Coevolutionary theory predicts that the distribution of obligately symbiotic organisms will be determined by the dispersal ability and ecological range of both partners. We examined this prediction for lichen-forming fungi that form obligate symbioses with cyanobacteria.
We compared genotypes of both partners of 250 lichens collected at multiple spatial scales in British Columbia, Canada. Multilocus sequence data collected from a subset of 128 of the specimens were used to determine the degree of recombination within the cyanobacterial populations.
We found that six distinct clusters of cyanobacterial genotypes are distributed throughout the known global phylogeny of the genus Nostoc, and that each appears to be evolving clonally. Fungal specialization is high, with each species associating with either one or two of the cyanobacterial clusters, while cyanobacterial specialization varies, with clusters associating with between one and 12 different fungal species. Specialization also varies geographically, with some combinations restricted to a single site despite the availability of both partners elsewhere.
Photobiont association patterns are determined by a combination of genetically based specificity, spatial population structure, and ecological factors and cannot be easily predicted by photobiont dispersal syndromes.
Lichens are symbiotic associations between a fungus (mycobiont) and one or more photosynthetic partners (photobionts), usually green algae and/or cyanobacteria. This symbiotic association is generally thought to be obligate for the mycobiont (Honegger, 1996), although there do appear to be rare cases of species that are capable of facultative lichenization, a phenomenon that seems associated with loss of lichenization in lineages such as in the Stictidaceae (Lutzoni et al., 2001; Wedin et al., 2004). It has long been argued that Trebouxia and closely related Asterochloris, some of the most common lichen photobionts, are also obligately symbiotic (Ahmadjian, 1967). More recent observations have challenged this view (Bubrick et al., 1984; Mukhtar et al., 1994), but the ecological significance of free-living stages remains controversial. Nostoc and Scytonema, the most common cyanobacterial lichen photobionts, are known to occur free-living in a wide variety of habitats but it has been difficult to determine if the same species occur both free-living and symbiotically using traditional phenotypic classification. Recent molecular phylogenetic evidence has shown that free-living and lichenized Scytonema form evolutionary independent lineages (Lücking et al., 2009), but Nostoc appears to be facultatively symbiotic (O'Brien et al., 2005).
For most lichen-forming fungal species, reproduction is predominantly sexual, though a variety of asexual propagules are also produced by different species (Bowler & Rundel, 1975). Many of these asexual propagules are specialized structures that facilitate the codispersal of the mycobiont and photobiont (Büdel & Scheidegger, 1996), resulting in the vertical transmission of the photobiont from one generation to the next. While codispersal of sexually produced ascospores and photobiont cells has been documented for some species (Pyatt, 1973), ascospores are generally dispersed independently of the photobiont. The association must then be reconstituted with algae or cyanobacteria in the vicinity of the germinating ascospore, resulting in the horizontal transmission of the photobiont for each new generation. Given that about one-fifth of all known fungal species are lichenized (Kirk et al., 2008), this presents a paradox: how can a horizontally transmitted obligate symbiosis give rise to such an evolutionarily successful and specialized group of organisms?
Several strategies have been proposed that might allow sexually reproducing lichen-forming fungi to reconstitute a lichen symbiosis in nature from one generation to the next. These include extracting photobiont cells from the thalli of other lichens (Friedl, 1987; Ott, 1987) or from the asexual propagules of other species (Rikkinen, 2003); and persistence by forming temporary associations with incompatible photobionts (Gaßmann & Ott, 2000) or in a free-living state (Etges & Ott, 2001) until a compatible photobiont is encountered. It has also been suggested that effective dispersal via ascospores is rare and that most dispersal is via asexual propagules or thallus fragments (Bowler & Rundel, 1975), an hypothesis that was recently confirmed for the lichen Lobaria pulmonaria (Dal Grande et al., 2012). For lichens where sexual reproduction appears to be the predominant mode of dispersal, such as Peltigera (O'Brien et al., 2009), selection on the mycobiont is likely to favor increases in the taxonomic range of compatible photobiont partners (reduced specificity; Smith & Douglas, 1987) in order to increase the probability that a compatible partner is encountered.
In the case of green algal lichens, most mycobiont species examined appear to be highly specific about which photobionts they lichenize with (Kroken & Taylor, 2000; Helms et al., 2001; Piercey-Normore & DePriest, 2001; Hauck et al., 2007). However, the degree of photobiont specificity and selectivity (preferential association with certain species/genotypes; Galun & Bubrick, 1984), collectively referred to as specialization (O'Brien et al., 2005), can vary dramatically among closely related mycobiont species (Yahr et al., 2004) and can also vary within a species across its range (Yahr et al., 2006; Fernández-Mendoza et al., 2011). Early molecular studies on cyanolichens reported that the mycobionts exhibited high photobiont specificity at the species level (Paulsrud et al., 1998, 2000). More recent studies suggest that this may not be a general result (Wirtz et al., 2003; O'Brien et al., 2005; Elvebakk et al., 2008), although it does appear to hold for some species (Myllys et al., 2007; Otálora et al., 2010; Fedrowitz et al., 2011). At higher taxonomic levels, it has been shown that the same lineage of Nostoc participates in symbiotic associations with a wide range of lichen taxa, as well as plant groups including cycads, bryophytes, and the angiosperm Gunnera (O'Brien et al., 2005), although epiphytic cyanolichens appear to be restricted to a lineage nested within Nostoc (Rikkinen et al., 2002; Elvebakk et al., 2008).
The contradictory results obtained for species-level photobiont specialization by mycobionts of cyanolichens may be partly a result of the failure to explicitly account for the local availability of different photobiont genotypes. The most definitive studies of photobiont specialization have compared the photobionts of each lichen species with the pool of photobiont genotypes represented by co-occurring species at the same locations (Yahr et al., 2004; Fedrowitz et al., 2011). By genotyping both partners, it is also possible to determine the role of vertical photobiont transmission in shaping population structure (Yahr et al., 2004; Fernández-Mendoza et al., 2011; Dal Grande et al., 2012), which can mimic the patterns resulting from specialization. This study applies such an approach to study photobiont specialization in the lichen-forming fungal genera Peltigera and Nephroma. Most species in both genera form mostly bimembered lichens, where the mycobiont is associated with the cyanobacterium Nostoc (although see Henskens et al., 2012). However, some species form trimembered lichens, where the green alga Coccomyxa is the main photobiont and Nostoc is an accessory photobiont which is restricted to specialized structures called cephalodia (Miadlikowska & Lutzoni, 2000, 2004). Samples were collected from five sites in British Columbia, Canada, separated by 15–150 km. Both fungal and cyanobacterial partners were genotyped using highly variable sequence markers. Three additional photobiont markers were sequenced from a subset of specimens in order to assess the extent of recombination. Photobiont sequences were compared with a global sampling of Nostoc sequences obtained from GenBank to determine which Nostoc lineages could act as photobionts for each mycobiont species (specificity). Photobiont sequences were also compared with the photobionts of all other species at the same site to quantify selectivity. Finally, photobiont genotypes were compared among mycobiont haplotypes within species to determine if patterns of association could be explained by vertical photobiont transmission within fungal clones.
Materials and Methods
All cyanolichen specimens were collected from moss-covered ground, rocks and tree trunks within 1-m-diameter circular plots separated by 5–60 m at five sites in British Columbia, Canada, as previously described (see Supporting Information, Fig. S1; O'Brien et al., 2009). Since the internal transcribed spacer (ITS) is highly variable at the species level in Peltigera (O'Brien et al., 2009), it was used to characterize mycobiont populations. Photobiont populations were characterized by sequencing the cyanobacterial gene rbcX and the intergenic spacer between rbcL and rbcX for each specimen (Rudi et al., 1998). In addition to the specimens used by O'Brien et al. (2009), genomic DNA was isolated and fungal ITS sequences were obtained for 46 additional Peltigera specimens as well as 15 Nephroma specimens from the same sites, using the PUREGENE Ultrapure DNA extraction Kit (Qiagen) and the primers ITS1F and ITS4 (see Table 1). In total, our dataset included 250 ITS sequences representing 25 species (see Table S1). rbcLX was sequenced using a published reverse primer and a redesigned forward primer that annealed within the last 50 bp of the rcbL gene (Table 1). For a subset of specimens, we obtained sequences for three additional cyanobacterial genes: nifV1 (121 specimens), rpoC2 (116 specimens) and trnL (118 specimens) (see Table S1 for details). See O'Brien et al. (2009) for details about PCR and sequencing conditions. All sequences were assembled using Sequencher 4.2 (Gene Codes, Ann Arbor, MI, USA) and representatives of each unique sequence type were aligned using MacClade 4.0 (Maddison & Maddison, 2000). Sequences have been deposited in GenBank (accession numbers KC437590–KC437650 (ITS), KC437651–KC437917 (rbcLX), KC437918–KC438038 (nifV), KC438039–KC438154 (rpoC2), KC438155–KC438272 (trnL).
Table 1. Primers for amplification and sequences of fungal (internal transcribed spacer (ITS)) and cyanobacterial markers
For each photobiont alignment, maximum-likelihood (ML) phylogenetic analyses were conducted using the HKY85 substitution model with gamma distributed rate variation (four categories), with a BioNJ starting tree and nearest neighbor interchange (NNI) branch swapping using PhyML (Guindon & Gascuel, 2003). Branch support was assessed using the approximate likelihood ratio test for branches (aLRT; Guindon et al., 2010). The P6b loop of trnL sequences was considered for discrimination of allele types, but excluded from phylogenetic analyses. The extent of interlocus recombination was characterized using Multilocus 1.2 (Agapow & Burt, 2001) as described in O'Brien et al. (2009).
Representatives of each unique rbcLX sequence type were aligned with published sequences from symbiotic Nostoc and free-living Nostoc commune (Rudi et al., 1998; O'Brien et al., 2005; Rajaniemi et al., 2005; Stenroos et al., 2006; Myllys et al., 2007), with intergenic spacers excluded from the alignment. This dataset was analyzed with ML, using the general time-reversible (GTR) substitution model with gamma-distributed rate variation (four categories), and an estimated proportion of invariant sites using Garli (Zwickl, 2006), with default tree search settings. Branch support was assessed with 100 bootstrap replicates. The resulting phylogeny was rooted with Anabaena PCC7120.
Separate ITS alignments were made for each of the 25 mycobiont species, so that all nucleotide positions could be unambiguously aligned. Haplotype networks were inferred using statistical parsimony with multibase indels coded as single characters with TCS (Clement et al., 2000).
Significance of associations between photobiont sequence types and mycobiont species and between photobiont clusters and sampling sites were assessed with Fisher's Exact Test using R (R Development Core Team, 2012), with redundant specimens from the same plot with identical mycobiont and photobiont sequences removed. Host specialization and spatial population structure were also quantified by grouping photobiont genotypes by mycobiont species or by sampling site and calculating linearized pairwise FST values. Only groups represented by 10 or more specimens were included in the comparisons. Significance was assessed using permutation tests. All population structure analyses were carried out using the program Arlequin 3.01 (Excoffier et al., 2005).
All sequencing reactions produced clean reads with no evidence of secondary peaks, suggesting that a single photobiont genotype predominated in each lichen thallus. When analyzed with rbcLX sequences from a global sampling of Nostoc strains from a variety of symbiotic associations obtained from GenBank, photobiont strains from British Columbia (BC) comprised 30 distinct alleles distributed throughout the core lineage of Nostoc that includes most free-living and symbiotic isolates (clade II of Svenning et al., 2005/Nostoc II of O'Brien et al., 2005; Fig. 1a, see Fig. S2 for details). Twenty six of the 30 alleles formed six clusters of closely related alleles (labeled I–VI in the figures). One cluster of six alleles (IV) was made up exclusively of BC photobionts and a few rare alleles (13, 26) were on long branches, but the majority of BC photobiont sequences were very closely related to lichen photobionts from Europe and other locations. Ten BC photobiont alleles were identical to photobionts from lichens collected in Europe (1, 2, 3, 5, 7, 9, 12, 14, 15, and 17), including the four most common types (Fig. S2).
For the 128 specimens for which multiple photobiont loci were sequenced, 18 different alleles were recovered for rbcLX, 21 for nifV1, 12 for rpoC2, and 15 for trnL (Fig. 1). The same rbcLX allele was recovered from 80 of the 128 specimens (allele 2 in Fig. 1a). These specimens had three different rpoC2 alleles, eight different nifV1 alleles, and five different trnL alleles. The frequency of the most common allele was 55 for rpoC2, 54 for trnL, and 32 for nifV1. There was one other rbcLX allele (type 4) that comprised two or more alleles in all other phylogenies, while alleles 1 and 9 comprised two nifV1 alleles each. There were also cases where strains with different rbcLX alleles had the same allele in another locus (alleles 1 and 11 in all three loci; alleles 2, 6 and 30 for nifV1; alleles 2 and 3 for rpoC2; alleles 2 and 6 and 25 and 27 for trnL). However, in all of these cases, strains with identical alleles at one locus had alleles that clustered together at all other loci and there were no cases where all four possible combinations of two alleles at each of two loci were present (the four gamete test; Hudson & Kaplan, 1985). Each of the six clusters identified in the rbcLX phylogeny were also monophyletic in the other gene trees. There was some incongruence between phylogenies at deeper nodes, including the phylogenetic positions of two singleton alleles (26 and 29), but the only topological conflict that was supported by bootstrapping involved the relationships among clusters II, III, and IV in rbcLX and nifV1. This congruence was reflected in the linkage disequilibrium calculations, as index of association values were extremely high (IA = 1.51), even when the sample was clone-corrected (IA = 0.78). These values were highly significant when compared with randomized datasets (P <0.001).
The most abundant mycobiont species present in the plots sampled was Peltigera leucophlebia (section Chloropeltigera), which is subdivided into three distinct clades following O'Brien et al. (2009; Fig. 2b). For the most abundant clade in this group, clade I, 44 mycobiont sequences were divided among eight different haplotypes (Fig. 2b). The two most common haplotypes comprised 33 of the 44 specimens, while the remaining six included four or fewer. The second most abundant group, clade III, was represented by 24 specimens divided among three fungal haplotypes, two of which were observed in a single specimen each, while clade II consisted of five specimens comprising three fungal haplotypes (Fig. 2b). Of these 72 lichen specimens, 58 had identical photobiont rbcLX sequences (allele 2) and six others had related photobionts in the same cluster (VI). Four additional photobiont alleles were recovered from the other eight specimens, which belonged to two other clusters (I and IV in Fig. 2a). Cluster I photobionts were found in both of the common P. leucophlebia clades while cluster IV photobionts were restricted to clade I.
The second most common mycobiont species in the sampling areas was Peltigera aphthosa (section Peltidea). Of 28 specimens examined, 23 had a single fungal haplotype, with the five other specimens each having a unique haplotype that differed from the common haplotype by one or two mutations (Fig. 2b). Allele 2 rbcLX sequences were recovered from the photobionts of 13 specimens and a closely related photobiont allele was found in two others. The remaining 13 specimens had one of four related photobiont alleles from cluster IV. The sister species to P. aphthosa, Peltigera britannica, was only sampled 16 times, but one more fungal haplotype was recovered than for P. aphthosa (Fig. 2b). The fungal haplotypes were also more divergent, with a total tree length of 11 steps compared with six for P. aphthosa. Allele 2 photobionts were present in nine specimens and two related alleles were recovered from three others. A single cluster IV allele was recovered from the photobiont of three P. britannica specimens and a single specimen had a cluster I photobiont.
The third mycobiont species from section Peltidea present in the sample plots was Peltigera malacea. Unlike the other members of this section, which form trimembered thalli with cephalodia, P. malacea forms bimembered thalli with exclusively cyanobacterial photobionts. Three mycobiont haplotypes were recovered from 16 specimens for this species, with two of these haplotypes found in a single specimen each (Fig. 2b). Five photobiont rbcLX alleles were recovered from this species. Four of them clustered together (cluster III), while the fifth (allele 25), which was recovered from a single specimen, belonged to the related cluster IV (Fig. 2a).
For Peltigera section Peltigera, mycobiont species richness was much higher, with representatives of seven described species and three undescribed species present (Fig. 2b). In addition, Peltigera ponojensis was represented by two fungal haplotypes that differed by 12 mutations, suggesting the presence of an additional cryptic species. Peltigera canina and Peltigera kristinssonii were the most common species from this group, represented by 19 and 20 specimens respectively. Mycobiont haplotype diversity varied, from Peltigera cinnamomea, which was represented by five specimens, each with a different haplotype, to P. kristinssonii for which 17 of 20 specimens had identical haplotypes. Photobionts with rbcLX alleles belonging to cluster VI were present in all specimens of four species in this group (Peltigera ‘fuscopraetextata’, P. kristinssonii, P. ponojensis, and P. sp. A), including the most common species (P. kristinssonii). All but one of the specimens of a fifth species (Peltigera praetextata) had cluster VI photobionts. Three other mycobiont species had photobionts with rbcLX sequences from two distinct clusters (V and VI for P. canina and P. cinnamomea; I and V for P. ‘neocanina’) while Peltigera membranacea photobionts had rbcLX sequences from cluster V and two different unclustered alleles (14 and 15). Finally, Peltigera degenii was represented by a single specimen with allele 14 photobiont.
Six additional Peltigera mycobiont species from three sections were also collected, and two of these (Peltigera neopolydactyla and Peltigera venosa) also appeared to have additional cryptic mycobiont species based on ITS sequence divergence (Fig. 2b). These species were all uncommon, being represented by between two and five specimens and either one or two fungal haplotypes. Four of these species were associated with photobionts from a single rbcLX sequence cluster each, including all three species in Peltigera section Horizontales (Peltigera collina, Peltigera horizontalis and Peltigera neckeri), none of which associated with cluster VI photobionts, the most common photobiont cluster in 11 other Peltigera species. P. venosa, a trimembered species from Peltigera section Phlebia, was associated exclusively with cluster VI photobionts, while Peltigera polydactylon and P. neopolydactyla (Peltigera section Polydactylon), were both associated with photobionts with rbcLX sequences from cluster VI and one other cluster or unique allele.
For Nephroma, four mycobiont species were sampled. Nephroma resupinatum was represented by a single specimen, while Nephroma bellum and Nephroma arcticum were sampled twice each, and Nephroma parile was sampled 10 times (Fig. 2c). A single fungal ITS haplotype was recovered from each species except N. bellum, where each of the two specimens had a different haplotype. All 10 N. parile specimens had the same photobiont rbcLX allele, while N. bellum and N. resupinatum shared photobionts with the same allele. The photobiont rbcLX alleles from these three species clustered together (cluster II) and were not found in any of the Peltigera species. The two specimens of N. arcticum, a trimembered species, had photobionts with different rbcLX sequences that did not group with those of the other Nephroma species. One was part of cluster IV, while the other (allele 26) was intermediate between clusters III and IV.
A large majority of specimens were collected from the two sites near Clearwater BC (southern sites), with 169 specimens collected from the Spahats Creek site and 56 from the Battle Creek site compared with between six and 19 specimens from the three northern sites in the Cariboo mountains (Barkersville, Cameron Ridge and Ghost Lake; Fig. S1). There were large mycobiont species composition differences between the sites, with nine species collected exclusively from the southern sites and five collected only in the northern sites (Fig. 2). Of the 12 species that were collected at both southern and northern sites, seven had identical mycobiont ITS haplotypes recovered from both. One to three mutations separated northern and southern haplotypes for all other mycobiont species, except for P. venosa where haplotypes were separated by 12 mutations and likely represent different species. Eight mycobiont species had identical photobiont rbcLX alleles in southern and northern sites while two of the others had photobionts from the same sequence clusters. Two of the species were associated with different photobiont clusters in southern and northern sites; however, P. britannica was associated exclusively with cluster VI photobionts in the southern sites and with photobionts with alleles 4 (cluster I) and 19 (cluster IV) at the northern sites. Likewise, P. membranacea was associated with allele 1 (cluster V) photobionts at the southern sites and photobionts with rbcLX alleles 14 and 15 at the northern sites.
Photobiont population structure
When photobionts from different lichen species were compared using FST, the photobionts of N. parile and P. malacea were highly structured in comparison to all other species (FST = 0.59–1.00; Table 2). These values were highly significant. P. aphthosa photobionts were also significantly different from those of all other species, except for the closely related species P. britannica (FST = 0.10 for P. britannica, FST = 0.30–0.61 for other species). FST values for all other comparisons ranged from −0.01 to 0.16. Some of these values were significant, but none of the other species with n ≥ 10 had photobionts that were significantly different from all other species. Pairwise FST values were significant for all comparisons between sites, but values were much lower for the comparison between Spahats Creek and Battle Creek (both southern sites; FST = 0.04) than for any of the other comparisons (FST = 0.15–0.42; Table 3).
Table 2. Pairwise FST values for photobionts, structured by mycobiont species (only species with ≥ 10 specimens included)
P. leu. I
P. leu. III
Values in bold indicate significant (P ≤ 0.05) values as assessed by Bonferroni corrected permutation tests.
N. parile (n =10)
P. aphthosa (n =28)
P. britannica (n =16)
P. canina (n =18)
P. leucophlebia Clade I (n =44)
P. ‘fuscopraetextata’ (n =13)
P. kristinssonii (n =20)
P. leucophlebia Clade III (n =24)
P. malacea (n =16)
Table 3. Pairwise FST values for photobionts, structured by site (Cameron Ridge excluded)
Values in bold indicate significant (P ≤ 0.05) values as assessed by Bonferroni corrected permutation tests.
Spahats Creek (n =168)
Battle Creek (n =56)
Ghost Lake (n =22)
Barkersville (n =14)
Multilocus sequence data from the photobionts of the lichens collected for this study strongly indicate that the population structure of symbiotic Nostoc is clonal. All multilocus genotypes observed were compatible with clonal descent, and linkage disequilibrium as measured by the Index of Association is high (0.78 after clone correction). The phylogenies from the different loci are also highly congruent, as the only topological conflict that received significant bootstrap support involved the relationships among clusters II, III, and IV in rbcLX and nifV1. The finding of a clonal population structure for Nostoc contradicts reports of incongruence between loci from other studies (Rudi et al., 1998; O'Brien et al., 2005; Kaasalainen et al., 2012). It may be that genetic exchange is so rare that no recombinants were present in our sample, or the frequency of recombination may vary geographically or be higher in free-living populations.
The lichen photobionts sampled from British Columbia represent a large fraction of the worldwide phylogenetic diversity of Nostoc (Figs 1a, S2), but the number of different sequence types present at each site was quite low, with a maximum of 17 sequence types at the most extensively sampled site. There was some evidence of isolation by distance, as there was much less population differentiation between the Spahats Creek and Battle Creek sites, which are located within 15 km of each other, than between any other pair of sites (Table 2). These results contradict the traditional view that microbes are not dispersal-limited (Beijerinck, 1913; Baas-Becking, 1934) but are similar to what has been observed for other groups of cyanobacteria (Papke et al., 2003; Gugger et al., 2005).
For many of the Peltigera species sampled, the most common photobiont rbcLX allele type was the same (allele 2). Most of these mycobiont species also associate with some of the closely related photobiont alleles from cluster VI. However, these photobiont types were completely absent from seven of the Peltigera species sampled. Many of these species were sampled at low frequency, and in some cases they were sampled only from the northern sites where photobiont type 2 was less common. Nevertheless, given the high frequency of rbcLX allele 2 in the species where it does occur, its absence from these seven species is highly significant (P < 10−5). Similarly, the second most common photobiont type (rbcLX allele 1), which was the predominant photobiont type for four Peltigera species, did not occur in any of the species that use allele 2 photobionts, except P. canina (Fig. 2b).
A third case of a cluster of photobionts being restricted to a subset of species is cluster IV, which was only recovered from N. arcticum, P. aphthosa, P. britannica, P. malacea and one of the three clades of P. leucophlebia that were shown to be reproductively isolated from each other by O'Brien et al. (2009). These photobiont rbcLX alleles are found at high frequency in specimens collected from the northern sites; in the southern sites, they were found only in P. aphthosa and in one specimen of P. leucophlebia clade I. These differences in photobiont association between the two regions are highly significant (P =0.0004 when all trimembered species are pooled). Such changes in patterns of host association with geography have also been documented for green algal lichens (Yahr et al., 2006) and are a strong prediction of the geographic mosaic theory of coevolution (Thompson 2005). It is interesting to note that this third group of photobiont types (cluster IV) is only found in trimembered lichens (except for one P. malacea specimen, discussed later), meaning that these Nostoc types were found almost exclusively as accessory photobionts isolated in cephalodia.
In addition to the photobiont specialization outlined earlier, there were also several cases of high reciprocal specificity, where all specimens of a lichen species (or of two related species) contained a single photobiont type or a single cluster of photobiont types that were not recovered from any other lichen (P. malacea, N. parile and N. bellum/N. resupinatum). Furthermore, all photobiont rbcLX sequences from specimens representing these three species collected in locations other than British Columbia, including eastern North America, Asia and Europe are identical to the sequences obtained here, or cluster with them to the exclusion of all other sequence types (see Fig. S2). In the case of the Nephroma spp., however, Myllys et al. (2007) found Lobaria pulmonaria photobionts with identical rbcLX sequences to both Nephroma photobiont types, and Parmeliella triptophylla photobionts that were identical to the N. bellum/N. resupinatum photobiont type. Neither of these species occurred in our study sites. This suggests that a similar pattern of low but variable specialization by the photobionts and higher specialization by the mycobionts is occurring in Peltigera and Nephroma, but that the different Nephroma photobiont lineages are more closely related to each other.
In contrast to Nephroma, 15 of 16 P. malacea photobionts recovered in this study grouped together with nine of 10 photobionts from P. malacea from eastern North America, Europe and Asia, forming a well-supported cluster to the exclusion of all other Nostoc sequences, while the remaining two sequences of the Nostoc found in P. malacea (one from Europe and British Columbia type 25), fall within a well supported group including the P. malacea-specific cluster and cluster IV (see Fig. S2). P. malacea is therefore an example of extreme one-to-one reciprocal specificity, a phenomenon observed in five different species of Collema and Leptogium (Otálora et al., 2010). In the case of Collema/Leptogium, high reciprocal specificity was restricted to species with specialized codispersal structures (isidia), while P. malacea does not produce isidia or soredia. P. malacea also rarely produces ascospores and reproduction appears to be primarily clonal, probably through thallus fragmentation (O'Brien et al., 2009). However, it seems unlikely that codispersal of the partners via thallus fragments could generate the pattern observed here, because of the diversity of different sequence types for both partners. It is more likely that some form of genetically determined specificity is operating.
The cluster containing P. malacea photobionts forms a strongly supported monophyletic group with cluster IV, which is restricted to trimembered Peltigera and Nephroma lichens, as well as photobionts of two Sticta species (Fig. S2). P. malacea is closely related to these trimembered Peltigera lichens (Miadlikowska & Lutzoni, 2004) and is very similar in morphology, chemistry and ecology (Vitikainen, 1994). These trimembered species can form bimembered cyanomorphs under the appropriate environmental conditions (Brodo & Richardson, 1978; Stocker-Wörgötter & Türk, 1994; Stocker-Wörgötter, 1995), suggesting that P. malacea may have evolved from the cyanomorph of a trimembered species. If the common ancestor of photobiont clusters III and IV evolved high specificity for trimembered Peltigera spp., possibly accompanied by reduced motility restricting these Nostoc strains to cephalodia (Paulsrud et al., 2001), they may have cospeciated along with the mycobiont.
When photobiont association patterns are compared with the phylogenetic relationships of the mycobionts (Miadlikowska & Lutzoni, 2000; Miadlikowska et al., 2003; O'Brien et al., 2009), there are some examples of phylogenetic conservatism, where sister mycobiont species associate with the same photobiont types (photobiont rbcLX allele 14 found only in P. degenii and P. membranacea; allele 1 found in P. cinnamomea and P. ‘neocanina’). However, most patterns of photobiont specialization did not reflect the phylogenetic relationships of the mycobionts, suggesting that it must be either an evolutionarily plastic trait or driven by ecological differences.
It has been suggested that epiphytic and terricolous lichens use different lineages of photobiont (Rikkinen et al., 2002; Rikkinen, 2003; Elvebakk et al., 2008), but this seems unlikely in this case because different lichen-forming fungal species were collected from the same plot and ecological niche, that is, associated with mosses growing on soil and rocks. Ecological conditions were also unable to explain the patterns of photobiont specificity observed by Myllys et al. (2007) and Stenroos et al. (2006). Most lichenized Nostoc strains studied by Rikkinen et al. (2002) fell into two clades, with one comprising photobionts of epiphytic lichens (Nephroma, Parmeliella, Lobaria) and the other comprising those of terricolous lichens (primarily Peltigera and two terricolous species of Nephroma). However, the latter clade had low phylogenetic support in their analysis and has not been recovered as monophyletic by others (O'Brien et al., 2005; Stenroos et al., 2006; Elvebakk et al., 2008). The ‘Nephroma guild’ lineage is recovered as monophyletic in our analysis (see clade B in Fig. S2), but Peltigera photobionts form at least six distinct lineages. The photobionts of N. arcticum (a terricolous Nephroma species forming a trimembered association; Nostoc rbcLX alleles 26 and 27) did not group with the other Nephroma photobionts, but they were also distinct from Peltigera photobionts. However, both N. arcticum specimens were collected at the Ghost Lake site, where sampling was not as extensive as for some of the other sites, so it is possible that the N. arcticum photobiont types may be present in some of the Peltigera spp. at that site.
All of the mycobiont species that we sampled form associations with a very restricted subset of the locally available photobiont genotypes. This high degree of selectivity in a horizontally transmitted obligate symbiosis suggests that availability of suitable cyanobacterial partners is not limiting, at least at the locations we sampled. This is likely a result, at least in part, of the fact that many mycobiont species can share the same photobiont genotypes, that is, photobionts being generalists, allowing facilitation through photobiont sharing (Rikkinen, 2003). Despite the overall high degree of specialization demonstrated by the mycobiont, most well-sampled Peltigera spp. do associate with at least two clusters of Nostoc sequence types. This flexibility may increase the probability that ascospores are dispersed to areas where compatible photobiont partners are present or may allow ecological specialization if different photobiont lineages are adapted to different microclimatic conditions. This may help to explain the extremely wide geographical range and habitat diversity of many species of Peltigera (Martínez et al., 2003).
We thank S. Diezmann, T. Goward, and E. O'Brien for field assistance, R. Vilgalys, J. Shaw, P. Manos, L. Kohn and D. Henk as well as three anonymous reviewers for comments on the manuscript, and L. Bukovnik for assistance with DNA sequencing. This work was supported by National Science Foundation (NSF) Doctoral Dissertation Improvement Grant (DDIG) DEB-03-09299 to H.E.O. and F.L., NSF REVSYS grant DEB-10-25930 to J.M. and F.L., and by Duke University. H.E.O. was also supported by a Canadian National Sciences and Engineering Research Council Postgraduate Scholarship.