Photobiont switching causes changes in the reproduction strategy and phenotypic dimorphism in the Arthoniomycetes

Phylogenetic analyses using mtSSU and nuITS sequences of Buellia violaceofusca (previously placed in Lecanoromycetes), a sterile, sorediate lichen having a trebouxioid photobiont, surprisingly prove that the species is conspecific with Lecanographa amylacea (Arthoniomycetes), a fertile, esorediate species with a trentepohlioid photobiont. These results suggest that L. amylacea and B. violaceofusca are photomorphs of the same mycobiont species, which, depending on the photobiont type, changes the morphology and the reproduction strategy. This is the first example of a lichenized fungus that can select between Trebouxia (Trebouxiophyceae) and trentepohlioid (Ulvophyceae) photobionts. Trebouxia photobionts from the sorediate morphotype belong to at least three different phylogenetic clades, and the results suggest that Lecanographa amylacea can capture the photobiont of other lichens such as Chrysothrix candelaris to form the sorediate morphotype. Phylogenetic analyses based on rbcL DNA data suggest that the trentepohlioid photobiont of L. amylacea is closely related to Trentepohlia isolated from fruticose lichens. The flexibility in the photobiont choice enables L. amylacea to use a larger range of tree hosts. This strategy helps the lichen to withstand changes of environmental conditions, to widen its distribution range and to increase its population size, which is particularly important for the survival of this rare species.


Results
Fungal Phylogeny. The mtSSU dataset consisted of 32 specimens and 590 unambiguously aligned sites.
In our mtSSU phylogenetic tree (Fig. 1), the Lecanographaceae are composed of four mainly supported lineages corresponding to the Plectocarpon-Alyxoria and Phacographa-Opegrapha brevis clades and the genera Lecanographa Egea & Torrente and Zwackhia Körb. The five specimens of Buellia violaceofusca are nested in the genus Lecanographa and intermixed with the specimens of Lecanographa amylacea. The nuITS dataset consisted of 10 specimens and 637 unambiguously aligned sites. In our nuITS phylogenetic tree (Fig. 2), the four specimens of Buellia violaceofusca cluster in a polytomy with the three specimens of L. amylacea because the nuITS sequences of these seven specimens are identical. These results prove that B. violaceofusca is conspecific with L. amylacea. Therefore, B. violaceofusca is synonymized with L. amylacea. L. amylacea is morphologically very different from B. violaceofusca notably by a thick, white-powdery thallus often dotted with brown flecks, containing a trentepohlioid photobiont (i.e. filamentous algae containing large amounts of carotenoid pigments, causing the algae to appear yellow orange in color), the absence of soralia and the occasional presence of apothecia. Our results prove thus that B. violaceofusca and L. amylacea represent two photomorphs of the same ascomycete species. These morphotypes are so distinct that they were described as two different species belonging to two different classes (Lecanoromycetes versus Arthoniomycetes).
Algal phylogenies. The Trebouxia algal dataset consisted of 99 nuITS sequences and 651 unambiguously aligned sites. The Bayesian tree (Harmonic mean was −4936.69) is shown in Fig. 3 with added bootstrap supports from RaxML analysis (ML Optimization Likelihood was −4490.471038). The newly sequenced photobionts revealed in three different clades within Trebouxia spp. The most common photobiont in B. violaceofusca thalli is Trebouxia sp. that groups together with Trebouxia sp. strain TR9 previously reported from Ramalina farinacea (L.) Ach. 19 . It was found in all Polish specimens (Kukwa 12916, 12998 and 15628) and four specimens from Sweden (i.e. Thor 31255, 31380, 31385 and 31389). A second lineage is related to the T. solaris clade and consists of Trebouxia sp. found in two Swedish specimens (i.e. Thor 31382 and 31385) as well as in four Chrysothrix candelaris specimens (i.e. Thor 31379, 31380, 31382, 31385) growing together with 'B. violaceofusca'-morphotype. These molecular data strongly support our hypothesis that Lecanographa amylacea can capture the photobiont of Chrysothrix candelaris to form 'Buellia violaceofusca' thalli. This possibility was also concluded from the morphological observations of our specimens, because of the presence of Buellia violaceofusca at the margin of Lecanographa amylacea (with trentepohlioid photobiont) thalli in contact with Chrysothrix candelaris. A third lineage consists of two sequences from a specimen of Buellia violaceofusca (Thor 31376) that are nested within the Trebouxia jamesii clade. The latter is a common photobiont reported from e.g. Candelariella vitellina (Hoffm.) Müll. Arg., Lecanora spp., Protoparmelia sp., Ramalina spp., Rhizocarpon geographicum (L.) DC., Rhizoplaca sp. 36,37 . Interestingly, among the studied samples, we found that two different Trebouxia photobionts may be present in Buellia violaceofusca thalli. In specimen Thor 31385, we detected Trebouxia sp. TR9 and Trebouxia sp. related to T. solaris clade (Fig. 3).
The rbcL dataset of trentepohlioid photobionts consisted of 60 sequences and 668 unambiguously aligned sites. The Bayesian tree (Harmonic mean was −9624.90) is shown in Fig. 4 38 as belonging to clade R. However, their proper identification to the species level is not possible due to insufficient data resolution in GenBank for this group of photobionts. The most closely related algae identified to the species level is a free-living Trentepohlia flava (Kützing) De Toni that was isolated from bark 39

Discussion
Here we report Lecanographa amylacea as a first example of a lichen that uses either Trebouxia (Trebouxiophyceae) or trentepohlioid algae (Ulvophyceae) as primary photobiont. This is also the first example of a lichen with two very distinct photomorphs, both containing distantly related green algae. Indeed, none of the studies that previously recorded green algae switch within single lichen species suggested an impact on the thallus morphology (see introduction). The high degree of dimorphism involving alterations in the thallus anatomy and morphology was so far only known from lichens in which the same mycobiont associates with either a green alga or a cyanobacterium (e.g. Dendriscocaulon-Lobaria 5,7,8 ). Moreover, photobiont switching between different species of green algae so far mainly concerned lichens belonging to the class Lecanoromycetes. One exception is Coniocarpon cinnabarinum DC., which belongs to the Arthoniomycetes and associates with the trentepohlialean algae Printzina  Newly generated sequences are in bold, with collecting numbers of the authors following the species names. In case of specimens for which more than one sequence was generated the collecting numbers are followed with the number of sequence (i.e. seq. 1 or seq. 2). Specimen of 'Buellia violaceofusca' morphotype Thor 31385 is marked with asterisk because two completely different Trebouxia strains were identified in its thallus. GenBank accession numbers of sequences retrieved from GenBank precede the species names. Clades with photobionts from Lecanographa amylacea and Chrysothrix candelaris are highlighted. lagenifera (Hildebrand) R.H. Thompson & Wujek and one of its close relative, probably belonging to the same genus 18 . However, no impact on the thallus appearance has ever been recorded before when the closely related green algae exchange in the same lichen species. The two types of photobionts (trentepohlioid versus Trebouxia) with which Lecanographa amylacea associates are obviously responsible for the strong dimorphism observed. Once L. amylacea has gained one of the Trebouxia photobionts, the lichen species can disseminate by means of soredia that propagate jointly both symbionts and may form independent sorediate thalli previously recognized as Buellia violaceofusca. The Trebouxia photobiont species might be maintained in this sterile sorediate photomorph as this reproduction strategy should allow preservation of the relationship among symbionts. However, maintenance of the symbiotic associations seems to be rather an option than a strict consequence of joint symbiont dispersal in lichens 41 and horizontal transmissions with other lichen species are possible as proved e.g. for the genus Lepraria Ach., which disperses solely by soredia-like granules 14 .
In addition to the alteration of morphology, the type of photobiont might influence the sexuality of the mycobiont. The Trebouxia-morphotype of L. amylacea is not known to form ascomata, contrary to its trentepohlioid morphotype, which is on the other hand never sorediate. When L. amylacea is fertile, ascospores, which disseminate only the fungal partner, could switch photobionts by the capture of Trebouxia algae from other lichens or from free-living, non-lichenized Trebouxia (despite these latter appear to be less frequently encountered), leading to the subsequent development into the sorediate morphotype. Buschbom & Mueller 31 hypothesized that switches between reproductive modes are triggered through symbiont interactions. They suggested that the asexual reproductive mode is advantageous when the relationship between the mycobiont and the photobiont is optimal in a given environment allowing the rapid dispersion of both partners: in this situation, it is advantageous for the mycobiont not to switch algae. In disadvantageous environmental situations, sexual reproduction of the mycobiont is preferred because this would increase the chance to "escape" from its current partner. The sexual reproduction allows the mycobiont to acquire new partners as well as generates variability through recombination, with the development of new genotypes that may be better adapted to local and changing conditions. Similarly, Ellis & Coppins 42 suggested that the asexually reproducing crustose lichens are better adapted to stable habitats. However, it is unclear if this might also be the case for the asexual/sexual morphotypes of L. amylacea. Studies at local and global scales should be performed to better understand which might be the combination of environmental factors that drive the asexual/sexual reproduction in L. amylacea and those responsible of the photobiont switches.
A higher flexibility in the photobiont choice is considered as a strategy to increase ecological tolerance, enabling the lichen to withstand changes of environmental conditions (e.g. shifts in light or humidity regimes), to colonize diverse microhabitats, and to widen its distribution range 8,12,16,21,22 . As a consequence, it increases the population size of the fungal partner 15 , which is particularly important for the survival of a rare species such as L. amylacea. Since its description from Sweden in 1991, 'Buellia violaceofusca' (=the Trebouxia-morphotype of L. amylacea) has been recorded from several European countries 43,44 . This morphotype is even present in regions where the trentepohlioid morphotype has never been recorded (e.g. Belgium or Białowieża Forest in Poland) suggesting that the Trebouxia photobionts may offer higher fitness in some habitats or in climatically different regions. The Trebouxia morphotype might also be more easily dispersed with successful new colonization as soredia are produced in large quantities and disseminate the fungal and algal symbionts together. The trentepohlioid morphotype of L. amylacea appears to be exclusively confined to old Quercus trees 40,[45][46][47][48] , while the Trebouxia-morphotype can sometimes colonize other phorophytes, mainly Fraxinus 43,47-49 and Acer 44,50 in addition to Quercus. The use of a larger range of tree hosts is an obvious advantage that might explain the success of the Trebouxia-morphotype in some regions. In addition, the Trebouxia-morphotype has the capacity to colonize much younger Quercus trees than the trentepohlioid morphotype, the latter only occurring on Quercus over 200 years old with > 60 mm bark crevices as shown in Sweden 45,51,52 . As a consequence, the Trebouxia-morphotype has a wider ecological niche on Quercus trees compared to the trentepohlioid morphotype, with a much higher number of Quercus trees being thus suitable for colonization by the Trebouxia-morphotype 51,52 . Despite the ability to switch photobionts, L. amylacea (including 'B. violaceofusca') is rare throughout its distributional range, being often known from small populations (on a single or a few trees). With other epiphytic species occurring on old Quercus trees in semi-open landscapes, it disappears notably because of shading by developing secondary woodland 52 , but the Trebouxia-morphotype of L. amylacea appears to have the capacity to survive in more shaded conditions, and might thus replace the trentepohlioid morphotype in more dense forests. Studies on the forest types and of the environmental parameters, particularly light and humidity, which favour each morphotypes should be performed to better understand their distribution.
The identification of trentepohlioid algae in lichens are difficult on the basis of morphological features even at the generic level, especially as sexual stages of algae are usually suppressed in lichens 18 . As a consequence, we used molecular methods to investigate the identity of the trentepohlioid alga of L. amylacea (Fig. 4). Unfortunately, the taxonomic coverage in GenBank for this group of algae is still very limited and uneven, and many deposited sequences are from samples identified to the genus level or above, with many genera being polyphyletic. Brooks et al. 53 summarized the most recent molecular studies and showed five well-supported lineages within Trentepohliales, of which none correspond to the morphologically defined genera. Lichenized trentepohlioid photobionts were found within most of these clades. It is estimated that approximately 23% of lichen-forming fungi associate with trentepohlioid algae 54 . In our study, the photobiont of Lecanographa amylacea was found to be closely related to Trentepohlia spp. isolated from different Roccella species (Fig. 4), which were shown to belong to clade R closely related to strains representing e.g.  (Fig. 4). Because molecular studies only started to elucidate the identity and diversity of trentepohlialean photobionts involved in lichen symbiosis 18,38,54 , more work is needed to investigate the selectivity and specificity (as defined by Yahr et al. 12 , selectivity denotes the frequency of association among partners, whereas specificity denotes the phylogenetic range of associated partners) with this type of photosynthetic partner. Although previous studies suggested that at least some bark-inhabiting lichens may switch their trentepohlialean photobionts 18,54 , none of these studies showed photobiont switch from coccoid green algae to trentepohlioid photobiont.
The most common trebouxioid photobiont detected in sorediate morphotype of L. amylacea thalli is Trebouxia sp. TR9 previously reported only from Ramalina farinacea 19 . In one of these thalli, the same photobiont as in Chrysothrix candelaris was detected together with Trebouxia sp. TR9 (Thor 31385; Fig. 3). Such multiple algal genotypes have been previously found in a single thallus of different species 13,15,20 , which may be beneficial in terms of lichen's ability to respond to environmental changes or to occupy diverse microenvironments 22 . Moreover, we also found the Trebouxia-morphotype of L. amylacea to contain the same photobiont as found in the neighboring thallus of C. candelaris (Thor 31382 ; Fig. 3). This suggests that Lecanographa amylacea may obtain photobionts from different lichens, one of those being probably Chrysothrix candelaris. In the specimen Thor 31382, 'B. violaceofusca'-like soralia were present along the margins of the trentepohlioid morphotype of L. amylacea thalli contiguous to those of Chrysothrix candelaris, without borders between them (see also Fig. 6g,h for specimen Thor 31391). It suggests that the growing thallus of L. amylacea is able to capture the photobiont from C. candelaris by entering its neighboring thalli (so called 'horizontal transmission' allowing acquisition of new symbionts). Moreover, thalli of C. candelaris often become paler and appear to be killed in close vicinity of the 'B. violaceofusca'-morphotype thalli. Phenomenon in which the mycobiont captures the photobiont of other lichen species has been already described long ago ( 10 for Diploschistes muscorum (Scop.) R. Sant. in the Ostropales in Lecanoromycetes). However, it is known that photobionts with identical nuITS sequences are present in lichens growing within the same lichen community, without necessarily implying that the photobiont was taken over from other lichens, as free-living Trebouxia can also be captured 55 . A study of the photobionts of entire lichen communities as well as their free-living algae will be needed to identify the different photobiont sources that might explain the existence of the multiple trebouxioid-algal taxa detected in the 'B. violaceofusca'-morphotype.
Composite or joint thalli of the two morphotypes of L. amylacea have not been reported so far thus explaining why connection between the two previously recognized taxa, B. violaceofusca and L. amylacea, has never been suggested before, nor a possible placement of the former one in the Arthoniomycetes. Both morphotypes have even been used as different indicator species of old trees in several ecological studies 51,52,56 .
Lecanographa amylacea belongs to the Lecanographaceae, a family of species that are either lichenized with a trentepohlioid photobiont or that are lichenicolous 57,58 . The presence of a morph with Trebouxia algae in this family is therefore surprising. Interestingly, Lecanographa insolita Lendemer & K. Knudsen was also recently described as having a trebouxioid photobiont 59 . However, this lichen has never been found fertile and has not been sequenced, so the species was only tentatively assigned to the Lecanographaceae on basis of secondary chemistry and of the anatomy of immature ascomata observed in only one specimen. Its taxonomic affiliation should be verified by sequencing to determine whether it is a second example of trebouxioid photomorph in the family. Species of Arthoniales having unicellular green algae are otherwise only known in the Arthoniaceae (e.g. Arthonia phlyctiformis Nyl.) and the Chrysotricaceae (e.g. Chrysothrix candelaris) 58 . Molecular studies are needed to determine the identity of the green unicellular algae living in symbiosis with Arthoniales, our study being the first step in this field. L. amylacea represents an interesting model to explore the evolutionary mechanisms that generate fungal specificity and selectivity in the Arthoniomycetes and lichens in general. Amplification reactions were prepared for a 50 µl final volume containing 5 µl 10 × DreamTaq Buffer (Thermo Scientific), 1.25 µl of each of the 20 µM primers, 5 µl of 2.5 mg mL −1 bovin serum albumin (Thermo Scientific), 4 µl of 2.5 mM each dNTPs (Thermo Scientific), 1.25 U DreamTaq DNA polymerase (Thermo Scientific), and 1 µl of template genomic DNA or, in case of direct PCRs, lichen fragments. For amplification of algal DNA (Trebouxia and trentepohlioid photobionts) StartWarm HS-PCR Mix (A&A Biotechnology) was used. A targeted fragment of about 0.8 kb of the mtSSU rDNA was amplified using primers mrSSU1 and mrSSU3R 62 and a fragment of about 0.6 kb of the nuITS rDNA (ITS1 + 5.8 S + ITS4) using primers ITS1F and ITS4 63 . We also amplified a fragment of algal nuITS rDNA (ITS1 + 5.8 S + ITS4) using Trebouxia-specific primers AL1500bf 64 and ITS4M 13 and a fragment of trentepohlioid chloroplast rbcL gene using the following primers a-ch-rbcL-203-5′MPN and a-ch-rbcL-991-3′-MPN 54 . The yield of the PCRs was verified by running the products on a 1% agarose gel using ethidium bromide. PCR products were purified using Clean-Up Concentrator (A&A Biotechnology). Both strands were sequenced by Macrogen ® using amplification primers. Sequence fragments were assembled with Sequencher version 5.3 (Gene Codes Corporation, Ann Arbor, Michigan). Sequences were subjected to 'megablast' searches to verify their closest relatives and to detect potential contaminations (no contaminations were found).

Methods
Phylogenetic analyses of fungi. Because the first mtSSU sequences obtained from Buellia violaceofusca from Belgium and Poland were identical based on BLASTN search to the single mtSSU sequence of Lecanographa amylacea present in GenBank (viz. KF707650), more samples of the two lichens were collected in Sweden and sequenced. In addition, mtSSU and nuITS sequences were also generated for four more species of Lecanographa. A total of thirteen new mtSSU sequences for this study were obtained for placing the newly sequenced samples in a phylogeny of the family (Table 1). Members representing all genera currently accepted in the Lecanographaceae and for which a mtSSU sequence was available, were selected from Ertz & Tehler 57 , Frisch et al. 58 and Van den Broeck et al. 65 . In order to test if B. violaceofusca is conspecific with L. amylacea, we also sequenced a more variable locus i.e. 10 nuITS (ITS1 + 5.8 S + ITS4; Table 1). Opegrapha vulgata (Ach.) Ach. was chosen as the outgroup species for the mtSSU tree and Lecanographa atropunctata for the nuITS tree. The sequences were aligned using MAFFT v6.814b 66 within Geneious and improved manually using Mesquite 3.04. No regions were excluded from the nuITS dataset. Ambiguous regions corresponding to a total of 371 aligned sites were delimited manually following Lutzoni et al. 67 and excluded from the mtSSU dataset. These regions were mainly due to a long insertion in  70 . The TVM + I + G and the HKY + G models were selected respectively for the mtSSU and the nuITS datasets. For each of them, two parallel MCMCMC runs were performed, each using four independent chains and 20 million generations, sampling trees every 1000 th generation. Tracer v. 1.6 71 was used to ensure that stationarity was reached by plotting the log-likelihood values of the sample points against generation time. Convergence between runs was also verified using the PSRF (Potential Scale Reduction Factor), where all values were equal or close to 1.000. Posterior probabilities (PP) were determined by calculating a majority-rule consensus tree generated from the 30002 post-burnin trees of the 40002 trees sampled by the two MCMCMC runs using the sumt option of MrBayes. In addition, a Maximum Likelihood (ML) analysis was performed using RAxML-HPC2 v.8.2.10 72 with 1000 ML bootstrap iterations (ML-BS) and the GTRGAMMA model.
The RAxML tree did not contradict the Bayesian tree topology for the strongly supported branches. Therefore, only the RAxML tree is shown for the mtSSU dataset with the bootstrap support values and the posterior probabilities of the Bayesian analysis added near the internal branches (Fig. 1), whereas only the Bayesian tree is shown for the nuITS dataset with the posterior probabilities and the bootstrap support values of the RAxML analysis added near the internal branches (Fig. 2). ML-BS ≥ 70 and PP ≥ 0.95 were considered to be significant. Bayesian analyses were carried out for algal nuITS dataset using the Metropolis-coupled Markov chain Monte Carlo (MCMCMC) method in MrBayes v. 3.2.6 on the CIPRES Web Portal 68 . GTR + I + G best-fit evolutionary model was selected based on Akaike Information Criterion (AIC 69 ) as implemented in MrModelTest2 76 . Two parallel MCMCMC runs were performed, each using four independent chains and 20 million generations, sampling trees every 1000th. Tracer v. 1.6 71 was used to ensure that stationarity was reached by plotting the log-likelihood values of the sample points against generation time. Convergence between runs was also verified using the PSRF (Potential Scale Reduction Factor), where all values were equal or close to 1.000. Posterior probabilities (PP) were determined by calculating a majority-rule consensus tree generated from the 60002 post-burnin trees of the 80002 trees sampled by the two MCMCMC runs using the sumt option of MrBayes. In addition, a Maximum Likelihood (ML) analysis was performed using RAxML-HPC2 v.8.2.10 72 with 1000 ML bootstrap iterations (ML-BS) and the GTRGAMMA model.
The RAxML tree did not contradict the Bayesian tree topology for the strongly supported branches, therefore, only the Bayesian tree is shown for algal nuITS rDNA dataset with the bootstrap support values and posterior probabilities of the Bayesian analysis added near the branches (Fig. 3). ML-BS ≥ 70 and PP ≥ 0.9 were considered to be significant and are shown near the branches.
Phylogenetic analyses of trentepohlioid algae. For the trentepohlioid rbcL gene analysis, the newly generated sequences were aligned together with all sequences of Trentepohliales (200 records) available in GenBank using Seaview software 77,78 employing muscle option and excluding ambiguous positions in Gblocks 0.91b 74,75 using less stringent settings (i.e. allow smaller final blocks, gap positions within the final blocks and less strict flanking positions). The preliminary Bayesian analysis was carried out using the Metropolis-coupled Markov chain Monte Carlo (MCMCMC) method in MrBayes v. 3.2.6 on the CIPRES Web Portal 68 . Based on the tree obtained, we decided to reduce our dataset to 60 sequences using selected representatives (one or two sequences) of all main lineages. Moreover, sequences that were significantly shorter were also excluded from analyses. Terminal ends of the alignment corresponding to positions 1-321 and 990-1427 in the sequence KM464712 of Trentepohlia annulata were excluded from the alignment.
Bayesian analyses were carried out for trentepohlioid rbcL dataset using the Metropolis-coupled Markov chain Monte Carlo (MCMCMC) method in MrBayes v. 3.2.6 on the CIPRES Web Portal 68 . GTR + I + G best-fit evolutionary model was selected based on Akaike Information Criterion (AIC 69 ) as implemented in MrModelTest2 76 . Two parallel MCMCMC runs were performed, each using four independent chains and 10 million generations, sampling trees every 1000th. Tracer v. 1.6 71 was used to ensure that stationarity was reached by plotting the log-likelihood values of the sample points against generation time. Convergence between runs was also verified using the PSRF (Potential Scale Reduction Factor), where all values were equal or close to 1.000. Posterior probabilities (PP) were determined by calculating a majority-rule consensus tree generated from the 15002 post-burnin trees of the 20002 trees sampled by the two MCMCMC runs using the sumt option of MrBayes. In addition, a Maximum Likelihood (ML) analysis was performed using RAxML-HPC2 v.8.2.10 72 with 1000 ML bootstrap iterations (ML-BS) and the GTRGAMMAI model.
The RAxML tree did not contradict the Bayesian tree topology for the strongly supported branches, therefore, only the Bayesian tree is shown for trentepohlioid rbcL dataset with the bootstrap support values and posterior probabilities of the Bayesian analysis added near the branches (Fig. 4). ML-BS ≥ 70 and PP ≥ 0.9 were considered to be significant and are shown near the branches.
Phylogenetic trees were visualized using FigTree v. 1.4.2 73 . The DNA sequences generated during the current study are available in the GenBank repository (https://www. ncbi.nlm.nih.gov; see Table 1) and from the corresponding author on reasonable request.