Diversity and Distribution Patterns of Endolichenic Fungi in Jeju Island, South Korea

Lichens are symbiotic organisms containing diverse microorganisms. Endolichenic fungi (ELF) are one of the inhabitants living in lichen thalli, and have potential ecological and industrial applications due to their various secondary metabolites. As the function of endophytic fungi on the plant ecology and ecosystem sustainability, ELF may have an influence on the lichen diversity and the ecosystem, functioning similarly to the influence of endophytic fungi on plant ecology and ecosystem sustainability, which suggests the importance of understanding the diversity and community pattern of ELF. In this study, we investigated the diversity and the factors influencing the community structure of ELF in Jeju Island, South Korea by analyzing 619 fungal isolates from 79 lichen samples in Jeju Island. A total of 112 ELF species was identified and the most common species belonged to Xylariales in Sordariomycetes. The richness and community structure of ELF were significantly influenced by the host taxonomy, together with the photobiont types and environmental factors. Our results suggest that various lichen species in more diverse environments need to be analyzed to expand our knowledge of the diversity and ecology of ELF.


Introduction
Lichens are symbiotic organisms in which the mycobiont (lichen-forming fungi) and the photobiont (green algae and/or cyanobacteria) live together in a mutualistic relationship: the mycobiont protects the photobiont against external environmental stress by forming a thallus, and the photobiont provides photosynthetic carbon as a reward [1,2]. Lichens play ecologically important roles in ecosystems as food and habitats for animals, as well as participants in nutrient cycling and soil formation [3,4]. In addition, they produce numerous secondary metabolites that are industrially or pharmaceutically effective compounds such as antibiotics, anti-tumor agents, and antioxidants [5]. Recent studies have shown that diverse microorganisms exist within lichen thalli, and they can influence the physiology of host lichen in a similar manner to the influence of endophytes on the host plant [6][7][8]. Endolichenic fungi (ELF) are non-mycobiont fungal species living in the lichen thallus [9]. They are distinct from the lichenicolous fungi in terms of symptomless characteristics. As for the relationship between ELF and lichens, it is unclear whether they have any kind of intimate association [8]. Given that the endophytic fungi influence the plant physiology and increase a tolerance against environmental stress (e.g., high temperature, drought, or pathogens) [10][11][12], it is suggested that ELF promote the biological function of lichens in the ecosystem [8]. Therefore, the biodiversity and distribution pattern of ELF is important not only for understanding the ecology and physiology of lichens, but also for the maintenance of ecosystem sustainability in the era of global climate change.
The ELF diversity has been studied in various biomes from tropical areas to polar regions using culture-dependent [13][14][15][16][17][18][19][20] or independent approaches [19,[21][22][23][24][25]. ELF are phylogenetically diverse, covering all lineages of Ascomycota and a minority of Basidiomycota and Mucoromycota [9,14]. The evolution of ELF is not well understood, but it is thought to be polyphyletic, and ELF have been suggested as the origin of endophytism, which would explain the similar phylogenetic range [9]. The ELF diversity is different from the co-existing endophytic diversity, which indicates that ELF are a distinctive ecological group differing from endophytic fungi [14,26]. However, the ELF diversity has been studied mainly in Europe [13,27,28], North America [9,14], and Southern Asia [17,29,30], while the ELF in Eastern Asia are largely unexplored [31]. Therefore, it is expected that exploring the ELF diversity in Eastern Asia can expand the knowledge of fungal diversity associated with lichens.
Several abiotic and biotic factors drive the pattern of fungal diversity and community structures [32][33][34]. It is well known that endophytic fungal communities are influenced by the climate [15,35], geographic position of the region [36], and host taxonomy [37,38]. The factors influencing the ELF community structures have rarely been investigated, but a few studies have suggested that geographic characteristics and the host taxonomy are important factors shaping the ELF community structure [9,14,15,27,28]. The ELF diversity was high in tropical or subtropical regions, probably due to favorable climatic conditions [9,15]. The lichen host is one of the factors that determine the ELF communities [14,15,27,28]. However, other studies showed no effect of the host taxonomy [39], which suggests that the pattern of ELF diversity and factors governing the ELF community may vary depending on the geographical region and spatial scales.
In this study, we investigated the ELF diversity and community structure in Jeju Island (South Korea). Jeju Island is composed of warm lowlands and subalpine (Mt. Halla, 1947 m above sea level) biomes with deciduous-evergreen broadleaved and coniferous forests [40], as well as many reported lichens [41][42][43][44]. The oreums, a special structure of parasitic volcanoes, are distributed from the lowland to the mountain area [45,46] and are covered by forests with a high diversity of animals [47,48], plants [49,50], and microorganisms [51,52]. Although study of lichen flora in oruems is limited, several studies have revealed novel records of lichen species from oreums [53,54]. However, the diversity of ELF is largely unexplored in South Korea as well as in oreums. We collected different species of lichens from several oreums in Jeju Island, and investigated the diversity of ELF by culture isolation. Moreover, we analyzed the pattern of the diversity and community structure of ELF to understand the effect of various factors (host taxonomy, photobiont type, and ecoregions) on the relationship between ELF and lichens.

Lichen Sample Collection and ELF Isolation
Lichen specimens were collected from sampling sites at 29 oreums covering the whole island from the lowland near the coast to the high mountain area in Mt. Halla, from April to September 2017 ( Figure 1; Table S1 (Supplementary Materials)). We focused on the foliose lichens, except for Stereocaulon (fruticose), because the wide area of the thallus is helpful to acquire sufficiently large thalli and to avoid contamination. Healthy and fresh thalli of lichens on rock, soil, and trees were collected (Table A1) and transferred to the laboratory in paper bags. The lichen thalli were subjected to sterilization. Before sterilization, the litter and debris attached on the thalli were removed using a needle of a syringe, and the thalli were washed in running tap water. The surface of the thalli was sterilized by a modified method used in the previous study [14]: 95% ethanol for 30 s, 0.5% NaOCl for 2 min, 70% ethanol for 30 s, and rinsed three times with sterile distilled water. The thalli were dried on sterilized paper and cut into 1 cm 2 pieces using sterilized scissors. For each specimen, a total of 20 fragments were placed on four plates of 90 mm Petri dishes containing a potato dextrose agar (PDA; Difco Laboratories, Detroit, MI, USA), and incubated at 25 °C for up to 12 weeks. The plates were checked every day, and fungal isolates were transferred to PDA medium for pure culture. Additionally, ELF strains previously isolated from Jeju Island using the same procedure and deposited in the Korean Lichen Research Institute (KoRLI) were included in the analysis to increase the volume of the dataset. To recover the strains from KoRLI, they were cultured on PDA medium at 25 °C.

Molecular Identification of Fungal Strains
Genomic DNA of all isolated fungal strains was extracted using the method of Park et al. [55]. For molecular identification, the nuclear ribosomal internal transcribed spacer (ITS) region was amplified using ITS1F and ITS4 [56]. PCR amplification was performed using an AccuPower PCR PreMix kit (Bioneer, Daejeon, South Korea) with 1 µL of DNA, 1 µL of each primer, and 17 µL of sterilized distilled water in the following condition: 5 min at 95 °C, 30 cycles of 30 s at 94 °C, 30 s at 56 °C, and 30 s at 72 °C, and a final extension for 5 min at 72 °C. PCR products were checked on 1% agarose gel electrophoresis, and sequencing was performed at GenoTech (Daejeon, South Korea).
Sequences were checked on FinchTV 1.4.0 (Geospiza Inc., WA, USA). Preliminary molecular operational taxonomic units (mOTUs) were constructed by 99% sequence similarity using Vsearch v. 2.14.1 [57] and its taxonomic candidates were assigned using BLAST against GenBank. Final taxonomy was assigned based on phylogenetic analysis with reference sequences. Sequences were aligned using MAFFT v.7 [58] and checked manually on MEGA v.5 [59]. The phylogenetic tree was constructed using a maximum likelihood (ML) analysis using RAxML v. 8.0.2 [60] with the GTRGAMMA model and 1000 bootstrap replicates. Finally, preliminary mOTUs with the same The lichen thalli were subjected to sterilization. Before sterilization, the litter and debris attached on the thalli were removed using a needle of a syringe, and the thalli were washed in running tap water. The surface of the thalli was sterilized by a modified method used in the previous study [14]: 95% ethanol for 30 s, 0.5% NaOCl for 2 min, 70% ethanol for 30 s, and rinsed three times with sterile distilled water. The thalli were dried on sterilized paper and cut into 1 cm 2 pieces using sterilized scissors. For each specimen, a total of 20 fragments were placed on four plates of 90 mm Petri dishes containing a potato dextrose agar (PDA; Difco Laboratories, Detroit, MI, USA), and incubated at 25 • C for up to 12 weeks. The plates were checked every day, and fungal isolates were transferred to PDA medium for pure culture. Additionally, ELF strains previously isolated from Jeju Island using the same procedure and deposited in the Korean Lichen Research Institute (KoRLI) were included in the analysis to increase the volume of the dataset. To recover the strains from KoRLI, they were cultured on PDA medium at 25 • C.

Molecular Identification of Fungal Strains
Genomic DNA of all isolated fungal strains was extracted using the method of Park et al. [55]. For molecular identification, the nuclear ribosomal internal transcribed spacer (ITS) region was amplified using ITS1F and ITS4 [56]. PCR amplification was performed using an AccuPower PCR PreMix kit (Bioneer, Daejeon, South Korea) with 1 µL of DNA, 1 µL of each primer, and 17 µL of sterilized distilled water in the following condition: 5 min at 95 • C, 30 cycles of 30 s at 94 • C, 30 s at 56 • C, and 30 s at 72 • C, and a final extension for 5 min at 72 • C. PCR products were checked on 1% agarose gel electrophoresis, and sequencing was performed at GenoTech (Daejeon, South Korea).
Sequences were checked on FinchTV 1.4.0 (Geospiza Inc., WA, USA). Preliminary molecular operational taxonomic units (mOTUs) were constructed by 99% sequence similarity using Vsearch v. 2.14.1 [57] and its taxonomic candidates were assigned using BLAST against GenBank. Final taxonomy was assigned based on phylogenetic analysis with reference sequences. Sequences were aligned using MAFFT v.7 [58] and checked manually on MEGA v.5 [59]. The phylogenetic tree was constructed using a maximum likelihood (ML) analysis using RAxML v. 8.0.2 [60] with the GTRGAMMA model and 1000 bootstrap replicates. Finally, preliminary mOTUs with the same species name were combined to single mOTUs. From there, mOTUs were referred as species names, but species identification was solely based on ITS sequences, without considering the morphological and cultural characteristics of the isolated strains. All generated sequences were deposited in GenBank under the accession numbers MN341225-MN341843.

Analysis of Diversity and Community Structure
Statistical analyses were conducted on R v.3.5.1 [61]. The taxonomy of the host lichen (genus), photobiont types, and ecoregions of the sampling sites were analyzed as the factors influencing the ELF diversity and community. For the lichen taxonomy, only the lichen genera represented by more than two samples were used for further analysis. The ecoregions were categorized using a cluster package [62] with partitioning around medoids (PAM) based on the elevation and climate data. Elevation data were obtained from a NASA shuttle radar topographic mission (SRTM) dataset [63] as a digital elevation model (DEM). For climate data, 19 bioclimatic variables were acquired from WorldClim database v.2 (30 arc second) [64]. Among the bioclimatic variables, BIO01 (Annual mean temperature), BIO03 (Isothermality), BIO07 (Temperature annual range), BIO12 (Annual precipitation), BIO14 (Precipitation of the driest month), and BIO15 (Precipitation seasonality) were chosen by a stepwise backward variable selection after removing high variance inflation factor (VIF > 10) using the usdm package [65]. The clustering of sampling sites according to their bioclimatic variables resulted in the delineation of three ecoregions based on the silhouette coefficient [66] and Elbow method [67]. A higher value of silhouette coefficient indicates the better quality of clustering [66]. The Elbow method suggests the optimal number of clusters when the additional number of clusters does not decrease the within-cluster variation (total within sum of square, WSS) much [67]. We computed the silhouette coefficient and WSS from the various number clusters (1-10 clusters), and choose the best number of clusters with the highest silhouette coefficient and determined by the Elbow method. The sampling sites grouped into the ecoregions were represented on the map created by QGIS v.3.8 [68] with DEM data.
The richness levels of ELF from lichens belonging to different genera, with different photobiont types, and from different ecoregions were compared using ANOVA with a Tukey's multiple comparison test as a post-hoc test. Community structures were analyzed and visualized using non-metric multidimensional scaling (NMDS) based on binary Jaccard distance with the phyloseq package [69]. The samples with more than two ELF species were used for the community analysis. The influence of the lichen host taxonomy, photobiont types, and ecoregion effect on the community structures were tested using Adonis with the vegan package [70] and pairwise Adonis with the pairwiseAdonis package [71]. Barplots and scatter plots were drawn using the ggplot2 package [72]. Indicator species analysis was performed using the vegan package to detect the ELF species showing group-specific distribution. Only the ELF species with more than five occurrences in the total dataset were accepted as an indicator species. Network analysis was conducted to detect the host preference of the ELF species that have significantly higher interactions with a specific host lichen in comparison to that of the null networks (n = 100) using an econullnetr [73]. Among the significant results from the network analysis, we excluded the results of ELF species isolated from the low number of lichens (<5 specimens). A bipartite network plot was drawn using a bipartite package [74].

Lichen Diversity
The 79 lichen specimens collected from 29 oreums in Jeju Island (Table A1) were identified morphologically as belonging to three orders (Caliciales, Lecanorales, and Peltigerales), eight families, 15 genera, and 44 species. The number of specimens was largest in Parmeliaceae (30 specimens), followed by Stereocaulaceae (12) and Physciaceae (10). At the genus level, Parmotrema was most frequently collected (18 specimens), followed by Stereocaulon (12) and Myelochroa (10). At the species level, Stereocaulon japonicum was the most common species (12 specimens), followed by Parmotrema tinctorum (6) and Myelochroa entotheiochroa (5). Most species (38/44 species) were collected only once or twice. Three photobiont types were characteristic for the collected lichens: green algae, cyanobacteria, and both of green algae and cyanobacteria (Table S2 (Supplementary Materials)). All specimens belonging to Caliciales and Lecanorales (except for Stereocaulon japonicum) as well as Lobaria discolor and Lobaria japonica (Peltigerales) had a green algal photobiont. All other specimens belonging to Peltigerales had a cyanobacterial photobiont, except for Peltigera leucophlebia. Stereocaulon japonicum and Peltigera leucophlebia had a photobiont consisting both of green algae and cyanobacteria.

Ecoregion Clustering
The sampling sites were clustered to three ecoregions using PAM based on environmental factors (elevation and six bioclimatic data) (Figure 2a). Ecoregion G1 is characterized by high elevation and precipitation (BIO12, BIO14, and BIO15) ( Table 1). Ecoregion G2 generally had a moderate level of environmental values, except for high isothermality (BIO03) and annual range of temperature (BIO07). Ecoregion G3 had a low elevation with high mean temperature (BIO10). The distribution of lichen specimens varied among the ecoregions ( Figure 2b). The genera belonging to Peltigerales (e.g., Collema, Leptogium, and Peltigera) were abundant in the G1 ecoregion, while most genera in Lecanorales (e.g., Cladonia, Myelochroa, Parmotrema, and Stereocaulon) were abundant in the G2 and G3 ecoregions. In the G2 ecoregion, Cladonia and Stereocaulon were abundant, while Parmotrema and Physcia were abundant in the G3 ecoregion. only once or twice. Three photobiont types were characteristic for the collected lichens: green algae, cyanobacteria, and both of green algae and cyanobacteria (Table S2 (Supplementary Materials)). All specimens belonging to Caliciales and Lecanorales (except for Stereocaulon japonicum) as well as Lobaria discolor and Lobaria japonica (Peltigerales) had a green algal photobiont. All other specimens belonging to Peltigerales had a cyanobacterial photobiont, except for Peltigera leucophlebia. Stereocaulon japonicum and Peltigera leucophlebia had a photobiont consisting both of green algae and cyanobacteria.

Ecoregion Clustering
The sampling sites were clustered to three ecoregions using PAM based on environmental factors (elevation and six bioclimatic data) (Figure 2a). Ecoregion G1 is characterized by high elevation and precipitation (BIO12, BIO14, and BIO15) ( Table 1). Ecoregion G2 generally had a moderate level of environmental values, except for high isothermality (BIO03) and annual range of temperature (BIO07). Ecoregion G3 had a low elevation with high mean temperature (BIO10). The distribution of lichen specimens varied among the ecoregions (Figure 2b). The genera belonging to Peltigerales (e.g., Collema, Leptogium, and Peltigera) were abundant in the G1 ecoregion, while most genera in Lecanorales (e.g., Cladonia, Myelochroa, Parmotrema, and Stereocaulon) were abundant in the G2 and G3 ecoregions. In the G2 ecoregion, Cladonia and Stereocaulon were abundant, while Parmotrema and Physcia were abundant in the G3 ecoregion.
The ELF richness was compared between the host genera, photobiont types, and ecoregions ( Table 2). The richness was significantly different between the host genera (p = 0.017), photobiont types (p = 0.024) and ecoregions (p = 0.006). The only pairwise comparison showed significant difference: ELF richness in Stereocaulon was significantly lower than that of Parmotrema (p = 0.023). For the photobiont type, the lichens with both green algae and cyanobacteria had significantly lower richness compared to that with green algae only (p = 0.019). For the ecoregions, G3 contained significantly higher ELF richness compared to G2 (p = 0.006).  The ELF richness was compared between the host genera, photobiont types, and ecoregions ( Table 2). The richness was significantly different between the host genera (p = 0.017), photobiont types (p = 0.024) and ecoregions (p = 0.006). The only pairwise comparison showed significant difference: ELF richness in Stereocaulon was significantly lower than that of Parmotrema (p = 0.023). For the photobiont type, the lichens with both green algae and cyanobacteria had significantly lower

Structure of ELF Communities
Ordination analysis based on binary Jaccard dissimilarities revealed separation of communities according to the lichen taxonomy, photobiont, and ecoregions ( Figure 4). Adonis analysis showed a significant effect of the host genera (R 2 = 0.179, p = 0.001), photobiont types (R 2 = 0.066, p = 0.001), and ecoregions (R 2 = 0.073, p = 0.001). For lichen genera, several pairwise comparisons, including Collema, Leptogium, Parmotrema, and Peltigera, showed significant differences (Table 3). All pairs in the photobiont types showed significant differences and ecoregion G1 was significantly different from G2 and G3. Variation partitioning analysis showed that the effects of the photobiont type and ecoregion were confounded to that of the host genera, explaining 5.83% of the total variation ( Figure S3  (Supplementary Materials)). Only the effect of host genera had unique explanatory power, even after removing the other effects.    The indicator species analysis detected ELF species associated with definite lichen genera, photobiont types, and ecoregions (Table 4). Sordaria cf. fimicola was chosen as an indicator species for both Caliciales (Phaeophyscia and Physcia) and Lecanorales (Myelochroa, Parmotrema, and Stereocaulon). For the photobiont type, four indicator species were detected and the type of green algae and cyanobacteria had the highest number of indicator species (Hypoxylon fragiforme, Sordaria cf. fimicola, and Nigrospora chinensis). Seven indicator species were detected for the ecoregions. Network analysis showed similar results with indicator species analysis, in that five ELF species occurred with significantly higher frequency in specific lichen host compared to null networks simulated by the same number of interactions ( Figure 5). Biscogniauxia petrensis and Nigrospora aurantiaca favored Parmotrema, Hypoxylon fragiforme and Sordaria cf. fimicola preferred Stereocaulon, and Daldinia eschscholtzii favored Leptogium.

Discussion
On Jeju Island, diverse ELF biota comprising 112 species was isolated from 79 lichen specimens. These ELF biota can be considered as one of the most numerous found at the regional level. Most studies identified approximately 20-60 species at the regional scale [13,16,17,22,28,30,39,75,76]. At the continental scale, Arnold et al. [9] isolated approximately 200 species, which was the highest number of ELF species found in a single study. Most ELF species isolated in our study belonged to Ascomycota, which coincides with the previous studies showing the dominance of Ascomycetes among ELF species [9,14,22,30]. Among the Ascomycota, Sordariomycetes had the highest number of ELF species (71.8%), followed by Dothideomycetes (13.6%). Some previous studies also showed a dominance of Sordariomycetes in ELF species (in China, France, Japan, North America, Norway, and Sri Lanka) [9,22,28,31,39,75]. By contrast, other studies showed a dominance of Dothideomycetes, Leotiomycetes or Pezizomycetes (in Antarctica, Germany, Italy, and USA) [13,14,27,77]. However, there is no specific pattern of ELF dominance depending on the country, continent, or climate zone. The species of Daldinia, Sordaria, and Nigrospora were the most frequently detected ELF species in Jeju Island (Figure 3), which were frequently isolated from lichens in the previous studies [17,28,30,39,75]. These genera are well-known wood-decaying fungi (Daldinia, Hypoxylon, and Xylaria) and soil-borne saprotroph (Chaetomium, Nigrospora, and Trichoderma). However, lichens collected in this study did not show any symptom of decomposition or disease, which suggests that many ELF species have a saprotrophic lifestyle but it is inactive within lichen thallus. Similar phenomena have been found from endophytic fungi in plants; Many endophytic fungi in leaves turn into saprotrophic fungi when leaves fall on the ground [78]. Considering a taxonomic similarity between ELF and endophytic fungi [79,80], ELF may be a latent saprotroph, like some endophytic fungi, waiting for a suitable environment where they start to decompose. In addition, some ELF are known as coprophilous fungi (Sordaria and Preussia) living in the feces [81,82]. Given that feces are a nitrogen-rich environment, coprophilous fungi may prefer lichens containing cyanobacteria because they can fix nitrogens from

Discussion
On Jeju Island, diverse ELF biota comprising 112 species was isolated from 79 lichen specimens. These ELF biota can be considered as one of the most numerous found at the regional level. Most studies identified approximately 20-60 species at the regional scale [13,16,17,22,28,30,39,75,76]. At the continental scale, Arnold et al. [9] isolated approximately 200 species, which was the highest number of ELF species found in a single study. Most ELF species isolated in our study belonged to Ascomycota, which coincides with the previous studies showing the dominance of Ascomycetes among ELF species [9,14,22,30]. Among the Ascomycota, Sordariomycetes had the highest number of ELF species (71.8%), followed by Dothideomycetes (13.6%). Some previous studies also showed a dominance of Sordariomycetes in ELF species (in China, France, Japan, North America, Norway, and Sri Lanka) [9,22,28,31,39,75]. By contrast, other studies showed a dominance of Dothideomycetes, Leotiomycetes or Pezizomycetes (in Antarctica, Germany, Italy, and USA) [13,14,27,77]. However, there is no specific pattern of ELF dominance depending on the country, continent, or climate zone. The species of Daldinia, Sordaria, and Nigrospora were the most frequently detected ELF species in Jeju Island (Figure 3), which were frequently isolated from lichens in the previous studies [17,28,30,39,75]. These genera are well-known wood-decaying fungi (Daldinia, Hypoxylon, and Xylaria) and soil-borne saprotroph (Chaetomium, Nigrospora, and Trichoderma). However, lichens collected in this study did not show any symptom of decomposition or disease, which suggests that many ELF species have a saprotrophic lifestyle but it is inactive within lichen thallus. Similar phenomena have been found from endophytic fungi in plants; Many endophytic fungi in leaves turn into saprotrophic fungi when leaves fall on the ground [78]. Considering a taxonomic similarity between ELF and endophytic fungi [79,80], ELF may be a latent saprotroph, like some endophytic fungi, waiting for a suitable environment where they start to decompose. In addition, some ELF are known as coprophilous fungi (Sordaria and Preussia) living in the feces [81,82]. Given that feces are a nitrogen-rich environment, coprophilous fungi may prefer lichens containing cyanobacteria because they can fix nitrogens from the atmosphere [83]. Abundance of melanin-producing fungi (e.g., Daldinia, Sordaria, and Xylaria) was another important feature of ELF communities isolated in this study [84][85][86]. In a harsh environment, melanin-producing fungi are abundant because melanin has a protective role against abiotic stress such as UV radiation, extreme pH or temperature, and drought [87]. Therefore, the melanin-producing ability of ELF can be the adaptive feature for survival of lichens in a harsh environment. In addition, the dominance of Daldinia childiae in lichens may be associated with its adaptation to a hostile environment. Genome sequence of Daldinia eschscholtzii showed that carbon assimilation in nutrient-limited conditions (acid trehalases) and the heat stress response gene (ATP-dependent molecular chaperone) are core gene families in this species [88].
The patterns of ELF richness and community structure were significantly influenced by the host taxonomy, photobiont types, and ecoregions, while their effects were mixed together (Table 2; Figure 4 and Figure S3 (Supplementary Materials)). The effects of the photobiont types and ecoregions were confounded to the host effect and the unique effect on community structure was found only for the lichen taxonomy ( Figure S3 (Supplementary Materials)). Generally, the photobiont type is specific for a specific lichen taxon, which explains why the effect of the photobiont types was confounded to the effect of the lichen taxonomy. For the effect of the ecoregions, the distribution of lichens varied among the ecoregions (Figure 2b), which led to a mixed effect of the host taxonomy and ecoregions. In the G1 ecoregion (high elevation and precipitation), Peltigerales species were abundant; Cladonia and Stereocaulon were abundant in G2 (moderate elevation and climate with wide temperature range), and Parmotrema and Physcia were abundant in G3 (lowlands with high average temperature) [41].
The ELF richness was comparable among the lichen genera, except for Stereocaulon and Parmotrema (Table 2); Myelochroa and Parmotrema had high ELF richness, while Stereocaulon had significantly low ELF richness. The high richness of ELF in Myechora and Parmotrema may be associated with the elevation in which they were frequently collected (lowlands) (Table A1), because for endophytic fungi, their richness increased in the low elevation region [35,89]. By contrast, the lowest richness of ELF in Stereocaulon may be due to its morphological characteristics. Among the lichens we collected, Sterocaulon was the only genus with fruticose thallus [90]. Because fruticose lichens are generally narrow and thin compared to foliose lichens, their thallus harbors lower number of ELF species than foliose lichens. The ELF community structures were significantly influenced by the taxonomy of host lichens ( Figure 4; Table 3). Previous studies also showed a difference of ELF communities depending on the lichen taxonomy [14]. Although it was significant, only a small part of the variations was explained by the host genus (6%). In addition, frequently detected ELF have numerous hosts showing host generality, which suggests that the host generality in the overall ELF communities is stronger than the host specificity. Chagnon et al. [26] also found that ELF showed higher host generality compared to endophytic fungi living in the plants. The low level of host specificity may be due to the harsh environment of the lichen thallus. Generally, lichens are poikilohydric organisms that frequently experience alternating dry and wet conditions [91] and damages by the reactive oxygen species (ROS) [92], which can be a harsh condition for ELF living inside of the thallus. In a harsh environment, the fungal community showed low host specificity. For example, the host specificity of fungal symbionts in plants (e.g., endophytes, ectomycorrhizal, or ericoid fungi) was decreased in alpine and arctic environments [93][94][95][96].
A previous study showed that the location of ELF in the lichen thallus is at the photobiont layer [9] and ELF can acquire nutrients from the photobiont [8], which suggests that the photobiont type can be important to the ELF composition in lichens. In line with this assumption, the photobiont type significantly influenced the ELF richness and communities' composition. The lichens with two kinds of photobionts showed significantly low ELF richness (Table 2). In addition, ELF community structures were also significantly different between lichens with different photobiont type (Figure 4; Table 3). Peltigerales and Stereocaulon are different compared to other lichens in terms of the photobiont type. Most Peltigerales species have cyanobacteria (e.g., Nostoc) as the primary photobiont, whereas Stereocaulon has green algae as the primary photobiont with cyanobacteria as the secondary photobiont [90]. Cyanobacteria can fix atmospheric nitrogen; thus, the total nitrogen concentration in the thallus was higher in these lichens [83], which can influence the ELF community and be attractive to coprophilous fungi such as Sordaria cf. fimicola (Table 4; Figure 5).
The ecoregions classified by the elevation and bioclimatic variables were one of the factors influencing the ELF richness and community structures (Table 2; Figure 4). The richness in the G3 ecoregion (high mean temperatures in lowlands) was significantly higher than that in the other ecoregions, which agrees with a previous observation that the ELF richness was increased following the length of the growing season that is associated with the annual temperature [15]. The ecoregion effect on the community structures was significant but totally confounded to the host effects; none of the explanatory power was unique for the ecoregions. Generally, endophytic fungal communities also showed variations according to the variations in such environmental factors as elevation, annual precipitation, and temperature [35,[97][98][99]. The lack of unique environmental patterns in this study may have arisen because the regional scale of environmental variations may be insufficient to reveal substantial variations in ELF diversity. In addition, elevation did not generally influence the presence or absence of ELF species but affected their abundance [100]; meanwhile, our study focused on the presence and absence of ELF species.
Although diverse ELF biota were isolated in our study, this study has some fundamental limitations that raise caution in interpreting the study results. First, we used a single method for isolation, which limited the number of isolated ELF species. Recent studies have shown that ELF diversity can vary depending on the isolation protocol [18,75]. Various media constituting the different nutrient compositions and different sterilization methods can expand the ELF diversity. Second, this study only covered cultivatable ELF. Given that many culture-independent studies have revealed a deeper fungal diversity [32,101], the culture-dependent approach may not have detected the whole ELF diversity within the lichen. A recent metabarcoding study for ELF diversity in Arctic and Antarctic regions showed higher ELF diversity [21,22], which suggests that the ELF diversity in Jeju Island may also exceed the diversity revealed in our study. Finally, the lichen taxonomy as a factor influencing the ELF communities was analyzed at the genus level due to the insufficient number of specimens. The low resolution of the lichen taxonomy may have ignored the ecological patterns at a fine level. Further studies with specific hypothesis and experimental designs are needed. Despite these study limitations, the culture-dependent approach can reveal remarkable diversity and consistent patterns of ELF communities. In addition, obtaining ELF isolates is important considering the usefulness of ELF as a potential source of natural products. Thus, understanding the ecological pattern of ELF isolates is meaningful to obtain various ELF species that can have potential experimental or industrial applications.

Conclusions
A high diversity of ELF species was identified from the lichens in Jeju Island. The most common ELF species belonged to Sordariomycetes, which is similar to the ELF diversity reported in other countries. In addition, many species were newly identified as ELF species in South Korea. The ELF richness and community composition were significantly influenced by the combination of host characteristics and environmental factors. Our results suggest that in order to reveal a higher diversity of ELF species, we need to collect a higher variety of lichen species from different environments and to combined the culture-based methodology with a culture-independent approach such as a metabarcoding method.
Supplementary Materials: The following are available online at http://www.mdpi.com/2071-1050/12/9/3769/s1. Figure S1: Phylogenetic tree based on ML analysis for (A) Xylariales and (B) other Sordariomycetes; Figure S2: Phylogenetic tree based on ML analysis for (A) other Ascomycota (non-Sordariomycetes) and (B) Basidiomycota and Mucoromycta; Figure S3: Variation partitioning plot for ELF communities based on binary Jaccard dissimilarities. The adjusted explanatory powers are presented for each factor; Table S1: Information about sampling locations; Table S2: Information of the ELF species isolated from this study.