The Community Ecology of Plant Parasites: from Coinfections to Metacommunities

THE COMMUNITY ECOLOGY OF PLANT PARASITES: FROM COINFECTIONS TO METACOMMUNITIES

Fletcher William Halliday

A dissertation submitted to the faculty at the University of North Carolina at Chapel Hill in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Biology.

Chapel Hill 2017

Approved by:

Charles Mitchell

Allen Hurlbert

James Umbanhowar

Rytas Vilgalys

Peter White

ABSTRACT

FLETCHER WILLIAM HALLIDAY: The Community Ecology of Plant Parasites: from Coinfections to Metacommunities (Under the direction of Charles E. Mitchell)

New emerging diseases and methodological advances have generated a recent surge in disease ecology research and renewed interest in identifying the ecological processes that structure parasite communities. Yet ecologists still lack a general framework for understanding the drivers of parasite diversity. Metacommunity theory is a general ecological framework that has been used to understand patterns of community composition in many ecological systems. This dissertation leverages one key insight from metacommunity theory – that multiple processes operate across different spatial and temporal scales to control the composition of local communities – to understand parasite communities within hosts. In this work, I used experimental studies to examine the community ecology of parasites that infect wild host plants over space and time.

At the smallest spatial scale, I explored how interactions among parasites in the same host leaf during coinfection alter parasite epidemics. Within host leaves, parasite growth was influenced by coinfections, but coinfections were often prevented by the sequence of parasite infection, generating priority effects within hosts. Coinfections, priority effects, and the severity of infections were altered by host immunity. Scaling up, I found that parasite phenology, which operates across host individuals, altered host susceptibility to secondary infections, parasite interactions, and ultimately the magnitude of parasite epidemics.

iii At the largest spatial scale, I explored how characteristics of host communities influence the diversity of parasite metacommunities. Parasite diversity across host communities depended on host diversity and resource supply to hosts. Host richness alone could not explain most changes in parasite diversity. However, shifting host composition allowed disease amplification, depending on parasite transmission mode. These effects also varied over time: the structure of host communities changed in response to initial host diversity and resource supply to hosts, leading to altered parasite richness and abundance. Together, these results highlight the utility of multiscale approaches to disease ecology. Specifically, integrating concepts from community ecology with information about infectious diseases and host-parasite interactions provides insight into the general mechanisms that control the diversity of parasites across space and time.

For Simone: my anchor during this journey through academia.

ACKNOWLEDGEMENTS

The completion of this dissertation was only achieved thanks to many collaborators, friends, and family members. First, I wish to thank my committee members, for challenging me to think more critically about my research questions, design, and implementation. Rytas Vilgalys helped me find traction in an unexplored study system. Allen Hurlbert and James Umbanhowar provided statistical and theoretical advice. James rarely missed an opportunity to challenge an assumption, question a relationship, or suggest and alternative approach, and this research is better for it. Peter White’s inquisitiveness and insight is infectious, and Peter’s unanswerable questions about generality in ecology guided my research approach during graduate school.

Charles Mitchell is the type specimen of a great mentor, giving me space and resources to pursue countless projects early on when time wasted was time well spent; motivating me to finally analyze my results and advising me on how to write them up when it was time to finish; and patiently listening when I talked and talked and talked for what must have seemed like and endless and unrelenting amount of time. Most importantly, Charles has been my advocate during this particularly vulnerable time in my career.

I would like to acknowledge my funding sources. Much of my research was funded by the NSF-USDA joint program in Ecology and Evolution of Infectious Diseases (NSF grant DEB-

1015909 and USDA-NIFA AFRI grant 2016-67013-25762). The NSF Graduate Research

Fellowship Program provided three-years of funding allowing me to focus entirely on my graduate studies, rather than paying my bills. UNC’s Alma Holland Beers Scholarship provided

vi summer funding which allowed me to perform full-time field work each summer after the GRFP expired. UNC’s Mrs. Coker Fellowship provided me a stipend for the Fall 2010 and Fall 2011 semesters. UNC’s WC Coker Fellowship provided summer funding and a stipend for the Fall

2017 semester. UNC’s Graduate School Dissertation Completion Fellowship provided a stipend for Fall 2015 and Spring 2016. During my remaining time at UNC, the Department of Biology provided TA-ships to support me.

I have benefitted tremendously from my labmates. Rob Heckman, Megan Rúa, Miranda

Welsh, and Kayleigh O’Keeffe provided sage advice and critical feedback. They improved my writing, stopped me from pursuing many half-baked research ideas, prevented mid-season fieldwork meltdowns, and generally improved my outlook on science. Bradley Saul changed the way I think about causation in research. I owe James Cronin for much of my early intellectual development during graduate school. I am thoroughly convinced that without his guidance, I would not have received the graduate research fellowship that opened so many doors for me.

Erin Mordecai was a role model, demonstrating how to balance a productive work life with a fulfilling personal one.

Peter Wilfahrt and Rob Heckman are wonderful collaborators and even better friends. We began a collaborative experiment on a whim, and through their tutelage and hard work, this project ultimately formed two chapters in this dissertation, and the backbone of my academic career.

During my time in graduate school, I have benefitted from coffee, and conversations, hikes, writing workshops, arguments, and board games with Peter, Rob, Erin, Bianca Lopez,

Miranda Maring, Alissa Brown, Colin Maxwell, Kriti Sharma, Kyle Palmquist, and Jes Coyle.

vii There are no words to describe how valuable these friendships have been to me. I’ve particularly benefitted from knowing Andrew Loyd, who taught me everything that I know about plant pathology, gardening, and brewing.

My research benefitted from assistance and discussions with many undergraduate students, particularly Briana Whitaker, Markus Le, Kristina Jacobs, Robert Price, Ben Robb,

Anita Simha, Prahlada Papper, and a group of 15 students from Santa Rosa Junior College, who designed and implemented a novel study on an emerging pathogen.

I want to thank my family. Louise Abrahams, Ashley Halliday and Ann Halliday provided roofs over my head and warm dinners during months of field work that ultimately did not make it into this dissertation. Ashley was also a tireless field assistant during a time when I would have otherwise been completely alone. Hugo, Stanley, and Penelope were helpful distractions when I needed them most. Thank you to Ivo, who was born just as this document came together. I am so grateful for the countless sleepless nights spent worrying about you rather than that damned final dissertation chapter.

Finally, I want to thank my spouse Simone. In 2010, we tied the knot and promptly moved across the country to a strange place with few personal connections. It has turned out to be the greatest adventure of my life, and I could not have done it without your guidance and understanding as my partner. Graduate school was an opportunity to find my personal and professional limits. How many hours can I sit in front of an r console without taking a break?

How many plants can I survey before they become indistinguishable from one another? How many consecutive episodes of Dr. Who can I watch in one sitting, when academics have become

viii too much? Simone facilitated my exploration of these limits, but also served as an anchor, preventing me from over-stepping them.

TABLE OF CONTENTS

LIST OF TABLES ...... xiv

LIST OF FIGURES ...... xvi

CHAPTER 1 : INTRODUCTION ...... 1

CHAPTER 2 : INTERACTIONS AMONG SYMBIONTS OPERATE ACROSS SCALES TO INFLUENCE PARASITE EPIDEMICS ...... 7

Introduction ...... 7

Within-host interactions that determine parasite growth rates ...... 8

Within-host priority effects ...... 9

Parasite phenology may alter within-host interactions and epidemics ...... 10

Methods ...... 11

Study system ...... 11

Experimental Design ...... 12

Survey ...... 14

Data analysis ...... 15

Results ...... 17

Does parasite phenology alter epidemics? ...... 17

Does parasite phenology alter within-host priority effects? ...... 18

Does parasite phenology alter within-host interactions that determine parasite

x growth rates? ...... 19

Discussion ...... 21

Conclusions ...... 24

CHAPTER 3 : HOST IMMUNITY MODIFIES INTERACTIONS AMONG PARASITES AND ALTERS PARASITE EPIDEMICS ...... 32

Introduction ...... 32

Methods ...... 35

Experimental design ...... 37

Survey ...... 38

Disease metrics ...... 39

Data analysis ...... 39

Results ...... 42

Does host immunity alter within-host priority effects? ...... 42

Do immune-mediated priority effects alter parasite prevalence? ...... 43

Does host immunity alter within-host parasite growth? ...... 43

What are the consequences for immune-mediated parasite interactions on parasite

epidemics? ...... 44

Discussion ...... 45

CHAPTER 4 : A MULTIVARIATE TEST OF DISEASE RISK REVEALS CONDITIONS LEADING TO DISEASE AMPLIFICATION ...... 52

Introduction ...... 52

Methods ...... 57

xi Host composition and species richness ...... 57

Resource supply treatment ...... 59

Quantification of host abundance, parasite abundance and parasite richness ...... 59

Data analysis ...... 61

Results ...... 63

Discussion ...... 66

CHAPTER 5 : ASSEMBLY OF THE HOST COMMUNITY INFLUENCES PARASITE RICHNESS AND ABUNDANCE IN A PLANT DIVERSITY EXPERIMENT ...... 74

Introduction ...... 74

Changes in host species richness during community assembly ...... 75

Changes in exotic host abundance during community assembly ...... 76

Changes in host phylogenetic diversity during community assembly ...... 77

Methods ...... 78

Host composition and species richness ...... 79

Resource supply treatment ...... 80

Quantification of host community assembly ...... 80

Quantification of parasite abundance and parasite richness ...... 81

Data analysis ...... 83

Results ...... 85

Discussion ...... 87

CHAPTER 6 : CONCLUSION ...... 96

xii APPENDIX A: SUPPLEMENTARY MATERIAL FOR CHAPTER 2...... 100

A1. Surveys to determine parasite order of arrival into the host population ...... 100

A2. Details of planting treatments and leaf surveys ...... 101

A3. Details of data analysis ...... 102

A4. Supplemental Figures ...... 106

A5. Supplemental tables ...... 108

APPENDIX B: SUPPLEMENTARY MATERIAL FOR CHAPTER 3...... 114

B1. Supplemental tables ...... 114

APPENDIX C: SUPPLEMENTARY MATERIAL FOR CHAPTER 4...... 117

C1. Supplemental tables ...... 117

APPENDIX D: SUPPLEMENTARY MATERIAL FOR CHAPTER 5...... 123

D1. Supplemental tables ...... 123

D2. Supplemental figures ...... 128

REFERENCES ...... 131

xiii LIST OF TABLES

Table 2.1 – Biology and ecology of the focal symbionts ...... 26

Table 5.1– Test of mediation for the final (reduced) model. Fully mediated model includes paths from experimental treatments to mediators only. Partially mediated models include paths from experimental treatments to mediators and responses...... 92

Table A2.1 – Cohort 1 survival analysis ANOVA ...... 108

Table A2.2 – Cohort 2 survival analysis ANOVA ...... 109

Table A2.3 – Cohort 3 survival analysis ANOVA ...... 109

Table A2.4 – Reduced model coefficients. Models were reduced from a full model using likelihood ratio tests to remove non-significant interactions. Estimates are only provided if they were included in the reduced model. Significant effects (p<0.05) from those likelihood ratio tests are indicated in bold. Coefficients are exponentiated...... 110

Table A2.5 – Cohort 1 longitudinal linear mixed model ANOVA ...... 111

Table A2.6 – Cohort 2 longitudinal linear mixed model ANOVA ...... 112

Table A2.7 – Cohort 3 longitudinal linear mixed model ANOVA ...... 112

Table A2.8 – Reduced model coefficients. Models were reduced from a full model using likelihood ratio tests to remove non-significant interactions. Estimates are only provided if they were included in the reduced model. Significant effects (p<0.05) from those likelihood ratio tests are indicated in bold. Coefficients are on a log scale...... 113

Table B3.1 – Reduced model coefficients. Models were reduced from a full model using likelihood ratio tests to remove non-significant interactions. Estimates are only provided if they were included in the reduced model. Coefficients are exponentiated for Cox mixed models and on a log scale for longitudinal mixed models ...... 114

Table B3.2 – Disease risk [Cox mixed model] ANOVA ...... 115

Table B3.3 – Parasite Prevalence ANOVA ...... 115

Table B3.4 – Infection severity [longitudinal linear mixed model] ANOVA ...... 116

Table B3.5 – Coinfection ANOVA ...... 116

xiv Table B3.6 – Parasite Burden ANOVA ...... 116

Table C4.1 – ANOVA for parasite abundance and parasite richness models. A) Parasite abundance, B) Rarefied parasite richness...... 117

Table C4.2 – ANOVA for multi-response regression models. “Response” is the multivariate response of insect and microbial parasite abundance and richness ...... 118

Table C4.3 – Estimated model terms for the transformed responses insect abundance, insect richness, microbial abundance, and microbial richness. (a) fixed coefficients (averaged over block), (b) the variance covariance matrix (left) and correlations (right) ...... 119 Table C4.4 – Parasite morphospecies (symptoms in brackets) and associated hosts ...... 120

Table D5.1 – Parasite morphospecies and their associated host species grouped into four categories: A) planted host species, B) native colonizing host species, C) exotic colonizing host species, and D) host species with unknown geographic provenance. Parasite morphospecies is presented in the leftmost column, with parasite type in brackets and genbank accession numbers in parentheses...... 124

Table D5.2 – Piecewise SEM goodness of fit Test. A) Conditional independence claims for a direct separation test using the full model. B) Results of the direct separation test (p<0.05 indicates that the model should be rejected) ...... 125

Table D5.3 – Coefficient estimates from the full model. Estimates are standardized to a common scale to facilitate comparisons. Correlations among dependent variables are indicated by ~~...... 126

Table D5.4 – Coefficient estimates from the final (reduced) model. Estimates are standardized to a common scale to facilitate comparisons. Significant predictors (p<0.05) are indicated in bold. Correlations among dependent variables are indicated by ~~...... 127

xv LIST OF FIGURES

Figure 2.1 - Hypothetical interaction network between a vertically transmitted fungal endophyte and subsequently colonizing fungal parasites. Blue arrows represent positive interactions (e.g., facilitation). Red clubs represent negative interactions (e.g., inhibition)...... 27

Figure 2.2 - Parasite sequence of arrival into the host population during 2013 – 2015 surveys. Points represent the average first date that at least 1% of host leaves in 2013 and 2014 and 1% of plots in 2015 were infected by each parasite. Error bars represent the earliest and latest date of first infection across the surveys...... 28

Figure 2.3 - Seasonal epidemics in three experimental cohorts of hosts. Points represent parasite prevalence (the proportion of leaves infected across all sentinel hosts) in each survey. Illustrative lines are LOESS fit to the data with span=0.9. Vertical lines represent the dates that each cohort was placed into the field, highlighting the overlap in time among cohorts...... 29

Figure 2.4 – Summary of symbiont interaction network indicated by the interaction models. Blue arrows represent positive interactions (e.g., facilitation). Red clubs represent negative interactions (e.g., inhibition). Dashed lines represent hypothesized interactions that were not supported by the models (p>0.05). “E” stands for Epichloë coenophiala, “P” stands for Puccinia coronata, “C” stands for Colletotrichum cereale, “R” stands for Rhizoctonia solani. a) Within-host priority effects from Cox-proportional hazards models. b) Interactions influencing within-host growth from longitudinal linear mixed models...... 30

Figure 2.5 – Model-estimated relative risk of infection as leaves age. Plots are results of the reduced Cox mixed models. A value above zero indicates that previous infection of a leaf by Colletotrichum (red), Puccinia (blue), Rhizoctonia (green), or endophyte-infected seed (black) increased the risk of subsequent infection by the focal parasite. A value below zero indicates that previous infection decreased the risk of subsequent infection by the focal parasite. Vertical lines along the x-axis show the age of each leaf when it became infected by the focal parasite, colored by the infection status of that leaf by other parasites. X-axis values are jittered to show the data. a) Cohort 1 Colletotrichum infection risk. b) Cohort 1 Puccinia infection risk. c) Cohort 1 Rhizoctonia infection risk. d) Cohort 3 Puccinia infection risk. These model results are summarized in Figure 2.4a...... 31

xvi Figure 3.1 – Model-estimated relative risk of infection. Plots are results of the reduced Cox mixed models, and are on a logarithmic scale. Points represent the treatment mean and error bars represent the 95% confidence interval. A value above zero indicates that the experimental treatment, or previous infection of a leaf by the other parasite (red) increased the risk of subsequent infection by the focal parasite. A value below zero indicates that the treatment or previous infection decreased the risk of subsequent infection by the focal parasite. a) Rhizoctonia infection risk. b) Colletotrichum infection risk...... 49

Figure 3.2– Model-estimated effects of experimental treatments (grey = control, black = JA, red = SA) on epidemics of Colletotrichum and Rhizoctonia. Points represent the treatment mean and error bars represent the 95% confidence interval. a) Rhizoctonia infection prevalence, calculated as the proportion of leaves across the entire plant that were infected by each parasite. b) Colletotrichum infection prevalence. c) Rhizoctonia leaf burden, calculated as the area under the disease progress stairs for each parasite. d) Colletotrichum leaf burden. e) Coinfection frequency, calculated as the proportion of longitudinally surveyed leaves that became coinfected during the course of the experiment...... 50

Figure 3.3 – Model-estimated within-host parasite growth. Plots are results of the reduced longitudinal mixed models, showing the rate at which log-transformed infection severity by Rhizoctonia increased as leaves age (i.e., parasite growth rate), and the effects of previous Colletotrichum infection on that relationship. Colors and contour lines represent mode-estimated Rhizoctonia infection severity. Points represent individual observations of leaves over the course of the experiment...... 51

Figure 4.1 – Relationships among host diversity, resource supply to hosts, parasite richness, and parasite abundance can be decomposed into their component parts. Host diversity effects can be decomposed into those that are driven by variation in host composition and those driven by variation in host species richness. Parasite richness and abundance can be decomposed into characteristics of parasite species, such as parasite taxonomic groups (here, insects vs microbes)...... 71

Figure 4.2– Effects of host diversity (monoculture, polyculture) and resource supply to hosts (ambient, black circles; fertilized, red triangles) on parasite abundance, back-transformed from the inverse hyperbolic sine (top, panels A and B), and rarefied parasite richness (bottom, panels C and D). Error bars represent 95% confidence intervals. The left panels (A and C) show the overall effects of the host diversity treatment on parasite abundance and

xvii richness. The right panels (B and D) show the effects of host diversity after accounting for variation in host composition...... 72

Figure 4.3 - Effects of host diversity (mono = monoculture; poly = polyculture) and resource supply to hosts (ambient, black circles; fertilized, red triangles) on insect and microbial parasite abundance and richness, calculated using a multi-response regression with standardized response variables. The panels show the effects of host diversity on insect and microbial parasite abundance and richness before (top) and after (bottom) accounting for host composition. For example, the leftmost four points (top panel) show positive effects of host diversity (i.e., amplification effect) and negative effects of soil fertilization on insect abundance; the amplification effect becomes non-significant after accounting for composition (bottom panel). Estimates are from a reduced model omitting the non-significant interaction between host diversity and resource supply. Error bars represent 95% confidence intervals. Coefficients that share a letter do not differ significantly as determined by the Bonferroni correction, α* = 0.05/4 = 0.0125...... 73

Figure 5.1 – Hypothesized effects of host diversity and resource supply on parasite richness and abundance, mediated by future host community structure (i.e. community assembly). Straight arrows represent causal relationships, and curved arrows represent correlations. A) Conceptual metamodel. B) Statistical measurement (full) model: Residuals are denoted by ε for response variables and ζ for mediating variables. Each dependent variable may be altered by the experimental covariate of block, modeled as a linear combination of coefficients, where the number of coefficients equals the number of levels minus 1. These block effects are depicted with four covariates (BLK 2 – BLK 5) and brackets around the dependent variables. Paths are labeled a-o for reference in the text...... 93

Figure 5.2 – Piecewise structural equation model results for the final (reduced) model. Dashed lines are non-significant (p> 0.05). All coefficients are standardized. Correlations between errors are denoted with double-headed arrows. R2 is the marginal R2 from the linear mixed-model, which represents the variance explained by fixed effects in the model. *p = 0.056 ...... 94

xviii in the model (e.g., the left panel in (d) shows the effect of plant species richness on parasite richness after accounting for the effects of exotic abundance and plant phylogenetic diversity on parasite richness). Regression lines are drawn for significant relationships only...... 95

Figure A2.1 – Cohort 1 within-host parasite growth. Plots are results of the reduced longitudinal mixed models, showing the rate at which log-transformed infection severity among infected leaves increased as leaves age (i.e., parasite growth rate), and the effects of other symbionts on that relationship. a-c) Colletotrichum infection severity as a function of previous Puccinia infection severity, previous Rhizoctonia infection severity, and endophyte infection, respectively. Colors and contour lines represent model-estimated Colletotrichum infection severity. d-e) Puccinia infection severity as a function of previous Colletotrichum infection severity and endophyte infection, respectively. f) Rhizoctonia infection infection severity as a function of previous Colletotrichum infection severity. Points represent individual observations of leaves over the course of the experiment. “Neg effects” are the number of leaves where the model estimated a negative effect of the other parasite on the focal parasite. “Pos effects” are the number of leaves where the model estimated a positive effect. “Mean effect” is the model-estimated per-capita effect of the other parasite on the focal parasite. * denotes estimated effects for models where there was no interaction between leaf age and previous infection severity by the other parasite. † denotes estimated effects for models where the main effect was non- significant. These model results are summarized in Figure 2.4b...... 106

Figure A2.2 – Cohort 3 within-host parasite growth. Plots are results of the reduced longitudinal mixed models, showing the rate at which log-transformed infection severity among infected leaves increased as leaves age (i.e., parasite growth rate), and the effects of other symbionts on that relationship. a) Colletotrichum infection severity as a function of endophyte infection. b) Puccinia infection severity as a function of endophyte infection. c) Rhizoctonia infection severity as a function of previous Colletotrichum infection severity. † denotes estimated effects for models where the main effect was non-significant. These model results are summarized in Figure 2.4b...... 107

Figure C4.1 – Effects of host diversity (mono = monoculture plots; poly = polyculture plots) and resource supply to hosts (ambient, black circles; fertilized, red triangles) on insect and microbial parasite abundance and richness, calculated using a multi-response regression model, and standardized to a common variable. The top panel shows the overall effects of host diversity on insect and microbial parasite abundance and richness. The bottom panel shows the effects of host diversity after accounting for variation in host

xix composition. Coefficients that share a letter do not differ significantly as determined by multiple comparisons tests with the Bonferroni correction, α* = 0.05/4 = 0.0125. Estimates are from the full model that includes a non-significant interaction between host diversity and soil fertility...... 121

Figure C4.2 – Effects of host diversity (monoculture plots; polyculture plots) and resource supply to hosts (ambient, black; fertilized, red) on the abundance of each parasite morphospecies, standardized to a common variable. Violins show distribution of the data. Asterisks show the mean of each group...... 122

Figure D5.1 – The final (reduced) model with parasite richness rarefied to two host individuals per plot. A subsample of two host individuals represents the minimum number of host individuals sampled per plot. Dashed lines are non-significant (p

Figure D5.2 – The final (reduced) model with parasite richness rarefied to five host individuals per plot. A subsample of five host individuals represents the median number of host individuals sampled per plot. Dashed lines are non-significant (p

Figure D5.3 – Piecewise structural equation model results for the full model. Dashed lines are non-significant (p> 0.05). All coefficients are standardized. Correlations between errors are denoted with double-headed arrows. *p = 0.057 ...... 130

xx CHAPTER 1 : INTRODUCTION

The diversity of parasites – organisms that live in and on hosts, potentially causing disease

– may rival the diversity of all other organisms on earth (Dobson et al. 2008). Yet, until recently, parasite diversity comprised an undervalued component of global biodiversity. In natural systems, parasite diversity can influence disease risk and host community structure (Hersh et al. 2012,

Johnson et al. 2013a). Understanding the drivers of parasite diversity may be important for predicting the emergence and spread of infectious diseases, an increasingly urgent need due to the emergence of diseases that pose threats to human, wildlife, and ecosystem health (Daszak et al.

2000, Hatcher et al. 2012, Boyd et al. 2013). However, despite the importance of parasites, ecologists still lack a general framework for understanding the drivers of parasite diversity.

Metacommunity theory is a general framework for understanding the processes that shape ecological communities including those of parasites (Mihaljevic 2012, Richgels et al. 2013,

Dallas and Presley 2014). One key insight from metacommunity theory is the observation that multiple processes operate simultaneously across spatial and temporal scales to control the composition of local communities (Leibold et al. 2004, Holyoak et al. 2005, Logue et al. 2011).

This dissertation applies this insight from metacommunity theory to advance a more general understanding of the processes structuring parasite communities in nature.

At its simplest, metacommunity theory posits that local and regional processes jointly determine the structure of biological communities by influencing the abundance of species in a given location (Leibold et al. 2004, Holyoak et al. 2005, Logue et al. 2011). Interactions among individuals define the scale of “local processes” (Ricklefs 1987). Local processes can be strongly

1 influenced by “regional processes”, which occur across larger spatial and temporal gradients.

Furthermore, regional processes can influence the richness and abundance of species available to colonize the local scale, thereby defining the subset of interactions that can occur in a given place and time (Fukami 2015). Because of the intimate association between parasites and a single individual host, local processes are likely to operate within host individuals while regional processes should operate across hosts (Kuris et al. 1980, Dove and Cribb 2006, Bordes and

Morand 2008, 2009, Borer et al. 2016).

To explore this concept from metacommunity theory, the research presented in this dissertation utilizes a model system: fungal parasites that infect wild plant leaves. This research spans spatial scales, from individual coinfections within host leaves to entire parasite metacommunities across communities of hosts, to identify the processes that influence parasite diversity. At the smallest scale, I consider a plant leaf to represent a local habitat patch for parasites (follwing Tollenaere et al. 2015). At this scale, interactions occur among parasites as they compete for resources within host plants or alter host immune responses. Just as multiple habitat patches scale up to form metapopulations and metacommunities, so, too, do habitat patches within hosts, scaling up from host leaves to host individuals to host populations and communities (Borer et al. 2016). At the scale of a host community, characteristics of the host community that determine the movement or spread of parasites, such as the density of susceptible hosts or host competence, may alter parasite diversity (Seabloom et al. 2015). Host communities also change over time in a process known as community assembly

(HilleRisLambers et al. 2011), and these changes over time during host community assembly may underlie predictable shifts in parasite diversity (Liu et al. 2017).

2 Using these insights from metacommunity theory, this dissertation addresses two broad questions: 1) How do interactions among parasites and other microorganisms influence parasite epidemics across scales? 2) How do characteristics of host communities, which change over time during host community assembly, interact with characteristics of individual parasite species to alter the richness and abundance of entire parasite metacommunities? The results of this work show that interactions among parasites at the scale of host leaves can alter parasite epidemics across hosts, but these interactions are dependent on characteristics of an individual host’s immune system as well as parasite phenology across hosts. At the scale of host communities, host composition can interact with characteristics of individual parasites to alter parasite diversity, but this effect changes over time as host communities assemble. These results highlight the dynamic and often complex pathways that connect host and parasite communities across space and time.

CHAPTER SUMMARIES

In Chapter 2, I explore how interactions among symbionts (i.e., any organism that spends at least one life history stage living in or on a single host individual, including, but not limited to parasites) influence parasite epidemics across scales. Parasite epidemics may be influenced by interactions among symbionts, which can depend on past events at multiple spatial scales. Within host individuals, interactions can depend on the sequence in which symbionts infect a host, generating priority effects. Across host individuals, interactions can depend on parasite phenology. To test the roles of parasite interactions and phenology in epidemics, I performed a field experiment in collaboration with Charles Mitchell and James Umbanhowar. I embedded multiple cohorts of sentinel plants, grown from seeds with and without a vertically transmitted

3 symbiont, into a wild host population, and tracked foliar infections caused by three common fungal parasites. Within hosts, parasite growth was influenced by coinfections, but coinfections were often prevented by priority effects among symbionts. Across hosts, parasite phenology altered host susceptibility to secondary infections, symbiont interactions, and ultimately the magnitude of parasite epidemics. Together, these results indicate that parasite phenology can influence parasite epidemics by altering the sequence of infection and interactions among symbionts within host individuals.

In Chapter 3, I tested whether coinfections, priority effects, and the severity of infections were altered by host immunity. Parasite epidemics can be influenced by interactions among parasites. These interactions may result from the host immune response to prior infection, resulting in priority effects. I hypothesized that immune-mediated interactions and priority effects would depend on parasite feeding strategies. To test the role of host immunity on parasite interactions and epidemics, I applied plant immune-signaling hormones to sentinel plants, embedded into a wild host population, and tracked foliar infections caused by two common fungal parasites. Within hosts, parasite growth and priority effects were influenced by the immune-signaling hormone, Salicylic Acid (SA). Hosts treated with SA experienced fewer coinfections, lower prevalence of an endemic parasite, and increased severity of infection by an epidemic parasite. Together, these results indicate that host immunity can alter within-host priority effects and within-host parasite growth among infected hosts, resulting in shifts in parasite prevalence, the frequency of coinfection, and the severity of disease experienced by hosts.

In Chapter 4, I explore how characteristics of host communities interact with characteristics of individual parasite species to alter the richness and abundance of entire parasite

4 metacommunities (i.e., disease risk). Theory predicts that increasing biodiversity will dilute the risk of infectious diseases under certain conditions and will amplify disease risk under others. Yet, few empirical studies demonstrate amplification. This contrast may occur because few studies have considered the multivariate nature of disease risk, which includes richness and abundance of parasites with different transmission modes. To address this question, I designed a field experiment in collaboration with Rob Heckman, Peter Wilfahrt, and Charles Mitchell that manipulated host

(plant) richness, composition, and resource supply to hosts. Using a multivariate statistical model developed for biodiversity-ecosystem-multifunctionality, I revealed that (1) host richness alone could not explain most changes in disease risk, and (2) shifting host composition allowed disease amplification, depending on parasite transmission mode. Overall, this study demonstrates that multiple drivers, related to both host community and parasite characteristics, can influence disease risk. Further, it provides a framework for evaluating multivariate disease risk in other systems.

In Chapter 5, I used the same multi-factorial field experiment as in Chapter 4 to test whether the effects of host diversity and resource supply to hosts on parasite richness and abundance change over time as host communities assemble. Increased host diversity is commonly associated with a reduction in disease risk. However increased host diversity is also commonly associated with shifts in host species composition and host phylogenetic diversity. Many empirical studies show that these characteristics of host communities that are correlated with host diversity may actually drive the relationship between host diversity and disease risk. The association between these characteristics and host diversity, and their influence on disease risk may result from host community assembly over time. To test this, we planted herbaceous perennial communities at two levels of host richness (one- and five-species), and two levels of resource supply to hosts

(ambient, fertilized). We then allowed host communities to be naturally colonized for two years,

5 measured post-assembly host composition, and host species richness, and quantified post-assembly disease risk by measuring parasite richness and parasite abundance. We hypothesized that initial host richness and resource supply to hosts would alter parasite richness and abundance by altering post-assembly host richness, the abundance of exotic host species, and the phylogenetic diversity of the host community. Consistent with our hypothesis, the effects of initial host richness and resource supply to hosts depended on subsequent changes in these three measures of host richness and composition. These results support the growing body of evidence that parasite abundance is most strongly influenced by host composition and phylogenetic diversity. In contrast to previous studies, these results highlight a causal relationship between host richness and parasite abundance, despite the strong association between host composition and parasite abundance. Consequently, these results provide a new mechanism by which host richness may alter disease: host richness influenced host compositional changes, which subsequently altered disease. Together, these results provide insight into the multiple pathways that connect host and parasite communities during host community assembly.

CHAPTER 2 : INTERACTIONS AMONG SYMBIONTS OPERATE ACROSS SCALES TO INFLUENCE PARASITE EPIDEMICS

Introduction

The diversity of parasites and pathogens (hereafter, “parasites”) can influence parasite epidemics (Dobson et al. 2008, Hersh et al. 2012, Johnson et al. 2013b). More generally, epidemics may be driven by interactions among diverse parasites and other symbionts that share a host (Rynkiewicz et al. 2015, Susi et al. 2015, Busby et al. 2016). Many field-based studies of symbiont interactions employ a largely “deterministic” framework (following Fukami 2015), in which the strength and direction of these interactions are assumed to be fixed (Fenton et al.

2014). However, interactions among symbionts may also be contingent on past events. Within hosts, priority effects occur when interactions are contingent on the sequence in which symbiont species infect an individual host (e.g., Kennedy et al. 2009; Hoverman et al. 2013; Adame-

Avarez et al. 2014). Across hosts, symbiont species often differ in their phenology, thus emerging or arriving into a host population sequentially (Schmidt et al. 2007, Dumbrell et al.

2011, Mundt and Sackett 2012). Yet the influence of parasite phenology on parasite interactions and epidemics remains unmeasured under field conditions. This study experimentally tests how parasite phenology influences deterministic interactions and priority effects, and measures the consequences for parasite epidemics in the field.

7 Within-host interactions that determine parasite growth rates

A symbiont is any organism that spends at least one life history stage living in or on a single host individual; symbionts span the continuum from parasites, which reduce host fitness, to mutualists, which increase host fitness (Starr 1975). Understanding how interactions among symbionts within hosts influence parasite epidemics is an important frontier in disease ecology

(Rynkiewicz et al. 2015, Seabloom et al. 2015, Tollenaere et al. 2015). Coinfecting symbionts may interact by competing for limiting resources in the host or via their impacts on host physiology, including host immune responses (Lello et al. 2004, Tollenaere et al. 2015). These interactions, which can generate mixtures of inhibition and facilitation among coinfecting symbionts, can alter symbiont population dynamics in both plants and animals (Eswarappa et al.

2012, Tollenaere et al. 2015). However, whereas symbiont interactions can be readily measured in the lab (Graham 2008), measuring symbiont interactions in the field is more challenging

(Fenton et al. 2010, 2014, Zhan and McDonald 2013). Although many analytical approaches have been proposed, model validation using theoretical (Fenton et al. 2010) and field-collected data (Fenton et al. 2014) indicates that longitudinal mixed-models of within-host parasite growth provide the most reliable measurement of symbiont interactions during natural epidemics.

While these models can reliably measure interactions within hosts (Hellard et al. 2015), they implicitly assume that the strength and direction of pairwise interactions among coinfecting symbionts is the same regardless of the historical context in which they occur (i.e.,

“deterministic” following Fukami 2015). This assumption is true for some parasites (e.g., Sousa

1993), but interactions among coinfecting symbionts may also be contingent on the sequence of past events, generating priority effects within hosts (Fukami 2015). Because longitudinal mixed models average over multiple individual hosts, and individual hosts may have experienced

8 different past events, these models may fail to identify important interactions if the strength and direction of those interactions depend on the past events each host experienced. Longitudinal mixed-models may also miss important interactions when prior infection prevents coinfections altogether via induced resistance or other mechanisms of interference (Rohani et al. 2003,

Leventhal et al. 2015).

Within-host priority effects

Within hosts, interactions among symbionts may result from priority effects, in which the per-capita strength of inhibition or facilitation among symbionts is altered by their sequence of arrival (Hoverman et al. 2013, Werner and Kiers 2015, Mordecai et al. 2016). Among free-living species, priority effects are common and their impacts are well-established (Fukami 2015), and within-host priority effects may be similarly common among symbionts. Vannete & Fukami

(2014) posited that priority effects are most likely to occur when species exhibit high niche overlap. This occurs when species require similar resources, share natural enemies, or for symbionts, respond to similar host immune processes. Additionally, priority effects should be more common when the early arriving species have large impacts on that niche and when the late arriving species are highly sensitive to the availability of that niche (Vannette and Fukami 2014).

These requirements may be commonly fulfilled for symbionts sharing a host. Because all symbionts require host resources for survival, growth, and reproduction (Stearns 1992, Roff

1993), they may exhibit some niche overlap and high sensitivity to the availability of that niche when they coinfect the same host individual. The resulting interactions among coinfecting parasites are well documented in plants (Tollenaere et al. 2015), wild animals (Ezenwa 2016), and humans (Griffiths et al. 2014). Early arriving symbionts can influence the success of later

9 arriving symbionts by impacting host fitness (Randall et al. 2013) or altering host immunity

(Lello et al. 2004, Graham 2008, Cobey and Lipsitch 2013). Impacts on host fitness and immunity may be related to the symbiont’s feeding strategy (Newton et al. 2010). Thus, symbiont feeding strategies may underlie and predict their priority effects.

Within-host priority effects among symbionts have been studied relatively extensively using mathematical models (Rohani et al. 2003, Leventhal et al. 2015, Mordecai et al. 2016) and experimental laboratory inoculations (e.g., Kennedy et al. 2009; Natsopoulou et al. 2015; Werner

& Kiers 2015; Klemme et al. 2016). However, extrapolating from lab inoculations to natural epidemics can be challenging. For example, lab inoculations often require unrealistically high concentrations of symbiont inoculum, potentially generating unrealistic interactions among symbionts. Yet within-host priority effects have largely remained unmeasured under field conditions (but see Laine 2011).

Parasite phenology may alter within-host interactions and epidemics

Symbionts often vary in their phenology, causing them to emerge or arrive into a host population sequentially (Schmidt et al. 2007, Dumbrell et al. 2011). Variation in the timing of parasite emergence can alter the rate of parasite spread across a landscape (Mundt et al. 2009,

Tian et al. 2015), thereby directly altering parasite epidemics. Variation in phenology may also indirectly influence epidemics by altering within-host interactions. The sequence in which symbionts infect individual hosts may be altered by the sequence in which symbionts arrive into the host population, thereby altering within-host priority effects. Those changes in within-host priority effects may then prevent or allow epidemics of coinfecting parasites (Leventhal et al.

2015, Mordecai et al. 2016). Similarly, the sequence of arrival into the host population may

10 influence which symbiont interactions can influence parasite growth or reproduction within hosts, which may slow or accelerate epidemics of coinfecting parasites (Susi et al. 2015, Ezenwa

2016). Yet the degree to which symbiont interactions and epidemics are driven by parasite phenology remains untested.

This experiment utilized the host tall fescue and four co-occurring symbionts to examine how parasite phenology alters (1) parasite epidemics, (2) within-host priority effects among symbionts, and (3) within-host symbiont interactions that determine parasite growth rates. To measure these effects, we quantified parasite incidence and infection severity on three cohorts of sentinel tall fescue plants. The cohorts were distributed across the growing season such that each cohort experienced a different sequence of parasite arrival, with all three parasites arriving sequentially in the first cohort, two parasites arriving at the same time followed by a third in the second cohort, and all three parasites arriving simultaneously in the third cohort. Here, we experimentally measure for the first time how parasite phenology influences epidemics of coinfecting parasites in the field. We then show that the sequence of arrival due to parasite phenology modified both within-host priority effects and within-host interactions that determine parasite growth.

Methods

Study system

This experiment focused on four common fungal symbionts of the host, tall fescue

(Lolium arundinaceum): the parasites Puccinia coronata, Colletotrichum cereale, and

Rhizoctonia solani, and the vertically transmitted endophyte, Epichloë coenophiala (Table 1).

These symbionts and host are of agricultural importance, and many potential mechanisms of

11 within-host fungal interactions have been tested experimentally, leading to specific predictions for our system (Fig. 2.1). Biotrophic fungal parasites commonly facilitate necrotrophs, which inhibit biotrophs via a combination of competition for host resources and induced resistance

(Mundt et al. 1995, Al-Naimi et al. 2005, Spoel et al. 2007, Kliebenstein and Rowe 2008). We therefore expected Puccinia to facilitate Rhizoctonia, and Rhizoctonia to inhibit Puccinia.

Hemibiotrophs initially infect hosts as biotrophs, during which time we expected them to experience antagonism from both biotrophs and necrotrophs, to inhibit other biotrophs, and to facilitate necrotrophs. When they switch to a necrotrophic feeding strategy, we expected them to still experience antagonism from necrotrophs, but to be facilitated by biotrophs, and to inhibit other nectrotrophs and biotrophs. Integrating across this ontogenetic shift, we expected

Colletotrichum to inhibit both Puccinia and Rhizoctonia, Puccinia to have either (indicated by * in Fig. 2.1) a net positive or a net negative effect on Colletotrichum, and Rhizoctonia to have a net negative effect on Colletotrichum. Vertically transmitted fungal endophytes can facilitate or suppress infection by fungal parasites via resource competition and changes in host immunity, which depend on parasite feeding strategies (Potter 1980, Potter 1982, Liu et al. 2006, Saikkonen et al. 2013). Therefore, we expected the endophyte to facilitate biotrophs such as Puccinia, and to inhibit hemibiotrophs and necrotrophs such as Colletotrichum and Rhizoctonia.

Experimental Design

The study was carried out at Widener Farm, an old field in Duke Forest Teaching and

Research Laboratory (Orange County, NC, USA) that produced row crops until 1996. Since

1996, the site has been mowed to produce hay. It is dominated by tall fescue.

12 During the 2013 - 2015 growing seasons, we observed sequential arrival of each parasite species into the host population, with Colletotrichum arriving first, followed by Rhizoctonia, and then Puccinia (Fig. 2.2, Table 2.1, Appendix A1). Specifically, Colletotrichum infected at least

30% of leaves in every survey between 2013 and 2015. Rhizoctonia infections began appearing on leaves in July 2014 and June 2015. Puccinia emerged in September 2013 and 2014 and in

August 2015.

To evaluate the effects of the sequence of arrival of parasites into the host population, we placed three cohorts of uninfected, sentinel outplants into the existing vegetation at different times during the parasites’ natural epidemics during the 2015 growing season. We used surveys of existing plants to determine when an epidemic of one parasite had begun, and transplanted the next cohort of plants shortly after that. The first cohort was placed on 22 June 2015, during the

Colletotrichum epidemic, but before other parasite epidemics began. The second cohort was placed on 27 July 2015, shortly after the start of the Rhizoctonia epidemic, but before the

Puccinia epidemic began. The third cohort was placed on 21 September 2015, shortly after the start of the Puccinia epidemic, when all three parasites were present. Environmental conditions may drive parasite phenology. We therefore tracked infections on each cohort until the first hard freeze on 29 Oct 2015, allowing for the comparison of epidemics among the cohorts at the same time. Because roughly one new leaf emerged per week on each plant, each cohort included leaves of the same age that were exposed to the same environmental conditions during the time when multiple cohorts were in the field.

Each cohort consisted of 40 plants (20 from endophyte-infected seed and 20 from endophyte-free seed) that were propagated from seed in a greenhouse, treated with the systemic insecticide, Marathon (Imidacloprid 1% granular, OHP Inc, Mainland, PA), to prevent insect

13 herbivory, and transplanted into the field by burying each plant in its individual pot in a hole within an approximately 16m2 area that was fenced to exclude vertebrate herbivores. The relative location of individual plants in the field was randomized across all three cohorts, and plants were rearranged weekly to homogenize exposure to fungal parasites. Plants that failed to establish or that resulted from seed contamination by the wrong species were excluded from analyses. This resulted in a total of 30 plants from the first cohort (13 from endophyte-infected seed and 17 from endophyte-free seed), 40 plants from the second cohort (20 from endophyte-infected and 20 from endophyte-free seed), and 36 plants from the third cohort (19 from endophyte-infected and

17 from endophyte-free seed) that were evaluated for symbiont interactions (Additional details in

Appendix A2). All plants were harvested on 29 October 2015.

Survey

All leaves on one focal tiller (ramet) of each plant (genet) were surveyed weekly for infection by foliar parasites. Each leaf was surveyed from emergence to senescence, or until the end of the study. This yielded 303 total leaves in the first cohort, 206 leaves in the second cohort, and 204 leaves in the third cohort. On each leaf, the initial date of symptomatic infection by each parasite was recorded, and the percent of leaf area infected by that parasite (“infection severity”) was estimated by visually comparing leaves to reference images of leaves of known infection severity (Mitchell et al. 2002, 2003).

14 leaves were individually identified based on their vertical order on the tiller, with their age recorded as the days since age 0.

At the conclusion of the experiment, we tested endophyte infection via immunoblot

(Agrinostics Ltd. Co, Watkinsville, GA) and microscopy, but were unable to detect endophyte infection in the experimental plants. We also tested endophyte infection via immunoblot on 100 seeds from the seed lot used to propagate endophyte-infected hosts. 98% of the endophyte- infected and 0% of the endophyte-free seeds tested positive for endophyte infection. Endophyte infection in seeds was also confirmed via microscopy, but tests on plants grown from these seeds confirmed that host seeds were endophyte-infected, but that host plants were not.

Data analysis

Leaves were analyzed as hosts because each parasite infection is restricted to a single leaf. Each model analyzed one cohort of plants, and included one dependent variable pertaining to one parasite species (“the focal parasite”). Each model accounted for both nestedness (leaves nested within host plants) and temporal autocorrelation between surveys caused by seasonal changes in the environment (Appendix A3). We analyzed all data in R version 3.2.3 (R Core

Team 2015).

To evaluate the magnitude of epidemics, we used parasite prevalence. This was calculated as the proportion of host leaves infected by each parasite.

We modeled parasite interactions following the analytical framework described in Fenton et al (2014). Specifically, we modeled focal parasite infection at a given time as a linear function of leaf age, endophyte inoculation, infection by other foliar parasites during the previous survey of that leaf, and the interaction between leaf age and previous infection by other symbionts. We

15 applied this framework to both within-host priority effects and within-host interactions that determine parasite growth.

To evaluate within-host priority effects, we used a Cox-proportional hazards mixed model from the R package, coxme (Therneau 2012), to measure the probability of infection by the focal parasite. The dependent variable in each model was time to infection. This time to infection is modeled as emerging from a baseline rate of infection that is shared by all individuals and modified by a linear combination of predictor variables. Leaves that do not become infected are right-censored, meaning that time to infection is assumed to be greater than the time of observation. Exponentiated coefficients on predictor variables are interpreted as multiplicative changes in infection rate.

To evaluate within-host interactions that determine parasite growth rate, we modeled the growth of each focal parasite as its change in infection severity with respect to leaf age, using the nlme package for linear mixed effects models (Pinheiro et al. 2016). The dependent variable, severity, was log-transformed to increase linearity of the relationship to predictors and homoscedasticity of the residuals. This measure of focal parasite growth encompasses both lesion expansion and new infections within leaves.

For both within-host priority effects and within-host interactions that determine parasite growth, some models contained multiple interactions involving leaf age that were non-significant and had correlated parameter estimates, indicating that the interactions were redundant. To avoid such redundancy, non-significant interactions among fixed-effects were removed from models using likelihood ratio tests (Crawley 2007, Zuur et al. 2009), and overall impacts of symbiont interactions were then determined by evaluating the parameter estimates from the reduced models. For models with interactions between continuous variables (such as interactions between

16 leaf age and previous infection severity), the overall impact of each symbiont was assessed by evaluating model-estimated values of each predictor variable over the range of observed values, weighted by the number of observations of each value. This way, we avoided extrapolating model results into areas where there was no data.

Results

Does parasite phenology alter epidemics?

Parasite sequence of arrival had a profound impact on the epidemics of all three parasites

(Fig. 2.3). These impacts can be seen in terms of the peak prevalence of each parasite.

Colletotrichum prevalence was highest (nearly 70%) in the first cohort, when

Colletotrichum was able to colonize hosts in the absence of other parasites. In the two later cohorts, when at least one other parasite species was present in the field when plants were transplanted, Colletotrichum peak prevalence decreased to less than 20%.

Rhizoctonia prevalence was lowest (less than 30%) in the first cohort, when

Colletotrichum was able to colonize hosts first. In the second cohort, when plants were transplanted into the field after the Rhizoctonia epidemic had begun, Rhizoctonia prevalence increased substantially, to a peak of 85%. Although still higher than in the first cohort, peak prevalence was reduced to 73% in the third cohort, indicating a potential negative effect of

Puccinia on Rhizoctonia when they were able to colonize host individuals at the same time.

Puccinia prevalence was highest in the first cohort, when Colletotrichum was able to colonize hosts first. This pattern differs from Colletotrichum and Rhizoctonia prevalence, which peaked in cohorts when other parasites did not arrive first. Puccinia prevalence was also lowest in the second cohort, when plants were transplanted into the field after the Rhizoctonia epidemic

17 had begun. Thus, rather than responding to arrival of Puccinia into the host population, Puccinia prevalence appeared to respond to the magnitude of the Rhizoctonia epidemic, which in turn was modulated by the sequence of arrival of Rhizoctonia and Colletotrichum into the host population.

Does parasite phenology alter within-host priority effects?

For each cohort, we measured within-host priority effects by evaluating how previous infection of a leaf by other symbionts (parasites and endophyte) influenced the risk of subsequent infection by each focal parasite. Within-host priority effects varied among cohorts (Fig. 2.4a).

This is consistent with the hypothesis that parasite phenology modifies within-host priority effects.

In the first cohort (Fig. 2.5a-c), where hosts were exposed to Colletotrichum before the other parasites, previous infection by Colletotrichum was initially associated with increased relative risk of subsequent infection by Puccinia that switched to become increasingly negative as leaves aged (X2=11.59, df=1, p=0.0007). Previous infection by Puccinia was associated with an increased risk of subsequent infection by Colletotrichum that weakened as leaves aged

(X2=17.50, df=1, p<0.0001), and an increased risk of subsequent infection by Rhizoctonia that weakened as leaves aged (X2=6.76, df=1, p=0.0093). Previous infection by Rhizoctonia was associated with an increasingly negative relative risk of subsequent infection by Colletotrichum as leaves aged (X2=7.09, df=1, p=0.0078). Finally, the endophyte did not influence infection risk of any focal parasite in the first cohort (Table A2.1, A2.4).

In the second cohort, where uninfected hosts were exposed to Rhizoctonia and

Colletotrichum at the same time, no significant relationships were identified between

Rhizoctonia, Colletotrichum, or the endophyte (Table A2.2). Puccinia only established a single

18 infection in the second cohort, and so its interactions with other parasites were not evaluated statistically. However, the lack of infections by Puccinia may be evidence of a strong priority effect by Rhizoctonia, mediated by host mortality. In the second cohort, 94% of Rhizoctonia infections occurred on leaves uninfected by any other parasite, and Rhizoctonia reached its highest severity (up to 80% of leaf area damaged) in those infections. Consequently, 40% of plants in the second cohort died by the time the Puccinia epidemic began, and host mortality nearly doubled in the following two weeks, essentially eliminating the chance of Puccinia establishing an infection in that cohort.

In the third cohort (Fig. 2.5d), where uninfected hosts were exposed to all three parasites simultaneously, previous infection by Rhizoctonia was initially associated with an increased relative risk of subsequent infection by Puccinia that switched to become increasingly negative as leaves aged (X2=14.97, df=1, p=0.0001). Previous infections by Puccinia and Colletotrichum were not associated with any changes in risk of infection by other parasites (Table A2.3, A2.4).

Finally, the endophyte was associated with an increased risk of infection by Puccinia (X2=4.80, df=1, p=0.029).

Does parasite phenology alter within-host interactions that determine parasite growth rates?

For each cohort, we evaluated how previous infection (presence/absence for the endophyte, log+1-transformed infection severity during the previous survey for other parasites) influenced the log-transformed infection severity of each focal parasite in infected leaves and the rate at which infection severity increased as leaves aged (i.e., the focal parasite growth rate following Fenton et al 2014). Within-host interactions that determine parasite growth rates

19 varied among cohorts (Fig. 2.4b, Table A2.5-A2.8). This is consistent with the hypothesis that parasite phenology moderates these interactions.

In the first cohort, (Fig. A2.1), previous infection severity of Colletotrichum was associated with increased Puccinia growth (F1,127=9.22, p<0.01), generating a positive per-capita effect of

Colletotrichum on Puccinia across 78% of infected leaves. Previous infection severity of

Colletotrichum was marginally significantly associated with decreased Rhizoctonia growth

(F1,109=3.08, p=0.05), and previous infection severity of Puccinia and Rhizoctonia were each associated with decreased Colletotrichum growth (F1,355=3.44, p=0.04; and F1,355=9.16, p<0.01, respectively). Together these effects generated negative per-capita effects of Colletotrichum on

Rhizoctonia, and of both Puccinia and Rhizoctonia on Colletotrichum, across more than 99% of infected leaves. Finally, the endophyte had a negative effect on Colletotrichum growth (F1,25=6.4, p=0.02) and a positive effect on Puccinia growth (F1,23=10.58, p<0.001; Table A2.5, A2.8).

In the second cohort, no significant relationships were identified between Rhizoctonia,

Colletotrichum, or the endophyte, and because Puccinia only established a single infection in the second cohort, its interactions with other parasites were not evaluated (Table A2.6, A2.8).

In the third cohort, previous Colletotrichum severity was associated with decreased

Rhizoctonia growth (F1,135=4.64, p=0.03; Table A2.7, A2.8, Fig. A2.2), generating a negative per-capita effect of Colletotrichum on Rhizoctonia across 99% of infected leaves. The endophyte facilitated Puccinia growth (F1,29=8.87, p=0.01; Fig. A2.2), and inhibited Colletotrichum growth

(F1,39=4.97 p=0.03; Fig. A2.2). Rhizoctonia and Puccinia did not significantly affect the growth of other parasites (Table A2.7, A2.8).

20 Discussion

This experiment aimed to evaluate whether parasite phenology alters symbiont interactions and consequently parasite epidemics. The results indicate that parasite phenology can be an important source of historical contingency for parasites, though some caution should be used when interpreting specific relationships between symbionts, owing to a lack of replication over multiple sites and years. Historical contingency has recently been applied to describe any ecological outcome influenced by the order or timing of past events (Fukami 2015).

This can include processes like priority effects, ecological succession, and community assembly.

Here, experimental manipulation of parasite sequence of arrival into the host population modified within-host priority effects, within-host interactions that determine parasite growth rates, and parasite prevalence. Together, these results indicate that historical contingency can profoundly influence parasite epidemics and interactions.

In this system, parasite phenology acted similarly to regional-scale processes in free- living communities. Interactions among individuals define the scale of “local processes”, including both deterministic interactions and priority effects (Ricklefs 1987). Local processes can be strongly influenced by processes at larger spatial scales, termed “regional processes”.

Regional processes can influence the potential for priority effects by altering the sequence of arrival in a local patch (Fukami 2015). Similarly, variation in parasite sequence of arrival into the host population may have influenced within-host priority effects by altering the sequence of infection within leaves. In the first cohort, 74% of all leaves that became infected by any parasite were infected by Colletotrichum first. That number was reduced to 16% and 27% in the second and third cohorts, respectively. Consequently, Colletotrichum exhibited within-host priority effects by preempting other parasites only in the first cohort. Symbionts often arrive into a host

21 population sequentially (Schmidt et al. 2007, Mundt et al. 2009, Dumbrell et al. 2011, Tian et al.

2015), and these results indicate that this can be a regional source of historical contingency that influences local interactions among symbionts.

The three experimental cohorts also profoundly altered patterns of parasite prevalence.

This effect may have arisen from altered susceptibility to secondary infection driven by changes in the sequence of infection within individual hosts. For example, hosts experienced high mortality in the second cohort, where plants were exposed to epidemics of both Rhizoctonia and

Colletotrichum. The high mortality was apparently due to increased colonization of healthy leaves by Rhizoctonia, and precluded infection by other parasites. Similar resource preemption can occur when species with adequate propagule supply are able to rapidly colonize available habitat, and then prevent potential competitors from colonizing (Rohani et al. 2003, Tilman

2004, Limberger and Wickham 2011, Livingston et al. 2012).

Across the three experimental cohorts, and consistent with many laboratory inoculation studies (e.g., Adame-Avarez et al. 2014; Natsopoulou et al. 2015; Klemme et al. 2016), some interactions among symbionts were contingent on the sequence of infection within individual hosts, generating priority effects within those hosts. Furthermore, among those interactions that were contingent on the sequence of infection, the strength or direction of that priority effect was often influenced by leaf age. In other words, within-host priority effects experienced their own contingencies.

This contingency of priority effects on leaf age was largely consistent with mechanisms of interactions among parasites differing in feeding strategy. Biotrophic parasites can facilitate necrotrophs via immune-mediated crosstalk (Spoel et al. 2007, Kliebenstein and Rowe 2008), which occurs when up-regulation of one immune signaling pathway leads to down-regulation of

22 another (Glazebrook 2005, Thaler et al. 2012). However, this crosstalk is temporary and spatially restricted within plants (Spoel et al. 2007, Koornneef et al. 2008). Consequently, facilitation by biotrophs may weaken as leaves age and are exposed to more necrotrophs (Vos et al. 2015). This experiment provides some support for this hypothesis. The facilitative effect of Puccinia, an obligate biotroph, on both Colletotrichum, a hemibiotroph, and Rhizoctonia, a necrotroph, weakened as leaves aged in the first cohort. In contrast, necrotrophic parasites often antagonize biotrophs via resource preemption when they kill host cells (Al-Naimi et al. 2005). This may strengthen over time if necrotroph growth within the host reduces the availability of live host cells that biotrophs can infect. Antagonism by necrotrophs may therefore strengthen as leaves age. Our experiment supports this hypothesis as well. The antagonistic effect of Rhizoctonia on

Colletotrichum increased as leaves aged in the first cohort. Our results are also consistent with effects of leaf age that are relatively independent of parasite feeding strategy. Some interactions may experience a lag time between infection by the first parasite and the biochemical changes that induce resistance to subsequent infections. In these circumstances, priority effects should increase as leaves age (e.g., the effect of Rhizoctonia on Colletotrichum in Cohort 1). Induced resistance can also weaken over time (Underwood 1998, Laine 2011). In these circumstances, priority effects should decrease as leaves age (e.g., the effect of Puccinia on both Rhizoctonia and Colletotrichum in Cohort 1). These results indicate that parasite feeding strategy may be a key factor influencing parasite interactions.

We hypothesized that the endophyte could alter parasite infection and within-host growth by two mechanisms. First, the endophyte may compete with parasites for resources (e.g., Pańka et al. 2013). This mechanism is unlikely to have operated in this study because host leaves were not infected with the endophyte. Host seeds were infected with the endophyte, indicating that the

23 endophyte was unable to leave the seeds and colonize leaf tissue. Nonetheless, the endophyte influenced parasite infection and within host growth. The second hypothesized mechanism is that the endophyte alters inducible host defenses against the parasites (e.g., Saikkonen et al. 2013).

This is hypothesized to occur via two host defense systems, each responding to different parasite feeding strategies, and between which there is cross-talk; together these may allow the endophyte to inhibit Rhizoctonia, facilitate Puccinia, and facilitate infection but inhibit growth of

Colletotrichum. Defense priming (e.g., Conrath et al. 2006), further enhances these responses to parasites, is systemic, and persists long after exposure to symbionts (Pieterse et al. 2014) . Our results are consistent with the effects of the endophyte on the parasites being mediated by these host defense systems.

Finally, we found that parasite phenology can alter the sequence of parasite arrival into a host population. This may subsequently alter the sequence of arrival onto individual hosts.

Within-host priority effects occur when that sequence of arrival onto host individuals influences the probability of coinfection (e.g., Mordecai 2011, Fukami 2015, Fukami et al. 2016).

Consequently, these results demonstrate a mechanism by which parasite phenology can alter within-host priority effects.

Conclusions

Symbiont interactions influenced natural epidemics in multiple ways, often simultaneously. Within hosts, we found evidence of priority effects that prevented coinfection.

When hosts became coinfected, deterministic interactions influenced parasite growth.

Meanwhile, across hosts, we found evidence that epidemics are driven by parasite phenology.

The importance of phenology and priority effects highlight the potential role of historical

24 contingency in epidemics. Moreover, these results demonstrate that symbiont interactions can influence parasite epidemics across scales. These results may advance a more general understanding of how regional processes influence local interactions. Regional processes are rarely measured because they often occur over experimentally and observationally intractable time scales (but see Viana et al. 2016). This experiment supports the growing body of literature suggesting that symbionts represent tractable models for studying ecological processes that operate across scales (e.g., Mihaljevic 2012, Johnson et al. 2015b, Borer et al. 2016,

Penczykowski et al. 2016).

25 Table 2.1 – Biology and ecology of the focal symbionts

Parasite Disease caused Feeding Strategy Transmission Seasonality Colletotrichum anthracnose Hemibiotroph - Mucilaginous Infections occur on cereale initially colonizes spores, most plants and extracts primarily throughout the resources from dispersed by growing season living cells, but rain splash then switches to kill living cells and extract resources from the dead tissue

Puccinia crown rust Obligate biotroph - Windborne A single epidemic coronata can only infect and spores begins around survive on living early September host tissue and increases until the growing season ends

Rhizoctonia brown patch Facultative Hyphal growth A single epidemic solani AG1-1A necrotroph - can and starts around July survive in the soil fragmentation, and tapers off as as a saprobe, and not spores. temperatures cool when it infects in the fall plants, it kills living cells and extracts resources from the dead tissue.

Epichloë none Intercellular Vertical Individual hosts do coenophiala endophyte - transmission not gain or lose restricted to living via seedborne infection plant tissue and mycelium systemic through aboveground tissues.

Vertically Transmitted Fungal Endophyte (e.g., Epichloë coenophiala)

Biotrophic Fungal * Hemibiotrophic Fungal Necrotrophic Fungal Parasite Parasite Parasite (e.g., Puccinia (e.g., Colletotrichum (e.g., Rhizoctonia solani) coronata) cereale)

27 Figure 2.2 - Parasite sequence of arrival into the host population during 2013 – 2015 surveys. Points represent the average first date that at least 1% of host leaves in 2013 and 2014 and 1% of plots in 2015 were infected by each parasite. Error bars represent the earliest and latest date of first infection across the surveys.

Puccinia

Rhizoctonia ●

Colletotrichum ● ●

Apr Jun Aug Oct Arrival date

28 Figure 2.3 - Seasonal epidemics in three experimental cohorts of hosts. Points represent parasite prevalence (the proportion of leaves infected across all sentinel hosts) in each survey. Illustrative lines are LOESS fit to the data with span=0.9. Vertical lines represent the dates that each cohort was placed into the field, highlighting the overlap in time among cohorts.

Cohort 1 ● Colletotrichum 1.0 Rhizoctonia Puccinia ● ● ● ● ● 0.5 ● ● ● ● ● ● ● ● ●

0.0 ● Cohort 2 1.0

0.5 ● ● ● ● ● ● 0.0 ● ● ● ● ● Cohort 3

Proportion of leaves infected infected Proportion of leaves 1.0

0.5 ● ● ● 0.0 ● ● Jul Aug Sep Oct Nov

Date

29 Figure 2.4 – Summary of symbiont interaction network indicated by the interaction models. Blue arrows represent positive interactions (e.g., facilitation). Red clubs represent negative interactions (e.g., inhibition). Dashed lines represent hypothesized interactions that were not supported by the models (p>0.05). “E” stands for Epichloë coenophiala, “P” stands for Puccinia coronata, “C” stands for Colletotrichum cereale, “R” stands for Rhizoctonia solani. a) Within- host priority effects from Cox-proportional hazards models. b) Interactions influencing within- host growth from longitudinal linear mixed models.

Within-host priority effects a) E E E

C R P C R P C R

Cohort 1 Cohort 2 Cohort 3 Interactions determining within-host growth b) E E E

C R P C R P C R

Cohort 1 Cohort 2 Cohort 3

30 Figure 2.5 – Model-estimated relative risk of infection as leaves age. Plots are results of the reduced Cox mixed models. A value above zero indicates that previous infection of a leaf by Colletotrichum (red), Puccinia (blue), Rhizoctonia (green), or endophyte-infected seed (black) increased the risk of subsequent infection by the focal parasite. A value below zero indicates that previous infection decreased the risk of subsequent infection by the focal parasite. Vertical lines along the x-axis show the age of each leaf when it became infected by the focal parasite, colored by the infection status of that leaf by other parasites. X-axis values are jittered to show the data. a) Cohort 1 Colletotrichum infection risk. b) Cohort 1 Puccinia infection risk. c) Cohort 1 Rhizoctonia infection risk. d) Cohort 3 Puccinia infection risk. These model results are summarized in Figure 2.4a.

Cohort 1: Colletotrichum Infection Risk Cohort 1: Puccinia Infection Risk A) B) 4

−2 Relative Risk (log) Relative −4

0 20 40 0 20 40 60 Leaf age (days) Leaf age (days)

Cohort 1: Rhizoctonia Infection Risk Cohort 3: Puccinia Infection Risk C) D) 4

−2 Relative Risk (log) Relative −4

0 20 40 60 0 10 20 30 40 Leaf age (days) Leaf age (days)

None Colletotrichum infected Rhizoctonia infected Colletotrichm and Rhizoctonia Endophyte Puccinia infected Colletotrichum and Puccinia Puccinia and Rhizoctonia

CHAPTER 3 : HOST IMMUNITY MODIFIES INTERACTIONS AMONG PARASITES AND ALTERS PARASITE EPIDEMICS

Introduction

Interactions among parasites and pathogens (hereafter, “parasites”) may alter parasite epidemics and host health (Rynkiewicz et al. 2015, Tollenaere et al. 2015). Some of these interactions depend on the sequence in which parasites infect host individuals, generating priority effects among co-occurring parasites (Hoverman et al. 2013, Halliday et al. 2017b). This contingency of interactions may result from host immune responses to parasite infection (Lello et al. 2004, Tollenaere et al. 2015). Yet the indirect consequence of host immunity on epidemics of interacting parasites remains unmeasured under field conditions. This study experimentally tests whether host immune signaling pathways alter interactions among parasites, and measures the consequences for parasite epidemics in the field.

Interactions among parasites may alter parasite epidemics. When propagules disperse to and establish in a local patch, their interactions with the resident community may structure species composition. This process is a cornerstone of community ecology (MacArthur 1958,

Chesson 2000, HilleRisLambers et al. 2011) and has gained renewed interest for understanding microbes within hosts (Costello et al. 2012, Fierer et al. 2012, Cobey and Lipsitch 2013).

Similarly, parasites that are able to establish in a given host are often subjected to interactions with the resident community during simultaneous infections, known as coinfections (Griffiths et al. 2014, Tollenaere et al. 2015, Ezenwa 2016). Coinfecting parasites may interact for limiting resources or can interact indirectly via their impacts on host physiology, including host immune

responses (Lello et al. 2004, Mideo 2009, Chung et al. 2012). These interactions, which can result in both inhibition and facilitation among coinfecting parasites, can also alter parasite epidemics (Eswarappa et al. 2012, Tollenaere et al. 2015).

Some interactions among parasites may result from priority effects. Much like free-living

(i.e., non-parasitic) organisms, the strength and direction of interactions among some parasites is the same regardless of the historical context in which they occur (e.g., Sousa 1993). However, interactions among coinfecting parasites may also be contingent on the sequence of past events

(i.e., historically contingent, following Fukami 2015), generating priority effects within hosts

(e.g., Hoverman et al. 2013). Within hosts, priority effects occur when the per-capita strength of inhibition or facilitation among parasites is altered by their sequence of arrival (Hoverman et al.

2013; Mordecai et al. 2016). Yet, ecologists still lack a general framework for predicting when such contingency in species interactions will occur (Vannette and Fukami 2014).

For parasites, this contingency of interactions may result from the host immune response to infection (Lello et al. 2004, Graham 2008, Cobey and Lipsitch 2013). Specifically, early arriving parasites often activate immune responses within hosts, which may then alter host susceptibility to later arriving parasites. These host immune mediated interactions largely fall into two categories: (1) immune-mediated cross protection and (2) immune-mediated crosstalk.

Immune-mediated cross protection (also referred to as induced resistance or cross- immunity) occurs when the immune response to infection by one parasite confers immunity to another (Jenner 1923, Fulton 1986, Pieterse et al. 2014), resulting in a decrease in the frequency of coinfection (Biere and Goverse 2016). This is most commonly reported among closely related parasite species (e.g., Fulton 1986, Adams et al. 1989), but can also occur when parasites respond to similar immune-signaling pathways (e.g., Van Loon 1997). Immune-mediated

crosstalk occurs when up-regulation of one immune signaling pathway leads to down-regulation of another (Thaler et al. 2012), facilitating subsequent infection, and consequently increasing the frequency of coinfection (Spoel et al. 2007, Ezenwa et al. 2010). This is most commonly reported among parasites that exhibit different feeding strategies, or that elicit and respond to distinct immune signaling pathways (Glazebrook 2005, Thaler et al. 2012, Ezenwa 2016). Both mechanisms of immune-mediated interactions among parasites have been reported in plant and animal hosts (Glazebrook 2005, Ezenwa et al. 2010, Pieterse et al. 2014).

This study focuses on the salicylic acid (SA) and jasmonic acid (JA) immune-signaling pathways of plants as potential mediators of within-host parasite interactions. Plants and their associated parasites are a useful model system for studying immune-mediated interactions among parasites, because the parasites can be quickly and reliably identified using visual surveys and are amenable to small-scale manipulations that can be difficult or unethical to accomplish with animal hosts (Antonovics et al. 2002, Power and Mitchell 2004, Roy et al. 2014).

Furthermore, the SA and JA pathways can be experimentally manipulated using direct application of plant-signaling hormones (Traw and Bergelson 2003, Cipollini et al. 2004, Sutter and Müller 2011).

In plant hosts, immune-mediated interactions are thought to chiefly occur through the SA and JA pathways (Kliebenstein and Rowe 2008, Vlot 2009). These interactions may depend on parasite feeding strategies (Glazebrook 2005, Spoel et al. 2007). Plant parasite feeding strategies occupy a continuum, from biotrophic parasites, which feed and reproduce in living host tissue to necrotrophic parasites, which kill living cells and extract resources from the dead tissue

(Mendgen and Hahn 2002, van Kan 2006). The SA pathway is expected to confer resistance to biotrophic parasites, while the JA pathway is expected to confer resistance against necrotrophic

parasites and insect herbivores (Glazebrook 2005, Spoel et al. 2007, Vlot et al. 2009).

Consequently, parasites that share the same feeding strategy (e.g., are both biotrophs) may suffer from immune-mediated cross protection, while parasites that differ in feeding strategies may benefit from immune-mediated crosstalk between the two pathways (Glazebrook 2005, Thaler et al. 2012, Vos et al. 2015) - a relationship that bears some analogy to the well-known mutual inhibition between the Th1 and Th2 immune responses in vertebrates (Graham 2008, Ezenwa et al. 2010). In addition to downregulating JA, SA signaling activates defense genes linked to host cell death and systemic acquired resistance (Vlot et al. 2009), thereby further facilitating infections by necrotrophic parasites (Glazebrook 2005).

Immune-mediated interactions may alter parasite interactions and epidemics (Rohani et al. 2003, Lello et al. 2004), but experimentally manipulating host immunity and measuring the consequences for parasites remains challenging outside of the lab (Zhan and McDonald 2013,

Pedersen and Fenton 2015). This experiment overcomes this limitation by manipulating host immunity and then measuring the consequences in shaping within-host interactions, and ultimately parasite epidemics in the field. Using the host plant, tall fescue, and two co-occurring foliar parasites, we show that host immunity can alter within-host priority effects and within-host parasite growth among infected hosts, resulting in shifts in parasite prevalence, the frequency of coinfection, and the severity of disease experienced by hosts.

Methods

This experiment was carried out at Widener Farm, an old field in Duke Forest Teaching and Research Laboratory (Orange County, NC, USA) that produced row crops until 1996. Since

1996, the site has been mowed to produce hay. It is dominated by the host, tall fescue (Lolium

arundinaceum). The study focused on two common fungal parasites of tall fescue:

Colletotrichum cereale and Rhizoctonia solani AG1-1A.

Colletotrichum is the cause of anthracnose of cool-season grasses. It is a hemibiotrophic parasite, meaning that it initially infects its host and extracts resources from living cells (a biotrophic feeding strategy), but then it switches its mode of parasitism to kill living cells and extract resources from the dead tissue (a necrotrophic feeding strategy). It is transmitted by mucilaginous spores that are dispersed primarily by rain splash. In this system, Colletotrichum prevalence and disease severity are relatively stable throughout the growing season (Halliday et al. 2017b). We therefore consider Colletotrichum to be an endemic fungal parasite of tall fescue.

Rhizoctonia is the cause of many diseases, including brown patch of tall fescue. It is a facultative necrotrophic parasite, meaning that it can survive in the soil as a saprobe, and when it infects plants, it kills living cells and extracts resources from the dead tissue. It is transmitted almost exclusively by hyphae (growth and fragmentation), not spores. In this system,

Rhizoctonia is best characterized by a single epidemic, beginning between late June and early

July, and peaking in mid to late September, after which prevalence and severity decline (Halliday et al. 2017b).

Many potential mechanisms of within-host interactions among fungal parasites have been tested experimentally, leading to specific predictions based on parasite feeding strategies

(Halliday et al. 2017b). Specifically, biotrophs often facilitate necrotrophs via immune-mediated crosstalk, while necrotrophs inhibit biotrophs via competition for host resources (Mundt et al.

1995, Al-Naimi et al. 2005, Spoel et al. 2007, Kliebenstein and Rowe 2008). Colletotrichum, a hemibiotroph, initially infects hosts as a biotroph. During this biotrophic phase, we expected

Rhizoctonia to antagonize Colletotrichum via competition for resources, and Colletotrichum to

facilitate Rhizoctonia via immune-mediated crosstalk. When Colletotrichum switches to a necrotrophic feeding strategy, we expected Rhizoctonia to still antagonize Colletotrichum, but

Colletotrichum to also antagonize Rhizoctonia via a combination of competition for resources and cross resistance.

Experimental design

To evaluate the effects of immune-mediated interactions on parasite epidemics, we experimentally manipulated the immune-signaling pathway on individual sentinel outplants, which were placed into the existing vegetation on 21 September 2015, at the peak of the

Rhizoctonia epidemic.

Following Schweiger et al (2014), each plant was randomly assigned to one of three treatments: The Jasmonic Acid treatment (JA) received an aqueous solution of Jasmonic Acid (J-

2500, Sigma-Aldrich, St. Louis) diluted to a .5mM solution. The Salicylic Acid treatment (SA) received an aqueous solution of Salicylic Acid (S-7401, Sigma-Aldrich) diluted to .5mM, and the control treatment (control) received distilled water, with half of those plants receiving distilled water adjusted to pH 3.1. These hormone concentrations were similar to those used in other studies (Traw and Bergelson 2003, Cipollini et al. 2004, Sutter and Müller 2011). Once per month, 1mL of solution was applied evenly across the surface of each plant using a handheld atomizer.

Each treatment consisted of 20 plants that were propagated from endophyte-free seed in a greenhouse for 33 days, then treated with the systemic insecticide, Marathon (Imidacloprid 1% granular, OHP Inc, Mainland, PA), to prevent insect herbivory, and transplanted into the field by burying each plant in its individual pot in a hole within an approximately 16m2 area that was

fenced to exclude vertebrate herbivores. The relative location of individual plants in the field was randomized, and plants were rearranged weekly to homogenize exposure to fungal parasites.

Plants that failed to establish or that resulted from seed contamination by the wrong species were excluded from analyses. This resulted in a total of 19 plants in the SA treatment group, 18 plants in the JA group, and 17 plants in the control group that were evaluated for symbiont interactions.

All plants were harvested on 29 October 2015.

Survey

All leaves on one focal tiller (ramet) of each plant (genet) were surveyed longitudinally for infection by foliar parasites (Following Halliday et al. 2017b). Each leaf was surveyed weekly from emergence to senescence, or until the end of the study. This yielded 107 total leaves in the SA group, 102 leaves in the JA group, and 98 leaves in the control group. On each leaf, the initial date of symptomatic infection by each parasite was recorded, and the percent of leaf area infected by that parasite (“infection severity”) was estimated by visually comparing leaves to reference images of leaves of known infection severity (Mitchell et al. 2002, 2003).

Leaf age was used as a proxy for exposure to parasite propagules. When plants were transplanted into the field and initially surveyed, pre-existing leaves were assigned age 0. Each subsequent survey, newly emerged leaves were recorded as age 0, and previously surveyed leaves were individually identified based on their vertical order on the tiller, with their age recorded as the days since age 0.

To evaluate the effects of experimental treatments on parasite prevalence, all leaves of each plant (genet), including leaves of the focal tiller, were surveyed for infection by foliar parasites at the conclusion of the experiment. This yielded 503 total leaves in the SA group, 536

leaves in the JA group, and 434 leaves in the control group. On each plant, the total number of leaves and the number of leaves infected by each parasite were recorded.

Disease metrics

To evaluate the magnitude of parasite epidemics, we used parasite prevalence, coinfection, and leaf burden. Parasite prevalence was calculated as the proportion of host leaves infected by each parasite across an entire host plant. Coinfection was calculated on each longitudinally surveyed leaf and coded one if that leaf became coinfected by both parasites, and zero if that leaf was never coinfected.

Leaf burden was also calculated individually on each longitudinally surveyed leaf as the area under the disease progress stairs using the agricolae package (de Mendiburu and de

Mendiburu 2016). Leaf burden includes the period of time that the leaf is uninfected, integrating the development of disease progress experienced by each leaf over the course of the experiment

(Madden et al. 2007). Consequently, area under the disease progress stairs provides a single value of disease burden that includes the effects of all within-host interactions, including priority effects. We used area under the disease progress stairs instead of the more traditional area under the disease progress curve, as it improves estimates of the first and last observations (Simko and

Piepho 2012).

Data analysis

Data analysis was performed using R version 3.2.3 (R Core Team 2015). Leaves were analyzed as hosts because each parasite infection is restricted to a single leaf.

To analyze the magnitude of epidemics, we used parasite prevalence, leaf burden, and coinfection as dependent variables and parasite species, experimental treatment, and their interaction as independent variables. Parasite prevalence was modeled as a grouped binomial response, using a logit link, with leaves grouped by host ID, using the lme4 package for generalized linear mixed effects models (Bates et al. 2014). Leaf burden was modeled as a linear combination of the independent variables, with leaves nested in host plants using the nlme package for linear mixed effects models (Pinheiro et al. 2016). The probability of coinfection was modeled as a binomial response, using a logit link, with leaves nested in host plants, using the lme4 package.

We assessed the treatment effects with the two control treatments separated (four levels: acid control, water control, JA, SA) and with the two control treatments combined (three levels: control, JA, SA) on the measures of parasite epidemics, separated by each parasite species. These models differed minimally, and generally favored the model with the control treatments grouped

(Colletotrichum burden ∆AICc = 2.43; Rhizoctonia burden ∆AICc = -1.95; Colletotrichum prevalence ∆AICc = 1.68, Rhizoctonia prevalence ∆AICc = 2.37, Probability of coinfection

∆AICc = 2.04). We therefore grouped the two control treatments into a single variable for all subsequent analyses to facilitate comparisons among the treatments of interest (JA vs control, SA vs control).

To model within-host interactions, we constructed a series of models following Halliday et al (2017b). Each model included one dependent variable pertaining to one parasite species

(“the focal parasite”) and accounted for both nestedness (leaves nested within host plants) and temporal autocorrelation between surveys.

To evaluate within-host priority effects, we used a Cox-proportional hazards mixed model from the R package, coxme (Therneau 2012), to measure the probability of infection by the focal parasite. The dependent variable in each model was time to infection. This time to infection was modeled as emerging from a baseline rate of infection shared by all individuals and modified by a linear combination of the experimental treatment, previous infection by the other parasite, and their interaction. A third parasite, Puccinia coronata, commonly co-occurs with

Colletotrichum and Rhizoctonia in this system, but is not considered as a focal parasite in this study. However, because Puccinia can generate within-host priority effects (Halliday et al.

2017b), previous infection by Puccinia was included as a fixed effect in each model. Leaves that did not become infected were right-censored, meaning that time to infection is assumed to be greater than the time of observation. Exponentiated coefficients on predictor variables are interpreted as multiplicative changes in infection rate.

To evaluate within-host parasite growth, we modeled the growth of each focal parasite as its change in infection severity with respect to leaf age, using the nlme package. The dependent variable, severity, was log-transformed to increase linearity of the relationship to predictors and homoscedasticity of the residuals. This measure of focal parasite growth encompasses both lesion expansion and new infections within leaves. The independent variables were leaf age, the experimental treatment, log-plus-one transformed infection by the other foliar parasite during the previous survey of that leaf, and their interactions. As in models of within-host priority effects, previous infection by Puccinia was included as a fixed effect in each model. Models that contained multiple interactions involving leaf age that were non-significant and had correlated parameter estimates had those redundant interactions removed using likelihood ratio tests

(following Halliday et al. 2017b).

Pairwise comparisons of the fixed coefficients were performed using the lsmeans package (Lenth 2013). Non-significant interactions make pairwise comparisons of the fixed coefficients difficult to interpret, and were therefore removed using likelihood ratio tests

(following Zuur et al. 2009) before performing pairwise comparison tests. To limit the number of comparisons, only pairwise comparisons of ecological relevance were conducted. Specifically, for models evaluating the magnitude of epidemics, pairwise comparisons were made separately for each parasite species and limited to comparisons between each treatment and the control group using the pairs function. For models of within-host interactions, two sets of pairwise comparisons were conducted: differences between each treatment and the control were assessed in leaves that were not previously infected by other parasites using the contrast function, and the effect of previous infection by other parasites was assessed separately for each treatment group using the pairs function.

Results

Does host immunity alter within-host priority effects?

Neither SA nor JA altered the risk of a healthy leaf becoming infected by Rhizoctonia (p

= 0.89 and 0.62 respectively; Fig 3.1a; Table B3.1A, Table B3.2A). However, SA did reduce the risk of Rhizoctonia infection via its interaction with previous Colletotrichum infection (p =

0.048). Specifically, Colletotrichum exhibited a priority effect over Rhizoctonia, reducing host risk of infection by Rhizoctonia by 94%, but only in SA-treated hosts (p = 0.033). This result is counter to the expectation that SA would strengthen the facilitative effect of hemibiotrophic

Colletotrichum on necrotrophic Rhizoctonia.

This within-host priority effect of Colletotrichum on Rhizoctonia among SA-treated leaves was at least partially offset by the marginally significant effect of the experimental treatment on Colletotrichum infection risk (p = 0.050; Fig. 3.1b; Table B3.1B, Table B3.2B).

Specifically, as expected, SA treatment reduced the Colletotrichum infection risk by 67% (p =

0.057). In addition to SA, previous infection by Rhizoctonia also reduced Colletotrichum infection risk by 60% (p = 0.01), though there was no interaction between experimental treatments and this priority effect of Rhizoctonia over Colletotrichum (p = 0.78). This result is consistent with competition for within-host resources driving the priority effect of necrotrophic

Rhizoctonia over hemibiotrophic Colletotrichum (e.g., Al-Naimi et al. 2005).

Do immune-mediated priority effects alter parasite prevalence?

To evaluate the outcome of immune-mediated priority effects on parasite epidemics, we analyzed effects on parasite prevalence. SA did not significantly influence Rhizoctonia prevalence (p = 0.29; Table B3.3, Fig. 3.2a), contrasting with the negative effect of prior

Colletotrichum infection on Rhizoctonia infection risk in SA hosts. Consistent with the effects of

SA on Colletotrichum infection risk, SA application reduced Colletotrichum prevalence by 51% relative to control (p = 0.030, Fig. 3.2b). Thus even though previous Colletotrichum infection reduced Rhizoctonia infection risk among SA hosts, the effect of SA on Colletotrichum risk appears to have offset this effect, resulting in no net effect of SA on Rhizoctonia prevalence.

Does host immunity alter within-host parasite growth?

Because host immune-signaling pathways may influence lesion expansion or within-host replication, we also tested whether experimental treatment of SA and JA altered within-host

parasite growth. Whereas neither experimental treatment altered Rhizoctonia infection risk, SA increased within-host growth of Rhizoctonia (p < 0.001; Fig. 3.3, Table B3.1A, B3.4A).

However, this effect was reduced in leaves that were coinfected with Colletotrichum (p = 0.048), indicating a potentially negative effect of Colletotrichum on within-host Rhizoctonia growth, moderated by SA.

In addition to the negative effect of SA on Colletotrichum infection risk, SA also reduced

Colletotrichum growth (p = 0.048; Table B3.1B, Table B3.4B). This resulted in lower severity of infection by Colletotrichum in SA-treated host, limiting the overall antagonistic effect of

Colletotrichum on Rhizoctonia in SA-treated hosts (Fig. 3.3).

What are the consequences for immune-mediated parasite interactions on parasite epidemics?

To evaluate the outcome of immune-mediated within-host priority effects and within-host parasite growth for parasite epidemics, we measured whether or not a leaf became coinfected and leaf burden by each parasite (calculated as the area under the disease progress stairs).

Coinfection frequency captures the outcome of interactions among parasites, including competitive exclusion, facilitation, and priority effects among parasites (Sousa 1993, Mordecai et al. 2016, Halliday et al. 2017b), while leaf burden integrates over all interactions that take place on a leaf during the course of the experiment. The experimental treatment significantly influenced the probability of coinfection and parasite burden (p = 0.016 and p < 0.001 respectively; Table B3.5, B3.6). Specifically, SA increased Rhizoctonia burden by 44% compared to control (p = 0.002; Fig 3.2c). SA also reduced the probability of a leaf becoming coinfected by 72% compared to control (p = 0.032; Fig 3.2e). Together, these results indicate

that by altering within-host interactions among parasites, SA can dramatically alter parasite epidemics.

Discussion

This study advances the hypothesis that host immunity may represent a general mechanism of within-host priority effects. Vannete & Fukami (2014) posited that priority effects are most likely to occur when species exhibit high niche overlap. This may occur when species require similar resources, share natural enemies, or for parasites, respond to similar host immune processes. Additionally, priority effects should be more common when the early arriving species have large impacts on that niche and when the late arriving species are highly sensitive to the availability of that niche (Vannette and Fukami 2014). These requirements may be commonly fulfilled for parasites sharing a host, and driven by host immunity. Parasites, by definition, require host resources for survival, growth, and reproduction (Lafferty and Kuris 2002), and may therefore exhibit some niche overlap and high sensitivity to the availability of that niche when they coinfect the same host individual. Early arriving parasites can strongly influence the success of later arriving parasites by altering host immunity (Lello et al. 2004, Graham 2008, Cobey and

Lipsitch 2013), and the impacts on host fitness and immunity may be related to the parasite’s virulence (Newton et al. 2010). Such immune-mediated interactions and priority effects are well documented using observational studies and mathematical models (Rohani et al. 2003, Lello et al. 2004). This study advances this hypothesis by experimentally measuring the consequences of immune-mediated priority effects on parasite epidemics in the field. Specifically, immune- mediated interactions altered parasite prevalence, coinfection frequency, and parasite burdens of both an endemic and an epidemic parasite.

Host-mediated interactions may depend on parasite feeding strategies. Specifically, immune-mediated cross protection, which reduces the frequency of coinfection, is expected to occur more commonly among parasites with similar feeding strategies, whereas immune- mediated crosstalk, which increases the frequency of coinfection, may be more common among parasites with different feeding strategies (Glazebrook 2005, Spoel et al. 2007, Pieterse et al.

2012, Biere and Goverse 2016). This dependence of host-mediated interactions on parasite feeding strategies may explain the effect of SA on interactions among Colletotrichum, a hemibiotroph, and Rhizoctonia, a necrotroph. Host immunity altered the effect of previous

Colletotrichum infection on Rhizoctonia. However, this interaction did not occur in the direction that we hypothesized. We hypothesized that Colletotrichum would facilitate Rhizoctonia via immune-mediated crosstalk, by down-regulating the JA pathway, and that experimental application of SA would strengthen the facilitative effect of Colletotrichum on Rhizoctonia.

However, we instead observed evidence of immune-mediated cross protection: Colletotrichum inhibited Rhizoctonia, but only among SA-treated hosts. Because cross protection is expected more commonly among parasites with similar feeding strategies (Van Loon 1997), and

Colletotrichum is only expected to antagonize Rhizoctonia during the necrotrophic phase of growth (Halliday et al. 2017b), this result may indicate that SA hastened the switch in

Colletotrichum from biotrophy to necrotrophy.

Plant hosts face many tradeoffs when responding to natural enemies, including parasites

(Stamp 2003). One of the most well-studied tradeoffs is the growth-defense tradeoff (Herms and

Mattson 1992). A growth-defense tradeoff occurs when resource allocation to defense comes at a cost to continued growth (Herms and Mattson 1992, Lind et al. 2013). This growth-defense tradeoff may be influenced by host immunity (van Hulten et al. 2006, Todesco et al. 2010,

Karasov et al. 2017, Douma et al. 2017). The SA pathway activates pathogen resistance genes, leading to host cell death and systemic acquired resistance to biotrophic parasites (Vos et al.

2015). However, this resistance comes at a cost: maintaining and activating these pathogen resistance genes can have negative consequences on plant performance in the absence infection by biotrophs (Todesco et al. 2010). This may result in selection against activation of the SA pathway under many ecological conditions. Consequently, tradeoffs in hosts’ allocation of resources may underlie interactions among parasites that infect those hosts.

Immune-mediated tradeoffs likely depend on exposure to parasites of specific feeding strategies. Activation of the SA pathway leads to cell death and to deactivation of the JA pathway, potentially increasing host susceptibility to necrotrophic parasites via immune- mediated crosstalk. This may have fitness consequences for hosts if parasite feeding strategy is related to parasite impacts. The results of this study are consistent with this hypothesis. SA reduced the prevalence of Colletotrichum, an endemic parasite that can reduce Rhizoctonia prevalence via priority effects (Halliday et al. 2017b). Consequently, SA can eliminate the protective effect of Colletotrichum over its host, resulting in increased burdens by the epidemic, necrotrophic parasite, Rhizoctonia.

As a consequence of immune-mediated tradeoffs, some hosts are primed to respond more strongly to JA (Conrath et al. 2006, Mauch-Mani et al. 2017). Defense priming is systemic, can persist long after exposure to microbial symbionts (Pieterse et al. 2014), and can be passed from one generation to the next (Mauch-Mani et al. 2017). Our results indicate that hosts may be similarly primed for a JA response in this system. Experimental application of JA had no effect across all responses measured, while experimental application of SA had negative consequences for the host, increasing burdens of the most damaging parasite species. Together, these results

underscore the complexity involved in understanding interactions and epidemics of co-occurring parasites. Epidemics of co-occurring parasites may be influenced by their interactions (Halliday et al. 2017b). These interactions may result from host immunity (Lello et al. 2004, Ezenwa

2016), which changes depending on parasite feeding strategy(Glazebrook 2005), and is limited by host resource requirements for growth and reproduction (van Hulten et al. 2006, Douma et al.

2017).

Predicting the magnitude of historical contingency in species interactions remains a challenge for ecologists (Vannette and Fukami 2014). Understanding the mechanisms that generate priority effects may therefore advance a more general understanding of interactions among species. This study indicates that for plant parasites, host immunity may underlie within- host priority effects, altering parasite epidemics. This study therefore supports the growing body of literature suggesting that parasites represent tractable models for studying ecological processes (e.g., Johnson et al. 2015c, Borer et al. 2016, Penczykowski et al. 2016, Halliday et al.

2017b).

Figure 3.1 – Model-estimated relative risk of infection. Plots are results of the reduced Cox mixed models, and are on a logarithmic scale. Points represent the treatment mean and error bars represent the 95% confidence interval. A value above zero indicates that the experimental treatment, or previous infection of a leaf by the other parasite (red) increased the risk of subsequent infection by the focal parasite. A value below zero indicates that the treatment or previous infection decreased the risk of subsequent infection by the focal parasite. a) Rhizoctonia infection risk. b) Colletotrichum infection risk.

− Colletotrichum free Colletotrichum infected 2 − − − − − 0 −

−2 Rhizoctonia risk Rhizoctonia (log) −

−4 A)

control JA SA

− Rhizoctonia free 1 − Rhizoctonia infected − − 0 − − − −1

Colletotrichum risk (log) − −2

control JA SA

Rhizoctonia Colletotrichum

0.6 − − − 0.2 − 0.4 − − Control 0.1 0.2 − JA −

Infection prevalence prevalence Infection − SA A) B) (backtransformed from logit) (backtransformed 0.0 0.0

Rhizoctonia Colletotrichum 0.04 0.008

− 0.006 0.03 − 0.004 − − 0.02 − 0.002 0.000 − 0.01 Leaf burden (log) Leaf burden −0.002 0.00 C) D)

0.3 − 0.2 −

0.1

Coinfection frequency Coinfection − E) (backtransformed from logit) (backtransformed 0.0 Control JA SA

Figure 3.3 – Model-estimated within-host parasite growth. Plots are results of the reduced longitudinal mixed models, showing the rate at which log-transformed infection severity by Rhizoctonia increased as leaves age (i.e., parasite growth rate), and the effects of previous Colletotrichum infection on that relationship. Colors and contour lines represent model-estimated Rhizoctonia infection severity. Points represent individual observations of leaves over the course of the experiment. Control Jasmonic acid Salicylic acid

4 Focal Parasite 4 4 Rhizoctonia severity 100

0.1

2 2 2 Previous Colletotrichum severity Colletotrichum severity Previous

0 0 0

0 10 20 30 40 0 10 20 30 0 10 20 30 40 Leaf age (days) Leaf age (days) Leaf age (days)

CHAPTER 4 : A MULTIVARIATE TEST OF DISEASE RISK REVEALS CONDITIONS LEADING TO DISEASE AMPLIFICATION

Introduction

Many of the impacts of pathogens and parasites (hereafter, “parasites”) on human, wildlife, and ecosystem health have been attributed to shifts in the diversity and abundance of parasites (Dobson et al. 2008, Hersh et al. 2012, Johnson and Hoverman 2012). Parasite diversity and abundance are often highly variable (Torchin et al. 2015), and understanding their ecological drivers has therefore become a central goal of disease ecology (Johnson et al. 2015a). Two elements of host communities that commonly influence the diversity and abundance of parasites are host diversity and resource supply to hosts (Mitchell et al. 2003, Liu et al. 2016). These relationships may be influenced by characteristics of host and parasite species (Borer et al.

2016). Moreover, within host communities, the total number of parasite species and the abundance of parasites per host may be positively correlated (Watve and Sukumar 1995).

Comparing shifts in parasite richness and abundance may also reveal ecological processes connecting host and parasite communities. Shifts in parasite richness may be attributable to changes in parasite colonization among communities, whereas shifts in parasite abundance may be attributable to changes in among-host transmission within communities (Johnson et al.

2015a). Yet, because these characteristics of host and parasite communities are tightly connected, understanding parasite richness and abundance requires careful evaluation of complex, non-independent relationships among hosts and parasites.

Resource supply to hosts may influence parasite richness and abundance. In communities of free-living plants and animals (i.e., non-parasitic organisms or life stages), increasing resource supply by fertilizing intact communities often reduces species richness across trophic levels

(Rosenzweig 1971, Borer et al. 2014b). If increasing resource supply to hosts increases the availability of nutrients to the parasites that feed on those hosts, then this could reduce parasite richness within hosts by a similar mechanism (e.g., removing resource limitation). Similarly, if resource supply increases host biomass or host tissue becomes more nutritious (Veresoglou et al.

2013), then this could lead to an increase in parasite abundance.

Host diversity may also influence parasite richness and abundance. Parasite richness in plant and animal hosts often increases with host richness, because higher host richness represents a more diverse pool of resources for parasites to colonize (Rottstock et al. 2014, Kamiya et al.

2014, Johnson et al. 2016). Unlike parasite richness, parasite abundance may respond positively or negatively to increasing host diversity. First, increasing host diversity may decrease the abundance of one or more parasite species by reducing transmission within communities, representing a dilution effect (Mitchell et al. 2002, e.g. LoGiudice et al. 2003). Second, increasing host diversity may increase parasite abundance representing an amplification effect

(Power and Mitchell 2004, Young et al. 2013). Empirical evidence exists for both dilution and amplification, though dilution effects have been observed more frequently (Civitello et al. 2015).

The influence of host diversity on parasite abundance may depend on characteristics of host species, and thus the composition of host communities. For example, a dilution effect is often expected when the most competent hosts persist as host species richness declines. Thus, increasing richness decreases the relative abundance of the most competent hosts, reducing their effect on disease transmission (Keesing et al. 2006). Alternatively, an amplification effect may

occur when the most competent hosts do not persist as host species richness declines, though this pattern lacks empirical support (Civitello et al. 2015).

Amplification may also result when changes in composition are random with respect to species richness (Joseph et al. 2013). This explanation for the amplification effect is paralleled in the biodiversity-ecosystem function literature by the sampling effect. A sampling effect may occur when communities with higher species richness, by chance alone, contain the most productive species within a local species pool (Hector et al. 2002). This may explain why experimental communities, constructed from random draws of species, are more productive at higher levels of species richness (Cardinale et al. 2006). Similarly, randomly assembled host communities of higher richness are more likely to contain the most competent hosts by chance, which can result in disease amplification (Joseph et al. 2013). Thus, a sampling effect may generate a positive relationship between host diversity and parasite abundance due to variation in composition, even when host composition varies independently of host richness.

Variation in host composition may also explain how host diversity and resource supply interactively influence parasite abundance. For plant and animal parasites, the most competent hosts may be adapted to high-resource environments (Johnson et al. 2012, Welsh et al. 2016, Liu et al. 2017), potentially linking host composition and host resource supply to changes in parasite abundance. Host composition can also vary predictably with host richness (LoGiudice et al.

2003), and the association of both composition and richness with resource supply may explain why experimental fertilization of an alpine meadow weakened the dilution effect in one study

(Liu et al. 2016), but strengthened the dilution effect in another (Liu et al. 2017). Distinguishing effects of diversity that are attributable to host composition from those attributable to host richness has been a challenge for disease ecologists (Randolph and Dobson 2012, Johnson et al.

2015a). This problem can be overcome by constructing communities that replicate multiple species compositions within species richness levels (Schmid et al. 2002), allowing researchers to attribute changes in ecosystem function to changes in species richness, composition, or both.

Effects of host diversity and resource supply on parasite richness and abundance may also depend on the parasite species’ characteristics. One key parasite characteristic is whether its life cycle includes a mobile free-living life stage or vector, because this can determine whether the parasite experiences more density- or frequency-dependent transmission. Parasites with a mobile free-living life stage or vector may be able to find hosts even at low host density (i.e., they experience more frequency-dependent transmission), while for other parasites, the rate at which they contact hosts may decline with host density (Anderson May, R.M. 1991, i.e., they experience more density-dependent transmission McCallum et al. 2001). Disease amplification requires any diluting effects of decreased host density to be overcome by changes in host composition (Keesing et al. 2006), and frequency-dependent transmission can allow parasites to persist when host density declines with increasing host richness (O’Regan et al. 2015).

Moreover, frequency-dependent transmission may make parasite abundance more sensitive to changes in host composition (Rudolf and Antonovics 2005), contributing further to amplification effects.

Given the complexity of pathways connecting host and parasite diversity and composition, a promising method for simultaneously quantifying these pathways is the ecosystem multifunctionality framework (Dooley et al. 2015, Lefcheck et al. 2015). While early studies of ecosystem functioning evaluated the effect of biodiversity on each function separately, the ecosystem multifunctionality framework aims to describe how biodiversity alters relationships among multiple non-independent functions (Hector and Bagchi 2007). This

approach can also be used to measure how insect and microbial parasite diversity and abundance respond to host diversity and resource supply to hosts (Fig. 4.1). Here, we follow the approach advocated by Dooley et al (2015), by analyzing ecosystem multifunctionality with multivariate- response regressions. This allows for the comparison of treatment effects across multiple non- independent responses.

This study used the ecosystem multifunctionality framework to evaluate how host diversity and resource supply interact to influence parasite abundance and richness. Specifically, we experimentally crossed host species richness and composition with resource supply to hosts, in an herbaceous, perennial-dominated old field in North Carolina, to address three questions: (1)

Do host diversity and resource supply to hosts interact to influence parasite richness and abundance? (2) Does host composition influence parasite abundance, or are the effects of host diversity attributable to changes in host richness alone? (3) Do the effects of host diversity and resource supply differ among insect and microbial parasites?

We hypothesized that parasites experiencing frequency-dependent transmission would be more likely to undergo disease amplification than species experiencing density-dependent transmission. From this general hypothesis, we derived specific predictions for different types of parasites infecting plants. Insect parasites of plants, such as galling and leaf-mining insects, spend a larval life history stage parasitizing a single host individual, but transmission is by free- living adults that can seek out host plants for their offspring, which is expected to lead to more frequency-dependent transmission (Antonovics et al. 1995). In contrast, many microbial parasites of plants that lack insect vectors are passively transmitted, which is expected to lead to more density-dependent transmission. Consequently, we predicted insect parasites to experience disease amplification more commonly than microbial parasites. In contrast to parasite

abundance, effects of host diversity and resource supply on parasite richness have not been hypothesized to depend strongly on density- vs. frequency dependent transmission of the parasites. Hence, we did not expect treatment effects on parasite richness to differ between insect parasites and microbial parasites.

Methods

We performed this study at Widener Farm, an old field in Duke Forest Teaching and

Research Laboratory (Orange County, NC, USA) that produced row crops until 1996. Since

1996, the site has been mowed to produce hay. It is dominated by perennial, herbaceous plants.

The study employed a randomized complete block design with three factorial treatments: (1) we manipulated native plant (i.e., host) richness with multiple native community compositions at each level of richness; (2) access by foliar fungal parasites and insect herbivores; and (3) soil resource supply. The full details of these experimental treatments can be found in Heckman et al

(2017), which reported the effects of the experimental treatments on colonization of the plots by exotic plants, but not effects on parasites. Here, we used only plots in which access by foliar fungal parasites and insect herbivores was unmanipulated. We therefore report results from the manipulation of host richness, community composition, and soil resource supply treatments only, and we do not report results from the manipulation of access by foliar fungal parasites and insect herbivores. This yielded a study that comprised 120 plots (5 replicate blocks × 2 resource supply levels × 2 host richness levels × 6 native community compositions).

Host composition and species richness

In May 2011, we established five spatial blocks; each block included 64 1 × 1 m plots with 1 m aisles between plots. In each block, 16 plots were not planted and are not included in

this study. The existing vegetation was removed from each plot using glyphosate herbicide

(Riverdale® Razor® Pro, Nufarm Americas Inc, Burr Ridge, IL). We did not apply herbicide to aisles between plots. Two weeks after herbicide application, we removed dead vegetation and covered all plots with landscape fabric.

Each plot was assigned to one of two levels of host species richness: monoculture or five- species polyculture. From a pool of six species, we assembled 12 planted communities: six monocultures and six five-species polycultures where one species was excluded from each polyculture community. Host species were selected from a pool of six native herbaceous perennials already present at Widener Farm. We selected host species that were present locally to ensure site suitability and to increase the likelihood that pathogens and herbivores capable of exploiting them were present locally. Our species pool included three grasses—Andropogon virginicus, Setaria parviflora, Tridens flavus, and three forbs—Packera anonyma, Scutellaria integrifolia, Solidago pinetorum.

In summer 2011, plants were propagated in the greenhouse at the University of North

Carolina at Chapel Hill for 8-12 weeks then transplanted into the soil through a small hole in the landscape fabric covering the plot. Each plot contained 41 individual plants, spaced approximately 10 cm from its nearest neighbors in a checkerboard pattern. Polycultures contained nine individuals of one randomly chosen species and eight individuals of the other four species. In early summer 2012, we replaced all individual plants that had not survived the previous winter. Setaria parviflora was planted only in 2012 because it replaced a species that was planted in 2011 but failed to establish in any plot. In July 2012, we removed landscape fabric from all plots, and removed non-planted individuals by hand. From that point forward, plant community richness and composition were unmanipulated.

Resource supply treatment

We began resource supply treatments in July 2012, soon after we completed planting. To manipulate soil resource supply, each plot was assigned to one of two resource supply treatments

(fertilized with 10 g N m-2 as slow-release urea, 10 g P m-2 as triple super phosphate, and 10 g K m-2 as potassium sulfate vs. not fertilized), hereafter referred to as the fertilization treatment.

This level of fertilization was chosen to alleviate nutrient limitation and has been used in other field studies (e.g. Borer et al. 2014a). We applied slow-release forms of each nutrient in order to elevate nutrient supply to experimental communities throughout the growing season.

Quantification of host abundance, parasite abundance and parasite richness

We visually quantified the percent cover of all planted species in each plot in July 2012 using a modified Daubenmire method (Daubenmire 1959, Borer et al. 2014a, Heckman et al.

2017). To account for plot-level edge effects, we quantified the absolute cover of each species in a marked 0.75 × 0.75 m subplot in the center of each plot.

In late September of 2012, when parasite abundance was greatest (Halliday unpublished), we quantified parasite richness and abundance by haphazardly surveying five individuals of each planted species in each plot, for a total of five host individuals in each monoculture plot and 25 host individuals in each polyculture. We visually inspected the five oldest leaves per host individual for damage by foliar parasites. On each leaf, insect and fungal parasites were categorized into morphospecies based on symptom morphology and fungal fruiting body structures when visible (Rottstock et al. 2014, e.g. Liu et al. 2016). We then visually estimated the percent of leaf area damaged by each foliar parasite morphospecies by visually comparing damage on leaves to reference images of leaves of known damage severity (James 1971,

Mitchell et al. 2002, 2003). Insect parasite morphospecies vouchers were verified by the North

Carolina State University plant disease clinic. We limited our data to the subset of insect herbivores that are leaf parasites by including only insects that spend an entire life history stage parasitizing a single host leaf. This included leaf mining and galling insects, and tent-forming caterpillars. Fungal parasite morphospecies vouchers were either verified by the North Carolina

State University plant disease clinic or by culturing fungal isolates from surface-sterilized lesions in 2% malt-extract agar. The cultured fungal isolates were sorted using morphological characteristics, then total genomic DNA was extracted from each unique culture using a RED-

Extract-N-Amp Plant kit (Sigma-Aldrich, St. Louis, Missouri, USA). The ITS region was amplified from the extracted DNA using the fungal-specific primers ITS 1F and ITS 4. The sequences obtained were compared with those from GenBank using the Basic Local Alignment

Search Tool (BLAST, http://blast.ncbi.nlm.nih.gov), yielding genus names for a subset of morphospecies (Table C4.4). Sequences associated with specific fungal morphospecies are deposited in NCBI GenBank under accession numbers MG016006– MG016021.

In order to compare parasite richness between monocultures, where we sampled five host individuals, and polycultures, where we sampled 25 host individuals, we performed “site-based” rarefaction on the count of parasite morphospecies (Gotelli and Colwell 2001), treating host individuals as sites. Specifically, in each polyculture plot, we randomly sampled five host individuals (one of each host species) without replacement, and counted the total number of parasite morphospecies in that subsample. We then permuted this 999 times and took the average rarefied parasite richness for each plot across those 999 permutations. This rarefied estimate of parasite richness, as well as the use of parasite morphospecies instead of parasite taxonomic species, produces a conservative estimate of the parasite species richness in each plot.

We quantified parasite abundance in a plot by calculating the mean leaf area damaged by each parasite on a host, averaged over all host leaves, including uninfected leaves, multiplied by the relative abundance of that host, and then summed across all hosts in the plot (i.e. Mitchell et al. 2002, Heckman et al. 2016). Unlike raw damage, this is a measure of parasite density in a plot, and is therefore independent of variation in host biomass, which can respond strongly to host diversity and resource supply to hosts. This measurement of parasite abundance represents the outcome of transmission among hosts (Johnson et al. 2015a).

Data analysis

We analyzed all data in R version 3.2.3 (R Core Team 2015). To model the effects of host diversity, fertilization, and their interaction on parasite abundance and richness, we used the nlme package for linear mixed effects models (Pinheiro et al. 2016). Each model included fertilization, host diversity, and interactions between these factors, as well as block, as categorical fixed effects.

In order to meet assumptions of homoscedasticity and normality of residuals, we hyperbolic-arcsine transformed parasite abundance. We used the hyperbolic arcsine instead of a conventional logarithmic transformation because unlike the log-transformation, which amplifies small differences in near-zero values, the hyperbolic arcsine approximates a linear- transformation for small values, while still providing a nearly logarithmic transformation of high values (Burbidge et al. 1988, Kirchner and Neal 2013). When we were unable to fully remove heteroscedasticity via transformation, we also included an identity variance structure that modelled residual variance separately by treatment level using the varIdent function in package nlme (Zuur et al. 2009, Pinheiro et al. 2016). We visually inspected the residuals of each model,

separately modelled the variances of the treatments that contained the most heteroscedasticity, and then replotted residuals to confirm that heteroscedasticity was eliminated (e.g. Heckman et al. 2016). For all such cases, residual variance was modelled separately by host diversity treatment.

Our multi-response regression model included four dependent variables: insect parasite richness, microbial parasite richness, insect parasite abundance, and microbial parasite abundance. We standardized the response variables to the same scale by dividing each observation by the maximum value for that response. To model whether effects differed among these responses, we constructed a multi-response regression model of k responses, where the equation for the kth response is of the form:

yk = bk + bik xi + … bnk xn + ek where bik is the main effect for predictor i for response k, and the variance–covariance matrix is a block diagonal matrix with a k × k block for each plot; within each k × k block, the diagonal entries are the individual response variances and off diagonal entries are the covariances between the errors of each pair of responses (e.g. Dooley et al. 2015).

The multivariate response model can be evaluated in two ways. A multivariate ANOVA can be used to test whether the aggregate response is influenced by predictors in the model; this analysis should be interpreted with caution, however, because the individual response with the most complex interaction structure will determine the results of the test (Dooley et al. 2015).

Alternatively, multivariate response modeling is well-suited for pairwise comparisons of fixed coefficients among responses and treatments, and therefore, we base our inferences on these pairwise comparisons.

For each multivariate regression model, we performed a multivariate ANOVA to determine whether host diversity and resource supply interactively influenced the aggregate response. Non-significant interactions make pairwise comparisons of the fixed coefficients difficult to interpret in this multivariate framework (Dooley et al. 2015). We therefore simplified the model to remove interactions that were non-significant in the multivariate ANOVA

(following Zuur et al. 2009). Pairwise comparisons (t-tests) of the fixed coefficients in the final model were averaged over the block effect using the lsmeans package (Lenth 2016). Coefficients were determined to differ significantly with unadjusted p < α*, determined by the Bonferroni correction for four response variables, α* = 0.05/4 response variables = 0.0125, as in Dooley et al. (2015).

To model the effects of species richness alone, we replicated all analyses with one key change: following Schmid et al. (2002) and others (e.g. Heckman et al. 2017), we included planted community composition as a random effect in each model. This conservative test allows us to ascribe differences to richness only when differences in a response within a host richness level (i.e., polycultures or monocultures) are smaller than differences between richness levels

(Schmid et al. 2002). In other words, this analysis tests the effect of host species richness after accounting for variation in host composition.

Results

Across insect and microbial parasites, parasite abundance increased by 16% with host diversity (p = 0.032), but did not significantly respond to host resource supply (p = 0.40), nor the interaction between resource supply and host diversity (p = 0.91; Table C4.1, Fig. 4.2A). After accounting for variation in host composition, the effect of host diversity on parasite abundance

became marginally non-significant (p = 0.054; Fig. 4.2B). This suggests that host diversity influenced parasite abundance via changes in host composition as well as host richness.

Across insect and microbial parasites, parasite richness was influenced interactively by host diversity and resource supply to hosts (p = 0.0022); parasite richness increased by 103% with increasing host diversity in fertilized plots, and 151% in ambient plots (Table S1, Fig.

4.2C). The effect of host diversity on parasite richness and its interaction with resource supply to hosts were both attributable more to host richness than to host composition, as they remained significant even after accounting for variation in host composition (Diversity, p < 0.0001;

Diversity × Resources, p = 0.0005; Fig. 4.2D).

To compare the responses of insect and microbial parasites to host diversity and resource supply, we used multivariate-response regression. In an ANOVA of the aggregate response, there was no interaction between host diversity and resource supply (p = 0.25; Table C4.2), so we reduced the model by dropping that interaction. In the reduced model, the multivariate response was additively influenced by host diversity (p < 0.0001) and resource supply (p =

0.0011). These effects did not change when accounting for variation in host composition

(Diversity, p < 0.0001; Resources, p = 0.0003; Diversity × Resources, p = 0.09). Following

Dooley et al. (2015), we do not draw inference from these ANOVA results; rather, our inferences are based on pairwise comparisons of fixed coefficients from the reduced model, averaged over the block effect, with a Bonferroni correction of α* = 0.05/4 = 0.0125 (Fig. 4.3, Table C4.3,

Pairwise comparisons from the full model are presented in Fig. C4.1), as follows.

Insect richness and abundance were positively correlated (r = 0.68), as were microbial richness and abundance (r = 0.48; Table C4.3). To our knowledge, a correlation between richness and abundance has not been previously reported for plant parasites.

Richness of both insect and microbial parasites increased with host diversity (α* =

0.0125; p = 0.0002 and p < 0.0001, respectively). These effects remained significant after accounting for host composition (α* = 0.0125; p = 0.0043 and p < 0.0001 respectively), indicating that this effect could be attributed to variation in host richness. Although both groups responded to host diversity, microbial parasite richness responded more strongly to the treatment than insect parasite richness: insect and microbial richness were not significantly different in monocultures (α* = 0.0125; p > 0.35), but in polycultures microbial parasite richness increased by 143%, while insect richness increased by 71%. As a result, microbial parasite richness was significantly higher than insect richness in polycultures (α* = 0.0125; p < 0.0001). Neither insect parasite richness nor microbial parasite richness responded significantly to fertilization (α* =

0.0125; p = 0.12 and p = 0.070, respectively).

The effects of host diversity and resource supply on parasite abundance were largely attributable to insect parasites. Resource supply to hosts decreased insect parasite abundance by

44% (α* = 0.0125; p = 0.0031), but did not influence microbial parasite abundance (α* = 0.0125; p = 0.85). Moreover, insect parasite abundance increased 90% with host diversity (α* = 0.0125; p = 0.0016), while microbial parasite abundance did not (α* = 0.0125; p = 0.053), indicating that insect parasites experienced disease amplification, but that microbial parasites did not. This is consistent with the hypothesis that parasites experiencing frequency-dependent transmission, but not density-dependent transmission, would undergo disease amplification. The effect of host diversity on insect parasite abundance became non-significant after accounting for host composition (α* = 0.0125; p = 0.033), indicating that this effect was attributable more to host composition than host richness. This is consistent with the hypothesis that parasites experiencing frequency-dependent transmission would undergo disease amplification due to changes in host

composition, rather than host richness.

Discussion

Theory predicts that parasites experiencing frequency-dependent transmission should respond more strongly to changes in host composition than parasites experiencing density- dependent transmission (Rudolf and Antonovics 2005, Mihaljevic et al. 2014). Because disease amplification requires a stronger response to host composition than host density (Keesing et al.

2006), we expected that insect parasites with free-living lifestages that can actively search for hosts would be more likely to experience disease amplification than parasites exhibiting density- dependent transmission. Our results support this hypothesis. For insect parasites, randomly assembled polycultures favored amplification rather than dilution, and this effect could not be attributed to richness. We also expected that parasites experiencing density-dependent transmission would experience a balance between dilution due to changes in host density (e.g.

Mitchell et al. 2002) and amplification due to changes in host composition (e.g. Power and

Mitchell 2004). Previous studies of microbial parasites, specifically foliar fungal parasites

(Mitchell et al. 2002, e.g. Rottstock et al. 2014) have found dilution effects that were the product of shifts in host density. In contrast, microbial parasite abundance in our system did not respond to host diversity generally, nor to host richness alone. We suggest that this contrasting result is explained by the effect of host composition being stronger than in the previous studies, thus balancing the effect of host density and resulting in no net effect of host diversity on microbial parasite abundance. In our experimental design, each host species in polyculture experienced a five-fold decrease in stem density relative to monocultures, similar to the previous studies.

Unlike the previous studies, our parasites may have experienced stronger effects of host

composition because our parasites had broader host ranges (Table C4.4), and a broader host range increases the potential influence of host composition. Together, using predictions grounded in fundamental theory about frequency- versus density-dependent transmission (Rudolf and Antonovics 2005), these results may help to resolve a contentious issue (Wood and Lafferty

2013, Johnson et al. 2015a): the generality of the relationship between host diversity and disease.

Disease amplification is expected to be largely driven by shifts in host composition, not richness alone (Joseph et al. 2013). Consistent with this expectation, richness alone was insufficient to explain disease amplification in this study. This result held at the parasite morphospecies scale as well (Fig C4.2). Even among individual parasite morphospecies experiencing amplification, the variance in monoculture plots was greater than the variance in polyculture plots, consistent with the sampling effect of biodiversity that was detected at the host community scale. Specifically, because host species were added randomly, polyculture plots were more likely to contain the most heavily infected host species by chance alone, generating an amplification effect at the scale of the host community. This is in contrast to the effect of host richness on parasite richness, indicating parasite colonization (Johnson et al. 2015a), which was not influenced by host composition.

Across all parasites, parasite richness was interactively influenced by host diversity and resource supply to hosts, and this effect was at least partially attributable to variation in host richness. The increase in parasite richness from monoculture to polyculture plots is consistent with many other studies (Kamiya et al. 2014, e.g. Johnson et al. 2016). All parasite species, even broad host generalists, show some degree of host specificity (Poulin et al. 2011) and consequently, species-rich host communities should represent a more heterogeneous pool of resources for parasites, supporting a greater number of parasite species (Kamiya et al. 2014).

These results are consistent with this hypothesis. Host species richness strongly increased parasite richness even after accounting for variation in host community composition.

The effect of host diversity on parasite richness was reduced by experimental fertilization. Eutrophication can reduce species richness by altering competitive interactions among species (Borer et al. 2014b). If experimental fertilization influences the nutrient content of host tissue (e.g. Veresoglou et al. 2013), then parasite richness may decline in experimentally fertilized plots via the same mechanism. This mechanism, however, should reduce parasite richness across all levels of host richness. Alternatively, experimental fertilization may alter competitive outcomes among hosts (Borer et al. 2014b, Harpole et al. 2016, DeMalach et al.

2017), causing one or a few host species to dominate fertilized polycultures, even when host richness is maintained (Reich et al. 2001). Even in such situations where fertilization does not decrease host richness, if it strongly decreases host diversity, then fertilization may also consequently reduce parasite richness.

The risk of infectious disease (i.e., disease risk) is multivariate (Keesing et al. 2006). For example, measures of disease risk may include factors such as disease transmission (e.g. Salkeld et al. 2013), infection severity (Han et al. 2015, e.g. Liu et al. 2016), and mortality (e.g. Han et al. 2015), which are influenced by the abundance of parasites in a community. Measures of disease risk can also include colonization of hosts in a community (e.g. Morand et al. 2014), and emergence of new infectious diseases (e.g. Plowright et al. 2014), particularly infection of novel host species (e.g. Manley et al. 2015), which may be determined by the richness of parasites in a community. Importantly, our results indicate that even when positively correlated (e.g. Watve and Sukumar 1995), parasite abundance and richness may still respond differently to the same characteristics of host communities (Johnson et al. 2013a, 2015a, Rottstock et al. 2014). To our

knowledge, a correlation between abundance and richness has only been made in two other host- parasite systems (Watve and Sukumar 1995, Johnson et al. 2013a), and never for plant parasites.

Effects of host diversity on parasite abundance and richness may depend jointly on parasite transmission mode (Rudolf and Antonovics 2005) and on the composition of the host community (Randolph and Dobson 2012). Models examining this hypothesis have been interpreted as making the prediction that parasites experiencing frequency-dependent transmission are more likely to experience dilution than those experiencing density-dependent transmission (Dobson 2004, e.g. Rudolf and Antonovics 2005, Mihaljevic et al. 2014). However, these models have included the assumption that increasing host richness reduces the frequency of the most competent host (i.e, competence and species richness are non-random, with the most competent hosts most likely to persist at low host richness). Accounting for this assumption suggests another interpretation of that prediction: that parasites experiencing frequency- dependent transmission are more sensitive to host competence than those experiencing density- dependent transmission. This interpretation, in combination with the observation that random addition of host species tends to increase community competence (Joseph et al. 2013), leads to the following prediction. When community assembly is random, a greater sensitivity of frequency-dependent parasites to host competence and reduced sensitivity to host density will drive a stronger amplification effect for frequency-dependent parasites than their density- dependent counterparts. In our experiment, community assembly was random, and the results support this prediction. More diverse communities experienced higher abundance of insect parasite (with more frequency-dependent transmission), but not microbial parasites (with more density-dependent transmission), and this effect could not be attributed to host richness alone, suggesting it was driven by host competence.

For these reasons, the complex relationships that link host diversity and infection risk are often difficult to explore mechanistically (Johnson et al. 2015a). To our knowledge, this study represents the first multivariate analysis of infection risk in a host community. Specifically, we used a multi-response model with an experimental host-parasite system to examine mechanisms driving parasite diversity, improving on past studies. Many studies examining the relationship between host diversity and parasite abundance have been criticized for focusing too little on the mechanisms that drive these relationships (Wood and Lafferty 2013, Civitello et al. 2015,

Johnson et al. 2015a). The multi-response models employed in this experiment examine how characteristics of parasite species influence their response to host diversity and resource supply to hosts, providing mechanistic insight into the processes that structure parasite communities.

Furthermore, this analytical framework, which is grounded in the diversity-ecosystem multifunctionality literature, is generalizable across responses (Dooley et al. 2015), and may therefore represent a new tool for evaluating mechanisms that drive the relationship between biodiversity and disease more broadly.

Host Diversity Resource Supply to Hosts

Host Host Experimental Richness Composition Fertilization

Parasite Richness Parasite Abundance

Insect Microbial Insect Microbial Parasites Parasites Parasites Parasites

Single−response regression Single−response regression accounting for composition

● Ambient 4 Fertilized ● ●

● ● 2 Parasite abundance abundance Parasite

(% leaf area damaged) A) B) 0

6 ● ●

3 ● ● Rarefied parasite richnessRarefied parasite C) D) 0 Monoculture Polyculture Monoculture Polyculture

Multivariate response regression 80 j ij ● Ambient ● Fertilized hi 60 h gh gh ● fgh ● efg ● def 40 de cde bcde ● bcd b ● bc ● 20 ● a

Standardized abundance or richness (%) abundance Standardized Marginal R2 = 0.5115 0

Multivariate response regression accounting for composition k 80 jk ● ij i 60 hi hi ● ghi ● fgh defg 40 ● c ef def bcdef bcde ● ab d bc ● ● 20 ● a

Marginal R2 = 0.4929

Standardized abundance or richness (%) abundance Standardized 2 0 Conditional R = 0.6886 Mono Poly Mono Poly Mono Poly Mono Poly Abundance: Insects Abundance: Microbes Richness: Insects Richness: Microbes

CHAPTER 5 : ASSEMBLY OF THE HOST COMMUNITY INFLUENCES PARASITE RICHNESS AND ABUNDANCE IN A PLANT DIVERSITY EXPERIMENT

Introduction

The richness and abundance of pathogens and parasites (hereafter, “parasites”) that infect wild hosts are important drivers of disease risk and host ecology (Dobson et al. 2008, Hersh et al.

2012, Johnson and Hoverman 2012). At the scale of the host community, two important drivers of parasite richness and abundance are host diversity and resource supply to hosts (e.g., Keesing et al. 2006, Kamiya et al. 2014, Johnson et al. 2016, Liu et al. 2016). Yet, the mechanisms connecting host diversity and resource supply to parasite richness and abundance remain the subject of considerable debate (e.g., Randolph and Dobson 2012, Veresoglou et al. 2013). This debate may result from covariance among host diversity, resource supply, and characteristics of host communities, and disentangling the drivers of this covariance may be hampered by the dynamic nature of ecological communities (Johnson et al. 2015a, Strauss et al. 2016).

The structure of ecological communities changes over time during community assembly

(HilleRisLambers et al. 2011). Specifically, host diversity and resource supply vary naturally, are changing due to anthropogenic drivers, and can alter competitive outcomes among host species, driving host community assembly (Reich et al. 2001, Harpole et al. 2016). Host community assembly may in turn drive the relationships among host diversity, resource supply, parasite richness, and parasite abundance. However, observing accompanying responses in parasite communities is rare (but see Liu et al. 2017). Three characteristics of host communities that

change during host community assembly and may alter parasite richness and abundance are host species richness, exotic host abundance, and host phylogenetic diversity (Figure 5.1a). Here, we test whether changes in each of these three characteristics over time mediate the long-term effects of host diversity and resource supply on parasite richness and abundance.

Changes in host species richness during community assembly

Host diversity and resource supply to hosts may influence the diversity and abundance of parasites during community assembly by altering how host species richness changes over time

(Liu et al. 2017). Increased resource supply often reduces host species richness by decreasing the number of limiting resources that species compete for (Fig. 5.1b, path g; Harpole et al. 2016,

DeMalach et al. 2017). Furthermore, communities that assemble from higher initial diversity may experience a legacy effect, so that host communities of higher richness maintain higher richness during community assembly (Fig. 5.1b, path a; Mouquet et al. 2003).

Parasite richness often increases with host richness, because higher host richness represents a more diverse pool of resources for parasites (Fig. 5.1b, path j; Kamiya et al. 2014,

Johnson et al. 2016, Liu et al. 2016). Parasite abundance, conversely, can respond positively or negatively to host richness (Fig. 5.1b, path k): a dilution effect occurs when increasing host diversity decreases the abundance of one or more parasite species (e.g., Mitchell et al. 2002;

LoGiudice et al. 2003), while an amplification effect occurs when increasing host diversity increases parasite abundance (Power and Mitchell 2004, Young et al. 2013). Empirical evidence exists for dilution and amplification, though dilution effects have been observed more frequently

(Civitello et al. 2015). Thus, because more diverse communities maintain higher richness during community assembly (Mouquet et al. 2003), higher host diversity in the early phases of

community assembly may indirectly increase parasite richness (Fig. 5.1b, paths a and j) and indirectly reduce parasite abundance (Fig. 5.1b, paths a and k) during host community assembly.

And because resource rich communities often lose host species over time, higher resource supply to hosts may indirectly reduce parasite richness (Fig. 5.1b, paths g and j) and may indirectly increase parasite abundance (Fig. 5.1b, paths g and k) via changes in host species richness.

Changes in exotic host abundance during community assembly

Host diversity and resource supply to hosts may influence the diversity and abundance of parasites by altering the abundance of exotic host species over time. Exotic abundance is often reduced by increased species richness (Fig. 5.1b, path b; Levine and D’Antonio 1999, Fargione and Tilman 2005), and increased by increased resource supply (Fig. 5.1b, path h; Huenneke et al.

1990, Davis et al. 2000, Heckman et al. 2017). Changes in exotic host abundance could have important effects on parasite richness and abundance because successful exotic species often escape the parasites that infected them in their native range (Mitchell and Power 2003, Mitchell et al. 2010, Heger and Jeschke 2014), potentially leading to lower parasite abundance and richness in exotic-dominated communities (Fig. 5.1b, paths l and n). Thus, because more diverse communities often become less heavily invaded, higher host diversity may indirectly increase parasite richness and abundance during community assembly via changes in exotic abundance

(Fig. 5.1b, paths b, l, and m). And because resource-rich communities often become more heavily invaded, higher resource supply to hosts may indirectly reduce parasite richness and abundance during community assembly via changes in exotic abundance (Fig. 5.1b, paths h, l, and m).

However, introduced hosts can also acquire infections from closely related native hosts

(Parker et al. 2015) or via repeated introductions over time (Mitchell et al. 2010, Stricker et al.

2016). Because successful exotic species may often be more competent hosts for the parasites that can infect them (Han et al. 2015, Young et al. 2017), this could lead to increased parasite abundance in communities dominated by exotic hosts (Fig. 5.1b, path m). Thus, when exotic hosts are not released from enemies, higher host diversity may indirectly reduce parasite abundance (Fig. 5.1b, paths b and m), and higher resource supply to hosts may indirectly increase parasite abundance (Fig. 5.1b, paths h and m) via changes in exotic abundance.

Changes in host phylogenetic diversity during community assembly

Host diversity and resource supply to hosts may influence the richness and abundance of parasites by altering host phylogenetic diversity, independent of host richness. Specifically, increased host richness may increase richness-independent host phylogenetic diversity by promoting colonization by species from different clades, with low niche overlap (Fig. 5.1b, path c; Mayfield and Levine 2010, Pavoine and Bonsall 2011). Conversely, traits related to nutrient uptake and allocation may be phylogenetically conserved (Verboom et al. 2017), and increased resource supply may therefore reduce phylogenetic diversity by favoring clades with specific resource uptake and allocation strategies (Fig. 5.1b, path i; Mayfield and Levine 2010).

Host species that are evolutionarily more distantly related are less likely to share pathogen species (Gilbert and Webb 2007). Consequently, higher host phylogenetic diversity, independent of species richness, may support a larger number of parasite species (Fig. 5.1b, path n). Moreover, as a consequence of this phylogenetic signal in host range, as host phylogenetic diversity increases, the abundance of parasites is expected to decrease, due to a reduction in the density of hosts (Fig. 5.1b, path o; Parker et al. 2015, Liu et al. 2016). Thus, because more

diverse communities are expected to increase richness-independent phylogenetic diversity during community assembly, higher initial host diversity may indirectly increase parasite richness and indirectly reduce parasite abundance via changes in host phylogenetic diversity (Fig. 5.1b, path c, n, and o). And because increased resource supply is expected to reduce phylogenetic diversity, higher resource supply to hosts may indirectly reduce parasite richness and indirectly increase parasite abundance via changes in host phylogenetic diversity (Fig. 5.1b, path i, n, and o).

Here, we examine how host diversity and resource supply to hosts influence natural host community assembly in experimental host communities in a North Carolina old field, and how host assembly processes in turn affect parasite richness and abundance. This represents the first study, to our knowledge, that experimentally links host and parasite community assembly, clarifying the effects of host diversity and resource supply on parasite richness and abundance.

Methods

We performed this study in an old field in Duke Forest Teaching and Research

Laboratory (Orange County, NC, USA). Since 1996 it has been dominated by perennial, herbaceous plants and mowed to produce hay. The study employed a randomized complete block design, consisting of five spatial blocks, each 15 × 15 m (225 m2). In each block, we established

64 plots, each 1 × 1 m with 1 m aisles between plots. In May 2011, the existing vegetation was removed from each plot using glyphosate herbicide (Riverdale® Razor® Pro, Nufarm Americas

Inc, Burr Ridge, IL), but we did not apply herbicide to aisles between plots. Two weeks after herbicide application, we removed dead vegetation and covered all plots with landscape fabric.

Each plot was assigned to a combination of three factorial treatments: We manipulated native plant (i.e., host) richness with multiple native community compositions at each level of richness;

access by foliar fungal parasites and insect herbivores; and soil nutrient supply. In each block, 16 plots were not planted and are not included in this study. The full details of these experimental treatments can be found in Heckman et al (2017), which reported the effects of each treatment on colonization of the plots by exotic plants, but did not present effects on parasites, except for the effect of manipulating access by foliar fungal parasites and insect herbivores. Here, we report results only in plots where access by foliar fungal parasites and insect herbivores was unmanipulated. This yielded a study that comprised 120 plots (5 replicate blocks × 2 resource supply levels × 2 host richness levels × 6 native community compositions).

Host composition and species richness

Each plot was assigned to one of two levels of host species richness: monoculture or five- species polyculture. From a pool of six species, we assembled 12 planted communities (i.e. combinations of host richness and community composition): six monocultures and six five- species polycultures where one species was excluded from each polyculture community. Host species were selected from a pool of six native herbaceous perennials already present at Widener

Farm. We selected host species that were present locally to ensure site suitability and to increase the likelihood that parasites capable of exploiting them were present locally. Our species pool included three grasses—Andropogon virginicus, Setaria parviflora, Tridens flavus, and three forbs—Packera anonyma, Scutellaria integrifolia, Solidago pinetorum.

Plants were propagated from seed in the greenhouse at the University of North Carolina at Chapel Hill then transplanted into the soil through a small hole in the landscape fabric covering the plot. Each plot contained 41 individual plants, spaced approximately 10 cm from its nearest neighbors in a checkerboard pattern. Polycultures contained 9 individuals of one

randomly chosen species and 8 individuals of the other 4 species. In early summer 2012, we replaced all individual plants that had not survived the winter. Setaria parviflora was planted in

2012 instead of 2011, replacing a species that failed to establish in any plot in 2011. In July

2012, we removed landscape fabric from all plots, and removed non-planted individuals by hand.

Because the goal of this study was to examine how plant and parasite community composition changes over time, we did not weed plots to maintain richness (Fargione and

Tilman 2005, Heckman et al 2017). Thus, the species richness treatments represent initial conditions and not the richness of a plot after July 2012.

Resource supply treatment

We began resource supply treatments in July 2012 after we completed planting. Each plot was assigned to one of two resource supply treatments (fertilized with 10 g each N, P, and K m-2 vs. not fertilized), hereafter referred to as the fertilization treatment. We applied slow-release forms of each nutrient each spring thereafter in order to alleviate nutrient limitation within experimental communities during the growing season.

Quantification of host community assembly

After two years of colonization by the plant species pool, we visually quantified the percent cover of all plant species in each plot in September 2014 using a modified Daubenmire method (Daubenmire 1959, Borer et al. 2014). To avoid plot-level edge effects, we quantified the absolute cover of each species in a marked 0.75 × 0.75 m subplot in the center of each plot. We quantified three components of host community structure to evaluate how the experimental treatments influenced host community assembly: plant species richness, exotic plant abundance,

and the phylogenetic diversity of plant species.

We calculated plant species richness as the number of plant species in each plot. To quantify exotic plant abundance, we classified species as exotic or native to eastern North

America using the USDA Plants Database. We excluded several rare species (amounting to less than 1% of total cover in any plot) which were not identifiable to species. We then assessed the relative abundance of exotic species (hereafter, exotic abundance) as the ratio of the absolute exotic cover to the total cover of all species within a plot.

A phylogeny of all non-tree species was constructed using ‘phyloGenerator’ (Pearse and

Purvis 2013) with options -gene rbcL, matK –alignment mafft –phylogen RAxML – integrated

Bootstrap 1000, and constraint tree topology following Smith (2011). Sequence data were not available for Asclepias syriaca (which amounted to less than 1% of total cover in any plot), so the phyloGenerator function THOROUGH was used to replace sequences with the most closely related taxon using NCBI taxonomy. Plant phylogenetic diversity was calculated using the ses.mpd function in R package Picante (Kembel et al. 2009). Specifically, it was quantified using a null-modeling approach that measures the degree to which a plot is more or less phylogenetically diverse than random, given the number of host species and weighted by their relative abundance. To do this, we generated a z-score comparing the mean-pairwise- phylogenetic-distance between taxa in a plot to a randomly assembled plot with the same number and relative abundance of host species, permuted 1000 times. This allowed us to generate an estimate of host phylogenetic diversity that is independent of host species richness.

Quantification of parasite abundance and parasite richness

Parasite richness and abundance were surveyed in late September of 2014, which is the

period of greatest parasite abundance in this system (Halliday unpublished). In each plot, we measured parasite richness and abundance by haphazardly surveying one individual of the most abundant species, and then the next most abundant species, iterating until the sampled species’ summed cover accounted for at least 80% of the plot’s total plant cover. In addition, we surveyed one individual of all six planted host species, regardless of cover. Surveys consisted of visually inspecting the five oldest leaves per host individual for damage by foliar parasites, including insect and fungal parasites that spend a full life-history stage infecting a single host individual.

On each leaf, parasites were categorized into morphospecies based on symptom morphology and fruiting body structures when visible (e.g., Liu et al. 2016, Halliday et al. 2017a). We then estimated the percent of leaf area damaged by each foliar parasite morphospecies by visually comparing damage on leaves to reference images of leaves of known damage severity (James

1971, Mitchell et al. 2002). Parasite morphospecies vouchers were verified by the North Carolina

State University plant disease clinic or by culturing fungal isolates from surface-sterilized lesions in 2% malt-extract agar. The cultured isolates were sorted using morphological characteristics, then DNA was extracted from each unique culture using a RED-Extract-N-Amp Plant kit

(Sigma-Aldrich, St. Louis, Missouri, USA). The ITS region was amplified from the extracted

DNA using the fungal-specific primers ITS 1F and ITS 4. The sequences were compared with those from GenBank using the UW-BLAST program (BLAST, http://blast.ncbi.nlm.nih.gov), yielding species names for a subset of fungal morphospecies (Table D5.1).

Although insect and microbial parasites may respond differently to host diversity and resource supply to hosts (Halliday et al. 2017a), post-assembly data on insects were not sufficient to test for differences from microbes. Therefore, parasite richness and abundance were calculated across all parasites, including insects and microbes. Parasite richness was calculated as the sum

of all parasite morphospecies per plot. Parasite abundance was the mean leaf area damaged by all parasites on a host, multiplied by the relative abundance of that host, and then summed across all hosts in the plot (Mitchell et al. 2002, Heckman et al. 2016, Halliday et al. 2017a).

Data analysis

To model the indirect effects of host diversity, fertilization, and their interaction, on parasite abundance and richness via their impacts on plant community assembly, we performed path analysis using the piecewiseSEM package for piecewise structural equation models

(Lefcheck 2016). Piecewise structural equation modeling uses local estimation of each path allowing the researcher to incorporate hierarchical model structure (Lefcheck 2016). Each path in the SEM was fit using the nlme package for linear mixed effects models (Pinheiro et al. 2016).

Each path in the model included the experimental covariate of block as a linear combination of coefficients, where the number of coefficients equals the number of levels minus

1. In order to meet assumptions of homoscedasticity and multinormality, we logit transformed exotic abundance, following Heckman et al (2017), and hyperbolic-arcsine transformed parasite abundance. We used the hyperbolic arcsine instead of a conventional logarithmic transformation because unlike the log, which amplifies small differences in near-zero values, the hyperbolic arcsine approximates a linear-transformation for small values while still providing a nearly logarithmic transformation of high values (Burbidge et al. 1988, Kirchner and Neal 2013). When we were unable to fully remove heteroscedasticity via transformation, we also included an identity variance structure that modelled residual variance separately by treatment level using the varIdent function in package nlme (Zuur et al. 2009, Pinheiro et al. 2016). We visually inspected the residuals of each model, separately modelled the variances of the treatments that contained

the most heteroscedasticity, and then replotted residuals to confirm that heteroscedasticity was eliminated (e.g., Heckman et al. 2016). For each path, we allowed variances to differ by host diversity treatment. Following Schmid et al. (2002) and others (e.g., Hector et al. 2011, Heckman et al 2017), we included planted community composition as a random effect in each path. This conservative test allows us to ascribe differences to host richness only when differences in a response within a richness level (i.e., polycultures or monocultures) are smaller than differences between richness levels (Schmid et al. 2002). In other words, this analysis tests the effect of the host diversity treatment after accounting for variation in host composition.

Inspection of the model coefficients revealed 9 nonsignificant block effects. We therefore reduced the model by removing these nonsignificant coefficients and refitting the model

(Following Kline 2010, Cronin et al. 2014). In order to facilitate comparisons among responses and clarify relationships among predictors, we also simplified the final model to remove non- significant interactions among treatments (following Crawley 2007, Zuur et al. 2009).

We tested whether the effects of host diversity and resources on parasite richness and abundance were fully or partially mediated by changes in plant composition by comparing AIC between fully mediated and partially mediated models (Shipley 2009), and evaluating model coefficients in the partially mediated models (following Zhao et al. 2010).

The number of host species, and thus individuals, surveyed varied between plots (min = 2, median = 5, max = 11 host individuals). We therefore performed “site-based” rarefaction on the count of parasite morphospecies (Gotelli and Colwell 2001) and then replicated the analyses. In each plot, we randomly sampled a set number of host individuals, and counted the number of parasite morphospecies in that subsample. We then permuted this 999 times and took the average rarefied parasite richness for each plot across those 999 permutations. The results were not

robust to rarefaction to two individuals per plot (Fig. D5.1), but were robust to five individuals per plot (Fig. D5.2). Based on this level of robustness, the Results present unrarefied results.

Results

We first tested whether initial host diversity and resource supply to hosts interactively influenced parasite richness and abundance via their impacts on host richness, exotic abundance, and phylogenetic diversity as the host community assembled (Fig 5.1b). The data were well fit by this model (Fisher’s C=7.01, df=14, p=0.93; Table S2). In this model, initial host diversity and fertilization each influenced parasite richness and abundance via their impacts on host species richness, exotic abundance, and phylogenetic diversity (Table D5.3, Fig D5.3). However, initial host diversity and resource supply to hosts did not interactively influence any of the response variables (p>0.05). Therefore, we removed these interactions, yielding a reduced model

(Fig 5.2).

We next tested whether the effects of initial host diversity and fertilization on parasite richness and abundance were fully mediated by changes in host richness, exotic abundance, and phylogenetic diversity (i.e., whether there were no other paths by which initial host diversity and fertilization affected parasite richness and abundance). Specifically, if the effects of initial host diversity and fertilization on parasite richness and abundance were mediated by unmeasured variables, these effects would be detected as part of direct paths from initial host diversity or fertilization to parasite richness or abundance. We tested this by comparing the reduced model to models also including these direct paths (Table 5.1). The fully mediated model was the best model based on AIC (Δ AICc > 10 for all partially mediated models; Table 5.1). These results, specifically the lack of any direct paths from the treatments to parasite richness or abundance,

support the hypothesis that the long-term effects of initial host diversity and resource supply to hosts on parasite richness and abundance are determined by changes in host community structure. These results therefore document a novel pathway by which host diversity and resource supply can alter parasite richness and abundance: via altered host community assembly.

Parasite richness and abundance were positively correlated, although only marginally significantly (p = 0.056). This result corroborates two previous reports for plant (Halliday et al.

2017a) and animal (Watve and Sukumar 1995) parasites.

Parasite richness was indirectly influenced by initial host diversity and fertilization, via their impacts on host richness (Fig 5.2; Table D5.4). Initial host diversity had an indirect positive effect on parasite richness. Specifically, initial host diversity increased subsequent host richness by 10% (p = 0.02; Fig 5.3a), which, in turn, positively influenced parasite richness (p < 0.001;

Fig 5.3d). Resource supply to hosts had an indirect negative effect on parasite abundance: fertilization reduced host richness by 26% (p < 0.001; Fig 5.3a), which positively influenced parasite richness (p < 0.001; Fig 5.3d). Parasite richness was not significantly influenced by exotic host abundance or host phylogenetic diversity (p = 0.28 and 0.061; Fig 5.3e and f respectively). Together with block effects, these indirect effects explained 46% of the variance in parasite richness.

Parasite abundance was indirectly influenced by initial host diversity and fertilization, via their impacts on exotic host abundance and host phylogenetic diversity (Fig 5.2; Table D5.4).

Initial host diversity had an indirect negative effect on parasite abundance via its impacts on exotic host abundance and host phylogenetic diversity. Specifically, increasing initial host diversity increased richness-independent phylogenetic diversity of host species by 70% (p =

0.002; Fig 5.3c), which, in turn, negatively influenced parasite abundance (p = 0.043; Fig 5.3i).

Increasing initial host diversity reduced exotic host abundance by 19% (p = 0.009; Fig 5.3b), which in turn positively influenced parasite abundance (p = 0.004; Fig 5.3h). Resource supply to hosts had an indirect positive effect on parasite abundance: fertilization increased exotic host abundance by 60% (p < 0.001; Fig 5.3b), which positively influenced parasite abundance (p =

0.004; Fig 5.3h). Parasite abundance was not significantly influenced by host richness (p = 0.72;

Fig 5.3g). Together with block effects, these indirect effects explained 38% of the variance in parasite abundance.

Discussion

In this study, the effects of host diversity and resource supply to hosts on parasite richness and abundance were indirect, being fully mediated by changes in the structure of the host community, i.e. host community assembly. Previous studies have focused almost exclusively on static drivers of parasite communities (e.g., Mitchell et al. 2002, Keesing et al.

2006, but see Liu et al. 2017), ignoring how these relationships change over time, and consequently perhaps missing important mechanisms underpinning these relationships. This research demonstrates that host community assembly processes, particularly legacy effects of initial host richness and changes in the composition of host communities independent of host richness, are critical for understanding how parasite communities change over time.

Nutrient addition influenced parasite richness and abundance by reducing host species richness and increasing exotic host abundance. Increased resource supply to hosts may alter parasite richness and abundance via two mechanisms. First, resource supply can indirectly affect parasite richness and abundance by shifting host community composition (Liu et al. 2017), as indicated by the significant mediating paths in our analysis. Second, increased resource supply to

hosts can affect parasite richness and abundance by altering the supply of nutrients to parasites within host individuals (Mitchell et al. 2003, Liu et al. 2016), which our path analysis would detect as a direct effect of resource supply on parasite richness or abundance. Our analysis found no significant direct effect of resource supply on parasite abundance or richness, supporting the hypothesis that effects of resource addition that are mediated by host composition may be more important than effects of resource addition driven by host stoichiometry (Veresoglou et al. 2013).

Initial host diversity influenced parasite richness and abundance by increasing host species richness and host phylogenetic diversity and reducing exotic host abundance. These effects of initial host richness on host composition may represent a legacy of initial host richness during host community assembly (Mouquet et al. 2003). A similar legacy effect on parasites would be detected in the path analysis as direct effects of initial host richness on parasite richness and abundance. These direct effects were not detected, indicating that the composition of parasite communities can rapidly equilibrate to changes that occur during host community assembly. This result adds a temporal dimension to the relationship between host diversity and disease risk. In many systems, increased host richness is associated with changes in parasite richness and abundance (Civitello et al. 2015, Liu et al. 2016). However, this relationship may be driven by other characteristics of host communities that are correlated with host richness, such as host phylogenetic diversity (Parker et al. 2015, Liu et al. 2016) and host composition (LoGiudice et al. 2003, Strauss et al. 2016, Halliday et al. 2017a). These results indicate that by determining how these factors change over time, initial host richness can also indirectly alter disease risk.

Host richness and resource supply to hosts indirectly altered parasite abundance via changes in exotic host abundance. In this system, exotic hosts were largely introduced by humans from fertilized pastures (Fridley 2008), benefit most strongly from experimental

fertilization (Heckman et al. 2016, 2017), and are most sensitive to initial host richness

(Heckman et al. 2017). Our results indicate that exotic hosts also contributed most to parasite abundance in communities that they dominated. Exotic species can alter parasite communities through multiple mechanisms. Exotic species are often released from their enemies upon introduction into a new habitat (Mitchell and Power 2003), which may reduce parasite richness and abundance in communities dominated by exotic hosts. In this experiment, three exotic hosts,

Leucanthemum vulgare, Rumex acetosella, and Plantago lanceolata, experienced no damage, indicating that some exotic species may have experienced release from their natural enemies.

However, the majority of exotic hosts, which ultimately contributed the most to exotic abundance, did not show evidence of release from natural enemies. We predicted that exotic hosts that were not released from their enemies would contribute positively to parasite abundance because they often exhibit characteristics of more competent hosts (Han et al. 2015,

Young et al. 2017). The positive relationship between exotic abundance and parasite abundance supports this prediction.

This study used path analysis to model host and parasite community assembly. This methodological approach provides an empirical framework for understanding linkages between host assembly and disease more generally, by evaluating host traits relevant to the system of interest. Our study system is characterized by high but variable abundance of exotic species

(Heckman et al. 2016), and host geographic provenance is linked to host competence (Young et al. 2017), making exotic abundance useful for connecting parasite abundance to host community assembly. However, different traits may be valuable for extending this approach to other systems. For example, a study of co-occurring amphibian species in California linked natural variation in host body size, growth rate, and lifespan to host competence and parasite

transmission (Johnson et al. 2012). Since shifts in host competence during community assembly may determine whether host richness increases or decreases parasite abundance (Joseph et al.

2013), integrating traits linked to host competence into a community-assembly framework may help predict whether host diversity increases or decreases disease in other systems.

Together, these results demonstrate a novel mechanism by which host diversity can affect parasite abundance, generating a dilution effect. A growing body of literature suggests that the association between host diversity and parasite abundance may result from covariance between host diversity and other characteristics of the host community, such as host density, competence, or phylogenetic diversity, that more directly alter parasite abundance in those communities

(Johnson et al. 2015a, Parker et al. 2015, Young et al. 2017). In these situations, the causal relationship between host diversity and parasite abundance can be obscured (Strauss et al. 2016).

One way to overcome this challenge is to look at processes over time, because this can establish causation (Imai et al. 2010). Here, we combined an experimental manipulation of host diversity with post-assembly data on host community structure and parasite abundance. Our results indicate that initial host richness can indirectly alter parasite abundance by determining the trajectory of that community as it assembles over time. This represents an important step forward in providing a temporal mechanism for the effects of host diversity on disease. Previous studies have documented that by altering community assembly, host diversity can alter host composition

(Fargione and Tilman 2005, Mayfield and Levine 2010) , and the effect of host composition on parasite abundance is well documented (LoGiudice et al. 2003, Johnson et al. 2013b, Parker et al.

2015). However, to our knowledge, this is the first study that explicitly links these processes together, documenting assembly-mediated effects of host richness on parasite abundance.

Specifically, while past studies have not considered diversity to be a possible cause of

composition (e.g., LoGiudice et al. 2003, Strauss et al. 2016), during community assembly, diversity at one time can alter future composition, which may impact disease (Joseph et al.

2013). Our results indicate that increased host richness can dilute disease by altering the competence and phylogenetic diversity of host communities as they assemble.

Table 5.1– Test of mediation for the final (reduced) model. Fully mediated model includes paths fromTable 1. experimental Test of mediation for the final (reduced) model. Fully mediated model includes paths from treatments to mediators only. Partially mediated models include paths from experimentalexperimental treatments to mediators only. Partially mediated models include paths from experimental treatments to mediators and responses. treatments to mediators and responses.

A) Model Comparison directed separation test model comparison Mediation Direct effect Fisher's C df p K n AIC AICc Δ AICc Full None 18.73 28 0.906 46 116 110.73 173.4 -- Partial Resource supply 18.09 24 0.799 48 116 114.09 184.3 10.902 Partial Host diversity 18.19 24 0.793 48 116 114.19 184.4 11.002 Partial Resource supply, host 17.64 20 0.611 50 116 117.64 196.1 22.705 diversity Partial Resource supply x host 27.19 26 0.4 52 116 131.19 218.7 45.285 diversity

B) Direct effect coefficients. Parasite richness response Parasite abundance response Resources Diversity Interaction Resources Diversity Interaction Mediation Direct effect Coef p Coef p Coef p Coef p Coef p Coef p Full None ------Partial Resource supply 0.00 0.99 ------0.04 0.74 ------Partial Host diversity -- -- -0.06 0.86 ------0.01 0.91 -- -- Partial Resource supply, host diversity 0.01 0.98 -0.06 0.86 -- -- -0.04 0.76 0.01 0.95 -- -- Partial Resource supply x host diversity 0.22 0.69 0.13 0.79 -0.39 0.56 0.09 0.54 0.11 0.40 -0.25 0.19 df: Chi-squared test degrees of freedom for the Fisher's C statistic; p: significance test derived from a Chi- squared distribution; K: Likelihood degrees of freedom, n: sample size; Δ AICc: difference from the fully mediated model

Initial Host Diversity X Resource Diversity Resources supply to hosts

Future Host Community Structure

Parasite Parasite richness abundance

Initial Host Diversity X Resource Experimental Diversity Resources supply to hosts Treatments (Established 2012) d f c g a b e h i

Host Host Exotic Host Host Community Phylogenetic Species Abundance Structure 2. How doRichness different assembly processesDiversity influence ζ (Measured 2014) BLK 2 parasite diversity?ζ1 2 ζ3

BLK 3 j k l m n o

BLK 4

BLK 5 Parasite Diversity Parasite Parasite Response richness abundance (Measured 2014)

ε1 ε2 B)

Figure 5.2 – Piecewise structural equation model results for the final (reduced) model. Dashed lines are non-significant (p> 0.05). All coefficients are standardized. Correlations between errors are denoted with double-headed arrows. R2 is the marginal R2 from the linear mixed-model, which represents the varianceFINAL PATH MODEL explained by fixed effects- PiecewiseSEM in the model. *p = 0.056

Initial Host Resource Diversity supply to hosts

Host Host -0.30 Exotic Host -0.21 Phylogenetic Species Abundance Richness Diversity BLK 2 R2=0.38 R2=0.27 R2=0.06

BLK 3

BLK 4

BLK 5

Parasite Parasite richness abundance R2=0.46 R2=0.38 0.15*

Dashed lines are non-significant Coefficients are standardized R2 is the marginal R294 * P = 0.056

Figure 5.3 – Bivariate relationships among modelled parameters represented in the Piecewise SEM path diagram. Model estimated effects of resources supply and initial diversity on a) Plant species richness, b) Exotic abundance, c) Plant phylogenetic diversity. Effects of plant species richness, exotic abundance, and plant phylogenetic diversity on d) parasite richness and e) parasite abundance. Parasite abundance and richness are residuals accounting for all other paths in the model (e.g., the left panel in (d) shows the effect of plant species richness on parasite richness after accounting for the effects of exotic abundance and plant phylogenetic diversity on parasite richness). Regression lines are drawn for significant relationships only.

● Ambient 0.8 Fertilized 0.0 ● 0.7 14 ●

● 0.6 −0.4 12 score)

0.5 − ● ● (z −0.8 10 0.4 Exotic abundance Exotic abundance ● Plant Species Richness Plant phylogenetic diversity diversity Plant phylogenetic a) from logit) (backtransformed 0.3 b) c) 8 −1.2 Monoculture Polyculture Monoculture Polyculture Monoculture Polyculture

●

● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ●● ●● ● ● ● ● ●●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●●● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● Parasite richness residuals Parasite ● Accounting for all other paths Accounting for d) e) f)

● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● Parasite abundance residuals abundance Parasite Accounting for all other paths Accounting for g) h) i) ● 5 10 15 20 −2.5 0.0 2.5 −3 −2 −1 0 1 Plant phylogenetic diversity Exotic abundance Plant Species Richness (z−score) (logit transformed)

CHAPTER 6 : CONCLUSION

Understanding the drivers of parasite diversity may be enhanced by using ecological concepts developed for non-parasitic organisms (Johnson et al. 2015c, Seabloom et al. 2015,

Borer et al. 2016). However, even though research into the ecological drivers of parasite diversity has a long history (Kuris et al. 1980, Sousa 1993, Kuris and Lafferty 1994, Esch et al.

2001, Dove and Cribb 2006, Poulin 2007), much of this research has been neglected, because many of these studies used heuristic indices and novel terminology, largely independent of existing ecological literature (Holmes and Price 1986, Bush et al. 1997, Dove 1999, Poulin

2001). As a result, the ecological drivers of parasite diversity are still debated (Poulin 2007,

Poulin and Morand 2011, Kamiya et al. 2013, Johnson et al. 2015c, Seabloom et al. 2015, Borer et al. 2016). The research presented in this dissertation aimed to improve on past studies by integrating a general ecological framework with host-parasite interactions to better understand what determines the composition of foliar parasite communities.

In this dissertation, I leveraged one key insight from ecology’s metacommunity theory – that multiple processes operate simultaneously across different spatial and temporal scales to control the composition of local communities – to advance a more general understanding of parasite diversity. At its simplest, metacommunity theory posits that local and regional processes jointly determine the structure of biological communities by influencing the abundance of species in a given location (Leibold et al. 2004, Holyoak et al. 2005, Logue et al. 2011).

Local processes are likely to operate within host individuals (Kuris et al. 1980). Chapter 2 treated host leaves as habitat patches, and documented local interactions among three co-

occurring parasites and a defensive mutualist. Consistent with studies of non-parasitic organisms

(e.g., Fukami 2015), within-host interactions depended on the sequence of arrival of parasites onto host individuals. Exploring this contingency further, Chapter 3 found that host immunity may underlie within-host parasite interactions and priority effects, supporting predictions about general causes of historical contingency (Vannette and Fukami 2014). These local interactions altered parasite burdens, the probability of coinfection, and even parasite epidemics, revealing the possibility that local interactions can feed back to influence regional processes.

Regional processes can influence the richness and abundance of species able to colonize local patches or hosts (Ricklefs 1987), thereby defining the subset of interactions that can occur locally (Fukami 2015). In Chapter 2, local interactions and epidemics were strongly influenced by parasite phenology, a process that occurs across spatial and temporal gradients. Chapters 4 and 5 explored how characteristics of host and parasite communities influence regional processes for parasites and found that host diversity and resource supply can alter parasite richness and abundance, depending characteristics of the host community and parasite species.

In Chapter 4, host diversity and resource supply to hosts altered parasite richness and abundance, consistent with ecological filtering in non-parasitic metacommunities (Logue et al.

2011). Specifically, host communities planted at higher diversity supported more parasite species, but this effect was reduced by fertilization. This result is consistent with the idea that individual hosts represent important ecological filters, determining which parasites can infect them (Kamiya et al. 2014), and that these ecological filters can be altered by the abiotic environment (Liu et al. 2016). Host communities planted at higher host diversity also exhibited increased parasite abundance, because polyculture plots were more likely to contain the most competent hosts, again consistent with the idea that some hosts represent more suitable habitat

for parasites. An amplification effect could similarly be attributed to spillover from the most competent host or hosts in a community, a process that would be consistent with the concept of mass effects in non-parasitic metacommunities (Logue et al. 2011, Cornell and Harrison 2013).

However, the effects documented in Chapter 4 were also contingent on characteristics of individual parasite species. Microbial parasite abundance did not respond as predicted.

Specifically, microbial parasites did not exhibit evidence of dispersal limitation at high host richness due to reductions in host density, as has been found in other studies of plant disease

(Mitchell et al. 2002, Rottstock et al. 2014, Liu et al. 2016). In contrast, insect parasite abundance increased at high host diversity, because sites planted with multiple species were more likely to contain the most competent hosts by chance, an effect similar to a sampling effect of biodiversity (Hector et al. 2002, Cardinale et al. 2006). After two years, host diversity did ultimately result in a reduction in parasite abundance as predicted (Chapter 5). However, this effect was contingent on shifts in host composition and phylogenetic diversity during host community assembly, a phenomenon that lacks a clear analog in non-parasitic metacommunities but is perhaps akin to gradual changes in regional climate patterns over the geologic time scale or post-glacial primary succession (Chapin et al. 1994, Davis and Shaw 2001).

These findings raise the question: Is there a small set of fundamental laws or first-order principles that govern all ecological patterns and processes? Or is ecology better viewed as a library of case studies, where the outliers, contingencies, and other special circumstances define the natural world?

Like all imposed dichotomies, the reality probably lies somewhere in the middle: There likely are a set of fundamental laws that govern ecological processes, but we need specific knowledge of the conditions in which they occur in order to truly understand them. While

Chapters 2 and 3 did show that local processes may generally operate on parasites within host individuals, these studies also showed that local interactions are strongly influenced by the specific historical context in which they occur. Parasite arrival sequence at local and regional scales influenced parasite epidemics (Chapter 2), and this contingency was attributable to characteristics of individual parasites and the host immune response to infection (Chapter 3).

Similarly, while Chapters 2, 4, and 5 all point to the general phenomenon of regional processes that operate across hosts in a community, these regional processes were contingent on the composition of hosts and characteristics of parasites in a community (Chapter 4) and changed over time as host communities assembled (Chapter 5).

Collectively, these generalities and contingencies may represent the most valuable contribution of this research. Ecological concepts like those from metacommunity theory can inform our understanding of parasite diversity, generally, and may provide a much-needed framework for predicting patterns in unexplored host-parasite systems. At the same time, parasite-specific contingencies may advance a more nuanced understanding of disease ecology, and perhaps ecology in general. This is particularly evident in Chapter 2, where we manipulated parasite phenology – a regional process – and monitored the effect on local species interactions.

This kind of experiment would be intractable for most non-parasitic or otherwise symbiotic organisms, and indicates that the interaction between phenology and dispersal should be explored further in other ecological communities. Consequently, while disease ecologists have much to gain by viewing the world through the lens of a community ecologist, perhaps the community ecologist can also benefit from viewing the world through the lens of the parasitologist.

APPENDIX A: SUPPLEMENTARY MATERIAL FOR CHAPTER 2.

A1. Surveys to determine parasite order of arrival into the host population

We observed seasonal variation in parasite order of arrival into the host community during surveys that took place between 2013 and 2015 (Figure 2.2).

In July, August, and September 2013, we surveyed host leaves for Colletotrichum and

Puccinia in unmanipulated, 1m2 plots, within a field experiment located at Widener farm

(Heckman et al in review). In July, over 50% of leaves were already infected with

Colletotrichum, but Puccinia had not yet emerged. In August, Puccinia still had not emerged, but by September, Puccinia infection had increased to more than 10% of leaves. Plants were not surveyed for Rhizoctonia in 2013. Colletotrichum has infected at least 1% of host leaves in every survey to date. The 2013 data for Colletotrichum were excluded from Figure 2 because the first survey was not until July, when Colletotrichum already infected more than 50% of leaves.

In March, June, July, September, and October 2014, we surveyed host leaves throughout

Widener farm using a stratified random sampling design. In March, more than 30% of leaves were already infected with Colletotrichum, but no host leaves were infected by Puccinia or

Rhizoctonia. In July, Rhizoctonia infected 4% of host leaves, but Puccinia still had not emerged.

In September, Puccinia infected 17% of host leaves.

Finally, between May and October 2015, the presence or absence of parasites was recorded in 1m2 plots distributed throughout the Widener farm field site (O’Keeffe unpublished data). In May, 100% of plots had Colletotrichum infections in them, but no Rhizoctonia or

Puccinia infections. By June, Rhizoctonia was present in 20% of plots, and by August, Puccinia was present in at least 3% of plots.

100

A2. Details of planting treatments and leaf surveys

Endophyte-infected and endophyte-free seed from the KY-31 cultivar of tall fescue

(Lolium arundinaceum) was obtained from the University of Kentucky. The endophyte-free line of the cultivar was created prior to the year 2000 by heat-treating endophyte-infected seeds. The

University of Kentucky has maintained endophyte-free and endophyte-infected KY-31 lines in the field since that time.

In the first cohort, plants were grown in the greenhouse for 30 days, then treated with a foliar insecticide (es-fenvalerate, Asana® XL, Dupont, Wilmington, DE), a well as Marathon to prevent insect herbivory. Asana is a contact insecticide that degrades over time. It has no direct effects on fescue growth in the greenhouse (Heckman et al 2016). Plants were then placed in a shady area for 7 days to harden off before being transferred into the field on 22 June 2015.

During the hardening-off period, 10 of the plants died and were excluded from data analysis. In total from the first cohort, 30 plants (13 from endophyte-infected seed and 17 from endophyte- free seed) were evaluated for symbiont interactions. In the second cohort, plants were grown in the greenhouse for 27 days, then treated with Marathon only and transferred directly into the field on 27 July 2015. In total from the second cohort, 40 plants (20 from endophyte-infected seed and 20 from endophyte-free seed) were evaluated for symbiont interactions. In the third cohort, plants were grown in the greenhouse for 33 days, then treated with Marathon only and transferred directly into the field on 21 September 2015. Four plants from the third cohort were excluded from analysis either because they failed to establish or resulted from seed contamination by the wrong species (Dactylis glomerata). In total from the third cohort, 36 plants (19 from endophyte-infected seed and 17 from endophyte-free seed) were evaluated for symbiont interactions.

101

Leaves were surveyed weekly for infection by foliar parasites. If a tiller died, a new tiller was haphazardly selected from the same host plant. The existing leaves on this tiller were assigned ages in weekly increments from youngest to oldest. On average, one tall fescue leaf emerges per week, so the youngest leaf was assigned an age of 0 days, the second leaf was assigned an age of 7 days, etc. Median leaf lifespan was three weeks, with a maximum observed lifespan of nine weeks.

A3. Details of data analysis

Models of within-host priority effects

To model within-host priority effects, we used a Cox-proportional hazards mixed model from the R package, coxme (Therneau 2012), to predict the probability of a leaf transitioning from uninfected to infected. To account for seasonal variation (a major source of temporal autocorrelation), we modeled the proportional hazards for each focal parasite as the transition rate from uninfected to infected as a function of survey date. We modeled how the probability that a leaf became infected deviated from that baseline rate as a function of leaf age, presence or absence of endophyte infection, and presence or absence of previous infection by other parasites.

We also modeled the interactions between endophyte infection and leaf age and previous parasite infection and leaf age to account for the possibility that a priority effect could be contingent on the age of a leaf. The general model for a focal parasite had the following form:

Fixed effects: Age + Age * Day + Pi + E + Age * E+ Age * Pi

Random effects: Random intercepts of Leaf ID nested in Host ID, and by-host

random slopes for Age,

102

where Age represents the age of a leaf on a given survey, the effect of which can vary seasonally

(Age*day), E represents the infection status by the endophyte (0,1), and Pi represents the infection status by the other parasite species during the previous survey of that leaf (0,1).

For each focal parasite, we fit a model that included the other two parasites, with two exceptions. First, in the second cohort, there were too few Puccinia infections to include

Puccinia infection status as a covariate in any model, or to model Puccinia as a focal parasite.

Second, in the third cohort, there were only 4 instances where Puccinia infected a leaf before

Rhizoctonia. Therefore, we could not test for the interaction between Puccinia and Age in the third cohort model of the focal parasite, Rhizoctonia.

We used Cox-proportional hazards mixed models instead of more traditional logistic regression models for three reasons. First, complex interactions among predictors led to complete separation in logistic models, which makes approximating the likelihood surface unreliable

(Gelman and Hill 2007). Cox mixed-models estimate deviation from a baseline transition rate, and therefore do not suffer from complete separation. Second, generalized linear models can not explicitly account for temporal autocorrelation, while Cox mixed-models can. Finally, once a leaf becomes infected by a focal parasite, that leaf remains infected until it senesces, thus estimating the probability that a leaf is infected will be biased by the duration of infection when using logistic regression. Cox mixed-models instead only track leaves until they become infected by the focal parasite, overcoming this limitation.

Models of within-host Interactions that determine parasite growth

103

To evaluate whether within-host interactions determine parasite growth rate, we modeled the growth of each focal parasite as its change in log-transformed infection severity with respect to leaf age, analyzing only infected leaves, and using the nlme package for linear mixed effects models (Pinheiro et al. 2016). This measurement encompasses both individual-level (lesion expansion) and population-level (new infections) growth within hosts. We employed the model presented by Fenton et al (2010), modified to account for temporal autocorrelation by including a continuous autoregressive structure of order 1 (CAR 1) in each model (Zuur et al. 2009). The general model for a focal parasite had the following form:

Fixed effects: Age + ln(Pi +1) + E + Age * E + Age * ln(Pi+1)

Random effects: Random intercepts of Leaf ID nested in Host ID, by-leaf random

slopes of Age, and by-host random slopes of a spline fit to Age

(e.g., Fenton et al 2010),

where Age represents the age of a leaf on a given survey (a proxy for exposure to parasite propagules), E represents the infection status by the endophyte (0,1), and Pi is the severity of infection by other parasite species during the previous survey of that leaf. Here, the main effect of other parasites and the endophyte are interpreted as their impact on the average leaf during its first week after emergence (i.e., between leaf emergence and the first survey of that leaf), and interactions between those variables and leaf age represent their impacts on focal parasite growth after the first week.

The overall impact of each symbiont was assessed by evaluating model-predicted values of each predictor variable over the range of observed values, weighted by the relative number of

104

observations of each value. This way, we avoided extrapolating from model results into areas of parameter space where there was no data. Specifically, for each observation of each leaf, we calculated the per capita effect of each parasite P on the focal parasite growth rate as bP + bP:A*Age, evaluated at the observed age, where bP is the modelled main effect of the other parasite on the focal parasite, and bP:A is the modelled effect for the interaction between previous parasite infection severity and leaf age. We then took the average across all observations on each leaf to calculate the average modeled per capita effect of each parasite on the focal parasite per leaf, and finally averaged across all leaves to calculate the average overall per capita effect of each parasite on the focal parasite.

In all models, we used the following imputation method to determine the previous infection severity of each foliar parasite during the first survey of each leaf: If a leaf was first observed to be uninfected (89% of all leaves surveyed), then its previous infection severity was assigned a value of 0. Each leaf that was first observed to be infected (11% of leaves) was assigned a previous infection severity that was one-half of the infection severity during the first survey of that leaf.

105

A4. Supplemental Figures

Figure A6.1 – Cohort 1 within-host parasite growth. Plots are results of the reduced longitudinal mixed models, showing the rate at which log-transformed infection severity among infected leaves increased as leaves age (i.e., parasite growth rate), and the effects of other symbionts on that relationship.In all models, we used the following imputation method to determine the previous a-c) Colletotrichum infection severity as a function of previous Puccinia infectioninfection severity of each foliar parasite during the first survey of each leaf: If a leaf was first severity, previous Rhizoctonia infection severity, and endophyte infection, respectively. Colorsobserved to be unin and contour linesfected (89% of all leaves surveyed), then its previous infection severity was represent model-estimated Colletotrichum infection severity. d-e) Pucciniaassigned a value of 0. Each leaf that was first observed to be infected (11% of leaves) was infection severity as a function of previous Colletotrichum infection severity and endophyteassigned a previous infection severity that was one infection, respectively. f) Rhizoctonia infection-half of the infection severity dur infection severity as a functioning the first of previoussurvey of that leaf. Colletotrichum infection severity. Points represent individual observations of leaves over the course of the experiment. “Neg effects” are the number of leaves where the model estimatedReferences cited in this appendix but not the manuscript: a negative effect of the other parasite on the focal parasite. “Pos effects” are the numberGelman, of A. leaves & Hill, where J. (2007). the model Data estimatedanalysis using a positive regression effect. and “Mean multilevel/hierarchical effect” is the model models- . estimatedCambridge per University-capita effect Press of New the otherYork, parasiteNY, USA on the focal parasite. * denotes estimated effects for models where there was no interaction between leaf age and previous infection severity by the other parasite. † denotes estimated effects for models where the main effect was non- significant.Appendix I V. TheseSupplemental Figures model results are summarized in Figure 2.4b.

Colletotrichum Severity Focal Parasite Focal Parasite Colletotrichum Colletotrichum 20 severity 10 severity Endophyte free 10 Endophyte infected 0.6 6 0.6 6

1 2

2 Colletotrichum severity severity Colletotrichum Previous Puccinia severity severity Puccinia Previous Previous Rhizoctonia severity severity Rhizoctonia Previous

0 0 Neg effects:226 Pos effects:1 Mean effect:-0.3667 Neg effects:227 Pos effects:0* Mean effect:-0.263* 0.1 0 20 40 60 0 20 40 60 0 20 40 60 a) b) Leaf age (days) c)

Puccinia Severity Rhizoctonia Severity Focal Parasite Puccinia Focal Parasite 5 severity 50 Endophyte free 10 Endophyte infected Rhizoctonia severity 0.3 11 1.2 6

Puccinia severity severity Puccinia 1

1 Previous Colletotrichum severity severity Colletotrichum Previous severity Colletotrichum Previous

0.1 0 0 Neg effects:17 Pos effects:59 Mean effect:0.3221 Neg effects:83 Pos effects:1 Mean effect:-0.3964† 0 20 40 60 0 20 40 60 0 20 40 60 d) e) Leaf age (days) f) Figure S1. Cohort 1 within-host parasite growth. Plots are results of the reduced longitudinal mixed models, showing the rate at which log-transformed infection severity among infected leaves increased as leaves age (i.e., parasite growth rate), and the effects of other symbionts on that relationship. a-c) Colletotrichum infection severity as a function of previous Puccinia infection severity, previous Rhizoctonia infection severity, and endophyte infec106 tion, respectively. Colors and contour lines represent model-estimated Colletotrichum infection

severity. d-e) Puccinia infection severity as a function of previous Colletotrichum infection severity and endophyte infection, respectively. f) Rhizoctonia infection infection severity as a Figurefunction of previous A6.2 – Cohort Colletotrichum3 within-host parasite infection severity growth. Plots. Points represent individual observations are results of the reduced longitudinal mixedof leaves over the course of the experiment. “Neg effects” are the number of leaves where the models, showing the rate at which log-transformed infection severity among infected leavesmodel increasedestimated as a negative effect leaves age (i.e., of the other parasite on the focal parasite. “Pos effects” are parasite growth rate), and the effects of other symbionts on thatthe number of leaves where the model relationship. a) Colletotrichum infectionestimated severity a positive effect. “Mean effect” is as a function of endophyte infection.the model b) - Pucciniaestimated infection per-capita effect of the other parasite on the focal parasite. severity as a function of endophyte infection. c) Rhizoctonia * denotes infectionestimated severity aseffects a function for models where there was no interaction between leaf age and previous infection of previous Colletotrichum infection severity. † denotes estimated effects for modelsseverity by the other parasite. where the main effect was† denotes estimated effects for models where the main effect non-significant. These model results are summarized in Figure 2.4b.was non -significant. These model results are summarized in Figure 4b.

Colletotrichum Severity Puccinia Severity Rhizoctonia Severity Focal Parasite Rhizoctonia Endophyte free 100 Endophyte free 5 severity Endophyte infected Endophyte infected 5 1.6 15

1 1

1 severity Puccinia Colletotrichum severity severity Colletotrichum Previous Colletotrichum severity severity Colletotrichum Previous 0.1 0.5 0 Neg effects:136 Pos effects:1 Mean effect:-0.5451† 0 10 20 30 40 10 20 30 0 10 20 30 40 a) b) Leaf age (days) c) Figure S2. Cohort 3 within-host parasite growth. Plots are results of the reduced longitudinal mixed models, showing the rate at which log-transformed infection severity among infected leaves increased as leaves age (i.e., parasite growth rate), and the effects of other symbionts on that relationship. a) Colletotrichum infection severity as a function of endophyte infection. b) Puccinia infection severity as a function of endophyte infection. c) Rhizoctonia infection severity as a function of previous Colletotrichum infection severity. † denotes estimated effects for models where the main effect was non-significant. These model results are summarized in Figure 4b.

107

A5. Supplemental tables Appendix V. Supplemental Tables

TableTable S1. A6.1Cohort 1 survival analysis ANOVA – Cohort 1 survival analysis ANOVA

108

Table S2. Cohort 2 survival analysis ANOVA TableTable S2. A6.2Cohort 2 – Cohort survival analysis ANOVA 2 survival analysis ANOVA

TableTable S3. Table S3. A6.3Cohort 3 survival analysis ANOVA Cohort 3 survival analysis ANOVA– Cohort 3 survival analysis ANOVA

109

Table A6.4 – Reduced model coefficients. Models were reduced from a full model using likelihood ratio tests to remove non-significant interactions. Estimates are only provided if they were included in the reduced model. Significant effects (p<0.05) from those likelihood ratio tests are indicated in bold. Coefficients are exponentiated.

110

Table A6.5 – Cohort 1 longitudinal linear mixed model ANOVA Table S5. Cohort 1 longitudinal linear mixed model ANOVA

111

Table S6. TableTable S6. A6.Cohort 2 longitudinal linear mixed model ANOVA6Cohort 2 longitudinal linear mixed model ANOVA – Cohort 2 longitudinal linear mixed model ANOVA

Table S7. TableTable S7. A6.Cohort 3 longitudinal linear mixed model 7Cohort 3 longitudinal linear mixed model – Cohort 3 longitudinal linear mixed modelANOVAANOVA ANOVA

112

Table A6.8 – Reduced model coefficients. Models were reduced from a full model using likelihood ratio tests to remove non-significant interactions. Estimates are only provided if they were included in the reduced model. Significant effects (p<0.05) from those likelihood ratio tests are indicated in bold. Coefficients are on a log scale.

113

APPENDIX B: SUPPLEMENTARY MATERIAL FOR CHAPTER 3.

B1. Supplemental tables

Table B6.9 – Reduced model coefficients. Models were reduced from a full model using likelihood ratio tests to remove non-significant interactions. Estimates are only provided if they were included in the reduced model. Coefficients are exponentiated for Cox mixed models and on a log scale for longitudinal mixed models A) Rhizoctonia B) Colletotrichum Coefficient SE Coefficient SE Disease risk (Cox mixed model) Treatment [JA] 1.18 0.200 1.02 0.46 Treatment [SA] 1.07 0.20 0.33 0.52 Puccinia infection 3.23 0.93 0.55 0.6 Colletotrichum infection 0.52 0.74 Rhizoctonia infection 0.4 0.35 Treatment [JA] × Colletotrichum 3.55 0.950 Treatment [SA] × Colletotrichum 0.12 1.490 Treatment [JA] × Rhizoctonia Treatment [SA] × Rhizoctonia

Infection severity (Longitudinal Linear Mixed Model) (Intercept) 0.66 0.18 -0.10 0.24 age 0.44 0.05 0.23 0.08 Treatment [JA] -0.092 0.26 -0.48 0.32 Treatment [SA] 0.18 0.25 -0.12 -0.51 Colletotrichum severity -0.30 0.57 Puccinia severity 0.020 0.075 1.86 0.71 Rhizoctonia severity 0.057 0.070 age × Treatment [JA] -0.046 0.073 0.17 0.10 age × Treatment [SA] 0.044 0.069 -0.13 0.15 Treatment [JA] × Colletotrichum 0.46 0.95 Treatment [SA] × Colletotrichum 13 5.34 Treatment [JA] × Rhizoctonia Treatment [SA] × Rhizoctonia age × Colletotrichum age × Puccinia -0.42 -0.17 age × Rhizoctonia age × Treatment [JA] × Colletotrichum -0.10 0.24 age × Treatment [SA] × Colletotrichum -3.16 1.31 age × Treatment [JA] × Rhizoctonia age × Treatment [SA] × Rhizoctonia

114

Table B6.10 – Disease risk [Cox mixed model] ANOVA A) Rhizoctonia B) Colletotrichum DF Χ2 p-value DF Χ2 p-value Treatment 2.00 7.36 0.03 2 40.04 0.00 Colletotrichum infection 1.00 2.28 0.13 Puccinia infection 1.00 0.10 0.75 1 1.09 0.30 Rhizoctonia infection 1 8.40 0.00 Treatment × Colletotrichum 2.00 6.25 0.04 Treatment × Rhizoctonia 2 0.49 0.78

Table B6.11 – Parasite Prevalence ANOVA DF Χ2 p-value Parasite 1 213.14 0.00 Treatment 2 0.19 0.91 Treatment × Parasite 2 17.31 0.00

115

Table B6.12 – Infection severity [longitudinal linear mixed model] ANOVA B) A) Rhizoctonia Colletotrichum DF F-value p-value DF F-value p-value Intercept 1, 193 1282.15 0.00 1, 38 27.68 0.00 age 1, 193 134.94 0.00 1, 36 51.46 0.00 Treatment 2, 51 11.54 0.00 2, 29 3.02 0.06† Puccinia severity 1, 193 0.02 0.89 1, 36 1.12 0.30 Colletotrichum severity 1, 193 3.98 0.05 Rhizoctonia severity 1, 36 1.44 0.24 age × Treatment 2, 193 0.83 0.44 2, 36 2.76 0.08† age × Puccinia 1, 193 0.07 0.79 1, 36 6.77 0.01 age × Colletotrichum 1, 193 0.08 0.77 age × Rhizoctonia 1, 36 0.25 0.62 Treatment × Colletotrichum 2, 193 0.20 0.82 Treatment × Rhizoctonia 2, 36 1.25 0.30 age × Treatment × Colletotrichum 2, 193 2.95 0.05† age × Treatment × Rhizoctonia 2, 36 0.87 0.43 † significant at p < 0.05 after model reduction

Table B6.13 – Coinfection ANOVA DF Χ2 p-value Treatment 2 8.28 0.02

Table B6.14 – Parasite Burden ANOVA DF F-value p-value Intercept 1, 304 262.60 0.00 Parasite 2, 51 5.82 0.01 Treatment 1, 304 182.40 0.00 Treatment × Parasite 2, 304 11.38 0.00

116

APPENDIX C: SUPPLEMENTARY MATERIAL FOR CHAPTER 4.

C1. Supplemental tables

Table C6.15 – ANOVA for parasite abundance and parasite richness models. A) Parasite abundance, B) Rarefied parasite richness. A) Parasite B) Rarefied abundance parasite richness DF F-value p-value F-value p-value Models not accounting for composition Intercept 1, 110 1240.75 <0.0001 1887.56 <0.0001 Block 4, 110 1.98 0.10 2.43 0.052 Resources 1, 110 0.71 0.40 10.33 0.0017 Diversity 1, 110 4.75 0.032 256.03 <0.0001 Resources × Diversity 1, 110 0.013 0.91 9.87 0.0022

Models accounting for composition Intercept 1, 110 1240.75 <0.0001 497.63 <0.0001 Block 4, 110 1.98 0.10 2.33 0.061 Resources 1, 110 0.71 0.40 10.75 0.0014 Diversity 1, 10 4.75 0.054 76.98 <0.0001 Resources × Diversity 1, 110 0.013 0.91 12.80 0.0005

117

Table C6.16 – ANOVA for multi-response regression models. “Response” is the multivariate response of insect and microbial parasite abundance and richness DF F-value p-value Insects vs microbes Block 5, 452 285.25 <0.0001 Response 3, 452 70.92 <0.0001 Response × Resources 4, 452 4.66 0.0011 Response × Diversity 4, 452 39.64 <0.0001 Response × Resources × 4, 452 1.35 0.25 Diversity

Insects vs microbes accounting for composition Block 5, 102 45.01 <0.0001 Response 3, 340 80.59 <0.0001 Response × Resources 4, 340 5.34 0.0003 Response × Diversity 4, 340 24.37 <0.0001 Response × Resources × 4, 340 2.04 0.09 Diversity

118

Table C6.17 – Estimated model terms for the transformed responses insect abundance, insect richness, microbial abundance, and microbial richness. (a) fixed coefficients (averaged over block), (b) the variance covariance matrix (left) and correlations (right) Response Measurement (a) Fixed coefficients Insect Abundance Insect Richness Microbial Microbial (%) (%) Abundance (%) Richness (%) Treatment Est SE Est SE Est SE Est SE Fertilized monoculture 6.87 3.08 1 a 19.05 3.92 1 b 44.91 3.74 1 c 23.88 3.92 1 b Ambient monoculture 16.41 3.03 2 a 25.79 3.85 12 B 45.67 3.67 1 C 31.51 3.85 1 b Fertilized polyculture 17.33 2.59 2 a 34.94 3.30 23 b 52.70 3.15 1 c 63.46 3.30 2 d Ambient polyculture 26.87 2.59 3 a 41.68 3.30 3 b 53.46 3.15 1 b 71.08 3.30 2 c

119

(b) The variance- Variances and covariances Correlations covariance matrix (left) and correlations Insect Insect Microbial Microbial Insect Microbial Microbial (right) abundance richness abundance richness richness abundance richness Insect abundance 446.01 0.68 -0.23 -0.38 Insect richness 290.51 434.16 -0.19 -0.26 Microbial abundance -101.35 -83.34 446.01 0.48 Microbial richness -162.45 -113.97 205.08 434.16 Coefficients that are not significantly different across treatments, within a response (i.e., in the same column), have a number in common. Coefficients that are not significantly different across responses, within a treatment (i.e., in the same row), have a letter in common. The level of significance is determined by the Bonferroni correction is α*=0.05/4 = 0.0125.

Table C6.18 – Parasite morphospecies (symptoms in brackets) and associated hosts

Parasite Morphospecies Andropogon Setaria Tridens Scutellaria Packera Solidago (GenBank Accession #) virginicus parviflora flavus integrifolia anonyma pinetorum Microbial Parasites Balansia sp. [stromata] X Colletotrichum cereale [anthracnose] (MG016017, MG016018) X X Colletotrichum sublineolum [anthracnose] (MG016014) X Drechslera / Aureobasidium / Gaueumannomyces complex [leaf spot] X (MG016008, MG016019, MG016021) Drechslera / Curvularia complex [leaf spot] (MG016006, MG016007) X Drechslera sp. [leaf spot] (MG016015) X X X Mycosphaerellacea / Colletotrichum cereale complex [leaf spot] (MG016018, X MG016020) Mycosphaerellaceae [leaf spot] (MG016010, MG016011, MG016013) X X Phyllachora graminis [tar spot] X Sclerotinia sp. [white rot] (MG016009) X Stagonospora sp. [leaf spot] X X Coleosporium sp. [rust] X Unidentified bacterium [leaf curl] X Unidentified microbe [black leaf spot] X X Unidentified microbe [black lesion] X Unidentified fungus [brown rust] X Unidentified fungus [orange rust] X Unidentified fungus [epiphyte] X X Unidentified microbe [chlorosis] X Unidentified microbe [dieback] X Unidentified microbe [scorch] X X Unidentified fungus 1 [leaf spot] X X Unidentified fungus 2 [leaf spot] X Unidentified fungus 3 [leaf spot] X Insect Parasites Unidentified leaf mining insect X X X Unidentified galling insect 1 X Unidentified galling insect 2 X Unidentified galling insect 3 X Unidentified galling insect 4 X Unidentified tent caterpillar X X X

120

C2. Supplemental figures

Figure C6.3 – Effects of host diversity (mono = monoculture plots; poly = polyculture plots) and resource supply to hosts (ambient, black circles; fertilized, red triangles) on insect and microbial parasite abundance and richness, calculated using a multi-response regression model, and standardized to a common variable. The top panel shows the overall effects of host diversity on insect and microbial parasite abundance and richness. The bottom panel shows the effects of host diversity after accounting for variation in host composition. Coefficients that share a letter do not differ significantly as determined by multiple comparisons tests with the Bonferroni correction, α* = 0.05/4 = 0.0125. Estimates are from the full model that includes a non-significant interaction between host diversity and soil fertility.

Multivariate response regression

g 80 ● Ambient ● Fertilized fg

ef ef 60 de de ● de

● ● cd 40 bc bc c abc bc ● ● ab ab ● 20 a ●

Standardized abundance or richness (%) abundance Standardized Marginal R2 = 0.5134 0

Multivariate response regression accounting for composition i 80 ● hi gh gh 60 fgh fgh ● efg

● ● def bcde 40 cd bcde bcd bcd abc ● ab ● ● 20 a ● Marginal R2 = 0.4964 Standardized abundance or richness (%) abundance Standardized 2 0 Conditional R = 0.6898 Mono Poly Mono Poly Mono Poly Mono Poly Abundance: Insects Abundance: Microbes Richness: Insects Richness: Microbes

121

Figure C6.4 – Effects of host diversity (monoculture plots; polyculture plots) and resource supply to hosts (ambient, black; fertilized, red) on the abundance of each parasite morphospecies, standardized to a common variable. Violins show distribution of the data. Asterisks show the mean of each group.

Balansia sp. [stromata] Coleosporium sp. [rust] Colletotrichum cereale Colletotrichum sublineolum Drechslera / Aureobasidium / Drechslera / Curvularia [anthracnose] [anthracnose] Gaueumannomyces complex complex [leaf spot] [leaf spot] 100 75 * Ambient 50 * Fertilized 25 0 * * * * * * * * * * * * * * * * * * * * * * * * Drechslera sp. [leaf spot] Mycosphaerellacea / Mycosphaerellaceae [leaf spot] Phyllachora graminis [tar spot] Sclerotinia sp. [white rot] Stagonospora sp. [leaf spot] Colletotrichum cereale complex [leaf spot] 100 75 50 25 * 0 * * * * * * * * * * * * * 122 * * * * * * * * * * Unidentified bacterium [leaf curl] Unidentified fungus [brown rust] Unidentified fungus [epiphyte] Unidentified fungus [orange rust] Unidentified fungus 1 [leaf spot] Unidentified fungus 2 [leaf spot]

100 75 50 25 0 * * * * * * * * * * * * * * * * * * * * * * * * Unidentified fungus 3 [leaf spot] Unidentified galling insect 1 Unidentified galling insect 2 Unidentified galling insect 3 Unidentified galling insect 4 Unidentified leaf mining insect Standardized load (% of max) Standardized

100 75 50 25 * * * 0 * * * * * * * * * * * * * * * * * * * * * Unidentified microbe Unidentified microbe [black lesion] Unidentified microbe [chlorosis] Unidentified microbe [dieback] Unidentified microbe [scorch] Unidentified tent caterpillar [black leaf spot]

100 75 50 25 0 * * * * * * * * * * * * Monoculture* * Polyculture Monoculture Polyculture Monoculture* * Polyculture* * Monoculture* Polyculture* Monoculture Polyculture Monoculture* * Polyculture* *

Host Richness

APPENDIX D SUPPLEMENTARY MATERIAL FOR CHAPTER 5.

D1. Supplemental tables

123

Table D6.19 – Parasite morphospecies and their associated host species grouped into four categories: A) planted host species, B) native colonizing host species, C) exotic colonizing host species, and D) host species with unknown geographic provenance. Parasite morphospecies is presentedTable S1. Parasite morphospecies and their associated host species grouped into four categories: A) planted host species, B) native colonizing host species, C) exotic colonizing in the leftmost column, with parasite type in brackets and genbank accession numbers host species, and D) host species with unknown geographic provenance. Parasite morphospecies is presented in the leftmost column, with parasite type in brackets and ingenbank accession numbers in parentheses. parentheses.

A) Planted host species Andropogon Setaria Scutellaria Packera Solidago virginicus parviflora Tridens flavus integrifolia anonyma pinetorum Balansia sp. [stromata] Colletotrichum cereale [anthracnose] (XXX) Drechslera / Aureobasidium / Gaueumannomyces complex [leaf spot] (XXX) Drechslera / Curvularia complex [leaf spot] (XXX) Drechslera sp. [leaf spot] (XXX) Mycosphaerellaceae [leaf spot] (XXX) Phyllachora graminis [tar spot] Rhizoctonia solani [leaf spot] Stagonospora sp. [leaf spot] Unidentified fungus [choke] Unidentified galling insect 1 Unidentified galling insect 3 Unidentified leaf mining insect

Unidentified microbe [black leaf spot] Unidentified microbe [dieback] Unidentified microbe [scorch] Unidentified tent caterpillar

B) Native colonizing host species Conyza Dichanthelium Eragrostis Oxalis Paspalum Schizachyrium Solanum Symphiotrichum canadensis dichotomum Erigeron annuus canadensis dillenii notatum scoparium carolinense pilosum Alternaria alternata [leaf spot] Colletotrichum cereale [anthracnose] (XXX) Colletotrichum sp. 2 [anthracnose] (XXX) Drechslera sp. [leaf spot] (XXX) Phyllachora graminis [tar spot] Unidentified fungus [leaf spot] Unidentified galling insect 3

Unidentified microbe [black leaf spot] Unidentified microbe [scorch] Unidentified tent caterpillar

C) Exotic colonizing host species Holcus Lespediza Leucanthemum Loniscera Plantago Rumex Schedonorus Sorghum lanatus cuneata vulgaris japonica lanceolata acetosella arundinaceus halpiensis

Bipolaris drechsleri [leaf spot] (XXX) Colletotrichum cereale [anthracnose] (XXX) Colletotrichum sublineolum / Alternaria / Aureobasidium complex [leaf spot] (XXX) Didymella glomerata [leaf spot] (XXX) Drechslera sp. [leaf spot] (XXX) Pestalotiopsis / Diaporthe complex [leaf spot] (XXX) Puccinia coronata [rust] Puccinia graminis [rust] Rhizoctonia solani [leaf spot] Sclerotinia sp. [white rot] (XXX) Unidentified fungus [leaf spot] Unidentified tent caterpillar

D) Host species with unknown geographic provenance Unknown Dichanthelium Poaceae 2 Carex sp. Juncus sp. sp. 2 Phyllachora sp. [tar spot] Unidentified microbe [chlorosis] Unidentified fungus [yellow rust]

124

Table D6.20 – Piecewise SEM goodness of fit Test. A) Conditional independence claims for a direct separation test using the full model. B) Results of the direct separation test (p<0.05 indicates that the model should be rejected) A) Missing path estimate se df critical value p Par rich ~ Resources -0.0851 0.4239 102 -0.2008 0.8412 Par abund ~ Resources -0.0418 0.119 102 -0.3511 0.7262 Par rich ~ Diversity -0.0206 0.3559 102 -0.0578 0.954 Par abund ~ Diversity -0.0125 0.0991 102 -0.1265 0.8996 Par rich ~ Diversity*Resources -0.4504 0.6772 100 -0.6652 0.5075 Par abund ~ Diversity*Resources -0.2443 0.1873 100 -1.3045 0.195 Host phylo div ~ Host richness 0.1761 0.3155 102 0.5582 0.5779

B) Fisher C df p AIC AICc K n 7.01 14 0.934 123.01 243.08 58 116

125

Table D6.21 – Coefficient estimates from the full model. Estimates are standardized to a common scale to facilitate comparisons. Correlations among dependent variables are indicated by ~~.

Response Predictor Estimate Std Error p Parasite Richness Host Richness 0.565 0.087 0.000 Block 5 -0.780 0.245 0.002 Block 2 -0.697 0.221 0.002 Host Phylogenetic Diversity -0.157 0.074 0.037 Block 4 -0.415 0.233 0.078 Block 3 -0.304 0.233 0.195 Host Exotic Abundance 0.042 0.091 0.645 Parasite Abundance Block 3 0.997 0.252 0.000 Host Exotic Abundance 0.268 0.099 0.008 Host Phylogenetic Diversity -0.157 0.082 0.057 Block 2 -0.450 0.239 0.063 Host Richness 0.034 0.095 0.725 Block 5 0.063 0.264 0.813 Block 4 -0.027 0.252 0.914 Host Richness Resources -0.993 0.213 0.000 Block 3 -0.886 0.224 0.000 Block 4 -0.812 0.217 0.000 Block 5 -0.718 0.217 0.001 Block 2 -0.436 0.217 0.047 Diversity 0.399 0.201 0.050 Resources X Diversity -0.106 0.281 0.706 Host Exotic Abundance Resources 1.171 0.254 0.000 Block 5 -0.797 0.205 0.000 Block 4 -0.505 0.205 0.016 Resources X Diversity -0.431 0.297 0.149 Block 2 -0.295 0.206 0.156 Block 3 0.183 0.211 0.390 Diversity -0.179 0.212 0.400 Host Phylogenetic Diversity 0.466 0.255 0.071 Diversity Block 2 0.328 0.251 0.195 Resources -0.220 0.305 0.472 Resources X Diversity 0.210 0.358 0.558 Block 4 0.073 0.249 0.771 Block 3 -0.052 0.257 0.841 Block 5 0.017 0.249 0.945 ~~ Parasite Richness ~~ Parasite Abundance 0.179 NA 0.026 ~~ Host Richness ~~ Host Exotic Abundance -0.307 NA 1.000 ~~ Host Exotic Abundance ~~ Host Phylogenetic Diversity -0.194 NA 0.982

126

Table D6.22 – Coefficient estimates from the final (reduced) model. Estimates are standardized to a common scale to facilitate comparisons. Significant predictors (p<0.05) are indicated in bold. Correlations among dependent variables are indicated by ~~.

Response Predictor Estimate Std Error p Parasite Richness Host Richness 0.631 0.080 0.000 Block 5 -0.201 0.076 0.010 Block 2 -0.189 0.073 0.012 Host Phylogenetic Diversity -0.140 0.074 0.061 Host Exotic Abundance 0.093 0.086 0.282 Parasite Abundance Block 3 0.381 0.078 0.000 Host Exotic Abundance 0.260 0.087 0.004 Block 2 -0.186 0.078 0.020 Host Phylogenetic Diversity -0.162 0.079 0.043 Host Richness 0.030 0.083 0.724 Host Richness Resources -1.054 0.138 0.000 Block 3 -0.341 0.086 0.000 Block 4 -0.329 0.088 0.000 Block 5 -0.291 0.088 0.001 Diversity 0.344 0.140 0.015 Block 2 -0.177 0.088 0.046 Host Exotic Resources 0.857 0.132 0.000 Abundance Block 5 -0.359 0.073 0.000 Block 4 -0.241 0.073 0.001 Diversity -0.396 0.149 0.009 Block 2 -0.158 0.073 0.034 Host Phylogenetic Diversity 0.576 0.180 0.002 Diversity Resources -0.070 0.156 0.654 ~~ Parasite Richness ~~ Parasite Abundance 0.148 NA 0.056 ~~ Host Richness ~~ Host Exotic Abundance -0.301 NA 1.000 ~~ Host Exotic Abundance ~~ Host Phylogenetic Diversity -0.208 NA 0.988

127

D2. Supplemental figures

Figure D6.5 – The final (reduced) model with parasite richness rarefied to two host individuals per plot. A subsample of two host individuals represents the minimum number of host individuals sampled per plot. Dashed lines are non-significant (p

128

Figure D6.6 – The final (reduced) model with parasite richness rarefied to five host individuals per plot. A subsample of five host individuals represents the median number of host individuals sampled per plot. Dashed lines are non-significant (p

129

Figure D6.7 – Piecewise structural equation model results for the full model. Dashed lines are non-significant (p> 0.05). All coefficients are standardized. Correlations between errors are denoted with double-headedFULL PATH arrows. *pMODEL = 0.057- PiecewiseSEM

Initial Host Diversity X Resource Diversity Resources supply to hosts

Host Host -0.31 Exotic Host -0.19 Phylogenetic Species Abundance Richness Diversity BLK 2

BLK 3

BLK 4

BLK 5

Parasite Parasite richness abundance

0.18

Dashed lines are non-significant Coefficients are standardized R2 is the marginal R2 * P = 0.056

130

REFERENCES

Adame-Avarez, R. M., J. Mendiola-Soto, and M. Heil. 2014. Order of arrival shifts endophyte- pathogen interactions in bean from resistance induction to disease facilitation. FEMS Microbiology Letters 355:100–107. Adams, D. B., B. H. Anderson, and R. G. Windon. 1989. Cross-immunity between Haemonchus contortus and Trichostrongylus colubriformis in sheep. International Journal for Parasitology 19:717–722. Al-Naimi, F. A., K. A. Garrett, and W. W. Bockus. 2005. Competition, facilitation, and niche differentiation in two foliar pathogens. Oecologia 143:449–457. Anderson May, R.M., R. M. 1991. Infectious diseases of humans : dynamics and control. Page Oxford University Press, London 1991. Wiley Online Library. Antonovics, J., M. Hood, and J. Partain. 2002. The ecology and genetics of a host shift: microbotryum as a model system. The American naturalist 160 Suppl:S40-53. Antonovics, J., Y. Iwasa, and M. P. Hassell. 1995. A Generalized Model of Parasitoid, Venereal, and Vector-Based Transmission Processes. The American Naturalist 145:661–675. Bates, D., M. Mächler, B. Bolker, and S. Walker. 2014. Fitting Linear Mixed-Effects Models using lme4. Journal of Statistical Software 67:1–48. Biere, A., and A. Goverse. 2016. Plant-Mediated Systemic Interactions Between Pathogens, Parasitic Nematodes, and Herbivores Above-and Belowground. Annu. Rev. Phytopathol 54:499–527. Bordes, F., and S. Morand. 2008. Helminth species diversity of mammals: parasite species richness is a host species attribute. Parasitology 135:1701–1705. Bordes, F., and S. Morand. 2009. Parasite diversity: an overlooked metric of parasite pressures? Oikos 118:801–806. Borer, E. T., W. S. Harpole, P. B. Adler, E. M. Lind, J. L. Orrock, E. W. Seabloom, and M. D. Smith. 2014a. Finding generality in ecology: a model for globally distributed experiments. Methods in Ecology and Evolution 5:65–73. Borer, E. T., A.-L. Laine, and E. W. Seabloom. 2016. A multiscale approach to plant disease using the metacommunity concept. Annual Review of Phytopathology 54:397–418. Borer, E. T., E. W. Seabloom, D. S. Gruner, W. S. Harpole, H. Hillebrand, E. M. Lind, P. B. Adler, J. Alberti, T. M. Anderson, J. D. Bakker, L. Biederman, D. Blumenthal, C. S. Brown, L. A. Brudvig, Y. M. Buckley, M. Cadotte, C. Chu, E. E. Cleland, M. J. Crawley, P. Daleo,

131

E. I. Damschen, K. F. Davies, N. M. DeCrappeo, G. Du, J. Firn, Y. Hautier, R. W. Heckman, A. Hector, J. HilleRisLambers, O. Iribarne, J. A. Klein, J. M. H. Knops, K. J. La Pierre, A. D. B. Leakey, W. Li, A. S. MacDougall, R. L. McCulley, B. A. Melbourne, C. E. Mitchell, J. L. Moore, B. Mortensen, L. R. O’Halloran, J. L. Orrock, J. Pascual, S. M. Prober, D. A. Pyke, A. C. Risch, M. Schuetz, M. D. Smith, C. J. Stevens, L. L. Sullivan, R. J. Williams, P. D. Wragg, J. P. Wright, and L. H. Yang. 2014b. Herbivores and nutrients control grassland plant diversity via light limitation. Nature 508:517–520. Boyd, I. L., P. H. Freer-Smith, C. A. Gilligan, and H. C. J. Godfray. 2013. The Consequence of Tree Pests and Diseases for Ecosystem Services. Science 342:1235773–1235773. Burbidge, J. B., L. Magee, and A. L. Robb. 1988. Alternative Transformations to Handle Extreme Values of the Dependent Variable. Journal of the American Statistical Association 83:123. Busby, P. E., K. G. Peay, and G. Newcombe. 2016. Common foliar fungi of Populus trichocarpa modify Melampsora rust disease severity. New Phytologist 209:1681–1692. Bush, A., K. Lafferty, J. Lotz, and A. Shostak. 1997. Parasitology meets ecology on its own terms: Margolis et al. Revisited. The Journal of parasitology 83:575–583. Cardinale, B. J., D. S. Srivastava, J. Emmett Duffy, J. P. Wright, A. L. Downing, M. Sankaran, and C. Jouseau. 2006. Effects of biodiversity on the functioning of trophic groups and ecosystems. Nature 443:989–992. Chapin, F. S., L. R. Walker, C. L. Fastie, and L. C. Sharman. 1994. Mechanism of Primary Succession Following Deglaciation at Glacier Bay, Alaska. Ecological Monographs 64:149–175. Chesson, P. 2000. Mechanisms of maintenance of species diversity. Annual review of Ecology and Systematics. Chung, E., E. Petit, J. Antonovics, A. B. Pedersen, and M. E. Hood. 2012. Variation in resistance to multiple pathogen species: anther smuts of Silene uniflora. Ecology and evolution 2:2304–14. Cipollini, D., S. Enright, M. B. Traw, and J. Bergelson. 2004. Salicylic acid inhibits jasmonic acid-induced resistance of Arabidopsis thaliana to Spodoptera exigua. Molecular Ecology 13:1643–1653. Civitello, D. J., J. Cohen, H. Fatima, N. T. Halstead, J. Liriano, T. a. McMahon, C. N. Ortega, E. L. Sauer, T. Sehgal, S. Young, and J. R. Rohr. 2015. Biodiversity inhibits parasites: Broad evidence for the dilution effect. Proceedings of the National Academy of Sciences 112:201506279.

132

Cobey, S., and M. Lipsitch. 2013. Pathogen diversity and hidden regimes of apparent competition. The American naturalist 181:12–24. Conrath, U., G. J. M. Beckers, V. Flors, P. García-Agustín, G. Jakab, F. Mauch, M.-A. Newman, C. M. J. Pieterse, B. Poinssot, M. J. Pozo, A. Pugin, U. Schaffrath, J. Ton, D. Wendehenne, L. Zimmerli, and B. Mauch-Mani. 2006. Priming: Getting Ready for Battle. Molecular Plant-Microbe Interactions 19:1062–1071. Cornell, H. V., and S. P. Harrison. 2013. Regional effects as important determinants of local diversity in both marine and terrestrial systems. Oikos 122:288–297. Costello, E. K., K. Stagaman, L. Dethlefsen, B. J. M. Bohannan, and D. A. Relman. 2012. The application of ecological theory toward an understanding of the human microbiome. Science (New York, N.Y.) 336:1255–62. Crawley, M. J. 2007. The R Book. Page Journal of the American Statistical Association. John Wiley & Sons, Ltd, West Sussex, England. Cronin, J. P., M. a Rúa, and C. E. Mitchell. 2014. Why is living fast dangerous? Disentangling the roles of resistance and tolerance of disease. The American naturalist 184:172–87. Dallas, T., and S. J. Presley. 2014. Relative importance of host environment, transmission potential and host phylogeny to the structure of parasite metacommunities. Oikos 123:866– 874. Daszak, P., a a Cunningham, and a D. Hyatt. 2000. Emerging infectious diseases of wildlife-- threats to biodiversity and human health. Science (New York, N.Y.) 287:443–9. Daubenmire, R. 1959. A canopy-coverage method of vegetational analysis. Northwest Science 33:43–64. Davis, M. A., J. P. Grime, and K. Thompson. 2000. Fluctuating resources in plant communities: a general theory of invasibility. Journal of Ecology 88:528–534. Davis, M. B., and R. G. Shaw. 2001. Range Shifts and Adaptive Responses to Quaternary Climate Change. Science 292:673–679. DeMalach, N., E. Zaady, and R. Kadmon. 2017. Light asymmetry explains the effect of nutrient enrichment on grassland diversity. Ecology Letters 20:60–69. Dobson, A. 2004. Population Dynamics of Pathogens with Multiple Host Species. The American Naturalist 164:S64–S78. Dobson, A., K. D. Lafferty, A. M. Kuris, R. F. Hechinger, and W. Jetz. 2008. Homage to Linnaeus: How many parasites? How many hosts? Proceedings of the National Academy of Sciences 105:11482–11489.

133

Dooley, A., F. Isbell, L. Kirwan, J. Connolly, J. A. Finn, and C. Brophy. 2015. Testing the effects of diversity on ecosystem multifunctionality using a multivariate model. Ecology Letters 18:1242–1251. Douma, J. C., P. J. Vermeulen, E. H. Poelman, M. Dicke, and N. P. R. Anten. 2017. When does it pay off to prime for defense? A modeling analysis. New Phytologist 216:782–797. Dove, A. D. M., and T. H. Cribb. 2006. Species accumulation curves and their applications in parasite ecology. Trends in Parasitology 22:568–574. Dove, A. D. 1999. A new index of interactivity in parasite communities. International journal for parasitology 29:915–20. Dumbrell, A. J., P. D. Ashton, N. Aziz, G. Feng, M. Nelson, C. Dytham, A. H. Fitter, and T. Helgason. 2011. Distinct seasonal assemblages of arbuscular mycorrhizal fungi revealed by massively parallel pyrosequencing. New Phytologist 190:794–804. Esch, G. W., L. a Curtis, and M. a Barger. 2001. A perspective on the ecology of trematode communities in snails. Parasitology 123 Suppl:S57-75. Eswarappa, S. M., S. Estrela, and S. P. Brown. 2012, June 21. Within-host dynamics of multi- species infections: Facilitation, competition and virulence. Public Library of Science. Ezenwa, V. O. 2016. Helminth–microparasite co-infection in wildlife: lessons from ruminants, rodents and rabbits. Parasite Immunology 38:527–534. Ezenwa, V. O., R. S. Etienne, G. Luikart, A. Beja-Pereira, and A. E. Jolles. 2010. Hidden consequences of living in a wormy world: nematode-induced immune suppression facilitates tuberculosis invasion in African buffalo. The American naturalist 176:613–24. Fargione, J. E., and D. Tilman. 2005. Diversity decreases invasion via both sampling and complementarity effects. Ecology Letters 8:604–611. Fenton, A., S. C. L. Knowles, O. L. Petchey, and A. B. Pedersen. 2014. The reliability of observational approaches for detecting interspecific parasite interactions: comparison with experimental results. International journal for parasitology 44:437–45. Fenton, A., M. E. Viney, and J. Lello. 2010. Detecting interspecific macroparasite interactions from ecological data: patterns and process. Ecology letters 13:606–15. Fierer, N., S. Ferrenberg, G. E. Flores, A. González, J. Kueneman, T. Legg, R. C. Lynch, D. McDonald, J. R. Mihaljevic, S. P. O’Neill, M. E. Rhodes, S. J. Song, and W. A. Walters. 2012. From Animalcules to an Ecosystem: Application of Ecological Concepts to the Human Microbiome. Annual Review of Ecology, Evolution, and Systematics 43:137–155. Fridley, J. D. 2008. Of Asian Forests and European Fields: Eastern U.S. Plant Invasions in a

134

Global Floristic Context. PLoS ONE 3:e3630. Fukami, T. 2015. Historical contingency in community assembly : integrating niches, species pools, and priority effects. Annual Review of Ecology Evolution and Systematics 46:1–23. Fukami, T., E. A. Mordecai, and A. Ostling. 2016. A framework for priority effects. Journal of Vegetation Science 27:655–657. Fulton, R. W. 1986. Practices and Precautions in the Use of Cross Protection for Plant Virus Disease Control. Annual Review of Phytopathology 24:67–81. Gelman, A., and J. Hill. 2007. Data analysis using regression and multilevel/hierarchical models. Cambridge University Press New York, NY, USA. Gilbert, G. S., and C. O. Webb. 2007. Phylogenetic signal in plant pathogen-host range. Proceedings of the National Academy of Sciences of the United States of America 104:4979–83. Glazebrook, J. 2005. Contrasting Mechanisms of Defense Against Biotrophic and Necrotrophic Pathogens. Annual Review of Phytopathology 43:205–227. Gotelli, N. J., and R. K. Colwell. 2001. Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecology Letters 4:379–391. Graham, A. L. 2008. Ecological rules governing helminth-microparasite coinfection. Proceedings of the National Academy of Sciences of the United States of America 105:566–70. Griffiths, E. C., A. B. Pedersen, A. Fenton, and O. L. Petchey. 2014. Analysis of a summary network of co-infection in humans reveals that parasites interact most via shared resources. Proceedings of the Royal Society B: Biological Sciences 281:20132286–20132286. Halliday, F. W., R. W. Heckman, P. A. Wilfahrt, and C. E. Mitchell. 2017a. A multivariate test of disease risk reveals conditions leading to disease amplification. Proceedings of the Royal Society B: Biological Sciences 284:20171340. Halliday, F. W., J. Umbanhowar, and C. E. Mitchell. 2017b. Interactions among symbionts operate across scales to influence parasite epidemics. Ecology Letters 20:1285–1294. Han, B. A., J. L. Kerby, C. L. Searle, A. Storfer, and A. R. Blaustein. 2015. Host species composition influences infection severity among amphibians in the absence of spillover transmission. Ecology and Evolution 5:1432–1439. Harpole, W. S., L. L. Sullivan, E. M. Lind, J. Firn, P. B. Adler, E. T. Borer, J. Chase, P. A. Fay, Y. Hautier, H. Hillebrand, A. S. MacDougall, E. W. Seabloom, R. Williams, J. D. Bakker, M. W. Cadotte, E. J. Chaneton, C. Chu, E. E. Cleland, C. D’Antonio, K. F. Davies, D. S.

135

Gruner, N. Hagenah, K. Kirkman, J. M. H. Knops, K. J. La Pierre, R. L. McCulley, J. L. Moore, J. W. Morgan, S. M. Prober, A. C. Risch, M. Schuetz, C. J. Stevens, and P. D. Wragg. 2016. Addition of multiple limiting resources reduces grassland diversity. Nature 537:93–96. Hatcher, M., J. Dick, and A. Dunn. 2012. Disease emergence and invasions. Functional Ecology. Heckman, R. W., F. W. Halliday, P. A. Wilfahrt, and C. E. Mitchell. 2017. Effects of native diversity, soil nutrients, and natural enemies on exotic invasion in experimental plant communities. Ecology 97:3337–3345. Heckman, R. W., J. P. Wright, and C. E. Mitchell. 2016. Joint effects of nutrient addition and enemy exclusion on exotic plant success. Ecology 97:3337–3345. Hector, A., and R. Bagchi. 2007. Biodiversity and ecosystem multifunctionality. Nature 448:188–190. Hector, A., E. Bazeley-White, M. Loreau, S. Otway, and B. Schmid. 2002. Overyielding in grassland communities: testing the sampling effect hypothesis with replicated biodiversity experiments. Ecology Letters 5:502–511. Heger, T., and J. M. Jeschke. 2014. The enemy release hypothesis as a hierarchy of hypotheses. Oikos 123:741–750. Hellard, E., D. Fouchet, F. Vavre, and D. Pontier. 2015. Parasite-Parasite Interactions in the Wild: How To Detect Them? Elsevier Ltd. Herms, D. A., and W. J. Mattson. 1992. The Dilemma of Plants: To Grow or Defend. The Quarterly Review of Biology 67:283–335. Hersh, M. H., R. Vilgalys, and J. S. Clark. 2012. Evaluating the impacts of multiple generalist fungal pathogens on temperate tree seedling survival. Ecology 93:511–20. HilleRisLambers, J., P. B. Adler, W. S. Harpole, J. M. Levine, and M. M. Mayfield. 2011. Rethinking Community Assembly Through the Lens of Coexistence Theory. Annual Review of Ecology, Evolution, and Systematics 43:120830113150004. Holmes, J. C., and P. W. Price. 1986. Communities of parasites. Pages 187–213in D. J. Anderson and J. Kikkawa, editors.Community ecology: patterns and processes. Blackwell Scientific., Oxford, U.K. Holyoak, M., M. a. Leibold, and R. D. Holt. 2005. Metacommunities: Spatial Dynamics and Ecological Communities. University of Chicago Press. Hoverman, J. T., B. J. Hoye, and P. T. J. Johnson. 2013. Does timing matter? How priority effects influence the outcome of parasite interactions within hosts. Oecologia 173:1471–

136

1480. Huenneke, L. F., S. P. Hamburg, R. Koide, H. A. Mooney, and P. M. Vitousek. 1990. Effects of Soil Resources on Plant Invasion and Community Structure in Californian Serpentine Grassland. Ecology 71:478–491. van Hulten, M., M. Pelser, L. C. van Loon, C. M. J. Pieterse, and J. Ton. 2006. Costs and benefits of priming for defense in Arabidopsis. Proceedings of the National Academy of Sciences 103:5602–5607. Imai, K., L. Keele, and D. Tingley. 2010. A general approach to causal mediation analysis. Psychological Methods 15:309–334. James, C. 1971. A manual of assessment keys for plant diseases. Page A manual of assessment keys for plant diseases. American Phytopathological Society, St. Paul, MN. Jenner, E. 1923. An Inquiry into the Causes and Effects of the Variolae Vaccinae, a Disease Discovered in Some of the Western Countries of England, Particularly Gloucestershire, and Known by the Name of “The Cow Pox” (1798). Page Reprinted by Milan: R Lier & Co. Milan, R. Lier & Co., London. Johnson, P. T. J., and J. T. Hoverman. 2012. Parasite diversity and coinfection determine pathogen infection success and host fitness. Proceedings of the National Academy of Sciences of the United States of America 109:9006–11. Johnson, P. T. J., R. S. Ostfeld, and F. Keesing. 2015a. Frontiers in research on biodiversity and disease. Ecology Letters 18:1119–1133. Johnson, P. T. J., D. L. Preston, J. T. Hoverman, and B. E. Lafonte. 2013a. Host and parasite diversity jointly control disease risk in complex communities. Proceedings of the National Academy of Sciences of the United States of America 110:16916–21. Johnson, P. T. J., D. L. Preston, J. T. Hoverman, and K. L. D. Richgels. 2013b. Biodiversity decreases disease through predictable changes in host community competence. Nature 494:230–3. Johnson, P. T. J., J. R. Rohr, J. T. Hoverman, E. Kellermanns, J. Bowerman, and K. B. Lunde. 2012. Living fast and dying of infection: host life history drives interspecific variation in infection and disease risk. Ecology letters 15:235–42. Johnson, P. T. J., J. C. de Roode, and A. Fenton. 2015b. Why infectious disease research needs community ecology. Science 349:1259504. Johnson, P. T. J., J. C. de Roode, and A. Fenton. 2015c. Why infectious disease research needs community ecology. Science 349.

137

Johnson, P. T. J., C. L. Wood, M. B. Joseph, D. L. Preston, S. E. Haas, and Y. P. Springer. 2016, July. Habitat heterogeneity drives the host-diversity-begets-parasite-diversity relationship: evidence from experimental and field studies. Joseph, M. B., J. R. Mihaljevic, S. A. Orlofske, and S. H. Paull. 2013. Does life history mediate changing disease risk when communities disassemble? Ecology Letters 16:1405–1412. Kamiya, T., K. O’Dwyer, S. Nakagawa, and R. Poulin. 2013. What determines species richness of parasitic organisms? A meta-analysis across animal, plant and fungal hosts. Biological reviews of the Cambridge Philosophical Society. Kamiya, T., K. O’Dwyer, S. Nakagawa, and R. Poulin. 2014. Host diversity drives parasite diversity: meta-analytical insights into patterns and causal mechanisms. Ecography 37:689– 697. van Kan, J. a L. 2006. Licensed to kill: the lifestyle of a necrotrophic plant pathogen. Trends in plant science 11:247–53. Karasov, T. L., E. Chae, J. J. Herman, and J. Bergelson. 2017. Mechanisms to Mitigate the Trade-Off between Growth and Defense. The Plant cell 29:666–680. Keesing, F., R. D. Holt, and R. S. Ostfeld. 2006. Effects of species diversity on disease risk. Ecology letters 9:485–98. Kembel, S. W., D. D. Ackerly, S. P. Blomberg, W. K. Cornwell, P. D. Cowan, M. R. Helmus, H. Morlon, and C. O. Webb. 2009. R tools for integrating phylgenies and ecology. Http:/Picante.R-Forge.R-Project.Org R package:1463–1464. Kennedy, P. G., K. G. Peay, and T. D. Bruns. 2009. Root tip competition among ectomycorrhizal fungi: Are priority effects a rule or an exception? Ecology 90:2098–2107. Kirchner, J. W., and C. Neal. 2013. Universal fractal scaling in stream chemistry and its implications for solute transport and water quality trend detection. Proceedings of the National Academy of Sciences of the United States of America 110:12213–8. Klemme, I., K. R. Louhi, and A. Karvonen. 2016. Host infection history modifies co-infection success of multiple parasite genotypes. Journal of Animal Ecology 85:591–597. Kliebenstein, D. J., and H. C. Rowe. 2008. Ecological costs of biotrophic versus necrotrophic pathogen resistance, the hypersensitive response and signal transduction. Plant Science 174:551–556. Kline, R. B. 2010. Principles and practice of structural equation modeling. 2011. Page New York: Guilford Press Google Scholar. Guilford publications. Koornneef, A., A. Leon-Reyes, T. Ritsema, A. Verhage, F. C. Den Otter, L. C. Van Loon, and C.

138

M. J. Pieterse. 2008. Kinetics of Salicylate-Mediated Suppression of Jasmonate Signaling Reveal a Role for Redox Modulation. Plant Physiology 147:1358–1368. Kuris, A. M., A. R. Blaustein, and J. J. Alio. 1980. Hosts as Islands. The American Naturalist 116:570–586. Kuris, A. M., and K. D. Lafferty. 1994. Community Structure: Larval Trematodes in Snail Hosts. Annual Review of Ecology and Systematics 25:189–217. Lafferty, K. D., and A. M. Kuris. 2002. Trophic strategies, animal diversity and body size. Tree 17:507–513. Laine, A. L. 2011. Context-dependent effects of induced resistance under co-infection in a plant- pathogen interaction. Evolutionary Applications 4:696–707. Lefcheck, J. S. 2016. piecewiseSEM : Piecewise structural equation modelling in r for ecology, evolution, and systematics. Methods in Ecology and Evolution 7:573–579. Lefcheck, J. S., J. E. K. Byrnes, F. Isbell, L. Gamfeldt, J. N. Griffin, N. Eisenhauer, M. J. S. Hensel, A. Hector, B. J. Cardinale, and J. E. Duffy. 2015. Biodiversity enhances ecosystem multifunctionality across trophic levels and habitats. Nature Communications 6:6936. Leibold, M. a., M. Holyoak, N. Mouquet, P. Amarasekare, J. M. Chase, M. F. Hoopes, R. D. Holt, J. B. Shurin, R. Law, D. Tilman, M. Loreau, and a. Gonzalez. 2004. The metacommunity concept: a framework for multi-scale community ecology. Ecology Letters 7:601–613. Lello, J., B. Boag, A. Fenton, I. R. Stevenson, and P. J. Hudson. 2004. Competition and mutualism among the gut helminths of a mammalian host. Nature 428:840–844. Lenth, R. V. 2016. Least-Squares Means: The {R} Package {lsmeans}. Journal of Statistical Software 69:1–33. Leventhal, G. E., A. L. Hill, M. A. Nowak, and S. Bonhoeffer. 2015. Evolution and emergence of infectious diseases in theoretical and real-world networks. Nature Communications 6:6101. Levine, J. M., and C. M. D’Antonio. 1999. Elton Revisited: A Review of Evidence Linking Diversity and Invasibility. Oikos 87:15. Limberger, R., and S. A. Wickham. 2011. Competition-colonization trade-offs in a ciliate model community. Oecologia 167:723–732. Lind, E. M., E. Borer, E. Seabloom, P. Adler, J. D. Bakker, D. M. Blumenthal, M. Crawley, K. Davies, J. Firn, D. S. Gruner, W. Stanley Harpole, Y. Hautier, H. Hillebrand, J. Knops, B. Melbourne, B. Mortensen, A. C. Risch, M. Schuetz, C. Stevens, and P. D. Wragg. 2013.

139

Life-history constraints in grassland plant species: a growth-defence trade-off is the norm. Ecology Letters 16:513–521. Liu, X., S. Lyu, D. Sun, C. Bradshaw, and S. Zhou. 2017. Species decline under nitrogen fertilization increases community-level competence of fungal diseases. Proceedings of the Royal Society B: Biological Sciences 284:20162621. Liu, X., S. Lyu, S. Zhou, and C. J. A. Bradshaw. 2016. Warming and fertilization alter the dilution effect of host diversity on disease severity. Ecology 97:1680–1689. Liu, Y., Z. Kang, and H. Buchenauer. 2006. Ultrastructural and Immunocytochemical Studies on Effects of Barley Yellow Dwarf Virus - Infection on Fusarium Head Blight, Caused by Fusarium graminearum, in Wheat Plants. Journal of Phytopathology 154:6–15. Livingston, G., M. Matias, V. Calcagno, C. Barbera, M. Combe, M. A. Leibold, and N. Mouquet. 2012. Competition-colonization dynamics in experimental bacterial metacommunities. Nature Communications 3:1234. LoGiudice, K., R. S. Ostfeld, K. A. Schmidt, and F. Keesing. 2003. The ecology of infectious disease: effects of host diversity and community composition on Lyme disease risk. Proceedings of the National Academy of Sciences of the United States of America 100:567–71. Logue, J. B., N. Mouquet, H. Peter, and H. Hillebrand. 2011. Empirical approaches to metacommunities: a review and comparison with theory. Trends in ecology & evolution 26:482–91. Van Loon, L. C. 1997. Induced resistance in plants and the role of pathogenesis-related proteins. European Journal of Plant Pathology 103:753–765. MacArthur, R. 1958. Population ecology of some warblers of northeastern coniferous forests. Ecology. Madden, L. V., G. Hughes, and F. van den Bosch. 2007. The study of plant disease epidemics. American Phytopathological Society. Manley, R., M. Boots, and L. Wilfert. 2015, April. Emerging viral disease risk to pollinating insects: Ecological, evolutionary and anthropogenic factors. Mauch-Mani, B., I. Baccelli, E. Luna, and V. Flors. 2017. Defense Priming: An Adaptive Part of Induced Resistance. Annual Review of Plant Biology 68:485–512. Mayfield, M. M., and J. M. Levine. 2010. Opposing effects of competitive exclusion on the phylogenetic structure of communities. Ecol. Lett. 13:1085–1093. McCallum, H., N. Barlow, and J. Hone. 2001. How should pathogen transmission be modelled?

140

Trends in ecology & evolution 16:295–300. Mendgen, K., and M. Hahn. 2002. Plant infection and the establishment of fungal biotrophy. Trends in Plant Science 7:352–356. de Mendiburu, F., and M. F. de Mendiburu. 2016. Package “agricolae.” Statistical procedures for agricultural research. Mideo, N. 2009. Parasite adaptations to within-host competition. Trends in parasitology 25:261– 8. Mihaljevic, J. R. 2012. Linking metacommunity theory and symbiont evolutionary ecology. Trends in ecology & evolution (Personal edition) 27:323–329. Mihaljevic, J. R., M. B. Joseph, S. A. Orlofske, S. H. Paull, and M. Killilea. 2014. The Scaling of Host Density with Richness Affects the Direction, Shape, and Detectability of Diversity- Disease Relationships. PLoS ONE 9:e97812. Mitchell, C. E., D. Blumenthal, V. Jarošík, E. E. Puckett, and P. Pyšek. 2010. Controls on pathogen species richness in plants’ introduced and native ranges: roles of residence time, range size and host traits. Ecology letters 13:1525–35. Mitchell, C. E., and A. G. Power. 2003. Release of invasive plants from fungal and viral pathogens. Nature 421:625–7. Mitchell, C. E., P. B. Reich, D. Tilman, and J. V. Groth. 2003. Effects of elevated CO 2 , nitrogen deposition, and decreased species diversity on foliar fungal plant disease. Global Change Biology 9:438–451. Mitchell, C., D. Tilman, and J. Groth. 2002. Effects of grassland plant species diversity, abundance, and composition on foliar fungal disease. Ecology 83:1713–1726. Morand, S., S. Jittapalapong, Y. Suputtamongkol, M. T. Abdullah, and T. B. Huan. 2014. Infectious Diseases and Their Outbreaks in Asia-Pacific: Biodiversity and Its Regulation Loss Matter. PLoS ONE 9:e90032. Mordecai, E. A. 2011. Pathogen impacts on plant communities: unifying theory, concepts, and emirical work. Ecological Monographs 81:429–441. Mordecai, E. A., K. Gross, and C. E. Mitchell. 2016. Within-Host Niche Differences and Fitness Trade-offs Promote Coexistence of Plant Viruses. The American Naturalist 187:E13–E26. Mouquet, N., P. Munguia, J. M. Kneitel, and T. E. Miller. 2003. Community assembly time and the relationship between local and regional species richness. Oikos 103:618–626. Mundt, C. C., L. S. Brophy, and M. S. Schmitt. 1995. Disease severity and yield of pure-line wheat cultivars and mixtures in the presence of eyespot, yellow rust, and their combination.

141

Plant Pathology 44:173–182. Mundt, C. C., and K. E. Sackett. 2012. Spatial scaling relationships for spread of disease caused by a wind-dispersed plant pathogen. Ecosphere (Washington, D.C) 3:art24. Mundt, C. C., K. E. Sackett, L. D. Wallace, C. Cowger, and J. P. Dudley. 2009. Aerial dispersal and multiple-scale spread of epidemic disease. EcoHealth 6:546–52. Natsopoulou, M. E., D. P. McMahon, V. Doublet, J. Bryden, R. J. Paxton, M. Woolhouse, D. Haydon, R. Antia, T. Rigaud, M. Perrot-Minnot, M. Brown, F. Cox, A. Read, L. Taylor, A. Griffin, S. West, A. Buckling, S. West, A. Buckling, R. Poulin, A. Escribano, T. Williams, D. Goulson, R. Cave, J. Chapman, P. Cabellero, S. Brown, M. Hochberg, B. Grenfell, R. Paul, V. Nu, A. Krettli, P. Brey, N. Mideo, S. Telfer, X. Lambin, R. Birtles, P. Beldomenico, S. Burthe, S. Paterson, M. Begon, A. Blackwell, M. Martin, H. Kaplan, M. Gurven, L. Abu-Raddad, P. Patnaik, J. Kublin, J. Lello, B. Boag, A. Fenton, I. Stevenson, P. Hudson, L. Pollitt, T. Churcher, E. Dawes, S. Khan, M. Sajid, M. Basáñez, N. Colegrave, S. Reece, L. Taylor, D. Walliker, A. Read, L. Chao, K. Hanley, C. Burch, C. Dahlberg, P. Turner, M. Thomas, E. Watson, P. Valverde-Garcia, F. Ben-Ami, T. Rigaud, D. Ebert, J. Lohr, M. Yin, J. Wolinska, J. Jackson, R. Pleass, J. Cable, J. Bradley, R. Tinsley, A. Karvonen, O. Seppälä, E. T. Valtonen, R. Alford, H. Wilbur, J. Hoverman, B. Hoye, P. Johnson, K. Haag, E. Traunecker, D. Ebert, P. Keeling, N. Fast, C. Texier, C. Vidau, B. Viguès, H. El Alaoui, F. Delbac, M. Higes, R. Martín-Hernández, A. Meana, M. Higes, A. Meana, C. Bartolomé, C. Botías, R. Martín-Hernández, R. Paxton, J. Klee, S. Korpela, I. Fries, I. Fries, F. Feng, A. da Silva, S. Slemenda, N. Pieniazek, I. Fries, S. Gisder, K. Hedtke, N. Möckel, M. Frielitz, A. Linde, E. Genersch, E. Forsgren, I. Fries, I. Fries, L. Malone, D. Stefanovic, E. Forsgren, I. Fries, W. Huang, L. Solter, I. Fries, D. VanEngelsdorp, A. Field, J. Miles, Z. Field, D. Bates, M. Maechler, T. Hothorn, F. Bretz, P. Westfall, M. Higes, M. Higes, R. Martín, A. Meana, D. Pilarska, L. Solter, M. Kereselidze, A. Linde, G. Hoch, G. Hoch, A. Schopf, J. Maddox, R. Malakar, J. Elkinton, A. Hajek, J. Burand, M. Higes, P. García-Palencia, R. Martín-Hernández, A. Meana, A. Graham, J. de Roode, M. Helinski, M. Anwar, A. Read, C. Dussaubat, R. Martín-Hernández, C. Botías, L. Barrios, A. Martínez-Salvador, A. Meana, C. Mayack, M. Higes, G. Williams, D. Shutler, K. Burgher-MacLellan, R. Rogers, K. Antúnez, R. Martín-Hernández, L. Prieto, A. Meana, P. Zunino, M. Higes, R. Martín-Hernández, C. Botías, E. Bailón, A. Martínez-Salvador, L. Prieto, A. Meana, M. Higes, J. Fernández, F. Puerta, M. Cousinou, R. Dios-Palomares, F. Campano, L. Redondo, M. Meixner, and J. Klee. 2015. Interspecific competition in honeybee intracellular gut parasites is asymmetric and favours the spread of an emerging infectious disease. Proceedings. Biological sciences / The Royal Society 282:20141896. Newton, A. C., B. D. L. Fitt, S. D. Atkins, D. R. Walters, and T. J. Daniell. 2010. Pathogenesis, parasitism and mutualism in the trophic space of microbe–plant interactions. Trends in

142

Microbiology 18:365–373. O’Regan, S. M., J. E. Vinson, and A. W. Park. 2015. Interspecific Contact and Competition May Affect the Strength and Direction of Disease-Diversity Relationships for Directly Transmitted Microparasites. The American Naturalist 186:480–494. Pańka, D., C. P. West, C. A. Guerber, and M. D. Richardson. 2013. Susceptibility of tall fescue to Rhizoctonia zeae infection as affected by endophyte symbiosis. Annals of Applied Biology 163:257–268. Parker, I. M., M. Saunders, M. Bontrager, A. P. Weitz, R. Hendricks, R. Magarey, K. Suiter, and G. S. Gilbert. 2015. Phylogenetic structure and host abundance drive disease pressure in communities. Nature 520:542–544. Pavoine, S., and M. B. Bonsall. 2011. Measuring biodiversity to explain community assembly: a unified approach. Biological Reviews 86:792–812. Pearse, W. D., and A. Purvis. 2013. phyloGenerator: an automated phylogeny generation tool for ecologists. Methods in Ecology and Evolution 4:692–698. Pedersen, A. B., and A. Fenton. 2015. The role of antiparasite treatment experiments in assessing the impact of parasites on wildlife. Trends in parasitology 31:200–11. Penczykowski, R. M., A. L. Laine, and B. Koskella. 2016. Understanding the ecology and evolution of host-parasite interactions across scales. Evolutionary Applications 9:37–52. Pieterse, C. M. J., D. Van der Does, C. Zamioudis, A. Leon-Reyes, and S. C. M. Van Wees. 2012. Hormonal modulation of plant immunity. Annual review of cell and developmental biology 28:489–521. Pieterse, C. M. J., C. Zamioudis, R. L. Berendsen, D. M. Weller, S. C. M. Van Wees, and P. A. H. M. Bakker. 2014. Induced Systemic Resistance by Beneficial Microbes. Annual Review of Phytopathology 52:347–375. Pinheiro, J., D. Bates, S. DebRoy, D. Sarkar, and R Core Team. 2016. {nlme}: Linear and Nonlinear Mixed Effects Models. Plowright, R. K., P. Eby, P. J. Hudson, I. L. Smith, D. Westcott, W. L. Bryden, D. Middleton, P. A. Reid, R. A. McFarlane, G. Martin, G. M. Tabor, L. F. Skerratt, D. L. Anderson, G. Crameri, D. Quammen, D. Jordan, P. Freeman, L.-F. Wang, J. H. Epstein, G. A. Marsh, N. Y. Kung, and H. McCallum. 2014. Ecological dynamics of emerging bat virus spillover. Proceedings of the Royal Society of London B: Biological Sciences 282. Potter, L. R. 1982. Interaction between barley yellow dwarf virus and rust in wheat, barley and oats, and the effects on grain yield and quality. Annals of Applied Biology 100:321–329.

143

Potter, L. R. 1980. The effects of barley yellow dwarf virus and powdery mildew in oats and barley with single and dual infections. Annals of Applied Biology 94:11–17. Poulin, R. 2001. Interactions between species and the structure of helminth communities. Parasitology 122 Suppl:S3-11. Poulin, R. 2007. Are there general laws in parasite ecology? Parasitology 134:763–76. Poulin, R., B. R. Krasnov, and D. Mouillot. 2011. Host specificity in phylogenetic and geographic space. Trends in parasitology 27:355–61. Poulin, R., and S. Morand. 2011. The Diversity of Parasites 75:277–293. Power, A. G., and C. E. Mitchell. 2004. Pathogen spillover in disease epidemics. The American naturalist 164 Suppl:S79–S89. R Core Team. 2015. R: a language and environment for statistical computing | GBIF.ORG. Vienna, Austria. Randall, J., J. Cable, I. a Guschina, J. L. Harwood, and J. Lello. 2013. Endemic infection reduces transmission potential of an epidemic parasite during co-infection. Proceedings. Biological sciences / The Royal Society 280:20131500. Randolph, S. E., and A. D. M. Dobson. 2012. Pangloss revisited: a critique of the dilution effect and the biodiversity-buffers-disease paradigm. Parasitology 139:847–863. Reich, P. B., J. Knops, D. Tilman, J. Craine, D. Ellsworth, M. Tjoelker, T. Lee, D. Wedin, S. Naeem, D. Bahauddin, G. Hendrey, S. Jose, K. Wrage, J. Goth, and W. Bengston. 2001. Plant diversity enhances ecosystem responses to elevated CO2 and nitrogen deposition. Nature 410:809–810. Richgels, K. L. D., J. T. Hoverman, and P. T. J. Johnson. 2013. Evaluating the role of regional and local processes in structuring a larval trematode metacommunity of Helisoma trivolvis. Ecography 36:854–863. Ricklefs, R. E. 1987. Community Diversity: Relative Roles of Local and Regional Processes. Science 235:167–171. Roff, D. 1993. The evolution of life histories: Theory and analysis. Springer Science & Business Media. Rohani, P., C. J. Green, N. B. Mantilla-Beniers, and B. T. Grenfell. 2003. Ecological interference between fatal diseases. Nature 422:885–888. Rosenzweig, M. L. 1971. Paradox of enrichment: destabilization of exploitation ecosystems in ecological time. Science 171:385–387.

144

Rottstock, T., J. Joshi, V. Kummer, and M. Fischer. 2014. Higher plant diversity promotes higher diversity of fungal pathogens, while it decreases pathogen infection per plant. Ecology 95:1907–17. Roy, B. a, K. Hudson, M. Visser, and B. R. Johnson. 2014. Grassland fires may favor native over introduced plants by reducing pathogen loads. Ecology 95:1897–906. Rudolf, V. H. W., and J. Antonovics. 2005. Species coexistence and pathogens with frequency- dependent transmission. The American naturalist 166:112–118. Rynkiewicz, E. C., A. B. Pedersen, and A. Fenton. 2015. An ecosystem approach to understanding and managing within-host parasite community dynamics. Trends in Parasitology 31:212–221. Saikkonen, K., P. E. Gundel, and M. Helander. 2013. Chemical Ecology Mediated by Fungal Endophytes in Grasses. Journal of chemical ecology 39:962–8. Salkeld, D. J., K. A. Padgett, and J. H. Jones. 2013. A meta-analysis suggesting that the relationship between biodiversity and risk of zoonotic pathogen transmission is idiosyncratic. Ecology Letters 16:679–686. Schmid, B., A. Hector, M. A. Huston, P. Inchausti, I. Nijs, P. W. Leadley, and D. Tilman. 2002. The design and analysis of biodiversity experiments. Pages 61–75Biodiversity and ecosystem functioning: synthesis and perspectives. Oxford University Press, Oxford. Schmidt, S. K., E. K. Costello, D. R. Nemergut, C. C. Cleveland, S. C. Reed, M. N. Weintraub, A. F. Meyer, and A. M. Martin. 2007. BIOGEOCHEMICAL CONSEQUENCES OF RAPID MICROBIAL TURNOVER AND SEASONAL SUCCESSION IN SOIL. Ecology 88:1379–1385. Seabloom, E. W., E. T. Borer, K. Gross, A. E. Kendig, C. Lacroix, C. E. Mitchell, E. A. Mordecai, and A. G. Power. 2015. The community ecology of pathogens: Coinfection, coexistence and community composition. Ecology Letters 18:401–415. Shipley, B. 2009. Confirmatory path analysis in a generalized multilevel context. Ecology 90:363–368. Simko, I., and H. Piepho. 2012. The Area Under the Disease Progress Stairs: Calculation, Advantage, and Application. Analytical and Theoretical Plant Pathology 102:381–389. Sousa, W. P. 1993. Interspecific antagonism and species coexistence in a diverse guild of larval trematode parasites. Ecological Monographs 63:103–128. Spoel, S. H., J. S. Johnson, and X. Dong. 2007. Regulation of tradeoffs between plant defenses against pathogens with different lifestyles. Proceedings of the National Academy of Sciences of the United States of America 104:18842–7.

145

Stamp, N. 2003. Out Of The Quagmire Of Plant Defense Hypotheses. The Quarterly Review of Biology 78:23–55. Starr, M. P. 1975. A generalized scheme for classifying organismic associations. Pages 1– 20Symposia of the Society for Experimental Biology. Stearns, S. C. 1992. The evolution of life histories. Oxford University Press Oxford. Strauss, A. T., M. S. Shocket, D. J. Civitello, J. L. Hite, R. M. Penczykowski, M. A. Duffy, C. E. Cáceres, and S. R. Hall. 2016. Habitat, predators, and hosts regulate disease in Daphnia through direct and indirect pathways. Ecological Monographs 86:393–411. Stricker, K. B., P. F. Harmon, E. M. Goss, K. Clay, and S. Luke Flory. 2016. Emergence and accumulation of novel pathogens suppress an invasive species. Ecology Letters 19:469–477. Susi, H., B. Barrès, P. F. Vale, and A.-L. Laine. 2015. Co-infection alters population dynamics of infectious disease. Nature communications 6:5975. Sutter, R., and C. Müller. 2011. Mining for treatment-specific and general changes in target compounds and metabolic fingerprints in response to herbivory and phytohormones in Plantago lanceolata. New Phytologist 191:1069–1082. Thaler, J. S., P. T. Humphrey, and N. K. Whiteman. 2012. Evolution of jasmonate and salicylate signal crosstalk. Trends in plant science 17:260–70. Therneau, T. M. 2012. Mixed Effects Cox Models [R package coxme version 2.2-5]. Tian, H., S. Zhou, L. Dong, T. P. Van Boeckel, Y. Cui, S. H. Newman, J. Y. Takekawa, D. J. Prosser, X. Xiao, Y. Wu, B. Cazelles, S. Huang, R. Yang, B. T. Grenfell, and B. Xu. 2015. Avian influenza H5N1 viral and bird migration networks in Asia. Proceedings of the National Academy of Sciences 112:172–177. Tilman, D. 2004. Niche tradeoffs, neutrality, and community structure: a stochastic theory of resource competition, invasion, and community assembly. Proceedings of the National Academy of Sciences of the United States of America 101:10854–61. Todesco, M., S. Balasubramanian, T. T. Hu, M. B. Traw, M. Horton, P. Epple, C. Kuhns, S. Sureshkumar, C. Schwartz, C. Lanz, R. A. E. Laitinen, Y. Huang, J. Chory, V. Lipka, J. O. Borevitz, J. L. Dangl, J. Bergelson, M. Nordborg, and D. Weigel. 2010. Natural allelic variation underlying a major fitness trade-off in Arabidopsis thaliana. Nature 465:632–636. Tollenaere, C., H. Susi, and A. L. Laine. 2015. Evolutionary and Epidemiological Implications of Multiple Infection in Plants. Trends in Plant Science 21:80–90. Torchin, M. E., O. Miura, and R. F. Hechinger. 2015. Parasite species richness and intensity of interspecific interactions increase with latitude in two wide-ranging hosts.

146

Ecology:150527153446003. Traw, M. B., and J. Bergelson. 2003. Interactive effects of jasmonic acid, salicylic acid, and gibberellin on induction of trichomes in Arabidopsis. Plant physiology 133:1367–1375. Underwood, N. 1998. The timing of induced resistance and induced susceptibility in the soybean Mexican bean beetle system. Oecologia 114:376–381. Vannette, R. L., and T. Fukami. 2014. Historical contingency in species interactions: Towards niche-based predictions. Ecology Letters 17:115–124. Verboom, G. A., W. D. Stock, and M. D. Cramer. 2017. Specialization to Extremely Low- Nutrient Soils Limits the Nutritional Adaptability of Plant Lineages. The American naturalist 189:684–699. Veresoglou, S. D., E. K. Barto, G. Menexes, and M. C. Rillig. 2013. Fertilization affects severity of disease caused by fungal plant pathogens. Plant Pathology 62:961–969. Viana, D. S., J. Figuerola, K. Schwenk, M. Manca, A. Hobæk, M. Mjelde, C. D. Preston, R. J. Gornall, J. M. Croft, R. A. King, A. J. Green, and L. Santamaría. 2016. Assembly mechanisms determining high species turnover in aquatic communities over regional and continental scales. Ecography 39:281–288. Vlot, A. 2009. Salicylic acid, a multifaceted hormone to combat disease. Annual review of …. Vlot, A. C., D. A. Dempsey, and D. F. Klessig. 2009. Salicylic Acid, a multifaceted hormone to combat disease. Annual review of phytopathology 47:177–206. Vos, I. A., L. Moritz, C. M. J. Pieterse, and S. C. M. Van Wees. 2015. Impact of hormonal crosstalk on plant resistance and fitness under multi-attacker conditions. Frontiers in Plant Science 6:639. Watve, M. G., and R. Sukumar. 1995. Parasite abundance and diversity in mammals - Correlates with host ecology. Proceedings of the National Academy of Sciences of the United States of America 92:8945–8949. Welsh, M. E., J. P. Cronin, and C. E. Mitchell. 2016. The role of habitat filtering in the leaf economics spectrum and plant susceptibility to pathogen infection. Journal of Ecology 104:1768–1777. Werner, G. D. A., and E. T. Kiers. 2015. Order of arrival structures arbuscular mycorrhizal colonization of plants. New Phytologist 205:1515–1524. Wood, C. L., and K. D. Lafferty. 2013. Biodiversity and disease: a synthesis of ecological perspectives on Lyme disease transmission. Trends in Ecology & Evolution 28:239–247. Young, H., R. H. Griffin, C. L. Wood, and C. L. Nunn. 2013. Does habitat disturbance increase

147

infectious disease risk for primates? Ecology Letters 16:656–663. Young, H. S., I. M. Parker, G. S. Gilbert, A. Sofia Guerra, and C. L. Nunn. 2017. Introduced Species, Disease Ecology, and Biodiversity–Disease Relationships. Trends in Ecology & Evolution 32:41–54. Zhan, J., and B. a McDonald. 2013. Experimental measures of pathogen competition and relative fitness. Annual review of phytopathology 51:131–53. Zhao, X., J. G. Lynch Jr., and Q. Chen. 2010. Reconsidering Baron and Kenny: Myths and Truths about Mediation Analysis. Journal of Consumer Research 37:197–206. Zuur, A. F., E. N. Ieno, N. J. Walker, A. A. Saveliev, and G. M. Smith. 2009. Mixed effects models and extensions in ecology with R. Gail M, Krickeberg K, Samet JM, Tsiatis A, Wong W, editors. New York, NY: Spring Science and Business Media.

148