, F., Figueroa, D., Jacob, U., Jones, J.I., Lauridsen, R.B., Ledger, M.E., Lewis, H.M., Olesen, J.M., Veen, F.J.F. van, Warren, P.H. & Woodward, G. (2008) Ecological networks - beyond food webs. Journal of Animal Ecology, 78, 253-269. Johnson, M.T.J. & Agrawal, A.A. (2005) Plant genotype and environment interact to shape a diverse arthropod community on evening primrose (Oenothera biennis). Ecology, 86, 874–885. Johnson, M.T.J., Lajeunesse, M.J. & Agrawal, A.A. (2006) Additive and interactive effects of plant genotypic diversity on arthropod communities and plant fitness. Ecology Letters, 9, 24–34.  114 Johnson, M.T.J. & Agrawal, A. (2007) Covariation and composition of arthropod species across plant genotypes of evening primrose (Oenothera biennis). Oikos, 116, 941–956. Johnson, M. & Stinchcombe, J. (2007) An emerging synthesis between community ecology and evolutionary biology. Trends in Ecology & Evolution, 22, 250–257. Johnson, M.T. (2008) Bottom-up effects of plant genotype on aphids, ants, and predators. Ecology, 89, 145–154. Johnson, M.T.J., Agrawal, A.A., Maron, J.L. & Salminen, J.-P. (2009) Heritability, covariation and natural selection on 24 traits of common evening primrose (Oenothera biennis) from a field experiment. Journal of Evolutionary Biology, 22, 1295–1307. Josse, J., Chavent, M., Liquet, B. & Husson, F. (2012) Handling missing values with regularized iterative multiple correspondence analysis. Journal of Classification, 29, 91–116. Jump, A.S., Marchant, R. & Peñuelas, J. (2009) Environmental change and the option value of genetic diversity. Trends in Plant Science, 14, 51–58. Julkunen-Tiitto, R. & Sorsa, S. (2001) Testing the effects of drying methods on willow flavonoids, tannins, and salicylates. Journal of Chemical Ecology, 27, 779–789. Kareiva, P. & Sahakian, R. (1990) Tritrophic effects of a simple architectural mutation in pea plants. Nature, 345, 433–434. Kerwin, R., Feusier, J., Corwin, J., Rubin, M., Lin, C., Muok, A., Larson, B., Li, B., Joseph, B., Francisco, M., Copeland, D., Weinig, C. & Kliebenstein, D. (2015) Natural genetic variation in Arabidopsis thaliana defense metabolism genes modulates field fitness. eLife, 4, e05604. Kondoh,  M. (2003) Foraging adaptation and the relationship between food-web complexity and stability. Science, 299, 1388–1391.  115 Lawton, J.H. (1983) Plant architecture and the diversity of phytophagous insects. Annual Review of Entomology, 28, 23–39. Lamit, L., Busby, P., Lau, M., Compson, Z., Wojtowicz, T., Keith, A., Zinkgraf, M., Schweitzer, J., Shuster, S. & Gehring, C. (2015) Tree genotype mediates covariance among communities from microbes to lichens and arthropods. Journal of Ecology, 103, 840–850. Lefcheck, J. (2015) piecewiseSEM: Piecewise structural equation modelling in r for ecology, evolution, and systematics. Methods in Ecology and Evolution. Legendre, P. & Gallagher, E.D. (2001) Ecologically meaningful transformations for ordination of species data. Oecologia, 129, 271–280. Leimu, R. & Koricheva, J. (2006) A meta-analysis of genetic correlations between plant resistances to multiple enemies. The American Naturalist, 168, E15–E37. Leinonen, T., McCairns, S.R., O’Hara, R.B. & Merilä, J. (2013) QST-FST comparisons: evolutionary and ecological insights from genomic heterogeneity. Nature Reviews Genetics, 14, 179–190. Levene, H. (1953) Genetic equilibrium when more than one ecological niche is available. The American Naturalist, 87, 331-333. Lewinsohn, T.M., Novotny, V., Basset, Y. (2005) Insects on plants: diversity of herbivore assemblages revisited. Annual Review of Ecology, Evolution, and Systematics, 36, 597-620. Lortie, C. & Cushman, H. (2007) Effects of a directional abiotic gradient on plant community dynamics and invasion in a coastal dune system. Journal of Ecology, 95, 468–481. Luck, G.W., Daily, G.C. & Ehrlich, P.R. (2003) Population diversity and ecosystem services. Trends in Ecology & Evolution, 18, 331–336.  116 Lynch, M. & Walsh, B. (1998) Genetics and Analysis of Quantitative Traits. Sinauer Associates, Sunderland. MacArthur, R. (1955) Fluctuations of animal populations and a measure of community stability. Ecology, 36, 533–536. MacArthur, R. (1972) Geographical ecology: Patterns in the distribution of species. Princeton University Press, Princeton.  Maddox, G.D. & Root, R.B. (1987) Resistance to 16 diverse species of herbivorous insects within a population of goldenrod, Solidago altissima: genetic variation and heritability. Oecologia, 72, 8–14. Maddox, G.D. & Root, R.B. (1990) Structure of the encounter between goldenrod (Solidago altissima) and its diverse insect fauna. Ecology, 71, 2115–2124. Matsuki, M. & MacLean Jr., S.F. (1994) Effects of different leaf traits on growth rates of insect herbivores on willows. Oecologia, 100, 141–152. McArt, Cook-Patton & Thaler. (2012) Relationships between arthropod richness, evenness, and diversity are altered by complementarity among plant genotypes. McCann. (2000) The diversity-stability debate. Nature, 228–33. McCann, K. (2012) Food webs. Princeton University Press, Princeton. Melián, C.J., Vilas, C., Baldó, F., González-Ortegón, E., Drake, P. & Williams, R.J. (2011) Eco-evolutionary dynamics of individual-based food webs. Advances in Ecological Research, 45, 225–268. Miller, W. & Weis, A. (1999) Adaptation of Coyote Brush to the Abiotic Environment and Its Effects on Susceptibility to a Gall-Making Midge. Oikos, 84, 199-208.   117 Mooney, K. & Agrawal, A. (2008) Plant genotype shapes ant-aphid interactions: implications for community structure and indirect plant defense. The American Naturalist, 171, E195–205. Moreira, X. & Mooney, K.A. (2013) Influence of plant genetic diversity on interactions between higher trophic levels. Biology Letters, 9, 20130133. Morse, D.R., Lawton, J.H., Dodson, M.M. & Williamson, M.H. (1985) Fractal dimension of vegetation and the distribution of arthropod body lengths. Nature, 314, 731–733. Moya-Laraño, J. (2011) Genetic variation, predator-prey interactions and food web structure. Philosphical Transactions of the Royal Society B: Biological Sciences, 366, 1425–1437. Moya-Laraño, J., Verdeny-Vilalta, O., Rowntree, J., Melguizo-Ruiz, N., Montserrat, M. & Laiolo, P. (2012) Climate change and eco-evolutionary dynamics in food webs. Advances in Ecological Research, 47, 1–80. Munns, R. & PrometheusWiki Contributors (2010) Plant water content and relative water content. URL http://www.publish.csiro.au/prometheuswiki/tiki-pagehistory.php?page=Plant water content and relative water content&preview=6 [accessed 6 November 2013] Murdoch, W.W., Briggs, C.J., Nisbet, R.M. (2003) Consumer-resource dynamics. Princeton University Press, Princeton. Nakagawa, S. & Schielzeth, H. (2013) A general and simple method for obtaining R2 from generalized linear mixed‐effects models. Methods in Ecology and Evolution, 4, 133–142. Nichols-Orians, C.M., Fritz, R.S. & Clausen, T.P. (1993) The genetic basis for variation in the concentration of phenolic glycosides in Salix sericea: clonal variation and sex-based differences. Biochemical Systematics and Ecology, 21, 535–542. Nuismer, S.L. & Gandon, S. (2008) Moving beyond common-garden and transplant designs: insight into the causes of local adaptation in species interactions. The American Naturalist, 171,  118 658–668. Nybakken, L. & Julkunen-Tiitto, R. (2013) Gender differences in Salix myrsinifolia at the pre-reproductive stage are little affected by simulated climatic change. Physiologia plantarum, 147, 465–476. O’Hara, R.B. & Kotze, D.J. (2010) Do not log-transform count data. Methods in Ecology and Evolution, 1, 118–122. Paine, R. (1966) Food Web Complexity and Species Diversity. The American Naturalist, 100, 65–75. Paine, R. (1969) A Note on Trophic Complexity and Community Stability. The American Naturalist, 103, 91–93. Palo, R.T. (1984) Distribution of birch (Betula spp.), willow (Salix spp.), and poplar (Populus spp.) secondary metabolites and their potential role as chemical defense against herbivores. Journal of Chemical Ecology, 10, 499–520. Poisot, T., Canard, E., Mouillot, D., Mouquet, N. & Gravel, D. (2012) The dissimilarity of species interaction networks. Ecology Letters, 15, 1353–1361. Poisot, T., Stouffer, D.B. & Gravel, D. (2015) Beyond species: why ecological interaction networks vary through space and time. Oikos, 124, 243–251. Post, D.M., Palkovacs, E.P., Schielke, E.G. & Dodson, S.I. (2008) Intraspecific variation in a predator affects community structure and cascading trophic interactions. Ecology, 89, 2019–2032. Price, P.W. (1980) Evolutionary Biology of Parasites. Princeton University Press, Princeton. Price, P.W. (1991) The plant vigor hypothesis and herbivore attack. Oikos, 62, 244–251. R Core Team (2013) R: A language and environment for statistical computing.  119 Raghu, S., Drew, R. A. I. & Clarke, A.R. (2004) Influence of host plant structure and microclimate on the abundance and behavior of a tephritid fly. Journal of Insect Behavior, 17, 179–190. Reddy, T.B.K., Thomas, A., Stamatis, D., Bertsch, J., Isbandi, M., Jansson, J., Mallajosyula, J., Pagani, I., Lobos, E. & Kyrpides, N. (2015) The Genomes OnLine Database (GOLD) v.5: a metadata management system based on a four level (meta)genome project classification. Nucleic Acids Research, 43, D1099–D1106. Roche, B.M. & Fritz, R.S. (1997) Genetics of resistance of Salix sericea to a diverse community of herbivores. Evolution, 51, 1490–1498. Roininen, H. & Tahvanainen, J. (1989) Host selection and larval performance of two willow-feeding sawflies. Ecology, 70, 129–136. Rohr, R.P., Saavedra, S. & Bascompte, J. (2014) On the structural stability of mutualistic systems. Science, 345, 1253497. Rudman, S., Rodriguez-Cabal, M., Stier, A., Sato, T., Heavyside, J., El-Sabaawi, R. & Crutsinger, G. (2015) Adaptive genetic variation mediates bottom-up and top-down control in an aquatic ecosystem. Proceedings of the Royal Society B: Biological Sciences, 282, 20151234. Ruff, C.P., Schindler, D.E., Armstrong, J.B., Bentley, K.T., Brooks, G.T., Holtgrieve, G.W., McGlauflin, M.T., Torgersen, C.E. & Seeb, J.E. (2011) Temperature-associated population diversity in salmon confers benefits to mobile consumers. Ecology, 92, 2073–2084.  Sacchi, C.F. & Price, P.W. (1988) Pollination of the arroyo willow, Salix lasiolepis: role of insects and wind. American Journal of Botany, 75, 1387–1393. Scheiner, S. (1993) Genetics and Evolution of Phenotypic Plasticity. Annual Review of Ecology,  120 Evolution and Systematics, 24, 35–68. Scheiner,  SM & Goodnight,  CJ. (1984) The comparison of phenotypic plasticity and genetic variation in populations of the grass Danthonia spicata. Evolution, 38, 845-855. Schindler, D.E., Hilborn, R., Chasco, B., Boatright, C.P., Quinn, T.P., Rogers, L.A. & Webster, M.S. (2010) Population diversity and the portfolio effect in an exploited species. Nature, 465, 609–612. Schoener. (2011) The Newest Synthesis: Understanding the Interplay of Evolutionary and Ecological Dynamics. Science, 331, 426–429. Shipley B (2000) A New Inferential Test for Path Models Based on Directed Acyclic Graphs. Structural Equation Modeling: A Multidisciplinary Journal, 7, 206–218. Silfver, T., Rousi, M., Oksanen, E. & Roininen, H. (2014) Genetic and environmental determinants of insect herbivore community structure in a Betula pendula population. F1000Research, 1–13. Sipura, M. (1999) Tritrophic interactions: willows, herbivorous insects and insectivorous birds. Oecologia, 121, 537–545. Stouffer, D.B. & Bascompte, J. (2011) Compartmentalization increases food-web persistence. Proceedings of the National Academy of Sciences of the USA, 108, 3648–3652. Strong, D.R., Lawton, J.H. & Southwood, S.R. (1984) Insects on Plants: Community Patterns and Mechanisms. Harvard University Press, Cambridge. Tack, A., Johnson, M. & Roslin, T. (2012) Sizing up community genetics: it’s a matter of scale. Oikos, 121, 481–488. Tahvanainen, J., Julkunen-Tiitto, R. & Kettunen, J. (1985) Phenolic glycosides govern the food selection pattern of willow feeding leaf beetles. Oecologia, 67, 52–56.  121 Thompson, J.N. (2005) The Geographic Mosaic of Coevolution. University of Chicago Press, Chicago. Tilman, D., Reich, P.B. & Knops, J.M.H. (2006) Biodiversity and ecosystem stability in a decade-long grassland experiment. Nature, 441, 629–632. Tuomisto H (2012) An updated consumer’s guide to evenness and related indices. Oikos, 121, 1203–1218. Villéger S, Mason N, Mouillot D (2008) New multidimensional functional diversity indices for a multifaceted framework in functional ecology. Ecology, 89, 2290–2301. Violle, C., Enquist, B.J., McGill, B.J., Jiang, L., Albert, C.H., Hulshof, C., Jung, V. & Messier, J. (2012) The return of the variance: intraspecific variability in community ecology. Trends in Ecology & Evolution, 27, 244–252. Walsh, M., DeLong, J., Hanley, T. & Post, D. (2012) A cascade of evolutionary change alters consumer-resource dynamics and ecosystem function. Proceedings of the Royal Society of London B: Biological Sciences, 279, 3184–3192.  Wardle, D.A., Bardgett, R.D., Klironomos, J.N., Setälä, H., Putten, W.H. van der & Wall, D.H. (2004) Ecological linkages between aboveground and belowground biota. Science, 304, 1629–1633. Whitham, T.G., Bailey, J.K. & Jennifer, J.A. (2006) A framework for community and ecosystem genetics: from genes to ecosystems. Nature Genetics, 7, 510-523. Whitham, T.G., Gehring, C.A., Lamit, L.J., Wojtowicz, T., Evans, L.M., Keith, A.R. & Smith, D.S. (2012) Community specificity: life and afterlife effects of genes. Trends in Plant Science, 17, 271–281.  122 Whitlock. (2014) Relationships between adaptive and neutral genetic diversity and ecological structure and functioning: a meta‐ analysis. Journal of Ecology, 102, 857-872. Woodward, G., Blanchard, J., Lauridsen, R.B., Edwards, F.K., Jones, J.I., Figueroa, D., Warren, P.H. & Petchey, O.L. (2010) Individual-based food webs: species identity, body size and sampling effects. Advances in Ecological Research, 43, 211–266. Yoshida, T., Jones, L.E., Ellner, S.P., Fussmann, G.F. & Hairston Jr, N.G. (2003) Rapid evolution drives ecological dynamics in a predator–prey system. Nature, 424, 303–306. Züst, T., Heichinger, C., Grossniklaus, U., Harrington, R., Kliebenstein, D. & Turnbull, L. (2012) Natural Enemies Drive Geographic Variation in Plant Defenses. Science, 338, 116–119. Züst, T. & Agrawal, A. (2016) Mechanisms and evolution of plant resistance to aphids. Nature Plants, 2, 15206.    123 Appendices Appendix A    A.1 Microsatellite analysis  Microsatellite loci used to genotype clones of Salix hookeriana in a common garden experiment. SB80 had 1-5 alleles (allele size range 109-137 bp) and SB194 had 1-4 alleles (allele size range 106-144 bp). Genotype ID Locus: SB80 Locus: SB194 Unique? * 125 106,118,122 Noa A 125 106,118,122 No B 109,115,125,129 106,116,120,128 Yes C 109,113,129 108,112 Yes D 109,125,133,137 106,112,124 Yes E 109,133 106,128 Yes F 109,125,137 106,120,128 Yes G 115,125 106,116 Yes H 109,125 106,120,128 Yes I 133 108 Yes J 109,125 106,116 Yes K 109,125,129 108,118,122 Yes L 109,111,125,133 108,122 Yes M 109,115,125 108,112 Yes N 115,125,133 106,120,128 Yes O 109,125,131,137 108,116,122 Yes P 109,125 108 Yes Q 109,119 106,116,142 Yes R 109,125,129 108,112 Yes S 109,125 108,112,132 Yes T 109,125,133 106,116 Yes U 125,133 108 Yes V 109,125 106 Yes W 109,115,125 108,116,122 Yes X 109,115,125,133 108,116,124,144 Yes Y 109,111,125 106,132 Yes Z 109,125 108,144 Yes Notes: aGenotype dropped from analysis.   124 A.2 Non-tannin phenolic compounds  List of non-tannin phenolic compounds found within leaves of Salix hookeriana with corresponding wavelength, retention time and response factor. Phenolic Compounds Wavelength (nm) Retention time (min) Response factor Salicylates      Salicin 220 2.269 0.00175   Tremulacin 220 32.311 0.002053   HCH-tremulacin  220 37.722 0.002053   Salicortin 270 13.051 0.01692   Disalicortin 270 35.87 0.01692   Cinnamoyl salicortin 270 35.247 0.01692   Diglucoside of SaOH 220 2.105 0.0008474 Phenolic acids      Cinnamic acid der. 1 270 14.017 0.002552   Cinnamic acid der. 2 270 22.108 0.002552   Cinnamic acid der3 270 31.394 0.002552   p-OH-cinnamic acid 320 12.03 0.000284   p-OH-cinnamic acid der. 1 320 8.184 0.000748   p-OH-cinnamic acid der. 2 320 8.692 0.000748   Chlorogenic acid 320 9.806 0.000748   Neochlorogenic acid 320 5.864 0.000748   Gentisic acid 220 3.215 0.00175 Flavones      Luteolin der. 1 320 25.901 0.00112   Luteolin der. 2 320 29.075 0.00112   Luteolin glycoside der. 1 320 23.306 0.00112   Luteolin glycoside der. 2 320 24.617 0.00112   Luteolin 5-glucoside 320 20.512 0.00112   Luteolin 7-glucoside 320 21.185 0.00112   Methylluteolin glycoside  320 24.079 0.00112 Flavonols      Myricitrin 320 20.931 0.001985   Quercetin der. 1 320 24.906 0.001598   Quercetin der. 2 320 26.72 0.001598 Miscellaneous Flavonoids      Eriodictyol 7-glucoside 270 22.536 0.001925   Ampelopsin 320 10.9 0.002953   Ampelopsin der. 320 17.078 0.002953    125 A.3 Diagram of minimum convex hull  Calculated minimum convex hull using ImageJ to quantify foliage density.  minimum convex hull 126 A.4 Heatmap of phenotypic trait correlations  Heatmap of phenotypic trait correlations (Pearson's r, sample size range = 115-140) for Salix hookeriana measured in a common garden experiment. Increasingly red colors indicate strong positive correlations, whereas increasingly blue shades indicated strong negative correlations.    127 A.5 Principal component analysis of leaf phenolic compounds Results from principal component analysis (PCA) on correlation matrix of different groups of phenolic compounds identified from leaves of Salix hookeriana in a common garden experiment. Bolded values indicate that more than 10% of the particular principal component is explained by the corresponding trait. Transformations on secondary metabolites prior to PCA are given in brackets next to their respective PCA.  Salicylates/Tannins PCA [√x] PC 1  Condensed tannins     Soluble condensed tannins 0.38    Insoluble condensed tannins -0.22  Salicylates     Salicin -0.39    Salicortin -0.40    Tremulacin -0.38    HCH-tremulacin -0.37    Diglucoside of SaOH -0.28    Cinnamoyl salicortin -0.17    Disalicortin -0.32  Variance explained by PC 57.3%      Phenolic acids PCA [log(x+1)] PC 1 PC 2   Neochlorogenic acid 0.21 0.54   p-OH-cinnamic acid der. 1 0.25 0.44   p-OH-cinnamic acid der. 2 0.41 -0.06   Gentisic acid 0.14 -0.37   Chlorogenic acid -0.02 -0.41   p-OH-cinnamic acid 0.11 -0.42   Cinnamic acid der. 1 0.51 -0.05   Cinnamic acid der. 2 0.49 -0.03   Cinnamic acid der. 3 0.44 -0.16 Variance explained by PC 40.1% 24.4%          128 Flavonoids PCA [√x] PC 1 PC 2 Flavones     Luteolin 5-glucoside -0.34 0.09   Luteolin 7-glucoside -0.24 -0.52   Luteolin glycoside der. 1 -0.28 0.00   Methylluteolin glycoside -0.28 -0.56   Luteolin glycoside der. 2 0.26 0.39   Luteolin der. 1 -0.37 0.31   Luteolin der. 2 -0.41 0.29 Flavonols     Myricitrin 0.27 0.04   Quercetin der. 1 0.31 -0.22   Quercetin der. 2 0.35 -0.16 Variance explained by PC 32.2% 17.4%    Miscellaneous flavonoids PCA [√x] PC 1  Ampelopsin 0.51  Ampelopsin der. 0.60  Eriodicytol 7-glucoside 0.62  Variance explained by PC 61.5%    A.6 Ordination of herbivore community response Ordination of herbivore community response to Salix hookeriana genotype with key herbivore species labeled. The centroid of each genotype is plotted with its corresponding letter in grey. Note that this ordination is the same as Figure 1C, but with species labeled.   129         130 Appendix B   B.1 Statistical models testing the genetic specificity of the plant-insect food web. Response df F or 2 P Gall size1       Leaf gall 23,57 2.17 0.009   Bud gall 21,44 0.98 0.504   Apical-stem gall 16,12 0.29 0.988 Gall abundance2 25,119 202.40 0.001 Leaf gall  74.60 0.001 Bud gall  55.02 0.006 Apical-stem gall  44.47 0.042 Mid-stem gall  28.27 0.295 Composition of gall community3 22,89 1.96 0.001 Abundance of gall-parasitoid interactions2 25,119 357.10 0.001 Leaf gall    Platygaster sp.  79.51 0.001 Mesopolobus sp.  50.00 0.009 Torymus sp.  60.11 0.001 Tetrastichus sp.  32.96 0.105 Mymarid sp. A  6.37 0.448 Bud gall    Platygaster sp.  18.04 0.276 Mesopolobus sp.  6.37 0.497 Torymus sp.  39.81 0.079 Tetrastichus sp.  18.09 0.492 Lestodiplosis sp.  16.05 0.552 Apical-stem gall    Torymus sp.  23.13 0.048 Mid-stem gall    Platygaster sp.  6.64 0.452 Composition of gall-parasitoid interactions3 12,45 1.57 0.007 Proportion of galls parasitized4    Leaf gall 23,58 75.79 <0.001 Platygaster sp.  93.47 <0.001 Mesopolobus sp.  42.56 0.008 Torymus sp.  42.92 0.007 Tetrastichus sp.  29.55 0.163 Mymarid sp. A  3.97 0.999 Bud gall 21,46 49.84 0.072 Apical-stem gall 18,12 15.69 0.614 Composition of trophic interactions in the plant-insect food web3  22,89 1.90 0.001  131 Notes: 1GLM (error distribution = Gaussian, link function = identity), log-transformed; 2multivariate GLM (error distribution = negative binomial, link function = log); 3PERMANOVA on Bray-Curtis dissimilarities (999 permutations); 4GLM (error distribution = binomial, link function = logit). P-values in bold (P < 0.05), italics (P < 0.10), and normal font (P > 0.10) denote degree of statistical significance.     132  B.2 Statistical models explaining insect food web responses  Statistical models explaining insect food web responses to genetic variation in coastal willow (Salix hookeriana). We report the coefficients of all predictor variables that were included in the final statistical models, which were determined using AIC and likelihood-ratio tests. Response Predictors Gall size1 Salicylates/ Tannins PC1 Flavones/ Flavonols PC1   Leaf gall -0.20 -0.26   Gall abundance2 C:N Flavanones/ Flavanonols PC1 Plant size  Leaf gall 0.04 -0.03 -0.36  Bud gall 0.08 -0.07 -1.01  Apical-stem gall 0.01 0.46 0.26  Mid-stem gall 0.02 -1.81 -4.77  Abundance of gall-parasitoid interactions2 Leaf gall size Leaf gall abundance Bud gall abundance Apical-stem gall abundance Leaf gall     Platygaster sp. -0.22 1.22 0.20 -0.15 Mesopolobus sp. -0.27 0.90 -0.26 0.44 Torymus sp. 0.19 0.76 -0.30 0.72 Tetrastichus sp. -0.24 0.71 0.45 -1.09 Mymarid sp. A -1.67 20.83 -2.07 3.35 Bud gall     Platygaster sp. 0.43 0.23 5.81 -14.25 Mesopolobus sp. 0.16 0.30 0.77 1.95 Torymus sp. -0.17 0.31 1.39 -0.43 Tetrastichus sp. 0.15 0.51 1.83 0.08 Lestodiplosis sp. 0.04 -0.61 1.46 1.75 Apical-stem gall     Torymus sp. -0.12 0.05 -0.64 4.09 Mid-stem gall     Platygaster sp. 1.54 -15.03 0.53 -9.23 Notes: 1GLM (error distribution = Gaussian, link function = identity), log-transformed; 2multivariate GLM (error distribution = negative binomial, link function = log). P-values  133 in bold (P < 0.05), italics (P < 0.10), and normal font (P > 0.10) denote degree of statistical significance.   B.3 Generalized linear models of leaf gall parasitism  Generalized linear models (error distribution = binomial, link function = logit) explaining the proportion of leaf galls parasitized. Final models were determined using AIC and likelihood-ratio tests. Response Predictor df 2 P Total parasitism Gall size 1,79 22.28 <0.001 Platygaster sp. Gall size 1,77 17.58 <0.001  Gall abundance 1,77 0.73 0.394  Gall size x abundance 1,77 8.71 0.003 Mesopolobus sp. Gall size 1,77 7.28 0.007  Gall abundance 1,77 0.29 0.588  Gall size x abundance 1,77 4.21 0.040 Torymus sp. Gall size 1,78 3.83 0.050  Gall abundance 1,78 5.24 0.022  B.4 Relatedness and functional-trait diversity of willow genotypes The matrix of microsatellite markers for the 26 willow genotypes used in this study was published in Table S1 of (1); however, since the willow genotyping was only based on 2 markers, they were unable to infer the relatedness among genotypes. If certain genotypes are more closely related to each other, and consequently have very similar phenotypes, this could introduce spurious confidence in our associations between willow traits and gall abundances/phenotypes. We can examine this phenotypic similarity by measuring the functional evenness and divergence of the 26 willow genotypes in multivariate trait space  (2). To do this, we calculated the average trait value for each of the 40 traits we measured for each willow genotype. We then calculated functional evenness and functional  134 divergence using the ‘FD’ package in R. For both indices, values close to zero correspond to functional redundancy, while values close to one indicate functional distinctiveness. We found that functional evenness and divergence were equal to 0.94 and 0.87, respectively, suggesting that the phenotypes (in multivariate trait space) of each genotype are quite distinct from each other. Therefore, we argue that not knowing the relatedness among the 26 genotypes probably introduces little bias in our trait associations with the abundances and sizes of galls.  B.5 Sampling interactions in gall-parasitoid network  The total number of potential gall-parasitoid interactions in this bipartite network is 24 (i.e. each of the 4 galls could interact with each of the 6 parasitoids, 6*4 = 24). Interspecific differences among gall species (e.g. differences in gall morphology, phenology, plant part galled) and sampling effort likely constrain the number of potential interactions observed to considerably less than 24. While it was not the focus of our study to examine interspecific differences, it is important to demonstrate that we have sampled the majority of interactions in the gall-parasitoid network. To demonstrate this, we considered unique gall-parasitoid interactions as ‘species’ and used Chao 1 (3) to estimate the total number of interactions. While we documented 12 unique gall-parasitoid interactions, Chao 1 estimated the number of interactions to be 14.98 (std. error = 4.49), suggesting that we have sampled the majority of interactions in the gall-parasitoid network.   135 B.6 Calculating quantitative-weighted linkage density (food-web complexity) Quantitative-weighted linkage density, 𝐿𝐷𝑞, was calculated using the following equations  (4). Given an s-by-s food web matrix b = [𝑏𝑖𝑗], with 𝑏𝑖𝑗 corresponding to the number of individuals of species j (galls or parasitoids) emerging from species i (willow or galls) per willow branch over a single growing season, 𝑏𝑖. is the sum of row i, 𝑏.𝑗 is the sum of column j, and 𝑏.. is the total sum. The Shannon indices for the prey and predatory interactions were calculated as, 𝐻𝑗 = − ∑𝑏𝑖𝑗𝑏.𝑗𝑠𝑖=1ln𝑏𝑖𝑗𝑏.𝑗 𝐻𝑖 = − ∑𝑏𝑖𝑗𝑏𝑖.𝑠𝑗=1ln𝑏𝑖𝑗𝑏𝑖.  The effective number of prey and predatory interactions were calculated as 𝑁𝑗∗ =exp(𝐻𝑗) and 𝑁𝑖∗ = exp(𝐻𝑖), respectively. Finally, quantitative-weighted link density was calculated as,   𝐿𝐷𝑞 =  12𝑏. .(∑ 𝑏𝑖. 𝑁𝑖∗  +  ∑ 𝑏.𝑗 𝑁𝑗∗𝑠𝑗=1𝑠𝑖=1)  B.7 Asymptotic vs. non-asymptotic models We fit both asymptotic and non-asymptotic phenomenological models (5) to extrapolate our estimates of food-web complexity. While more sophisticated and accurate models have been developed to extrapolate species richness (3), nothing has been developed for extrapolating food-web complexity. These phenomenological models have the advantage that they make no assumptions about the processes generating the data (3); therefore, they  136 are likely a good starting point for extrapolating food-web complexity.   For our asymptotic model we used a scaled and shifted Michaelis-Menten function (6) of the form,  𝐿𝐷𝑞,𝑁 =𝑎(𝑁−1)(𝑏+(𝑁−1)+ 𝐿𝐷𝑞,1̅̅ ̅̅ ̅̅ ̅,  where N represents either the number of plants (sampling effort simulation) or the number of genotypes (genetic variation simulation). LDq,N is the predicted complexity at N, while a and b are phenomenological parameters that scale LDq,N  and N, respectively. 𝐿𝐷𝑞,1̅̅ ̅̅ ̅̅ ̅ is a constant parameter, representing the average complexity for mixtures of either 1-genotype 1-plant (sampling effort simulation) or 1-genotype 4-plants (genetic variation simulation). Adding the constant, 𝐿𝐷𝑞,1̅̅ ̅̅ ̅̅ ̅, and subtracting the constant, 1, shift the function so that when N = 1, 𝐿𝐷𝑞,𝑁 = 𝐿𝐷𝑞,1̅̅ ̅̅ ̅̅ ̅ . We used non-linear least squares to estimate parameters a and b. For the non-asymptotic models, we fit log-log (log(𝐿𝐷𝑞,𝑁) = m ∗ log(𝑁) + 𝑖𝑛𝑡𝑒𝑟𝑐𝑒𝑝𝑡) and log-linear (𝐿𝐷𝑞,𝑁 = m ∗ log(𝑁) +𝑖𝑛𝑡𝑒𝑟𝑐𝑒𝑝𝑡) models. The asymptotic and non-asymptotic models we chose have been widely used for extrapolating species richness (5), which is why we used them for food-web complexity.  B.8 Results for simulations of sampling effort and genetic variation  We fit the asymptotic and non-asymptotic models to our sampling effort simulations of 1-genotype mixtures of 1 to 4 plants (1,000 estimates per level of sampling effort, details in Materials and Methods). We found that all of the models gave a similar fit to the data; however, they gave very different predictions for the complexity of 1-genotype 100-plant  137 mixtures (Appendix B.4). Therefore, to evaluate which of these models was more realistic, we re-fit these models to our genetic variation simulations of 1 to 25 genotypes (grey circles in Fig. 6 of main text). We found that the asymptotic model provided a much better fit (R2 = 0.96) and more accurate predictions than either of the non-asymptotic models (Appendix B.5). In particular, the asymptotic model’s predicted complexity of 25-genotype 100-plant mixtures deviated less than a tenth of 1% from the observed average (LDq = 2.209), whereas the non-asymptotic models overestimated complexity by 2.4% (log-linear) and 3.1% (log-log).   Comparing asymptotic and non-asymptotic models for predicting the complexity of 1-genotype 100-plant mixtures. Note that for these data (sampling effort simulation), N represents the number of plants. Model type Equation R2 Predicted LDq 1-genotype 100-plant mixture Asymptotic  (Michaelis-Menten) 𝐿𝐷𝑞,𝑁 =0.62(𝑁 − 1)(3.62 + (𝑁 − 1)+ 1.25 0.885 1.84 Non-asymptotic (log-log) log(𝐿𝐷𝑞,𝑁) = 0.15 ∗ log(𝑁) + 0.22 0.881 2.45 Non-asymptotic  (log-linear) 𝐿𝐷𝑞,𝑁 = 0.20 ∗ log(𝑁) + 1.24 0.884 2.17  Comparing asymptotic and non-asymptotic models for predicting the complexity of 25-genotype 100-plant mixtures. The observed complexity of the 25-genotype 100-plant mixture was 2.209. Note that for these data (genetic variation simulation), N represents the number of genotypes.   138 Model type Equation R2 Predicted LDq 25-genotype 100-plant mixture Asymptotic (Michaelis-Menten) 𝐿𝐷𝑞,𝑁 =0.76(𝑁 − 1)(2.25 + (𝑁 − 1)+ 1.52 0.96 2.210 Non-asymptotic (log-log) log(𝐿𝐷𝑞,𝑁) = 0.10 ∗ log(𝑁) + 0.50 0.87 2.277 Non-asymptotic (log-linear) 𝐿𝐷𝑞,𝑁 = 0.19 ∗ log(𝑁) + 1.65 0.89 2.262  B.9 Assessing the accuracy of the asymptotic model After we identified the asymptotic model as the most appropriate for our data, we wanted to evaluate whether the model was likely to over- or under-estimate the complexity of 1-genotype 100-plant mixtures. To do this, we took advantage of the complete data set we had for the genetic variation simulation. Specifically, we refit the asymptotic model with smaller fractions of data to examine how accurately it extrapolated to predict the complexity of 25-genotype 100-plant mixtures. When we did this, we found that the model began to increasingly overestimate food-web complexity, but only slightly. For example, using only the first 40% of the data (i.e. 1 to 10 genotypes), the model overestimated food-web complexity by less than 0.5%, while, using only the first 16% of the data (e.g. 1 to 4 genotypes), the model overestimated food-web complexity by 0.9%. Since our asymptotic model for the sampling effort simulation is extrapolating based on 4% of the potential data (4 of 100 plants), the predicted complexity of 1-genotype 100-plant mixtures is likely an overestimate. This suggests that the reported effect of 20% is a conservative estimate of the additive effects of genetic variation.   139 B.10 Structural equation model of food-web complexity For our plant-insect food web, complexity is principally determined by three components: (i) the effective number of gall species per willow (i.e. Shannon diversity of galls); (ii) the effective number of parasitoid species per gall (vulnerability, Vq); and (iii) the effective number of gall species per parasitoid (generality, Gq) (4). Increases in any of these 3 components, all else equal, will directly increase food-web complexity. Moreover, the total abundance and diversity of galls may indirectly affect complexity by influencing the vulnerability and generality of the gall-parasitoid network. Therefore, we built our structural equation model to incorporate these different pathways. In addition, since species diversity is determined by both the evenness and richness of a community, we partitioned gall diversity into its evenness (E1 = exp(Shannon diversity)/richness) and richness components (7) before building the model. Given the non-linear relationship between genetic variation and food-web complexity (Fig. S1), we restricted our analysis to the first 4-levels of genetic variation. We feel this was justified for two reasons: (i) this was the portion of the relationship that increased the most; and (ii) this was the only portion of the relationship that was mostly linear with constant variance, thereby satisfying the assumptions of the linear regression models that made up our structural equation model. Finally, we used a test of directed separation (8), which essentially tests whether there are any significant paths missing from the model. For tests of directed separation, P > 0.05 indicates that the model provides a good fit to the data (i.e. no missing paths), whereas P < 0.05 indicates a model with missing paths.    140 Fig. S1 shows the data from the one replicate simulation that we used to evaluate the structural equation model in Fig. S2. We found that this model provided a good fit to the data (Fisher C = 11.61, k = 6, P = 0.071). In particular, we found that genetic variation increased food-web complexity primarily by: (i) an increase in gall richness that directly increased complexity (0.40*0.52 = 0.21); and (ii) an increase in gall abundance that indirectly increased complexity by increasing gall vulnerability (0.57*0.55*0.83 = 0.26). Interestingly, gall evenness had a small overall negative effect on complexity ((-0.18*0.39) + (-0.18*-0.32*0.83) + (-0.18*0.25*0.28) = -0.03).   Fig. S1: One of 50 replicate simulations, showing the positive relationship between willow genetic variation and food-web complexity. Grey circles represent estimates of food-web complexity for specific samples, whereas blue circles represent the average complexity at each level of genetic variation. These data were used in the structural equation model (Fig. S2). 1.52.01 5 10 15 20 25No. of willow genotypesFood−web complexity (LDq) 141  Fig. S2: Structural equation model of the paths by which genetic variation influences food-web complexity. Blue and red arrows indicate positive and negative relationships, respectively. One-way arrows indicate modelled paths, whereas double-headed arrows indicate correlated relationships. Numerical values in the middle of each path represent the standardized path coefficients and can be used to determine the magnitude of direct and indirect effects.        −0.000.010.130.13−0.180.220.250.270.28−0.32−0.32−0.360.390.40 0.520.550.570.83GallevennessGallrichnessGallabundanceVulnerabilitygall−parasitoidGeneralitygall−parasitoidFood−webcomplexityGeneticvariation 142 Appendix C   C.1 Abundance responses of key arthropod guilds in ant-aphid experiment Summary of abundance responses of key arthropod guilds in the ant-aphid experiment. We analyzed all of these responses using generalized linear mixed-effect models (error distribution = Poisson, link function = log). We report the likelihood-ratio test statistic and include the degrees of freedom for each test as a subscript next to each predictor. Arthropod abundances Genotype (G)(9) Eaphid(1) Eant(1) G×Eaphid(9) G×Eant(9) Eaphid×Eant(1) G×Eaphid×Eant(9) Leaf-mining moth (Gracilliaridae) 26.78 0.32 2.35 13.56 20.31 4.32 - non-A. farinosa aphids (Aphididae) 24.43 0.01 0.04 23.16 6.99 3.63 8.16 Leafhopper (Cicadellidae) 21.92 0.84 0.01 7.29 11.54 1.67 - Spiders (Araneae) 16.24 0.01 0.10 15.39 11.34 0.01 - non-F. obscuripes ants (Formicidae) 22.43 17.70 1.52 5.21 7.07 0.73 - Leaf-tiering moth (Tortricidae) 23.79 0.81 9.79 - - 3.77      143 C.2 Abundance responses of key arthropod guilds in wind experiment Summary of abundance responses of key arthropod guilds in the wind experiment. We analyzed all of these responses using generalized linear mixed-effect models (error distribution = Poisson, link function = log). We report the likelihood-ratio test statistic and include the degrees of freedom for each test as a subscript next to each predictor. Arthropod abundances Genotype (G)(9) Ewind(1) G×Ewind(9) Eyear(1) G×Eyear(9) Ewind×Eyear(1) G×Ewind×Eyear(9) Leaf-mining moths  (Gracilliaridae) 17.15 9.42 - 4.26 - 0.09 - Gall midges  (Cecidomyiidae) 19.26 17.59 - 38.07 - - - Leaf-tiering moths  (Tortricidae) 24.78 1.34 11.50 117.19 - 2.65 - Aphids2012 (Aphididae) 4.31 0.32 - - - - - Spiders (Araneae) 8.27 6.04 - 6.54 - 0.20 -    144  C.3 Principal components analysis of aboveground plant traits Summary of loadings and variance explained by first two components from separate principal components analysis (PCA) of aboveground plant traits. All traits were scaled to mean = 0 and SD = 1 prior to PCA to give them equal weight in the analysis. Individual traits Trait PC1 Trait PC2 Ant-aphid experiment   Plant height 0.51 -0.49 Shoot count 0.43 0.25 Shoot length 0.64 -0.14 Leaf trichome density -0.12 -0.72 Leaf water content -0.36 -0.39 Variance explained 39% 29% Wind experiment, 2012   Plant height 0.55 -0.18 Shoot count 0.47 0.29 Shoot length 0.68 -0.07 Leaf trichome density -0.08 0.64 Leaf water content 0.09 0.69 Variance explained 36% 24% Wind experiment, 2013   Plant height 0.46 -0.29 Shoot count 0.49 -0.27 Shoot length 0.45 -0.20 Leaf water content -0.36 -0.52 Specific Leaf Area (SLA) -0.46 -0.42 Leaf C:N -0.04 0.59 Variance explained 45% 26%         145 C.4 Principal components analysis of soil properties in wind experiment Summary of loadings and variance explained by first two components from separate principal components analysis (PCA) of soil properties. All soil characteristics were scaled to mean = 0 and SD = 1 prior to PCA to give them equal weight in the analysis.   Soil properties Soil PC1 Soil PC2 NO3--N -0.25 -0.09 NH4+-N -0.15 -0.02 Ca2+ 0.23 0.45 Mg2+ 0.29 0.34 K+ 0.23 -0.20 H2PO4--P 0.23 0.09 Fe3+ 0.33 -0.20 Mn2+ 0.19 0.16 Cu2+ 0.24 -0.30 Zn2+ 0.17 -0.40 B(OH)4--B 0.20 -0.26 SO4--S 0.26 0.10 Pb2+ 0.21 -0.10 Al3+ 0.27 -0.23 Cd2+ 0.05 0.40 Organic matter (%) 0.35 0.14 Soil moisture 0.30 0.04 Variance explained 38% 14% 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0314221/manifest

Comment

Related Items