Sharing of photobionts in sympatric populations of Thamnolia and Cetraria lichens: evidence from high-throughput sequencing

In this study, we explored the diversity of green algal symbionts (photobionts) in sympatric populations of the cosmopolitan lichen-forming fungi Thamnolia and Cetraria. We sequenced with both Sanger and Ion Torrent High-Throughput Sequencing technologies the photobiont ITS-region of 30 lichen thalli from two islands: Iceland and Öland. While Sanger recovered just one photobiont genotype from each thallus, the Ion Torrent data recovered 10–18 OTUs for each pool of 5 lichen thalli, suggesting that individual lichens can contain heterogeneous photobiont populations. Both methods showed evidence for photobiont sharing between Thamnolia and Cetraria on Iceland. In contrast, our data suggest that on Öland the two mycobionts associate with distinct photobiont communities, with few shared OTUs revealed by Ion Torrent sequencing. Furthermore, by comparing our sequences with public data, we identified closely related photobionts from geographically distant localities. Taken together, we suggest that the photobiont composition in Thamnolia and Cetraria results from both photobiont-mycobiont codispersal and local acquisition during mycobiont establishment and/or lichen growth. We hypothesize that this is a successful strategy for lichens to be flexible in the use of the most adapted photobiont for the environment.

and an advantageous strategy for lichens that disperse by meiospores is to use as immediate source photobiont cells from already established lichens, dying lichens or vegetative propagules 9,10 . Accordingly, lichen lineages do not constitute cospeciating symbiotic fungi and algae; rather, horizontal transmission (i.e., switching) of photobionts between different fungal lineages frequently occurs 11,12 . Photobiont switching seems to be an ubiquitous phenomenon in lichens and appears to play a vital role in lichens adaptation to environment 8,13 . Furthermore, this is not a lichen specific phenomenon, but was also shown in reef corals, which can survive elevate temperatures and recover from bleaching episodes if they succeed to associate with a specifically adapted photobiont lineage 2,14 . Although the photobiont adaptive hypothesis was only occasionally experimentally tested in lichens 15,16 , several studies have shown a strong correlation between the photobiont genotype and ecology or geography of the lichen 4,17,18 . Additionally, it was shown that photobiont genotypes can be shared by diverse mycobionts at the same location 17,[19][20][21][22][23][24][25][26][27][28][29][30] . All of these findings advocate that lichen forming fungi associate with local adapted photobiont strains that will be selected and hence increase in frequency within population 19 . Lichens photobionts do not show a significant population structure between localities, but identical photobiont genotypes were shown to be encountered across wide geographic ranges 4,31 , suggesting that the vegetative dispersing mode of lichens is an important contributor on photobiont distribution 23,32 .
At the same time, it is yet not clarified how many photobiont genotypes individual lichen thalli can contain. Traditional Sanger sequencing of PCR products containing amplicons of ITS-region has been used for the molecular identification of photobionts in lichen thalli (e.g. 31 ) and this methods is not optimal for detecting multiple photobionts in a tissue. However, the results from studies using different methodological approaches, including High-Throughput Sequencing (HTS), challenge the dogma of single photobiont genotypes in a lichen thallus, and suggest the presence of a heterogeneous photobiont population within a lichen individual 15,[33][34][35][36][37] .
The aims of this study were (1) to explore the diversity of photobionts in lichen symbioses in nature, and (2) to compare the sensitivity of Sanger and IonTorrent HTS technologies for detecting photobiont genotypes in lichens. For this purpose, we used a simple study design to investigate photobionts of two cosmopolitan, ecological similar but distantly related 38 lichenized fungal genera, Thamnolia Ach. ex Schaer. and Cetraria Ach. We investigated the diversity of photobionts in sympatric populations of Thamnolia and Cetraria from two geographic localities (Iceland and Öland). We also investigated whether the retrieved local photobiont genotypes can be encountered in Thamnolia and Cetraria individuals from other parts of the world. Our HTS data suggest that multiple photobiont genotypes coexist within lichen thalli and revealed sharing of photobionts in sympatric population. This study increases our understanding of the nature of symbioses formed by heterotrophs and autotrophs, and may spur other researchers to use HTS methodology to examine the photobiont community of lichens.

Materials and Methods
Study system. Thamnolia and Cetraria were selected since they are often found growing together in nature.
Furthermore, there are previous reports of symbiont flexibility in Thamnolia 31,39 and photobiont switching has previously been correlated with the colonization of climatically different regions in Cetraria 4 . Both Thamnolia and Cetraria can be encountered in high alpine and tundra environments, but Cetraria species have a wider distribution and are also found in the boreal forest or even warm, arid habitats 4 . Both genera appear to mainly reproduce through vegetative propagules, via lichen fragments as in Thamnolia [40][41][42] or through diaspores (isidia: small outgrowths of lichen thallus) produced by Cetraria 4 . In addition, both genera produce fungal spores that will need to re-associate with local photobionts to form new lichens 4,43 . Field sampling and morphological identification. The lichen samples used for this study were collected from two islands representing two different habitats: the melur of Iceland (sandy, gravelly hills with sparse vegetation) 44 and the alvar of Öland, Sweden (limestone plain with very thin soil layer) 45 . On Iceland one site was sampled (Gardabaer), while on Öland two sites were sampled (Öland 1 and Öland 2) (Fig. 1, Table 1). There was a noticeable difference in the soil color at the different alvar sites; while Öland 1 had a yellowish color, indicating goethite enrichment and an increased moisture availability, Öland 2 had a reddish color, which indicates a good soil drainage and the presence of unhydrated iron oxides 46 . The field sampling was performed as follows: within an area of 400 m 2 , five circles with one-meter radius and with at least two meters distance between each other were selected (Fig. 1). Within each circle, Thamnolia and Cetraria specimens (individual lichen thalli) were collected and identified as Thamnolia subuliformis (Ehrh.) W.L. Culb., Cetraria aculeata sensu lato (Schreb.) Fr. or Cetraria islandica (L.) Ach by using morphological characters. For further analyses, we used one specimen of each genus from each circle (Fig. 1, Table 1).

DNA extraction.
For each selected specimen, we divided the lichen tissue in two halves: one half to be analysed using Sanger sequencing and one using HTS based Ion Torrent technology (see Supplementary Fig. S1). For Sanger sequencing, DNA extractions were separately performed on all 30 specimens listed in Table 1. For Ion Torrent sequencing we pooled for each locality the five specimens of Thamnolia into one sample, and the five specimens of Cetraria in another sample (Table 1, see Supplementary Fig. S1). We used 5 mg tissue of each specimen for DNA extraction. For both sequencing technologies, DNA was isolated with DNeasy Plant Mini Kit (Qiagen, GmbH, Hilden, Germany).
Sanger sequencing of the mycobiont and photobiont ITS regions. PCR reaction was executed with the Thermo-Scientific Green Phusion High-Fidelity DNA polymerase kit (Thermo Fisher Scientific Inc. Waltham, MA, USA), according to the manufacturer's instructions. From each of the 30 samples of Table 1, the ITS region of the mycobiont was amplified with the general primers ITS1F and ITS4, and the ITS region of the photobiont was amplified using the general primer pair ITS1T and ITS4T (Table S1). The PCR products were cleaned of unwanted dNTPs and primers using Exonuclease I (Thermo Fisher Scientific, SE) and Shrimp Alkaline SCIENtIfIC REPORTS | (2018) 8:4406 | DOI:10.1038/s41598-018-22470-y Phosphatase (Affymetrix). The BigDye Teminator v 3.1 Cycle Sequencing and BigDye XTerminator Purification Kits (Thermo Fisher Scientific Inc. Waltham, MA, USA) were used to prepare the sample to be sequenced with Sanger technology by the ABI 3730 XL machine of Evolutionary Biology Centre, Uppsala University. All PCR products were sequenced from both directions using the same primers that were used for PCR amplification. The chromatograms of the sequences were visually inspected and curated by using the program Sequencher v5.2.3 (Gene Codes Corporation, Ann Arbor, MI, USA) and Geneious R9 (http://www.geneious.com) 47 . Sequencher was used to identify heterozygous sites in the sequences generated by Sanger technology: when two peaks with over 80% area overlap were encountered in a sequence, the site was identified as heterozygous, and the IUPAC ambiguity code was used for that particular site.

Verification of mycobionts in the pooled samples.
To verify that we did not have any cross-contamination between samples of Thamnolia and Cetraria in the six pooled samples of lichens used for Ion Torrent sequencing (Table 1), we used genus specific primers to verify the mycobiont. As control, we mixed DNA from the two genera in ratios of 1:9, 1:5 and 1:1 and performed two rounds of amplification on all of these DNA samples by using either Thamnolia and Cetraria fungal ITS specific primers designed for this study (Table S1).
Ion Torrent sequencing of the ITS1 region of photobionts. PCR was executed with the Thermo-Scientific Phusion High-Fidelity DNA polymerase kit to amplify a part of the algal ITS1 region of maximum size 350 pb, using barcoded primers specifically designed for this study by the bioinformatics support from the Ion Torrent Sequencing center at SciLifeLab in Uppsala (Table S1). We also used two negative controls in which the template DNA was replaced by sterile water. Each sample for PCR (including the negative control) had its own barcoded primer combination. Gel electrophoresis was used to confirm the presence of PCR products of the expected size in the samples with DNA and the absence of PCR products in the negative controls. We used DNA Clean and Concentrator kit (Zymo Research, Orange, CA) to clean the amplicons from unwanted primers and dNTPs. The PCR product concentrations were measured with Qubit fluorometric quantification (Life Technologies), where after a single solution with all barcoded amplicons represented in equal amounts, and the two negative controls, was created. The library preparations with fusion primers were performed according to the manufacturers instructions. Uppsala SciLifeLab sequenced the pooled amplicons samples on an Ion Torrent PGM platform (Ion Torrent Xpress) using the 314 TM Chip KIT and 400 bp reads. The raw reads are deposited to the European Nucleotide Archive (ENA) (see Table 1 for the ENA accession numbers).
Delineating and quantifying operational taxonomic units from Ion Torrent data. The Ion Torrent reads were processed to trim adapter and low-quality sequences using TorrentSuite BaseCaller 4.0-6/76303. To verify adapter removal, we constructed an adapter sequence list using Ion Torrent documentation, choosing A and P1 adapters along with the common prefix of barcode adapters, and scanned FastQ-format read sequences for each sample using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/). None of the candidate adapter sequences were detected. To cluster amplicon sequences into operational taxonomic units (OTUs) for photobiont delineation and quantification, we used USEARCH 8.1.1861 48 following recommended practices for Roche 454 reads (http://drive5. com/usearch/manual/upp_454.html, accessed 2016/06/01), which have an error profile similar to data obtained by Ion Torrent sequencing. All OTU-related trimming, clustering and quantification steps were performed using USEARCH. We explored a variety of read trim lengths and expected-error counts using the -fastq_eestats2 option, and selected trimming to 200 bp with 1.5 expected errors per read; this allowed us to retain sufficient length for each OTU to make comparisons with our Sanger sequences and sequences downloaded from databases. FastQ reads for all samples were pooled, dereplicated, and clustered into OTUs using default settings, with the additional constraint of requiring each OTU to be supported by at least two reads (−minsize 2). Thus, OTUs were formed using reads from all samples analysed together, with singleton OTUs excluded. Abundances of OTUs within each pooled set of reads were then estimated via global alignment of sampled-tagged reads, with minimum identity 97%. This method ensures continuity of OTUs across samples, enables formation of OTUs that might only have sufficient read support in aggregate while avoiding formation of singleton OTUs, and allows for OTUs to be quantified within samples in which they are rare. The script implementing this pipeline together with the candidate adapter sequences used during the scan with FastQC are available at https://github. com/douglasgscofield/pubs/tree/master/Brannstrom-et-al-1.
To example the dependence of OTU richness estimates on read counts, which varied somewhat among samples (mean 36740 ± 8300 [s.d.], range 22486-43110), we determined rarefaction curves using estimated_obser-vation_richness.py in QIIME 1.9.1 49,50 . Richness estimates differed little throughout the range of observed read counts (see Supplementary Fig. S2), so we made no further correction to OTU diversity estimates.  Taxonomic verification of mycobionts and photobionts. To assign the taxonomic identity of the mycobiont for each specimen identified as Thamnolia or Cetraria based on morphological characters, we compared the newly generated mycobiont and photobiont ITS sequences, both Sanger and Ion Torrent data, to sequences available at the GenBank nucleotide database 51 . We identified the best hits based on query coverage, sequence identity and e-value. The best hits that also had a species name assigned, using additional supporting evidence, were retrieved from GenBank. Subsequently, we employed a phylogenetic approach (see below) to investigate the relationship among the retrieved GenBank sequences and our newly generated data.
Phylogenetic analyses. For each independent analysis, the sequences were aligned with mafft 52 and trimmed in AliView v1.18 53 . All sequence alignments and the tree files resulted from this study analyses were deposited to https://github.com/douglasgscofield/pubs/tree/master/Brannstrom-et-al-1.
We chose the most appropriate strategy for phylogenetic analyses based on the presence of heterozygous sites and the amount of variable sites within each alignment.
Since we identified heterozygous sites in the data from photobiont Sanger sequences of the complete ITS region, the program RRHS version 1.0.0.2 54 was used to randomly draw (with replacement) 1000 phased alignments of the trimmed alignment of the 30 photobiont Sanger sequences. For each of the 1000 phased alignments, phylogenetic trees were generated using the maximum likelihood (ML) with garli 2.01 55 by running for each alignment five independent searches. Kimura 80 was selected as model for sequence evolution using the jMod-elTest. The 1000 trees with the best likelihood score were compiled in one phylogeny that represents the 50% majority-rule consensus tree using the program SumTrees 56 .
Thamnolia and Cetraria are highly divergent mycobiont genera from Lecanoromycetes 38 . Therefore we performed a separate phylogeny for samples of each genus based on the mycobiont Sanger sequences of the complete ITS region of our newly generated data together with GenBank references of sequences. For Thamnolia we chose representative ITS sequences for the previously discovered lineages A, B and C 31 with the following accession numbers: KY550180, KY550083, KY550130, KY550098, KY550079, KY550170, KY550163, KY550143, KY550153, KY550127, KY550194. For Cetraria we used ITS sequences previously identified as C. aculeata, C. muricata and C. islandica (Accession numbers: AF228302, AF192409, EF373589, EF373567, AF228299). Both sets of sequences were aligned in AliView with mafft, and TrNef + G was chosen with JModel Test as sequence evolution model. Maximum likelihood phylogenies were performed with garli 2.01 together with 500 bootstrap trees. SumTrees script was used to place the bootstrap values on the branches.
The alignment of our newly generated photobiont sequences (ITS1 region) amplified with Sanger and Ion Torrent and the GenBank references (Table S2) presented many ambiguously aligned regions. To be able to use all sequence information, the software BAli-Phy 2.3.8 57 , which simultaneously estimates the alignment and the phylogeny using Bayesian inference, was used. We used JModel test 58  Sharing of photobiont OTUs. Non-metric multidimensional scaling (NMDS) was used to visualize similarities among the lichen samples with regard to photobiont diversity. The analyses were performed in R employing the function metaMDS in Vegan 60 using the Bray-Curtis dissimilarity index. Before this analysis, the data sets were resampled (using the function rrarefy from Vegan) to an equal number of sequences per sample (22486). The NMDS analysis was run for 2 dimensions and 2000 iterations and the plot was generated in R using ggplot2 61 . The script used for this analysis is available at https://github.com/douglasgscofield/pubs/tree/master/ Brannstrom-et-al-1.
The number of shared and specific OTUs between Thamnolia and Cetraria for each of the three localities was visualized as Venn diagrams that were constructed using the following web tool: http://bioinformatics.psb.ugent. be/webtools/Venn/.
The relative quantities of photobiont OTUs shared at each sampling location between Thamnolia and Cetraria were visualized by a hierarchical cluster analysis. The OTU read counts per sample were normalized to range from 0-1 within each sample, and were then clustered by similarity separately among OTUs and among site/ genus pairs using hclust and dist in R. Distances were calculated via the Canberra method, as this permitted clustering to remain sensitive to similarities among more rare OTUs, unlike Euclidean distances, while retaining quantitative sensitivity unlike Boolean presence/absence distances. For display, OTU quantities were scaled by their cube-root, and colored by clade membership as determined by the phylogenetic analysis described above. The R script implementing this analysis is available at https://github.com/douglasgscofield/pubs/tree/master/ Brannstrom-et-al-1.
Data Availability. All the herbarium vouchers are deposited in UPS (The Museum of Evolution Herbarium -Uppsala, Sweden) and their herbarium accessions numbers can be found in Table 1. All the DNA sequences are deposited in GenBank and ENA. The accession numbers can be found in Table 1 (for the Sanger sequences) and

Results
DNA extractions, PCR, Sanger and Ion Torrent sequencing. By using PCR and Sanger sequencing, we successfully amplified and sequenced the ITS region of the mycobiont from 29 of the 30 lichen specimens, and the ITS region of the photobiont from all 30. For the Ion Torrent sequencing of the photobionts, DNA was extracted and the ITS1 region was successfully amplified from all six pooled samples ( Table 1). None of the negative controls showed amplicons by gel electrophoresis. Only one sequence of ITS per thallus was obtained by Sanger technology, while with Ion Torrent we discovered reads of multiple photobiont OTUs: after delineation and quantification of the photobiont amplicon sequences, 32 OTUs were identified. Each OTU ID together with the number of reads for all OTUs and their prevalence in each sample is shown in Table 2. Sequencing of the negative control resulted in 1 read (Table 2).

No cross-contamination of mycobionts in the pooled samples. PCR-reactions using Thamnolia or
Cetraria specific primers (Table S1) verified that single genera were found in each pooled tissue-sample, and thus, that the samples were not cross-contaminated prior to the Ion Torrent sequencing step. Specifically, we amplified a PCR product from the Cetraria samples when Cetraria specific primers were used, but not when Thamnolia specific primers were used, and vice versa. In contrast, all samples that were mixed in different proportions with DNA originating from both Cetraria and Thamnolia tested positive with both primer combinations, which suggests that any cross contamination at the level of 10% would have been detected (data not shown).  Taxonomic identification of the mycobiont using Sanger ITS data. The mycobiont phylogeny based on the Sanger sequences of the ITS region of Thamnolia (Fig. 2a) shows that all newly generated sequences group together with Thamnolia GenBank sequences of a clade previously identified as Thamnolia Lineage C 31 and subsequently named Thamnolia subuliformis s. str. 62 . In contrast, our Cetraria phylogeny (Fig. 2b) splits our samples into four supported lineages. Based on the GenBank references sequences, shown in black font in Fig. 2b, all samples identified as C. islandica group together with the reference of C. islandica. However, the specimens that were morphologically identified as C. aculeata are paraphyletic, two groups cluster with C. aculeata reference and one with the one of C. muricata (Fig. 2b).

OTUs ID
Taxonomic identification of the photobiont using Sanger and Ion Torrent ITS data. In the phylogenetic analyses of the Sanger and Ion Torrent sequences of the ITS region of the photobionts, all 5 MCMC chains converged and were combined in the 50% majority consensus rule tree. The Trebouxia impressa clade was  Table 1). The accession numbers and the species/lineage name of the sequences retrieved from GenBank are shown in black. The support (<60%) for each branch is given as bootstrap values shown above branches. The scale bar indicates the branch length.
Most of the Ion Torrent sequences were found in four of the clades: 'T. simplex 2′, 'T. simplex C1', 'T. simplex 1′, and 'T. jamesii-vulpinae'. The exceptions were OTU 29 and OTU 14 which groups in the 'T. gigantean-vagua' clade together with a photobiont sequence amplified from a Thamnolia lichen specimen from China (accession number EU715061: Fig. 3). Notably, this sequence was amplified with Sanger technology by another research group and retrieved by us from GenBank.
Due to its divergence from the other sequences, OTU 25 was not included in the phylogeny and was identified based on a nucleotide blast search in the NCBI database. The best hit was Hemichloris antarctica which had a 100% query cover, an e-value of 1e-37 and an identity score of 80% was (GenBank accession number: HG972970). OTU 25 was found in low abundance in Cetraria (Öland 1, with 1 read) and Thamnolia (Öland 1, with 9 reads) ( Table 2).

Sharing of photobionts between Thamnolia and Cetraria.
Both the Sanger and the Ion Torrent datasets indicate that the photobiont community is more similar between Thamnolia and Cetraria on Iceland than on Öland. First, the photobiont phylogenetic analyses of the complete ITS region obtained with Sanger technology (see Supplementary Fig. S3) showed that the photobionts of Thamnolia and Cetraria from Iceland associate with closely related photobionts. Most of the photobionts of Cetraria specimens from Öland are closely related with the lichens from Iceland, while the Thamnolia photobionts form a separate clade (see Supplementary Fig. S3).
The data from Ion Torrent shows the same trend but at higher resolution, since more genotypes were recovered by this analysis. Both clustering analyses of the Ion Torrent data (Figs 4 and 5) together with the phylogenetic analyses of the amplicons of the ITS1 region (Fig. 3) group the photobiont communities of the two lichen genera from Iceland together. The Venn diagrams of Fig. 5b show that circa 67% of the photobiont OTUs are common between Thamnolia and Cetraria at melur (Iceland). Most of this shared photobiont OTUs belong to a clade identified as Trebouxia simplex 1 (Fig. 5a), which was also encountered in Cetraria specimens from Öland. The Icelandic Thamnolia and Cetraria were also found associated with a low frequency OTU (OTU 1) that was identified as Trebouxia simplex 2 (Fig. 5a). Noteworthy, OTU1 was found in high abundance in Thamnolia samples from Öland ( Fig. 5a; Table 2).
These overall results of differential sharing hold when applying more stringent filtering in requiring each OTU to be supported by at least 3, instead of 2 reads (see Supplementary Figs S4 and S5).

Discussion
Thamnolia and Cetraria associate with diverse, and partly overlapping, Trebouxia photobionts. Our results confirm the findings of previous studies indicating that mycobionts of Thamnolia and Cetraria are flexible towards their photobionts in that they can associate with different Trebouxia lineages 4,31,39 (Fig. 3). Although three of these photobiont lineages were detected only in either Thamnolia (Trebouxia 'impressa',  Fig. 3) or Cetraria (Trebouxia 'jamesii-vulpinae', Trebouxia 'angustilobata', Fig. 3), all the other clades contained photobionts retrieved from both lichen genera. Our results were based on observations from three sites, but nevertheless showed that Thamnolia and Cetraria, which are often found growing together, can harbour an overlapping spectrum of genetically diverse photobionts in their thalli.
It is worth pointing out that we cannot with this study design confirm without doubt that the identified algal OTUs have their origin from within the thalli, and thus, that they represent true photobionts (c.f. refs 36,63 ). Other studies, which were specifically interested in investigating the intrathalline photobiont diversity in several lichen species employed a more stringent approach by washing the lichen thalli with different solutions 36,63 . However, the only certain confirmation of the algal origin from within the thalli is the culturing of single cells of intrathalline photobionts 33 , or specific labelling, such as for example the one used in the recent study by Spribille et al. 2016 64 . Both of these last two approaches require a priori knowledge of the present intra-lichen diversity, which was not available at the onset of our study. The only taxon that appears to be epithallic from our data is OTU 25, which is divergent (Hemichloris or similar, i.e., not a known lichen photobiont 38 ) and was found in a small fraction of samples only on Öland 1, suggesting that external contamination is not a major issue for our study. Nevertheless, we are left with essentially the same result: sharing of a broad diversity of photobiont taxa, whether endo-or epithallic, which differs between species and sites. HTS sequencing suggests that multiple photobiont genotypes exist within lichen thalli. The presence of a single photobiont in a lichen thallus usually found when using PCR and Sanger sequencing of molecular markers (e.g. ref. 25 ) could be biologically true for certain lichens, but recent work indicates that it might be an artefact of the traditional Sanger method, which is biased towards producing a predominant genotype as its single amplification product. Accordingly, with Sanger, we identified one genotype per individual thallus investigated herein, while the Ion Torrent data suggests the coexistence of several algal genotypes within single lichen individuals. Because our samples were pooled before Ion Torrent sequencing, we were not able to directly investigate the distribution of genotypes within individual thalli with this data. However, the finding of more photobiont genotypes (10 to 18 OTUs) than the number of studied thalli (5) in each sample suggests the simultaneous presence of multiple photobionts in Thamnolia and Cetraria lichens. The finding of more OTUs than number of thalli holds also when applying more stringent filtering in requiring each OTU to be supported by at least 3 reads, indicating that this result is not due to, e.g., rare OTUs originating from sequencing errors. Furthermore, we verified the coexistence of multiple photobiont genotypes within single lichen thalli by using specific primers to amplify and Sanger-sequence several Trebouxia lineages ('T. simplex', 'T. impressa' and 'T. vagua-gigantea' - Fig. 3) (data not shown).
The idea that symbiotic organisms are not a homogenous population of autotrophic symbionts but rather a mosaic of local adapted photobiont strains was also shown in some corals systems 65 , and is consistent with several lichen studies that have used other molecular techniques such as restriction fragment polymorphism 37 , microsatellites 34 or HTS technologies to show that individual thalli contain more than one photobiont genotype 36,66 .
In conclusion, all these recent studies are challenging the paradigm that lichen thalli contain homogeneous photobiont populations and highlight the importance of using more sensitive techniques than PCR and Sanger sequencing to the study of lichen photobionts. For example, if macrolichens such as Thamnolia and Cetraria contain heterogeneous intrathalline photobiont populations, approaches that reveal only the most abundant genotype (c.f. refs 4,31,39,67 ) can give misleading results, e.g., infer photobiont switching and replacement in cases where instead dynamic fluctuations of photobiont within lichen thalli is taking place.

Photobiont sharing in sympatric populations of Thamnolia and Cetraria. Data from both
sequencing technologies showed convincingly that the photobionts associated with Thamnolia and Cetraria from Iceland are closely related with each other, and with photobionts from Cetraria individuals from Öland (Figs 3  and 5a). Based on these genetic similarities we suggest that on Iceland, Thamnolia and Cetraria share photobionts, while on Öland the two lichen genera associate with distinct photobiont communities, although few OTUs were shared. Sharing of photobionts between closely related mycobiont species with similar ecologies was previously shown in lichens 23,24 but rarely described between distantly related mycobiont genera 66 . Together with other studies, which show that lichen communities from harsh climatic conditions share photobionts 20,25,66 , our results support the hypothesis that mycobionts with similar ecological requirements, such as those specific to arctic environments, might share a common pool of photobionts 21,12,17,35 . Furthermore, the presence of similar genotypes across wide geographic ranges, combined with occurrence of multiple photobiont genotypes within lichen thalli, suggest a continuous exchange of photobionts between Thamnolia and Cetraria, by yet unknown mechanisms. The significance of differential photobiont abundance in lichen thalli. If the proportion of reads generated by Ion Torrent sequencing is used to assess the predominant genotype within a lichen, we see that, as an example, on Öland one genotype (OTU3; T. simplex C1) is dominant in the Cetraria thalli, while on Iceland a genetically different photobiont is most abundant in Cetraria (OTU2; T. simplex 1) (Fig. 5a). Thamnolia sampled on Iceland also harbours OTUs belonging to T. simplex 1, but based on the proportion of reads shows more abundantly photobionts of OTU 24. The amount of the different photobiont genotypes in a lichen population may be a result of chance alone, or simply reflect the demographic history and/or the specificity of the mycobiont towards their photobionts 68 . However, one may also hypothesize that the photobionts confer a phenotypic impact on the lichen and that the abundant photobiont provides a fitness advantage of the individual thallus, (i.e., that . NMDS clustering of the photobiont composition, based on Ion Torrent data, of the six pooled lichen samples: one Thamnolia (depicted with circles) and one Cetraria (depicted with rhombs) from each of the three localities: Iceland (blue), Öland 1 (yellow) and Öland 2 (red). For all samples the reads were pooled, dereplicated, and clustered into OTUs using the constraint of requiring each OTU to be supported by at least two reads (−minsize 2). The Bray-Curtis dissimilarity was used and the stress value was 1.
SCIENtIfIC REPORTS | (2018) 8:4406 | DOI:10.1038/s41598-018-22470-y the abundance of photobionts is driven by the adaptation to the local environment). Experiments investigating differential fitness of lichen thalli in response to relative abundance of photobionts have to our knowledge not been carried out. However, in line with this hypothesis, when photobiont lineage specific primers have been used in combination with monocultures of photobiont cells from lichen thalli, it has been shown that coexisting algal genotypes with complementary physiological performances 15 could be found within a lichen individual 33,69 . Such selection of genetically similar but distinct photobionts within a thallus fits within the 'fungal-farmers hypothesis' , in which mycobionts acquire available photobionts and subsequently select the appropriate genotype as Hierarchical clusters based on photobiont OTU relative quantities. The number of reads for each OTU in each sample is shown by bar thickness, which is scaled via its cube-root for display. The left dendrogram clusters OTUs by similarity in composition and abundances, and the top dendrogram clusters site/genus pairs by similarities in composition and abundances. The colors of the bars reflect the genotypic identity of each OTU based on the ITS1 photobiont phylogeny (Fig. 3). For all samples the reads were pooled, dereplicated, and clustered into OTUs using the constraint of requiring each OTU to be supported by at least two reads (−minsize 2). (b) Venn diagrams showing the proportion of shared and exclusive OTUs between Thamnolia and Cetraria pooled samples within each locality.
SCIENtIfIC REPORTS | (2018) 8:4406 | DOI:10.1038/s41598-018-22470-y the farming observed in ant growing fungi 19 or the use of environmentally adapted photobionts in corals 2,70 . Yet another, not mutually exclusive, adaptive model is that the selection is acting on the photobiont, so that it increases in abundance in the thallus, but that this process is independent of the mycobiont.
The effect of lichen dispersal on photobiont community in thalli. Photobiont sharing between distantly related lichens suggest that horizontal transmission of photobiont occurs 11 and our data supports this notion. However, the mechanism of photobiont recruitment in macrolichens such as Thamnolia and Cetraria with a predominantly or even exclusively vegetative dispersal mode is still not understood. Does photobiont switching take place only when fungal spores spread by themselves and resynthesize the lichen by acquiring a photobiont obtained from an external source? Independent dispersal of the symbionts to Iceland and local reassembling of the lichen in that area could explain some of the observed sharing of photobiont populations between Thamnolia and Cetraria and why Thamnolia specimens from Öland seem to contain a different spectrum of photobionts (Fig. 5a). Furthermore, our data of identical, or very similar, photobiont genotypes from different parts of the world (Fig. 3) suggest that the photobiont community is also affected by long-distance codispersal of the symbionts, which is likely to take place by vegetative propagules. One third factor that may affect the photobiont distribution in lichens is the continuous acquisition of photobionts after the lichen is formed, a phenomenon shown to take place in corals 70 , but never proven in lichens. The combination of these routes of lichen dispersal and photobiont acquisition, together with the ability to maintain a diverse set of photobionts within a thallus, may be advantageous for exclusively or predominantly vegetative spreading lichen species to adapt to new environments.

Conclusion
Our results confirm that lichens may contain heterogeneous populations of photobionts that can be shared between species 15,[33][34][35][36][37] . The lichen guild hypothesis postulates that distantly related but sympatric lichenized fungi may share a common photobiont pool 21,24,34 , and it was also shown that geographically separated lichens that occupy similar niches may have similar photobionts 4,12,17,18 . Based on this earlier work, together with our findings of similar genotypes across wide geographic ranges and the occurrence of multiple photobiont genotypes within Thamnolia and Cetraria thalli, we suggest an extended hypothesis regarding lichens adaption to the environment. We speculate that, just as reef-forming corals 70 , predominantly vegetatively dispersing lichens acquire photobionts during both incipient and adult life stages (idea also proposed in 2010 by Asplund and co-workers 71 ). Depending on the environment, the most adapted photobiont increases in relative frequency within the thallus, at the same time as the accumulated intrathalline diversity is retained. Dispersal through vegetative propagules that maintain the photobiont diversity of the thallus will increase the chance for a successful establishment of lichen individuals in the new environments. Additional data from a diverse range of lichen species addressing for example the distribution of photobionts within individual thalli combined with fitness experiments, is needed in order to confirm this adaptive hypothesis and prove its generality in lichen symbiosis.