An Abstract of the Dissertation Of

AN ABSTRACT OF THE DISSERTATION OF

Edward Gilman Barge for the degree of Doctor of Philosophy in Botany and Plant Pathology presented on December 13, 2019.

Title: Structure and Function of Foliar Fungal Communities of Populus trichocarpa Across its Native Range, Pacific Northwest, USA.

Abstract approved: ______Posy E. Busby

Foliar fungi – pathogens, endophytes, epiphytes – form taxonomically diverse communities that affect plant health and productivity. The composition of foliar fungal communities is variable at spatial scales both small (e.g., individual plants) and large (e.g., continents). However, few studies have focused on how environmental factors and host plant traits influence the composition and temporal variability of these communities. Moreover, predicting how nonpathogenic members of these communities affect the plant host remains a challenge.

In Chapter two we used ITS metabarcoding to characterize foliar fungal communities of Populus trichocarpa in two consecutive years at the same sites located across its native range in the Pacific Northwest of North America. We used multivariate analyses to test for and differentiate spatial and environmental factors affecting community composition, and tested whether the magnitude of year-to-year variation in community composition varied among environments. We found that climate explained more variation in community composition than geographic

distance, although the majority of variation was shared, and that the year-to-year variability of communities depended on the environmental context, with greater variability in the drier sites located east of the Cascade Range.

In Chapter three we used ITS metabarcoding and multivariate analyses to test whether the influence of intraspecific host genetic variation on the foliar fungal community diminished over the course of one growing season. We utilized 12 P. trichocarpa genotypes that vary in two functional traits: phenology of bud-burst

(early vs. late) and Sphaerulina leaf spot resistance (resistant vs. susceptible). We found that both of these traits drove differences in community composition among trees, but that the strength of the effect diminished through time as the communities converged. Our results suggest that for the leaves of deciduous plants, intraspecific host genetic variability may have its strongest impact on microbial community composition early in assembly.

In Chapter four we tested whether Cladosporium endophyte phylogeny can be used to predict endophyte effects on P. trichocarpa leaf rust disease severity caused by Melampsora × columbiana. We used multilocus sequence typing to infer phylogenetic relationships among 96 Cladosporium endophyte isolates collected from wild P. trichocarpa trees throughout its native range. We then conducted a double- inoculation leaf-disk assay (endophyte inoculated first, then rust pathogen) for a subset of 50 Cladosporium isolates to characterize disease modification for the endophyte isolates; data on endophytes parasitizing rust was collected simultaneously for each isolate. We found that Cladosporium phylogeny was a significant predictor of rust disease severity and was also correlated with rust mycoparasitism,

demonstrating that fungal endophyte phylogenetic relatedness can help predict differences in endophyte function.

Structure and Function of Foliar Fungal Communities of Populus trichocarpa Across its Native Range, Pacific Northwest, USA

by Edward Gilman Barge

A DISSERTATION

submitted to

Oregon State University

in partial fulfillment of the requirements for the degree of

Doctor of Philosophy

Presented December 13, 2019 Commencement June 2020

Doctor of Philosophy dissertation of Edward Gilman Barge presented on December 13, 2019

APPROVED:

Major Professor, representing Botany and Plant Pathology

Head of the Department of Botany and Plant Pathology

Dean of the Graduate School

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

Edward Gilman Barge, Author

ACKNOWLEDGEMENTS

I am indebted to too many people, things, ideas, places to name them all. But, here’s a start. I would like to thank the Botany and Plant Pathology department as a whole for providing a warm, welcoming, and stimulating environment for my dissertation. I would like to thank everyone in both the Busby and Spatafora labs.

Thank you Posy Busby for being an excellent advisor and mentor. You were very gracious and helped me grow in so many ways. Shawn Brown helped kick things off.

Devin Leopold, Kyle Gervers, Gillian Bergmann and Sabrina Heitmann were always there to chat and trouble-shoot and push me to think about things in different ways. I would like to thank the Stuntz Foundation, Oregon Mycological Society, Cascade

Mycological Society, and the Ben Woo Research Grant for funding. I would like to thank Richard Tehan for being a great friend and partner in crime (mushroom hunting). I would like to acknowledge and thank the beaches, forests, mushrooms, and beer of Oregon. Truly a wonderful state. I would like to thank Ekhart Tolle. And last but not least, I would like to thank my family, my mother, father, brother aunt

Patty, and my wife, Charissa Bujak for providing unconditional love and support throughout.

One Love,

Edward G. Barge

CONTRIBUTION OF AUTHORS

Chapter 2: Posy E. Busby and George Newcombe conceived the idea for the study.

Posy E. Busby collected and processed the samples with assistance from Kabir Peay.

Devin R. Leopold assisted with data analysis. Posy E. Busby, George Newcombe,

Kabir Peay, and Devin R. Leopold advised and edited the manuscript.

Chapter 3: Posy E. Busby and Sean P. Brown conceived the idea for the study, assisted with data analysis and edited the manuscript.

Chapter 4: Posy E. Busby and Rytas Vilgalys assisted in the design of the study.

Alejandro Rojas conducted a portion of the multi-locus sequence typing. Devin R.

Leopold assisted with data analysis. Posy E. Busby, Alejandro Rojas, and Devin R.

Leopold edited the manuscript.

TABLE OF CONTENTS

Page

Chapter 1: Introduction...……………………………………………………………...1

References……………………………………………………………………..5

Chapter 2: Differentiating spatial from environmental effects on foliar fungal

communities of Populus trichocarpa...………………………………………….…… 7

Abstract………………………………………………………………………..8

Introduction………………………………………………………………...….9

Materials and Methods…………………………………………………….…12

Results………………………………………………………………………..18

Discussion……………………………………………………………...…….20

References……………………………………………………………………25

Chapter 3: Intraspecific host genetic effects on the foliar fungal microbiome of Populus

trichocarpa diminish through the growing season….....………………...…………...47

Abstract………………………………………………………………………48

Introduction………………………………………….……………………….49

Materials and Methods……………………………………………………….52

Results………………………………………………………………………..57

Discussion……………………………………………………...…………….59

References……………………………………………………………………65

Chapter 4: Phylogenetic relatedness among Cladosporium leaf endophytes

predicts their ability to reduce the severity of a poplar leaf rust disease …………....78

Abstract………………………………………………………………………79

TABLE OF CONTENTS (Continued)

Page

Introduction……..…………………………………………..………………..80

Materials and Methods……………………………………………………….82

Results………………………………………………………………………..90

Discussion……………………………………………………………………92

References……………………………………………………………………99

Chapter 5: Conclusion………………………………………………………………128

LIST OF FIGURES

Figure Page

Figure 2.1 Map of study sites………………………………….…………..…………29

Figure 2.2 Taxonomic distribution of the 200 most abundant foliar fungal OTUs….30

Figure 2.3 NMDS ordination of foliar fungal community composition and scatterplots of site-level Bray-Curtis dissimilarity vs. geographic distance...…..…...31

Figure 2.4 Associations between OTUs and sites over the two years……………….32

Figure 2.5 Box plots of foliar fungal Shannon diversity and Chao1 richness...... …..33

Figure 2.6 Venn diagram of variation in community composition explained by spatial vs. climatic factors………………………………………….……..………….34

Figure 2.7 Bar graph showing the proportion of variation in community composition explained by year at each site…………………………………………..35

Figure 2.S1 Monthly precipitation and mean temperature at sites in 2013 and 2014……………………………………………………………………………...36

Figure 2.S2 Variation in sequencing depth by year and region…………………...…37

Figure 2.S3 Within-site and within-region betadispersion………………..…………38

Figure 2.S4 PCA plot of abiotic environmental variation among sites…………...…39

Figure 3.1 Verification of leaf phenology and disease resistance phenotypes…………………………………………………………………………...69

Figure 3.2 Convergence in community composition (mean pairwise Bray-Curtis dissimilarity) between disease resistant vs. susceptible trees and between early vs. late phenology trees…………………………………………..70

Figure 3.3 Convergence in community composition (PCoA axes) over time……………………………………………………………………..………71

Figure 3.S1 Map of study sites and genotypes and their associated phenology and disease resistance phenotypes…………………………………………………...72

Figure 3.S2 Variation in community composition (PCoA axes 1–4) over time and between early vs. late phenology trees and between disease resistant vs. susceptible trees……………………………………………………...…73

LIST OF FIGURES (Continued)

Figure Page

Figure 3.S3 Variation in richness and Shannon diversity over time and between early vs. late phenology trees and between disease resistant vs. susceptible trees…………………………………………………………………..74

Figure 4.1 Map of study sites………………………….………...………………….103

Figure 4.2 Images of experimental leaf disks displaying healthy Melampsora uredinium, putative mycoparasitism by Cladosporium, and disease antagonism………………………………………………………...…...104

Figure 4.3 Multigene maximum likelihood phylogeny of the 50 Cladosporium isolates used in rust disease modification assays…………………...105

Figure 4.4 Regression of Cladosporium phylogenetic PCoA axes against rust disease severity………………………………………………………………...106

Figure 4.5 Regression of Cladosporium phylogenetic PCoA axis 1 against rust mycoparasitism……………………………………………………………...... 107

Figure 4.6 Effects of autoclaved, filtered, and live spore slurry on rust severity……………………………………………………………….………...108

Figure 4.S1 Multigene maximum likelihood phylogeny of all 96 Cladosporium isolates………………………………………………………………109

Figure 4.S2 Maximum likelihood phylogeny of the tub locus……………………..110

Figure 4.S3 Maximum likelihood phylogeny of the act locus……………………...111

Figure 4.S4 Maximum likelihood phylogeny of the ITS region……………….…...112

Figure 4.S5 Maximum likelihood phylogeny of the tef1 locus…………………….113

Figure 4.S6 Maximum likelihood phylogeny of the rpb2 locus……………………114

Figure 4.S7 PCoA plot of phylogenetic distances among the 50 Cladosporium isolates used in disease modification assays……………………...... 115

LIST OF TABLES

Table Page

Table 2.S1 Region effects on betadispersion…………………………...………….. 40

Table 2.S2 Climate variables used in principal components analysis………….……41

Table 2.S3 Results of distance-based redundancy analysis and forward model selection of spatial and environmental predictors…………………...42

Table 2.S4 Results of PERMANOVA testing effect of temporal and spatial/environmental factors on community composition…………………….…….43

Table 2.S5 Fungal indicator species for areas either east or West or East of the Cascades………………………………………………………………….44

Table 2.S6 Year, region, and sequencing depth effects on foliar fungal alpha diversity……………………………………………………………...…45

Table 2.S7 Results of PERMANOVA testing effect of year on community composition at each site………………………………………………....46

Table 3.1 Results of linear mixed-effects models testing phenology, disease resistance and time effects on PCoA axes 1–4……………………………………...75

Table 3.2 Results of linear mixed-effects models testing phenology, disease resistance and time effects on richness and Shannon diversity……………………...76

Table 3.S1 Results of linear mixed-effects models testing phenology, disease resistance and time effects on disease severity………………………………………77

Table 4.S1 Site metadata……………………………………………………………116

Table 4.S2 Primers used in Cladosporium multilocus sequence typing………………..……………………………………………………….……..117

Table 4.S3 Optimal gene partitioning scheme and best evolutionary model for each gene partition………...…………………………………………….118

Table 4.S4 GenBank accession numbers and top BLAST match to the act locus of each Cladosporium isolate…………………………………………119

Table 4.S5 The number of SNPs, mean length, and SNPs/100 bp for each locus………………………………………………………………….……122

LIST OF TABLES (Continued)

Table Page

Table 4.S6 Final model test for whether Cladosporium phylogeny was correlated with rust disease severity…………………………………….……..123

Table 4.S7 Final model test for whether Cladopsorium phylogeny was correlated with rust mycoparasitism…………………………………………...124

Table 4.S8 Results of tests for difference in rust disease severity among spore slurry treatments……………………………………………………...125

Table 4.S9 Tukey contrasts for difference in rust disease severity among spore slurry treatments……………………………………………………...126

Table 4.S10 Physical evidence of mycoparasitism on rust uredinia in second leaf disk assay…………………………………………………..127

Chapter 1. Introduction

Plant leaves are colonized by a vast diversity of fungi. A small fraction of this diversity is made up of pathogens, which cause plant disease. Plant pathogenic fungi have long been known to impact the structure and composition of plant communities by causing disease (Gilbert, 2002).

However, the majority of fungal diversity on and in leaves is made up of epiphytes and endophytes; these organisms live on (epi-) and in (endo-) plant leaves without causing disease.

Epiphytes and endophytes are taxonomically and functionally diverse and can be mutualists (Clay, 1988; Arnold et al., 2003; Arnold, 2007; Rodriguez et al., 2009; Busby et al.,

2016a), latent saprotrophs (Promputtha et al., 2010; Sun et al., 2010), latent pathogens (Carroll,

1988) or commensal symbionts (May, 2016). However, most epiphyte/ and endophyte/plant relationships are poorly understood. Thus, developing a better understanding of the functional diversity of endophytes and epiphytes and how foliar fungal communities (pathogens, endophytes, epiphytes) vary through space and time may help to explain variation in plant and ecosystem function, and facilitate the development of endophytes and epiphytes for crop improvement and plant conservation.

As in macroecological communities (Legendre et al., 2005), the composition of foliar fungal communities is driven in part by dispersal limitation and abiotic habitat filters. For example, foliar fungal communities vary over broad latitudinal gradients (Arnold & Lutzoni,

2007), and climatic gradients (U’Ren et al., 2012; Zimmerman & Vitousek, 2012; Giauque &

Hawkes, 2016). However, few studies have conducted repeat sampling over time, and it is not known whether the magnitude of year-to-year variation in foliar fungal communities is consistent across contrasting environments.

Another important factor that influences the assembly of foliar fungal communities is the plant host itself. Foliar fungi can vary among host plant species (Wearn et al., 2012; Kembel &

Mueller, 2014; Saptoka et al., 2015), and among genotypes (Bálint et al., 2013; Hunter et al.,

2015) of the same host plant species. A number of recent studies have identified plant traits which may be driving differences in foliar fungal communities among plants. For example, leaf aluminum content (Kembel & Mueller, 2014), and traits related to biochemical defense

(Saunders & Kohn, 2009) have been shown to be important drivers of foliar endophyte community assembly. However, the majority of plants and plant traits have not been investigated. Further, few studies have conducted repeat sampling over time, and it is not known whether the influence of host plant traits on community assembly varies over time. Identifying when plant traits have their strongest impact on foliar fungal community composition is an important first step in elucidating how genotypic variation in host plants impacts the microbiome

(Busby et al., 2017).

In addition to efforts to elucidate the diversity and distribution of foliar fungi, improving our ability to predict endophyte function is of great interest, both for understanding the evolution and ecology of endophytes, but also for deploying endophytes for crop improvement. Many studies have demonstrated that endophytes can modify host plant disease severity (Arnold et al.,

2003; Kiss, 2003; Busby et al., 2016a). However, it is also now clear that endophyte effects on plant disease severity are context-dependent and vary by endophyte, pathogen, host, and the biotic and abiotic environment (Kiss, 2003; Kurose et al., 2012; Busby et al., 2016b). Thus, given the high diversity of fungi, plants, and environments, broad functional predictions are mostly lacking. Phylogenetic relatedness and function have been shown to correlate in a few recent studies of plant associated fungi (Powell et al., 2009; McGuire et al., 2010; Kia et al.,

2017; Hoeksema et al., 2018), suggesting the possibility of using taxonomy to predict function.

However, fungal function is typically not well conserved at high taxonomic levels due to

4 convergent evolution (Zanne et al., 2019). At lower taxonomic levels (e.g., genera), it is less clear if phylogenetic relationships can help predict ecological function. Thus, the use of phylogeny to predict function in more recently diverged lineages containing endophytes could be particularly useful.

This dissertation investigates how spatial, temporal, and host factors impact the assembly of foliar fungal communities. Further, we explore whether foliar endophyte phylogeny can be used as a means to predict endophyte function in planta. We conducted this research in the model tree Populus trichocarpa. Chapter two focuses on teasing apart spatial from environmental drivers of foliar fungal community composition and whether the magnitude of year-to-year variation in communities is consistent in contrasting environments. Chapter three tests whether the influence of intraspecific plant host genetic variation on foliar fungal community composition diminishes over the course of the growing season. Chapter four tests whether phylogenetic relatedness among Cladosporium leaf endophytes can be used to predict their ability to reduce the severity of a poplar leaf rust disease.

REFERENCES:

Arnold, A. E. (2007). Understanding the diversity of foliar endophytic fungi: progress, challenges, and frontiers. Fungal Biology Reviews, 21(2), 51–66. Arnold, A. E., Mejía, L. C., Kyllo, D., Rojas, E. I., Maynard, Z., Robbins, N., & Herre, E. A. (2003). Fungal endophytes limit pathogen damage in a tropical tree. Proceedings of the National Academy of Sciences, 100, 15649–15654. Arnold, A. E., & Lutzoni, F. (2007). Diversity and host range of foliar fungal endophytes: are tropical leaves biodiversity hotspots? Ecology, 88(3), 541–549. Bálint, M., Tiffin, P., Hallström, B., O’Hara, R. B., Olson, M. S., Frankhauser, J. D., Piepenbring, M., & Schmitt, I. (2013). Host genotype shapes the foliar fungal microbiome of balsam poplar (Populus balsamifera). PLoS One, 8(1), e53987. Busby, P. E., Peay, K. G., & Newcombe, G. (2016a). Common foliar fungi of Populus trichocarpa modify Melampsora rust disease severity. New Phytologist, 209(4), 1681– 1692. Busby, P. E., Ridout, M., & Newcombe, G. (2016b). Fungal endophytes: modifiers of plant disease. Plant Molecular Biology, 90, 645–655. Busby, P. E., Soman, C., Wagner, M. R., Friesen, M. L., Kremer, J., Bennett, A., Morsy, M., Eisen, J. A., Leach, J. E., & Dangl, J. L. (2017) Research priorities for harnessing plant microbiomes in sustainable agriculture. PLoS biology, 15(3), e2001793. Carroll, G. C. (1988). Fungal endophytes in stems and leaves: from latent pathogen to mutualistic symbiont. Ecology, 69, 2–9. Clay, K. (1988). Fungal endophytes of grasses: a defensive mutualism between plants and fungi. Ecology, 69, 10–16. Hoeksema, J. D., Bever, J. D., Chakraborty, S., Chaudhary, V. B., Gardes, M., Gehring, C. A., Hart, M. M., Housworth, E. A., Kaonongbua, W, Klironomos, J. N. & Lajeunesse, M. J. (2018). Evolutionary history of plant hosts and fungal symbionts predicts the strength of mycorrhizal mutualism. Communications biology, 1(1), 116. Gilbert, G. S. (2002). Evolutionary ecology of plant diseases in natural ecosystems. Annual Review of Phytopathology, 40(1), 13–43. Giauque, H., & Hawkes, C. V. (2016). Historical and current climate drive spatial and temporal patterns in fungal endophyte diversity. Fungal Ecology, 20, 108–114. Hunter, P. J., Pink, D. A., & Bending, G. D. (2015). Cultivar-level genotype differences influence diversity and composition of lettuce (Lactuca sp.) phyllosphere fungal communities. Fungal Ecology, 17, 183–186. Kembel, S. W., & Mueller, R. C. (2014). Plant traits and taxonomy drive host associations in tropical phyllosphere fungal communities. Botany, 92(4), 303–311. Kia, S. H., Glynou, K., Nau, T., Thines, M., Piepenbring, M., & Maciá-Vicente, J. G. (2017). Influence of phylogenetic conservatism and trait convergence on the interactions between fungal root endophytes and plants. The ISME journal, 11(3), 777. Kiss, L. (2003). A review of fungal antagonists of powdery mildews and their potential as biocontrol agents. Pest Management Science, 59(4), 475–483. Kurose, D., Furuya, N., Tsuchiya, K., Tsushima, S., & Evans, H. C. (2012). Endophytic fungi associated with Fallopia japonica (Polygonaceae) in Japan and their interactions with Puccinia polygoni-amphibii var. tovariae, a candidate for classical biological control. Fungal biology, 116(7), 785–791.

Legendre, P., Borcard, D., & Peres-Neto, P. R. (2005) Analyzing beta diversity: partitioning the spatial variation of community composition data. Ecological monographs, 75(4), 435–50. May, G. (2016). Here come the commensals. American Journal of Botany, 103(10), 1709–1711. McGuire, K. L., Bent, E., Borneman, J., Majumder, A., Allison, S. D., & Treseder, K. K. (2010). Functional diversity in resource use by fungi. Ecology, 91(8), 2324–2332. Powell, J. R., Parrent, J. L., Hart, M. M., Klironomos, J. N., Rillig, M. C., & Maherali, H. (2009). Phylogenetic trait conservatism and the evolution of functional trade-offs in arbuscular mycorrhizal fungi. Proceedings of the Royal Society B: Biological Sciences, 276(1676), 4237–4245. Promputtha, I., Hyde, K. D., McKenzie, E. H., Peberdy, J. F., & Lumyong, S. (2010). Can leaf degrading enzymes provide evidence that endophytic fungi become saprobes? Fungal Diversity, 41(1), 89–99. Rodriguez, R. J., White, Jr. J. F., Arnold, A. E., & Redman, R. S. (2009). Fungal endophytes: diversity and functional roles. New Phytologist, 182(2), 314–330. Saptoka, R., Knorr, K., Jørgensen, L. N., O’Hanlon, K. A., & Nicolaisen, M. (2015). Host genotype is an important determinant of the cereal phyllosphere mycobiome. New Phytologist, 207(4), 1134–1144. Saunders, M., & Kohn, L. M. (2009). Evidence for alteration of fungal endophyte community assembly by host defense compounds. New Phytologist, 182, 229–238. Sun, X., Guo, L. D., & Hyde, K. D. (2010). Community composition of endophytic fungi in Acer truncatum and their role in decomposition. Fungal Diversity, 47, 85–95. U’Ren, J. M., Lutzoni, F., Miadlikowska, J., Laetsch, A. D., & Arnold, A. E. (2012). Host and geographic structure of endophytic and endolichenic fungi at a continental scale. American Journal of Botany, 99(5), 898–914. Wearn, J. A., Sutton, B. C., Morley, N. J., & Gange, A. C. (2012). Species and organ specificity of fungal endophytes in herbaceous grassland plants. Journal of Ecology, 100(5), 1085– 1092. Zanne, A. E., Abarenkov, K., Afkhami, M. E., Aguilar-Trigueros, C. A., Bates, S., Bhatnagar, J. M., Busby, P. E., Christian, N., Cornwell, W., Crowther, T. W., & Moreno, H. F. (2019). Fungal functional ecology: Bringing a trait-based approach to plant-associated fungi. EcoEvoRxiv, doi:10.32942/osf.io/a7f6g. Zimmerman, N. B., & Vitousek, P. M. (2012). Fungal endophyte communities reflect environmental structuring across a Hawaiian landscape. Proceedings of the National Academy of Sciences, 109(32), 13022–13027.

Chapter 2. Differentiating spatial from environmental effects on foliar fungal communities of Populus trichocarpa

Edward G. Barge, Devin R. Leopold, Kabir G. Peay, George Newcombe, Posy E. Busby

Journal of Biogeography Volume 46, Issue 9

ABSTRACT:

Foliar fungi – pathogens, endophytes, epiphytes – form taxonomically diverse communities that affect plant health and productivity. The composition of foliar fungal communities is variable at spatial scales both small (e.g., individual plants) and large (e.g., continents), yet few studies have attempted to tease apart spatial from climatic factors influencing these communities. Moreover, few studies have sampled in more than one year to gauge interannual variation in community structure. In two consecutive years, we used DNA metabarcoding to characterize foliar fungal communities of Populus trichocarpa across its geographic range, which encompasses a sharp climatic transition as it crosses the Cascade

Mountain Range. We used multivariate analyses to: 1) test for and differentiate spatial and environmental factors affecting community composition; and 2) test for temporal variation in community composition across spatial and environmental gradients. In both study years, we found that foliar fungal community composition varied among sites and between regions (east vs. west of the Cascades). We found that climate explained more variation in community composition than geographic distance, although the majority of the variation explained by each was shared. We also found that inter-annual variation in community composition depended on environmental context: communities located in the dry, eastern portion of the tree’s geographic range varied more between study years than those located in the wet, western portion of the tree’s range. Our results suggest that the environment plays a greater role in structuring foliar fungal communities than dispersal limitation.

INTRODUCTION:

Foliar fungi exist in and on living plant leaves. The most well-known members of these communities are pathogenic fungi, which cause plant disease and thereby affect the structure and composition of plant communities (Gilbert, 2002). Less well understood are the non-pathogenic epiphytic and endophytic fungi. These fungi are functionally diverse and can be mutualists

(Clay, 1988; Arnold et al., 2003; Arnold, 2007; Rodriguez et al., 2009; Busby et al., 2016a), latent saprotrophs (Promputtha et al., 2010; Sun et al., 2010), or commensal symbionts (May,

2016). Because of their effects on plants, foliar fungi are consequential for ecosystems. Given the ecological roles played by foliar fungi, understanding how these communities vary through space and time may help to explain variation in plant and ecosystem function.

Many factors can result in spatial patterns in foliar fungal community composition. For example, foliar fungi display biogeographic patterns similar to animals and plants, such as distance decay (Meiser et al., 2014), negative associations between richness and latitude (Arnold

& Lutzoni, 2007; Meiser et al., 2014) and variation over elevation and climatic gradients (Arnold

& Lutzoni, 2007; Zimmerman & Vitousek, 2012; Vacher et al., 2016). Additionally, foliar fungi can vary between host plant species (Wearn et al., 2012), and among individuals (Christian et al.,

2016) or genotypes (Bálint et al., 2013) of the same host plant. At smaller spatial scales, foliar fungi can even vary by the position (Harrison et al., 2016) or age (Arnold & Herre, 2003) of leaves on a plant.

While appreciation for the importance of spatial and environmental factors in structuring foliar fungal communities has deepened with recent studies, particularly those utilizing high- throughput DNA sequencing, few studies have attempted to tease apart spatial distance from abiotic environmental effects. Distinguishing between these two drivers is important for

10 understanding whether communities respond more to dispersal limitation or environmental filters. U'Ren et al. (2012) found that climate was a better predictor than spatial distance for foliar endophyte communities at a continental scale. Further, they found that endophyte isolation frequency increased as a function of both growing season length and annual precipitation (U’Ren et al., 2012), suggesting that climate plays a strong role in structuring these communities.

Focusing on a smaller, landscape level scale (elevation gradient on Mauna Loa, Hawaii) and a single host plant (Metrosideros polymorpha), Zimmerman & Vitousek (2012) also found that among-site variation in foliar endophyte community composition varied with temperature and precipitation, even after controlling for geographic distance. Giauque & Hawkes (2016) also found that endophyte communities associated with one grass species in Texas were controlled primarily by a climate gradient. All three of these studies suggest that climate may play a stronger role in structuring foliar fungal communities than spatial distance. However, U’Ren et al. (2012) and Giauque & Hawkes (2016) employed culture-based methods in their study, which do not capture as much diversity as culture-free methods, and thus may fail to detect certain taxa.

Further, Zimmerman & Vitousek (2012) focused their study over a small area, which may have failed to capture differences in community composition due to spatial distance and dispersal limitation.

Foliar fungal communities have also been shown to vary temporally, although less is known about the drivers of this variation. Seasonal variation in communities has been shown to track fungal life-cycles, physical and chemical changes associated with leaf aging, and seasonal changes in weather conditions (Suto, 1999; Kaneko et al., 2003; Osono, 2008; Jumpponen &

Jones, 2010; Fort et al., 2016). Year-to-year variability in community composition has been found in both tropical evergreen (Higgins et al., 2014; Del Olmo-Ruiz & Arnold, 2014) and

11 temperate deciduous systems (Giauque & Hawkes, 2016), although the drivers are poorly understood. Year-to-year variability of foliar fungal communities, at least in deciduous plants, likely stems from the fact that foliar fungi are primarily horizontally transmitted (Rodriguez, et al., 2009) and communities must reassemble each year on the newly emerging leaves of annual or deciduous plants. However, whether the magnitude of this temporal variation is uniform across spatial and environmental gradients is not known. For example, dry environments may provide fewer opportunities for leaf colonization, which generally requires moisture on the leaf surface, potentially resulting in more stochastic community assembly outcomes and thus higher temporal variability in community composition (Fukami, 2005, 2015). Alternatively, very harsh environments may impose a severe abiotic filter, restricting the set of potential colonists and resulting in more deterministic assembly outcomes.

To address these knowledge gaps in our understanding of spatial and temporal patterns in foliar fungi, we sampled leaves of the model deciduous tree species, Populus trichocarpa, in two consecutive years from 10 sites, spanning the trees natural range. P. trichocarpa occupies a broad geographic area in the Pacific Northwest of North America which is bisected by the

Cascade Mountains (Fig 1). This physical barrier arose in the Pliocene and it marks a sharp transition from a marine, wet and mild climate (USDA plant hardiness zones 8 and 9) to a dry and more continental climate (USDA plant hardiness zones 6 and 7) from west to east as well as acting as a potential barrier to dispersal of some plants and animals (Daubenmire, 1975;

Brunsfeld et al., 2001). The sharp climatic transition created by the Cascades thus sets up a good study system for exploring environmental effects on community structure as it partly decouples space and environment. We used DNA metabarcoding and multivariate analyses to: 1) test for patterns of spatial structure in foliar fungal community composition arising from geographic

12 distance or environmental conditions; and 2) test whether the magnitude of temporal variation in community composition is uniform in contrasting environments.

MATERIALS AND METHODS:

Study location and sample collection:

In October 2013 and 2014, leaves were collected from 10 P. trichocarpa populations, 5 west and 5 east of the Cascade Range in the Pacific Northwest of North America (Fig. 2.1). Each population is associated with an independent river valley; western rivers drain into Puget Sound, eastern rivers drain into the Columbia River. The weather patterns (monthly mean temperature and precipitation) were broadly similar between the two study years, and similar to 30-year averages (Supplementary Fig. 2.S1). However, precipitation was generally lower both east and to a lesser extent west of the Cascades in the winter prior to 2013 sampling (Supplementary Fig.

2.S1). Three representative leaves per tree (i.e., not necessarily asymptomatic) were sampled from five trees per population, targeting lower canopy leaves of a standardized age (n = 2 years x

10 sites x 5 trees x 3 leaves = 300). Because our intention was to explore variability of communities between study years at the level of regions and tree populations, not at the level of individual trees, we did not sample the same trees between study years.

Leaves collected in the field were transported to the laboratory in coolers and processed within 24 h. Leaves were surface sterilized in a laminar flow hood by soaking in a 1% sodium hypochlorite (NaClO) solution for 2 min followed by two rinses (1 min each) in sterile deionized water (Raghavendra & Newcombe, 2013) and then air-dried in a laminar flow hood and lyophilized.

Molecular methods:

We used the same protocols for DNA extraction and library preparation in both study years, with minor exceptions noted. We extracted DNA from approximately 10 mg of lyophilized tissue per leaf using Qiagen DNeasy 96 Plant Kits (Qiagen, Valencia, CA, USA), following the manufacturer’s protocols. For the 2013 samples, we used an ethanol precipitation to remove impurities; DNA was eluted in 200 µl elution buffer. For the 2014 samples, DNA was eluted directly into 100 µl elution buffer with no ethanol precipitation.

We amplified and sequenced the fungal ITS1 region for each leaf sample using a modified version of the primers ITS1F and ITS2 (Smith & Peay, 2014) and the Illumina MiSeq platform (Illumina, San Diego, CA, USA). In 2013, PCR reactions consisted of 6 µl genomic

DNA, 0.5 µl of each 10 µM primer, 5 µl of OneTaq Standard Reaction Buffer (New England

BioLabs, Ipswitch, MA, USA), 0.5 µl of 10 mM dNTPs (New England BioLabs), 0.63 units of

OneTaq Hot Start DNA polymerase (New England Biolabs) and water up to 25 µl. In 2014, PCR reactions consisted of 1-2 µl genomic DNA, 0.5 µl of each 10 µM primer, 5 µl of OneTaq

Standard Reaction Buffer (New England BioLabs, Ipswitch, MA, USA), 0.5 µl of 10 mM dNTPs

(New England BioLabs), 0.63 units of OneTaq Hot Start DNA polymerase (New England

Biolabs) and water up to 25 µl. PCR conditions were the same in both years: initial denaturation at 94°C for 1 min; 35 cycles of 30 s at 94°C, 30 s at 52°C and 30 s at 68°C; followed by a 7 min final extension at 68°C. We visualized all PCR products using gel electrophoresis.

In 2013, PCR reactions were cleaned using the Agencourt Ampure XP kit (Beckman

Coulter, Brea, CA, USA), and DNA yield was quantified for normalization using the Qubit hs-

DS-DNA kit (Invitrogen, Carlsbad, CA, USA) on a Tecan Infinite F200 Pro plate reader (Tecan,

Morrisville, NC, USA). In 2014, PCR products were cleaned and normalized in one step using

Just-a-Plate™ 96 PCR Purification and Normalization Kits (Charm Biotech, San Diego, CA,

USA). Both libraries were sequenced at the Stanford Functional Genomics Facility on Illumina

MiSeq (250-bp paired-end in 2013, 300-bp paired-end in 2014). In 2013, three samples yielded no sequence data; raw reads for 147 samples are deposited in NCBI’s Short Read Archive

(accession no. SRP064132; Busby et al., 2016b). In 2014 all 150 samples yielded sequence data; raw reads for 150 samples are deposited in NCBI’s Short Read Archive (accession no.

SRP137181).

Bioinformatics:

Raw reads were processed first with cutadapt v1.4 (Martin, 2011) to trim low-quality 5’ tails and 10 bp from the 3’ end. Operational taxonomic units (OTUs) were identified using usearch v10.0.240 (Edgar, 2010) by merging paired-end reads, quality filtering (max expected error < 1) and denoising with unoise3 (Edgar, 2016a). Denoised reads were then abundance- sorted and clustered at 97% similarity. Spurious OTUs were removed by excluding those with

<65% match to Kingdom Fungi in the UNITE database (Kõljalg et al., 2005). Predicted taxonomy was assigned first with SINTAX (Edgar, 2016b), querying the Warcup v2 fungal ITS region database (Deshpande, et al., 2016), and then curated using manual blastn queries of

GenBank (Clark et al., 2016). We used the R package metacoder (Foster et al., 2017) to visualize taxonomic composition in both years for the 200 most abundant OTUs. To filter the data prior to conducting analyses we removed OTUs making up less than 0.1% of total reads in each sample, and removed samples with fewer than 1000 reads (n = 29). We normalized variable sequencing depth by calculating the proportional abundance of each OTU in each sample (McMurdie & Holmes, 2014). To further ensure that variable sequencing depth among

15 sites (Supplementary Fig. 2.S2) was not driving spatial patterns of variation in community composition, we also ran analyses on samples rarefied to 1000 reads and found that patterns were the same as for non-rarefied data. Thus we chose to present only the results of analyses focusing on non-rarefied data to avoid the loss of information that comes with rarefying (McMurdie &

Holmes, 2014).

Community composition:

Community analyses were carried out using the R package vegan (Oksanen et al., 2007) unless noted otherwise. To visualize fungal community composition among sites we used non- metric multidimensional scaling (NMDS) of Bray-Curtis dissimilarities among trees, and displayed site-level means. To test whether community composition varied with sampling year, region (east or west of the Cascade Range) and among sites within regions, we used permutational analysis of variance (PERMANOVA) of Bray-Curtis dissimilarity among trees

(leaf samples pooled to avoid pseudoreplication). Year was specified first in the model to account for year effects prior to testing for region and site effects. To visualize the relative effects of climate and spatial distance on community composition we constructed a scatterplot of pairwise geographic distance between sites vs. Bray-Curtis dissimilarity, and color-coded each point as a west-west, east-east or east-west comparison. Because the heterogeneity of group variances can influence PERMANOVA results, we calculated betadispersion both within sites and within each region using the function ‘betadisper’. To test for a difference in within-site and within-region betadispersion between regions, we fit a linear mixed-model for each response

(within-site betadispersion, within-region betadispersion) using the R package lme4, function

‘lmer’ (Bates et al., 2015). Predictor variables in each model included the fixed effect region and

16 the random effects year + site. Statistical significance of fixed predictors was assessed using

Type III ANOVA in the R package lmerTest (Kuznetsova, 2016). Betadispersion did not significantly differ between regions (within-site P = 0.074; within-region P = 0.726)

(Supplementary Table 2.S1, Supplementary Fig. 2.S3). However, the betadispersion results suggest that the PERMANOVA may have been influenced by differences in within-site betadisperion between regions.

To explore whether particular OTUs were more common or abundant east or west of the

Cascades in each year, we conducted indicator species analysis of the 25 most abundant OTUs using PC-ORD (McCune & Mefford, 2015). P-values were corrected for multiple hypothesis testing using the method of Holm (1979) at alpha = 0.05. The relative abundances of the 10 most abundant OTUs overall were visualized for each study site using a bipartite network, linking sites and OTUs with the R package bipartite, function ‘plotweb’ (Dormann et al., 2009).

In order to determine whether alpha-diversity varied by year and region (east vs. west of the Cascades) we first estimated Shannon Diversity and species richness (Chao1) for each sample using the R package phyloseq, function ‘estimate_richness’ (McMurdie & Holmes,

2013). Diversity metrics were calculated using non-normalized sequence counts for all OTUs, including low abundance OTUs filtered for other analyses (Chao, 1984). To test whether alpha- diversity varied by year, region, and year × region, we fit a linear mixed-effects model using the

R package lme4, function ‘lmer’ (Bates, et al., 2015). Log-transformed sequencing depth was included as a fixed-effect to account for unequal sampling effort and both tree and site were included as random effects to account for the non-independence of samples. Statistical significance of fixed effects was assessed using Type III ANOVA in the R package lmerTest

(Kuznetsova, 2016).

Partitioning spatial and environmental effects:

We used variation partitioning to explore the relative effects of spatial versus climatic effects on community composition. To control for the differences between years, we first performed a distance-based redundancy analysis (function ‘dbrda’) using Bray-Curtis dissimilarity among sites as the response variable and year as the predictor, and extracted the residuals for further analysis. We interpolated 30-year average climate variables for each site using ClimateNA (Wang et al., 2016) (Supplementary Table 2.S2). The climate variables selected were mean annual precipitation (MAP), mean annual temperature (MAT), continentality

(i.e., difference in mean temperatures of the warmest and coldest months; TD), and mean number of frost free days (NFFD). Because climate variables were highly correlated we performed principal components analysis (PCA), retaining the first two PC axes, representing 96% of the variation among sites and clearly separating the sites by region (east vs. west of Cascades)

(Supplementary Fig. 2.S4). To assess spatial patterns of fungal community composition we used trend-surface analysis (Gittins, 1968), with second-degree orthogonal polynomials generated from site coordinates using the ‘poly’ function in R. We then determined which spatial and climate axes most significantly explained variation in Bray-Curtis community dissimilarity among sites using distance-based redundancy analysis and forward model selection with functions ‘dbrda’ and ‘ordiR2step’. Forward model selection retained one spatial predictor (the linear east-west-axis), and one principle component of climate data (PC1) (Supplementary Table

2.S3). Finally, we partitioned the variation in community composition among sites by climate versus spatial factors using function ‘varpart’. We then tested the significance of the unique fractions of variation in community composition explained by climate versus spatial factors using partial distance-based redundancy analysis and permutation tests.

Between year variation:

To investigate variation in community composition between study years, we pooled samples at the tree level and conducted PERMANOVA analyses separately for each site, using

Bray-Curtis dissimilarity among trees as the response variable and year as the predictor variable in each analysis. We used R2 (proportion of variation explained by year) to compare the relative strength of the year effect among sites/regions.

RESULTS:

Community composition:

P. trichocarpa hosted a taxonomically diverse community of foliar fungi representing 50 identifiable orders and 1,216 OTUs (786 OTUs in 2013 and 849 OTUs in 2014). Communities were dominated in both years by Ascomycota (2013 = 92%, 2014 = 95%), followed by

Basidiomycota (2013 = 3%, 2014 = 4%), and Mucoromycota (2013 = 5%, 2014 = 1%) (Fig. 2.2).

Dominant classes in both years included Dothideomycetes, Leotiomycetes, and Sordariomycetes in the Ascomycota (Fig. 2.2). The most abundant genus sampled in both years was Sphaerulina

(Ascomycota, Capnodiales), which contains the known P. trichocarpa pathogen S. populicola

(Fig. 2.2). Other genera containing Populus pathogens that were dominant in both years included

Venturia (Pleosporales) and Marssonina (Helotiales). Ascomycetes that are non-pathogenic on

P. trichocarpa, and which were abundant in both years included species of Cladosporium and

Ramularia (Capnodiales), Aureobasidium (Dothideales), Alternaria, Articulospora, and

Epicoccum (Pleosporales). Other non-pathogens, such as Trichoderma (Hypocreales) and

Phialocephala (Helotiales), both in the Ascomycota, and Mortierella (Mortierellales) in the

Mucoromycota were abundant in 2013 but nearly non-existent in 2014. Basidiomycota was represented primarily by the rust pathogen Melampsora (Pucciniales) in both years (Fig. 2.2).

The community composition of fungi associated with P. trichocarpa leaves varied among sites and between regions (east vs. west of Cascades) in both years (Supplementary Table 2.S4,

Fig. 2.3). Differences in community composition between east-west site pairs were large even at short spatial distances, suggesting a sharp turnover in community composition on opposite sides of the Cascades (Fig 2.3). In both years, 13 of the 25 most abundant OTUs were associated with a geographic region; 10 west and 3 east (Supplementary Table 2.S5, Fig. 2.4). Shannon diversity and estimated richness (Chao1) were higher in the region west of the Cascades in both years

(Supplementary Table 2.S6, Fig. 2.5).

Partitioning spatial and environmental effects:

We found that spatial distance (the linear east-west-axis) and climate (PC1) explained

1.4% and 8.3% of unique variation in P. trichocarpa foliar fungal community composition respectively, although the majority of overall variation explained by each was shared between the two sets of predictors (10.1%) (Fig. 2.6). The unique variation explained by climate (PC1) was significant (F=2.435, P=0.033) but the fraction of unique variation explained by spatial distance was not (F=1.099, P=0.307).

Between year variation:

We found that fungal community composition significantly varied between study years

(Fig. 2.3); however, the magnitude of this effect varied among study sites. In particular, when community composition was analyzed at the site-level, it varied with year at all 5 eastern sites,

20 but only 1 of the 5 western sites (Supplementary Table 2.S7, Fig. 2.7). Shannon diversity and estimated richness (Chao1) also varied from between study years (both were lower in 2014), and this variation was not region-dependent (i.e. year × region interaction was not significant)

(Supplementary Table 2.S6, Fig. 2.5).

DISCUSSION:

Spatial/environmental structure:

P. trichocarpa occupies a broad geographic area in Western North America, encompassing a wide range of environmental conditions. Within the distribution of P. trichocarpa, the Cascade Range has, since the Pliocene, marked a particularly sharp transition in the abiotic and biotic environment, moving from a wet and mild climate to the west into a dry and more continental climate to the east, as well as acting as a potential physical barrier to dispersal of plants and animals (Daubenmire, 1975; Brunsfeld et al., 2001). For example, the coastal and inland varieties of Douglas-fir are consistent with divergence since the Pliocene

(Gugger et al., 2010). We found that foliar fungal community composition was highly variable across the range of P. trichocarpa. Specifically, communities varied among sites and between regions (east vs. west of the Cascades), and fungal alpha-diversity was higher in the wetter, milder sites west of the Cascades.

While the observational nature of this study precluded us from explicitly separating the effects of space and climate on foliar fungi, variation partitioning allowed us to address their relative effects. We found that fungal community composition was primarily explained by the correlated components of space and climate. However, we also found that additional variation in fungal community composition could be explained by variation in climate that was independent

21 of spatial distance among sites. This finding makes sense given the study system, in which the

Cascade Range marks a sharp transition in climate over a short geographic distance. This finding is also in agreement with previous studies of foliar endophyte communities that have examined turnover in foliar endophyte community composition both at a continental scale (U’Ren et al.,

2012) and at a smaller landscape scale over an elevation gradient (Zimmerman & Vitousek,

2012). Thus it appears that spatial patterns of foliar fungal communities may be different than soil fungal communities, which display greater dispersal limitation (Talbot et al., 2014), and environmental filtering by the soil matrix (David et al., 2015). The Cascade Range may also act as a physical barrier to spore dispersal, which may have led to some of the differences in community composition we observed in this study. However, because there is evidence of pollen

(not seed) exchange across the Cascades (Gugger et al., 2010), and pollen is wind-born like spores of fungi, we suspect that the Cascades do not entirely prohibit fungal dispersal.

The biotic environment, or host plant population, may have also influenced spatial variation in foliar fungal communities (Bálint et al., 2013). Previous studies have found evidence for P. trichocarpa population genetic structure at the scale of our study, but along a north-south gradient (Evans et al., 2014), as well as differences in P. trichocarpa leaf traits (e.g., defense, epidermal structure; Dunlap & Stettler, 1996; 2001) between eastern and western tree populations. Because leaf traits can influence colonization by foliar fungi (Kembel & Mueller,

2014), phenotypic variation among tree populations could have contributed to observed variation in foliar fungi in our study. Future, manipulative studies (e.g., reciprocal common garden design) are needed to disentangle the effects of host genetic variation, geography, and climatic gradients on foliar fungi.

Between year variation:

Sampling the same P. trichocarpa populations in two consecutive years allowed us to investigate variation in fungal community composition between study years. Our observational study did not directly address the specific factor(s) driving site-level community shifts between years. However, we found greater variation in foliar fungal community composition between study years in dry, continental (eastern) vs. wet, maritime (western) sites, suggesting that environmental context may influence temporal variability. Major differences among eastern and western sites that could have led to the pattern we observed include moisture availability, landscape connectivity, and foliar fungal diversity. Overall, a more stochastic, and therefore less deterministic community assembly process is predicted for each of these characteristics: moisture limitation and habitat isolation (Fukami, 2005, 2015), and limited diversity (Levine &

D’antonio, 1999), all of which characterize eastern sites. However, given that sampling was conducted in only two years, we cannot draw general conclusions regarding the drivers of year- to-year variation. Additional sampling (i.e., multi-year) would be necessary to reveal whether differences in the magnitude of variation in community composition that we observed between these two years are representative, and whether the variation has a deterministic cause (e.g., climate). Additionally, manipulative experiments would be necessary to identify and tease apart the factors playing a prominent role in year-to-year variation of the foliar fungal community.

Differences in fungal community composition, the number of OTUs, sequencing depth and Shannon and Chao1 between years may also have been influenced by differences in sampling and sequencing methodology. For example, different trees in the study populations were sampled in the different years, and we made minor modifications to DNA extraction protocols, PCR library preparation, and sequencing in the second study year. In particular, we

23 noticed a difference in the number of observed OTUs (786 OTUs in 2013 and 849 OTUs in

2014) and alpha-diversity between years, which could be attributed to differences in molecular methods between years. However, these methodological differences would have affected all samples within each year, limiting potential effects on variation in the magnitude of year-to-year variation among sites. Additionally, despite methodological changes, we identified many of the same dominant taxa in each year, and found consistent patterns of variation in community composition and alpha-diversity.

Conclusions:

While substantial progress has been made in understanding how foliar fungi are distributed across space (Arnold & Lutzoni, 2007; U’Ren et al., 2012, Zimmerman & Vitousek,

2012; Bálint et al., 2013; Christian et al., 2016), few studies have attempted to tease apart spatial from environmental factors influencing these communities, or tested the degree to which communities vary from year-to-year in the leaves of deciduous plants. Answers to these question are important given the critical functional roles played by foliar fungi in both wild and crop plant systems. For example, common foliar endophytes of P. trichocarpa are capable of modifying the expression of foliar disease (Busby et al., 2016b). Our results reveal that the community composition of P. trichocarpa foliar fungi is spatially variable throughout the plant’s native range. In particular, we found distinct communities of foliar fungi when comparing sites on opposite sides of the Cascade Range, which marks a sharp climatic transition. This division is consistent with our finding that spatial patterns of fungal community composition among sites were also largely correlated with variation in climate. However, climate also explained additional variation in community composition that was not associated with spatial distance. Moreover, the

24 degree of site-level, variation in community composition between study years depended on environmental context. Sites in the drier, more continental region were associated with higher variation in fungal composition between study years in comparison to sites in the wetter, more maritime region.

REFERENCES:

Arnold, A. E. (2007). Understanding the diversity of foliar endophytic fungi: progress, challenges, and frontiers. Fungal Biology Reviews 21(2), 51–66. Arnold, A. E., & Herre, E. A. (2003). Canopy cover and leaf age affect colonization by tropical fungal endophytes: ecological pattern and process in Theobroma cacao (Malvaceae). Mycologia 95(3), 388–398. Arnold, A. E., Mejía, L. C., Kyllo, D., Rojas, E. I., Maynard, Z., Robbins, N., & Herre, E. A. (2003). Fungal endophytes limit pathogen damage in a tropical tree. Proceedings of the National Academy of Sciences 100, 15649–15654. Arnold, A. E., & Lutzoni, F. (2007). Diversity and host range of foliar fungal endophytes: are tropical leaves biodiversity hotspots? Ecology 88(3), 541–549. Bálint, M., Tiffin, P., Hallström, B., O’Hara, R. B., Olson, M. S., Frankhauser, J. D., … Schmitt, I. (2013). Host genotype shapes the foliar fungal microbiome of balsam poplar (Populus balsamifera). PLoS One 8(1), e53987. Bates, D., Maechler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects models using lme4. Journal of Statistical Software 67, 1-48. Brunsfeld, S. J., Sullivan, J., Soltis, D. E., & Soltis, P. S. (2001). A comparative phylogeography of northwestern North America: a synthesis. In J Silvertown & J Antonovics (Eds.), Integrating Ecology and Evolution in a Spatial Context (pp. 319–340). Oxford, UK: Blackwell Science. Busby, P. E., Ridout, M., & Newcombe, G. (2016a). Fungal endophytes: modifiers of plant disease. Plant Molecular Biology 90, 645–655. Busby, P. E., Peay, K. G., & Newcombe, G. (2016b). Common foliar fungi of Populus trichocarpa modify Melampsora rust disease severity. New Phytologist 209(4), 1681– 1692. Chao, A. (1984). Non-parametric estimation of the number of classes in a population. Scandinavian Journal of Statistics 11, 265–270. Christian, N., Sullivan, C., Visser, N. D., & Clay, K. (2016). Plant host and geographic location drive endophyte community composition in the face of perturbation. Microbial Ecology 72(3), 621–632. Clark, K., Karsch-Mizrachi, I., Lipman, D. J., Ostell, J., & Sayers, E. W. (2016). GenBank. Nucleic Acids Research 44, D67–D72. Clay, K. (1988). Fungal endophytes of grasses: a defensive mutualism between plants and fungi. Ecology 69, 10–16. David, A.S., Seabloom, E.W., & May, G. (2015). Plant host species and geographic distance affect the structure of aboveground fungal symbiont communities, and environmental filtering affects belowground communities in a coastal dune ecosystem. Microbial Ecology 71, 912–926. Daubenmire, R. (1975). Floristic plant geography of eastern Washington and northern Idaho. Journal of Biogeography 2, 1–18. Del Olmo-Ruiz, M. & Arnold, A. E. (2014). Interannual variation and host affiliations of endophytic fungi associated with ferns at La Selva, Costa Rica. Mycologia 106, 8–21. Deshpande, V., Wang, Q., Greenfield, P., Charleston, M., Porras-Alfaro, A., Kuske, C. R., … Tran-Dinh, N. (2016). Fungal identification using a Bayseian classifier and the Warcup training set of internal transcribed spacer sequences. Mycologia 108(1), 1–5.

Dormann, C. F., Fründ, J., Blüthgen, N., & Gruber, B. (2009). Indices, graphs and null models: Analyzing bipartite ecological networks. Open Ecology Journal 2, 7–24. Dunlap, J. M., Stettler, R. F. (1996). Genetic variation and productivity of Populus trichocarpa and its hybrids. IX. Phenology and Melampsora rust incidence of native black cottonwood clones from four river valleys in Washington. Forest Ecology and Management 87(1), 233–256. Dunlap, J. M. & Stettler, R. F. (2001). Variation in leaf epidermal and stomatal traits of Populus trichocarpa from two transects across the Washington Cascades. Canadian Journal of Botany 79(5), 528–536. Edgar, R. C. (2010). Search and clustering orders of magnitude faster than BLAST. Bioinformatics 26(19), 2460–2461. Edgar, R. C. (2016a). UNOISE2: improved error-correction for Illumina 16S and ITS amplicon sequencing. bioRxiv 082157. Edgar, R. C. (2016b). SINTAX: a simple non-Bayesian taxonomy classifier for 16S and ITS sequences. bioRxiv 074161. Evans, L. M., Slavov, G. T., Rodgers-Melnick, E., Martin, J., Ranjan, P., Muchero, W., … Tuskan, G. A. (2014). Population genomics of Populus trichocarpa identifies signatures of selection and adaptive trait associations. Nature Genetics 46(10), 1089–1096. Fort, T., Robin, C., Capdevielle, X., Delière, L., & Vacher, C. (2016). Foliar fungal communities strongly differ between habitat patches in a landscape mosaic. PeerJ 4, e2656. Foster, Z. L., Sharpton, T. J., & Grünwald, N. J. (2017). Metacoder: An R package for visualization and manipulation of community taxonomic diversity data. PLoS Computational Biology 13(2), e1005404. Fukami, T. (2005). Integrating internal and external dispersal in metacommunity assembly: preliminary theoretical analyses. Ecological Research 20(6), 623–631. Fukami, T. (2015). Historical contingency in community assembly: integrating niches, species pools, and priority effects. Annual Review of Ecology, Evolution, and Systematics 46(1), 1–23. Gilbert, G. S. (2002). Evolutionary ecology of plant diseases in natural ecosystems. Annual Review of Phytopathology 40(1), 13–43. Gittins, R. (1968). Trend-surface analysis of ecological data. Journal of Ecology 56(3), 845–869. Gugger, P. F., Sugita, S., & Cavender-Bares, J. E. (2010). Phylogeography of Douglas-fir based on mitochondrial and chloroplast DNA sequences: testing hypotheses from the fossil record. Molecular Ecology 19(9), 1877–1897. Harrison, J. G., Forister, M. L., Parchman, T. L., & Koch, G. W. (2016). Vertical stratification of the foliar fungal community in the world’s tallest trees. American Journal of Botany 103(12), 2087–2095. Higgins, K. L., Arnold, A. E., Coley, P. D., & Kursar, T. A. (2014). Communities of fungal endophytes in tropical forest grasses: highly diverse host- and habitat generalists characterized by strong spatial structure. Fungal Ecology 8, 1–11. Holm, S. (1979). A simple sequentially rejective multiple test procedure. Scandinavian Journal of Statistics 6, 65–70. Jumpponen, A. & Jones, K. L. (2010). Seasonally dynamic fungal communities in the Quercus macrocarpa phyllosphere differ between urban and nonurban environments. New Phytologist 186(2), 496–513.

Kaneko, R., Kakishima, M., & Tokumasu, S. (2003). The seasonal occurrence of endophytic fungus, Mycosphaerella buna, in Japanese beech, Fagus crenata. Mycoscience 44, 277– 281. Kembel, S. W. & Mueller, R. C. (2014). Plant traits and taxonomy drive host associations in tropical phyllosphere fungal communities. Botany 92(4), 303–311. Kõljalg, U., Larsson, K. H., Abarenkov, K., Nilsson, R. H., Alexander, I. J., Eberhardt, U., … Pennanen, T. (2005). UNITE: a database providing web-based methods for the molecular identification of ectomycorrhizal fungi. New Phytologist 166(3), 1063–1068. Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2016). lmerTest: Tests in Linear Mixed Effects Models. R package version 2.0-30. Levine, J. M. & D’antonio, C. M. (1999). Elton revisited: a review of evidence linking diversity and invasibility. Oikos 87, 15–26. Martin, M. (2011). Cutadapt removes adapter sequences from high-thoughput sequencing reads. EMBnet.journal 17(1), 10. May, G. (2016). Here come the commensals. American Journal of Botany 103(10), 1709–1711. McCann, K. S. (2000). The diversity-stability debate. Nature 405(6783), 228–233. McCune, B. & Mefford, M. J. (2015). PC-ORD. Multivariate Analysis of Ecological Data. Version 7.287 beta. Gleneden Beach, OR, USA: MjM Software. McMurdie, P. J. & Holmes, S. (2013). Phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data. PLoS One 8(4), e61217. McMurdie, P. J. & Holmes, S. (2014). Waste not, want not: why rarefying microbiome data is inadmissible. PloS computational biology 10(4), e1003531. Meiser, A., Bálint, M., & Schmitt, I. (2014). Meta-analysis of deep-sequenced fungal communities indicates limited taxon sharing between studies and the presence of biogeographic patterns. New Phytologist 201(2), 623–635. Oksanen, J., Kindt, R., Legendre, P., O’Hara, B., Stevens, M. H. H., Oksanen, M. J., & Suggests, M. A. (2007). The vegan package. Community Ecology Package 10, 631–637. Osono, T. (2008). Endophytic and epiphytic phyllosphere fungi of Camelia japonica: seasonal and leaf age-dependent variations. Mycologia 100, 387–391. Peay. K. G., Kennedy, P. G., & Talbot, J. M. (2016). Dimensions of biodiversity in the Earth mycobiome. Nature Reviews Microbiology 14(7), 434–447. Promputtha, I., Hyde, K. D., McKenzie, E. H., Peberdy, J. F., Lumyong, S. (2010). Can leaf degrading enzymes provide evidence that endophytic fungi become saprobes? Fungal Diversity 41(1), 89–99. Raghavendra, A. K. & Newcombe, G. (2013). The contribution of foliar endophytes to quantitative resistance to Melampsora rust. New Phytologist 197(3), 909–918. Rodriguez, R. J., White Jr., J. F., Arnold, A. E., & Redman, A. R. (2009). Fungal endophytes: diversity and functional roles. New Phytologist 182(2), 314–330. Smith, D. P. & Peay, K. B. (2014). Sequence depth, not PCR replication, improves ecological inference from next generation DNA sequencing. PLoS One 9(2), e90234. Sun, X., Guo, L. D., & Hyde, K. D. (2010). Community composition of endophytic fungi in Acer truncatum and their role in decomposition. Fungal Diversity 47, 85–95. Suto, Y. (1999). Mycosphaerella chaenomelis sp. nov.: the teleomorph of Cercosporella sp., the causal fungus of frosty mildew in Chaenomeles sinensis, and its role as the primary infection source. Mycoscience 40, 509–516.

Talbot, J. M., Bruns, T. D., Taylor, J. W., Smith, D. P., Branco, S., Glassman, S. I., … Peay, K. G. (2014). Endemism and functional convergence across the North American soil mycobiome. Proceedings of the National Academy of Sciences 11(17), 6341–6346. U’Ren, J. M., Lutzoni, F., Miadlikowska, J., Laetsch, A. D., Arnold, A. E. (2012). Host and geographic structure of endophytic and endolichenic fungi at a continental scale. American Journal of Botany 99(5), 898–914. Vacher, C., Cordier, T., & Vallance, J. (2016). Phyllosphere fungal communities differentiate more thoroughly than bacterial communities along an elevation gradient. Microbial Ecology 72(1), 1–3. Wang, T., Hamann, A., Spittlehouse, D., & Carroll, C. (2016). Locally downscaled and spatially customizable climate data for historical and future periods for North America. PLoS One 11(6), e0156720. Wearn, J. A., Sutton, B. C., Morley, N. J., & Gange, A. C. (2012). Species and organ specificity of fungal endophytes in herbaceous grassland plants. Journal of Ecology 100(5), 1085– 1092 Zimmerman, N. B. & Vitousek, P. M. (2012). Fungal endophyte communities reflect environmental structuring across a Hawaiian landscape. Proceedings of the National Academy of Sciences 109(32), 13022–13027.

Figure 2.1: Map of study sites located across the core of the native range of Populus trichocarpa in the Pacific Northwest of North America. The study area spans a rainfall gradient from west (wet) to east (dry) of the Cascade Range. Sites west of the Cascades are colored green and sites east of the Cascades are colored brown.

Figure 2.2: Taxonomic distribution of the 200 most abundant foliar fungal OTUs (displayed to genus) associated with Populus trichocarpa throughout its native range in the Pacific Northwest in 2013 and 2014. Node width indicates the proportional abundance of sequence reads assigned to a given taxonomic classification. Labels display the taxonomy of the 25 most abundant genera overall, and label size corresponds to relative abundance.

Figure 2.3: (a) Non-metric multidimensional scaling of foliar fungi associated with Populus trichocarpa, showing site-level means (+/- standard error). Communities sampled west of the Cascades are colored green and communities sampled east of the Cascades are colored brown. Communities sampled in 2013 are represented as circles and communities sampled in 2014 are represented as triangles. Text adjacent to each point indicates the site name. Results of PERMANOVA testing temporal (year) and spatial/environmental (site, region) effects on community composition are also displayed. *significant at alpha = 0.05. (b) Scatterplot of Bray- Curtis dissimilarity (y axis) vs. geographic distance (x-axis) for all pairwise comparisons among plots, color-coded by whether comparisons are west-west (green), east-east (brown) or east-west (yellow). East-west sites separated by small geographic distances are just as dissimilar as east- west sites separated by large geographic distances, suggesting that distance alone is not driving differences in community composition between eastern and western sites.

Figure 2.4: Associations between the 10 most abundant Populus trichocarpa foliar fungal OTUs (% of total OTU abundance) and sites over two years. Line width indicates the relative abundance of a fungal OTU (central nodes) in amplicon sequencing data for each site (outer nodes). The color of site nodes indicates whether they are located east (brown) or west (green) of the Cascades. Colored OTU nodes indicate taxa that were significant indicators of the study regions, east (brown) or west (green) of Cascades. Art=Articulospora, Mel=Melampsora, Clad=Cladosporium, Sph=Sphaerulina, Mars=Marssonina, Phi=Phialocephala, Epi=Epicoccum, Ram=Ramularia, Ven=Venturia.

Figure 2.5: Box plots of Shannon diversity (a) and Chao1 richness (b) of foliar fungi associated with Populus trichocarpa in 2013 and 2014. Points represent richness and diversity estimates of individual samples. Box color indicates whether samples came from east (brown) or west (green) of the Cascades. *significant difference in the response at alpha = 0.05.

Figure 2.6: Results of analyses partitioning the variation in Populus trichocarpa foliar fungal community composition explained by spatial vs. climatic factors. *unique fraction significant at alpha = 0.05.

Figure 2.7: Bar graph showing the proportion of variation in Populus trichocarpa foliar fungal community composition (PERMANOVA R2) explained by year at each study site. Bar color indicates whether study site is east (brown) or west (green) of the Cascades. *significant at alpha = 0.05.

Supplementary Figure 2.S1: Monthly precipitation at sites west (a) and east (b) of the Cascades and mean monthly temperature at at sites west (c) and east (d) of the Cascades during winter, spring, summer and fall of the year leading up to October sampling in 2013 (year 1) and 2014 (year 2), and 30-year averages. Climate data was interpolated using the PRISM climate group web interface (www.prism.oregonstate.edu/).

Supplementary Figure 2.S2: Variation in sequencing depth by year and region (east vs. west of Cascades) after trimming low abundance OTUs (making up less than 0.1% of reads in each sample) and samples with less than 1000 reads. Based on Welch’s ANOVA, sequencing depth significantly varied by site (F9,97=4.28, P=0.0001), but not by region (F1,199=2.13, P=0.146) or year (F1,250=1.57, P=0.212).

Supplementary Figure 2.S3: Populus trichocarpa foliar fungal community within-site (a) and within-region (b) betadispersion for each region (East versus West of the Cascades). Points represent per-tree distance to median.

Supplementary Figure 2.S4: Principal components analysis (PCA) of abiotic environmental variation among sites showing the first two PC axes which explained 96% of the variation (PC1 = 75%, PC2=21%). MAP=mean annual precipitation, MAT=mean annual temperature, TD=continentality, NFFD=mean number of frost free days.

Supplementary Table 2.S1: Region (east vs. west of Cascades) effects on betadispersion of foliar fungal communities associated with Populus trichocarpa. Year-to-year and site-to-site variation was accounted for by specifying year and site as random effects.

Within-site betadispersion Within-region betadispersion Region F1,17=3.63 F1,17=0.13 P=0.074 P=0.726

Supplementary Table 2.S2: Climate variables selected for principal components analysis: mean annual precipitation (MAP), mean annual temperature (MAT), continentality (TD), and mean number of frost free days (NFFD). Continentality is defined as the difference in the mean temperatures of the warmest and coldest months. Climate variables were predicted for the 10 sampling locations using ClimateNA (Wang et al. 2016).

Site MAP (mm) MAT (°C) TD NFFD DO 1128 10.8 14.4 304 CAR 1094 10.4 14.1 299 NM 978 10.1 14.8 283 SNO 1620 9.8 14.7 286 SK 1717 9.6 15.5 279 TIE 586 6.2 19.3 182 YK 267 8.8 23 213 KR 473 8.1 24.6 208 CW 400 11.4 23.2 264 LS 523 8.3 24.1 202

Supplementary Table 2.S3: Results of distance-based redundancy analysis and forward model selection of spatial (UTM) and environmental (PC1) predictors of variation in Populus trichocarpa foliar fungal community composition. NS = non-significant in forward model selection. *significant at alpha = 0.05.

Adjusted R2 DF AIC F P PC1 0.113 1 33.89 3.43 0.002* PC2 0.007 NS NS NS NS x 0.072 1 34.79 2.47 0.002* y 0.013 NS NS NS NS x2 0.011 NS NS NS NS y2 0.013 NS NS NS NS xy -0.006 NS NS NS NS

Supplementary Table 2.S4: Results of PERMANOVA testing spatial/environmental (site, region) factors influencing the composition of foliar fungal communities associated with Populus trichocarpa. Samples are pooled among trees; site is nested in region. Year was specified first in the model to account for year effects prior to testing for region and site effects. Region = east or west of Cascades. *significant at alpha = 0.05.

Source d.f. SS MS F R2 P Year 1 2.126 2.126 9.193 0.074 0.001* Region 1 2.837 2.837 12.271 0.098 0.001* Site 8 4.292 0.537 2.320 0.148 0.001* Residuals 76 19.655 0.231 0.680 Total 95 28.911 1

Supplementary Table 2.S5: Significant Populus trichocarpa foliar fungal indicator species for areas either West or East of the Cascades in both years (2013 and 2014) and within the 25 most abundant OTUs overall in order of decreasing abundance (P’ = Holm-corrected p-value significant at alpha = 0.05).

OTU Region P’ (2013/2014) Cladosporium OTU4 West 0.005/0.005 Articulospora OTU3 East 0.005/0.005 Epicoccum OTU14 West 0.005/0.005 Ramularia OTU12 West 0.005/0.005 Melampsora OTU15* East 0.005/0.005 Cladosporium OTU11 East 0.006/0.027 Venturia OTU22* West 0.005/0.005 Aureobasidium OTU17 West 0.005/0.012 Neofabraea OTU18 West 0.005/0.016 Hymenoscyphus OTU45 West 0.005/0.005 Ramularia OTU38 West 0.005/0.005 Alpinaria OTU26 West 0.008/0.048 Ramularia OTU62 West 0.005/0.005 *Genus contains known leaf pathogens of Populus trichocarpa

Supplementary Table 2.S6: Year, Region (east vs. west of Cascades) and sequencing depth effects on leaf-level alpha-diversity (Shannon, Chao1) of foliar fungal communities associated with Populus trichocarpa. Site and tree were included as random effects to account for non- independence of samples. *significant at alpha = 0.05.

Shannon Chao1 Year F1,15=13.33 F1,88=28.67 P=0.0024* P=6.75e-07* Region F1,15=21 F1,90=21.58 P=0.0003* P=1.15e-05* Year × Region F1,15=0.004 F1,88=3.12 P=0.95 P=0.081 log Sequencing Depth F1,262=7.5 F1,242=36.11 P=0.0066* P=6.82e-09*

Supplementary Table 2.S7: Results of PERMANOVA analyses testing the effect of year on Populus trichocarpa foliar fungal community composition at each site. Samples are pooled among trees. *significant at alpha = 0.05.

Site Region F R2 P SK West 1.96 0.197 0.091 SNO West 2.87 0.264 0.011* CAR West 1.68 0.173 0.133 DO West 1.25 0.135 0.31 NM West 1.17 0.127 0.287 TIE East 2.87 0.264 0.024* LS East 6.08 0.432 0.012* KR East 5.83 0.493 0.013* CW East 19.41 0.708 0.009* YK East 5.93 0.497 0.028*

Chapter 3. Intraspecific host genetic effects on the foliar fungal microbiome of Populus trichocarpa diminish through the growing season

Edward G. Barge, Shawn P. Brown, Posy E. Busby

Target Journal: The ISME Journal

ABSTRACT:

Intraspecific plant genetic variation is known to impact the composition of the plant microbiome; however, these effects are typically modest and almost always measured at a single time point. Evaluating host genetic effects on the microbiome throughout the process of its assembly could help to reveal when the host exerts its greatest impact on microbiome composition, richness, and diversity. Here, we used ITS metabarcoding to characterize foliar fungal communities of twelve Populus trichocarpa genotypes in a common garden environment in Corvallis Oregon, USA at 21 timepoints (every 10 days) spanning the tree’s growing season.

Our experiment was fully factorial, with tree genotypes selected to vary in two plant functional traits that are expected to impact leaf microbiome composition: susceptibility to the leaf spot pathogen Sphaerulina populicola, and the timing (phenology) of leaf flush in the spring. We identified differences in the fungal community between early- versus late-phenology tree genotypes and between Sphaerulina leaf spot resistant versus susceptible tree genotypes; however, these differences diminished through time as foliar fungal communities converged. Our results suggest that for the leaves of deciduous plants, intraspecific host genetic variability may have its strongest impact on microbial community composition early in assembly.

INTRODUCTION:

The microbiome is integral to the health and development of its host (Cho & Blaser,

2012; Berendsen et al., 2012; Stefka et al., 2014; Bates et al., 2018). The degree to which a host exerts control over the composition of its microbiome is therefore a question of great importance in both basic and applied plant microbial ecology (Busby et al., 2017). In plants, several studies have shown that intraspecific genetic variation within host plants can shape the composition of both fungal and bacterial communities in the microbiome (Saunders & Kohn, 2009; Bulgarelli et al., 2012; Lundberg et al., 2012; Rastogi et al., 2012; Bálint et al. 2013, Peiffer et al., 2013;

Cregger et al., 2018), with some further demonstrating that microbiome composition can be viewed as a complex, heritable trait (Horton et al., 2014; Wagner et al., 2016; Wallace et al.

2018).

Plants differentially recruit or exclude microbes through functional trait variation. For example, varying the quantity and quality of plant resources available to microbes (e.g. root and leaf exudates) can influence microbiome membership (Berendsen et al., 2012). Additionally, both host plant physical defenses and immune responses can restrict entry into the microbiome.

Leaf cuticle thickness (Reisberg et al., 2013), defensive compounds (Saunders & Kohn, 2009) and genetic resistance to pathogens (Lebeis et al., 2015; Mendes et al., 2018, Wagner et al. 2019) have each been shown to impact plant microbiome composition. However, in the majority of studies characterizing intraspecific host genetic effects on the microbiome, the impacts are modest compared to spatial and environmental effects (Berendsen et al., 2012; Rastogi et al.,

2012; Bálint et al. 2013; Wagner et al., 2016; Wallace et al., 2018), and host effects are almost always quantified at a single time point in microbiome assembly (but see Rastogi et al., 2012;

Wagner et al., 2016; 2019). Thus it is unclear whether host genetic effects on the microbiome are

50 magnified at any point in the assembly process. If genetic effects on microbiome composition are predictably stronger at certain times in assembly, those timepoints could represent windows of opportunity for the management or manipulation of plant-microbiome interactions in agricultural systems.

The leaves of deciduous plants are an excellent study system for examining how plant genetic effects on the microbiome vary during community assembly because leaves re-assemble a new community each growing season. New leaves are a putatively “blank slate” for the colonization of horizontally transmitted microbes found in the environment (Rodriguez et al.,

2009). However, examining host genetic effects on the microbiome for deciduous plant leaves during the assembly process also requires considering that these effects occur against a backdrop of seasonal variation in several factors which also influence microbiome membership. First, the host will age over the course of a growing season, which may change how plant functional traits influence microbiome composition. For example, plant defensive chemistry can change as a leaf ages (Boege et al., 2007). Second, microbes have variable lifecycles that will determine when different species colonize leaves; but overall, the number of microbial species is generally expected to increase after leaves flush and microbes colonize mostly microbe-free plant tissues

(Jumpponen & Jones, 2010; Maignien et al., 2014). Third, abiotic environmental factors that impact microbe colonization and composition, like temperature and humidity, will change over the growing season. Thus the importance of plant functional traits for microbiome composition could vary over the growing season as a result of interactions with seasonal variation in the plant microbiome and the external abiotic environment.

Our study asks whether the impact of plant functional traits on foliar fungal microbiome characteristics (e.g., composition, richness, diversity) varies over the course of the growing

51 season within the model deciduous tree, Populus trichocarpa. Previous studies in this system have found evidence for strong spatial and environmental effects on the foliar fungal microbiome

(Barge et al. 2019), and highlighted the functional importance of fungal leaf endophytes for the severity of foliar diseases (Busby et al. 2016). Susceptibility to foliar pathogens has been suggested as a key plant functional trait impacting leaf microbiome composition in species of

Populus (Cregger et al. 2018), but this hypothesis has not been formally tested.

We examined the foliar fungal microbiome of Populus trichocarpa genotypes that varied in two functional traits that are expected to influence microbiome composition: Sphaerulina leaf spot disease resistance (resistant vs. susceptible) and the phenology of spring leaf flush (early vs. late). Resistance to Sphaerulina populicola, a common pathogen of P. trichocarpa (Callan,

1998), could impact the foliar fungal community directly by precluding its colonization and possibly other fungi (Cregger et al. 2018, Wagner et al. 2019), or indirectly via higher order microbial interactions. Because this pathogen infects leaves early in the growing season (Callan,

1998), we expected that differential genetic resistance to the pathogen would have its strongest impact on microbiome characteristics early in the growing season. We also expected that variation in the second plant functional trait that we examined, leaf phenology, would have its strongest impact on microbiome characteristics early in the growing season. This would occur if early emerging leaves are exposed to a different airborne species pool (source of inoculum) than later emerging leaves, leading to the assembly of initially different communities; in contrast, all leaves are increasingly exposed to the same airborne species pool as the growing season progresses. Thus, our overall hypothesis is that intraspecific variation in disease resistance to an early infecting pathogen and phenology of spring leaf flush have stronger host effects on the microbiome earlier in its assembly compared to later. To test this hypothesis, we characterized

52 foliar fungal communities using DNA metabarcoding of leaves from the same 12 trees (12 different genotypes), sampled every 10 days throughout the growing season (from leaf burst to senescence) in a common garden in Corvallis, OR, USA.

MATERIALS AND METHODS:

Field sampling:

Our study took advantage of an established field planting of a Populus trichocarpa genome wide association study (GWAS) population in Corvallis, OR, USA (Supplementary Fig.

3.S1). Using existing phenotype data for the GWAS population, we identified 12 tree genotypes that varied in two traits: 1) phenology of bud burst (early vs. late; Evans et al., 2014); and 2)

Sphaerulina leaf spot resistance (resistant vs. susceptible; LeBoldus, unpub.). Prior to bud-burst in 2017, we labeled 40 unopened buds per tree with plastic twist-ties. We selected buds that were at roughly the same position on each branch (4–6 buds down from the branch-tip). This enabled us to keep track of and sample leaves that emerged in the spring and aged together throughout the growing season. We sampled leaves (1 leaf per tree) from the same trees every 10 days from bud-burst to leaf senescence. We sampled at 21 time points from March 27 to October 20, 2017 for a total of 238 leaf samples.

Leaf processing and disease scoring:

Immediately after collecting leaves they were placed into clean quart-size zip-top plastic bags and transported to the lab for processing. Sphaerulina leaf spot disease severity was scored for each leaf by calculating the percent leaf area showing disease symptoms using a gridded, translucent quilters grid (1cm2). Leaves were washed for 1 min in 1% Triton X-100 (v:v),

53 followed by three successive rinses in three separate beakers of deionized water (30 sec each rinse) to remove epiphytic communities and surface debris (Brown et al., 2018). After cleaning, nine leaf disks (0.6 cm diameter) per leaf were excised for DNA extraction using a sterilized hole punch. Leaf disk locations were haphazardly selected with four disks from one side of the mid- vein and five from the other side. For the youngest, smallest leaves (early spring), whole leaves were used for extractions. Leaf disks were placed into 1.5mL microcentrifuge tubes and were frozen at -20° C until further processing. All leaves were collected and processed within 4 hours.

Leaf phenology and disease resistance phenotypes:

To verify the pre-determined leaf phenology phenotype (Evans et al., 2014), we recorded the date at which all tagged buds on each tree had flushed. We conducted a students t-test to determine whether the timing of full leaf flush differed between the projected early versus late phenology trees. To verify the pre-determined disease severity phenotype (LeBoldus, unpub.), we tested whether observed disease severity varied by disease resistance (resistant vs. susceptible), leaf phenology (early vs. late), time and all combinations of interactions with a linear mixed-effects model using the R package lme4, function ‘lmer’ (Bates, et al., 2015). Tree was included as a random effect to account for the non-independence of samples. Statistical significance of fixed effects was assessed using Type III ANOVA, and of the random effect using a likelihood ratio test in the R package lmerTest (Kuznetsova, 2016).

DNA extraction, PCR, and sequencing:

DNA was extracted from leaf disks using the 96 well SynergyTM DNA Extraction Kit

(OPS Diagnostics, Bridgewater, NJ, USA) following the manufacturer’s instructions. For the

54 sample homogenization stage of the DNA extraction process, we used a GenoGrinder 2010

(Spex SamplePrep, Metuchen, NJ, USA) to homogenize leaf tissue with a 12 min beat. We included two negative controls (extraction blanks) for each 96 well extraction plate (3 plates total).

We amplified the fungal ITS2 region of the rRNA gene operon using the primers

ITS3_kyo1 (Toju et al., 2012) and ITS4 (White et al., 1990) using a 2-stage PCR. Stage one PCR amplified template out of DNA using gene region-specific primers with Illumina overhang adapters. Stage two PCR attached sample-specific indices and Illumina sequencing adapters. To decrease amplification of the poplar ITS region during stage one PCR, we added a solution of peptide nucleotide acid (PNA) blockers oligos (PNA Bio Inc., Thousand Oaks, CA, USA), based on Cregger et al. (2018), which targets the Populus 5.8S rRNA gene upstream of the ITS2 region. Reactions during stage one PCR were 25 µl and consisted of 12.5 µl MyFiTM Mix DNA polymerase (Bioline, USA), 1.25 µl forward primer, 1.25 µl reverse primer, 2.5 µl PNA clamp

(10 µM), 5 µl molecular grade water, and 2.5 µl template DNA. Reactions during stage two PCR were also 25 µl and consisted of 12.5 µl Kapa HiFi HotStart ReadyMix (Kapa Biosystems,

Woburn, MA, USA), 1.25 µl forward primer, 1.25 µl reverse primer, 9 µl molecular grade water, and 1 µl template from stage one. Thermal cycler conditions for stage one PCR were 3 min at

95° C, followed by 28 cycles of 95° C for 30 sec, 78° C for 5 sec, 50° C for 30 sec, and 72° C for

30 sec, followed by a final elongation at 72° C for 3 min. Thermal cycler conditions for stage 2

PCR were 98° C for 45 sec, followed by 8 cycles of 98° C for 15 sec, 67° C for 30 sec, and 72°

C for 30 sec, followed by a final elongation at 72° C for min. We purified and normalized stage 2

PCR products using the Just-a-PlateTM 96 PCR cleanup and normalization kit (Charm Biotech,

San Diego, CA, USA). Libraries were sequenced at the Center for Genome Research and

Biocomputing at Oregon State University on Illumina MiSeq (one lane, 300-bp paired-end).

Bioinformatics:

Raw reads were demultiplexed using deML (Renaud et al., 2014). Gene primers and adapter sequences were then removed using cutadapt v1.18 (Martin, 2011). Reads were then denoised, ITS sequences were extracted, identical OTUs were collapsed, and chimeras and sequences <75 bp were removed using DADA2 (Callahan et al., 2016). The resulting OTUs were not clustered under a similarity threshold, but were kept as unique sequence variants. Spurious

OTUs were removed by excluding those with <65% match to Kingdom Fungi in the UNITE database (Kõljalg, et al., 2005). Taxonomy was then assigned first at ≥ 50% bootstrap support using the Warcup v2 fungal ITS region database (Deshpande, et al., 2016), and then curated using manual BLASTn queries of GenBank (Clark et al., 2016).

Negative controls contained very few sequencing reads (0–65) and were thus removed from the dataset. To filter the data prior to conducting analyses we removed OTUs making up less than 0.1% of total reads in each sample, and removed samples with fewer than 500 reads (n

= 22). Because this removed many early-season samples (n = 21) due to low sequencing depth, we chose to restrict downstream sequence-based analyses to time point five (April 26, 2017) and beyond. Prior to calculating an abundance-based beta diversity metric (Bray-Curtis), we normalized variable sequencing depth by calculating the proportional abundance of each OTU in each sample. Prior to calculating alpha diversity metrics (Richness, Shannon diversity) and a presence-absence-based beta diversity metric (Jaccard), we rarefied samples to 500 reads.

Beta diversity and community composition:

We quantified beta diversity by calculating both Bray-Curtis and Jaccard dissimilarity among leaves using the R package phyloseq (McMurdie & Holmes, 2013). First, to directly test our hypothesis that disease resistance and leaf phenology would have a stronger impact on community composition early in the growing season, we fit two linear models: one included mean pairwise Bray-Curtis dissimilarity between early vs. late phenology trees at each timepoint as the response variable and time as the predictor variable, and the other included mean pairwise

Bray-Curtis dissimilarity between resistant vs. susceptible trees at each timepoint as the response variable and time as the predictor variable. We also conducted these analyses using mean pairwise Jaccard dissimilarity. Second, to further test whether beta diversity varied by time, disease resistance, leaf phenology and all interactions, we conducted principal coordinates analysis (PCoA) of Bray-Curtis and Jaccard dissimilarities among leaves using the R package vegan (Oksanen et al., 2007), and tested whether the top four PCoA axes of variation in community composition varied by time, disease resistance, leaf phenology and all interactions by fitting linear mixed-effects models (Bates, et al., 2015). Tree was included as a random effect to account for repeated sampling (non-independence of samples). Statistical significance of fixed effects was assessed using Type III ANOVA, and of the random effect using likelihood ratio tests in the R package lmerTest (Kuznetsova, 2016).

Alpha diversity:

We estimated species richness and Shannon diversity for each sample using the R package phyloseq, function ‘estimate_richness’ (McMurdie & Holmes, 2013). To test whether alpha diversity varied by time, disease resistance, leaf phenology and all interactions, we fit a

57 linear mixed-effects model using the R package lme4, function ‘lmer’ (Bates, et al., 2015).

Again, tree was included as a random effect to account for the non-independence of samples.

Statistical significance of fixed effects was assessed using Type III ANOVA, and of the random effect using a likelihood ratio test in the R package lmerTest (Kuznetsova, 2016).

RESULTS:

Leaf phenology and disease resistance phenotypes:

The first trees to achieve full bud-burst at all tagged buds did so by March 27, 2017. By

April 19, 2017, full bud-burst had occurred on all trees. Early phenology trees achieved full bud- burst an average of 11.2 days before late phenology trees (t8=-3.59, P=0.007) (Fig. 3.1). Disease severity varied between resistant and susceptible plants and was greater on susceptible plants

(P=0.003) (Fig. 3.1, Supplementary Table 3.S1). However, the effect of disease resistance on disease severity was also time-dependent, with the difference in disease severity between resistant and susceptible trees increasing over time (P<0.001) (Fig. 3.1, Supplementary Table

3.S1). Disease severity also generally increased over time (P<0.001) (Fig. 3.1, Supplementary

Table 3.S1). Phenology (nor any interaction including phenology) did not have an impact on disease severity (Fig. 3.1, Supplementary Table 3.S1).

Beta diversity and community composition:

We found support for our hypothesis that intraspecific variation in disease resistance to an early infecting pathogen and phenology of spring leaf flush have stronger host effects on microbiome composition early in assembly. We found that mean pairwise Bray-Curtis dissimilarity between early vs. late phenology trees and between disease resistant vs. susceptible

58 trees decreased over time (Fig. 3.2). We also found significant effects of time x phenology and time x disease resistance on several of the top PCoA axes (Table 3.1, Fig. 3.4, Supplementary

Fig. 3.S2), reflecting that differences in community composition between both early and late phenology trees and resistant and susceptible trees, diminished over time. We also found a strong overall effect of time on community composition, and a smaller effect of phenology and disease resistance; all four PCoA axes varied with time, whereas phenology by itself only predicted variation in PCoA2 and PCoA4, and disease resistance by itself only predicted variation in

PCoA2 (Table 3.1, Fig. 3.4, Supplementary Fig. 3.S2). Results were similar for the presence- absence based Jaccard dissimilarity metric (not shown).

Alpha diversity:

We found partial support for our hypothesis that intraspecific variation in disease resistance and phenology have stronger host effects on microbiome alpha diversity earlier in assembly compared to later. We found a significant effect of time x phenology on richness, with differences in richness between early vs. late phenology trees decreasing over the course of the growing season (Table 3.2, Supplementary Fig. 3.S3). However, the effect of time x phenology on Shannon diversity was not significant (Table 3.2, Supplementary Fig. 3.S3). We found a significant effect of disease resistance x phenology on both richness and Shannon diversity; however, without the clear pattern of convergence in richness as seen for early vs. late phenology trees (Table 3.2, Supplementary Fig. 3.S3). We also found that both richness and Shannon diversity increased until around the middle of the growing season (Table 3.2, Supplementary Fig.

3.S3).

DISCUSSION:

Over the course of a single growing season, we observed a variable microbiome within the leaves of twelve Populus trichocarpa trees varying in leaf phenology and resistance to

Sphaerulina leaf spot disease. Overall, fungal composition, richness, and diversity changed substantiality over time, which is in line with the findings of many previous studies (Thompson et al., 1993; Suryanarayanan & Thennarasan, 2004; Jumpponen & Jones, 2010; Ek-Ramos et al.,

2013; Fort et al., 2016; Giauque & Hawkes, 2016; Wagner et al., 2019). In particular, we found:

1) fungal richness and Shannon diversity increased over time, 2) community composition converged over time, and 3) the impact of host functional traits on fungal community composition diminished over time. These results support our hypothesis that the phenology of spring leaf flush and disease resistance to an early infecting pathogen have stronger impacts on the microbiome earlier in the growing season than later in the growing season. Our study thus raises important questions about the relationship between host genetic effects and community convergence within the plant microbiome.

Intraspecific host genetic variation and microbial community convergence:

Intraspecific host genetic variation has been linked to microbial community composition in many studies (Saunders & Kohn, 2009; Bulgarelli et al., 2012; Lundberg et al., 2012; Rastogi et al., 2012; Bálint et al. 2013, Peiffer et al., 2013; Horton et al., 2014; Wagner et al., 2016;

2019; Cregger et al., 2018; Wallace et al. 2018), with several studies further linking community composition to variation in plant functional traits such as leaf cuticle thickness (Reisberg et al.,

2013), defensive compounds (Saunders & Kohn, 2009) and genetic resistance to pathogens

(Lebeis et al., 2015; Mendes et al., 2018, Wagner et al. 2019). However, we are not aware of any

60 studies that have evaluated the relationship between plant host genetic effects and microbial community convergence. There are several ways in which diminishing host genetic effects could have been causally related to microbial community convergence in our study. Variation in one of the plant functional traits that we examined, genetic resistance to Sphaerulina leaf spot disease, may have had its strongest impact on microbial community assembly early in the growing season due to the timing of pathogen infection. The Sphaerulina populicola leaf spot pathogen infects P. trichocarpa leaves early in community assembly (Callan, 1998), thus host genetic variation in resistance to this pathogen may have had its strongest impact on fungal community assembly during this time, by deterministically selecting for or against colonization by the pathogen and possibly other fungi. The initial effect of disease resistance on early community assembly may have then decreased over time as the host immune response subsided and other forces favoring convergence (see below) took over.

Variation in the other plant functional trait that we examined, phenology of spring leaf flush, may have had its strongest impact on microbial community assembly early in the growing season as early emerging leaves may have initially been exposed to and inoculated by a different airborne microbial species pool than later emerging leaves. Priority effects, where early arriving species prevent the establishment of later arriving species (Fukami, 2015) may have played a role in driving some of the differences in community composition we saw between early and late phenology trees, early in the growing season. However, as time progressed, all leaves were likely increasingly exposed to the same airborne microbial species pool and communities converged as other assembly processes became dominant (see below).

Other factors driving convergence:

While intraspecific plant host genetic variation has been shown to play a role in microbial community composition, it is unclear whether there may be a causal relationship between diminishing host genetic effects and community convergence. For example, the reduction in the importance of host genetics that we observed may have coincided with other factors gaining in importance over time and pushing communities to converge. Abiotic environmental filtering driven by increasing homogeneity of key abiotic environmental factors over time has been implicated as one mechanism driving microbial community convergence. For example, Castle et al. (2016) found that convergence in soil organic carbon and soil organic chemistry along retreating glacier forefronts was associated with convergence in soil microbial community composition (Castle et al., 2016). We did not measure local abiotic variables in our study, however, inclusion of ultraviolet radiation, temperature, and moisture measurements in future studies should be done as these variables have been found to be important for leaf microbial communities (Fierer et al., 2010; Vorholt, 2012).

Microbial community convergence on plant leaves has also been suggested to be an emergent property, deterministically driven by the niche structure provided by leaves. For example, Maignien et al. (2014) found that phyllosphere bacterial communities on greenhouse- grown Arabidopsis thaliana plants initially mirrored airborne communities, but over time, the plants converged to a compositionally distinct community. The authors suggest this pattern may have been driven by the similar niche structures provided by the plant leaves leading to deterministic selection for community membership from the airborne species pool and ultimate convergence among plants (Maignien et al., 2014). A similar pattern was found by Copeland et al. (2015), who showed that leaf-specific microbial communities converged among several

62 species of annual plants over the course of the growing season. Thus, in the case of our study, the effects of similar niche structures (the leaf environment) across all sampled trees may have ultimately overridden the effects of host genetic variability on community assembly. However, divergence in leaf-associated microbial community composition over time has been found in other studies (Fort et al., 2016; Wagner et al., 2019). Fort et al. (2016) found that foliar fungal communities of adjacent vineyards and forest trees diverged from each other over the course of the growing season, which they attributed to selective pressures exerted in each habitat by the different host plants as well as differences in microclimate. While this study differed from ours in investigating multiple plant species, it suggests the possibility that strong enough intraspecific variability among plants could drive microbial community divergence over time. Wagner et al.

(2019) also found increasing beta diversity of foliar fungal communities of maize across two timepoints, although possible mechanisms underlying this divergence were not determined.

Further, leaf age has been shown to affect leaf chemistry (Boege et al., 2007), which in turn can affect microbial community composition (Kembel et al., 2014). Thus, leaf age may have also played an important role in the pattern of convergence we observed. Future studies should monitor leaf age-associated changes in leaf chemistry and other properties in relation to microbial community assembly.

Microbial community convergence has also been shown to be driven by more stochastic assembly processes. For example, increasing microbial dispersal can lead to mass effects, which make local communities more similar to each other and to the regional species pool (Leibold et al., 2004; Stegen et al., 2015; Evans et al., 2017). Mass effects are also known to be associated with increases in local alpha diversity, such that it approaches the diversity of the regional species pool (Evans et al., 2017), which in turn is known to decrease beta diversity and thus drive

63 convergence (Chase & Myers, 2011). We found increasing fungal alpha diversity (possibly driven by increased dispersal) over time in our study, and this may have been a key driver of the pattern of convergence we observed. However, in order to more fully tease apart drivers of foliar microbiome community convergence on annual and deciduous plants, manipulative experiments controlling for host genetic variation, changes in the airborne species pool, changes in the leaf environment due to leaf age, and changes in the abiotic environment are necessary.

Conclusions:

We found support for our hypothesis that the phenology of spring leaf flush and disease resistance to an early infecting pathogen have stronger impacts on the leaf microbiome of

Populus trichocarpa earlier in the growing season than later in the growing season. These results suggest that for the leaves of deciduous plants, intraspecific host genetic variability may have its strongest impact on microbial community composition early in assembly. However, whether this hypothesis can be generalized to include other plants species and other plant traits is not clear.

For one, it is necessary to investigate a much broader range of intraspecific host genetic variability and host functional traits than has been investigated to date. In the case of Populus trichocarpa, it is necessary to investigate traits which might be expected to have a stronger influence later in the growing season, such as resistance to a later emerging pathogen. Further, while at least two studies have pointed to the possibility that microbial community convergence on plant leaves is an emergent property (Maignien et al., 2014; Copeland et al., 2015), divergence in leaf-associated microbial communities has been observed as well (Fort et al., 2016;

Wagner et al., 2019). Thus, there remains a need to explore leaf-associated microbial community succession over a wider range of plant species and environments before general conclusions can

64 be made regarding the impact of intraspecific host genetic variability on microbial community assembly.

REFERENCES:

Bálint, M., Tiffin, P., Hallström, B., O’Hara, R. B., Olson, M. S., Frankhauser, J. D., Piepenbring, M., & Schmitt, I. (2013). Host genotype shapes the foliar fungal microbiome of balsam poplar (Populus balsamifera). PLoS One, 8(1), e53987. Barge, E. G., Leopold, D. R., Peay, K. G., Newcombe, G., & Busby, P. E. (2019). Differentiating spatial from environmental effects on foliar fungal communities of Populus trichocarpa. Journal of Biogeography, DOI:10.1111/jbi.13641. Bates, K. A., Clare, F. C., O’Hanlon, S., Bosch, J., Brookes, L., Hopkins, K., McLaughlin, E., Daniel, O., Garner, T. W. J., Fisher, M. C., & Harrison, X. A. (2018). Amphibian chytridiomycosis outbreak dynamics are linked with host skin bacterial community structure. Nature Communications, 9(1), 693. Bates, D., Maechler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67, 1–48. Berendsen, R. L., Pieterse, C. M., & Bakker, P. A. (2012). The rhizosphere microbiome and plant health. Trends in Plant Science, 17(8), 478–486. Boege, K., Barton, K. E., & Dirzo, R. (2011). Influence of tree ontogeny on plant-herbivore interactions. Size- and Age-Related Changes in Tree Structure and Function, Vol. 4 (eds F. C. Meinzer, B. Lachenbruch & T. E. Dawson), pp. 193–214. Springer, Dordrecht, The Netherlands. Brown, S. P., Leopold, D. R., & Busby, P. E. (2018). Protocols for investigating the leaf mycobiome using high throughput DNA sequencing. In: Ma, W. & Wolpert, T. (Eds.), Plant Pathogenic Fungi and Oomycetes: Methods in Molecular Biology, vol. 1848. Humana Press, New York, pp. 39–51. Bulgarelli, D., Rott, M., Schlaeppi, K., van Themaat, E. V. L., Ahmadinejad, N., Assenza, F., Rauf, P., Huettel, B., Reinhardt, R., Schmelzer, E., Peplies, J., Gloeckner, F. O., Amann, R., Eickhorst, T., & Schulze-Lefert, P. (2012). Revealing structure and assembly cues for Arabidopsis root-inhabiting bacterial microbiota. Nature, 488(7409), 91. Busby, P. E., Peay, K. G., & Newcombe, G. (2016). Common foliar fungi of Populus trichocarpa modify Melampsora rust disease severity. New Phytologist, 209(4), 1681– 1692. Busby, P. E., Soman, C., Wagner, M. R., Friesen, M. L., Kremer, J., Bennett, A., Morsy, M., Eisen, J. A., Leach, J. E., & Dangl, J. L. (2017). Research priorities for harnessing plant microbiomes in sustainable agriculture. PLoS Biology, 15(3), e2001793. Callahan, B. J., McMurdie, P. J., Rosen, M. J., Han, A. W., Johnson, A. J. A., & Holmes, S. P. (2016). DADA2: high-resolution sample inference from Illumina amplicon data. Nature Methods, 13(7), 581. Callan, B. E. (1998). Diseases of Populus in British Columbia: a diagnostic manual. Canadian Forest Service. Castle, S. C., Nemergut, D. R., Grandy, A. S., Leff, J. W., Graham, E. B., Hood, E., Schmidt, S. K., Wickings, K., & Cleveland, C. C. (2016). Biogeochemical drivers of microbial community convergence across actively retreating glaciers. Soil Biology and Biochemistry, 101, 74–84. Chase, J. M., & Myers, J. A. (2011). Disentangling the importance of ecological niches from stochastic processes across scales. Philosophical transactions of the Royal Society B: Biological Sciences, 366(1576), 2351–2363.

Cho, I. & Blaser, M. J. (2012). The human microbiome: at the interface of health and disease. Nature Review Genetics, 13(4), 260. Clark, K., Karsch-Mizrachi, I., Lipman, D. J., Ostell, J., & Sayers, E. W. (2016). GenBank. Nucleic Acids Research, 44, D67–D72. Copeland, J. K., Yuan, L., Layeghifard, M., Wang, P. W., Guttman, D. S. (2015). Seasonal community succession of the phyllosphere microbiome. Molecular Plant-Microbe Interactions, 28(3), 274–85. Cregger, M. A., Veach, A. M., Yang, Z. K., Crouch, M. J., Vilgalys, R., Tuskan, G. A., & Schadt, C. W. (2018). The Populus holobiont: dissecting the effects of plant niches and genotype on the microbiome. Microbiome, 6(1), 31. Deshpande, V., Wang, Q., Greenfield, P., Charleston, M., Porras-Alfaro, A., Kuske, C. R., Cole, J. R., Midgley, D. J., & Tran-Dinh, N. (2016). Fungal identification using a Bayseian classifier and the Warcup training set of internal transcribed spacer sequences. Mycologia, 108(1), 1–5. Ek-Ramos, M. J., Zhou, W., Valencia, C. U., Antwi, J. B., Kalns, L. L., Morgan, G. D., Kerns, D. L., & Sword, G. A. (2013). Spatial and temporal variation in fungal endophyte communities isolated from cultivated cotton (Gossypium hirsutum). PLoS One, 8(6), e66049. Evans, L. M., Slavov, G. T., Rodgers-Melnick, E., Martin, J., Ranjan, P., Muchero, W., Brunner, A. M., Schackwitz, W., Gunter, L., Chen, J.-G., Tuskan, G. A., & DiFazio, S. P. (2014). Population genomics of Populus trichocarpa identifies signatures of selection and adaptive trait associations. Nature Genetics, 46(10), 1089. Evans, S., Martiny, J. B., & Allison, S. D. (2017). Effects of dispersal and selection on stochastic assembly in microbial communities. The ISME Journal, 11(1), 176. Fierer, N., Nemergut, D., Knight, R., & Craine, J. M. (2010). Changes through time: integrating microorganisms into the study of succession. Research in Microbiology, 161(8), 635– 642. Fort, T., Robin, C., Capdevielle, X., Delière, L., & Vacher, C. (2016). Foliar fungal communities strongly differ between habitat patches in a landscape mosaic. PeerJ, 4, e2656. 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. Giauque, H., & Hawkes, C. V. (2016). Historical and current climate drive spatial and temporal patterns in fungal endophyte diversity. Fungal Ecology, 20, 108–114. Horton, M. W., Bodenhausen, N., Beilsmith, K., Meng, D., Muegge, B. D., Subramanian, S., Vetter, M. M., Vilhjálmsson, B. J., Nordborg, M., Gordon, J. I., & Bergelson, J. (2014). Genome-wide association study of Arabidopsis thaliana leaf microbial community. Nature Communications, 5, 5320. Jumpponen, A., & Jones, K. L. (2010). Seasonally dynamic fungal communities in the Quercus macrocarpa phyllosphere differ between urban and nonurban environments. New Phytologist, 186(2), 496–513. Kembel, S. W., O’Connor, T. K., Arnold, H. K., Hubbell, S. P., Wright, S. J., & Green, J. L. (2014). Relationships between phyllosphere bacterial communities and plant functional traits in a neotropical forest. Proceedings of the National Academy of Sciences, 111(38), 13715–13720.

Kõljalg, U., Larsson, K. H., Abarenkov, K., Nilsson, R. H., Alexander, I. J., Eberhardt, U., … Pennanen, T. (2005). UNITE: a database providing web-based methods for the molecular identification of ectomycorrhizal fungi. New Phytologist, 166(3), 1063–1068. Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2016). lmerTest: Tests in Linear Mixed Effects Models. R package version 2.0–3.0. Lebeis, S. L., Paredes, S. H., Lundberg, D. S., Breakfield, N., Gehring, J., McDonald, M., Malfatti, S., Del Rio, T. G., Jones, C. D., Tringe, S. G., & Dangl, J. L. (2015). Salicylic acid modulates colonization of the root microbiome by specific bacterial taxa. Science, 349(6250), 860–864. Leibold, M. A., Holyoak, M., Mouquet, N., Amarasekare, P., Chase, J. M., Hoopes, M. F., Holt, R. D., Shurin, J. B., Law, R., Tilman, D., & Loreau, M. (2004). The metacommunity concept: a framework for multi-scale community ecology. Ecology Letters, 7(7), 601– 613. Lundberg, D. S., Lebeis, S. L., Paredes, S. H., Yourstone, S., Gehring, J., Malfatti, S., Tremblay, J., Engelbrektson, A., Kunin, V., Del Rio, T. G., & Edgar, R. C. (2012). Defining the core Arabidopsis thaliana root microbiome. Nature, 488(7409), 86. Maignien, L., DeForce, E. A., Chafee, M. E., Eren, A. M., Simmons, S. L. (2014). Ecological succession and stochastic variation in the assembly of Arabidopsis thaliana phyllosphere communities. MBio 5(1), e00682–13. Martin, M. (2011). Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.Journal, 17(1), 10. McMurdie, P. J., & Holmes, S. (2013). Phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data. PLoS One, 8(4), e61217. Mendes, L. W., Raaijmakers, J. M., de Hollander, M., Mendes, R., & Tsai, S. M. (2018). Influence of resistance breeding in common bean on rhizosphere microbiome composition and function. The ISME Journal, 12(1), 212. Oksanen, J., Kindt, R., Legendre, P., O’Hara, B., Stevens, M. H. H., Oksanen, M. J., & Suggests, M. A. (2007). The vegan package. Community Ecology Package, 10, 631–637. Peiffer, J. A., Spor, A., Koren, O., Jin, Z., Tringe, S. G., Dangl, J. L., Buckler, E. S., & Ley, R. E. (2013). Diversity and heritability of the maize rhizosphere microbiome under field conditions. Proceedings of the National Academy of Sciences, 110(16), 6548–6553. Rastogi, G., Sbodio, A., Tech, J. J., Suslow, T. V., Coaker, G. L., & Leveau, J. H. (2012). Leaf microbiota in an agroecosystem: spatiotemporal variation in bacterial community composition on field-grown lettuce. The ISME Journal, 6(10), 1812. Reisberg, E. E., Hildebrandt, U., Riederer, M., & Hentschel, U. (2013). Distinct phyllosphere bacterial communities on Arabidopsis wax mutant leaves. PLoS One, 8(11), e78613. Renaud, G., Stenzel, U., Maricic, T., Wiebe, V., & Kelso, J. (2014). deML: robust demultiplexing of Illumina sequences using a likelihood-based approach. Bioinformatics, 31(5), 770–772. Rodriguez, R. J., White Jr, J. F., Arnold, A. E., & Redman, A. R. A. (2009). Fungal endophytes: diversity and functional roles. New Phytologist, 182(2), 314–330. Saunders, M., & Kohn, L. M. (2009). Evidence for alteration of fungal endophyte community assembly by host defense compounds. New Phytologist, 182, 229–238. Stefka, A. T., Feehley, T., Tripathi, P., Qiu, J., McCoy, K., Mazmanian, S. K., Tjota, M. Y., Seo, G. Y., Cao, S., Theriault, B. R., Antonopoulos, D. A., Zhou, L., Chang, E. B., Fu, Y. X.,

& Nagler, C. R. (2014). Commensal bacteria protect against food allergen sensitization. Proceedings of the National Academy of Sciences, 111(36), 13145–13150. Stegen, J. C., Lin, X., Fredrickson, J. K., & Konopka, A. E. (2015). Estimating and mapping ecological processes influencing microbial community assembly. Frontiers in Microbiology, 6, 370. Suryanarayanan, T. S., & Thennarasan, S. (2004). Temporal variation in endophyte assemblages of Plumeria rubra leaves. Fungal Diversity, 15, 197–204. Thompson, I. P., Bailey, M. J., Fenlon, J. S., Fermor, T. R., Lilley, A. K., Lynch, J. M., McCormack, P. J., McQuilken, M. P., Purdy, K. J., Rainey, P. B., & Whipps, J. M. (1993). Quantitative and qualitative seasonal changes in the microbial community from the phyllosphere of sugar beet (Beta vulgaris). Plant and Soil, 150(2), 177–191. Toju, H., Tanabe, A. S., Yamamoto, S., & Sato, H. (2012). High-coverage ITS primers for the DNA-based identification of ascomycetes and basidiomycetes in environmental samples. PloS One, 7(7), e40863. Vorholt, J. A. (2012). Microbial life in the phyllosphere. Nature Reviews Microbiology, 10, 828– 840. Wagner, M. R., Lundberg, D. S., Tijana, G., Tringe, S. G., Dangl, J. L., & Mitchell-Olds, T. (2016). Host genotype and age shape the leaf and root microbiomes of a wild perennial plant. Nature Communications, 7, 12151. Wagner, M. R., Busby, P. E., & Balint-Kurti, P. (2019). Analysis of leaf microbiome composition of near-isogenic maize lines differing in broad-spectrum disease resistance. New Phytologist, accepted. http://biorxiv.org/cgi/content/short/647446v1 Wallace, J. G., Kremling, K. A., Kovar, L. L., & Buckler, E. S. (2018). Quantitative genetics of the maize leaf microbiome. Phytobiomes Journal, 2(4), 208–224. White, T. J., Bruns, T., Lee, S., & Taylor, J. (1990). Amplification and direct sequencing of fungal ribosomal RNA genes for phylogenetics. PCR protocols: a guide to methods and applications, 18(1), 315–322.

Figure 3.1: Verification of leaf phenology and disease resistance phenotypes. (a) Date of full bud-burst at tagged buds differed for early and late phenology trees (P=0.007). (b) Sphaerulina leaf spot disease resistance, time, and disease resistance x time each predicted variation in disease severity (P=0.003; P<0.001, P<0.001). Error bars represent standard error of the mean.

Figure 3.2: Convergence in community composition (mean pairwise Bray-Curtis dissimilarity) (a) between early vs. late phenology trees over time and (b) between disease resistant vs. susceptible trees over time. Results of the linear model (lm) are displayed in the upper right hand corner of each panel. Gray area indicates 95% confidence interval around lm smoothed line.

Figure 3.3: (a) PCoA of Bray-Curtis dissimilarity among leaf samples displaying convergence in community composition over time. (b) Linear mixed-effects models revealed that community composition (certain PCoA axes) converged over time between early vs. late phenology trees (top panel) and between resistant vs. susceptible trees (bottom panel) (detailed statistics are found in Table 3.1). For brevity, only two examples are displayed (the full set of figures displaying the relationship between the top four PCoA axes, plant traits, and time are displayed in Supplementary Fig. 3.S2). Error bars represent standard error of the mean.

Supplementary Figure 3.S1: (a) Map showing the location of the study site (Populus trichocarpa genome-wide association study common garden) in Corvallis, Oregon, USA. (b) Populus trichocarpa genotypes sampled in this study and their associated phenology and disease resistance phenotypes.

Supplementary Figure 3.S2: Variation in community composition (PCoA axes 1-4) over time and (a) between early vs. late phenology trees and (b) between disease resistant vs. susceptible trees. Detailed statistics from linear mixed-effects models are found in Table 3.S1.

Supplementary Figure 3.S3: Variation in leaf-level foliar fungal richness (a) and Shannon diversity (b) over time and between early vs. late phenology trees. Variation in richness (c) and Shannon diversity (d) over time and between disease resistant vs. susceptible trees. Detailed statistics from linear mixed-effects models are found in Table 3.S2. Error bars represent standard error of the mean.

Table 3.1: Results of linear mixed-effects model testing Phenology, Disease resistance and Timepoint effects on foliar fungal community composition as represented by PCoA axes 1–4. Tree was included as a random effect to account for non-independence of samples. *significant at alpha = 0.05. Statistics for Phenology x Timepoint and Disease resistance x Timepoint are shown in bold.

Predictor PCoA1 PCoA2 PCoA3 PCoA4 (43%) (11%) (8%) (7%) Phenology F1,16=1.20 F1,27=8.12 F1,84=0.52 F1,26=15.17 P=0.29 P=0.008* P=0.47 P<0.001* Disease resistance F1,16=5.15 F1,27=18.69 F1,84=3.21 F1,26=0.42 P=0.06 P<0.001* P=0.08 P=0.52 Timepoint F1,195=774.54 F1,195=32.16 F1,195=11.36 F1,195=11.81 P<0.001* P<0.001* P<0.001* P<0.001* Phenology x Disease F1,16=14.25 F1,27=8.12 F1,84=0.39 F1,26=4.79 resistance P=0.002* P<0.008* P=0.54 P=0.04* Phenology x F1,195=0.38 F1,195=3.48 F1,195=1.70 F1,195=18.87 Timepoint P=0.54 P=0.06 P=0.19 P<0.001* Disease resistance x F1,195=4.34 F1,195=18.36 F1,195=0.28 F1,195=1.76 Timepoint P=0.04* P<0.001* P=0.60 P=0.19 Phenology x Disease F1,195=11.30 F1,195=8.41 F1,195=0.76 F1,195=2.42 resistance x Timepoint P=0.001* P=0.004* P=0.38 P=0.12 2 2 2 2 Tree Χ 1=70.27 Χ 1=32.75 Χ 1=4.56 Χ 1=35.83 P<0.001* P<0.001* P=0.03* P=<.001*

Table 3.2: Results of linear mixed-effects model testing Phenology, Disease resistance and Timepoint effects on leaf-level alpha diversity (Richness, Shannon) of foliar fungal communities. Tree was included as a random effect to account for non-independence of samples. *significant at alpha = 0.05. Statistics for Phenology x Timepoint and Disease resistance x Timepoint are shown in bold.

Predictor Richness Shannon Phenology F1,49=7.74 F1,30=0.12 P=0.008* P=0.73 Disease resistance F1,49=1.91 F1,30=5.49 P=0.17 P=0.026* Timepoint F1,195=48.82 F1,195=122.25 P<0.001* P<0.001* Phenology x Disease resistance F1,49=8.00 F1,30=23.28 P=0.007* P<0.001* Phenology x Timepoint F1,195=4.14 F1,195=0.004 P=0.04* P=0.95 Disease resistance x Timepoint F1,195=5.54 F1,195=9.37 P=0.02* P=0.003* Phenology x Disease resistance x Timepoint F1,195=4.81 F1,195=16.01 P=0.03* P<0.001* 2 2 Tree Χ 1=12.93 Χ 1=28.04 P<0.001* P<0.001*

Supplementary Table 3.S1: Results of linear mixed-effects model testing Phenology, Disease resistance and Timepoint effects on leaf-level Sphaerulina leaf spot disease severity. Tree was included as a random effect to account for non-independence of samples. *significant at alpha = 0.05.

Predictor Disease severity Phenology F1,72=0.01 P=0.91 Disease resistance F1,72=9.15 P=0.003* Timepoint F1,195=114.68 P<0.001* Phenology x Disease resistance F1,72=0.005 P=0.95 Phenology x Timepoint F1,195=0.007 P=0.93 Disease resistance x Timepoint F1,195=57.72 P<0.001* Phenology x Disease resistance x Timepoint F1,195=1.99 P=0.16 2 Tree Χ 1=6.90 P=0.009*

Chapter 4. Phylogenetic relatedness among Cladosporium leaf endophytes predicts their ability to reduce the severity of a poplar leaf rust disease

Edward G. Barge, Alejandro Rojas, Devin R. Leopold, Rytas Vilgalys, Posy E. Busby

Target Journal: Molecular Ecology

ABSTRACT:

More closely related organisms are expected to function more similarly than distantly related organisms due to shared ancestry and functional trait heritability. However, there have been few tests of this hypothesis for fungal leaf endophytes, which can modify host plant disease severity by a variety of mechanisms. We tested whether phylogenetic relatedness within

Cladosporium, a genus including many common fungal leaf endophyte species, predicts endophyte effects on cottonwood leaf rust disease severity caused by Melampsora × columbiana. First, we used multilocus sequence typing to infer phylogenetic relationships among

96 Cladosporium isolates collected from wild cottonwood trees growing in the Pacific Northwest of North America. Next, we conducted a double-inoculation leaf-disk assay (endophyte inoculated first, then rust pathogen) for a subset of 50 Cladosporium isolates to characterize disease modification for the endophyte isolates; data on endophytes parasitizing rust was collected simultaneously for each isolate. We used generalized linear models to link disease modification and mycoparisitic ability to endophyte phylogeny, while accounting for endophyte geographic origin. We recognized 17 distinct species of Cladosporium; all fifty isolates of

Cladosporium reduced rust disease severity in our leaf disk assay (by as much as 79% and as little as 45%). Cladosporium phylogeny was a significant predictor of rust disease severity and was also correlated with mycoparasitism. The geographic origin of the isolates explained only a small amount of the overall variation in disease reduction. Our results demonstrate that fungal endophyte phylogenetic relatedness can help predict differences in endophyte function.

INTRODUCTION:

Plant-associated fungi – e.g., pathogens, mycorrhizae, endophytes, saprotrophs – are integral to plant ecology and ecosystem function (Van Der Heijden et al., 1998; Bakker et al.,

2012), yet our ability to predict their function is constrained by limited data on the distribution of key fungal traits. Phylogenetic relatedness and functional similarity have been shown to correlate in a few studies of plant-associated fungi (Powell et al., 2009; McGuire et al., 2010; Kia et al.,

2017; Hoeksema et al., 2018), suggesting the possibility of using taxonomy to predict function.

However, fungal ecological guilds – groups defined by the resource use of members – are typically not well conserved at high taxonomic levels due to convergent evolution (Zanne et al.,

2019). Thus, while previous studies have identified broad correlations between fungal phylogeny and function across some fungal guilds, it is less clear if phylogenetic relationships can help predict ecological function within more recently diverged lineages of plant-associated fungi (e.g., genera). This approach could be particularly useful for helping to identify beneficial fungi in highly polyphyletic, and poorly characterized guilds, like foliar endophytes.

Fungal leaf endophytes are non-pathogenic fungi living within plant leaves. Many leaf endophytes are commensal symbionts; others can improve host fitness by promoting growth or providing protection from biotic and abiotic stress (Rodriguez et al., 2009). While endophytes do not cause disease, increasingly they are recognized as interacting with plants and pathogens in ways that modify plant disease severity (Arnold et al., 2003; Busby et al., 2016a). Taxonomically diverse endophytes can modify disease severity, indicating that this function is not phylogenetically constrained at a high taxonomic level (Pandey et al., 1993; Perello et al., 2002;

Hanada et al., 2010). At low taxonomic levels, i.e., isolates within a genus or species, variable effects on plant disease severity have also been observed (Busby et al., 2016b), but to our

81 knowledge, no attempt has been made to link patterns of phylogenetic relatedness to disease modification by fungal endophytes. Developing a framework for predicting endophyte function is critical for furthering our understanding of endophyte evolution and ecology, and advancing efforts to utilize endophytes for plant conservation and crop improvement (Busby et al., 2017).

In this study, we explored whether the phylogenetic relatedness of fungal leaf endophytes in the widespread genus Cladosporium (Bensch et al., 2012) is correlated with foliar disease modification in Populus trichocarpa, a model tree in genomics and ecology (Jansson & Douglas,

2007). P. trichocarpa hosts a diverse community of foliar endophytes (Barge, et al., 2019), some of which can modify the severity of disease caused by Melampsora leaf rust (Busby et al.,

2016b). Melampsora leaf rust diseases are a major concern in the poplar industry because they reduce the growth and survival of plantation trees (Widin & Schipper, 1981; Ostry et al., 1989).

We focused on endophytes in the genus Cladosporium, which can reduce Melampsora leaf rust disease severity by mycoparasitizing rust pathogens (Sharma & Heather, 1988; Assante et al.,

2004), though other mechanisms of rust disease antagonism by Cladosporium spp. are also possible (Moricca et al., 2005).

We tested the hypothesis that more closely related individuals within Cladosporium modify rust disease severity more similarly than distantly related individuals. We first characterized the phylogenetic relationships among 96 Cladosporium isolates collected from eight sites across the tree’s geographic range. Next we conducted a double-inoculation leaf-disk assay (endophyte inoculated first, then rust pathogen) for a subset of 50 Cladosporium isolates to quantify their effect on rust disease severity. Finally, we conduced a second double-inoculation leaf-disk assay that included an autoclaved spore slurry treatment and spore slurry filtrate treatment to better understand whether variation in disease severity could be attributed to

82 mycoparasitism, or other potential mechanisms underlying rust disease modification by

Cladosporium spp.

MATERIALS AND METHODS:

Field sampling and culturing:

To generate a culture collection of foliar Cladosporium spp. from P. trichocarpa, we sampled eight different tree populations in October, 2016. Half of the tree populations were located west, and the other half east, of the Cascade Range which bisects the tree’s geographic range (Fig. 4.1, Supplementary Table 4.S1). Melampsora disease pressure has historically been greater in the wet, mild environment west of the Cascade Range (Dunlap & Stettler, 1996), so we included the geographic origin of isolates in our models to additionally test whether disease pressure in the environment of origin influences rust disease modification. We sampled five trees per site and selected four leaves of roughly the same age (leaf plastochron index 5; Larson &

Isebrands, 1971) per tree, from five different lower canopy branches. Trees were a minimum of 5 m apart. Each leaf was sealed in a separate sterile ziplock bag and stored in a cooler on ice for 48 hrs until further processing. Under a laminar flow hood, each leaf was transferred into a sterile moist-incubation chamber, i.e., a UV sterilized ziplock bag with an autoclave-sterilized paper towel moistened with sterilized deionized water (Busby et al., 2016b). After four weeks of incubation at room temperature, we examined the leaves under a dissecting microscope, and used a sterilized needle to isolate Cladosporium conidiophores onto potato dextrose agar (PDA) amended with the antibiotics penicillin (100 units/mL) and streptomycin (100 µg/mL). Our methods capture both epiphytes and endophytes, as we expect both can interact with the rust pathogen in ways that modify disease severity. We isolated a total of 222 cultures, then selected

115 for single-spore isolates. A long-term collection of single-spore isolates were made by placing small plugs of each culture in sterile deionized water in 2 mL screw-cap tubes.

DNA extraction and sequencing:

We selected a subset of 94 single-spore isolates for DNA sequencing so that a roughly equal proportion of isolates were included from each of the eight study sites. Additionally, we included two isolates from a previous study which sampled the same sites (Busby et al., 2016b), for a total of 96 single-spore isolates. We extracted DNA from a small amount (~ 2 cm2) of fresh fungal tissue using the Mag-Bind Plant DNA DS 96 extraction kit (Omega Bio-Tek, Doraville,

GA) and the Thermo KingFisher® Flex automated extraction system (Thermo Fisher Scientific,

Waltham, MA). Multilocus sequence typing of the ITS-LSU region rDNA, and partial sequences of the protein coding genes beta-tubulin (tub), RNA polymerase II (rpb2), and translation elongation factor 1-alpha (tef1) was performed using the MLSTez pipeline (Chen et al., 2015).

Briefly, DNA samples were normalized to approximately 5 ng/µL and normalized samples were used as templates for PCR with gene target primers including adapters (Supporting Information

S2). Amplification consisted of initial denaturation at 95ºC for 3 mins, followed by 35 cycles consisting of 95ºC for 1 min, 52ºC (tub, rpb2) or 55ºC (ITS-LSU, tef1) for 30s, and 72ºC for 1 min 30s; and a final extension at 72ºC for 10 mins. Amplicons were used in a second round of

PCR to barcode the samples using 2µL of template diluted 5-fold and 2µL of 5µM barcoding primer. Samples were barcoded by individual (e.g. one barcode was used for the four target loci in a single isolate). Barcoding amplification cycling conditions were 95ºC for 30s, 35 cycles at

94ºC for 30s, 50ºC for 1 min, and 72ºC for 1 min 30s, and a final extension at 72ºC for 10 min.

Target loci were verified on 1% agarose gel and the four target loci were pooled by individual

84 based on amplification intensity similar to Rothfels et al. (2016). Pooled samples by individual were quantified using Quant-iT dsDNA HS kit (Thermo Fisher Scientific, Waltham, MA) and finally 1µg of each individual were pooled in a final library. Primers, dNTPs and other small fragments were removed through Mag-Bind TotalPure NGS purification (Omega Bio-tek,

Norcross, GA). The SMRT cell library was prepared and sequenced on 1 SMRT cell using

Sequel System Chemistry V3.0 (PacBio, Menlo Park, CA) at Duke GCB Genome Sequencing

Shared Resource (Durham, NC). Finally, for all isolates we amplified and Sanger sequenced a

~220 bp portion of the actin (act) gene. This portion of the act gene has been shown to be highly phylogenetically informative in previous studies of Cladosporium, typically capturing the same species-level clades as captured in multi-gene phylogenies (Schubert et al., 2007, Bensch et al.,

2010). All primer information is in Supplementary Table 4.S2.

Phylogenetic analyses:

Preliminary phylogenetic analyses indicated that all isolates belonged to two of the three major Cladosporium species complexes; the C. cladosporioides (Bensch et al., 2010) and C. herbarum species complexes (Schubert et al., 2007). Thus, we selected Cladosporium sphaerospermum of the C. sphaerospermum species complex as the outgroup for all phylogenetic analyses. Cladosporium sphaerospermum gene sequences were obtained by performing blastn searches with a representative sequence of each of our genes against the C. sphaerospermum UM843 genome on Mycocosm

(https://genome.jgi.doe.gov/programs/fungi/index.jsf).

We aligned the sequences using MUSCLE (Edgar, 2004) and manually adjusted the alignments in Geneious Pro 11 (Biomatters Ltd., Auckland, New Zealand). The portion of the

LSU rDNA we sequenced was 100% conserved across all isolates, thus we trimmed all of it except for a small portion at the 3’ end of the ITS region for downstream analyses. Protein coding genes (act, tub, rpb2, tef1) were partitioned into 1st, 2nd, and 3rd codon positions, and introns. The ITS region was partitioned into ITS1, ITS2, and rRNA (SSU, 5.8S, LSU).

PartitionFinder2 (Lanfear et al., 2016) was used to calculate the optimal partitioning scheme and best evolutionary model for each partition in each gene and for the concatenated dataset containing all 5 genes (Supplementary Table 4.S3). RAxML 8 (Stamatakis, 2014) on the Cipress

Science Gateway (Miller et al., 2010) was used to compute maximum likelihood trees for each individual gene and for the concatenated alignment using the GTRGAMMA model. We selected the GTRGAMMA model because RAxML allows only one model to be specified for each analysis and this was found to be the optimal model for the majority of partitions

(Supplementary Table 4.S3).

Species recognition:

We delimited species using the genealogical concordance phylogenetic species recognition criterion (GCPSR) (Taylor et al., 2000). Under this species recognition criterion, terminal clades present in the combined analysis (Supplementary Fig. 4.S1) that were also present in the majority of single-gene trees (Supplementary Figs. 4.S2–4.S6) were hypothesized to represent species. If a terminal clade present in the combined analysis was not present in the majority of single-gene trees, the next encompassing clade was examined using the same methods as outlined above to determine whether it should be recognized as a phylogenetic species. A total of 5 isolates yielded sequences which occupied divergent, isolated branches in all analyses for which gene content was present. Because of their divergent position in all analyses,

86 each of these isolates was also recognized as a distinct species. We did not include information from the ITS tree in the GCPSR analysis as the ITS region in Cladosporium has low variability and provides little fine-scale phylogenetic information (Bensch et al., 2010). To infer the taxonomy of our isolates we performed blastn queries on GenBank (Clark et al., 2016) using the act sequences and downloaded act sequences representing the top matching species to each of our phylogenetic species (Supplementary Table 4.S4). We generated an additional concatenated tree including one representative act sequence per species as downloaded from GenBank

(Supplementary Table 4.S4) and used the resulting tree to infer the taxonomy of our isolates

(multigene tree: Supplementary Fig. 4.S1; single-gene trees: Supplementary Figs. 4.S2–4.S6).

Disease modification:

To test the effect of each Cladosporium isolate on Melampsora x columbiana rust disease severity, we performed double-inoculation leaf disk assays in which the endophyte was inoculated first, followed by the rust pathogen (Raghavendra & Newcombe, 2013). We grew cuttings of a single rust-susceptible P. trichocarpa genotype collected along the Clearwater

River, Idaho, USA, in a greenhouse on the campus of Oregon State University for approximately four months. Plants were watered at the base to avoid leaf-wetting, which could facilitate contamination by non-target endophytes present in the greenhouse. We harvested leaves of roughly the same age (leaf plastochron index 3–5) and immediately placed the leaves in sterile bags moistened with 1 mL sterile deionized water to prevent desiccation. We punched 2 cm disks from leaves using a sterile cork borer and randomized disks in a sterile ziplock bag by shaking for 1 min. Leaf disks (three per petri dish) were then placed abaxial side up on sterile filter paper

87 moistened with 2 mL of a sterile aqueous solution of 100 ppm gibberellic acid (to delay senescence) in petri dishes.

We prepared Cladosporium inoculum (spore slurry) for each of the 50 isolates by pipetting 1 mL sterile deionized water onto a 14 day old culture, then dislodging spores by scraping the surface of the culture with the pipet tip. Spore slurry concentrations were quantified with a hemocytometer and then diluted to a final concentration of 2 x 106 spores/mL. Each

Cladosporium isolate was inoculated onto 12 leaf disks (4 plates) by spraying ~2 mL spore slurry using a sterile spray bottle. Plates were sealed and incubated at room temperature. A total of 36 leaf disks (12 plates) were left uninoculated and were sprayed with ~2 mL sterile deionized water to serve as a negative control. After incubating for 48 hrs, all leaf disks were sprayed with

~2mL of two-week-old urediniospores of the rust pathogen Melampsora x columbiana (1 x 104 spores/mL; Newcombe, 1998). The M. x columbiana strain was a single-urediniospore isolate collected near Moscow, Idaho, USA

(https://mycocosm.jgi.doe.gov/MecolCla_1/MecolCla_1.home.html). After inoculating leaf disks with M. x columbiana, the plates were sealed with parafilm and randomized on a bench in a growth chamber with 12 h day/night cycle. After 10 days, each plate was photographed (Fig. 4.2) and disease severity scored by counting the number of uredinia on each leaf disk (without differentiating between immature and mature uredinia).

Linking phylogeny to disease modification:

We tested whether endophyte phylogeny was correlated with rust disease severity

(number of uredinia) using generalized linear models and a quasi-poisson error distribution to account for over dispersion. Because disease severity could not be measured on some leaf disks

88 due to leaf senescence, we modeled the number of uredinia per leaf disk by including the final number of leaf disks for each treatment as an offset in the model. To generate the phylogenetic predictors used for generalized linear modeling, we extracted phylogenetic eigenvectors from the phylogeny (Diniz-Filho et al., 1998). In short, these eigenvectors were obtained by first calculating genetic distances among isolates on the maximum likelihood phylogeny using the R package ape, function ‘cophenetic’ (Paradis & Schliep, 2018), followed by principal coordinate analysis (PCoA) of the pairwise genetic distance matrix using ape, function ‘pcoa’. Using the broken-stick criterion, we selected the first two PCoA axes to include in our models, which represent the main topological features of the phylogeny (84% of the phylogenetic variation among isolates) (Diniz-Filho et al., 1998). We controlled for the geographic origin of endophyte isolates (east vs. west of Cascades), by including geographic origin, and the interaction between geographic origin and phylogeny, as predictors in the model. Beginning with the full model, we performed reverse model selection using F-tests and the R base function ‘drop1’. The significance of terms in the final model was reported using Wald tests with the function

‘summary’. To compare the relative effect size of terms in the final model we calculated variance-function based partial pseudo-R2 for the predictors in the final model using function

‘rsq.partial’ in the R package rsq (Zhang, 2017).

Mechanisms of disease modification:

At 18 days post-rust inoculation we scored putative mycoparasitism on each leaf disk using a dissecting microscope. Disks were recorded as displaying mycoparisitism if we observed

Cladosporium conidiophores emerging directly from M. x columbiana uredinia within a time frame of 30 seconds per leaf disk (Dolińska et al., 2011; Fig. 4.2). It was not tractable to inspect

89 all uredinium on every leaf disk. For each isolate, we recorded the number of leaf disks displaying mycoparasitism. Cladosporium conidiophores emerging from rust uredinial tissue were re-isolated from a random sample of 10 leaf disks and the partial act gene was re-sequenced in an attempt to confirm that these putative mycoparasites were the same species as those originally inoculated with. In all cases, the putative mycoparasites matched the original inoculation species (data not shown). We tested whether phylogeny (while controlling for geographic origin) was correlated with the number of leaf disks displaying evidence of mycoparasitism per isolate using binomial generalized linear models following the same model selection procedure as in Linking phylogeny to disease modification, except here reverse model selection was performed using chi-squared tests.

We conducted a second leaf disk assay to further evaluate whether disease antagonism could be attributed to mycoparasitism. We included a subset of seven Cladosporium isolates in this experiment, 4 from the C. cladosporioides species complex, and 3 from the C. herbarum species complex (3 plates per isolate x 3 leaf disks per plate = 9 replicate leaf disks per isolate).

This assay was conducted in the same way as the previous assay, except here we included two additional treatments: 1) autoclaved Cladosporium spore slurry to test whether dead

Cladosporium cellular debris and/or secondary metabolites or enzymes could modify rust severity (although heat can destroy thermolabile compounds, Azmir et al., 2013), and 2) filtered

Cladosporium spore slurry using a 0.2 µm sterile filter and sterile syringe and applied the filtrate to leaf disks to test whether secondary metabolites or enzymes alone could modify rust severity.

We also included a live Cladosporium spore slurry treatment for each isolate as a positive control, and a water only negative control (neg. control = 25 replicate leaf disks). We scored mycoparasitism and disease severity as before. We tested for a difference in disease severity

(number of uredinia) among treatments separately for each species complex (C. cladosporioides and C. herbarum) using quasi-poisson generalized linear models. Overall significance of a treatment effect was reported using chi-squared tests with the function ‘anova’. This was followed by Tukey contrasts between all pairwise combinations of treatments (negative control, autoclaved spore slurry, filtered spore slurry, live spore slurry (positive control)) within each species complex, using function ‘glht’ in the multcomp package in R (Hothorn et al., 2016)

RESULTS:

Phylogeny:

We obtained sequences for all 96 Cladosporium isolates, although we did not obtain all loci for every isolate (Supplementary Table 4.S4). Using genealogical concordance phylogenetic species recognition, we recognized 17 distinct Cladosporium species: 12 in the C. cladosporioides species complex (representing 77 isolates) and 5 in the C. herbarum complex

(representing 19 isolates) (multigene tree: Supplementary Fig. 4.S1; single-gene trees:

Supplementary Figs. 4.S2–4.S6). Based on BLAST matches to the partial act locus, most of the species appear to be known; however, at least 4 may be undescribed. The rpb2, tub, and tef1 loci recovered all species-level clades as recovered in the concatenated phylogeny. The act locus resolved 15 out of 17 species, and the ITS region only clearly resolved 5 of the species-level clades. The number of single-nucleotide polymorphisms (SNPs), mean length, and SNPs/base- pair are shown for each locus (Supplementary Table 4.S5). Sequence data have been deposited in

GenBank (MN054571–MN055029) and accession numbers are displayed in Supplementary

Table 4.S3.

Linking phylogeny to disease modification:

Cladosporium phylogeny was significantly correlated with disease severity (PCoA1

R2=0.28, P<0.001; PCoA2 R2=0.21, P<0.001; Figs. 4.3, 4.4; Supplementary Table 4.S6). The phylogenetic PCoA axes we tested clearly separated the C. cladosporioides complex from the C. herbarum complex (PCoA1), and C. perangustum from the rest of the C. cladosporioides complex (PCoA2) (Supplementary Fig. 4.S7), with disease antagonism tending to be greater in the core C. cladosporioides complex (mean 67% reduction) than in C. perangustum (mean 55% reduction) or the C. herbarum complex (mean 59% reduction). After accounting for phylogeny, geographic origin was also significantly correlated with disease severity (R2=0.13, P=0.009)

(Figs. 4.3, 4.4; Supplementary Table 4.S6), but accounted for only an average reduction of 1.15 uredinia per leaf disk for the Eastern vs. the Western isolates. The interaction between phylogeny and geographic origin was not significant.

Mechanisms of disease modification:

We found that our measure of putative mycoparasitism was correlated with the axis of phylogenetic variation (PCoA1) (R2=0.78, P<0.001) (Fig. 4.5; Supplementary Table 4.S7) that separated the C. cladosporioides complex from the C. herbarum complex (i.e. putative rust mycoparasitism and disease reduction were both higher in the C. cladosporioides complex than the C. herbarum complex). Phylogenetic PCoA2 (which separated C. perangustum from the rest of the C. cladosporioides complex and from the C. herbarum complex) and geographic origin were not significantly correlated with putative mycoparasitism.

In our second experiment exploring mechanisms of disease modification, we found that rust disease severity differed among leaf disks inoculated with a live Cladosporium spore slurry,

92 autoclaved spore slurry, filtered spore slurry, and water negative control in both species complexes (C. cladosporioides P<0.001, C. herbarum P<0.001; Supplementary Table 4.S8).

Inoculation with live spore slurry (positive control) reduced rust disease severity relative to the

Cladosporium-free negative control in both species complexes (Fig. 4.6, Supplementary Table

4.S9), confirming the result of our first experiment. In particular, the live spore slurry reduced rust relative to the negative control by 46% in the C. cladosporioides complex, and 32% in the C. herbarum complex. The live spore slurry reduced rust more than the filtrate in the C. cladosporioides complex, but not in C. herbarum complex (Fig. 4.6, Supplementary Table 4.S9).

The filtered spore slurry reduced rust severity relative to the Cladosporium-free negative control in both species complexes (Fig. 4.6, Supplementary Table 4.S9). The filtered spore slurry reduced disease severity relative to the negative control by 27% in the C. cladosporioides complex and by 28% in the C. herbarum complex. The autoclaved Cladosporium spore slurry did not have an effect on rust disease severity relative to the negative control for both the C. cladosporioides and C. herbarum species complexes (Fig. 4.6, Supplementary Table 4.S9).

Visual evidence of putative mycoparasitism on rust uredinia in the live spore slurry treatment mirrored the results of the earlier assay and was greater in the C. cladosporioides complex than the C. herbarum complex (Supplementary Table 4.S10).

DISCUSSION:

We found support for our hypothesis that more closely related Cladosporium isolates modify rust disease severity more similarly than distantly related Cladosporium isolates, revealing that phylogenetic relatedness can help predict differences in rust disease antagonism within this endophyte genus. In addition, Cladosporium phylogeny was correlated with our

93 measure of putative mycoparasitism, with isolates in the C. cladosporioides species complex displaying both greater mycoparasitism and rust disease antagonism than isolates in the C. herbarum complex. Together, these results suggest that at least some of the genotypic variation underlying the degree of rust mycoparasitism by Cladosporium spp. is phylogentically conserved and could be responsible for the patterns of rust disease antagonism we observed in our experiment.

Previous studies have also found that fungal functional traits can be phylogenetically conserved, and related to symbiont function (e.g., for arbuscular mycorrhizal fungi,

Powell et al., 2009). While there has been little research on this topic for endophytes, two recent studies examining endophyte function at higher taxonomic levels than examined here (fungal orders), found evidence for root endophyte phylogeny impacting function (Kia et al., 2017), but not leaf endophyte phylogeny (Giauque et al., 2019). Reconciling these results, and more generally improving our understanding of the relationship between endophyte evolutionary history and function, will depend on better characterizing the phylogenetic distribution of traits underlying endophyte functions. Ongoing efforts to develop a comprehensive fungal traits database should facilitate these advances (Zane et al., 2019).

Phylogeny:

We resolved 17 phylogenetic species of Cladosporium leaf endophytes. This represents a greater diversity than we expected given previous sampling in this geographic region using ITS1 metabarcoding, which recovered just two common OTUs of Cladosporium (Barge et al., 2019).

However, the diversity recovered through the use of additional protein coding genes agrees with other studies which used a combination of protein coding genes and morphological data

(Schubert et al., 2007; Bensch et al., 2010). It appears that most of the species we recovered have been previously described, and have global distributions (Schubert et al., 2007; Bensch et al.,

2010; Bensch et al., 2012). Broad spatial distributions are further supported by the fact that

Cladosporium spp. produce copious amounts of airborne asexual conidia and are one of the most common genera isolated from air (Bensch et al., 2012). However, a visual inspection of the phylogeny reveals that some species were found more frequently either east or west of the

Cascade mountains. Additional work is needed to better understand geographic structure within

Cladosporium.

In general, the phylogenetic trees in this study were well-resolved, with deeper relationships and species-level clades receiving high bootstrap support. In agreement with other studies of Cladosporium systematics (Schubert et al., 2007; Bensch et al., 2010), the ~220bp portion of the actin gene (act) resolved most species present in the multigene tree (16 out of 17) and the ITS region performed poorly, resolving only 5 species. While portions of the protein coding genes rpb2, tub, and tef1 have been widely used and found to be good markers in other studies of Cladosporium (Schubert et al., 2007; Bensch et al., 2010), the portions of these genes that we sequenced had not been widely used at the time of this study. However, we found them to be good species markers (each resolved all species present in the multigene tree).

Linking phylogeny to disease modification:

We found that phylogeny was a significant predictor of rust disease antagonism, providing support for our hypothesis that more closely related species within Cladosporium modify rust disease severity more similarly than distantly related species. The two phylogenetic

PCoA axes we tested represented the major topological features of the phylogeny (84% of the

95 phylogenetic variation among isolates). PCoA1 identified the deeper-level split between the C. herbarum and C. cladosporioides species complexes and PCoA2 identified a subclade (C. perangustum) within the C. cladosporioides complex. Weaker rust antagonism was found in both the C. herbarum complex and in C. perangustum than in the core C. cladosporioides complex.

Previous studies have also demonstrated rust disease antagonism by members of the C. cladosporioides complex (Morgan-Jones & McKemy, 1990; Moricca et al., 2001; Assante et al.,

2004; Moricca et al., 2005; Dolińska et al., 2011; Zhan et al., 2014; Anderson et al., 2016; Torres et al., 2017). We are not aware of any studies that have reported rust disease antagonism by members of the C. herbarum complex.

We found little support that Cladosporium geographic origin determines the level of rust disease antagonism. This result suggests that within the genus Cladosporium, rust antagonism may be tied to deeply conserved genomic elements that do not experience varying selection pressure across environments. Whether endophytes are adapted to habitats with elevated pathogen pressure has not been formally tested. However, Rodriguez et al. (2008) found that endophytic Colletotrichum species from high disease environments conferred disease resistance to hosts, whereas endophytic Fusarium and Curvularia isolates from low disease environments did not, suggesting the potential for habitat-adapted symbiosis. Future experimental studies in the

Cladosporium -Melampsora system integrating information on disease pressure in the environment of origin for individual endophytes could address habitat adapted symbiosis.

Additionally, integrating rust strains from different environments could enable testing whether endophytes are adapted to high disease pressure generally, or to rust isolates specifically. It is also possible that the mechanisms responsible for rust antagonism are not specific to

Cladosporium-rust interactions. For example, traits underlying mycoparasitism or the secretion

96 of secondary compounds could be linked to interactions with many other fungi, thus rust disease pressure alone may not be a strong selective force.

Mechanisms of disease modification:

We found that our measure of mycoparasitism was correlated with the dominant phylogenetic PCoA axis (PCoA1) which separated the C. cladosporioides from the C. herbarum complex and also explained the greatest amount of variation in rust disease severity. The observation that both mycoparastism and pathogen antagonism were higher in the C. cladosporioides complex suggests that mycoparasitism and pathogen antagonism are evolutionarily conserved and correlated in Cladosporium, and provides evidence that mycoparasitism is at least one of the mechanisms underlying rust disease antagonism. Rust mycoparasitism has also been previously reported for members of the C. cladosporioides complex such as C. tennuisimum (Moricca et al., 2001; Assante et al., 2004; Moricca et al.,

2005), C. cladosporioides (Zhan et al., 2014; Torres et al., 2017), C. pseudocladosporioides

(Torres et al., 2017), and C. uredinicola (Morgan-Jones & McKemy, 1990; Dolińska et al., 2011;

Anderson et al., 2016). On the other hand, to our knowledge, no study has reported rust mycoparasitism or antagonism by members of the C. herbarum complex. Members of the C. herbarum complex; however, have been reported to occur on or in association with other fungi

(Heuchert et al., 2005; Schubert et al., 2007), and this coupled with our finding of lower levels of putative rust mycoparasitism and disease antagonism in this group, suggests that they can be mycoparasitic.

We found that in both the C. cladosporioides and C. herbarum complexes, spore slurry filtrates reduced rust severity relative to the endophyte-free control, suggesting the presence of

97 rust antagonizing secondary metabolites or enzymes. Secondary compounds isolated from C. tenuissimum, known as cladosporols, which are β-1,3- glucan (a component of fungal cell walls) biosynthesis inhibitors, have been shown to inhibit rust spore germination and reduce rust disease severity (Nasini et al., 2004; Moricca et al., 2001; Assante et al., 2004). Further, growth of C. cladosporioides, C. pseudocladosporioides, and C. tenuissimum on media in which β-1,3- glucan (provided as laminarin) was the sole carbon source suggests that at least some

Cladosporium species also possess fungal cell wall degrading enzymes (Assante et al., 2004;

Torres et al., 2017). Interestingly, we found no difference in mean rust severity between the spore slurry filtrate and the live spore slurry treatments for the weakly mycoparasitic C. herbarum complex, whereas we found a small difference for the C. cladosporioides complex.

This result, in addition to our finding of greater putative mycoparisitism in the C. cladosporioides complex, suggests that secondary compounds produced in culture were drivers of rust antagonism for both species complexes, but that higher levels of mycoparasitism in the C. cladosporioides complex led to additional antagonism. Whether additional cladosporol or other compounds are being produced during mycoparasitism is unknown.

Conclusions:

Overall, we found that phylogenetic relatedness can help predict differences in endophyte function within Cladosporium. Cladosporium phylogeny was also correlated with our measure of putative mycoparasitism, suggesting that mycoparisitism was one mechanism underlying rust antagonism in our study. Niche competition or exclusion, where the endophyte precedes the pathogen and prevents or reduces colonization could also play a role in disease modification

(Vanier et al., 2019). Additionally, the response of the host plant to endophyte inoculation could

98 also play role in disease antagonism (Hacquard et al., 2017). Future work addressing these mechanisms will be critical for deepening our understanding of plant-pathogen-endophyte interactions, and for promoting these interactions in crop disease management.

REFERENCES:

Anderson, F. E., López, S. P. S., Sánchez, R. M., Fuentealba, C. G. R., & Barton, J. (2016). Puccinia araujiae, a promising classical biocontrol agent for moth plant in New Zealand: Biology, host range and hyperparasitism by Cladosporium uredinicola. Biological control, 95, 23–30. Arnold, A. E., Mejía, L. C., Kyllo, D., Rojas, E. I., Maynard, Z., Robbins, N., & Herre, E. A. (2003). Fungal endophytes limit pathogen damage in a tropical tree. Proceedings of the National Academy of Sciences, 100, 15649–15654. Assante, G., Maffi, D., Saracchi, M., Farina, G., Moricca, S., & Ragazzi, A. (2004). Histological studies on the mycoparasitism of Cladosporium tenuissimum on urediniospores of Uromyces appendiculatus. Mycological Research, 108(2), 170–82. Azmir, J., Zaidul, I. S., Rahman, M. M., Sharif, K. M., Mohamed, A., Sahena, F., Jahurul, M. H., Ghafoor, K., Norulaini, N. A., & Omar, A. K. (2013). Techniques for extraction of bioactive compounds from plant materials: A review. Journal of Food Engineering, 117(4), 426–436. Bakker, M. G., Manter, D. K., Sheflin, A. M., Weir, T. L. & Vivanco, J.M. (2012). Harnessing the rhizosphere microbiome through plant breeding and agricultural management. Plant and Soil, 360, 1–13. Barge, E. G., Leopold, D. R., Peay, K. G., Newcombe, G., & Busby, P. E. (2019). Differentiating spatial from environmental effects on foliar fungal communities of Populus trichocarpa. Journal of Biogeography, DOI:10.1111/jbi.13641. Bensch, K., Groenewald, J. Z., Dijksterhuis, J., Starink-Willemse, M., Andersen, B., Summerell, B. A., Shin, H. D., Dugan, F. M., Schroers, H. J., Braun, U., & Crous, P. W. (2010). Species and ecological diversity within the Cladosporium cladosporioides complex (Davidiellaceae, Capnodiales). Studies in Mycology, 67, 1–94. Bensch, K., Braun, U., Groenewald, J. Z., & Crous, P. W. (2012). The genus Cladosporium. Studies in Mycology, 72, 1–401. Busby, P. E., Ridout, M., & Newcombe, G. (2016a). Fungal endophytes: modifiers of plant disease. Plant Molecular Biology, 90, 645–655. Busby, P. E., Peay, K. G., & Newcombe, G. (2016b). Common foliar fungi of Populus trichocarpa modify Melampsora rust disease severity. New Phytologist, 209(4), 1681– 1692. Busby, P. E., Soman, C., Wagner, M. R., Friesen, M. L., Kremer, J., Bennett, A., Morsy, M., Eisen, J. A., Leach, J. E., & Dangl, J. L. (2017). Research priorities for harnessing plant microbiomes in sustainable agriculture. PLoS biology 15(3), e2001793. Carbone, I., & Kohn, L. M. (1999). A method for designing primer sets for speciation studies in filamentous ascomycetes. Mycologia, 91, 553–556. Chen, Y., Frazzitta, A. E., Litvintseva, A. P., Fang, C., Mitchell, T. G., Springer, D. J., Ding, Y., Yuan, G., & Perfect, J. R. (2015). Next generation multilocus sequence typing (NGMLST) and the analytical software program MLSTEZ enable efficient, cost- effective, high-throughput, multilocus sequencing typing. Fungal Genetics and Biology, 75, 64–71. Clark, K., Karsch-Mizrachi, I., Lipman, D., Ostell, J., & Sayers, E. W. (2016). GenBank. Nucleic Acids Research, 44, 67–72.

100

Diniz-Filho, J. A. F., de Sant’Ana, C. E. R, & Bini, L. M. (1998). An eigenvector method for estimating phylogenetic inertia. Evolution, 52, 1247–1262. Dolińska, A., Bartkowska, T. M., & Schollenberge, M. (2011). Light and scanning microscope observations of Cladosporium uredinicola growth on rust fungi. Phytopathologia, 61, 37–44. Dunlap, J. M., & Stettler, R. F. (1996). Genetic variation and productivity of Populus trichocarpa and its hybrids. IX. Phenology and Melampsora rust incidence of native black cottonwood clones from four river valleys in Washington. Forest Ecology and Management, 87(1), 233–256. Edgar, R. C. (2004). MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Research, 32(5), 1792–1797. Gardes, M., Bruns, T. D. (1993). ITS primers with enhanced specificity for basidiomycetes- application to the identification of mycorrhizae and rusts. Molecular Ecology, 2(2), 113– 118. Giauque, H., Connor, E. W., & Hawkes, C. V. (2019). Endophyte traits relevant to stress tolerance, resource use and habitat of origin predict effects on host plants. New Phytologist, 221(4), 2239-2249. Groenewald, J. Z., Nakashima, C., Nishikawa, J., Shin, H. D., Park, J. H., Jama, A. N., Groenewald, M., Braun, U., Crous, P. W. (2013). Species concepts in Cercospora: spotting the weeds among the roses. Studies in Mycology, 75, 115–170. Hacquard, S., Spaepen, S., Garrido-Oter, R., & Schulze-Lefert, P. (2017). Interplay between innate immunity and the plant microbiota. Annual review of phytopathology, 55, 565– 589. Hanada, R. E., Pomella, A. W. V., Costa, H. S., Bezerra, J. L., Loguercio, L. L., & Pereira, J. O. (2010). Endophytic fungal diversity in Theobroma cacao (cacao) and T. grandiflorum (cupuaçu) trees and their potential for growth promotion and biocontrol of black-pod disease. Fungal Biology, 114, 901–910. Heuchert, B., Braun, U., & Schubert, K. (2005). Morphotaxonomic revision of fungicolous Cladosporium species (hyphomycetes). Schlechtendalia, 13, 1–78. Hoeksema, J. D., Bever, J. D., Chakraborty, S., Chaudhary, V. B., Gardes, M., Gehring, C. A., Hart, M. M., Housworth, E. A., Kaonongbua, W, Klironomos, J. N. & Lajeunesse, M. J. (2018). Evolutionary history of plant hosts and fungal symbionts predicts the strength of mycorrhizal mutualism. Communications biology, 1(1), 116. Hothorn, T., Bretz, F., Westfall, P., Heiberger, R. M., Schuetzenmeister, A., Scheibe, S., & Hothorn, M. T. (2016). Package ‘multcomp’. http://cran.stat.sfu.ca/web/packages/multcomp/mulcomp.pdf. Jansson, S., & Douglas, C. J. (2007). Populus: a model system for plant biology. Annual Reviews in Plant Biology, 58, 435–458. Kia, S. H., Glynou, K., Nau, T., Thines, M., Piepenbring, M., & Maciá-Vicente, J. G. (2017). Influence of phylogenetic conservatism and trait convergence on the interactions between fungal root endophytes and plants. The ISME journal, 11(3), 777. Lanfear, R., Frandsen, P. B., Wright, A. M., Senfeld, T., & Calcott, B. (2016). PartitionFinder 2: new methods for selecting partitioned models of evolution for molecular and morphological phylogenetic analyses. Molecular Biology and Evolution, 34(3), 772–773. Larson, P. R., & Isebrands, J. G. (1971). The plastochron index as applied to developmental studies of cottonwood. Canadian Journal of Forest Research, 1, 1–11.

101

Liu, Y. J., Whelen, S., Hall, B. D. (1999). Phylogenetic relationships among ascomycetes: evidence from an RNA polymerse II subunit. Molecular Biology and Evolution, 16(12), 1799–1808. Martiny, J. B., Jones, S. E., Lennon, J. T., & Martiny, A. C. (2015). Microbiomes in light of traits: a phylogenetic perspective. Science, 350(6261), aac9323. McGuire, K. L., Bent, E., Borneman, J., Majumder, A., Allison, S. D., & Treseder, K. K. (2010). Functional diversity in resource use by fungi. Ecology, 91(8), 2324–2332. Miller, M. A., Pfeiffer, W., & Schwartz, T. (2010). Creating the CIPRES Science Gateway for inference of large phylogenetic trees. Proceedings of the Gateway Computing Environments Workshop (GCE), New Orleans, LA, pp. 1–8. Morgan-Jones, G., & McKemy, J. M. (1990). Studies in the genus Cladosporium sensu lato. I. Concerning Cladosporium uredinicola, occurring on telial columns of Cronartium quercuum and other rusts. Mycotaxon, 39, 185–200. Moricca, S., Ragazzi, A., Mitchelson, K. R., & Assante, G. (2001). Antagonism of the two- needle pine stem rust fungi Cronartium flaccidum and Peridermium pini by Cladosporium tenuissimum in vitro and in planta. Phytopathology, 91(5), 457–468. Moricca, S., Ragazzi, A., & Assante, G. (2005). Biocontrol of Rust Fungi by Cladosporium tenuissimum. In: Pei, M. H. & McCracken, A. R., eds. Rust diseases of willow and poplar. Wallingford, UK: CABI, 213–229. Nasini, G., Arnone, A., Assante, G., Bava, A., Moricca, S., & Ragazzi, A. (2004). Secondary mould metabolites of Cladosporium tenuissimum, a hyperparasite of rust fungi. Phytochemistry, 65(14), 2107–2111. Newcombe, G. (1998). Association of Mmd1, a major gene for resistance to Melampsora medusae f. sp. deltoidae, with quantitative traits in poplar rust. Phytopathology, 88, 114– 121. O’Donnell, K., Cigelnik, E. (1997). Two divergent intragenomic rDNA ITS2 types within a monophyletic lineage of the fungus Fusarium are nonorthologous. Molecular Phylogenetics and Evolution, 7, 103–116. Ostry, M. E., Wilson, L. F., McNabb, H. S., & Moore, L. M. (1989). A guide to insect, disease, and animal pests of poplars. U.S. Department of Agriculture Agriculture Handbook 677. Pandey, R. R., Arora, D. K., & Dubey, R. C. (1993). Antagonistic interactions between fungal pathogens and phylloplane fungi of guava. Mycopathologia, 124, 31–39. Paradis, E., & Schliep, K. (2018). ape 5.0: an environment for modern phylogenetics and evolutionary analyses in R. Bioinformatics, bty633. Perello, A., Simon, M. R., Arambarri, A. M. (2002). Interactions between foliar pathogens and the saprophytic microflora of the wheat (Triticum aestivum L.) phylloplane. Journal of Phytopathology, 150, 232–243. Powell, J. R., Parrent, J. L., Hart, M. M., Klironomos, J. N., Rillig, M. C., & Maherali, H. (2009). Phylogenetic trait conservatism and the evolution of functional trade-offs in arbuscular mycorrhizal fungi. Proceedings of the Royal Society B: Biological Sciences, 276(1676), 4237–4245. Raghavendra, A. K., & Newcombe, G. (2013). The contribution of foliar endophytes to quantitative resistance to Melampsora rust. New Phytologist, 197(3), 909–918. Rehner, S. A., Buckley, E. (2005). A Beauveria phylogeny inferred from nuclear ITS and EF1-α sequences: evidence for cryptic diversification and links to Cordyceps teleomorphs. Mycologia, 97(1), 84–98.

102

Rodriguez, R. J., Henson, J., Van Volkenburgh, E., Hoy, M., Wright, L., Beckwith, F., Kim, Y. O., & Redman, R. S. (2008). Stress tolerance in plants via habitat-adapted symbiosis. The ISME Journal, 2(4), 404. Rodriguez, R. J., White Jr., J. F., Arnold, A. E., Redman, A. R. (2009). Fungal endophytes: diversity and functional roles. New Phytologist, 182(2), 314–30. Rothfels, C. J., Pryer, K .M., & Li, F. W. (2016). Next-generation polyploid phylogenetics: rapid resolution of hybrid polyploid complexes using PacBio single-molecule sequencing. New Phytologist, 213(1), 413–429. Schubert, K., Groenwald, J. Z., Braun, U., Dijksterhuis, J., Starink, M., Hill, C. F., Zalar, P., De Hoog, G. S., & Crous, P. W. (2007). Biodiversity in the Cladosporium herbarum complex (Davidiellaceae, Capnodiales), with standardisation of methods for Cladosporium taxonomy and diagnostics. Studies in Mycology, 58, 105 –156. Sharma, I. K., & Heather, W. A. (1988). Light and electron microscope studies of Cladosporium tenuissimum, mycoparasitic on poplar leaf rust, Melampsora larici-populina. Transactions of the British Mycological Society, 90, 125–131. Stamatakis, A. (2014). RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies. Bioinformatics, 30(9), 1312–1313. Torres, D. E., Rojas-Martínez, R. I., Zavaleta-Mejía, E., Guevara-Fefer, P., Márquez-Guzmán, G. J., & Pérez-Martínez, C. (2017). Cladosporium cladosporioides and Cladosporium pseudocladosporioides as potential new fungal antagonists of Puccinia horiana Henn., the causal agent of chrysanthemum white rust. PloS one, 12(1), e0170782. Taylor, J. W., Jacobson, D. J., Kroken, S., Kasuga, T., Geiser, D. M., Hibbett, D. S., & Fisher, M. C. (2000). Phylogenetic species recognition and species concepts in Fungi. Fungal Genetics and Biology, 31, 21–32. Van Der Heijden, M. G., Klironomos, J. N., Ursic, M., Moutoglis, P., Streitwolf-Engel, R., Boller, T., Wiemken, A., & Sanders, I. R. (1998). Mycorrhizal fungal diversity determines plant biodiversity, ecosystem variability and productivity. Nature, 396(6706), 69. Vannier, N., Agler, M., & Hacquard, S. (2019). Microbiota-mediated disease resistance in plants. PLoS pathogens, 15(6), e1007740. Vilgalys, R., Hester, M. (1990). Rapid genetic identification and mapping of enzymatically amplified ribosomal DNA from several Cryptococcus species. Journal of Bacteriology, 172(8), 4238–4246. Widin, K. D., & Schipper, A. L. (1981). Effect of Melampsora medusae leaf rust infection on yield of hybrid poplars in north-central United States. European Journal of Forest Pathology, 11, 438–448. Zanne, A. E., Abarenkov, K., Afkhami, M. E., Aguilar-Trigueros, C. A., Bates, S., Bhatnagar, J. M., Busby, P. E., Christian, N., Cornwell, W., Crowther, T. W., & Moreno, H. F. (2019). Fungal functional ecology: Bringing a trait-based approach to plant-associated fungi. EcoEvoRxiv. doi:10.32942/osf.io/a7f6g Zhan, G., Tian, Y., Wang, F., Chen, X., Guo, J., Jiao, M., Huang, L., & Kang, Z. (2014). A novel fungal hyperparasite of Puccinia striiformis f. sp. tritici, the causal agent of wheat stripe rust. PloS One, 9(11), e111484. Zhang, D. (2017). rsq: R-Squared and Related Measures. R package version 1.0. https://CRAN.R-project.org/package=rsq.

103

Figure 4.1: Map of study sites located across the core of the native range of Populus trichocarpa in the Pacific Northwest of North America. The study area spans a rainfall gradient from west (wet) to east (dry) of the Cascade Range. Sites west of the Cascades are colored green and sites east of the Cascades are colored brown. GPS and climate data for each site are provided in Supplementary Table 4.S1.

104

Figure 4.2: (a) Healthy Melampsora rust uredinium displaying no signs of mycoparasitism. Scale bar = 80 µm. (b) Heavily mycoparasitized rust uredinium displaying many emerging Cladosporium cladosporioides (LST2L4C18) conidiophores. Scale bar = 80 µm. (c) Control endophyte-free leaf disk inoculated only with Melampsora rust displaying abundant uredinia. Scale bar = 0.5 cm. (d) Leaf disk inoculated first with C. cladosporioides (TIET4L1C27) displaying reduced disease severity. Scale bar = 0.5 cm.

105

Figure 4.3: Multigene maximum likelihood phylogeny of the 50 Cladosporium isolates used in rust disease modification assays displaying each isolates geographic origin, and rust disease severity (average # of uredinia per leaf disk). Only bootstrap values >50% are shown. Phylogenetic species identifications are displayed along the right hand margin.

106

Figure 4.4: Regression of the first two phylogenetic PCoA axes of the 50 Cladosporium isolates against rust disease severity (average # of uredinia per leaf disk). Results of final quasi-poisson generalized linear model (glm) are displayed on the right-hand side of figure. Point color indicates Cladosporium isolate geographic origin. Gray area indicates 95% confidence interval around glm smoothed line.

107

Figure 4.5. Regression of phylogenetic PCoA1 of the 50 Cladosporium isolates against putative rust mycoparasitism. Results of final binomial generalized linear model (glm) are displayed in the upper right-hand corner of figure. Gray area indicates 95% confidence interval around glm smoothed line.

108

Figure 4.6: Results of experiment examining effects of autoclaved and filtered spore slurry on rust severity in the (a) C. cladosporioides, and (b) C. herbarum species complexes. Bars with different letters denote significant differences at P<0.05. Neg. control = no endophyte water control, Pos. control = live endophyte spore slurry. Error bars represent standard error of the mean.

109

Supplementary Figure 4.S1: Maximum likelihood tree of ITS, act, tub, rpb2, tef1 loci, containing all 96 Cladosporium isolates and one representative act sequence per species as downloaded from GenBank. Thickened branches lead up to clades receiving >70% bootstrap support. Only bootstrap values >50% are shown. Phylogenetic species identifications are displayed along the right hand margin. Clade concordance used in genealogical concordance phylogenetic species recognition is displayed.

110

Supplementary Figure 4.S2: Maximum likelihood tree of the tub locus. Only bootstrap values >50% are shown. Phylogenetic species as recognized by genealogical concordance phylogenetic species recognition are delineated by the gray bars to the right of the clades with the numbers (1- 17) matching species delineations as displayed on the multigene tree (Supplementary Fig. 4.S1).

111

Supplementary Figure 4.S3: Maximum likelihood tree of the act locus. Only bootstrap values >50% are shown. Phylogenetic species as recognized by genealogical concordance phylogenetic species recognition are delineated by the gray bars to the right of the clades with the numbers (1- 17) matching species delineations as displayed on the multigene tree (Supplementary Fig. 4.S1).

112

Supplementary Figure 4.S4: Maximum likelihood tree of the ITS region. Only bootstrap values >50% are shown. Phylogenetic species as recognized by genealogical concordance phylogenetic species recognition are delineated by the gray bars to the right of the clades with the numbers (1- 17) matching species delineations as displayed on the multigene tree (Supplementary Fig. 4.S1).

113

Supplementary Figure 4.S5: Maximum likelihood tree of the tef1 locus. Only bootstrap values >50% are shown. Phylogenetic species as recognized by genealogical concordance phylogenetic species recognition are delineated by the gray bars to the right of the clades with the numbers (1- 17) matching species delineations as displayed on the multigene tree (Supplementary Fig. 4.S1).

114

Supplementary Figure 4.S6: Maximum likelihood tree of the rpb2 locus. Only bootstrap values >50% are shown. Phylogenetic species as recognized by genealogical concordance phylogenetic species recognition are delineated by the gray bars to the right of the clades with the numbers (1- 17) matching species delineations as displayed on the multigene tree (Supplementary Fig. 4.S1).

115

Supplementary Figure 4.S7. Principal coordinate analysis (PCoA) plot of phylogenetic distances among the 50 Cladosporium isolates used in disease modification assays. Axis 1 clearly separates the C. cladosporioides complex from the C. herbarum complex; axis 2 separates C. perangustum (bottom-center) from the rest of the C. cladosporioides complex (top- left). The number of points does not represent the total number of isolates included due to overlapping points (i.e. genetically identical isolates).

116

Supplementary Table 4.S1: Sampling site information.

a Site GPS location Region MAP (mm) MAT (°C)b DO N 47.68600, W 122.89845 West 1128 10.8 CAR N 47.09835, W 122.15285 West 1094 10.4 SNO N 47.530168, W 121.807082 West 1620 9.8 SK N 48.51154, W 121.90083 West 1717 9.6 TIE N 46.672984, W 121.039878 East 586 6.2 YK N 46.91901, W 120.509507 East 267 8.8 CW N 46.436367, W 116.903442 East 400 11.4 LS N 45.234189, W 116.325298 East 523 8.3 aMAP: Mean annual precipitation bMAT: Mean annual temperature

117

Supplementary Table 4.S2: Primers used in multilocus sequence typing (ITS-LSU, tub, rpb2, tef1) and Sanger sequencing (act). Universal adapter sequences for multilocus sequence typing (forward: CTGGAGCACGAGGACACTGA; reverse: GCTGTCAACGATACGCTACG).

Locus Primers Sense Primer sequence Reference ITS-LSU ITS1F F CTTGGTCATTTAGAGGAAGTAA Gardes & Bruns (1993) LR3 R GGTCCGTGTTTCAAGAC Vilgalys & Hester (1990) tub T1 F AACATGCGTGAGATTGTAAGT O’Donnell & Cigelnik (1997) TUB3Rd R TCVGWGTTSAGYTGACCNGGG Groenewald et al. (2013) rpb2 fRPB2-7cF F ATGGGYAARCAAGCYATGGG Liu et al. (1999) RPB2- R CAATCWCGYTCCATYTCWCC Liu et al. 11bR (1999) tef1 EF1-983F F GCYCCYGGHCAYCGTGAYTTYAT Rehner & Buckley (2005) EF1-2218R R ATGACACCRACRGCRACRGTYTG Rehner & Buckley (2005) act ACT-512F F ATGTGCAAGGCCGGTTTCGC Carbone & Kohn (1999) ACT-783R R TACGAGTCCTTCTGGCCCAT Carbone & Kohn (1999)

118

Supplementary Table 4.S3: Optimal partitioning scheme and best evolutionary model for each partition in the concatenated dataset containing all 5 genes.

Subset Best Model # sites Partition names 1 GTR+G 348 act 3rd codon, ITS1, ITS2 2 GTR+I+G 369 tef1 1st codon, act 2nd codon 3 GTR+I+G 837 act 1st codon, rRNA, tef1 2nd codon, tub 2nd codon 4 GTR+I+G 474 tub introns, act introns 5 GTR 238 tub 1st codon 6 GTR+G 238 tub 3rd codon 7 GTR+G 330 tef1 3rd codon 8 GTR+I+G 353 rpb2 1st codon 9 GTR+G 352 rpb2 2nd codon 10 GTR+G 352 rpb2 3rd codon

119

Supplementary Table 4.S4. GenBank accession numbers and top BLAST match to the act gene for each isolate. Sample site names are first two or three letters of culture ID; sites west of Cascades: DO, CAR, SNO, SK; sites east of the Cascades: TIE, YK, LS, CW.

GenBank Accession Numbers Top BLAST match to act

Culture ID ITS rpb2 tef1 tub act Taxon Accession % Sim.

TIET3L5C26 MN054749 MN054574 MN054844 MN054952 MN054664 C. allicinum MF473762 99

DOT2L5C14 MN054777 MN054613 MN054887 MN055025 MN054700 C. cf. cladosporioides 4 HM148550 100

SKT5L1C10 MN054778 MN054614 MN054888 MN055026 MN054701 C. cf. cladosporioides 4 HM148550 100

CWT1L1C2.1 MN054822 MN054642 MN054916 MN054990 MN054725 C. cladosporioides HM148533 100

CWT1L3C6 MN054785 MN054641 MN054915 MN054972 MN054726 C. cladosporioides HM148533 100

CWT3L3C12.1 MN054821 MN054640 MN054914 MN054975 MN054727 C. cladosporioides HM148533 100

CWT3L5C17 MN054820 MN054639 MN054913 MN054965 MN054728 C. cladosporioides HM148490 100

CWT3L5C18 MN054784 MN054637 MN054911 MN054982 MN054739 C. cladosporioides HM148490 100

CWT5L3C25.1 MN054814 MN054632 MN054906 MN054988 MN054736 C. cladosporioides HM148490 100

CWT5L3C26 MN054783 MN054631 MN054905 MN054979 MN054740 C. cladosporioides HM148490 100

DOT5L3C31 MN054807 MN054622 MN054896 MN054977 MN054720 C. cladosporioides HM148533 99

LST1L1C2 MN054781 MN054623 MN054897 MN054966 MN054723 C. cladosporioides MF473792 100

LST1L3C6 MN054825 MN054646 MN054920 MN054991 MN054743 C. cladosporioides HM148490 100

LST1L4C9 MN054811 MN054627 MN054901 MN054986 MN054734 C. cladosporioides HM148490 99

LST2L4C18 MN054806 MN054621 MN054895 MN054974 MN054719 C. cladosporioides HM148533 98

LST4L3C41 MN054819 MN054638 MN054912 MN054983 MN054729 C. cladosporioides HM148490 100

LST4L4C43 MN054786 MN054644 MN054918 MN054971 MN054742 C. cladosporioides HM148533 100

LST4L5C45 MN054815 MN054633 MN054907 MN054980 MN054737 C. cladosporioides HM148490 99

LST5L3C49 MN054812 MN054629 MN054903 MN054973 MN054735 C. cladosporioides HM148490 99

SKT5L2C12 MN054824 MN054645 MN054919 MN054976 MN054731 C. cladosporioides HM148490 100

TIET2L5C12 MN054782 MN054628 MN054902 MN054987 MN054730 C. cladosporioides HM148490 99

TIET4L1C27 MN054809 MN054625 MN054899 MN054967 MN054721 C. cladosporioides HM148533 100

TIET5L1C32 MN054823 MN054643 MN054917 MN054984 MN054724 C. cladosporioides HM148490 100

TIET5L3C37 MN054817 MN054635 MN054909 MN054970 MN054741 C. cladosporioides HM148522 99

YKT2L3C10 MN054826 MN054647 MN054921 MN054985 MN054732 C. cladosporioides HM148490 100

YKT2L4C6 MN054808 MN054624 MN054898 MN054978 MN054722 C. cladosporioides MF473792 100

YKT3L2C14 MN054810 MN054626 MN054900 MN054968 MN054733 C. cladosporioides HM148490 99

YKT4L2C24 MN054818 MN054636 MN054910 MN054989 MN054738 C. cladosporioides HM148490 100

CART1L4C4 MN054794 MN054605 MN054879 MN055010 MN054695 C. delicatulum HM148569 100

CART3L4C18 MN054798 MN054609 MN054883 MN055012 MN054699 C. delicatulum HM148568 100

DOT1L1C1 NA MN054604 MN054878 MN055009 MN054693 C. delicatulum HM148568 99

DOT3L2C18 MN054803 MN054616 MN054890 MN055018 MN054704 C. delicatulum KT600584 98

SKT3L3C5 MN054780 MN054619 MN054893 MN055017 MN054706 C. delicatulum MF473804 98

120

Supplementary Table 4.S4: Continued

GenBank Accession Numbers Top BLAST match to act

Culture ID ITS rpb2 tef1 tub act Taxon Accession % Sim.

SKT3L5C6 MN054795 MN054606 MN054880 MN055011 MN054696 C. delicatulum MF473803 100

SKT3L5C7 MN054804 MN054618 MN054892 MN055016 MN054702 C. delicatulum KT600584 98

SKT4L2C8 MN054796 MN054607 MN054881 MN055007 MN054697 C. delicatulum HM148568 100

SKT4L5C9 MN054793 MN054603 MN054877 MN055006 MN054694 C. delicatulum HM148568 100

SKT5L5C15 MN054797 MN054608 MN054882 MN055008 MN054698 C. delicatulum HM148568 100

SNOT2L1C13 MN054802 MN054615 MN054889 MN055014 MN054703 C. delicatulum MF473804 98

SNOT4L2C31 MN054779 MN054617 MN054891 MN055015 MN054705 C. delicatulum KT600584 98

DOT4L1C22 MN054789 MN054600 MN054873 MN055023 MN054689 C. inversicolor HM148596 100

SKT1L1C2 MN054792 NA MN054876 MN055022 MN054692 C. inversicolor HM148596 100

SKT5L1C11 MN054788 NA MN054872 MN055020 MN054691 C. inversicolor HM148596 100

SNOT1L2C5 MN054776 MN054599 MN054871 MN055019 MN054690 C. inversicolor MF473974 100

LST4L3C39 MN054748 MN054573 MN054843 MN054939 MN054662 C. limoniforme KT600592 100

TIET3L1C16 MN054746 MN054571 MN054841 MN054937 MN054661 C. limoniforme KT600591 100

SKKM520366 MN054801 MN054612 MN054886 MN055028 MN054688 C. lycoperdinum HM148603 99

SNOT3L3C27 MN054805 MN054620 MN054894 MN055029 MN054686 C. lycoperdinum HM148603 97

SNOT5L3C34 MN054800 MN054611 MN054885 MN055027 MN054687 C. lycoperdinum HM148603 99

CWT3L1C8 MN054750 MN054575 MN054846 MN054953 MN054665 C. macrocarpum EF679524 99

TIET5L2C35 MN054751 MN054576 MN054847 MN054954 MN054666 C. macrocarpum EF679524 100

TIET1L2C2.1 MN054770 MN054595 MN054866 MN054958 MN054678 C. perangustum HM148631 100

TIET4L2C30 MN054767 MN054591 MN054862 MN054957 MN054681 C. perangustum HM148631 100

YKT3L1C8 MN054771 MN054596 MN054867 NA MN054677 C. perangustum HM148631 99

YKT3L4C15 MN054774 MN054592 MN054863 MN054961 MN054683 C. perangustum HM148610 100

YKT3L5C19 MN054765 MN054589 MN054860 MN054956 MN054680 C. perangustum HM148610 100

YKT4L3C25 MN054768 MN054593 MN054864 NA MN054684 C. perangustum HM148631 100

YKT4L3C26 MN054769 MN054594 MN054865 MN054962 MN054682 C. perangustum HM148631 100

YKT5L3C30 MN054772 MN054597 MN054868 MN054959 MN054676 C. perangustum HM148631 99

YKT5L4C31 MN054766 MN054590 MN054861 MN054960 MN054679 C. perangustum HM148631 100

TIEKM520364 MN054775 NA MN054870 MN054963 MN054685 C. phyllophilum HM148644 96 C. CART3L3C16 MN054831 NA MN054927 MN055002 MN054716 pseudocladosporioides HM148645 100 C. CART3L5C19 MN054787 NA MN054925 MN055003 MN054717 pseudocladosporioides KF415078 100 C. CART4L2C20 MN054830 MN054651 MN054926 NA MN054718 pseudocladosporioides MF474070 100 C. SKT5L3C14 MN054829 MN054650 MN054924 MN055001 MN054715 pseudocladosporioides MF474070 100

DOT1L2C3 MN054756 MN054581 MN054852 MN054950 MN054669 C. ramotenellum MF474083 98

DOT1L4C6 MN054759 MN054584 MN054855 MN054949 MN054671 C. ramotenellum KT600622 100

DOT2L2C9 MN054764 MN054579 MN054850 MN054948 MN054668 C. ramotenellum MF474083 99

121

Supplementary Table 4.S4: Continued

GenBank Accession Numbers Top BLAST match to act

Culture ID ITS rpb2 tef1 tub act Taxon Accession % Sim.

DOT2L3C11 MN054762 MN054587 MN054858 MN054944 MN054673 C. ramotenellum MF474096 99

DOT3L1C15 MN054763 MN054588 MN054859 MN054945 MN054674 C. ramotenellum MF474096 100

DOT3L3C19 MN054760 MN054585 MN054856 MN054951 MN054675 C. ramotenellum MF474096 99

DOT5L2C30 MN054757 MN054582 MN054853 MN054942 MN054670 C. ramotenellum KT600622 99

LST4L2C35 MN054754 MN054578 MN054849 MN054941 MN054667 C. ramotenellum MF474099 97

YKT1L5C1 MN054761 MN054586 MN054857 MN054947 MN054672 C. ramotenellum MF474096 99

CWKM520365 MN054753 NA MN054845 MN054940 MN054663 C. tenellum EF679554 100

CART2L2C6 MN054832 MN054652 MN054928 MN054992 MN054707 C. uwebrauniana MF474157 99

CART2L3C7 MN054837 MN054657 MN054933 MN054996 MN054714 C. uwebrauniana MF474157 100

CART2L3C8 MN054838 MN054658 MN054934 MN054999 MN054709 C. uwebrauniana MF474157 100

CART2L5C10 MN054839 MN054659 MN054935 MN055000 MN054710 C. uwebrauniana MF474157 100

CART3L1C13 MN054834 MN054654 MN054930 MN054993 MN054712 C. uwebrauniana MF474157 100

CART3L3C15 MN054840 MN054660 MN054936 MN054997 MN054711 C. uwebrauniana MF474157 100

SKT2L3C3 MN054836 MN054656 MN054932 MN054995 MN054713 C. uwebrauniana MF474157 100

TIET5L3C38 MN054833 MN054653 MN054929 MN054998 MN054708 C. uwebrauniana MF474157 100

LST4L1C34 MN054828 MN054649 MN054923 MN055005 MN054745 C. xylophilum HM148720 99

LST5L3C48 MN054827 MN054648 MN054922 MN055004 MN054744 C. xylophilum HM148719 99

CWT1L1C4.1 MN054813 MN054630 MN054904 MN054969 NA NA NA NA

CWT5L2C19.1 MN054791 MN054602 MN054875 MN055024 NA NA NA NA

DOT1L3C4 MN054758 MN054583 MN054854 MN054946 NA NA NA NA

DOT4L5C28 MN054755 MN054580 MN054851 MN054943 NA NA NA NA

DOT5L5C33 MN054790 MN054601 MN054874 MN055021 NA NA NA NA

LST3L2C23 MN054816 MN054634 MN054908 MN054981 NA NA NA NA

SKT2L3C4 MN054835 MN054655 MN054931 MN054994 NA NA NA NA

SNOT1L1C1 MN054799 MN054610 MN054884 MN055013 NA NA NA NA

SNOT2L5C19 MN054773 MN054598 MN054869 MN054964 NA NA NA NA

TIET3L4C23 MN054752 MN054577 MN054848 MN054955 NA NA NA NA

YKT2L5C7 MN054747 MN054572 MN054842 MN054938 NA NA NA NA

122

Supplementary Table 4.S5: The number of single-nucleotide polymorphisms (SNPs), mean length, and SNPs/100 base-pairs (bp) for each locus.

Locus SNPs Avg. Length SNPs/100 bp act 198 218 90 tub 344 1001 34 rpb2 335 1084 31 tef1 133 1015 13 ITS 48 533 9

123

Supplementary Table 4.S6: Final model of test for whether Cladosporium phylogeny was correlated with rust disease severity (number of uredinia) using quasi-poisson generalized linear models. The significance of terms in the final model is reported using Wald tests with the function ‘summary’.

Estimate Std. Error t value P Intercept 3.38 0.039 87.69 <0.001* Region 0.14 0.054 2.62 0.012* PCoA1 0.51 0.118 4.36 <0.001* PCoA2 -0.83 0.235 -3.53 <0.001* *significant at alpha = 0.05

124

Supplementary Table 4.S7: Final model of test for whether Cladosporium phylogeny was correlated with putative rust mycoparasitism using binomial generalized linear models. The significance of terms in the final model is reported using Wald tests with the function ‘summary’.

Estimate Std. Error z value P Intercept 2.00 0.16 12.14 <0.001* PCoA1 -8.21 0.62 -13.15 <0.001* *significant at alpha = 0.05

125

Supplementary Table 4.S8: Results of tests for a difference in disease severity (number of uredinia) among treatments (live spore slurry, filtered spore slurry, autoclaved spore slurry, negative control) for each species complex (C. cladosporioides and C. herbarum) using quasi- poisson generalized linear models. Overall significance of treatment effect is reported using chi- squared tests with the function ‘anova’

C. cladosporioides complex Df Deviance Residual Df Residual P Deviance NULL 123 1220.23 Treatment 3 318 120 901.24 <0.001* C. herbarum complex Df Deviance Residual Df Residual P Deviance NULL 102 893.16 Treatment 3 168.86 99 724.31 <0.001*

126

Supplementary Table 4.S9: Results of Tukey contrasts for difference in rust severity (number of uredinia) among leaf disks treated with autoclaved spore slurry (autoclaved), spore slurry filtrate (filtrate), live spore slurry (pos. control), and an endophyte-free control. Tests were conducted separately for the C. cladosporioides and C. herbarum complexes.

C. cladosporioides complex Treatment Difference Std. Error z value P Pos. control-Filtrate -13.3 0.099 -2.87 0.021* Pos. control-Autoclaved -24.2 0.096 -4.90 <0.001* Pos. control-Neg. control -33.5 0.099 -6.09 <0.001* Filtrate-Autoclaved -10.9 0.085 -2.15 0.136 Neg. control-Filtrate 20.1 0.089 3.56 0.002* Neg. control-Autoclaved 9.3 0.085 1.57 0.397 C. herbarum complex Treatment Difference Std. Error z value P Pos. control-Filtrate -5.7 0.101 -0.71 0.893 Pos. control-Autoclaved -18.4 0.097 -3.23 0.007* Pos. control-Neg. control -24.3 0.096 -4.15 <0.001* Filtrate-Autoclaved -14.7 0.094 -2.57 0.05 Neg. control-Filtrate 20.6 0.093 3.51 0.002* Neg. control-Autoclaved 5.9 0.083 0.93 0.787

127

Supplementary Table 4.S10: Physical evidence of mycoparasitism on rust uredinia in the living Cladosporium spore slurry treatment in the second leaf disk assay.

C. cladosporioides complex Species Isolate Proportion of leaf disks displaying evidence of mycoparasitism C. cladosporioides TIET5L1C32 1 C. lycoperdinum SNOT5L3C35 1 C. perangustum YKT3L5C19 1 C. uwebraunianum SKT2L3C4 1 C. herbarum complex Species Isolate Proportion of leaf disks displaying evidence of mycoparasitism C. allicinum TIET3L5C26 0.11 C. macrocarpum TIET5L2C35 0.63 C. ramotenellum DOT3L1C15 0.33

128

Chapter 5. Conclusion

129

This dissertation broadens our understanding of the diversity, distribution, and function of foliar fungi. Through this work, we also identified many areas in need of more research. In

Chapter two we show that environmental forces play a stronger role in structuring foliar fungal communities than spatial distance across the native range of Populus trichocarpa. To expand upon this finding, future work could take an experimental approach and utilize spore traps and better controls for the confounded nature of spatial and environmental variation. We also found that communities located in the drier, more continental environment East of the Cascades changed more from year-to-year than communities located in the more stable, maritime environment West of the Cascades. This finding begs for future, long term (5 plus years) sampling to better understand this pattern, and whether long-term foliar fungal community stability is correlated with plant health.

In Chapter three, we show that the phenology of spring leaf flush and disease resistance to an early infecting pathogen have stronger impacts on the leaf microbiome of Populus trichocarpa earlier in the growing season than later in the growing season. These results suggest that for the leaves of deciduous plants, intraspecific host genetic variability may have its strongest impact on microbial community composition early in assembly. However, whether this hypothesis can be generalized to include other plants species and other plant traits is not clear. It remains necessary to investigate a much broader range of intraspecific host genetic variability and host functional traits across a broader range of environments.

In Chapter four, we show that Cladosporium endophyte phylogeny can be used to predict its effect on rust severity. While the explanatory power of phylogeny was modest, this opens the door for using this technique for other endophyte-containing lineages as a means to predict their function. This type of research can also be used as a starting point for conducting endophyte

130 population genomics studies and genome-wide-association studies to better understand the genetic basis for endophyte trait variation. Further, while conducting this study, the need was made more urgent to measure the plant response to endophyte inoculation, and to move this research into the field setting. Much can be learned from conducting highly controlled experiments, but nothing can be used successfully until field tested.