Deep divergence between island populations in lichenized fungi

Macaronesia is characterized by a high degree of endemism and represents a noteworthy system to study the evolutionary history of populations and species. Here, we compare the population-genetic structure in three lichen-forming fungi, the widespread Lobaria pulmonaria and two Macaronesian endemics, L. immixta and L. macaronesica, based on microsatellites. We utilize population genetic approaches to explore population subdivision and evolutionary history of these taxa on the Canary Islands, Madeira, Azores, and the western Iberian Peninsula. A common feature in all species was the deep divergence between populations on the Azores, a pattern expected by the large geographic distance among islands. For both endemic species, there was a major split between archipelagos. In contrast, in the widespread L. pulmonaria, divergent individuals were distributed across multiple archipelagos, suggesting a complex evolutionary history involving repeated migration between islands and mainland.

Microsatellite repeat data. Highly variable fungal microsatellite loci were utilized to quantify the regional population genetic structure in Macaronesia (for sampling sites, see Fig. 1). Diversity statistics are reported in Table 1. For L. immixta, population trees indicated that populations on the Canary Islands and Madeira were related to one another, but also isolated, as indicated by them being occupied by separate clades (Fig. 2A,D). In total, 13 genetic clusters were found by Bayesian analysis of population structure in BAPS. The five clusters found on the Canary Islands and Madeira were closely related, but only one cluster was found on Madeira, and it was not shared with the Canary Islands. Sites on the Azores were occupied by eight somewhat more divergent clusters, comprising three major groups. The first group of clusters was widespread throughout the Azores, while the remaining two groups were exclusively found in the central Azores islands. The single site on the Iberian Peninsula grouped with one of the two cluster groups distributed in the central Azores ( Figs. 2A,D, 3A). Furthering these patterns, results of Discriminant Analysis of Principal Components (DAPC) revealed a hierarchical steppingstone like model for microsatellite variation of Lobaria immixta between islands (Fig. 4A). This highlights gene flow between adjacent islands (Canary Islands & Madeira), with minimal gene flow between these groups and the Azores cluster. The Iberian genotypes grouped with those from the Azores (Fig. 4A). Of all three species investigated, L. immixta exhibited the lowest allele counts, e.g. with respect to private alleles (Fig. 4D).   Fig. 2B,E) and separate groups of genetic clusters from Bayesian analysis of population structure in BAPS (Fig. 3B). In total, 19 genetic clusters were found. Similar to the pattern in L. immixta, in L. macaronesica, the largest genetic difference resolved with BAPS was between the Azores vs. Canary Islands and Madeira. Other than in L. immixta, sites on Madeira and the Canary Islands hosted dissimilar groups of genetic clusters. Sites on the Azores harbored three distinct groups of genetic clusters with Eastern, central, and Western distributions on that archipelago. The site on the Iberian Peninsula grouped with sites on the Canary Islands (Figs. 2B,E, 3B). DAPC clustering reinforced the presence of highly distinct genetic clusters for the Azores, Madeira, and the Canary Islands. The Iberian genotypes grouped within the Canarian cluster of genotypes (Fig. 4B). Lobaria macaronesica had about twice as many private alleles on each archipelago as L. immixta (Fig. 4E).
In L. pulmonaria, the pattern was somewhat more complicated than in the endemic species. A total of 28 genetic clusters were found, reflecting the higher number of sites analyzed in this species and the more extensive (but not rangewide) geographic sampling. The site in Newfoundland, Canada showed some similarity to sites on the Iberian Peninsula and the single site investigated in Ireland showed similarity to the Azores (Fig. 2C,F). One cluster group was shared between the Azores and Canary Islands (Gran Canaria), and another between the Azores and Madeira (Fig. 3C). However, other than in the endemic species, in L. pulmonaria, individuals belonging to more or less divergent genetic groups of clusters co-occurred on all archipelagos, indicating a more complicated evolutionary history, with repeated geneflow between archipelagos. Nevertheless, some groups of clusters had the tendency to be more frequent on a single archipelago or were restricted to a single archipelago.  All species showed significant isolation by distance (Fig. 5), as indicated by highly significant relationships between pairwise F ST values and geographic distance in linear models. The relationships were strongest and explained the most variance in Lobaria immixta (R 2 = 0.48), followed by L. macaronesica (R 2 = 0.38). Lobaria pulmonaria had a statistically significant, but somewhat weaker signal of isolation by distance, where the geographic distance explained only 7% of the variance in pairwise F ST data.

Discussion
Comparison of genetic patterns across species. A unifying motive in the microsatellite repeat data of Lobaria pulmonaria and its two endemic relatives, L. immixta and L. macaronesica, was that all three species showed high divergence between archipelagos and significant isolation by distance. This pattern was expected considering the large geographic distance between archipelagos and the lack of stepping-stone habitats characteristic of oceanic island biota. Overall, the population models in DAPC provided evidence for a hierarchical island model. This is consistent with the fact that the studied archipelagos were far apart, supporting divergence, but some islands in closer proximity had the opportunity for occasional gene flow. Previous studies comparing Swiss and Canadian populations of L. pulmonaria have found substantial differentiation, as expected in populations situated on different continents 17 . So, not surprisingly, the occurrence and frequency of gene flow appears to decrease with geographic distance in L. pulmonaria, which is also evident in our data on isolation by distance. Significant differentiation over large geographic scales has been shown in other lichen fungi as well [18][19][20][21][22][23][24] . www.nature.com/scientificreports/ Taken together, our data indicate that L. pulmonaria has a complex population history in Macaronesia, the coexistence of divergent groups of genetic clusters on the same archipelago suggesting recurrent immigrations from sites on the European continent, but less from the North American continent. By comparison, the population histories of the two endemic species were more straightforward. Homogeneity of groups of genetic clusters within archipelagos rejects the hypothesis of repeated recent migration among archipelagos in the endemics. For both endemic species, each archipelago contained numerous private alleles. Thus, the populations on archipelagos evolved by genetic drift and/or by accumulation of new mutations, causing divergence.
Recent studies of lichen fungi and bryophytes emphasize the importance of Macaronesia for the colonization of sites in continental Europe. First, a study of lichen fungi in the genus Nephroma inferred the colonization of mainland sites from sites on the Macaronesian islands 4 . Second, an investigation of a liverwort showed that (1) sites in Macaronesia hosted a hidden diversity in the liverwort comparable to the radiation in higher plants at the genus level, and (2) low-diversity sites in western Europe were recolonized from Macaronesian refugial populations 25 . This diversity pattern was not evident in L. pulmonaria (the only species with more occurrences outside of than within Macaronesia)-sites on the mainland were highly diverse -see also 26 -and were differentiated from island sites. If mainland sites received L. pulmonaria from the islands, this migration must have occurred a long time ago because the observed substantial differentiation of the large island and mainland populations would require that many generations have passed.
In contrast, for the lichen Parmelina carporrhizans, unidirectional gene flow to the Macaronesian Islands was inferred 27 . Our data for Lobaria pulmonaria show a contrasting pattern with repeated gene flow between mainland and Macaronesian Islands and an overall higher differentiation among populations.
A previous study dated the origin of L. pulmonaria to 6.9-11.9 Myr BP when it diverged from its Asian relative Lobaria tuberculata, and the divergence of the two Macaronesian endemic species from L. pulmonaria to 5.5-9.9 Myr BP 28 . Thus, the endemic species are likely to represent neo-endemics that may have evolved on the Macaronesian Islands after diverging from the older and widespread species, L. pulmonaria. This is similar to what has been found for Macaronesian Nephroma species 4 .

Relationship between island and mainland sites in Lobaria.
In all species, we found substantial differentiation among archipelagos, indicating long-term isolation. This was also true for the widespread species, L.pulmonaria: most island populations were genetically distinct from populations on the mainland. This reflects a history of low migration to the islands and genetic drift and diversification of island populations, which seem to have evolved from the continental populations and accumulated private alleles, which might have resulted from mutation or from genetic drift leading to loss of variation. Populations of L. pulmonaria from the Canary Islands were related to the Iberian Peninsula and those from the Azores to the Irish population. Divergence between geographic regions was also confirmed by our analysis of private and shared alleles. All regions contained a comparatively high number of private alleles in L. pulmonaria, with few alleles shared between regions. Hence, gene flow between archipelagos or between continental populations and archipelagos is rather low. www.nature.com/scientificreports/ Previous studies of L. pulmonaria have found that differentiation between regional populations is mainly dependent on the geographic scale of sampling 29,30 . On a small spatial scale, individuals of this species showed spatial autocorrelation in genotypes and alleles due to local deposition of diaspores [31][32][33] , but no or little differentiation was found among forested areas within a pasture woodland 29 or even among sites tens of km apart within a region 17 . However, significant differentiation among geographic regions has been found when different refugial regions were included in the analysis 26,34 , or when populations growing in different habitats such as floodplain forests and mountain ridges were compared within one region 35 .
The lack of genetic differentiation between subsets of a landscape or a forested region in L. pulmonaria could be due to the overarching role of stepping-stone habitats leading to effective gene flow. In volcanic archipelagos situated off the continental shelf and surrounded by ocean, the islands themselves represent the few stepping stones, but there are no additional populations in between. Thus, we expected a substantial amount of genetic differentiation between sites on different archipelagos, which was indeed found.
However, genetic distance was not simply a function of geographic distance for some of the studied populations of L.pulmonaria. Interestingly, sites on the Iberian Peninsula unexpectedly grouped with the site investigated on Newfoundland (but not with the spatially more proximate site in Ireland). Moreover, the site in Ireland was more similar to Azorean populations, rather than to the geographically closer sites on the Iberian Peninsula.
A particularly interesting facet of our data set was the relationship between island and mainland sites in the endemic species. Both endemic species are likely to have evolved on the Macaronesian islands, given their divergence time 28 . In L. macaronesica, the single known small population of this species on the Iberian Peninsula is genetically distinct from the Macaronesian island populations and could be the remains of what might have been a much larger population during phases of past, moister climate.
Given its diversity pattern in the microsatellite data, the single site where L. immixta is known from the Iberian Peninsula grouped within the Azorean populations ( Fig. 2A,D). Therefore, it seems likely that this species has been introduced to the Iberian Peninsula by Man with plant material from the Azores in recent time. An additional consideration that makes this scenario even more likely is that all individuals of this species were found in a Royal palace garden, a site that certainly has received massive amounts of plants (e.g. exotic trees) from various regions in the world, including Macaronesia. Hence, the Iberian populations of the two endemic species of Lobaria appear to have different histories.

Similarity of geographic patterns with Macaronesian phorophytes.
For an oceanic archipelago, the Azores host relatively few endemic plant taxa, a pattern which has been ascribed to a less variable paleoclimate because climatic variability can drive radiations [36][37][38] . The endemic plant species of the Azores are widespread across the archipelago and exhibit little divergence between populations, relative to those of the Canary Islands 36 . In opposition to this pattern, we found substantially higher levels of divergence between Azorean than between Canarian populations in our data, a pattern consistent across all three species. This pattern could be explained by the greater distances among islands on the Azores as compared to the Canary Islands, leading to less geneflow among islands. In the variable paleoclimate of the Canary Islands with its frequent shifts in humidity, the Lobaria lichens would have become locally extinct during arid periods, and populations may have gone through repeated genetic bottlenecks, leading to lower genetic variability as observed in our data. In a study of eight plant taxa, substantial genetic diversification was found on the Azores, emphasizing that in some endemic Azorean species, diversification may exist 39 , a pattern consistent with our data on Lobaria lichen fungi. A study on Erica (phanerogams) had a similar result 40 .
Epiphytic lichens such as the investigated species of Lobaria are dependent on a woody plant community. Two phorophyte taxa of the lichens have also been investigated in a phylogeographic context in Macaronesia. In the heather Erica scoparia, a western European and Macaronesian shrub which is a frequent phorophyte of Lobaria www.nature.com/scientificreports/ sp. on the Macaronesian Islands, Azorean populations were far more diverse than those in western continental Europe 40 . The authors explained this pattern by recurrent migration to the Azores and extinctions on the mainland. In L. pulmonaria, populations on the Azores were diverse and consisted of divergent genetic clusters, but in contrast with the pattern of E. scoparia, populations located on the Iberian Peninsula, Newfoundland and Ireland showed high diversity as well. A study targeting the phylogeography of L. pulmonaria on the European continent using microsatellite repeat data on thousands of samples showed no tendency for continental populations to have low allelic richness 26 . This could be the case if this lichen spread efficiently into sites affected by Pleistocene glaciations.
The second phorophyte of our study species investigated in a phylogeographic context was laurel tree (Laurus nobilis L., Laurus azorica (Seub.) Franco). Laurus azorica is a common phorophyte of Lobaria lichens in Macaronesia, whereas the drought resistant Laurus nobilis on the mainland is only rarely hosting the species. In the Laurus species complex, chloroplast haplotypes of Macaronesian L. azorica were closely related to those of L. nobilis from the western Mediterranean and from Northwestern Africa. The 'Macaronesian' haplotype occupied northwestern Africa, the Canary Islands and Madeira. A second, closely related haplotype occupied the Azores, and two others the western Mediterranean. The authors inferred the existence of multiple refugia in Laurus, and strong range dynamics in the western Mediterranean part of the range 41 . Climatic reconstructions showed that large, suitable areas for Laurus existed in the Pliocene, but during the last glacial maximum, these were reduced to the Mediterranean Basin and the Macaronesian islands 42 . These areas could also have enabled the survival of Lobaria spp., given that the climate was moist enough. With some exceptions (e.g. Annaba, Algeria), current sites where Laurus is found in northern Africa do not harbor Lobaria spp. as the climate is too dry, indicating that the ecological niches of the lichens and their phorophyte do only partially overlap. The large differentiation between Azorean and Canarian/Madeiran populations in all of our study species appears to somewhat resemble the phylogeographic pattern of Laurus. We would not have expected a close similarity of genetic patterns due to the differences in life history including dispersal syndrome in trees vs. lichens and as we used different loci.
This study further elucidates the complex evolutionary history of Lobaria, revealing the role of geography on the extent of genetic divergence in widespread Lobaria pulmonaria and the Macaronesian endemics, and provides one of the few studies dedicated to understanding lichen fungal genetics in an island setting. In summary, the occurrence of highly divergent individuals of L. pulmonaria in Macaronesia point towards a complex history with multiple migrations between islands and the mainland. In contrast, the Macaronesian endemics L. immixta and L. macaronesica with a reduced geographic range on the mainland which likely originated in Macaronesia, exhibit high divergence between archipelagos. While the single population of L. macaronesica on the Iberian Peninsula may represent an old population that has diverged from the Macaronesian island populations, L. immixta appeared to have recently been introduced by Man to the Iberian Peninsula. Although this report improves our understanding of Lobaria's evolutionary history, these findings could be furthered by more extensive sampling of Lobaria populations both in Macaronesia and elsewhere, and the subsequent addition of more loci.

Materials and methods
Study area. Our study area included three volcanic archipelagos located in the North Atlantic-the five westernmost Canary Islands, Madeira, the Azores, and in adjacent areas. The vegetation of the study sites is characterized by moist forests: laurel forests, Pinus canariensis forests, and high altitudinal shrub vegetation. The study sites were located in high elevational areas that were often covered with fog or were located within the Passat cloud zone. On the Azores islands, we often found Lobaria spp. in the near-natural forests covering the margins of calderas and in high elevation forests.
The Canary Islands are situated c. 100 km NW of the coast of southern Morocco, an area which is currently too dry for Lobaria spp. In contrast, the Rif mountains of Morocco are known to harbor L. pulmonaria 43,44 , but they are much further away from the Canary Islands (ca. 1000 km, Fig. 1). Both the Azores and Madeira are far away from the Iberian Peninsula (3600 km and 1400 km). Due to the large distances, gene movement between archipelagos and mainland is expected to be very low.
The age of the three archipelagos is well documented. Island age decreases from East to West on the Canary Island archipelago: Fuerteventura, 20.6 Ma; Lanzarote, 15.5 Ma; Gran Canaria, 14.5 Ma; La Gomera, 12.0 Ma; Tenerife, ~ 7.5 Ma; La Palma, 2.0 Ma; El Hierro, 1.12 Ma 7 . Tenerife is a special case because it consists of three palaeoislands (Adeje, 12 Ma; Teno, 6 Ma; Anaga, 4 Ma), which were connected by a rather recent (1 Ma) event 8 . The Selvagem archipelago (27 Ma) is situated between the Canary Islands and Madeira and is characterized by a dry climate and of low elevation (≤ 153 m) vegetation lacking forest 45 . Due to lack of suitable forest habitat, Selvagem could not have served as a steppingstone between the other archipelagos for our study species in recent time. Madeira consists of two main islands: Madeira (< 5.6 Ma) and the older Porto Santo (14 Ma) [45][46][47] .
The Azores were formed along spreading midoceanic ridges at the joint of the African, American and European plates 48,49 . The studied islands are Santa Maria (8.12 Ma), Sao Miguel (4.01 Ma) 50 52 . The occurrence of a land bridge connecting Pico and Faial at the maximum of the last glaciation (18,000 BP) has been reported 53,54 . It seems unlikely that this land bridge would have led to increased connectivity between the high-altitude forest types characteristically inhabited by Lobaria lichens, unless the habitat suitable for Lobaria spp. occurred at considerably lower elevations in the past.
Sampling. From each collecting area, we collected specimens from one (Gran Canaria) to nine (Madeira) populations, depending on island size and availability of populations of Lobaria ( Fig. 1; Table 1). Prior to field sampling, collecting permits were obtained for all studied regions. All collected material was deposited in the publically available cryoherbarium Christoph Scheidegger at WSL Swiss Federal Research Institute, cryopreserved at − 20 °C for long-term storage. Voucher numbers and collection information is provided in the Sup- Samples were collected along a transect through the population in large populations. If the local population size was small, we deviated from this sampling scheme and sampled lichens from all trees situated within approximately 1 hectare of forest. We attempted to include three specimens per species from each sampled tree. Our species identifications in the field were sometimes difficult as many specimens were poorly developed. Hence, we had to rely on molecular species identifications by RealTime PCR 55 .
Study species. Our study species were Lobaria pulmonaria, L. immixta and L. macaronesica. Lobaria pulmonaria is a widespread epiphytic lichen in large parts of the northern hemisphere, with a few occurrences in the southern hemisphere 56 . The other two species are endemic to Macaronesia but are both known from one spot on the mainland in Portugal (Sintra). From this site, L. immixta had been reported by Christa and Josef Poelt in 1961 57 . This is the second report of this species from this site. Moreover, L. macaronesica was discovered in Sintra in 2007 by C. Scheidegger 13 . For the endemic species, all samples from Sintra were found inside of the Royal Gardens, while L. pulmonaria also occurred in natural forests located in the vicinity. Molecular analysis. DNA was extracted following the manufacturer's protocol using the DNeasy 96 plant kit (Qiagen, Hilden, Germany). Each thallus was inspected for the presence of apothecia and parasites, and only thallus parts free of these structures were utilized for DNA extractions. Field identifications of the lichen material was done by SW and CS, lab identifications by SW. As many specimens were not well developed and lacked the diagnostic propagules 13 , we utilized RealTime PCR for species identifications. The RealTime PCR used a small (10 bp), species-specific stretch of the Internal Transcribed Spacer (ITS) region. For protocols, see Werth et al. 55 .
The genotyping of microsatellites followed Werth et al. 58 . Only six out of eight loci worked for all species and these were used for the analyses to ensure comparability among data sets; these markers were fungus-specific according to Widmer et al. 59 . We recently developed additional microsatellite loci for L. pulmonaria 60 , but the variability in the present set of six markers was highly suitable for a regional-scale study 58 .
All markers were run in a single multiplex PCR. For two loci, we used primers optimized to work in all species. Each Data analysis. In order to compare regional genetic structures among endemic and widespread species, we utilized extensive microsatellite datasets of each species to quantify the overall genetic relationships between sites, to infer the grouping of populations, and to analyze the partitioning of genetic variability upon sites and geographic regions.
For each species, 1000 bootstrapped microsatellite allele frequency datasets were generated using the 'seqboot' module of PHYLIP version 3.69 62 . The genetic relationships between all studied sites were evaluated for each species using the chord distance D C 63 , as implemented in the 'gendist' module of PHYLIP. Neighbor-joining trees were constructed with 'neighbor' , and a majority-rule consensus tree was computed using the 'consense' module of PHYLIP. The consensus trees were midpoint rooted with FigTree version 1.2.1 64 and visualized on a map with the geo-spatial information system GenGIS version 2.4.1 65 .
In order to analyze the main grouping in the data of each species and to infer the number of genetic groups (K) in each species, sites were clustered in a mixture analysis using BAPS version 5.4 66,67 . After an initial run for values of K up to 39 (number of sites), an additional 10 runs were performed for the optimum partition in each species. Admixture analysis (data not shown) was run and none of 1402 individuals (L. immixta), three of 1499 (L. macaronesica), and nine of 1239 samples (L. pulmonaria) were found to be admixed when a Bonferroni correction for multiple testing had been applied to the data. Neighbor-joining trees depicting the relationships between clusters were generated in BAPS using Nei's genetic distance, averaged over loci 68,69 . The major groupings were mapped in ArcGIS version 10 (ESRI). To quantify genetic differentiation between geographic regions, population pairwise F ST values were computed in Arlequin version 3.5 70 for each of the three fungal species.
Discriminant Analysis of Principal Components (DAPC) can provide valuable information about the most likely population model of a set of sampling sites, as different geneflow patterns lead to different genetic clustering 71 . Thus, it is possible to distinguish e.g. island models of gene flow from hierarchical island models or stepping-stone models 72 www.nature.com/scientificreports/ To assess whether there was isolation by distance, pairwise F ST values were calculated in R using the 'hierfstat' package v. 0.04-22 and geographic distances between sites were calculated with the package 'geosphere' v. 1.5-10 using the 'distm' and 'distGeo' functions. Significance of isolation by distance was assessed with linear regression models, where F ST was the response variable and geographic distance the predictor.
Data. The current data set consists of six fungus-specific microsatellite data and all allele frequency are accessible in the Dryad repository 76 .