Foliar fungal endophyte community structure is independent of phylogenetic relatedness in an Asteraceae common garden

Abstract Phylogenetic distance among host species represents a proxy for host traits that act as biotic filters to shape host‐associated microbiome community structure. However, teasing apart potential biotic assembly mechanisms, such as host specificity or local species interactions, from abiotic factors, such as environmental specificity or dispersal barriers, in hyperdiverse, horizontally transmitted microbiomes remains a challenge. In this study, we tested whether host phylogenetic relatedness among 18 native Asteraceae plant species and spatial distance between replicated plots in a common garden affects foliar fungal endophyte (FFE) community structure. We found that FFE community structure varied significantly among host species, as well as host tribes, but not among host subfamilies. However, FFE community dissimilarity between host individuals was not significantly correlated with phylogenetic distance between host species. There was a significant effect of spatial distance among host individuals on FFE community dissimilarity within the common garden. The significant differences in FFE community structure among host species, but lack of a significant host phylogenetic effect, suggest functional differences among host species not accounted for by host phylogenetic distance, such as metabolic traits or phenology, may drive FFE community dissimilarity. Overall, our results indicate that host species identity and the spatial distance between plants can determine the similarity of their microbiomes, even across a single experimental field, but that host phylogeny is not closely tied to FFE community divergence in native Asteraceae.


| INTRODUC TI ON
Microorganisms have been intimately associated with macro-organisms for hundreds of millions of years (Krings et al., 2007;Ley et al., 2008). Research across plant and animal systems, combined with the advent of advanced sequencing technologies (Alivisatos et al., 2015), has revealed a wide range of taxonomic and functional diversity in their microbiomes (Christian et al., 2015). Despite the persistence and longevity of these symbioses over evolutionary timescales, the fact that most microbiota colonize hosts via horizontal transmission from their local environment raises the question of whether environmentally acquired microbiota reflect the phylogenetic relatedness of their hosts (Christian et al., 2017) as in many pathogen (Gilbert & Webb, 2007) and mutualistic systems (Hammerstein & Noë, 2016), or whether they simply reflect spatial structuring and environmental specificity.
Host-associated microbiomes are assembled and maintained via multiple mechanisms that act across a wide range of spatial and temporal scales. At the local level, bacterial and fungal symbionts may compete for common resources or to gain entry and occupy a particular spatial niche within the host (Hooper et al., 2012). Host phenotypic traits (e.g., morphology, physiology, tissue chemistry) can also influence microbial colonization (Laforest-Lapointe et al., 2017). At larger spatial scales, dispersal limitation can constrain which microbial community members are able to colonize hosts and can lead to patterns of spatial distance decay in community similarity across the landscape (Oono et al., 2017). Microbial preferences for particular environmental or climatic conditions can also be an important driver of microbial community composition (Giauque & Hawkes, 2013).
Likewise, the frequency and timing of disturbances can affect the ability of microbial communities to rebound to their previous composition (Shen et al., 2018).
Because the richness, diversity, and structure of horizontally transmitted microbial communities are shaped by mechanisms acting across such a wide range of spatial and temporal scales, it has been challenging to disentangle the role of evolutionary history in these symbioses from contemporaneous ecological mechanisms (e.g., species interactions, environmental conditions, dispersal limitation).
Previous studies across a variety of plant-associated microbiomes have found that host species identity or phylogenetic relatedness often plays a significant role in structuring the host microbiome.
For example, host species identity can drive the structure of foliar bacterial communities among forest tree species (Laforest-Lapointe et al., 2017) and the composition of pathogenic virus communities among grasses (Seabloom et al., 2013). The probability that foliar fungal pathogens will spill over onto introduced, non-native hosts in grassland communities increases with increasing phylogenetic relatedness and population densities of native plant species in the resident community (Parker et al., 2015). By contrast, root-associated communities may show a reduced effect of host species identity or phylogenetic distance compared to aboveground communities (David et al., 2016;Glynou et al., 2018;Wagner et al., 2016).
Foliar fungal endophytes (FFE) are a major component of the plant microbiome and have recently emerged as a useful model for testing the outcome of plant-microbial interactions due to their ease of cultivation and relative tractability under laboratory conditions (Hawkes & Connor, 2017). FFE have existed within plant hosts since land colonization (Krings et al., 2007) and can act as mutualists, commensalists, pathogens, or saprotrophs (Busby et al., 2016). Most FFE colonize via environmentally transmitted propagules from wind, rain, or the previous season's plant litter (Christian et al., 2015) and have both high spatial (David et al., 2016) and temporal (Bowsher et al., 2020) turnover. Previous research has provided contrasting evidence of both FFE host specificity (i.e., FFE community differences among host species or genotypes) and host phylogenetic drivers (i.e., correlation between FFE community distance and host phylogenetic distance). For example, FFE showed host-specificity to Populus genotypes (Bálint et al., 2015) and herbaceous plant species (Gange et al., 2007), while spatial distance was a stronger determinant of FFE community structure than host identity in other reports (i.e., Picea sp. and tropical forest grasses; Eusemann et al., 2016;Higgins et al., 2014). Similarly, one recent study found that FFE community structure was significantly related to host phylogeny among Ficus tree species in a mixed botanical garden (Liu et al., 2019), but no evidence for host phylogenetic drivers was found among families of trees in a tropical forest (Vincent et al., 2016). Thus, experiments that control for host phylogeny, but minimize broad regional differences in spatial and temporal sampling of host individuals, would be a strong test for the relative importance of host phylogeny versus local ecological conditions on the distribution of FFE.
Here we investigated the relative roles of host phylogenetic relatedness and spatial distance between hosts on FFE community richness, diversity, and structure in a multispecies common garden. The use of a common garden framework allowed us to control for potentially confounding spatial and temporal factors, such as regional-scale changes in environmental and climatic conditions, long-distance dispersal limitation of FFE, and host age, by spatially randomizing all species under similar environmental conditions (Kawecki & Ebert, 2004). We predicted that FFE community structure would differ significantly among host species (Laforest-Lapointe et al., 2017) and that FFE community dissimilarity would be positively associated with phylogenetic divergence among host taxa (Liu et al., 2019). Similarly, we expected to see a positive, though weaker, effect of spatial distance with FFE community dissimilarity across the plots in the common garden due to the limited distance between plots. Lastly, we predicted that FFE community richness and diversity would differ among host species in the common garden reflecting inherent differences among hosts in their suitability as habitat for FFE taxa (Gange et al., 2007).

| Study system
We analyzed FFE communities of native perennials within the family Asteraceae using a host phylogenetic framework. With approximately 24,000 recognized species, the Asteraceae is the largest family of vascular plants and represents 10% of all flowering plants (Funk et al., 2009). Geographic spread and adaptive radiations out of South America have extended the range of the Asteraceae to every continent except Antarctica (Funk et al., 2009). Previous classifications segregated the Asteraceae into five major lineages, but more recent phylogenetic analyses support twelve subfamilies, with the most basal clades being either wholly endemic, or largely constrained, to South America (Panero & Funk, 2008). FFE have previously been cultured from species of Asteraceae (Christian et al., 2016), and a comparison of FFE communities between two host species from co-occurring genera of Asteraceae growing in southern England showed host-specific differences in FFE diversity and abundance (Gange et al., 2007).

| Species selection and plant propagation
Individual plants from 28 species of Asteraceae spanning three subfamilies, seven tribes, and 22 genera (Funk et al., 2009) were planted into each of six replicated plots within the common garden. Additionally, two outgroup species were included from the Campanulaceae, a sister family to the Asteraceae within the order  Table A1). Because of the wide variation in seed size, porosity, and other seed surface characteristics among species, it was not logistically tractable to optimize surface sterilization techniques for all 30 species. Additionally, initial tests indicated potential seed death due to bleach for highly porous seeds. Therefore, we did not surface sterilize seeds. Seeds were cold stratified as necessary according to vendor specifications using pasteurized sand and sterile water. The

| Common garden
The common garden was established in April and May 2014 at the  Purpose 20-20-20, JR Peters Inc., Allentown PA) to improve initial establishment in the field, after which no additional watering or fertilization was applied. As necessary, plots were hand-weeded and individual transplants were protected from small mammal herbivory with wire mesh enclosures (a cylinder 15 cm tall, with 1 cm 2 mesh size) until established.

| Leaf collection
After 3 months of growth and exposure to natural sources of endo- To minimize differences in sampling method among the 19 host species, which varied widely in height, growth architecture, and leaf size (see Figure A2), three leaves per plant (or leaf sub-sections from large-leaved plant species) were selected from mid-stem height. Leaf samples were then stored at 4°C until processing, which occurred within 24 hr of collection. To minimize differences in leaf sterilization efficiency across species, we standardized the size and shape of leaf fragments sterilized (nine haphazardly selected 0.5 × 0.5 cm square fragments). Leaf fragments were surface sterilized for 3 min in 70% ethanol, 2 min in 0.5% sodium hypochlorite, 1 min in sterile water, and then allowed to air dry for 1 min (Mejía et al., 2008). Surfacesterilized leaf fragments were then placed in sterile 2-ml microcentrifuge tubes, flash-frozen in liquid nitrogen, and stored at −80°C.
Two additional leaf disks per plant (1 cm diameter) were sampled from similarly aged, proximate leaves 3-4 weeks after collecting leaf samples for Illumina sequencing to determine leaf mass per area (LMA). Leaf disks were oven-dried at 60°C for 3 days and weighed.  Figure 1a.

| Molecular analyses
The ellipses represent the centroid and standard deviation for each of the seven host tribes and the Campanulaceae family (i.e., Lobelia cardinalis) outgroup quality Illumina reads (Binet et al., 2017). First, primers NSA3 and NLC2 (Martin & Rygiewicz, 2005) (Smith & Peay, 2014). The reverse amplification primer also contained a twelve-bp Golay barcode sequence, which was read using a third sequencing primer in an additional cycle (Caporaso et al., 2010(Caporaso et al., , 2012

| Phylogenetic inference
To generate a measure of phylogenetic relatedness in our analysis of FFE communities, a single locus phylogeny was reconstructed for all 19 plant species using plant nuclear ITS sequences retrieved from GenBank (see Table A2 for accession numbers & species names).
Species lacking available sequences in GenBank were replaced with available ITS sequences from congeneric species (n = 3), or contribal species (n = 3), based on previous phylogenetic inference in the Asteraceae (Fu et al., 2016;Panero & Funk, 2008;Schmidt & Schilling, 2000;Urbatsch et al., 2000). Plant sequences were aligned using Muscle version 3.8.31 (Edgar, 2004) and converted to PHYLIP  (Panero & Funk, 2008). The final best scoring ML tree was used to provide a measure of phylogenetic distance between all species pairs in our data set.
Before conducting all analyses, we subsampled (rarefied) to 15,514 reads per sample, which was the sequencing depth of the lowest sample after quality filtering (Bowsher et al., 2020). The best scoring ML tree was imported into R using the "ggtree" package (Figure 1a

| FFE community structure analyses
We used the Bray-Curtis dissimilarity index to test for differences in FFE community structure using a permutational multivariate analysis of variance (PERMANOVA) and a marginal sum of squares method to compute pseudo F-statistics for hypothesis testing ("adonis2" function, "vegan" package, 3,000 permutations; Oksanen et al., 2017

| FFE richness & diversity analyses
We used linear models to test how FFE OTU richness and diversity varied among individual plants, the unit of replication in this study.
For these models, we used LMA to standardize FFE richness and diversity per unit leaf mass, because while the same area of leaf tissue was sampled for Illumina sequencing, leaf mass varied among host individuals and species ( Figure A3). LMA was also a significant correlate of individual plant size (Adj. R 2 = 0.64 from model including plant size, host species and plots; Figure A4). Linear models tested whether richness and diversity varied among host species and plots in the common garden. FFE richness was square-root transformed to meet assumptions of normality.

| FFE community distance analyses
A multivariate statistical framework was used to test the association of FFE community structure with host species phylogeny and plot location within the common garden (N = 106 plants). Specifically, we performed partial Mantel tests to examine the correlation between the rarefied and normalized Bray-Curtis dissimilarity matrix with pairwise host species phylogenetic distance while controlling for pairwise spatial distance among plots, and vice versa. A null distribution drawn from 9,999 permutations of the Bray-Curtis matrix was used to test for statistical significance. Pairwise phylogenetic distance among all species pairs was estimated using the "cophenetic" function of tree branch length in R. The distance between plots was always substantially greater than the distance between plants within plots. Therefore, pairwise spatial distance between the centroid of all field plots in the common garden was calculated using Google Earth v.7.1.7.2602. Additionally, we tested whether core or rare taxa were drivers of distance-decay patterns. We created occupancy-abundance plots ( Figure A5; Shade et al., 2018) using prevalence and mean relative abundance across the dataset. Core taxa were selected on the criteria that they were (a) present in more than four host plants and (b) had a mean relative abundance greater than 0.025%. This partitioned the full FFE community matrix into 178 core and 380 rare fungal OTUs.

| Basic sequencing results
The quality-filtered sequence dataset, based on 106 plant sam-  (Figure 1c).

| FFE community structure, richness, and diversity
After 3 months of exposure to natural inoculum sources, there was a significant effect of host species identity on the structure of FFE communities (pseudo-F 18,72 = 1.26, p = 0.0033; Figure 2; see Figure 1a,b for phylogenetically based color reference). There was also a significant effect of host tribe on the structure of FFE communities (pseudo-F 7,83 = 1.30, p = 0.0143; Figure 2), but there was no significant effect of host subfamily (p = 0.5448). There was a non-significant trend toward differences in FFE community structure between different common garden plots (pseudo-F 5,72 = 1.23, p = 0.06390; Figure A6). There were no significant differences among host plants with differing LMA (p = 0.2422). A variance partitioning analysis demonstrated that host species identity explained a significant amount of variation in FFE community structure (4.85%; p = 0.002). Neither common garden plot identity (1.40%; p = 0.078) or LMA significantly explained any variation in FFE community structure (0.23%; p = 0.256).
The average OTU richness per host was 40 OTUs after rarefying and ranged from 6 to 94 OTUs per host. However, FFE OTU richness did not differ among host species (p = 0.1830; Figure A7a) or among plots in the common garden (p = 0.2711; Figure A8a). Similarly, though the average OTU diversity per host was 1.65, it also did not differ among host species (p = 0.1439; Figure A7b) or among common garden plots (p = 0.2759; Figure A8b).

| FFE community distance
Phylogenetic distance among hosts ranged from 0 for individuals of the same species to 1.52 branch length units for individuals of the two most divergent species (i.e., L. cardinalis and Cacalia plantaginea; Figure 1b). The pairwise spatial distance between plots in the common garden ranged from 0 to 648 m. Results of the partial Mantel test showed that dissimilarity of FFE communities among hosts was not significantly correlated with host phylogenetic distance (p = 0.1719; Figure 3a). By contrast, pairwise distance between plots within the common garden was significantly and positively correlated with the dissimilarity of FFE communities among hosts, although it explained relatively little of the variation (r = 0.0610; p = 0.0299; Figure 3b).
When the FFE community was divided into core and rare components, the core community displayed significant spatial distance decay within the common garden (r = 0.0683; p = 0.0183), while the rare community did not (p = 0.0937). Neither the core nor rare community was significantly correlated with host phylogenetic distance (core p = 0.1509; rare p = 0.5119).

| D ISCUSS I ON
Our study represents the first to investigate host phylogenetic relationships for FFE microbiota using a common garden approach, where individual plants from different species of the same age were spatially randomized across replicate plots within the same local habitat. We found that FFE communities differed significantly among host species and host tribes but that, contrary to our original prediction, pairwise phylogenetic distance between host species was not significantly correlated with FFE community dissimilarity between individual hosts. By contrast, FFE community dissimilarity did show modest increases with increasing spatial distance in the common garden over the limited spatial scale of 648 m. Overall, our results do not support the hypothesis that more closely related Asteraceae species within a common garden share more similar FFE microbiomes, but do suggest that other differences among host species, unaccounted for by host phylogenetic distance, lead to divergence in FFE communities among host species.
Our common garden experiment was established with the explicit purpose of maximizing local effects (e.g., small-scale environmental conditions and local inocula sources), while minimizing regional effects (e.g., geographic differences in surrounding vegetation, climatic conditions, soil types) on FFE community structure.
As a result, host species in the common garden shared many of the same FFE OTUs (see overlap in Figure 2), despite exhibiting significant species-specific differences. Given the large differences across plant species in leaf traits ( Figure A2), plant architecture, chemistry, phenology, and other phenotypic traits, it is perhaps surprising that we did not see more variation in FFE community structure among species. Future research should examine how host traits interact with environmental factors to influence microbial colonization and FFE community assembly. For example, significant variation in leaf secondary chemistry is a well-documented feature of the Asteraceae family that is not well resolved by phylogenetic relationships (Calabria et al., 2007) and that could influence colonization and persistence of specific microbial taxa (Christian et al., 2020). Micro-environmental differences among leaves sampled from different host individuals F I G U R E 3 FFE community dissimilarity A) did not significantly correlate with host phylogenetic distance but B) did significantly correlate with spatial distance between common garden field plots. The regression line (blue-dashed lines) characterizes the relationship between pairwise FFE community distance with pairwise distance between plots in the common garden. Each point represents a single pairwise comparison (e.g., leaf age, variable UV exposure; Osono & Mori, 2003) could be another source of unexplained variation in FFE community structure identified in our study. In addition, recent evidence indicates that microbial colonization from conspecific and heterospecific neighboring plants can also affect microbial community structure in mixed forests (Laforest-Lapointe et al., 2017). We did not quantify variation in the vegetation surrounding the common garden plots or within the common garden itself, but the neighboring vegetation may have influenced microbial community structure via short-distance dispersal or via priority effects of initial colonizers onto newly emerged leaves (Adame-Álvarez et al., 2014).
Contrary to our original prediction, we did not find a significant role for phylogenetic relatedness in determining the structure of FFE communities. Moreover, relatively little of the observed variation in FFE community structure was explained by any of the factors tested here. Host phylogenetic relatedness has often been used as a proxy for understanding complex ecological and evolutionary processes because it is simpler to measure than an array of functional traits and relatively inexpensive with the increasing availability of genetic sequence data (Cavender-Bares et al., 2009).
The genotypic and phenotypic divergence among host species underlying phylogenetic relationships is complex and undoubtedly includes a wide array of traits that differentially influence colonization success of specific microbial taxa (Liu et al., 2019). For example, research on plant-pathogen interactions suggests that more closely related hosts share similar genetic pathways for cellular recognition of proteins and effector molecules during pathogen colonization and resistance (Barrett & Heil, 2012). Our results stand in contrast to previous studies demonstrating significant signatures of host phylogenetic relationships for foliar fungal pathogens (Gilbert & Webb, 2007;Parker et al., 2015), root endophytic fungi (Wehner et al., 2014), arbuscular mycorrhizal fungi (Anacker et al., 2014), and FFE communities among Ficus species in a botanical garden (Liu et al., 2019). The Asteraceae represent a speciose plant family that has evolved more recently than the Ficus clade (Stevens, 2019). Thus, there may be less evolutionary divergence in traits related to asymptomatic FFE colonization for the Asteraceae relative to Ficus. On the other hand, our insignificant phylogenetic results were consistent with studies on belowground fungi and bacteria (David et al., 2016;Wagner et al., 2016) and FFE from host systems of differing phylogenetic breadth (Vincent et al., 2016).
Disentangling local, host-based processes, such as genotypic and phenotypic differences among individual plants, from spatial processes, such as dispersal, remain a challenge in studies of microbiome assembly and function. For example, one study at the landscape level showed that FFE community structure varied predictably across a 400-km precipitation gradient but did not significantly differ between two congeneric grass species (Giauque & Hawkes, 2013).
While in another study, where both a very broad phylogenetic and spatial breadth of host plants was compared (i.e., angiosperms to bryophytes, across continents), host identity and host genetic distance were the dominant drivers of FFE community structure and dissimilarity while spatial drivers were not detected (U'Ren et al., 2019). Here we detected a significant spatial effect on FFE community structure across 648 m, even while local climatic and soil factors were relatively constant (i.e., less than 1m elevational change across plots, homogeneous soil from a former agricultural field with decades of tilling). Micro-environmental differences across plots were not measured here (e.g., soil moisture, surrounding vegetation, wind speed), but could be incorporated into future studies to better tease apart small-scale drivers of FFE community assembly. Similarly, to separate local and regional influences on FFE communities from host genetic effects, experimental plantings of species or genotypes across multiple spatial and environmental scales, followed by microbiome characterization, could be implemented. For example, several genotypes within a group of host species that vary in specific leaf traits, or immune responses, could be transplanted across soil fertility or spatial distance gradients.
Our results likely represent a conservative assessment of the relative importance of host identity and spatial distance in driving FFE communities. Specifically, the restriction of hosts to a single plant family, combined with the relatively rapid speciation of the Asteraceae overall, could have led to an underestimation of host identity importance relative to spatial scale. Additionally, inclusion of negative and positive controls in our sequencing efforts (Nguyen et al., 2015) may have increased the confidence in community differences between host species by more accurately determining the presence or absence of OTUs between samples (Palmer et al., 2018).
It is also possible that our PCR conditions (e.g., nested PCR) led to a reduced pool of fungal species for comparison across hosts (Yu et al., 2015). However, it is important to note that studies comparing environmental or spatial factors with host genetic identity in FFE often find a relatively low contribution of host genetic identity (Bálint et al., 2015;Lamit et al., 2014;Whitaker et al., 2018).
In conclusion, our results provide insights and raise additional questions about the assembly and structure of host-associated microbiota across phylogenetically divergent host lineages (Clay & Schardl, 2002;Ley et al., 2008;Russell et al., 2017). Most previous research on the role of host phylogenetic relatedness in structuring microbial communities has relied on field sampling of natural populations and communities where host taxa occurred in different locations or microenvironments (Eusemann et al., 2016;Tedersoo et al., 2013;Vincent et al., 2016;Wehner et al., 2014) and thus did not control for regional and temporal drivers of microbial community assembly. Our study controlled for those factors, and we demonstrate that host species identity and spatial distance are significant, albeit modest, drivers of FFE community structure. We would like to thank M. Zaret, K. Hoban, L. Cole, Z. Shearin, Q.
Zhu, T. Cheeke, and the IU Greenhouse and RTP Staff for assistance with the project, as well as the IU Herbarium Director Dr. E. Knox for assistance identifying native Asteraceae and the IU Center for Genomics and Bioinformatics for bioinformatics support. Finally, we would like to thank the Clay and Bever laboratories for feedback on the manuscript.

CO N FLI C T O F I NTE R E S T
We have no competing interests to declare.

F I G U R E A 5
Occupancy-Abundance Plot. The proportion of host individuals each FFE occurred in plotted against the average relative abundance of each FFE across all sampled hosts (both log 10 -transformed). To the right of the vertical-dashed line (black) indicates FFE that had an average relative abundance of greater than 0.025%, while above the horizontal dashed line (purple) indicates FFE that occupied greater than 4 out of 106 host individuals F I G U R E A 6 FFE community structure did not significantly vary among plots in the common garden. Ellipses depict the centroid and standard deviation for common garden plots. The two grey-shaded ellipses represent plots 1 & 2, the two green-shaded ellipses represent plots 3 & 4, and the two purple-shaded ellipses represent plots 5 & 6. The location of each host individual in ordination space is denoted by a filled circle and is color-coded in kind with the ellipses F I G U R E A 7 FFE community (a) richness and (b) diversity did not vary significantly by host species. Boxplots are displayed for each species. For both panels, species have been sorted from least to highest FFE diversity. The host species codes appear as in Table 1. Richness and diversity have both been standardized by Leaf Mass per Area