Phylogenetic relationships of rock-inhabiting black fungi belonging to the widespread genera Lichenothelia and Saxomyces

ABSTRACT Rock-inhabiting fungi (RIF) are adapted to thrive in oligotrophic environments and to survive under conditions of abiotic stress. Under these circumstances, they form biocoenoses with other tolerant organisms, such as lichens, or with less specific phototrophic consortia of aerial algae or cyanobacteria. RIF are phylogenetically diverse, and their plastic morphological characters hamper the straightforward species delimitation of many taxa. Here, we present a phylogenetic study of two RIF genera, Lichenothelia and Saxomyces. Representatives of both genera inhabit rather similar niches on rocks, but their phylogenetic relationships are unknown so far. The cosmopolitan genus Lichenothelia is recognized by characters of fertile ascomata and includes species with different life strategies. In contrast, Saxomyces species were described exclusively by mycelial characters found in cultured isolates from rock samples collected at high alpine elevations. Here, we use an extended taxon sampling of Dothideomycetes to study the phylogenetic relationships of both Lichenothelia and Saxomyces. We consider environmental samples, type species, and cultured isolates of both genera and demonstrate their paraphyly, as well as the occurrence of teleomorphs in Saxomyces. We applied three species delimitation methods to improve species recognition based on molecular data. We show the distinctiveness of the two main lineages of Lichenothelia (Lichenotheliales s. str.) and Saxomyces and discuss differences in species delimitation depending on molecular markers or methods. We revise the taxonomy of the two genera and describe three new taxa, Lichenothelia papilliformis, L. muriformis, and Saxomyces americanus, and the teleomorph of S. penninicus.


INTRODUCTION
Natural and anthropogenic lithic substrates are colonized frequently by black, rock-inhabiting fungi (RIF; Sterflinger 2006). RIF are cosmopolitan and known from a wide range of habitats that are hardly colonized by any other eukaryotic life forms, including both hot and cold deserts (Friedmann 1982;Staley et al. 1982;Onofri et al. 1999). Oligotrophic RIF seem to outcompete other fungi in part via their survival capacities, including a pronounced desiccation tolerance and resistance to diverse environmental stresses (Staley et al. 1982;Sterflinger 2006;Gorbushina 2007;Onofri et al. 2008;Gostincar et al. 2012;Selbmann et al. 2015).
Rock-inhabiting fungi represent a complex of polyphyletic taxa that evolved early and independently in the Dothideomycetes and Eurotiomycetes (Gueidan et al. 2008(Gueidan et al. , 2011. Close relatives of RIF are taxa that share similar polyextremotolerant features but have different lifestyles (Gostincar et al. 2011(Gostincar et al. , 2012Grube et al. 2013). They include plant and human pathogens, lichenicolous fungi, or fungi that form lichen-like structures with algae (Ruibal et al. 2005(Ruibal et al. , 2009Brunauer et al. 2007;Harutyunyan et al. 2008;Muggia et al. 2008Muggia et al. , 2013Muggia et al. , 2016Selbmann et al. 2005Selbmann et al. , 2013. Even if RIF occurring directly on exposed rocks do not necessarily form associations with algae, they may be able to do so in culture (Gorbushina et al. 2005;Gorbushina and Broughton 2009;Ametrano et al. 2017). These findings prompted the authors to speculate about phylogenetic links between RIF and lichenized fungi (Gostincar et al. 2012).
Notwithstanding the overall similarities in morphological characters, RIF present subtle variation in mycelial or microcolonial forms that are consistent with their genetic diversity (Ruibal et al. 2005(Ruibal et al. , 2009Muggia et al. 2015). This variation may be induced by spatial isolation or local environmental pressures that contribute to adaptive radiation , which complicates species delimitation (a cryptic species phenomenon). This also renders it difficult to assess the variation of species described in the past, especially in lineages represented by only few specimens, e.g., in the genus Lichenothelia (Henssen 1987). More recent descriptions of RIF species are based on cultured strains (e.g., Vermiconia, Egidi et al. 2014;Saxomyces, Selbmann et al. 2014). Nucleic sequence data help to identify them as species and to place them in larger phylogenetic frameworks. Such studies, using a broad taxon sampling (e.g., in Dothideomycetes) may also uncover lineages of RIF at higher taxonomic levels (e.g., orders) Ruibal et al. 2009;Schoch et al. 2009;Muggia et al. , 2015.
Here, we focused on two genera of RIF, i.e., Lichenothelia and Saxomyces, which have originally been described according two different approaches. Since their description, isolates obtained in culture and genetic sequence data have enabled further comparisons of their morphological traits and genetic diversity. Lichenothelia comprises 28 taxa (according to MycoBank, January 2018) and represents a group of usually fertile and globally distributed fungi. Lichenothelia species generally have been described by morphological characters studied in environmental samples (Hawksworth 1981;Henssen 1987;Atienza and Hawksworth 2008;Zhurbenko 2008;Etayo 2010;Muggia et al. 2015;Valadbeigi et al. 2016). These studies revealed Lichenothelia as a genus with diverse life strategies. Species may inhabit rocks or lichens or grow loosely associated with algae as borderline lichens (Henssen 1987;Muggia et al. 2015). For this reason, the genus was hypothesized to represent a link between rock-inhabiting and lichenized fungi (Hawksworth 1981;Muggia et al. 2013). Because of its versatile lifestyle, Lichenothelia also has been included in lichen surveys either as lichenized fungi (Etayo 2010) or as lichen parasites (Kocourková and Knudsen 2009). Muggia et al. (2015) recently focused on Lichenothelia species distributed in desert regions in California (USA) and characterized five species, revising the Lichenothelia species concept introduced by Henssen (1987) in the first survey of the genus. Saxomyces, on the other hand, has been isolated from rock samples from high altitudes in the Alps and was described exclusively from cultured isolates as anamorphic hyphomycetes for which only conidiophores and conidiospores were observed ).
We present a broad taxon sampling of Dothideomycetes in which we consider environmental samples and culture isolates of Lichenothelia and Saxomyces collected in multiple localities and including the types of Saxomyces and of some Lichenothelia species. We discuss their phylogenetic relationships, their identity, and their taxonomic treatment, aiming at (i) resolving in more detail (at species level) and improving phylogenetic support for the monophyletic lineage Lichenotheliaceae/Lichenotheliales as identified previously in Hyde et al. (2013) and Muggia et al. (2013Muggia et al. ( , 2015; (ii) testing how a morphologically based approach of species recognition reconciles with a phylogenetic species concept and species delimitation in Lichenothelia and Saxomyces; and (iii) evaluating the relationships of Saxomyces and Lichenothelia and revising their taxonomy in accordance to their anamorphic and teleomorphic states.

MATERIALS AND METHODS
Sampling.-Samples of Lichenothelia and Saxomyces were collected in seven localities across Europe (Italy and Czech Republic) and USA (California) in the period 2009-2016 (TABLE 1). The samples include thalli growing on different substrates and under different ecological conditions. Specimens were stored at room temperature until processing. A total of 49 samples were considered for morphological analyses. Of these, 47 were selected for molecular analyses, and of these 47 ten were further selected for culture isolations. The samples are stored in GZU, UCR, and the private Herbarium Mycologicum Kocourková & Knudsen (Hb. K & K) collections, with some duplicates distributed to other herbaria (B, H, KRAM, NY, PRM, UPS).

Morphological
analyses.-We analyzed the morphology of all samples and segregated species also based on the monophyly that the samples showed in the phylogenetic analyses. The descriptions of the two genera are not reported here, as they correspond to the original descriptions of Lichenothelia (and Lichenotheliaceae) by Hennsen (1986) and Hyde et al. (2013) and of Saxomyces by Selbmann et al. (2014), respectively. The following morphoanatomical traits were analyzed in the environmental samples and used for species delimitation as in Muggia et al. (2015): fertile stromata stipitate or not; presence or absence of slender interascal filaments; ascospore size and septation; and morphology of the thallus, especially of  superficial hyphae. Structures of samples were studied in water and 10% KOH (K). Amyloid reactions were tested in fresh and undiluted Lugol's iodine (I; Merck 109261; Sigma-Aldrich, Germany) without pretreatment with K, and ascus stains were studied with I with or without pretreatment with K. Ascospore measurements were made in water with an accuracy of 0.5 µm and given as (min.-)X1-X2-X3(max.), where min./max. are the extremes from all measurements, X1 is the lowest arithmetic mean observed for a specimen, X2 is the arithmetic mean of all observations, and X3 is the highest arithmetic mean observed for a specimen. They are followed by the number of measurements (n). The length/breadth ratio of ascospore is indicated as l/b and given in the same form.
Macrophotographs of environmental samples were taken with digital cameras (  Morphological analyses of the cultured strains were performed on 6-10-mo-old cultures. We considered the following characters as in Egidi et al. (2014), , and Muggia et al. (2015): form of growth, filamentous vs. yeast-like; branching of the hyphae; hyphal maturation and degree of melanization; and conidiogenesis. Small fragments of the mycelium were taken, and squashed sections were mounted in water. Images were acquired with a ZeissAxioCam MRc5 digital camera (Carl Zeiss AG, Oberkochen, Germany) fitted to the microscopes. Images of growth habit and hyphal structure were digitally optimized using the CombineZM software (open source image processing software available at www.hadle yweb.pwp.blueyonder.co.uk/). The figures were prepared with CorelDRAW X4 (Corel Corp.,Ottawa, Canada).
Culture isolation.-Fungal strains were isolated from environmental samples under a sterile hood. Several washing steps with a 1:10 dilution of Tween 80 solution and sterile deionized water were performed on a magnetic stirrer to decrease possible external contamination by bacteria and yeast (Bubrick and Galun 1986; 15 min washing in deionized sterile water, 15 min washing in 1:10 Tween 80 solution, and 5 min rinse in deionized sterile water). A final washing step was performed by pipetting the 1:10 Tween 80 solution directly on the selected area of the thallus and finally washing by pipetting with deionized sterile water. Up to 15 inocula were prepared for each sample on Bold basal medium (BBM; Bold 1949) amended with ampicillin. When specimens were fertile, the ascomata were dissected and the hymenial parts were used to inoculate the medium. For sterile specimens, pieces as tiny as possible of vegetative black hyphae were picked. Agar plates were incubated in a growth chamber at 20 C with a dark-light regime of 12:12 h. After 3 mo, inocula were subcultured on malt yeast medium (MY; Ahmadjian 1967). After 1-2 mo, a fragment of the subculture was taken for genetic identification, morphological analyses, and cryostock preparations. The cultured strains are maintained at the University of Graz in the culture collection Cultures of Lichens and Extremotolerant Organisms (CLEO) established by L.M., at the public culture collections of Culture Collection of Fungi from Extreme Environments (CCFEE) at the University of Viterbo, and at the Mycotheca Universitatis Taurinensis (MUT) at the University of Turin (Italy). Cultured strains are referenced as their DNA extraction number (culture identification numbers [LMCCxxxx] are reported in TABLE 1). The identity of the cultures was checked by sequencing the nuclear and mitochondrial loci (nuclear 28S and 18S rDNA, mitochondrial 12S rDNA) selected for the environmental samples.
DNA extraction, amplification, and sequencing.-The fungal material used for DNA extraction and culture isolation was taken from a single area of the mycelium and transferred into a 1.5-mL tube. DNA extractions were performed using either a small group of ascomata or, if these were rare or absent, about 0.5 cm 2 of the dry crustose, melanized thallus. The DNA of pure cultures and environmental samples was extracted according to the cetyltrimethylammonium bromide (CTAB) method (Maniatis et al. 1982) with some minor modifications. Fungal material from the axenic cultures was dried overnight in silica before grinding in liquid nitrogen. Quality and amount of the extracted DNA were checked by Nanodrop (Thermo Fischer Scientific Inc., Waltham, MA, USA). Polymerase chain reaction (PCR) amplifications were carried out following a touch-down protocol as in Muggia et al. (2013Muggia et al. ( , 2015. The nuclear ribosomal 28S (nucLSU [large subunit]) locus was amplified with primers LR3R and LR7 (Vilgalys and Hester 1990). The nuclear 18S (nucSSU [small subunit]) was amplified with primers NS1 (White et al. 1990) and nuSSU0852 (Gargas and Taylor 1992). The mitochondrial 12S (mtSSU) locus was amplified with primers mtSSU1KL (Lohtander et al. 2002) and MSU7 (Zhou and Stanosz 2001). PCR products were cleaned using Ezna Cycle Pure Kit (Omega Bio-tek Inc., Norcross, GA, USA) and sequencing was performed by Microsynth (Balgach, Austria). Amplification of the nuclear internal transcribed spacers and 5.8S gene (ITS), as the standard fungal DNA barcode (Schoch et al. 2012), was performed with primers ITF1F (Gardens and Bruns 1993) and ITS4 (White et al. 1990). However, most samples yielded multiple PCR products and were not sequenced successfully. Therefore, we exluded the ITS locus from the analyses.
Alignment and phylogenetic analyses.-Newly obtained sequences of Lichenothelia and Saxomyces sequences were added to a broad data set including representatives of most of the orders of Dothideomycetes (available up to August 2017) and previous data from Muggia et al. (2013Muggia et al. ( , 2015. As Arthoniomycetes are considered the sister group of Dothideomycetes according to recent phylogenetic inferences (Schoch et al. 2009;Egidi et al. 2014;Ertz et al. 2014), four species belonging to Arthoniales (Dendrographa leucophaea, Lecanactis abietina, Schismatomma decolorans, and Roccella fuciformis) were used as outgroups. Single-locus alignments were prepared in BioEdit 7.2.5 (Hall 1999). The multilocus alignment was prepared using Sequence Matrix 1.7.8. and partitioned by locus. jModelTest 2.1.10 (Darriba et al. 2012) was used to assess the best model of nucleotide substitution for each gene via the corrected Akaike information criterion (cAIC; Akaike 1974).
Both Bayesian and maximum likelihood (ML) approaches were used to reconstruct phylogenetic inferences. Bayesian phylogenies (single-and multilocus) were generated in MrBayes 3.2.6 on the CIPRES Science gateway 3.3 (Miller et al. 2010). The Bayesian/Markov chain Monte Carlo (B/MCMC) analyses were run with six chains simultaneously in two runs for 2 × 10 7 generations, and trees were sampled every 100 generations. Log-likelihood scores against generation time were plotted with Tracer 1.6 (Rambaut and Drummond 2007) and effective sample sizes (ESS) belonging to all parameters of the substitution models were checked. Burn-in was set at 2.5 × 10 6 generations to ensure that likelihood stationarity was reached. The consensus tree was calculated from the sampled trees. Convergence of the analyses was confirmed by the potential scale reduction factor (PSRF) approaching 1. Three independent runs of the same analysis were conducted. The ML phylogenies were generated on a local machine in RAxML 7.0.3 (Stamatakis et al. 2005;Stamatakis 2006). The analysis was performed using the GTRMIX model combined with 1000 bootstrap replicates. ViPhy 1.3.1 (Bremm et al. 2011) was used to compare tree topologies and highlight differences in sample composition of clusters. Tree topologies were compared via both the leaf-based and the element-based algorithms, which assign a similarity score (0-1) to tree nodes depending on the consistency of the trees relative to a user-defined reference tree. The leaf-based method only considers consistency of tips below a certain node, whereas the element-based method considers both tips and inner nodes to calculate the similarity score. The "best match" selection option was then used to highlight samples that were always within a certain clade (either Lichenotheliales s. str. or Saxomyces) in every phylogeny compared with the multilocus tree.  TABLE 1;  SUPPLEMENTARY TABLE 1). Each approach was performed on the three single-locus data sets, mtSSU, nucLSU, nucSSU, which contained a variable number of sequences (31-49) due to the different number of successfully sequenced loci for each sample. Further, to test whether outgroups influence species delimitation, the analyses were performed with two different outgroups. First, representatives of Arthoniales were selected (as reported above); second, the most basal lineages of the Lichenotheliales s. str. and Saxomyces clades were chosen. The latter two, as reported below, corresponded to the clade Cryomyces for Lichenotheliales s. str. and representatives of Myriangiales (Myriangium duriaei, Myriangiales sp. A554, and Myriangiales sp. A578) for the Saxomyces clade. Each distance matrix or phylogeny used for species delimitation analyses was generated also with a reduced data set, which only includes samples with three loci in order to avoid the different amounts of missing data, which could affect the number of delimited ESU.  ABGD (http://wwwabi.snv.jussieu.fr/public/abgd/ abgdweb.html) was applied to the distance matrix obtained from the data set alignments. This matrix was generated in MEGA7 with the following settings: Tamura-Nei distance, gamma distributed rates among sites, gamma shape parameter according to jModelTest 2.1.10, and pairwise deletion of missing data to retain the maximum amount of information (due to different length of the sequences).
PTP and GMYC were tested as single-and multirate PTP and single-and multithreshold GMYC. The analyses are identified as sPTP, mPTP, sGMYC, and mGMYC, respectively. The Bayesian GMYC (bGMYC) was further implemented, as it uses a random subset of posterior trees (after burn-in) generated by BEAST (Bayesian Evolutionary Analysis Sampling Trees; Rambaut and Drummond 2007). PTP analyses were performed on the corresponding Web server (http://species.h-its.org/; http:// mptp.h-its.org/), whereas GMYC analyses were performed in R (SPLITS and bGMYC packages). The input data were phylogenies generated using BEAST and MrBayes. Phylogenetic trees inferred by MrBayes were not ultrametric and therefore were used only in PTP. These phylogenies have been additionally smoothed using the R package APE 4.0 with the function "chronos" (Paradis et al. 2004), with λ = 0, model = "relaxed," in order to use them also in GYMC as well (as "chronos"-smoothed phylogenies were used as a third input data set for both PTP and GMYC analyses). Substitution models were set according to the cAIC criterion. Maximum clade credibility trees were obtained by three to four runs of 10 7 generations in BEAST and two runs of 5 × 10 6 generations in MrBayes.

Morphological analyses of environmental samples.
-According to morphological analyses, we recognize three new taxa, which are formally described as Lichenothelia muriformis, L. papilliformis, and Saxomyces americanus. We revise the circumscription of the species Lichenothelia intermixa Henssen, provide the new combination of Morphological analyses of culture isolates.-Fewer than 10% of inocula remained uncontaminated during cultivation, reflecting the biological complexity and diversity present in environmental samples. Those that were subcultured successfully were used for molecular and morphological analyses. We recovered culture isolates in four clades (FIG. 1): Lichenostigmatales, Lichenotheliales s. str., Capnodiales, and Saxomyces group. The strains L2282-L2284 within Lichenotheliales; L2291, L2293, and L2296 within Saxomyces group; and L2285 and L2289 within Lichenostigmatales were chosen for morphological inspection, as they developed a sufficient amount of mycelium. Strains L2282-L2284 represent Lichenothelia intermixta (FIG. 2a-k); they are characterized by a dense mycelium that abundantly secretes oil drops and by hyphae mainly constituted by isodiametric cells (FIG. 2c- (FIG. 2q).
Strains L2184 and L2299 are two isolates obtained from Lichenothelia dimelaenae infecting the lichen Dimelaena oreina JK8234. These cultures were checked for their genetic identity at the time the two inocula have grown. However, the inocula were preserved as cryostocks and not subcultured any longer. Due to the extremely slow growth of the replated cryostocks, these strains were not available for the morphological analyses to be included here.
Phylogenetic analyses.-A total of 113 new sequences were obtained. Four samples were represented only by a single locus, 21 samples by two loci, and 22 samples by all three markers.
jModelTest analyses set the best models for the gene partitions as follows: GTR+I+G for nucLSU, SYM+I+G for nucSSU, and HKY+I+G for mtSSU. All criteria gave the same result except the cAIC, which assigned the GTR+I+G model as the most suitable for each molecular marker. As AIC tends to overparameterize by picking models with more parameters than those strictly necessary (Davidson and MacKinnon 2004), models suggested by the other three decision criteria were set for the phylogenetic analyses.
Notwithstanding the low support obtained for the basal branches in Dothideomycetes, the phylogenetic inference (FIG. 1) is topologically congruent with previous analyses (Hyde et al. 2013;Wijayawardene et al. 2014;Liu et al. 2017). Orders and families are consistently monophyletic and highly supported. Samples representing both Lichenothelia and Saxomyces are recovered in distinct clades in both single-locus and multilocus analyses and are recognized as Lichenotheliales s. str., Lichenotheliales (1), and Saxomyces group. These three lineages present low support, hinting at a potentially undefined position within Dothideomycetes.
The Saxomyces lineage is recovered with a consistent composition of samples across the different MrBayes phylogenies. It splits into two clades with low support (FIG. 1;  SUPPLEMENTARY FIG. 1). The Lichenotheliales s. str. always resolve as monophyletic, but with low support in both Bayesian and ML analyses. The third, smaller clade Lichenotheliales (1) contains 11 samples. It is always identified in the Bayesian multilocus analyses (FIG. 1), but its monophyly is not supported by ML analyses.
Single-locus Bayesian phylogenies (SUPPLEMENTARY FIG. 2) were compared with the multilocus phylogeny (FIG. 1). Leaf-based phylogenetic inferences (SUPPLEM- ENTARY FIG. 3) have a higher similarity score than element-based ones (SUPPLEMENTARY FIG. 4), as clades are similar in sample composition but differ substantially in the internal topology of the subtrees. Both Lichenotheliales s. str. and Saxomyces group lineages are recognizable. The nucSSU-based phylogeny (SUPPLEMENTARY FIG. 3d) is the most consistent considering the Saxomyces clade. Conversely, the nucLSU-based phylogeny (SUPPLEM- ENTARY FIG. 3c), and to a lesser extent the mtSSU-  fungus (a, b). c-h, j, k. Hyphae are mainly constituted by isodiametric cells; abundant branching (arrow in d) is observed in younger parts of the colonies where hyphae are still rather hyaline (c-f). g, j, k. Mature parts of the colonies are highly melanized and mainly composed of dense agglomerates of isodiametric cells. l-s. Strains of Saxomyces americanus (Saxomyces group): L2291 (l-o) and L2293 (p-s); habitus of L2291 (l) and L2293 (p). m-o, q-s. Hyphae are overall highly melanized, composed of elongated (m-o, r, s) and also isodiametric cells, which form dense agglomerates (q) in older parts of the colony. t-w. Strains of Lichenostigmatales: L2285 (t, u) and L2289 (v, w); fungi develop both filamentous hyphae and yeast morphs (arrow in t, v). w. Filamentous hyphae with slightly elongated cells are observed in between of the dense agglomerates of isodiametric, highly melanized yeast-like cells. Bars: a, p, v = 2 mm; b, t = 1 mm; c-h, j, k, m-o, q-s, w = 20 µm; i, u = 4 mm; l = 8 mm.
based phylogeny (SUPPLEMENTARY FIG. 3b), has a slightly different taxon composition than the reference clade in the multilocus phylogeny ( SUPPLEMENTARY  FIG. 3a). The mtSSU phylogeny (SUPPLEMENTARY FIG. 3f) is the most consistent for the Lichenotheliales clade and carries the greatest part of the phylogenetic information among the considered markers, likely influencing the final topology of the multilocus phylogeny.
Correspondence between environmental samples and their culture isolates was confirmed for samples L2218 and L2219 (7), L1609 (5), L1798 (1), L1799 (3), L2180 (4), L2220 and L2221 (6) (FIG. 1; TABLE 1). Isolates obtained from the samples L2156 (2) and L2212 (8) did not correspond to the Lichenothelia fungus sequenced from the environmental samples but were recovered in Capnodiales (FIG. 1). Isolates from L2216 and L2217 (9) belong to the Saxomyces clade, but fungi corresponding to the environmental samples are recovered in Lichenotheliales s. str. The two cultured strains L2184 and L2299 isolated from a sample of Lichenostigma dimelaenae (parasitic on the lichen Dimelaena oreina) are within Lichenotheliales s. str. as well, but no molecular data were gained from the environmental specimen for comparison. The two samples L1717 and L2198 recognized as L. arida in Muggia et al. (2015) are placed in the present phylogeny at the base of Saxomyces group and within Capnodiales, respectively (FIG. 1).
Species delimitation analyses.-A total of 18 analyses was carried out for each PTP method (single-and multirate): 12 analyses using each GMYC method (single-and multithreshold) and 6 analyses using ABGD (SUPPLEMENTARY TABLES 2 and 3). The distribution of the number of delimited ESU is shown by box and whisker plots ( SUPPLEMENTARY FIG. 5), both using the complete data set ( SUPPLEMENTARY  FIG. 5a, b) and using the reduced data set ( SUPPLEMENTARY FIG. 5c, d). The two data sets provided similar results when the delimitation methods are compared; however, the reduced data set provided fewer ESU because 42% of the samples are devoid of missing data.
Delimitations of the Lichenotheliales s. str. and Saxomyces clades for every analysis, and their average values with SD, are reported in SUPPLEMENTARY TABLES 2 and 3). ABGD did not find an adequate barcode gap to identify any cluster in Lichenotheliales s. str. (SUPPLEMENTARY FIG. 5a, c). The only exception is the clade Saxomyces based on the mtSSU locus ( SUPPLEMENTARY FIG. 5b, d). In both the Lichenotheliales s. str. and Saxomyces clades, sPTP and mGMYC tend to delimit more ESU than do mPTP and sGMYC (SUPPLEMENTARY FIG. 5). Furthermore, mPTP and sGMYC show a narrower distribution of the number of delimited ESU than sPTP and mGMYC.
Most ABGD analyses and some mPTP and sGMYC analyses were not effective: they did not produce any valid species delimitation in our data set. For the nuclear loci in Saxomyces (SUPPLEMENTARY FIG.  6d, f), most of the methods (ABGD, mPTP, sGMYC) tend to lump the data set whereas sPTP split them markedly (Dayrat 2005;Rittmayer and Austin 2012). Analyses on the Lichenotheliales s. str. clade (SUPPLEMENTARY FIG. 6a, c, e) confirm the difficulties to find any valid threshold useful to delimit ESU. In the case of nucLSU (SUPPLEMENTARY FIG. 6c), three of the five methods did not find any cluster. As observed for the Saxomyces clade, sPTP always tends to delimit many ESU. The two GMYC methods are the most effective for Lichenotheliales s. str. mtSSU (SUPPLEMENTARY FIG. 6a), with an almost identical ESU delimitation to each other and a conceivable number of delimited ESU. Among the applied methods, only GMYC provides also the statistically significance of delimitation results. The likelihood-ratio test (LRT) highlighted a significantly higher value than the null model (one species) for both Lichenotheliales s. str. and Saxomyces mtSSU BEAST trees (P < 0.05), when delimited with both the single-and multithreshold GMYC.
Ecology and distribution: On both hard and soft calcareous rock, at elevations from 1200-2550 m. Not associated with trees and detritus, usually in full sun; not parasitic when in association with lichens. Source of nutrition unknown. Distributed in western North America, in southern and eastern California in the Mojave Desert and in the Basin and Range Province, in Inyo County (Darwin Wash and White Mountains) and San Bernardino County (Cactus Flats in San Bernardino Mountains).
Cultured strains: Cultured Lichenothelia intermixta (strains L2282-L2284; FIGS. 1, 5a-k) on MY medium are characterized by a dense mycelium that secretes abundantly oil drops and by hyphae of predominately isodiametric cells (FIG. 2c-h, j, k) with abundant branching (FIG. 2c-f) observed in younger parts of the colonies, where hyphae are still rather hyaline (FIG. 2c-f). Mature parts of the colonies are highly melanized (FIG. 2g, j, k) and mainly composed of dense agglomerates of isodiametric cells. Oil inspersion was particularly abundant and prevented a proper analysis of the colony structure in the strain L2284.
Notes: The thallus of Lichenothelia intermixta varies in size depending on elevation, with stromata and ascospores being smaller in the type collection in Darwin Wash in Inyo County (stromata 40-80 μm; ascospores 10-15 × 7-9 μm). In the San Bernardino Mountains in southern California at elevations of 1200-1850 m, we observed a larger size of stromata and ascospores than in samples from the White Mountains in the Basin and Range Province at elevations of 2500-2600 m (stromata 50-110 μm wide; ascopores 14-20 × 7-10 μm). Henssen's protolog of Lichenothelia intermixta (Henssen 1987) describes a single collection, collected at its lowest recorded elevation (1200 m), with most stromata growing solitary and immersed in the rock among dendrite Lichenothelia calcarea Henssen. From this single collection, Henssen (1987) described very small sterile stromata almost completely lacking septate epilithic branches and very small fertile stromata without stipes (80 μm). She did not describe the interior of infertile or fertile stromata or whether interascal gel was amyloid or not. Specimens were impossible to be positively identified before we completed our study. The inaccuracy of Henssen's descriptions from single specimens highlights the unreliability of morphological descriptions of new Lichenothelia species from single specimens. If a description is based on a comparison with a table of Henssen's species, the analysis should include the study of her types. She apparently conserved her types and probably did only one or two sections when she wrote the descriptions. In her description of Lichenothelia calcarea and L. convexa, further studies of type and topotype material revealed a wider ascospore variability than she described, which was later confirmed by phylogenetic studies of analyzed specimens (Kocourková and Knudsen 2009;Muggia et al. 2013). The DNA extraction numbers of the analyzed samples correspond to L2220, L2221, L2313, L2314, L2318, and L2322 (FIG. 1; TABLE 1 Diagnosis: Similar to Lichenothelia scopularia but differing in having large muriform ascospores, (19-) 22.1-25.5-28.8(-30) × (9-)11-12.6-14.2(-15) μm.
Notes: This is the first species described with multiple ostioles and locules in fertile stromata. As with Lichenothelia scopularia (Nyl.) D. Hawksw., Saxomyces americanus, and S. penninicus, fertile stromata emerge from an areolate or nonareolate thallus, which forms a foundation of sterile paraplectenchymatous hyphae. The DNA extraction numbers of the analyzed samples correspond to L2302, L2303, and L2308 (FIG. 1;  TABLE 1). Diagnosis: Similar to Lichenothelia scopularia but differing in production short aerial hyphae with stromata at their terminus, papillate areoles, and fertile stromata with blue-brown paraplectenchymatous tissue of the wall.
Ecology and distribution: Known only from the type locality on north-facing limestone slopes in pinyonjuniper woodland from 1849 to 1865 m. Sources of nutrition unknown.
Notes: Lichenothelia papilliformis forms dense colonies of sterile areoles with rare fertile stromata. It produces epilithic and aerial hyphae generating new stromata at the terminus and papillae on sterile stromata. Its morphology suggests that it is a species more successfully spread vegetatively than generatively, replicating by forming new stromata at the end of epilithic or aerial hyphae and by papillae breaking off rather than by low number of ascospores produced in rather rare fertile stromata with low number of asci. Aerial hyphae and stromata are broken off in microflooding or by grains of rock in high wind and allow the new stromata to lodge among the limestone crevices, even several centimeters away from originating stroma. Because we did not find aerial hyphae with only stromata broken off, we hypothesize that whole aerial units break away. Henssen (1987) described Lichenothelia globulifera Henssen from single specimen on granite from Seychelles. It had similar stroma on aerial hyphae but only the stroma broke off, leaving behind aerial hyphae, and the taxon was also fertile.
Etymology: The name refers to this taxon being the first Saxomyces described from North America.
Lichenothelia macrocarpa Henssen (Henssen 1987) was described from Mount Rosa in Italy. It also has an areolate thallus, nonamyloid interascal gel, and interascal filaments like Saxomyces penninicus, but differs in having larger, two-celled ascospores size of 19-24 × 8-15 μm. This may be earliest name of Saxomyces penninicus if the ascospores measured from our two specimens for this study represent too narrow a sampling. It could also be the earliest name of the teleomorph of S. alpina, which was more abundant than S. penninicus in environmental samples from Mount Rose (Selbmann et al. 2013).
Cultured strains: We obtained three culture isolates for this species forming well-defined lineage with the cultured type strain S. penninicus CCFEE 5495: samples L2304, L2305, and L2337, from California and Italy, respectively.

DISCUSSION
The relationships of Lichenothelia, Saxomyces, and other RIF in Dothideomycetes.-We used an expanded and targeted taxon sampling of dothidealean RIF, and the inclusion of anamorphs and teleomorphs, to shed light on the cryptic diversity of these fungi and to support the process of species delimitation. We also show that combining molecular data from environmental samples with morphological analyses of environmental material and cultured isolates improves the recognition of taxa.
Previous studies have reported incongruence between genetic data gained from environmental samples and their corresponding cultures (Muggia et al. , 2015. The species diversity entangled in both the epilithic and endolithic RIF assemblages can be misestimated by primer biases or by culture conditions that favor the amplification and the growth, respectively, of only certain strains and leave others undetected. So far, we have recovered RIF associated with morphologically identified Lichenothelia and Saxomyces thalli in four lineages of Dothideomycetes. Interestingly, six isolates form a lineage within Lichenostigmatales, an order sister to Arthoniomycetes established by Ertz et al. (2014) to accommodate the lichen parasitic genera Etayoa and Lichenostigma. Eight cultured isolates together with two environmental samples are recovered within Capnodiales as close relatives of other RIF isolated from calcareous rocks from the Mediterranean region (Ruibal et al. 2005(Ruibal et al. , 2009, from Antarctica , and as plant pathogens (reference). Two environmental samples (L2311 and L2320) constitute a small lineage with a fungus isolated from lichen thalli (Muggia et al. 2016), and the single sample L2307 is basal to all the orders of Dothideomycetes. These results support the idea that certain RIF strains are widely distributed and can be recovered from diverse ecologies.
The present phylogenetic inference resolves Lichenothelia and Saxomyces into three lineages. The first, Lichenotheliales s. str., represents the core of Lichenothelia species, corresponding to Lichenotheliaceae/Lichenotheliales as circumscribed by Hyde et al. (2013) and Muggia et al. (2013Muggia et al. ( , 2015. It now includes seven well-recognized species (L. arida, L. convexa, L. dimelaenae, L. intermixa, L. muriformis, L. papilliformis, and L. umbrophila), two cultured strains representing Lichenothelia dimelaenae (L2184 and L2299), and 12 samples for which no name has been assigned yet (FIG. 1). Among these, we find the cultured strain L1851, which was isolated from a thallus of Lichenostigma epirupestre infecting the lichen Pertusaria pertusa (Perez-Ortega S. 1433). For this unique sample, we refrain from proposing here a new combination and wait to gather further molecular and morphological data.
Samples of L. calcarea and an additional three samples of L. arida group together in the second lineage, named Lichenotheliales (1), basal to Tubeufiales, Patellariales, and additional orders. These samples grouped with the core taxa of Lichenotheliaceae/ Lichenotheliales in the previous analysis of Muggia et al. (2015). Also, in previous analyses (Muggia et al. , 2015Liu et al. 2017), the order Lichenotheliales has been poorly supported, consistent with the broken monophyly of Lichenotheliales recovered here. The relationship of the most basal lineages in Dothideomycetes is still not fully settled.
The third lineage groups the type species of Saxomyces alpinus, S. penninicus, and S. americanus. Here, S. americanus is described as a new species based on fertile environmental samples and their cultured isolates. Within this lineage, additional samples, represented by both fertile and sterile thalli and isolates, show a multiplicity of morphologies. Further collections are necessary to corroborate their species description.
Even though both Lichenotheliales s. str. and Saxomyces group are individually monophyletic, their clades do not receive statistical support, and the placement of few samples still impairs an understanding of this relationship. In fact, the environmental samples L2216 and L2217 are recognized in Lichenotheliales s. str. as L. intermixa and L. arida, respectively. However, their cultured isolates are related to the Saxomyces group, together with other environmental samples coming from the same locality and their corresponding cultures, which are here formally described as S. americanus. This discrepancy between environmental samples and culture isolates derived from them recalls biases that may be due to either primer specificity (universal fungal primers were used) or that environmental samples each had material of several fungal species, with other species being amplified or cultured by chance. The risk of culturing or amplifying other species than the target one, however, were drastically reduced for fertile specimens, whose inocula were taken from fruiting bodies.

Species delimitation within Lichenothelia and
Saxomyces.-In view of the difficulty to use morphological characters alone for species recognition in Lichenothelia and Saxomyces, we were interested to complement morphological information with molecular data. Unfortunately, we could not obtain ITS sequence data for the majority of the samples and therefore could not base our species delimitation analyses on the standard barcode for Fungi (Schoch et al. 2012).
The more conserved nuclear genes, nucLSU and nucSSU rDNA, proved to be not suitable for ESU delimitation at the species level. Significantly higher values than the null model likelihood values in the LRT were found indeed only for the mitochondrial SSU marker. Nevertheless, species delimitation performed on BEAST-inferred phylogenies turned out to be the most coherent across different markers (SUPPLEMENTARY TABLES 2 and 3).
As reported in the literature (Fujisawa and Barraclough 2013;Tang et al. 2014;Kapli et al. 2017), mPTP and sGMYC were robust methods, often able to find enough separation between putative speciation events. Alternatively, sPTP, and to a lesser extent mGMYG, often oversplits the data set. The ABGD method did not find any valid barcode gaps to split the data set into different clusters.
For our data set, species delimitation methods did not always provide number and sample composition of ESU consistent with the species recognized by morphological analyses and phylogenetic inference. However, some main clusters are consistently recovered and correspond to the taxa recognized by classical taxonomy. Lichenothelia arida, L. convexa, and L. intermixta are consistently identified by sGMYC, whereas most of the samples of S. alpinus and S. americanus are identified by both PTP and GMYC methods (from the mtSSUbased BEAST phylogeny). More samples from different localities, and informative loci at the species level, are needed to consistently delimit species within Lichenothelia and Saxomyces.
Integrative taxonomy and teleomorphs and anamorphs in RIF.-Comprehensive approaches to delimit and describe species by integrating data from multiple types of analyses are known as integrative taxonomy (Dayrat 2005;Will et al. 2005;Schlick-Steiner et al. 2010). In mycology, integrative taxonomy has been applied widely when the circumscription of many taxa has been hampered by cryptic speciation, inconspicuous taxonomically diagnostic characters, or the lack of sexual reproductive forms (anamorph; e.g., Crous et al. 2004;Muggia et al. 2008Muggia et al. , 2009Taylor 2011;Lücking et al. 2014).
More recently, integrated approached including molecular data proved successful for fungi to find correspondences between anamorphs and teleomorphs. Saxomyces is here a further example, as the teleomorphic state was unknown at the time of its description ). Our first report in this paper of the sexual morph for a pure rock-inhabiting fungus in the new species S. americanus fills this lack of knowledge. The species forms a conspicuous, nonareolate mycelium on rock surfaces and bears fertile stromata (see species description above), which have allowed detailed analyses of asci and ascospores. Saxomyces americanus has been collected primarily on siliceous rocks (only once on limestone), usually in full sun above 2000 m elevation in mountains in California. Due to the presence of fertile stromata, the collections were first determined as Lichenothelia sp. according to stromata morphology and spore septation. However, molecular results unequivocally placed the specimens next to the type species of S. alpinus and S. penninicus. Our molecular data suggest that also S. penninicus bears its teleomorphic state, represented by the samples L2304, L2305, and L2337. These are fertile areolate mycelia with fertile stromata collected in California and Italy, respectively, and they form a well-defined lineage with the cultured type strain S. penninicus CCFEE 5495. It is likely that any future collection of RIF in the same ecological settings of S. alpinus might find its corresponding teleomorph, and this could be contemplated for other, remarkable rock-inhabiting genera included in the family Teratosphaeriaceae (e.g., Recurvomyces, Elasticomyces, Cryomyces as well as the Antarctic endemic Friedmanniomyces) known so far from cultured isolates only.
Our integrative approach demonstrates that in the case of Lichenothelia and Saxomyces, the use of genetic data and the application of the phylogenetic species concept proved necessary to better circumscribe taxa. Unfortunately, descriptions of new species based on detailed morphological analyses cannot be always complemented by genetic data (Valadbeigi et al. 2016). However, previous studies (Henssen 1987 and the present one have shown that the subtle continuum of morphological differences may mirror intraspecific variability or a phenotypic plasticity (Handry 2016), necessitating examination of multiple specimens per taxon. Additionally, the lack of genetic data for putatively new taxa with the available frame of Lichenotheliales s. str. (Hyde et al. 2013;Muggia et al. 2013Muggia et al. , 2015 may introduce further biases when new species of this group are presented to science. In Henssen (1987), the descriptions in the protologs were based on few or even single samples and did not properly estimate the variation of diagnostic traits.