Do Photobiont Switch and Cephalodia Emancipation Act as Evolutionary Drivers in the Lichen Symbiosis? A Case Study in the Pannariaceae (Peltigerales)

Lichen symbioses in the Pannariaceae associate an ascomycete and either cyanobacteria alone (usually Nostoc; bipartite thalli) or green algae and cyanobacteria (cyanobacteria being located in dedicated structures called cephalodia; tripartite thalli) as photosynthetic partners (photobionts). In bipartite thalli, cyanobacteria can either be restricted to a well-delimited layer within the thallus (‘pannarioid’ thalli) or spread over the thallus that becomes gelatinous when wet (‘collematoid’ thalli). We studied the collematoid genera Kroswia and Physma and an undescribed tripartite species along with representatives of the pannarioid genera Fuscopannaria, Pannaria and Parmeliella. Molecular inferences from 4 loci for the fungus and 1 locus for the photobiont and statistical analyses within a phylogenetic framework support the following: (a) several switches from pannarioid to collematoid thalli occured and are correlated with photobiont switches; the collematoid genus Kroswia is nested within the pannarioid genus Fuscopannaria and the collematoid genus Physma is sister to the pannarioid Parmeliella mariana group; (b) Nostoc associated with collematoid thalli in the Pannariaceae are related to that of the Collemataceae (which contains only collematoid thalli), and never associated with pannarioid thalli; Nostoc associated with pannarioid thalli also associate in other families with similar morphology; (c) ancestors of several lineages in the Pannariaceae developed tripartite thalli, bipartite thalli probably resulting from cephalodia emancipation from tripartite thalli which eventually evolved and diverged, as suggested by the same Nostoc present in the collematoid genus Physma and in the cephalodia of a closely related tripartite species; Photobiont switches and cephalodia emancipation followed by divergence are thus suspected to act as evolutionary drivers in the family Pannariaceae.


Introduction
Several spectacular aspects of the lichen symbiosis have come to light recently, the most surprizing for the general public and the most promising for evolutionary studies being the multiple variations of the association between the mycobiont and photobiont partners. The lichen as the icon of consensual and stable symbiosis between two very different partners ''for better and for worse'' is not the model that molecular studies have produced in recent years. Indeed, some mycobionts can incorporate several algal genotypes in their thallus [1][2][3], or even different algal species [4][5]. Several phylogenetic studies have demonstrated that photobiont switching is rather widespread [6], even in obligatory sterile taxa where both partners are dispersed together, and may occur repeatedly over evolutionary timescales [7]. Studies of the genetic diversity of both partners within a geographical context revealed that mycobionts can recruit several lineages of photobionts, allowing for ecotypic differentiation and thus for colonization of different ecological niches and distribution [6,8]. Those multiple variations in the association between the partners involved in the lichen symbiosis may take part in their evolutionary trajectory and we here address that matter for a lichen family (the Pannariaceae) in which several very different types of thalli occur together with variation in the number of photobionts involved in their construction.
The Peltigerales, a strongly supported lineage within the Lecanoromycetes, contains many well-known lichen genera, such as Lobaria, Peltigera and Sticta, within 10 families [9][10][11][12], including the Collemataceae and the Pannariaceae, two families that will be mentioned in this paper.
Within the Peltigerales, symbiosis includes two different lineages of photobionts [10]: (a) cyanobacteria mostly belonging to the genus Nostoc, or to Scytonema, Hyphomorpha and other taxa in the Scytonemataceae and Rivulariaceae; (b) green algae, mainly assigned to the genera Coccomyxa, Dictyochloropsis, Myrmecia, all belonging to the Trebouxiophyceae. The number of photobionts associated with the mycobiont provides the ground for the distinction of bi-and tripartite lichens, the latter case being much more diverse in the way of allocating space for the cyanobacteria [13][14][15]: (a) association with a single photobiont partner, either a cyanobacteria or a green algae; these thalli are bipartite and are referable to the cyanolichens or the chlorolichens, respectively [16]; (b) association with two partners, a cyanobacteria and a green algae and corresponding thalli referred to as tripartite thalli [17]; the topological organization of the partners can vary : (b1) both photobionts can be present in a dedicated layer within the thallus (chloro-cyanolichen; see [16]); (b2) the green photobiont is present in a dedicated layer within the thallus whilst cyanobacteria are confined to dedicated and morphologically recognizable organs, named cephalodia [18]; (b3) production of two different thallus types, either living independently from one another or being closely associated, one with the cyanobacteria and the other one with the green algae; these structures are referred to as « photosymbiodemes », « photopairs » or « photomorphs » and can be morphologically rather similar or very much different one from the other -in the latter case the cyanomorph has a Dendriscocaulon-like morphology [14].
Further two different types of cyanobacterial bipartite thallus can be distinguished on the basis of their response to changes in water availability [19]. A first type is characterized by thalli that swell considerably and become very much gelatinous when wet, and return to a rather brittle and crumpled condition when dry, while the second type has thalli that do not radically change when water availability varies, albeit strong changes in color can occur. The first type is associated with a homoiomerous thallus anatomy, that is absence of a specialized photobiont layer, with chains of Nostoc with thick mucilaginous walls being easily recognized and present throughout the thallus thickness, an upper cortex being absent or present; it will be hereafter referred to as the collematoid thallus type. The second type of thallus is heteromerous, that is with a usually very distinct photobiont layer present under the upper cortex (which is always present) and Nostoc (or other genera) or green algal cells compacted and assembled in clusters. Within the second group, several morphotypes can be distinguished, ranging from nearly crustose to large foliose and dendroidfruticose; the pannarioid type refers to a squamulose to foliose thallus developed over a black prothallus. Within the Peltigerales, a thallus associated with cyanobacteria can either belong to the collematoid or to other types, incl. the pannarioid type; on the other hand, thalli associated with green algae never belong to the collematoid type.
In summary, the lichen family Pannariaceae includes genera with very different thalli, easily recognized by their morphology and anatomy and behavior to water availability, the collematoid and pannarioid thalli. We here wish : (1) to examine the phylogenetic relationships of the collematoid genera Kroswia and Physma, and to examine the phylogenetic relationships of the photobiont of these two taxa (both being lichenized with Nostoc); (2) to examine the phylogenetic relationships of the collematoid, pannarioid and tripartite thalli all across the family Pannariaceae, and to establish whether a photobiont switch can be associated with the transition towards from pannarioid thalli to collematoid thalli and vice versa; (3) to examine the phylogenetical position of an undescribed species with tripartite thallus, belonging to Pannaria s. l. (foliose species with a green algae in the thallus and developing squamulose cephalodia with Nostoc over its surface) and to assess the evolutionary significance of a thallus combining a green algae and a cyanobacteria.

Taxon Sampling
We assembled material belonging to the Pannariaceae from recent field trips in Madagascar (2008)

Molecular Data
Well-preserved lichen specimens lacking any visible symptoms of fungal infection were selected for DNA isolation. Extraction of DNA followed the protocol of Cubero et al. [50]. We sequenced the ribosomal nuclear loci ITS, using primers ITS1F [51] and ITS4 [52], and LSU with primers LR0R [53] and either LR7 [53] or LIC2044 [54], the mitochondrial ribosomal locus mtSSU, using primers SSU1 and SSU3R [55], and part of the protein-coding gene RPB1 with RPB1AF [56] and RPB1CR [57]. We sequenced the 16S ribosomal region of the Nostoc symbiont of 25 of this set of Pannariaceae as well as 2 additional Fuscopannaria leucosticta, 2 additional Physma and 4 from two other genera (Leptogium and Pseudocyphellaria) belonging to the Peltigerales, using the two primer pairs fD1 [58]-revAL [17] and f712 [59]-rD1 [58]. Amplicons were sequenced by Macrogenß or by the GIGA technology platform of the University of Liège.     concatenated the different loci. As several species are represented by sequences obtained from specimens collected in the different parts of the world, mostly with ITS, we further assembled a 3 loci dataset excluding ITS. We thus produced three matrices, two for a large sampling of the Pannariaceae including our target taxa (Kroswia, Physma and the undescribed species with a tripartite thallus), including the four loci 5.8S, mtSSU, LSU and RPB1 or including only the latter three, and one with the Nostoc 16S data.
For the concatenated analysis of the four loci, we partitioned the data in different subsets to optimize likelikood. We used PartitionFinder [65] to choose the best partition and determine the best models for the different subsets. We used BIC as the criterion to define the best partition, and compared all models implementable in MrBayes [66]. The partition tested for the analysis on the four loci was composed of 6 subsets: RPB1, 1 st codon position, RPB1, 2 nd codon position, RPB1 3 rd codon position, mtSSU, LSU, 5.8S. For the 16S analysis on Nostoc, we used MrModelTest version 2.3 [67] to determine the best model.

Maximum Likelihood and Bayesian Phylogenetical Analyses
For each matrix, we produced the best likelihood tree and bootstrapped for 1000 pseudoreplicates in the same run using RAxML version 7.4.2 [62][63] with the default settings and the GTRCAT model. We further ran a Bayesian analysis using MrBayes version 3.1.2 [66]. Each analysis consisted of 2 runs of 3 heated chains and 1 cold one. We assessed the convergence using Tracer version 1.5 [68] and stopped the runs after checking with AWYT [69] that convergence was reached for each run and that tree topologies have been sampled in proportion of their true posterior probability distribution. The analysis for the family Pannariaceae was stopped after 15610 6 generations, the analysis on Nostoc 16S after 37610 6 generations.

Ancestral State Reconstruction
We reconstructed ancestral character states using SIMMAP version 1.5.2 [70], with default settings, on the consensus Bayesian tree produced by the MrBayes analysis on the Pannariaceae 4 loci concatenated dataset, as well as on a subset of 20 trees (10 from each run of the Bayesian analysis) and with Mesquite version 2.75 [71][72] using the likelihood parameters and the default settings, calculating the average probabilities of the ancestral states based on the same subset of 20 trees.
We also used BayesTraits version 1.0 [73] on a set of 2 trees: the best tree produced by the ML analysis on the Pannariaceae 4 loci concatenated dataset and on the best tree of the concatenated analysis without 5.8S, as they were slightly different, to constrain some branches (ancestors) to be to a certain state. We compared the harmonic mean of the iterations, which is an approximation of the marginal likelihood of the model, calculating the Bayes Factor, which is twice the difference of likelihood between the models, with each state of ancestor, to see which state of the ancestor leads to the best likelihood of the model. A positive Bayes Factor suggests that the first character state tested has a better likelihood than the second one, and a Bayes Factor above 2 is considered significant (Bayestraits Manual, available at http://www. evolution.rdg.ac.uk/BayesTraits.html). We used reversible jump and a gamma hyperprior whose mean and variance vary between 0 and 10. We ran the program for 50610 6 iterations for each constrained state. The character reconstructed was the type of thallus, and the character states considered were tripartite, pannarioid bipartite and collematoid bipartite.

Topological Tests
We tested different tree topologies on the concatenated dataset of 4 loci for the Pannariaceae. We generated 8 constrained best trees with RAxML, with the same settings as above, and using the following constraints: (1) the 3 accessions of Kroswia forming a monophyletic group; (2) Kroswia as a monophyletic group basal to a group formed by Fuscopannaria ahlneri, F. confusa, F. leucosticta and  We computed the likelihood of 100 trees (the best constrained tree, the best unconstrained tree and a random sample of 98 bootstrap replicate trees from the unconstrained analysis), estimating parameters on a NJ tree, using an HKY model with a gamma rate of heterogeneity and 4 gamma categories (parameters choice and methodology suggested by [74]). We performed the 1sKH test [75][76][77], the SH test [75] and the ELW test [78] on the constrained tree using TreePuzzle v. 5.2. [79]. Due to its very low power (see for instance [74]), we did not consider the results of the SH test.

Molecular Data
We amplified ITS, mtSSU and RPB1 for all 36 selected specimens, except one for RPB1. We amplified LSU for 21 specimens, all 15 negative results being resolved in a single clade comprising all accessions of Physma, the Parmeliella mariana gr. (P. brisbanensis, P. mariana and P. stylophora), Parmeliella borbonica and the undescribed tripartite 'Pannaria' R969 (here annotated the tripartite R969). Wedin et al. [19] could amplify the LSU loci for three species of Physma, but, for unknown reasons, all our attempts to amplify LSU for this clade failed.

Matrix Assemblage and Concatenation
For the analysis on the Pannariaceae mycobiont, we could include the following newly sequenced specimens: 21 specimens with all 4 loci, 14 with 3 loci (lacking LSU) and 1 specimen with 2 loci (lacking LSU and RPB1). We added 46 taxa retrieved from GenBank to complete our sampling, 39 members of the Pannariaceae, and 7 outgroup taxa all belonging to the Peltigerales (3 Vahliellaceae, 1 Collemataceae, 1 Placynthiaceae, 1 Peltigeraceae). Those included either the 4 loci or a subset of them. Detailed information can be found in table 1. For the 16S dataset on Nostoc, we produced 36 new sequences; we added 93 Nostoc sequences retrieved from GenBank, chosen either on the phylogenetic position of their fungal partner or their nucleotide similarity to our sequences, based on megaBLAST searches [60], and 14 outgroup sequences, belonging to other genera, to complete our sampling.

Partitioning and Model Selection
For the analysis on the Pannariaceae mycobiont, PartitionFinder divided the partition in 4 subsets: one composed of RPB1 1 st and 2 nd codon positions with LSU, one with mtSSU only, one with 5.8S only and one with RPB1 3 rd codon position only. For the first subset, the model selected was GTR+I+G, as well as for mtSSU and RPB1 3 rd codon position; for 5.8S, the model selected was K80+I+G. For the analysis on the Nostoc 16S dataset, the model selected was GTR+I+G.

Phylogenetic Analyses
The 50% Bayesian consensus tree of the analysis of the Pannariaceae mycobiont dataset comprizing 4 loci is presented in Figure 1, with the bootstrap values of the ML analysis and the Bayesian PP values written above the branches. The same consensus tree obtained with the 3 loci dataset is available in the Supplementary Material ( Figure S1). The 50% Bayesian consensus tree of the analysis of the Nostoc 16S dataset is presented in Figure 2, with the bootstrap values of the ML analysis and the Bayesian PP values written above the branches.
Phylogeny of the Family Pannariaceae (Fig. 1) Topology of the family. The analysis of the 3 and 4 loci datasets yielded the same topology, albeit with less support for some branches for the former; as expected the 5.8S loci provides an interesting resolution power to discriminate branches at the generic and infrageneric level. We retrieved the Pannariaceae as a monophyletic group, divided into two strongly supported clades: the first one includes all Parmeliella accessions, incl. the genus type P. triptophylla, except for the P. mariana group and P. borbonica which are resolved with strong support in the other clade. The so-called Parmeliella s. str. clade further includes Degelia (here resolved as polyphyletic, as already detected by Wedin et al. [19]), Erioderma, Leptogidium and the monotypic Joergensenia which represents the only tripartite species in this clade. The second clade can be divided into three groups: (1) the first one is not supported in ML optimization but gets a PP = 0.95 in the Bayesian analysis; it is composed of Xanthopsoroma, Physma, the Parmeliella mariana group, Parmeliella borbonica and the tripartite species R969, and will be referred to as the Physma group; (2) a group not supported in ML optimization but getting a PP = 0.94 in Bayesian analysis, composed of Pannaria, Staurolemma, Ramalodium, Fuscoderma, Psoroma and Psorophorus, that will be referred to as the Pannaria group; and finally (3) a group composed of Fuscopannaria, Kroswia, Protopannaria, Leciophysma and Parmeliella parvula, that will be referred to as the Fuscopannaria group.
Wedin et al. [19] and Spribille and Muggia [10] retrieved the Parmeliella s. str. group, the Pannaria group and the Fuscopannaria group with similar topology as ours. However, in their studies, their single or multiple accessions of Physma is or are nested within the Pannaria group. With our dataset, which includes a larger sampling of Physma and representatives of the closely related Parmeliella mariana gr., P. borbonica and the tripartite R969, the hypothesis of the whole Physma group nested in the Pannaria group and the Fuscopannaria group as basal is strongly rejected by two topological tests (ELW and 1sKH tests; see table 2).

Monophyly of Several Genera
Our accessions of Kroswia crystallifera (the type species of the genus; [27]) gathered in Madagascar and Reunion are not resolved as a monophyletic group: they are nested within Fuscopannaria, and closely related to its type species F. leucosticta [38]. Even with the exclusion of species now referred to Vahliella [12,80], the genus Fuscopannaria is not resolved as monophyletic, unless F. sampaiana is excluded and Kroswia crystallifera included. Two strongly supported clades can be distinguished if the genus is so recircumscribed: one with F. ignobilis and F. mediterranea and the other with the type species and Kroswia crystallifera.
Pannaria is resolved as a diverse but nevertheless well-supported genus, including several tripartite species formally placed in the genus Psoroma and which were transferred to Pannaria following the detailed studies by Elvebakk [81][82][83][84], Elvebakk & Bjerke [85], Elvebakk & Galloway [86] and Elvebakk et al. [17]. Interestingly, our single accession of the tripartite Pannaria-like R969 is not resolved amongst other tripartite Pannaria but within the Physma clade with strong support. It therefore appears that the tripartite Pannaria-like species are more diverse than expected and that the tripartite habit is widespread amongst the Pannariaceae, being absent only in the Fuscopannaria group. Two recently described and tripartite genera Xanthopsoroma and Psorophorus, segregated from Psoroma [87], are retrieved as a part of the Physma gr. with support only in the Bayesian analysis for the former, and as sister to Psoroma s. str. in the Pannaria group for the latter.
Parmeliella (type species: P. triptophylla) is a well-supported monophyletic group if the Parmeliella mariana gr., Parmeliella borbonica and P. parvula are excluded. The latter is resolved with strong support within the Fuscopannaria gr. whilst the others are resolved within the Physma group, on a long and strongly supported branch. Further, P. borbonica appears nested inside Physma, which is therefore paraphyletic.
Nostoc Phylogeny (Fig. 2) We defined phylotypes (A to G) on the Nostoc tree based on wellsupported monophyletic groups containing sequences from our representatives of the Pannariaceae family. All our sequences are part of Nostoc clade 2 (sensu [59,88]) except phylotype G, which seems related to Nostoc clade 3 sensu Svenning et al. [59].
There is no evidence suggesting coevolution or cospeciation events between the mycobiont and the photobiont. The phylogeny of Nostoc involved in the lichen symbiosis does not match the phylogeny of the Pannariaceae.
Topological Uncertainties (Table 2) he tests do not reject the monophyly of Kroswia, either its position outside of the polytomy including i.a. Fuscopannaria leucosticta and F. praetermissa, although the difference of likelihood with the best unconstrained tree is relatively high (13.68). However, the position of Kroswia outside of Fuscopannaria s. str. (including F. mediterranea and F. ignobilis) is significantly rejected by the ELW and 1sKH tests. Therefore Kroswia crystallifera should be considered as part of Fuscopannaria. Concerning the position of the tripartite R969, the topological tests do not reject its position at the base of the Physma group as a whole. However, its position at the base of the Parmeliella mariana gr., with Physma basal to both of them, is significantly rejected by the ELW and 1sKH tests.
Concerning the position of Parmeliella borbonica, the topological tests do not reject its position neither as basal to Physma, nor as basal to the Parmeliella mariana gr., with Physma basal to both of them, although the difference of likelihood for the latter case is relatively high (10.29). We consider that the weak resolution of the test regarding the position of Parmeliella borbonica might be due to a large amount of missing data as only 2 loci are available for this accession, reducing its impact on the likelihood of the trees. More material should therefore be studied before the taxonomic status of P. borbonica can be revised.
As commented above, we also tested the topology proposed by Wedin & al. [19] and Spribille & Muggia [10] where their accessions of Physma are resolved within the Pannaria gr. Such a topology is rejected on our dataset by the ELW and 1sKH tests. (Fig. 1, Table 3)

Reconstruction of Ancestral States
Results of the SIMMAP reconstructions on the Bayesian consensus tree are shown in pie charts on Figure 1. Results of the BayesTraits and Mesquite reconstructions, as well as the SIMMAP reconstruction on 20 trees are shown in table 3.
Even though the probability values can vary quite widely from a reconstruction method to the other, the same ancestral character state is recovered for most branches.
For the Fuscopannaria group, a pannarioid ancestor is strongly supported, incl. for the Fuscopannaria s. str. clade (all Fuscopannaria except for F. sampaiana). Within the Pannaria group, two deep nodes are recovered with a tripartite ancestor (the unresolved clade with all accessions of Pannaria, and the clade including Fuscoderma, Psoroma and Psorophorus) as well as the node supporting the whole group. The node supporting both groups (the Fuscopannaria and the Pannaria gr.) also has tripartite thallus as the most likely ancestral type. For the clade comprizing Physma, the Parmeliella mariana gr., P. borbonica and the tripartite R969, reconstructions favor a pannarioid ancestor without much support, except the Bayes Factor that slightly favors a tripartite ancestor. However, for the whole group and thus including both accessions of Xanthopsoroma, reconstructions recover a tripartite ancestor with strong support. The node supporting the three groups (Fuscopannaria-, Pannaria-, and Physma-group) has most likely a tripartite thallus, as recovered by all four methods. The Parmeliella s. str. group most probably had a pannarioid ancestor, as well as the family Pannariaceae.

Discussion
Nostoc from Collematoid and Pannarioid Thalli (Fig. 2) Thalli belonging to the collematoid or pannarioid types never share the same Nostoc phylotype. Phylotypes A, E and F only contain symbionts from collematoid thalli. Moreover phylotype F also contains symbionts associated with the lichen genus Leptogium, a typical representative of the collematoid type, these accessions being resolved in a strongly supported clade together with the Kroswia symbionts. Phylotype E includes the photobiont of several Physma accessions together with that of the cephalodia of the tripartite R969, and these cephalodia have the same homoiomerous structure as the thallus of Physma byrsaeum (Fig. 3a, c).
Phylotypes B, C, D and G only contain symbionts from pannarioid thalli. Phylotype B which contains the photobiont of our accession of the terricolous Fuscopannaria praetermissa is closely related to sequences from terricolous-muscicolous Nephroma arcticum photobionts whereas phylotypes C and D contain Nostoc sequences from epiphytic Lobaria, Nephroma and Pseudocyphellaria, along with our accessions of epiphytic Pannariaceae with pannarioid thalli. This confirms that Nostoc from epiphytic heteroimerous thalli cluster together, although they group in a polyphyletic assemblage of different phylotypes [17,89,90]. These data strongly suggest that many pannarioid thalli share Nostoc strains between them and with other representatives of the Peltigerales that also have Nostoc in a well-defined thin layer. Furthermore collematoid thalli can share Figure 2. Phylogenetic relationships in the genus Nostoc, based on the best ML tree of the analysis on the 16S dataset. Values above branches represent ML bootstrap and Bayesian PP values, respectively. Names in bold are those for which DNA sequences were produced for this study. Color boxes represent phylotypes containing our sequences and defined by well-supported monophyletic groups. Colors in the taxa names represent the type of the thallus containing the Nostoc: in green tripartite thalli, in red pannarioid thalli and in blue collematoid thalli. Taxa names refer to the host of the Nostoc symbionts, when available.    Nostoc with representatives of the Collemataceae that also have Nostoc chains throughout their thallus. These results strongly suggest that the thallus type (collematoid versus pannarioid), and the organization of the Nostoc cells inside it, depend on the phylotype of the Nostoc with which the mycobiont associates. Therefore, it seems that in the family Pannariaceae, the Nostoc associated with the mycobiont would have more impact on the morphology of the thallus formed than the phylogenetic origin of the mycobiont. The corollary might be true as well, the Nostoc selection by the mycobiont is more affected by the morphological and ecophysiological characteristics of the association than by the phylogenetic position of the mycobiont. Extracellular polysaccharides substances (EPS) produced by many bacterial lineages, incl. cyanobacteria, are involved in the physiological and ecological characteristics of those organisms [91]; in Nostoc, the biochemistry and structure of the dense sheath of glycan strongly participate in the dessication tolerance of Nostoc commune [92]. Although no clear evidence is available, we suspect that variations in the glycan sheath characteristics amongst the various strains of Nostoc involved in the lichenization events within the Pannariaceae drive the differences between the collematoid and the pannarioid thallus types.
Occurrence of Collematoid Thalli All across the Pannariaceae (Fig. 1) We found collematoid thalli in the four main groups of the family. Kroswia and Leciophysma appear as part of the Fuscopannaria group, Kroswia being nested within Fuscopannaria s. str., excluding F. sampaiana; Staurolemma and Ramalodium are part of the Pannaria group and Pannaria santessonii was described as a collematoid thallus species; Physma is in the Physma group, along several taxa with pannarioid thalli; and finally Leptogidium is part of the Parmeliella s. str. group. These results suggest that thalli switched from pannarioid type to collematoid and possibly vice versa several times along the evolutionary history of the family.
These results also suggest that the thallus type organized by the association between a mycobiont and a photobiont is primarly driven by the identity of the latter, the Nostoc phylotype with which it associates rather than by the phylogenetic identity of the mycobiont. Indeed, unlike the original assumption that all collematoid thalli were part of the Collemataceae and all pannarioid thalli were part of the Pannariaceae, many collematoid thalli are actually members of the Pannariaceae, as already detected by Wedin et al. [19] and Otálora et al. [35]. Moreover, they do not form a monophyletic group inside the Pannariaceae, but are present all across the family, suggesting the absence of phylogenetic pattern of the mycobiont related to the collematoid morphological and anatomical thallus type.

Evidence for Coincidence between Photobiont Switch and Change of Thallus Type
The most spectacular and straightforward example lies with the type species of Kroswia which is nested inside Fuscopannaria s. str.: it exhibits a drastic change of morphology (see figure 3d-e) of the thallus (all representatives of this genus so far have typical pannarioid thalli), and it associates with a Nostoc phylotype (phylotype F) that is totally different from the one associating with the closely related Fuscopannaria leucosticta (phylotype D). Moreover, phylotype F has also been found associated with the typically collematoid Leptogium lichenoides. The duo Kroswia/ Fuscopannaria thus provides the best example of the influence of the Nostoc on the shape of the thallus. Actually, K. crystallifera is a species of Fuscopannaria with little genetic divergence with its related species such as F. leucosticta and F. praetermissa; this divergence however precludes any assumption that it could be considered as a photomorph of one of them. Its thallus is dramatically different because it switched to a different Nostoc, one that triggers the collematoid format for the thallus. Jørgensen [24], when studying the apothecia characters of the other species assigned to that genus (K. gemmascens), concluded that ''the characters of the hymenium and the chemistry of the thallus certainly place it close to Fuscopannaria (…)''. Quite interestingly another photobiont switch can be postulated in that group as the phylogenetic position of Moelleropsis nebulosa as sister to F. leucosticta has been retrieved by Ekman & Jørgensen [93] and more recently announced as confirmed [94]. This species exhibits granulose thalli with clusters of Nostoc interwoven and covered by short-celled hyphae and very much different from the pannarioid thallus type, and thus most probably associated with a different Nostoc phylotype.
Occurrence of Tripartite Thalli All across the Pannariaceae (Fig. 1) We could detect tripartite thalli in all main groups within the family, except in the Fuscopannaria group. This absence might be caused by incomplete sampling as the only tripartite species known in Fuscopannaria (F. viridescens, associated with a green algae and producing cephalodia; [95]) as well as both species of Degeliella (forming tripartite thalli; [42]) could not be included in our dataset. Psoroma, Psorophorus and the tripartite representatives of Pannaria are resolved in the Pannaria group, Xanthopsoroma and the tripartite R969 belong to the Physma group, and the characteristic Joergensenia is included in the Parmeliella group. Until the seminal papers by Elvebakk & Galloway [86] and Passo et al. [96], all tripartite Pannariaceae were assigned to a single genus (Psoroma) assumed to form a monophyletic group. Within the three main groups of the Pannariaceae where they are resolved, the species with tripartite thalli are mixed up with species with bipartite thalli, mainly of pannarioid type but also with collematoid type. These results suggest that several times through the history of the family, mycobionts switched from a tripartite to a bipartite thallus or vice versa.

Evidence for Cephalodia Emancipation
Switches from a tripartite to a bipartite thallus may involve the cephalodia and their emancipation from their green algaecontaining thalli. Although cephalodia are usually associated with rather small, firmly attached, or even included, structures, there are many examples of tripartite Pannaria and Psoroma in which cephalodia are large and easily detached, or proliferating and developing large squamules that can be easily detached from their ''host'' thalli (examples in [17,81,97,98]). The cephalodia of the tripartite R969 start their development as modest blue gray squamules over the thallus, but eventually grow up to 0.7 cm across and develop a foliose habit with denticulate to deeply lobulate margin (see figure 3a).
More interestingly, the Nostoc photobiont in several accessions of Physma byrsaeum (annotated R1, R2, R2846 and R2847; phylotype E) is very closely related to the one found in the cephalodia of the tripartite R969. As the latter is basal to the clade containing all accessions of Physma, it can be postulated that several species belonging to this genus arose from cephalodia emancipation from their common ancestor. Indeed, the common ancestor of the whole Physma clade is recovered as producing tripartite thallus. Furthermore, the disposition of the Nostoc cells inside the cephalodia of R969 is similar to the one inside Physma thalli (see figure 3a-c): they are enclosed in ellipsoid chambers delimited by medulla hyphae, these structures being responsible for the maculate upper surface of thalli (Physma) or cepahodia (R969).
Besides the tripartite R969, the clade included both accessions of the recently described genus Xanthopsoroma [87], which also develops tripartite thalli, with a green algae as the main photobiont and Nostoc included in cephalodia. The three species recognized within the Parmeliella mariana gr. may have arisen from cephalodia emancipation of their common tripartite ancestor or from a photobiont switch from a Physma ancestor. Quite interestingly, the pannarioid Parmeliella borbonica, nested within Physma, is associated with phylotype D of Nostoc, shared by most accessions of the Pannaria and Parmeliella s. str. groups (as well as other distantly related species of the Peltigerales), and not phylotypes C or G, chosen by all our accessions of its closely related species of the Parmeliella mariana gr. When excluding both accessions of Xanthopsoroma, the Physma gr. is a well-supported clade on a long branch and includes a tripartite species, species with pannarioid as well as collematoid thalli. The long branch may indicate that our sampling is too scarce and geographically too restricted. However, as both Physma and the Parmeliella mariana gr. have a pantropical distribution, we can confidently assume it would not collapse in future studies.
In figure 4, we illustrate the different possible scenarios to switch from tripartite to bipartite, and from collematoid to pannarioid thalli and vice versa, and emphasize on the possibility to obtain, with switches and time, the three types of thalli from the same tripartite ancestor.
As a matter of fact, earlier workers came close to the conclusion that cephalodia can emancipate and start their own evolutionary trajectory. Ekman & Jørgensen [93] pointed to the « homology » between the cephalodia of the green algae-containing Psoroma hypnorum and the thallus of the cyanobacterial autonomous species Santessoniella polychidioides; Passo et al. [96] retrieved the latter as sister to Psoroma aphthosum, a green algal species with coralloidsubfruticose cephalodia, very much akin the thallus of Santessoniella polychidioides. We strongly suspect this case represents a further case of cephalodia emancipation, and subsequent divergence. This scenario implies that emancipated cephalodia can reproduce sexually as most species of Physma and Santessoniella polychidioides produce apothecia and well-developped ascospores. There is indeed no reason to believe that thalli newly formed by cephalodia emancipation and containing only Nostoc as photobiont would not be able to produce apothecia, as only the mycobiont is involved in such formation. An interesting alternative would be that, when expelled out of the ascus, the ascospore produced by the mycobiont involved in the ancestral tripartite thallus, would collect or recapture the Nostoc of the cephalodia.
Several representatives of the Lobariaceae produce photomorphs, mainly within the genera Lobaria and Sticta [14,99]. These duos involving the same fungus lichenized either with a green algae or with a Nostoc comprize thalli morphologically rather similar or not (see Introduction), and living attached (thus forming tripartite thalli) or not. Although molecular studies on these duos have mainly sought to demonstrate the strict identity of the fungus involved in each part, the separation or ''living apart'' of one from the other has long been recognized in several taxa, such as Lobaria amplissima and its cyanomorph Dendriscocaulon umhausense and Sticta canariensis and its cyanomorph S. dufourii [100]. There is a priori no reason to exclude that the duos can separate on ''a permanent basis'' and thus emancipate; each morph would eventually run its own evolutionnary trajectory, as recently suggested for divergence patterns in Sticta photomorphs [101]. Such a scenario can be interpreted as a variant of cephalodia emancipation as advocated here for the evolution of thallus types within the Pannariaceae.
The alternative scenario for the complex phylogenies including bi-and tri-partite thalli implies that a cyanolichen would capture a green algae from the environment (or from another lichen), adopt it as its main photobiont and confine its Nostoc into cephalodia. This hypothesis has been suggested by Miadlikowska & Lutzoni [32] for the sect. Peltidea in the genus Peltigera but so far has not been confirmed. Our data and reconstruction of ancestral state do not support it in the Pannariaceae, with a possible exception for Joergensenia cephalodina, but a better sampling is needed in that group to reconstruct the ancestral states.

Conclusions and Perspectives
Field observations of the lichen species belonging to the widespread and well-known order Peltigerales on the tiny and remote island of Reunion in the Indian Ocean instigated our studies on the relationships between photomorphs in the Lobariaceae (14) and the present study on the Pannariaceae. Indeed, we were intrigued by the occurrence, several times at the same locality or even on the same tree, of representatives of that family with collematoid and pannarioid thalli, and more locally of tripartite thalli.
Collematoid and pannarioid thalli are represented throughout the Pannariaceae. Each thallus type mostly appears mingled within complex topologies. Switches between those thallus types are thus frequent throughout the family. We could demonstrate that both collematoid genera in the Pannariaceae we examined from Reunion material (Kroswia and Physma) are involved in photobiont switches. We suspect that such a scenario could be detected elsewhere in the Pannariaceae and may act as an important evolutionary driver within the whole family, and perhaps elsewhere within the fungi lineages containing lichenized taxa.
The tripartite thallus type is shown to be the ancestral state in the clade we could study (the Physma gr.). Although a larger sampling is needed before such an result could be confirmed, we can postulate that cephalodia emancipation and subsequent evolutionary divergence is the most likely scenario within that clade. The data available support the same scenario in other clades of the Pannariaceae, and it can be suspected in the Lobariaceae where it is represented by the separation and subsequent divergence of photomorphs.
The photomorph pattern in the Lobariaceae demonstrates that a single mycobiont can recognize and recruits phylogenetically unrelated photobiont partners and these associations result in morphologically differentiated thalli. We show here that the use of different lineages of Nostoc or the association with only one partner instead of two might lead to the same consequences. Recognition of compatible photobiont cells is carried out by specific lectins produced by the mycobiont, characterized by their ligand binding specificity [102]. Peltigera species have served as models in the studies of lectins and their involvment in the recognition of symbiotic partners [103][104][105][106]. A lectin detects compatible Nostoc cells at the initiation of cephalodium formation in P. aphthosa and this process is highly specific [107], as further demonstrated by experiment of inoculation of several Nostoc strains into the cephalodia of the same species [108]. The biochemical process sustaining the recognition of both partners in two lichen species associated with green algae has been elucidated by Legaz et al. [109] and extended to cyanolichens with collematoid thalli by Vivas et al. [110]. The genes coding for two lectins assumed to be involved in photobiont recognition have recently been identified [111][112]. Evaluation of the variation of those genes is of tremendous interest in the context of photobiont switching and cephalodia emancipation as lectins have been shown to be under selection pressure by the symbionts in corals [113][114] and a coevolutionary process could thus be highlighted and demonstrated in lichenized fungi. A preliminary study with Peltigera membranacea material from Iceland could demonstrate a significant positive selection in LEC-2 but not due to variation in photobiont partner [112].
Further research should thus assemble larger dataset of tripartite taxa within the Pannariaceae and reconstruct their evolutionary history, especially as to the fate of their cephalodia. Numerous methods for detecting genes under positive selection are available [115] and could be applied to the Pannariaceae. Genomics studies of lectins associated with photobiont recognition on tripartite taxa as well as those involved in obvious photobiont switches (pannarioid to collematoid and vice versa) could therefore bring to light a nice model of coevolution [116].
The taxonomical consequences of these results are published in a companion paper, dedicated to new taxa and new combinations.

Data Accessibility
All newly produced sequences are deposited in GenBank. All matrices used in the analyses are deposited in Treebase.