Disruption of Traditional Grazing and Fire Regimes Shape the Fungal Endophyte Assemblages of the Tall-Grass Brachypodium rupestre

The plant microbiome is likely to play a key role in the resilience of communities to the global climate change. This research analyses the culturable fungal mycobiota of Brachypodium rupestre across a sharp gradient of disturbance caused by an intense, anthropogenic fire regime. This factor has dramatic consequences for the community composition and diversity of high-altitude grasslands in the Pyrenees. Plants were sampled at six sites, and the fungal assemblages of shoots, rhizomes, and roots were characterized by culture-dependent techniques. Compared to other co-occurring grasses, B. rupestre hosted a poorer mycobiome which consisted of many rare species and a few core species that differed between aerial and belowground tissues. Recurrent burnings did not affect the diversity of the endophyte assemblages, but the percentages of infection of two core species -Omnidemptus graminis and Lachnum sp. -increased significantly. The patterns observed might be explained by (1) the capacity to survive in belowground tissues during winter and rapidly spread to the shoots when the grass starts its spring growth (O. graminis), and (2) the location in belowground tissues and its resistance to stress (Lachnum sp.). Future work should address whether the enhanced taxa have a role in the expansive success of B. rupestre in these anthropized environments.


INTRODUCTION
The study of the plant microbiome is a powerful tool that contributes to the understanding and prediction of the resilience of plant communities to global change (Vandenkoornhuyse et al., 2015). In extreme environments, where ecosystems are fragile and very sensitive to changes, the plant microbiota is crucial to understand the adaptive capacity of plant communities. Low temperatures, soil acidity, low rates of mineralization, and low nutrient availability are some of the common traits of artic and alpine ecosystems, where the study of the composition and function of the plantmicrobial consortium has been addressed in the last decades (Li et al., 2012;Poosakkannu et al., 2014;Rudgers et al., 2014;Bråthen et al., 2015;Kotilínek et al., 2017). Current research reports that for a particular ecosystem, the composition and diversity of the fungal microbiome of a plant host depends on its potential for carbon provision (plant abundance and biomass), on the fungal propagule availability, and on environmental factors, mainly climate, rather than on the taxonomy of the plant host (Higgins et al., 2014;Ranelli et al., 2015;Glynou et al., 2016;Kivlin et al., 2019). The first factor, potential for carbon provision, gives support to the resourcediversity hypothesis, the highest plant carbon determines the highest guest diversity (Wang et al., 2020), and this effect is modulated by the richness and diversity of the neighboring host plants and their own fungal populations Vannier et al., 2020). Regarding the climate, the historical and current precipitation regime is known to play a key role in fungal endophyte composition (Lau et al., 2013;Liu et al., 2017), and its consequences in a scenario of climate change are a growing subject of study in recent years (Giauque and Hawkes, 2016;Koide et al., 2017;Slaughter et al., 2018).
In addition to the climate change, plant communities in some cold environments are being affected by other significant mechanisms of change (Pauchard et al., 2009). Many high altitude grasslands undergo profound changes due to the disruption of the historical regime of disturbances that created and preserved them (Lasanta-Martínez et al., 2005;Komac et al., 2013). Fire and herbivory are two crucial disturbances shaping the landscape of natural grasslands worldwide (Archibald et al., 2005;Anderson and Hoffman, 2007). Extensive grazing by livestock in highlands has sharply declined in Europe in the last decades. Grazing is considered the major driver of plant guild composition and diversity in grassland communities (Milchunas et al., 1988;Canals and Sebastia, 2000;Frank, 2005;Eldridge and Delgado-Baquerizo, 2018). Grazing has profound effects on soils, promoting root exudation and carbon flow exchange to the rhizosphere (Dawson et al., 2000;Hamilton and Frank, 2001), enhancing soil microbial biomass and activating the soil biogeocycles (Bardgett and Wardle, 2003;Shaw et al., 2016), increasing and redistributing nutrients availability (Augustine et al., 2003;Liu et al., 2017) and affecting soil aeration by compaction (Jing et al., 2014). Through these effects on plants and soils, grazing may affect the structure of the fungal microbiome of plants (Schulz and Boyle, 2005;Wang et al., 2020).
Regarding fire, some high-altitude areas are currently experiencing an intensified regime of anthropic fires to reduce shrub encroachment and biomass build-ups caused by the relaxation of grazing activities (Köhler et al., 2005;Múgica et al., 2018). The effects of prescribed fires on soil properties have been documented in a range of contrasted habitats (Alcañiz et al., 2018), including high-altitude grasslands (in the Pyrenees, Armas-Herrera et al., 2016;San Emeterio et al., 2016). Although prescribed burnings do not reach the high temperatures of wildfires, their short and midterm effects (thermal shocking, aboveground plant combustion, ash deposition, nutrient mineralization...) may drive profound changes in the plant community composition if the fire regime increases in recurrence (Uys et al., 2004). Regarding microbial soil communities, even low-intensity fires may depress soil microbial biomass (Múgica et al., 2018), with complex consequences on the C and N cycles (Soong and Cotrufo, 2015;Shaw et al., 2016;Pellegrini et al., 2020). The composition of soil fungal communities has been documented to be disrupted by fires as well (Artz et al., 2009;Egidi et al., 2016;Semenova-Nelsen et al., 2019). Few studies have addressed the effects of fire on plant fungal assemblages (Bellgard et al., 1994;Eom et al., 1999;Mataix-Solera et al., 2002), although changes in species richness have been detected in foliar endophytes of burned trees (Huang et al., 2016), and specific plant-fungal mutualisms and pyrophilous fungal species have been described (Baynes et al., 2012;Raudabaugh et al., 2020).
The western Pyrenees encompass large areas affected by the decoupling of traditional fire and grazing practices. As a result of decreasing herbivore pressure and increasing burnings, a native perennial grass is expanding. Brachypodium rupestre (Host) Roem. & Schult (Schippmann, 1991;Schippmann and Jarvis, 1988) (=B. pinnatum subsp. rupestre (Host) Schübl. & Martens according to some authors), (Aizpuru et al., 1999), dominates grasslands and causes a severe loss of sympatric species (Canals et al., 2014;Canals, 2019), a phenomenon also observed in other European mountain ranges (Buckland et al., 2001;Holub et al., 2012;Tardella et al., 2018). In this natural setting, many questions arise concerning the response of the plant holobiont to the changed disturbance regime. To what extent the composition of the B. rupestre mycobiome responds to the altered disturbance regime? Does the species harbor a specific fire-adapted mycobiome in frequently burnt areas? Or on the contrary, is there sufficient inertia to maintain a similar mycobiome in a common climatic environment (cold temperatures, high rainfall and humidity), independently of the current disturbance regime? Since taxon-specific endophytes (such as Epichloë sp.) may confer a higher adaptive advantage in stressful situations than broadly distributed non-systemic endophytes (Hill et al., 1989;Malinowski and Belesky, 2000;Pereira et al., 2019;Harrison and Griffin, 2020), is the mycobiome of the plant responding to these expectations?
The main objective of this research was to characterize the culturable fungal endophyte community of B. rupestre across a well-defined gradient of disturbance. The gradient encompassed sites with different burning recurrences and levels of herbivory, which have led to different patterns of grassland structure and diversity. We also studied the fungal endophyte community of the aerial tissues of the most frequent companion species of B. rupestre in grasslands, Festuca rubra and Agrostis capillaris. The purpose was to estimate whether the major diversity of coexisting plant species (and potential fungal propagules) in diverse grasslands had a positive influence in the fungal assemblage of B. rupestre.

Study Site and Sampling Design
The study area was located in the Aezkoa Valley in the western Pyrenees (43 • 3 N1 • 13 W; 800-1900 m.a.s.l.), and occupies an area of 198 Km 2 . Because of the influence of the Atlantic ocean (55 km away in a straight line) the weather is cold and snowy during winter, and mild and foggy during summer. The mean annual temperature is 9.3 • C and mean annual precipitation is 1856 mm, according to data collected during 1989-2019 at the nearest climatic station, Irabia at 822 m.a.s.l. 1 . Soils are mainly derived from sandstones and calcareous clays, acidic with high organic matter and loamy or clay-loamy textures. The landscape is a mosaic of beech forests, gorse shrublands dominated by Ulex gallii Planch and Erica vagans L., and grasslands dominated by perennial grasses such as Festuca rubra gr., Agrostis capillaris L., Brachypodium rupestre (Host) Roem. & Schult, Danthonia decumbens (L.) and Avenula sulcata (J. Gay ex Boiss) Dumort. Other species are Galium saxatile L., Potentilla erecta (L.) Räeusch, Potentilla montana Brot., Hypochaeris radicata L., and the legume Trifolium repens L. The area is part of the European Protected Areas Network (Natura 2000) and was declared Special Area of Conservation (Roncesvalles-Selva de Irati, code ES0000126) in 2011 (Figure 1).
High-altitude grasslands are extensively grazed from May to October by sheep, cows and horses. Domestic herbivores have been present in the area since the Neolithic, as numerous prehistoric pastoral remains indicate. Where an appropriate extensive summer grazing exists, grasslands constitute highdiversity communities that are burned every 6-7 years by the bush-to-bush traditional practice to control localized shrub resprouting (San Emeterio et al., 2016). However, in the last decades, the number of grazers has sharply declined due to socioeconomical changes experienced in the valleys, and the use of repeated winter burnings to control the build-up of ungrazed biomass has increased (Durán et al., 2020). Nowadays, many areas are burned recurrently, every 1-2 years (Canals et al., 2017). Due to the frequent burnings and the lack of grazing, the grasslands are far less diverse, favored by the expansion and dominance of B. rupestre, which grows in dense and tall clumps. Consequently, the current open landscape is a combination of grassland communities with different degrees of cover by B. rupestre, which accurately reflects the level of herbivory and the burning recurrence.
Based on the cartographic information provided by the Management Plan of the area (Ferrer and Canals, 2008), we selected six grassland communities where B. rupestre was present at different coverages: three areas had above 75% cover and three areas under 25% cover ( Table 1 and Figure 1). The detailed floristic communities of the area are available in Durán et al., 2020. At each location, 40 turfs (400 cm 3 ) of B. rupestre (including shoots, rhizomes, roots and rhizosphere) were collected in summer 2018. We established a distance of ca 30 m among collected plants to avoid clonal individuals (Baba et al., 2012). In the three areas of diverse grasslands, 40 turfs of Festuca rubra and 40 turfs of Agrostis capillaris were also collected. Sampling points were georeferenced and the sampling grid covered ca 2.5 ha per location (Supplementary Figure 1). In total, 480 turfs were collected, placed in seedbed trays and transported to the UPNA laboratory. The plants in the turfs were processed for the isolation of fungi from shoots, rhizomes and roots in the same week.

Isolation and Identification of Fungi
We isolated fungi from shoots, rhizomes and roots of B. rupestre and from shoots of F. rubra and A. capillaris. Leaf sheaths and stems were cut into fragments of ca 5 mm, and surface-disinfected by immersion in a solution of 20% commercial bleach (1% active chlorine) containing 0.02% Tween 80 (v:v) for 10 min, and finally rinsed with sterile water. The fragments of rhizomes and roots were surface-sterilized with the same bleach-Tween 80 solution, but then treated with an aqueous solution of 70% ethanol for 30 s, and a final rinse with sterile water. About 10-12 tissue fragments of the same individual were plated in a Petri dish with potato dextrose agar (PDA) containing 200 mg/L of chloramphenicol to avoid the growth of endophytic bacteria. Petri dishes, kept at room temperature and ambient light, were checked daily for mycelium growth during 5 weeks. When mycelium emerged from a tissue fragment, a small amount was transferred to a new Petri dish to obtain a culture. Isolated fungi were grouped into morphotypes according to their morphological characteristics (colony color, exudates, growth type, and mycelium appearance). One or more isolates of each morphotype were genotyped for taxonomic purposes. To do this, a small amount of mycelium was scratched from the isolate culture and its DNA extracted using the Phire Plant Direct PCR Kit (Thermo Fisher Scientific). The ITS1, 5.8s and ITS2 regions were amplified using ITS4 and ITS5 primers (White et al., 1990). PCR amplification was done at 98 • C for 5 min, followed by 35 cycles of 95 • C for 5 s, 54 • C for 5 s, 72 • C for 20 s, and a final phase of 72 • C for 1 min. Amplicons were purified (Favor Prep TM Plant Genomic DNA Extraction Mini Kit, Favorgen) and sequenced by the Sanger method at an external sequencing service (STABvida).
Because the range of intraspecific variation in ITS sequences is unknown for most fungal species (Taylor et al., 2000), DNA sequences were clustered using the CD_HIT program (Li and Godzik, 2006;Huang et al., 2010) and those with 97% or more similarity were considered to belong to the same taxon. A representative sequence of each cluster was selected and used to search the database of the ITS region from fungi type and reference material (Schoch et al., 2014) at the National Center for Biotechnology Information (NCBI) using the BLAST algorithm. In addition, the UNITE database of fungal nucleotide sequences was used as a complement for sequences without type specimens in NCBI. We also used the database FUNGuild (Nguyen et al., 2016) to get information on the ecological guild of each taxon, and to estimate for their possible functional roles.

Data Analyses
To evaluate the efficiency of our sampling effort for measuring species richness, species accumulation curves were estimated with all the species, and also excluding singletons, species that appeared only once (Vegan package, Oksanen et al., 2017). We calculated the incidence of all endophyte species within each type of tissue (shoots, rhizomes and roots), grassland type (low and high-diversity) and host plant species, and determined richness and Shannon and Simpson diversity indexes (Vegan package, Oksanen et al., 2017). Differences in species richness and diversity indexes of B. rupestre endophytes were analyzed using twoway ANOVA with tissue and grassland type as fixed factors, and permutation tests was used to compare diversity indexes of B. rupestre, F. rubra, and A. capillaris shoots (Coin package, Hothorn et al., 2008).
Venn diagrams were used to represent the taxa shared among grassland types, plant tissues and grass species (Euler package, Chen and Boutros, 2011). Frequencies were calculated from the matrix of presence/absence of isolates and permutational analyses of variance (PERMANOVA) were used to evaluate the variability of the fungal endophyte assemblages of B. rupestre between grassland types and among plant tissues. For that purpose, Adonis function was used (Vegan package, Oksanen et al., 2017). Distances were calculated using Bray-Curtis dissimilarities, set the number of permutations to 9999 and constrained the permutations within each location. Since PERMANOVA analyses are very sensitive to heterogeneity of multivariate dispersions, homogeneous dispersion between treatment groups was tested using the betadisper function.
To identify which species were characteristic of a particular tissue and grassland type, the indicator species tests in the Labsdv package was used (Roberts, 2019). The indicator value of a species (indval) measures the fidelity and relative abundance of the species in a particular situation (Dufrêne and Legendre, 1997). General Linear Mixed Models (GLMM) was used to determine whether the indicator species for a particular grassland type presented differences in the probability of incidence among grassland types. The grassland type was included as the fixed factor, the location as the random factor, using a binomial distribution. In addition, GLMM was used to analyze whether the probability of incidence of some mycobiome core species from the aboveground tissues presented differences among host plant species. Host plant species was included as the fixed factor, location as the random factor, using a binomial distribution. GLMM's were done using the lme4 package (Bates et al., 2015).

Isolation and Identification of Fungal Endophytes
We plated ca 10000 tissue fragments of B. rupestre (shoots, rhizomes and roots), F. rubra (shoots) and A. capillaris (shoots) in 960 culture media plates (10-12 fragments per plate). We obtained 1151 isolates (ca 190 isolates per location) which were classified into a total of 95 morphotypes. One or more isolates of each morphotype were sequenced, obtaining 116 sequences. After the sequence clustering process, 61 different sequence types remained, which were identified into 53 different taxa using the NCBI and UNITE databases. Twenty-two of them were classified to species rank, 20 to genus, 6 to family, 3 to order, and the remaining two to class rank (Supplementary Table 1 contains the complete list of taxa identified and their accession number).

Diversity of the B. rupestre Mycobiome
Fungi were isolated from 58% of the B. rupestre plants in lowdiversity grasslands and from 43% in high-diversity grasslands ( Table 2). Endophyte incidence varied among tissues: 37% of shoots, 52% of rhizomes and 64% of roots harbored fungi. Forty-five different taxa were identified in B. rupestre. Endophyte species richness in B. rupestre shoots, rhizomes and roots were 9, 25, and 26 taxa, respectively. Endophyte species richness in B. rupestre growing in low-diversity grasslands ranged from 9 to 22, and from 12 to 18 taxa in high-diversity grasslands ( Table 2). Roots and rhizomes shared 23.9% of the species, whereas shoots with rhizomes and with roots shared 8.7 and 6.5%, respectively, all tissues had only 6.5% of all species in common (Figure 2). B. rupestre plants in low and high-diversity grasslands shared 39.1% of the species.
Omnidemptus graminis was more frequent in shoots than in rhizomes and absent in roots, whereas Albotricha sp., Lachnum sp. B, and Mollisiaceae sp. were more frequent in belowground tissues than in shoots. In addition to the Mollisiaceae sp. taxon, other dark septate endophytes (DSE) such as Cadophora sp., Microdochium bolleyi, Microdochium neoqueenslandicum, and Periconia sp. were found, although with a low incidence. Members of the Clavicipitaceae family, such as Metapochonia bulbillosa, Metarhizium carneum, and Epichlöe typhina in shoots and rhizomes also occurred. The frequencies of the remaining species were under 2% (Supplementary Table 2). Three trophic types (pathotroph, saprotroph, and symbiotroph) were found in the mycobiome of B. rupestre. The most common guilds were plant pathogens, endophytes and undefined saprotrophs. However, a remarkable number of taxa were animal pathogens (Clonostachys rosea, Tolypocladium album, and Trichoderma koningii are described as entomopathogens, Metapochonia bulbillosa and Sarocladium strictum as nematophagous). Fungi with antimicrobial activity (Fusarium circinatum, Glarea sp., Nemania sp., and Penicillium ortum) and species with a high source of bioactive compounds (Acremonium sp., Gaeumannomycella sp., and Lachnum sp.) also occurred (Supplementary Table 1).
The species-accumulation curves including all the endophyte taxa were non-asymptotic (Figure 3), but when singletons were excluded the curves reached a plateau, suggesting that an increase in the sampling effort would reveal few common taxa, and mostly rare species. Therefore, a sampling effort of 100 plants is adequate to flatten the curves of the different tissues when excluding the singletons (Figures 3B-D). A greater sampling effort would be needed to include all the tissues (Figure 3A), although low-diversity grasslands need more plants than highdiversity grasslands to characterize their endophyte richness (Figures 3E,F).

Distribution Patterns of the B. rupestre Mycobiome
A homogeneous dispersion was found within different tissues (Beta-disper; F2,15 = 2.195, p = 0.146) and grassland types (Beta-disper; F1,16 = 0.028, p = 0.870), contrasting with the significant effects on community assemblages. PERMANOVA analysis showed that tissue and grassland type had a significant effect on the endophytic community of B. rupestre plants (both p < 0.001). Tissue explained ca 50% of the variance of the model (r 2 = 0.508), whereas grassland type explained ca 5% (r 2 = 0.052). The NMDS (Non-Metric Multidimensional Scaling) plot discriminated the assemblages according to the tissues and included the indicator species (Figure 5) GLMM's showed that B. rupestre plants from low-diversity grasslands had a greater probability of having their roots infected by Lachnum sp. B (LRT = 4.5719, p = 0.032) and their shoots by Omnidemptus graminis (LRT = 4.679, p = 0.030) than plants from high-diversity grasslands (Figure 6). Regarding Albotricha sp., the patterns were not so clear and its probability of incidence in B. rupestre roots was not significantly different between grassland types (LRT = 0.2849, p = 0.594, Figure 6).

Aboveground Endophyte Assemblages of the Most Common Grasses in High-Diversity Grasslands
The incidence of endophytes varied among the shoots of the three grass species, ranging from 25% in B. rupestre and A. capillaris to 38% in F. rubra. For each plant species, a strong variability of endophyte incidence among locations was detected ( Table 4). Eight taxa not observed in the B. rupestre survey were isolated from F. rubra or A. capillaris (Supplementary Table 1).
A total of 23 taxa were isolated from the shoots of B. rupestre, F. rubra and A. capillaris (Supplementary Table 3). Endophyte species richness in B. rupestre, F. rubra, and A. capillaris shoots were 5, 10, and 16, respectively ( Table 4). F. rubra and A. capillaris shared more endophyte species between them than B. rupestre with either (Figure 7). The three grasses had two endophyte species in common, Lachnum sp. B and Omnidemptus graminis. The most frequent endophyte species in B. rupestre shoots were Omnidemptus graminis and Epichloë typhina, in F. rubra Epichlöe festucae and Mollisiaceae sp., and in A. capillaris Lachnum sp. B, Epichloë baconii and Mollisiaceae sp. (Table 5).
When including all endophyte species, the speciesaccumulation curves of the three host grasses were non-asymptotic (Figure 8), but when removing singletons the curves flattened with a sampling effort of 50 plants for B. rupestre and 60 plants for F. rubra and A. capillaris in high-diversity grasslands (Figure 8).
The mycobiome of A. capillaris appeared to be more rich and diverse (averaged per site richness AC = 7.0, Shannon AC = 1.85, Simpson AC = 0.83) than that of F. rubra and B. rupestre (averaged richness FR = 4.3, Shannon FR = 0.76, Simpson FR = 0.36; averaged richness BR = 2.7, Shannon BR = 0.49, Simpson BR = 0.26). However, the high variability among locations and the size of the sample (Figure 9) did not allow to detect significant differences in richness and diversity indexes among grasses in the permutations tests (richness, t = 1.7876, p = 0.1737; Shannon, t = 2.0697, p = 0.09614; Simpson, t = 2.0193, p = 0.1076).
Omnidemptus graminis and Epichloë typhina were species indicators of B. rupestre (p = 0.001; p = 0.008), whereas Epichloë festucae was a species indicator of F. rubra (p = 0.001) and Epichloë baconii of A. capillaris (p = 0.009). The GLMM showed that the probability of incidence of Omnidemptus graminis significantly varied among species (LRT = 38.194, p < 0.001), and had a higher probability of incidence in B. rupestre, despite its presence in the other grasses (Figure 10). On the contrary, Epichloë species were specific of each grass (E. typhina in B. rupestre, E. festucae in F. rubra and E. baconii in A. capillaris), but the probability of Epichloë infection was higher in F. rubra than in the rest of grasses (LRT = 34.581, p < 0.001; Figure 10).

Shoot Fungal Endophyte Assemblages of the Most Common Grasses
The richness and diversity of the mycobiomes from this survey was relatively low compared to other studies in grasses (Sánchez-Márquez et al., 2008Pereira et al., 2019). The study of fungal endophytes is highly method-dependent (Hyde and Soytong, 2008). In this research, we applied a conservative protocol, with along sterilization process -i.e., 10 min vs. 6 min in Pereira et al. (2019)-, and discarding the mycelia growing in the first day, which might influence the final counts of species (Burgdorf et al., 2014). Despite the low richness reported, interesting patterns emerged from the shoot assemblages of the grasses. B. rupestre hosted the poorest mycobiome, half of that present in F. rubra, and one third of that of A. capillaris. Only two of 23 endophyte species identified were common to all grasses (Omnidemptus graminis and Lachnum sp. B). Recent studies indicate that plant identity influences foliar fungal assemblages more than expected from previous literature (Kivlin et al., 2019). Since the three grasses are sympatric and share an identical abiotic environment (i.e., climate, soil, location), other factors might define their mycobiomes. Plant abundance and capacity for carbon provision (B. rupestre develops more biomass but it is generally less abundant in diverse grasslands than the other grasses), potential interactions among fungal species (i.e., Omnidemptus graminis, which is abundant in B. rupestre, can induce host defense against other fungi (Schulz and Boyle, 2005;Constantin et al., 2020) and the different degree of herbivory (B. rupestre is less palatable to livestock than A. capillaris and F. rubra), may play a definite role in the structure of fungal assemblages.
Regarding systemic endophytes, the three grasses were infected by species of the genus Epichloë. E. typhina, E. baconii, and E. festucae are highly specific for B. rupestre, A. capillaris, and F. rubra, respectively (Leuchtmann et al., 2014;Saikkonen et al., 2016). Epichloë species are known to produce longterm, systemic infections and have a narrow range of hosts, limited to a genus or to related genera (Schardl et al., 2009;Schirrmann et al., 2015). In the area of study, the incidence of Epichloë was variable depending on species and locations, and symbioses with E. festucae were by far the most abundant (up to 50% incidence in one location) compared to E. typhina and E. baconii (less than 8% incidence in all locations). This is in accordance with previous research showing that the incidence of Epichloë-grass infections change depending on the associations (Leuchtmann and Schardl, 1998). E. festucae mainly reproduces asexually by vertical transmission to seeds, while in E. typhina and E. baconii seed transmission is absent (Leuchtmann et al., 2014). B. sylvaticum, a species close to B. rupestre, is very common in forest undergrowth, and displays high levels of infection by E. sylvatica, a seed-transmitted species (Meijer and Leuchtmann, 1999;Brem and Leuchtmann, 2001). The results suggest that the capacity of systemic endophytes to infect host seeds and transmit vertically, which is the most common reproductive mechanism in Epichloë species, leads to higher levels of incidence than in sexual, horizontally transmitted Epichloë species. This makes sense because in vertically transmitted endophytes reproductive fitness is intimately tied to that of their hosts (Saikkonen et al., 2002). As a consequence, mutualisms are expected to be more successful and intense in these situations. Of the three Epichloë species identified, E. festucae has been extensively studied for its capacity to establish successful symbioses in the most extreme environments (Vázquez-de-Aldana et al., 2013b;Zabalgogeazcoa et al., 2013;Leinonen et al., 2019;Pereira et al., 2019), and for the variety of advantages that confers to the host plant (i.e., tolerance to stress, resistance to herbivores, plant fitness) (Malinowski and Belesky, 2000;Zaurov et al., 2001).

Fungal Tissue Assemblages and Core Mycobiome of B. rupestre
Roots and rhizomes of B. rupestre displayed different endophyte assemblages than aboveground tissues, had the greatest percentages of fungal infection (Figure 5), and a significantly high richness and diversity of taxa. Previous studies have shown that belowground tissues have greater endophyte diversity than aerial tissues, and this could be related to the level of tissue exposition to fungal inoculum (Sánchez-Márquez et al., 2010;Harrison and Griffin, 2020). In B. rupestre, as in many perennial grasses, most aboveground tissues die and renew annually, while belowground tissues such as rhizomes, survive and consequently have a larger time frame for fungal reinfection.
As in other grass species, the mycobiome of B. rupestre was constituted by a few core species and many rare species (Comby et al., 2016;Ofek-Lalzar et al., 2016;Pereira et al., 2019;Sun et al., 2020). Only two species were found in more than 20% of the B. rupestre plants and in most locations, Omnidemptus graminis in shoots and rhizomes and Lachnum sp. B in roots and rhizomes. O. graminis is a member of the Magnaporthaceae that has been described recently (Hernández-Restrepo et al., 2019). Although some Magnaporthaceae strains have shown plant-growth promoting activity (Yuan et al., 2010;Changyeol et al., 2017), the best known members of this family are pathogens of grasses, associated to roots (Gaeumannomyces graminis and Magnaporthe poae) and shoots (Magnaporthe oryzae) (Illana et al., 2013). M. oryzae is a hemibiotrophic fungus that causes the rice blast disease. This species switches from a biotrophic growth phase -it feeds from the host plant without killing its cellsin early infection to a necrotrophic stage (Kankanala et al., 2007). According to Talbot et al. (1997), the switch between phases may be triggered by the lack of nutrients within the host cell. In this survey, O. graminis was present in asymptomatic plants, suggesting at least the occurrence of a biotrophic phase of unknown duration. As many other fungal Lachnum is a large genus within the Hyaloscyphaceae family with more than 250 species described and distributed in a wide range of habitats and host species (Wu and Su, 2007;Nagao, 2008). Most Lachnum species are latent saprophytes, that may grow as endophytes in roots of perennial grasses (Sánchez-Márquez et al., 2010;Pereira et al., 2019) and form ericoid mycorrhiza coils in some Ericaceae (Walker et al., 2011). Many Lachnum sp. are bioactive, producing a wide range of biologically active compounds (Shan et al., 1997;Ondeyka et al., 2009;McMullin et al., 2017;Zong et al., 2017). In addition to the core species Omnidemptus graminis and Lachnum sp., the culturable mycobiota of B. rupestre was characterized by a remarkable number of taxa with recognized defensive activities, such as entomopathogenic fungi (Tolypocladium album and Sarocladium strictum), (Gera Hol et al., 2007;Quesada-Moraga et al., 2014;El-Sayed et al., 2020), nematophagous fungi (Metapochonia bulbillosa and Clonostachys rosea) (Sankaranarayanan et al., 1997;Ownley et al., 2010;Manzanilla-López and Lopez-Llorca, 2017), and fungi with antifungal activity (Fusarium circinatum, Penicillium ortum, Lachnum sp., Glarea sp., Trichoderma sp., Trichoderma koningii and Nemania sp.) (Ondeyka et al., 2009;Youssar et al., 2011;Zhang et al., 2014;Mousa et al., 2015;Kornsakulkarn et al., 2017). The known defensive functions of this group of microorganisms  may confer a decisive advantage to B. rupestre, which may use them for its own protection, as predators of its pathogens and pests, according to the bodyguard hypothesis (Elliot et al., 2000).
To what extent this particular defensive assemblage plays a key role in the success and expansion of B. rupestre with regard to other grasses is a matter of interest that needs further research. Low and high-diversity grasslands, generated by a distinctive disturbance regime, had similar fungal richness and diversity (Figure 4), but different endophyte assemblages (Figure 2 and PERMANOVA analysis). Differences were mostly due to the different probability of incidence of the core taxa (Figure 6). B. rupestre in low-diversity grasslands had a greater probability of incidence of Omnidemptus graminis and Lachnum sp. B than in high-diversity grasslands, and the two grassland types only shared 36% of the total fungal taxa identified. Considering that we sampled 120 turfs per grassland type, and that we did not reach the asymptote in the species accumulation curves (Figure 3), a more exhaustive sampling effort would be necessary to draw conclusions on a specific cohort of rare taxa per grassland type. With regard to the systemic endophyte E. typhina, no significant differences in the probability of incidence were found between grassland types. In consequence, the highest degree of incidence of the two core taxa, O. graminis (in shoots and rhizomes) and Lachnum sp. B (in roots and rhizomes), in low-diversity, recurrently burned grasslands was the most sound result of this part of the study. From the spectrum of core species infecting B. rupestre, O. graminis was only found in shoots and rhizomes, which indicates its affinity for both tissues, which are anatomically similar (de Kroon and Knops, 1990), and different from roots ( Table 3). When isolated and cultured in plates, we observed rapid mycelial growth of O. graminis compared to the rest of species. We hypothesize that fires destroy the aerial mycelium of O. graminis, together with the aboveground biomass of the plant, but the fungus remains in the rhizome reservoir. Since fire is applied in winter time in moist soils, the temperatures reached in the top soil are low (at 1 cm deep, soil temperature rise 9-10 • C and no change is measured at 5 cm depth -data unpublished), and the function of rhizomes remains unaffected. Consequently, recolonization of the aboveground tissues by the rhizome mycelium of O. graminis can occur rapidly, paralleling the regrowth of the plant, and conferring a decisive initial advantage to this fungal species.
Regarding Lachnum sp. B, the increased incidence of this fungal species in frequently burnt areas paralleled a consistent decrease of Albotricha sp. in roots and rhizomes (Table 3). These two endophytic taxa constitute a large part of the belowground mycobiome of B. rupestre and belong to the same family, Hyaloscyphaceae. The close phylogeny of both genera and the polyphyly of the Lachnum group has been demonstrated in genetic studies (Ye and Zhuang, 2003;Hosoya et al., 2010). The pattern of increased Lachnum sp. and decreased Albotricha sp. infection in the most burned sites, may indicate a negative interaction between both species, or a contrasting response among them to the fire disturbance. The capacity of Lachnum sp. to cope with environmental stress, by producing sclerotia and adopting latent forms, and its growth in belowground structures, which are less affected by fire than aboveground structures, may help to explain its success. In particular, in temperate shrublands close to the region of study, ascocarps of Lachnum pygmaeum have been observed on charred wood and roots of Ulex europaeus (unpublished results). It is known that the thermal shock produced by fire may enhance fungal fecundity and activate the development of sexual structures in some pyrophilous fungi (Raudabaugh et al., 2020). Also, fungal biomass may be increased by fires, as demonstrated for the pyrophilous fungi Morchella sp. infecting the grass Bromus tectorum (Baynes et al., 2012). To what extent the particular Lachnum sp. B identified in the area of study is a pyrophilous fungus merits additional study.
Although the incidences of O. graminis and Lachnum sp. B are favored in such anthropized fire-prone habitats, the extent to which these endophytes confer particular advantages to its host, B. rupestre, is unknown. Further experimental FIGURE 9 | Species richness and diversity indexes (Shannon, Simpson) of the mycobiome of shoots from three representative grasses of high-diversity grasslands from three locations. FIGURE 10 | Probability of incidence of Epichloë spp. and Omnidemptus graminis in the aboveground tissues of the most common grasses of high-diversity grasslands.
research is needed to evaluate whether O. graminis and/or Lachnum sp. B infected plants perform better after fires than non-infected plants and whether the relationship in such a disrupted environment relies in a mutualistic relationship. In this research, the systemic E. typhina did not display significant differences in incidence among grasslands, and its percentage of infection was low at all sites. Previous research on Epichloë endophytes of grasses did not find evidence of a mutualistic relationship associated to fire (Faeth et al., 2002;Hall et al., 2014), contrary to grazing intensities that have been positively related to the abundance of vertically transmitted Epichloë producers of toxic metabolites (Vázquez-de-Aldana et al., 2010;Hume et al., 2020).

CONCLUSION
The perennial tall-grass B. rupestre had a moderately diverse endophytic mycobiome consisting of a few core species and many rare species in assemblages that differed between aerial and belowground tissues. Recurrent grassland burnings, which eliminate the aerial biomass of the grass every 1-2 years, did not affect the richness and the diversity of the fungal community in B. rupestre, but the percentages of infection of two core taxa, Omnidemptus graminis and Lachnum sp. B, were significantly modified. The results indicate that although in frequently burnt areas the same core species of diverse grasslands subsist, Omnidemptus graminis and Lachnum sp. B are singularly benefited due to the proposed following mechanisms: (1) the capacity to survive belowground in rhizomes during the winter, and to spread rapidly to the shoots when the plant starts its spring growth (O. graminis), and (2) the location in belowground tissues (Lachnum sp. B) and the higher resistance to stress than other core root fungi, such as the related-taxon Albotricha sp. Following steps should address whether these two core taxa benefit the expansive success of B. rupestre in these antrophized, fire-prone environments, as well as to determine whether the cohort of less abundant fungi with well-defined defensive functions play a role too.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/ Supplementary Material.

AUTHOR CONTRIBUTIONS
IZ and BV introduced the training in culturable techniques for fungal identification to MD. MD and LS designed the experiments and analyzed the data. MD, LS, and LM collected the plants and isolated the fungi. MD and IZ identified the fungi. MD and RC wrote the manuscript. LS, IZ, and BV reviewed the manuscript. MD and RC got the projects that supported financially the research. All authors contributed to the article and approved the submitted version.

FUNDING
MD and LM got funding through two UPNA's doctorate scholarships. The research was supported by three projects financed by Ecoembes-SEO/Birdlife (Libera 2017), "la Caixa" Foundation and CAN foundation (LCF/PR/PR13/51080004), and the Spanish Ministry of Science and Innovation and FEDER funds (AGL2016-76035-C2-1R).

ACKNOWLEDGMENTS
We are very grateful to B. Soret, C. Paredero, and I. Maquirriain for offering technical support in the laboratory tasks. We thank the Aezkoa Valley Council for its interest and willingness to process the permissions necessary for sampling in the study sites.