A First Draft of the Core Fungal Microbiome of Schedonorus arundinaceus with and without Its Fungal Mutualist Epichloë coenophiala

Tall fescue (Schedonorus arundinaceus) is a cool-season grass which is commonly infected with the fungal endophyte Epichloë coenophiala. Although the relationship between tall fescue and E. coenophiala is well-studied, less is known about its broader fungal communities. We used next-generation sequencing of the ITS2 region to describe the complete foliar fungal microbiomes in a set of field-grown tall fescue plants over two years, and whether these fungal communities were affected by the presence of Epichloë. We used the Georgia 5 cultivar of tall fescue, grown in the field for six years prior to sampling. Plants were either uninfected with E. coenophiala, or they were infected with one of two E. coenophiala strains: The common toxic strain or the AR542 strain (sold commerically as MaxQ). We observed 3487 amplicon sequence variants (ASVs) across all plants and identified 43 ASVs which may make up a potential core microbiome. Fungal communities did not differ strongly between Epichloë treatments, but did show a great deal of variation between the two years. Plant fitness also changed over time but was not influenced by E. coenophiala infection.


Introduction
Plants are home to a wide variety of microorganisms which impact their survival and function. Research into the plant microbiome indicates that plants often contain hundreds to thousands of fungal and bacterial species, ranging from beneficial (mutualistic) to harmful (parasitic).
Tall fescue (Schedonorus arundinaceus (Schreb.) Dumort., formerly Festuca arundinacea Schreb., Lolium arundinaceum (Schreb.) Darbysh., Schedonorus phoenix (Scop.) Holub) is a cool-season grass native to Europe and northern Africa, and is widely cultivated throughout the temperate world for use as a forage and turf grass [1]. Tall fescue is commonly infected with the fungal endophyte Epichloë coenophiala (Morgan-Jones & W. Gams) C.W. Bacon & Schardl (formerly Neotyphodium coenophialum (Morgan-Jones & W. Gams) Glenn, C.W. Bacon & Hanlin [2]). This endophyte is asexual and strictly vertically transmitted, meaning that it can only be transmitted from mother to daughter via the seed. As a result of this vertical transmission, the relationship between tall fescue and E. coenophiala is mutualistic, although depending on the environmental conditions the benefits may not always outweigh the costs [3].
Benefits of E. coenophiala are attributed to its production of alkaloids (although other factors may also contribute [4][5][6]), and include resistance to insect and mammalian herbivory, drought, and some plant pathogens [7][8][9]. Resistance to mammalian herbivory in particular is caused by ergot alkaloids, which make grass containing the common toxic strain of E. coenophiala less desirable for use with livestock. This has led to the development of cultivars containing so-called "novel" endophytes (natural, but uncommon strains of E. coenophiala). Some of these novel endophytes do not produce the ergot alkaloids, but still confer some of the other benefits which improve plant growth and survival [10,11].
Fewer studies focus on the aboveground (or foliar) portion of the tall fescue fungal microbiome. Nissinen et al. [26] observed 54 fungal OTUs (operational taxonomic units) in tall fescue leaves and found that these communities were influenced by E. coenophiala infection. In the closely related grass and fungal species perennial ryegrass (Lolium perenne L.) and Epichloë festucae var.  [27] observed 247 OTUs but did not observe an effect of the Epichloë endophyte on fungal community composition. The communities were, however, strongly affected by study region and season. Liu et al. [28] observed 479 OTUs in drunken horse grass (Achnatherum inebrians (Hance) Keng) and found no effect of Epichloë gansuensis (C.J. Li & Na) Schardl (formerly Neotyphodium gansuense C.J. Li & Nan) on fungal communities, but did observe a difference between endophytic and epiphytic communities. From these few studies, it appears that cool-season grasses, like many other plants, contain diverse fungal microbiomes. It is not yet clear how important these communities are to their hosts, but research into other plant species has shown effects of foliar fungi on factors such as seed production, disease severity, and nitrogen uptake [29][30][31].
In this study, we sought to extend the work of Nissinen et al. [26] by examining tall fescue plants that had been established in the field for much longer (6 years vs. 3 months) and following individual plants for multiple years. We used next-generation sequencing of the fungal ITS2 region, and asked and answered the following questions.
What species were present in all or most plants over time (which we refer to as the draft "core microbiome")? 3.
Were these fungal communities influenced by the presence of the E. coenophiala endophyte? 4.
Were these fungal communities different between plants infected with different strains of the Epichloë endophyte? 5.
Were these fungal communities correlated with surrogates of plant fitness?

Field Samples
Fifty-one tall fescue plants (cv. Georgia 5; [32]) comprising three endophyte treatments (common-toxic strain, E+; novel strain, AR542, sold commercially as MaxQ; and an Epichloë-free control, E−) were grown from seed in a greenhouse. Georgia 5 seed lines were originally obtained from Donald Wood (University of Georgia, USA). Ten tillers from each plant were potted in the first week of May 2011 and grown in the greenhouse until 30 June, when they were transplanted to the field at the Guelph Turfgrass Institute (Guelph, ON, Canada; 43°32 56 N, 80°12 39 W). Details about the climate at this field site can be found in Figure S1. The plants were arranged in a 3 × 17 grid and were completely randomized (see Figure 1 for field layout). Since E. coenophiala is a strictly vertically transmitted endophyte, there was no risk of contamination of the Epichloë treatment between plants [2]. The plants were watered three times a week (M,W,F) until 8 August 2011, and the area around each plant was trimmed and mowed regularly throughout the growing season. After this initial set-up period, the plants were left to grow unmanaged with the exception of occasional mowing of the area immediately surrounding each plant. By May 2017, one E+ plant (E+125) and one AR542 plant (A160) had died, leaving 49 plants remaining. Tissue and seed samples were collected from each plant in July 2017 and 2018. An approximately 3-4 cm section of the pseudostem (see Figure 2) was collected from four tillers per plant and immediately flash frozen in liquid nitrogen, then freeze-dried and stored at −20°C until further processing. To assess plant fitness, each summer the number of tillers on each plant were counted and seed heads were collected weekly and counted in the lab.

DNA Extraction and Sequencing
Three out of the four tillers collected from each plant were pooled and ground in a Geno/Grinder (SPEX ® SamplePrep, USA). One tiller was kept as a backup in case of sample loss. DNA was extracted from 20 mg of this ground tissue using the DNeasy Plant Mini Kit (Qiagen Inc., Toronto, ON, Canada).
Of the 49 surviving plants, 48 were sequenced (omitted A64 due to low DNA concentration). After sequencing, several samples were left out of statistical analyses (E−250-S17 due to low read count, A34 and A49 due to loss of Epichloë over time, and E−189 due to potentially being a mislabelled E+ plant). Due to an issue with sequencing depth in the initial sequencing run, 10 samples were re-run. Data from the original run were omitted from statistical analyses for these 10 samples, but were retained when obtaining experiment-wide ASV totals and taxonomic information.

Epichloë Concentration
To estimate the concentration of E. coenophiala in samples, quantitative PCR (qPCR) was performed on the TefA (translation elongation factor 1-α) gene using primers specific to Epichloë (forward primer 5 -CAATGCAGCGAGTGAACATC-3 and reverse primer 5 -CACGTACTGACTGAAGCGTAGC-3 ) on a Roche LightCycler 480 (Roche Diagnostics, Rotkreuz, Switzerland). Each reaction contained 6 µL DNA (0.5 ng/µL), 7.5 µL SYBR Green PCR mix (Roche Diagnostics, Rotkreuz, Switzerland), and 0.75 µL of each primer (at 10 µmol concentration) and each sample had three technical replicates. Each qPCR plate also included a negative control (water). The PCR program was: 95°C for 5 min; 45 cycles of 95°C for 10 s, 64°C for 15 s, and 72°C for 15 s; followed by 95°C for 5 s, 65°C for 1 min, continuous acquisition at 97°C, and 40°C for 30 s to obtain the melt curve. For further details about the qPCR protocol, see Ryan et al. [35].

Bioinformatics
We used the DADA2 ITS Pipeline Workflow (v1.8, https://benjjneb.github.io/dada2 /ITS_workflow.html, accessed on 26 September 2022 [36]). Briefly, this workflow is an ITS-specific variation of version 1.8 of DADA2. After removing primers using cutadapt (v2.3) [37], reads were filtered, trimmed, and sorted into amplicon sequence variants (ASVs). ASVs are increasingly being used in microbiome research instead of operational taxonomic units (OTUs) due to their higher resolution and consistency across studies [38]. The end product is an ASV table providing the number of times each exact ASV was observed in each sample. It also assigns taxonomy to the output using the UNITE database v8.2 [39].

Statistical Analyses
PERMANOVA (using Bray-Curtis distances) was performed on rarefied ASV data (adonis function in vegan R package [40,41] (R v4.0.1, vegan v2.5-7) to identify whether community composition differed between endophyte treatments. The pairwiseAdonis package [42] in R was used for post hoc analysis. Because non-metric multidimensional scaling (NMDS) showed strong clustering of samples by year (Figure 3), and because our data contain repeated measures (which are not supported by the adonis function), PERMANOVA was performed separately for each year.
Numerous methods exist to normalize microbiome data. One of the most common, rarefaction, has been criticized in recent years due to the potential loss of statistical power that comes from discarding data [43,44]. This is primarily an issue when looking at α-diversity or identifying differentially abundant ASVs, where other methods and transformations may be more appropriate. When comparing overall community composition between samples, rarefied data can still be clustered accurately and may in fact be one of the best normalization methods [45,46]. There is strong separation by year, but a great deal of overlap between the Epichloë treatments.
Differentially abundant ASVs were identified using ANCOM-BC [47]. For the 43 core ASVs, we also performed repeated measures ANOVAs on rarefied read counts.
A repeated measures ANOVA was performed on seed count, tiller count, and endophyte concentration data to test for effects of endophyte treatment and year. Sample E−250 from 2018 was removed from this analysis to maintain a balanced design (due to the corresponding 2017 sample being removed earlier due to a low read count; see Section 2.2).
We follow Wasserstein et al. [48] in reporting exact P-values where practical and avoiding the use of the terms "significant" and "non-significant." Furthermore, we follow Greenland [49] by also reporting the Shannon information transformation, s = −log 2 (P).

Results
After filtering and trimming, there were 10,357,271 sequences. From these, we obtained 3487 amplicon sequence variants (ASVs) across all samples, including two ASVs corresponding to the two strains of E. coenophiala. After rarefaction, we had 2165 ASVs remaining.

Taxonomic Variability across Epichloë Treatments and Years
The 10 most common taxa at each level of organization (from phylum to species) are shown in Table 1. Dothideomycetes was far more common than any other class both in terms of ASV counts (≈41% of all ASVs) and total reads (≈65% of all reads; see Table 1). Pleosporales (Dothideomycetes) was the most common order overall (≈34% of all ASVs, ≈62% of all reads). Two other common orders were Helotiales (Leotiomycetes; ≈6% of all ASVs, ≈10% of all reads) and Tremellales (Tremellomycetes; ≈5% of all ASVs, ≈3% of all reads). The most common family was Phaeosphaeriaceae (Pleosporales; ≈20% of all ASVs, ≈29% of all reads). This was followed by Corticiaceae (Corticiales; ≈3% of all ASVs, ≈3% of all reads) and Bulleribasidiaceae (Tremellales; ≈3% of all ASVs) for ASV count and Didymosphaeriaceae (Pleosporales; ≈6% of all reads) and Didymellaceae (Pleosporales; ≈6% of all reads) for total reads. The most common genus was Septoriella (Phaeosphaeriaceae; ≈4% of all ASVs, ≈9% of all reads) for both ASV count and total reads, followed by Parastagonospora (Phaeosphaeriaceae) and Phaeosphaeria (Phaeosphaeriaceae; ≈2% of all ASVs) for ASV count and Paraphaeosphaeria (Didymosphaeriaceae; ≈6% of all reads) and Pyrenochaetopsis (Cucurbitariaceae; ≈6% of all reads) for total reads. Table 1. Most common taxonomic groups by amplicon sequence variants (ASVs) occurences or the number of reads, using either the raw data or the rarefied data. In each case the proportions of the totals are shown. For each taxonomic category and measure of abundance (columns), the most abundant taxon is highlighted in blue-green , the second most abundant is highlighted in rose and the third most abundant is highlighted in yellow . Note that species identities listed here are only tentative; see Section 4.5.

Phylum (P), Class (C), Order(O)
Family  Table 2 shows the pattern of ASV occurrences and co-occurrences (in the rarefied data) by Epichloë treatment and year. In Table 2, an 'occurrence' indicates that the ASV was present in at least one plant in the Epichloë × year combination. A 'co-occurrence' indicates that the ASV was present in at least one plant from each Epichloë pair or triplet. More than 70% of ASVs were present only in a single Epichloë treatment and single year ([264 + 272 + 232 + 315 + 264 + 191 = 1538]/2165). More than 13% of all ASVs were present in all Epichloë treatments and both years (302/2165). Approximately 80% of all ASVs occurred in only one year or the other but not in both years ([899 + 832]/2165). Together, these results suggest that a large segment of the plant's microbiome is 'transient.'

Fungal Community Structure
Non-metric multidimensional scaling (NMDS) indicated strong separation of fungal communities by year but a great deal of overlap between Epichloë treatments ( Figure 3). This was confirmed with PERMANOVA, which showed no difference in fungal communities between Epichloë treatments in 2017 and a small difference in 2018 (see Table 3). Post-hoc analysis could not identify which specific pair(s) of Epichloë treatments were different in 2018. The PERMANOVA also indicated differences in communities depending on the location of the plant within the field site (Row and Column). Table 3. PERMANOVA analyses by year. The row and column sources of variance refer to the physical layout of the experiment, see Figure 1. d f = degrees of freedom, SS = sum of squares, MS = mean squares, F = pseudo-F statistic, P = p-value; s denotes the Shannon Information Transformation (s = −log 2 (p)); see Section 2.5. The Epichloë treatment was important in 2018 but not 2017.

Draft Core Microbiome
We sought to characterize the core microbiome, which we define as those ASVs present in all or most of the plants in each Epichloë treatment and across both years. We refer to this as the 'draft core microbiome'. Table 4 shows the ASVs that meet these criteria. The strictest definition-ASV must be present in all plants and all years-picked out 13 ASVs that are strong candidates for membership in the core microbiome. Our least strict definition permitted an ASV to be absent from up to five plants in any treatment by year combination. Under this definition, only 43 ASVs met the inclusion criterion. This is <2% of the total ASVs we identified and yet these 43 ASVs together comprise 69% of the (post-rarefied) total reads.

Guilds
We attempted to assign guilds to the ASVs that comprised the draft core microbiome. We used the FungalTraits database [50]. For the ASVs that we were able to identify to the species level (tentatively; see Section 4.5), we were only able to match two to entries in this database: ASV32 and ASV95, both of which were classified as Epicoccum nigrum, a plant pathogen and endophyte. Ten core ASVs could be matched at the genus level, of which seven were classified as plant pathogens on at least some of their host plants, and six were classified as endophytes in at least some of their plant hosts. For those ASVs identified to the genus level that did not match anything in FungalTraits database, we searched the CABI Invasive Species Compendium [51]. Unfortunately, the guilds for many of these ASVs are not known. Across the dataset as a whole (not only the draft core), common guilds included plant pathogens, endophytes, animal pathogens, and wood saprotrophs.

Differential Reads
ANCOM-BC (analysis of compositions of microbiomes with bias correction, [47]) identified several ASVs which were differentially abundant between endophyte treatments and between years. These are shown in Table 5. Between years, 48 ASVs were differentially abundant, although there is no obvious common pattern in these results (33 ASVs were more common in 2017, 15 ASVs were more common in 2018). Between Epichloë treatments, only 5 ASVs were differentially abundant, which is not enough to draw any conclusions.
These numbers cannot reflect some of the interesting spatial patterns that emerge from the data. In the Supplemental information (Table S2)  Missing entries in the p-value columns indicate that the data could not be adequately transformed to meet the assumptions of the test. Note that these 43 ASVs account for 69% of all reads across both years. Consistent correlations denote ASVs that have consistent correlations (positive or negative) across all Epichloë treatments and years. For tiller and seed numbers, r < 0 suggest these ASVs are plant antagonists, and r > 0 suggest the ASVs are beneficial to the plant. For the Epichloë concentrations, r < 0 suggest a antagonistic relationship while values of r > 0 suggest a mutualistic relationship. ASVs denoted in brown are thought to be plant pathogens (see Guild column). For guild information from Fungal Traits [50], E denotes endophyte, PP denotes plant pathogen, AP denotes animal pathogen, and WS denotes wood saprophyte. For those ASVs identified to genus level, that did not match anything in FungalTraits database, we searched the CABI Invasive Species Compendium [51]. From this database, gpp denotes the genus contains plant pathogens (in this case Parastagonospora), and pp denotes plant pathogen.

Epichloë coenophiala Concentrations and Plant Fitness
The E. coenophiala concentrations are shown in Figure 4A. These concentrations were higher in 2018 (1109 copies ng −1 gDNA) than 2017 (457 copies ng −1 gDNA) but similar between the E+ and AR542 treatments.
Plants produced fewer seeds and tillers in 2018 (637 seeds and 137 tillers per plant) compared to 2017 (1935 seeds and 158 tillers per plant) ( Figure 4B,C). Seed and tiller numbers were similar between Epichloë treatments.

Relationship between Core ASVs, Plant Fitness, and Epichloë Concentrations
We calculated the Pearson's correlation coefficients between each of the ASVs in the draft core microbiome and the concentrations of the Epichloë, seed production, and tiller number for each year. The results are shown in Figure 5. These results are summarized in Table 4. For tiller numbers, five ASVs had consistently negative correlations and three had consistently positive correlations. For seed numbers, eight ASVs had consistently negative correlations while only three had consistently positive correlations. For Epichloë concentrations, four ASVs had consistently negative correlations, while six had consistently positive correlations.

Taxonomy
Many of the most common genera are known plant pathogens (Limonomyces, Parastagonospora), and some pathogens known to be associated with tall fescue and related grasses were also found in our dataset, e.g., Puccinia spp. (rust) [52,53] and Fusarium oxysporum [54]. Based on the FungalTraits database [50], other common fungal guilds in our dataset include animal pathogens, endophytes, wood saprotrophs, and lichen parasites. There were also many ASVs for which we could not obtain detailed taxonomic informa-tion; 47% of ASVs (making up 35% of overall reads) could not be identified at the genus level. Many of these may be fungi that are difficult to culture and therefore have not been classified taxonomically.
We identified 13 ASVs that were present in every plant across both years, and 43 ASVs which were absent from five or fewer plants per treatment-year combination (Table 4). We defined these 43 ASVs as the draft "core microbiome." Combined, these core ASVs make up nearly 70% of total reads. The vast majority of core ASVs were ascomycota (38/43), and more than half were dothidiomycetes (27/43). Of those with guild information from FungalTraits available, many were potential plant pathogens. This is consistent with our plant fitness data, which indicate a negative correlation between many core ASVs and seed and tiller counts ( Figure 5 and Table 4). Further work is required to see whether these particular ASVs are geographically specific and whether they persist over longer periods of time.

Effect of Epichloë and Comparisons with Previous Research
We observed diverse fungal communities in tall fescue. Of the ASVs we were able to obtain taxonomic information for, a relatively small proportion matched the taxa found in König et al. [27] and Liu et al. [28] (see Figure 6). This is perhaps not surprising given the different species (König et  used the 3rd and 4th leaf blades, and we used the pseudostem). Both of these studies also sequenced the ITS1 region rather than ITS2, although research has shown that results from both regions are comparable [55]. Finally, both of these studies also appear to exclude low-abundance OTUs, whereas we retained any ASVs with more than two reads, which likely explains why we observed the largest number of unique taxa. Nissinen et al. [26] looked at fungal communities in tall fescue, though as of writing the full data set is not available for comparison. Of their 11 most common taxa (excluding Epichloë; see Figure 1 in Nissinen et al.), two species were found in our data set (Puccinia coronata, Blumeria graminis; note however that we cannot be certain about species identification from molecular data alone, as discussed in Section 4.5). Of the remaining nine taxa, six were members of genera observed in our data (Pyrenophora, Podospora, Pyrenophora, Cryptococcus, Eutypa, Colletotrichum), and two were members of families observed in our data (Nectriaceae, Glomerellaceae). Figure 6. Comparison of the taxonomic diversity found in our study with that found in König et al. [27] and Liu et al. [28]. The Venn Diagrams show unique and shared numbers of phyla (A), classes (B), orders (C), families (D), genera (E) and species (F). It is clear that the overlap in diversity between the three studies is low. In (F), the 10 species that are common to all three studies are: Alternaria rosae, Cystofilobasidium macerans, Filobasidium magnum, Zymoseptoria verkleyi, Sporobolomyces roseus, Dioszegia hungarica, Buckleyzyma aurantiaca, Rhodotorula babjevae, Malassezia restricta, Blumeria graminis.
Presence of Epichloë did not consistently alter the fungal community composition, but there was some difference in 2018 when Epichloë concentration was highest. Epichloë have been found to have antifungal properties before. However, these results may be due to factors other than the alkaloids that confer many of its other effects, because Fernando et al. [56] found no antifungal activity for a variety of common Epichloë-produced alkaloids on several plant pathogens.
Research to date on the effect of Epichloë on fungal communities has been variable. König et al. [27] and Liu et al. [28] observed no effect of Epichloë festucae var. lolii and Epichloë gansuensis on the fungal microbiome of perennial ryegrass and drunken horse grass, respectively, whereas Nissinen et al. [26] found an effect of E. coenophiala on tall fescue fungal microbiomes. Using a different approach, Zabalgogeazcoa et al. [57]  Several ASVs in our study were differentially abundant between Epichloë treatments (see Table 5). Only one could be identified to the genus level (Limonomyces), which was absent from E+ plants. Previous research has shown some antifungal activity by E. coenophiala (and related Epichloë species) against Limonomyces roseipellis (pink patch), a fungal pathogen in grasses [58].

Variation over Time
There was a noticeable difference in fungal communities between the the two years measured, both in the overall communities ( Figure 3) and in specific differentially abundant ASVs (Table 5). This suggests that at least some of the fungal community of tall fescue is transient. This is consistent with previous research showing that phyllosphere (leaf surface) microbial communities are strongly affected by environmental changes like temperature and moisture [59,60]. According to historical climate data from a local weather station, the mean (±standard deviation) temperatures in the month leading up to sample collection were 17.87 ± 3.20°C in 2017 and 18.70 ± 3.96°C in 2018, and the relative humidity values were 74.13 ± 9.36 %RH in 2017 and 70.67 ± 9.88 %RH in 2018. For more detailed climate data see Figure S1 in the Supplemental Information.
Although communities fluctuated a great deal over time, we also observed some ASVs that were consistently associated with one or more specific plants across both years (see Figure S3 in Supplementary Information).

Plant Fitness
Epichloë infection was not associated with a change in either seed count or tiller count (see Figure 4). Previous research into the effect of Epichloë on plant fitness has produced mixed results; although many demonstrate a positive correlation between the two [61][62][63][64], other research has also shown no Epichloë effect [65,66] or even a negative association between Epichloë infection and fitness [67]. Fitness was, however, different between the two years measured. Both seed counts and tiller numbers were higher in 2017. These results may be due to environmental differences such as herbivore pressure, temperature or precipitation.

Limitations
The strong differences in fungal communities between the two years suggest that these communities fluctuate over time, so it would be interesting to observe them over more years to obtain a clearer longitudinal picture. Additionally, because we performed DNA extraction on the 2017 and 2018 samples at different times, it is possible that some of the changes observed between the two years could be due to this rather than true biological differences. However, the differences we observed in tiller and seed numbers between years suggests that there was something different between the years, and climate data in Section 4.3 suggests one possible source of this difference.
The taxonomic information included in this paper is limited to what is currently available in databases; many of the fungi that have been cultured and described are plant pathogens, and therefore our results may be biased towards these fungi. Therefore, caution is warranted in interpreting these guild data. Additionally, although we provide specieslevel identities for some ASVs, these are only tentative; without more information (such as sequences from additional barcode regions) we cannot be completely confident about species assignment from the ITS2 region alone [68].

Conclusions and Future Directions
Overall, the fungal communities we observed in tall fescue were diverse but not strongly affected by the presence of E. coenophiala. Although some fungi appeared to have a long term association with some or most plants, most were rare, and seemed to have a more transient relationship with their host as shown by the strong differences in the communities between years. In the future it would be interesting to observe these communities over a longer time period. It also might be helpful to investigate phyllosphere and endosphere communities separately, given that phyllosphere communities are likely to be more transient in nature than endophytic communities.
This research provides an initial assessment of the composition of the tall fescue fungal microbiome, but relatively little about how it might function in the plant. Next steps might include using other "omics" technologies (transcriptomics, proteomics, metabolomics) to clarify some of these mechanisms. Future research might also involve the use of plants that are genetic clones treated with a dilution series of a systemic fungicide (analogous to a gene knockout experiment) to further assess the fitness consequences of the fungal microbiome in this species.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/jof8101026/s1, Figure S1: Climate data, Figure S2: Rarefaction curves for all samples, Figure S3 Acknowledgments: This research was conducted on the treaty lands and territory of the Mississaugas of the Credit First Nation. We recognize that today this gathering place is home to many First Nations, Inuit and Métis peoples and acknowledging them reminds us of our collective responsibility to the land where we learn and work, and to our on-going efforts for reconciliation. We would like to thank Jeff Gross and the University of Guelph Genomics Facility staff for Illumina sequencing services, Terri Porter for bioinformatics advice and instruction, and Aurora Patchett for help with lab and bioinformatics methods. We would also like to thank Kim Bolton, Rie Kezia Matias, Carolyn Vandervelde, Jenn Roloson, Jordan Minigan, Charlotte Coates, Krista Nuziato, Misha Golin, and Holly Ivany for help with field work and seed counting. The authors thank four peer reviewers for their thoughtful and helpful comments on an earlier version of this paper.

Conflicts of Interest:
The authors declare no conflict of interest.