1 Introduction

Lichens have been iconic examples of symbiosis for the past 150 years (Honegger 2000). While a lichen was originally defined as a symbiotic relationship between a single fungus, the mycobiont, and a single or few species of green algae or cyanobacteria, the photobiont, studies have shown this is overly simplistic. It wasn’t until the late twentieth century that in vitro studies began to look at other fungi as potentially lichen-associated organisms rather than mere contaminants (Petrini et al. 1990; Crittenden et al. 1995; Girlanda et al. 1997). It is now apparent that lichens are ecologically complex, internally consistent, and self-sustaining symbiotic phenotypes composed of evolutionarily diverse microbes (Goward 2008; Honegger 1993; Honegger 2001; Spribille et al. 2020).

Advances in sequencing technologies have allowed for deeper investigation into the diversity of the lichen symbiosis. In addition to the predominant myco- and photobionts, additional fungal and green algal/cyanobacterial species are often associated with a lichen thallus, in addition to non-photosynthetic bacteria (Grube et al. 2009; Lawrey and Diederich 2003; Muggia et al. 2013). Photobiont diversity can be shaped by reproductive and dispersal strategies of the mycobiont (Cao et al. 2015; Steinova et al. 2019), geography (Muggia et al. 2014; Werth and Sork 2014; Leavitt et al. 2015), growth substrate (Bačkor et al. 2010; Leavitt et al. 2013b; Muggia et al. 2014), and macroclimate (Lu et al. 2018; Singh et al. 2018). The diversity of photobionts has been only recently explored by environmental DNA metabarcoding approaches and has focused on species within the Mediterranean basin to date (Moya et al. 2017; Dal Grande et al. 2018). In contrast to high-throughput sequencing approaches, traditional and largely applied DNA barcoding using Sanger sequencing was able to detect only the principal photobiont in the thalli (Paul et al. 2018). Additionally, many studies show that lichens are surrounded by a consortium of bacteria (Bates et al. 2011) that may change with substrate, altitude, and geography (Cardinale et al. 2012; Hodkinson et al. 2012; Fernandez-Brime et al. 2019). Potential functions of bacterial microbiomes include providing the host with nutrients, as well as protective and growth-regulating functions (Cernava et al. 2017). Furthermore, carbon exchange between lichen green algae and non-photosynthetic bacteria has recently been suggested (Kono et al. 2017).

The lichen mycobiome – fungal communities superficially associated with the lichen thallus, e.g. on or near the lichen’s surface/cortex, and within the thallus – can be made up of lichenicolous fungi (Lawrey and Diederich 2003) and endolichenic fungi (Arnold et al. 2009; U'Ren et al. 2010; Muggia et al. 2016). Lichenicolous fungi growing on a lichen thallus may or may not be parasitic, and are defined as “symptomatic” if they influence their host’s morphology (Lawrey and Diederich 2003; U'Ren et al. 2010; Fleischhacker et al. 2015). The majority of lichenicolous fungi are classified within lichen-dominated groups (e.g., Lecanoromycetes), while endolichenic fungi are common among all major primary nonlichenized lineages, e.g., Sordariomycetes, Dothideomycetes, Leotiomycetes, and Pezizomycetes (Arnold et al. 2009; Hibbett et al. 2007). Further, endolichenic fungi are largely asymptomatic or cryptic with the host thallus. Many of the fungi associated with lichens appear to not be accidental colonizers of lichens (Arnold et al. 2009; U'Ren et al. 2010). While some studies have found patterns in the lichen-associated mycobiome – e.g., changing with altitude (Zhang et al. 2015; Wang et al. 2016) – others have found little specificity between the lichens and their associated mycobiome (Fleischhacker et al. 2015; Fernandez-Mendoza et al. 2017; Yu et al. 2018).

Recently, basidiomycete yeasts have been called into question as a potential symbiotic partner in the lichen symbiosis with the discovery of Cystobasidiomycetes (Basidiomycota, Pucciniomycotina) in the cortices of lichens (Spribille et al. 2016). The presence of this group of fungi was previously discovered in association with two genera in the lichen-forming family Parmeliaceae, Hypogymnia, and Usnea by Millanes et al. (2016) who clarified the phylogenetic position and the monophyly of two lichen-inhabiting species which were accommodated in the new genus Cyphobasidium. Later Černajová and Škaloud (2019) found Cystobasidiomycete yeasts in 95% of Cladonia specimens collected across Europe, though they were suggested to be either part of a superficial biofilm or living within the thallus without associating with the cortex itself. In contrast, Lendemer et al. (2019) found them in just nine of the 339 species investigated. The question remains of how abundant and specific cystobasidiomycetes are in lichen symbioses, as well as how consistent the mycobiome might be among different lichen-forming fungal species, e.g., do evolutionary constraints of the mycobiont influence the range and composition of associated fungi?

Intrathalline photobiont diversity – multiple photobiont species within a single lichen thallus – has previously been observed in a number of lichen symbioses (Muggia et al. 2013; Dal Grande et al. 2014; Moya et al. 2017; Škaloud et al. 2018). In some cases, algae with different physiological performances are ever-present in lichen thalli potentially facilitating the success of these lichens in a wide range of habitats and geographic areas and/or in changing environmental conditions. However, PCR amplification and Sanger sequencing has been shown to consistently fail to effectively generate DNA sequence data from lichen specimens when multiple Trebouxia lineages occur within a single lichen thallus (Paul et al. 2018), potentially biasing the perspective of lichen photobiont associations. The prevalence of intrathalline photobiont diversity in lichens remains unclear and thus impacts our understanding of its ecological and evolutionary significance.

As lichens are a model of symbiosis, there is a need to better characterize their microbial partners and associations. With the increasing availability of metagenomic short-read data from lichens, it may be possible to utilize these data to explore novel questions relating to lichen symbioses. Currently available data has been generated using a wide array of methodological approaches, ranging from metagenomic and transcriptomic sequencing with reads obtained from samples from multiple species, each represented by a single thallus fragment (Leavitt et al. 2016; Spribille et al. 2016), to population-level samples with multiple thalli representing a single species pooled into replicate samples (Dal Grande et al. 2017; Dal Grande et al. 2018). Therefore, we used existing datasets of metagenomic shotgun sequences and implemented a bioinformatics pipeline to extract metagenomic reads representing the standard fungal DNA barcode region in an attempt to do the following: (i) characterize the lichen mycobiomes across multiple, phylogenetically distinct lichen groups, (ii) assess the prevalence of basidiomycete yeast, a putative symbiotic partner in some lichen symbioses, and (iii) investigate the potential for multiple species-level Trebouxia algal lineages within a single lichen thallus.

2 Materials and methods

2.1 Taxon sampling

Although the Code of Botanical Nomenclature anchors the name of the lichen to the Latin binomial of the mycobiont, whole lichens – the complete symbiotic phenotype or holobiont (all organisms found within a lichen thallus) – lack any formal taxonomic recognition (Goward 2008). Therefore, in this study when referring to the lichen holobiont, we use the appropriate taxonomic level of the mycobiont followed by ‘lichen’, e.g., ‘Rhizoplaca lichens’ refers to lichen holobionts associating with mycobionts in the genus Rhizoplaca Zopf and not to the mycobiont alone. Our sampling focused on five morphologically distinct lichen groups: (i) Rhizoplaca lichens (rock posy lichens; Fig. 1a & b), (ii) Xanthoparmelia lichens (green rock shields; Fig. 1c & d), (iii) Umbilicaria lichens (rock tripe lichens; Fig. 1e), (iv) Bryoria lichens (horsehair lichens; Fig. 1f), and (iv) representatives from the mycobiont family Physciaceae (shadow lichens; Fig. 1g & h) (Table 1). Rhizoplaca lichens were represented by three distinct forms from the closely related Rhizoplaca melanophthalma group, all occurring in western North America: the vagrant/erratic forms representing Rhizoplaca arbuscula Rosentr., St. Clair & Leavitt (Fig. 1b; n = 3) and R. melanophthalma subsp. crispa Rosentr. & B.D. Ryan (n = 3), in addition to R. melanophthalma (DC.) Leuckert, which is attached to rocks (Fig. 1a; n = 3) (Leavitt et al. 2013a). Xanthoparmelia lichens were also represented by three distinct forms occurring in western North America: vagrant forms representing Xanthoparmelia aff. chlorochroa (Tuck.) Hale (Fig. 1d; n = 3), isidiate (vegetative reproductive propagules) forms (Fig. 1c; n = 3), and the sexually reproducing taxon X. subcumberlandia Elix & T.H. Nash (n = 3) (Leavitt et al. 2011). Umbilicaria lichens were represented by two species collected in Spain: U. hispanica (Frey) Davydov, Peršoh & Rambold (3 populations) and U. pustulata (L.) Hoffm. (Fig. 1e; 3 populations). For the Umbilicaria lichens, each sample represents metagenomic reads from a pooled population – 100 lichen thalli/population (Dal Grande et al. 2017) – rather than reads from an individual lichen thallus. Bryoria lichens were represented by two species: Bryoria fremontii (Tuck.) Brodo & D. Hawksw. (Fig. 1f; n = 3) and B. tortuosa (G. Merr.) Brodo & D. Hawksw. (n = 3) (Velmala et al. 2009). Lichens associating with the mycobiont family Physciaceae were represented by Mobergia calculiformis (W.A. Weber) H. Mayrhofer & Sheard (Leavitt 16–697 [BRY-C]), Physcia aff. biziana (A. Massal.) Zahlbr. (Leavitt 17–611 [BRY-C]), Physciella aff. chloantha (Ach.) Essl. (Leavitt 17–586 [BRY-C]), Oxnerella safavidiorum S.Y. Kondr., B. Zarei-Darki, L. Lőkös & Hur (Leavitt 16–665 [BRY-C]), and Rinodina (Ach.) Gray sp. (Leavitt 16–665 [BRY-C]). For Rhizoplaca lichens, Xanthoparmelia lichens, and representatives of Physciaceae, specimens were collected in dry conditions, with subsamples for molecular study removed within 24 h of collection and frozen at −20 °C until DNA extraction. Sampling of Bryoria and Umbilicaria lichens were reported previously in Spribille et al. (2016) and Dal Grande et al. (2017, 2018), respectively.

Fig. 1
figure 1

Examples of lichens groups considered in this study, including Rhizoplaca lichens (a & b), Xanthoparmelia lichens (c & d), Umbilicaria lichens (e), Bryoria lichens (f), and Physciaceae lichens (g & h). a, Rhizoplaca melanophthalma – field image from La Sal Mountain Range, Utah, USA. b, Rhizoplaca arbuscula – collected from Lemhi Valley, Idaho, USA, voucher Leavitt 18–1017 (BRY-C). c, Xanthoparmelia cf. mexicana – field image from Snake Range, Nevada, USA. d, Xanthoparmelia aff. chlorochroa – field image from Awapa Plateau, Utah, USA. e, Umbilicaria pustulata – field image from La Coruña, Galicia, España. (source: https://commons.wikimedia.org/wiki/File:Lasallia_pustulata.001_-_Islas_Cies.JPG [CC BY-SA 4.0]). f, Bryoria fremontii – from Oppland, Norway (source: https://www.flickr.com/photos/aburgh/27323080245 [CC BY-NC-SA 2.0]). g, Physcia biziana – field image from vicinity of Santa Fe, New Mexico, USA (Hollinger 2492). h, Rinodina olivaceobrunnea – field image from vicinity of John Day, Oregon, USA. (Hollinger 7073). Note: the name listed for each lichen is for the mycobiont (main fungal partner) and does not account for the range of potential other associated symbionts. Permission to use photographs in panels ‘g’ and ‘h’ was kindly provided by Jason Hollinger

Full size image
Table 1 List of lichen specimens included in this study, including: non-scientific names referring to major lichen groups investigated here and the associated mycobiont and family affiliation; voucher number, NCBI Short Read Archive experiment accession number (SRX), or sample name from previous study; locality information or data source for each specimen; the total number of short reads (and read length) generated for each specimen; the number of reads mapped (and relative proportion) to the UNITE QIIME v.8 dynamic release ITS database for fungi; and the NCBI Short Read Archive project and accession numbers for each sample
Full size table

2.2 Metagenomic sequencing

Metagenomic short reads used in this study originated from a range of sources and sequencing methods (Table 1). Metagenomic reads from Rhizoplaca lichens were initially reported in Leavitt et al. (2016, 2019) and are available in NCBI’s Short Read Archive under project PRJNA576709. For newly generated metagenomic reads from Xanthoparmelia lichens and representatives of Physciaceae, total genomic DNA was extracted from a small portion of the lichen thallus (comprised of the mycobiont, photobiont, and other associated microbes) using the E.Z.N.A. Plant DNA DS Mini Kit (Omega Bio-Tek, Inc., Norcross, GA, USA) following the manufacturers’ recommendations. Total genomic DNA was prepared following the standard Illumina whole genome sequencing (WGS) library preparation process using Adaptive Focused Acoustics for shearing (Covaris), followed by an AMPure cleanup step. The DNA was then processed with the NEBNext Ultra™ II End Repair/dA-Tailing Module end-repair and the NEBNext Ultra™ II Ligation Module (New England Biolabs) while using standard Illumina index primers. Libraries were pooled and sequenced with the HiSeq 2500 sequencer in high output mode at the DNA Sequencing Center, Brigham Young University, Provo, Utah, USA, using either 250 cycle paired-end reads or 300 cycle paired-end reads. Reads from Xanthoparmelia lichens and representatives from the mycobiont family Physciaceae are available in NCBI’s Short Read Archive (Table 1). The reads from the Bryoria lichens are distinct because they are transcriptomic reads (Spribille et al. 2016), and we aimed to extract by-catch reads representing the internal transcribed spacer region (ITS). For the Umbilicaria lichens, each sample represents metagenomic reads from a pooled population (Pool-seq) – 100 lichen thalli/population (Dal Grande et al. 2017, 2018) – rather than reads from an individual lichen thallus.

2.3 Sequence analysis

All reads were filtered using TRIMMOMATIC v0.33 (Bolger et al. 2014) before mapping to remove low quality reads and/or included contamination from Illumina adaptors using the following parameters: ILLUMINACLIP; LEADING:3; TRAILING:3; SLIDINGWINDOW:4:15; and MINLEN:36. Previous studies have used assembled metagenomic contigs (Keepers et al. 2019) or mapped fungal reads to a fungal protein database (LaBonte et al. 2018) to provide crucial insight into fungal diversity in lichens and deciduous trees. Given the expected low coverage for fungi potentially co-occurring with a lichen thallus in short reads generated for this study, we chose to focus on the well-known repeat region which includes the standard fungal DNA barcode, the ITS region of the nuclear ribosomal DNA (nrDNA) (Schoch et al. 2012). Across fungi, nrDNA copy number has been shown to vary considerably, ranging from tens to over 1400 copies per genome (Lofgren et al. 2019; Bradshaw et al. 2020). Furthermore, a comparatively robust and well-curated ITS database exists for fungi (Nilsson et al. 2019).

For reference ITS sequences, we used the UNITE QIIME v.8 dynamic release for fungi (Nilsson et al. 2019), filtered to include only sequences between 300 to 800 base pairs (reduced from 70,512 to 69,872 ITS sequences). Following recommendations by QIIME 2 developers, flanking regions, e.g., portions of the18S and/or 28S, with ITS sequences in the UNITE database were retained to reduce erroneous classifications when using the naïve Bayes classifier (https://doi.org/10.7287/peerj.preprints.27295v2). The UNITE ITS database was supplemented with all Cystobasidiomycetes ITS sequences reported in Spribille et al. (2016). All sampled lichens are reported to associate with members of the genus Trebouxia as the primary lichen photobiont. In addition to assessing fungal diversity in short reads generated from lichen thalli, we also included representative sequences for each of the Trebouxia operational taxonomic units (OTUs) circumscribed in Leavitt et al. (2015). Although lichens are known to associate with a broader range of algae than the core photobionts (Muggia et al. 2013), we did not assess accessory algae outside of Trebouxia.

For each metagenomic library, reads were mapped back to the composite ITS database using the Geneious read mapper in Geneious Prime (Kearse et al. 2012), implementing ‘Medium-Low Sensitivity / Fast’ sensitivity, iterated two times and saving all successfully mapped reads. Exploratory analyses with other read mapping approaches consistently recovered lower quantities of successfully mapped reads (data not shown). For each sample, metagenomic reads successfully mapped to the ITS references were imported into QIIME 2 (Bolyen et al. 2019). Reads were dereplicated using Vsearch ‘dereplicate-sequences’ (Rognes et al. 2016), implementing default settings. The dereplicated sequences were clustered into de novo OTUs at a 97% similarity in Vsearch using ‘cluster-features-de-novo’ (McDonald et al. 2012; Rognes et al. 2016). A naïve Bayes taxonomic classifier was trained using the same ITS reference library in QIIME 2 (Bokulich et al. 2018). The OTUs were then taxonomically classified with the trained naive Bayes trainer using the QIIME 2 ‘feature-classifier classify-sklearn’ at a 0.95 confidence level to minimize false positives, with all other settings at default (McKinney 2010; Pedregosa et al. 2011; Bokulich et al. 2018).

Of the estimated 2.2 to 3.8 million fungal species, only 3–8% are currently named (Hawksworth and Lücking 2017), and a much smaller portion are represented in available DNA reference libraries. Exploratory analyses of our lichen mycobiome data revealed poor taxonomic resolution below class levels for the majority of OTUs inferred here. Therefore, fungal OTUs that were classified at the class level were retained, and others with less taxonomic resolution were excluded. Classification of fungal OTUs generated from reads mapped to the reference ITS database was summarized using the QIIME ‘Taxa Barplot’ feature (Caporaso et al. 2010). Data were managed, analyzed, and visualized in R (R Core Team 2019) using ggplot2 (Wickham 2016) and tidyr (Wickham et al. 2019). To assess the similarity of lichen mycobiomes within and among phylogenetically distinct mycobionts, a principle component analysis (PCA) was performed on the class-level taxonomic classification using tidyr (Wickham et al. 2019), with the command ‘prcomp’. While formal species-level taxonomy in the lichen photobiont Trebouxia remains woefully inadequate (Muggia et al. 2020), DNA sequence data representing a wide range of putative species-level lineages, with accompanying provisional names, is available (Leavitt et al. 2015). For Trebouxia (photobiont) OTUs, the classified reads were filtered at the ‘species’ level, based on the 69 putative species-level OTUs from Leavitt et al. (2015), using QIIME ‘taxa filter-table’ command to determine the range of Trebouxia diversity occurring within each sample. All code used in this experiment is provided as supplementary file S1.

3 Results

Between 0.41 and 3.68% of metagenomic reads from each sample were mapped back to the ITS reference library (Table 1). The primary lichen symbionts, the mycobiont and photobiont, accounted for ca. 50% of all ITS reads extracted from the metagenomic data on average (Fig. 2a). The relative abundance of ITS reads representing the mycobiont (inferred at the class level, e.g., Lecanoromycetes) was between 5.20% to 80.31% of ITS reads, with an average relative abundance of ca. 40%. The relative abundance of reads from the photobiont, Trebouxia spp., comprised between 0.68% to 35.09% of ITS reads, with an average relative abundance of ca. 10%.

Fig. 2
figure 2

Overview of lichen symbionts and associated fungi inferred from data from the internal transcribed spacer region extracted from metagenomic shotgun sequencing short reads sequenced from lichen thalli representing five different groups of lichens. a, proportion of reads assigned to the lichen symbionts – mycobiont (shown in blue) and photobiont (in red) – and other major fungal lineages. b, inferred membership of lichen mycobiomes (at class level); the main lichen symbionts are excluded – see panel ‘a’ – and relative abundance of remaining fungal classes were adjusted proportionally. Abbreviations by names of fungal groups in panel ‘b’ are: Asc. = Ascomycota, Bas. = Basidiomycota, Chy. = Chytridiomycota, Ent. = Entomophthoromycota, Glo. = Glomeromycota, and Kic. = Kickxellomycota. For Rhizoplaca lichens, ‘Rhar’ = R. arbuscula (vagrant lichen), ‘Rhpo’ = R. melanophthalma subsp. crispa (vagrant forms), and ‘Rhme’ = R. melanophthlama (rock-dwelling, fertile forms); for Xanthoparmelia lichens, ‘Xach’ = X. aff. chlorochroa (asexual, vagrant lichens), ‘X. isidiate’ = three different isidiate, rock-dwelling forms (see Table 1), and ‘Xasu’ = X. subcumberlandia (fertile, rock-dwelling forms); for Bryoria lichens, ‘Brto’ = B. tortuosa and ‘Brfr’ = B. fremontii; for Umbilicaria lichens, ‘Uspu’ = U. pustulata and ‘Ushi’ = U. hispanica. See Table 1 for a full list of sampled lichens

Full size image

Lichen-associated fungi made up a large fraction of metagenomic reads, representing a total of 22 fungal classes (Fig. 2b). Both in terms of abundance and diversity, Ascomycota OTUs were most frequently recovered and represented by 10 classes, excluding the mycobiont class Lecanoromycetes, followed by Basidiomycota represented by seven classes. Chytridiomycota (represented by two classes), Glomeromycota (one class), and Kickxellomycota (one class) were found in low abundance and diversity (Fig. 2). Overall, reads from Cystobasidiomycete yeasts were poorly represented in extracted ITS reads, found in only 5 of the 35 samples. Notably, ITS by-catch from the Bryoria fremontii transcriptomic data from which lichen-associated yeasts were reported in the cortex, resulted in the highest abundance of reads potentially representing cystobasidiomycete yeasts, with an average relative abundance of 0.7% of the ITS reads in the three B. fremontii samples. In the remaining two samples with evidence of Cystobasidiomycete yeasts, Physcia biziana and one sample of Xanthoparmelia chlorochroa (818F), had an average relative abundance of 0.03%.

Closely related mycobionts tended to have more similar mycobiomes (Fig. 3). The PCA revealed a general pattern of mycobiome similarity among samples representing mycobiont species, and relatively high levels of similarity among mycobiont congeners (Figs. 2b & 3). Differences in lichen mycobiomes are most distinct among different genera of lichen-forming fungi.

Fig. 3
figure 3

Principal component analysis (PCA) of lichen mycobiome diversity in five distinct lichen groups, each represented by multiple species. Sample codes shown in PCA plot are linked to specimens listed in Table 1

Full size image

Evidence supporting intrathalline Trebouxia photobiont diversity was observed in 16 of the 29 samples (Umbilicaria samples not considered) (Fig. 4). Intrathalline photobiont diversity in Umbilicaria pustulata and U. hispanica was described in detail in Paul et al. (2018). Short read data from Umbilicaria lichen samples analyzed in this study were generated from 100 pooled individual thalli from a single population. Thalli from representatives of Physciaceae and Rhizoplaca lichens consistently contained a dominant Trebouxia lineage with >90% relative abundance. Xanthoparmelia lichens associated with a broader range of Trebouxia species, with evidence of multiple Trebouxia species occurring within an individual lichen thallus. Two of the six Bryoria samples also provided evidence of multiple Trebouxia species occurring within individual thalli (Fig. 4).

Fig. 4
figure 4

Assessment of Trebouxia (photobiont) diversity for each metagenomic sample derived from a single lichen thallus. Distinct operational taxonomic units (OTUs) are color coded, and the proportional occurrence of each OTUs is reported for each metagenomic sample derived from a single thallus. For Rhizoplaca lichens, ‘Rhar’ = R. arbuscula (vagrant lichen), ‘Rhpo’ = R. melanophthalma subsp. crispa (vagrant forms), and ‘Rhme’ = R. melanophthlama (rock-dwelling, fertile forms); for Xanthoparmelia lichens, ‘Xach’ = X. aff. chlorochroa (asexual, vagrant lichens), ‘X. isidiate’ = three different isidiate, rock-dwelling forms (see Table 1), and ‘Xasu’ = X. subcumberlandia (fertile, rock-dwelling forms); and for Bryoria lichens, ‘Brto’ = B. tortuosa and ‘Brfr’ = B. fremontii. See Table 1 for a full list of sampled lichens. Trebouxia OTUs nomenclature follows Leavitt et al. (2015)

Full size image

4 Discussion

The broad range of organisms involved in lichen symbioses has recently been highlighted, including diverse algae (Muggia et al. 2013; Moya et al. 2017), non-photosynthetic bacteria (Cardinale et al. 2006; Grube et al. 2009; Hodkinson and Lutzoni 2009), and diverse fungal lineages (Lawrey and Diederich 2003; Spribille et al. 2016; Tuovinen et al. 2019). Using data mining of fungal ITS reads from metagenomic shotgun sequences of lichen thalli, we provide a coarse snapshot of unexpectedly diverse lichen-associated mycobiomes (Fig. 2). The accessory fungi accounted for a significant proportion of ITS reads extracted from metagenomic shotgun sequencing data (Fig. 2b), spanning multiple phyla – dominated by Ascomycota and Basidiomycota but with representatives from Entomophthoromycota, Chytridiomycota, Glomeromycota, and Kickxellomycota. While a number of the class-level lineages inferred from metagenomic ITS reads are known to associate with lichens, e.g., Agaricomycetes, Dothideomycetes, Eurotiomycetes, and Sordariomycetes, other classes do not include fungi previously known to associate with lichens, e.g. Entomophthoromytes. In contrast to recent studies highlighting the role of two basidiomycete lineages in some lichen symbioses, Tremella (Tuovinen et al. 2019) and Cystobasidiomycete yeasts (Spribille et al. 2016), these were recovered only sporadically and in very low abundance in our samples. Nonetheless, these basidiomycete fungi have often been reported as lichenicolous, growing on a number of lichen hosts (Diederich 1996; Millanes et al. 2016). Below we discuss the potential implications of our findings and potential ways to move forward.

The relative importance of host versus environment in determining the diversity of the lichen mycobiome is poorly understood. However, lichen mycobiomes appear to comprise stable and transient guilds, which to some extent correlate with the ecological conditions of the lichen habitats. Fernandez-Mendoza et al. (2017) proposed three ecological components of lichen mycobiomes: (i) generalist taxa common to the environmental pool of bio- and sapro-trophic fungi, (ii) lichenicolous and endolichenic fungi specific to each genus/species, and (iii) species which disperse and possibly germinate on, among, and within lichen thalli, but do not play a definite ecological role in the lichen community. Our results indicate that closely related mycobionts tend to have more similar mycobiomes (Fig. 3), even in cases where distinct lichens commonly co-occur, e.g. Xanthoparmelia and Rhizoplaca lichens. Furthermore, umbilicate and vagrant forms of Rhizoplaca lichens shared similar fungal communities, despite the perceived ecological differences between growing on rocks versus occurring free on the soil. These data support the perspective that a significant component of the lichenicolous and endolichenic fungal community are specific to different mycobiont genera/species and that evolutionary constraints of the mycobiont may influence the range and composition of associated fungi.

Differences in the growth form of lichens likely create distinct microhabitats that may influence the intrathalline microbiome, particularly on the underside of the thallus or other specialized morphological features. While both Rhizoplaca and Umbilicaria lichens are umbilicate with a central holdfast, we found that mycobiome communities of umbilicate lichens occurring on rocks were quite distinct. Whether this is the result of the evolutionary constraint of the mycobiont host or broader biogeographic patterns of lichen associated fungi (Rhizoplaca lichen samples were all collected from western North America, while Umbilicaria lichen samples originated from Europe) remains unknown. The influence of lichen secondary metabolites in shaping microbiome diversity also remains under-explored. Lichen secondary metabolites have broad antibacterial properties (Boustie and Grube 2005). Grube et al. (2015) hypothesized that the fungal partner plays an important role in regulating bacterial colonization of the thallus, and secondary metabolites may also impact other components of the lichen microbiome. Each of the lichen groups sampled here can be characterized by the production of distinct secondary metabolites, and exploring the role of secondary metabolite variation in structuring mycobiome communities was beyond the scope of this study.

Broadly speaking, Sordariomycetes and Leotiomycetes are frequently recovered from lichens occurring in humid, temperate, boreal environments, and Antarctic environments, representing lineages closely related to plant endophytes (Arnold et al. 2009; U'Ren et al. 2010; Yu et al. 2018). In contrast, Dothideomycetes and Eurotiomycetes are more frequently associated with rock-inhabiting lichens (Muggia and Grube 2018). In rock-inhabiting lichens, the lichen-associated fungi are usually melanized fungi comprising unknown and known hyphomycetous lineages, which show close affinities to some symptomatic lichenicolous fungi, extremotolerant rock-inhabiting fungi from oligotrophic environments, and to plant and animal pathogenic black yeasts (Muggia et al. 2016; Muggia and Grube 2018). These fungi are widely known as black fungi because they accumulate melanins in their cell walls, which enable them to grow in oligotrophic environments and resist multiple abiotic stresses, such as high doses of radiation, desiccation, and temperature extremes (Gostinčar et al. 2009). Black fungi are, therefore, usually recognized as (poly)extremotolerant organisms.

In our study, different lichen-forming fungal genera tended to associate with distinct fungal communities (Fig. 2b & 3). These relationships appear to be consistent across relatively broad geographic areas, at least for some lichens. Our results indicated that the mycobiomes of Xanthoparmelia lichen and Rhizoplaca lichen populations occurring across western North America were strikingly different (Fig. 2b). While disparate morphologies of Rhizoplaca lichens were shown to have relatively consistent mycobiomes, even in specimens collected across geographically distinct populations, differences in mycobiome communities of Xanthoparmelia lichens with different morphologies and reproductive strategies were observed (Fig. 2b). However, within Xanthoparmelia lichens, vagrant (obligately unattached specimens), rock-dwelling isidiate (reproducing via specialized asexual propagules), and rock-dwelling sexually reproducing forms tended to associate with distinct fungal communities, albeit inferred from limited sample sizes. Additional research will be required to more fully assess if distinct mycobiomes, or core subsets of the mycobiome, within lichen groups are maintained across geographic and ecological distances. If differing core mycobiome communities are found in association with distinct mycobionts, at what level does this specificity exist, e.g., mycobiont species, genera, etc.? Directed experimental design and broader sampling will be required to determine how lichen mycobiomes are structured at different evolutionary scales relative to the predominant mycobiont.

When investigating the potential for photobiont (Trebouxia spp.) diversity within a single lichen thallus, our results suggest that a single lichen thallus of some lichen groups, e.g., shadow lichens (members of the mycobiont family Physciaceae) and Rhizoplaca lichens, tend to associate with a single/one dominant Trebouxia lineage. For Umbilicaria lichens, Paul et al. (2018) observed a single pattern of a single dominant Trebouxia lineage per thallus. However, the metagenomic reads from Umbilicaria lichens used in the present study were generated from multiple lichen thalli pooled into a single population per site, and we were unable to corroborate these results reported. In contrast, it appears that Xanthoparmelia lichen (mycobiont = Xanthoparmelia) thalli consistently harbor multiple, distinct Trebouxia lineages. A previous study characterizing Trebouxia diversity associating with members of the mycobiont family Parmeliaceae also demonstrated distinct patterns of photobiont association between Rhizoplaca spp. and Xanthoparmelia spp., with Xanthoparmelia spp. associating with a much wider range of photobionts than Rhizoplaca spp. (Leavitt et al. 2015). Furthermore, these two mycobiont genera consistently associated with distinct Trebouxia lineages with very little overlap, and these results were corroborated by our findings (Fig. 4). By explicitly taking the potential for intrathalline photobiont diversity into consideration, we anticipate novel insight into different strategies of lichen symbiosis.

While our study provides novel insight into lichen symbioses and impetus for future research, there are a number of methodological limitations that potentially bias the results presented here. Metagenomic reads from lichen-forming fungi are expected to be dominated by reads from the major lichen symbionts, the myco- and photobionts (Pizarro 2019), and other eukaryotic microbial diversity associated with lichen thalli is likely found in much lower abundance in metagenomic short read data. Therefore, here we opted to target fungal reads from the multi-copy nuclear ribosomal cistron (nrDNA) in order to identify fungi that might be found in low relative abundance and likely overlooked using single copy regions and metagenomic binning approaches. Furthermore, while portions of the nrDNA are highly conserved across fungi, we focused on the ITS region due to the high variability and well-curated reference database (Schoch et al. 2012; Nilsson et al. 2019). However, nrDNA copy number varies by orders of magnitude across fungi, from tens to over 1400 copies per genome (Lofgren et al. 2019). Therefore, the relative abundance of fungal groups inferred in this study (e.g., Fig. 2b) does not accurately depict true relative abundance of lichen-associated fungi given the potential for a very wide range of nrDNA copy number of these fungi.

Another source of potential bias is from the bioinformatic pipeline implemented here. Even using relatively well-established pipelines for analyzing ITS amplicon-based metagenomic reads, bioinformatics analysis pipelines have been shown to vary greatly in their relative performance and accuracy in characterizing fungi from metagenomic data (Anslan et al. 2018). We would anticipate that the data mining approach implemented in this study may have introduced a number of unexpected and difficult to identify artifacts, ranging from potentially over- and under-representing different fungal lineages to erroneous taxonomic assignments. For example, in the present study, a significant proportion of reads from Xanthoparmelia lichens were assigned to the class Entomophthoromycetes, a lineage that has not previously been found in association with lichens. Whether the inferred prevalence of Entomophthoromycetes is biased by copy number variation of the nrDNA, an artifact of read mapping to the UNITE database, etc., or accurately represents a novel finding is unclear. Furthermore, only a small fraction of the estimated 2.2–3.8 million fungal species are represented in currently available curated databases. Therefore, in fungal metabarcoding studies, a large proportion of OTUs cannot be assigned to any meaningful taxonomic group, and these unclassifiable species hypotheses, or ‘dark taxa’, remain problematic in metagenomic studies of fungi (Nilsson et al. 2019).

How might differences in sample preparation, DNA extraction approaches, and library preparation influence the range of captured diversity? How comparable are by-catch rDNA reads from transcriptomic sequencing with metagenomic sequencing, and might one have an advantage for detecting organisms found in low abundance, e.g. Cystobasidiomycete yeasts? The impact of these potential methodological limitations is not clear. Furthermore, important aspects of the experimental design are lost when data mining available metagenomic reads. In this study, the only metagenomic lichen samples originating outside of North America were represented by all the Umbilicaria specimens. While our results clearly indicate that the fungal communities associating with Umbilicaria lichens are distinct from those associating with other lichen groups analyzed here, we have no reference to assess if these differences are related to evolutionary constraints of the mycobiont or continental-scale biogeographic factors shaping fungal communities. Similarly, our sampling of Physciaceae lichens represented broader ecological and evolutionary diversity than the other sampled lichen groups, likely masking important patterns in mycobiome communities associated with members of the Physciaceae and making comparisons of Physciaceae lichens with other lichen groups less direct. Taken together, our results highlight, on one hand, the presence of a highly diverse, seemingly lichen host-specific mycobiome, and on the other hand, the risk of applying overly simplistic techniques – such as phylum rank classifications – to tackle the diversity of these lichen-associated fungal communities.