Introduction

Pioneer colonisation by microorganisms is important for the later establishment of higher organisms on newly exposed substrata [1–4]. The modification of soil conditions includes increased water retention, stabilisation of micro- and macro-particles and incorporation of nutrients such as organic carbon by photosynthesis or nitrogen by nitrogen fixation [5–7]. These processes occur in a variety of environments, including recently deglaciated areas. Glacier retreat, visible nowadays in polar and subpolar areas of both hemispheres due to global warming [8–11], exposes to the atmosphere soils and rocks previously covered by ice. The first colonisation events consist of psychrophilic microorganisms, including cyanobacteria [12, 13] that may have been present in the crust under the ice (or within the ice itself [14]) or that reach the site just after exposure [15, 16]. This is followed by colonisation of exposed rocks and soil by other organisms such as lichens or bryophytes, and finally by the establishment of vascular plants [3, 4, 17–19]. In this process, symbioses (such as lichens, mycorrhizae or plant–bacteria symbiotic associations) play an important role, enabling or increasing the colonisation abilities of the organisms involved, mainly by means of nutrient acquisition [20]. There is a large range of plants and microorganisms able to establish symbiotic relationships (reviewed by [21–25]), including plants such as legumes or cycads [26–28] and microbes such as Proteobacteria (e.g. Rhizobium) and filamentous, heterocyst-forming cyanobacteria (genera Nostoc and Anabaena [29]). Of particular interest is the symbiotic relationship between the plant genus Gunnera (Gunneraceae) and its cyanobacterial endosymbiont, Nostoc spp., as it is the only documented specific association between angiosperms and cyanobacteria [29–33]. The genus Gunnera consists of 61 species distributed mainly in the Southern Hemisphere, with only two species present in Tierra del Fuego [34]. These plants are perennial herbs, ranging in size from 30 cm (Gunnera magellanica) to more than 5 m [34]. On the other hand, the genus Nostoc (Nostocaceae) comprises filamentous photosynthetic cyanobacteria that can form motile filaments (called hormogonia) and differentiate into atmospheric nitrogen-fixing cells (called heterocysts); these organisms can occur in a symbiotic state or free living, in both terrestrial and aquatic habitats, and show a cosmopolitan distribution [35]. The intracellular symbiosis of Nostoc within the host Gunnera and the process of infection have been extensively studied [29–31]. When the symbiosis is established, many of the vegetative Nostoc cells become heterocysts that fix and transfer nitrogen to the host cell [30]. The host plant in turn exports carbon to the cyanobacterial symbiont and allows the cyanobacteria to expand their ecological niche. Multiple Nostoc strains are known to be able to establish symbiotic associations with different Gunnera species [30, 36, 37]. However, it has been addressed that a single plant might be associated only with a single Nostoc strain [37] as a consequence of the timing and the mechanisms of the infection process.

The aims of the present study were to investigate the genetic diversity of symbiotic Nostoc strains associated with G. magellanica at different spatial scales: within the same individual host plant and within and among several populations along Tierra del Fuego (including retreated glaciers and nearby areas with well-established ecosystem). Finally, we also aimed to characterise the relationship between the symbionts.

Materials and Methods

Study Area

G. magellanica specimens were collected in nine localities along Tierra del Fuego (XII Region, Chile). Six localities were sampled in Isla Grande, corresponding to areas where the glacial ice has been retreating over a period of time [38]. One locality on Península Brunswick and two on Isla Navarino were also sampled (Table 1).

Table 1 Collecting data for each of the localities studied
Full size table

The area is dominated by a rugged landscape, with Cordillera Darwin as the most important mountain range (reaching altitudes of 2,580 m.a.s.l.); high-altitude plateaus are also present. These sites have a maritime climate along the coast, characterised by hurricane-force winds and dense cloud cover [39]. These features, combined with the influence of the Pacific Ocean and the peculiar, distinctive orography create a significant gradient of decreasing rainfall from the Atlantic towards the Pacific shore (500 mm/year max near easternmost sites 8 and 9 vs. 4,000 mm near westernmost site 5) and between north and south slope of Cordillera Darwin [40]. The average temperature is 5 °C with little seasonal changes near the seaside [41–44].

Maps and geographic representation were carried out using the software DIVA-GIS and the information from its webpage [45].

Biological Material

G. magellanica (Gunneraceae) is an herbaceous plant species present only in the southernmost region of the Southern Hemisphere. It is a perennial diploid (2n = 34) plant that can be monoecious or dioecious, being much smaller in size compared with other species of the genus, which is mainly tropical. Diminutive size may represent an adaptation to more challenging environments [34]. Cyanobacteria-infected regions (n = 181) inside the rhizome of complete G. magellanica specimens (n = 133) were collected and stored individually at −20 °C (samples from sites 1–7), while the samples from sites 8 and 9 were immersed in cetyltrimethylammonium bromide (CTAB) buffer [46], and transported to the laboratory.

From each plant, at least one symbiotic colony was extracted in aseptic conditions, avoiding as much as possible the inclusion of surrounding vegetal tissue, as well as fragments of non-cyanobacteria-infected rhizome tissue from four different specimens used as negative controls. The DNA extraction was carried out according to Cubero et al. [46], with a modification at the lysis step, extending it to a 12-h length.

PCR Amplification and Sequencing

Three different regions from the cyanobacterial genome were amplified: a fragment of the 16S ribosomal RNA gene (16S); the 16S–23S internal transcribed spacer (ITS); and the RuBisCO large subunit gene coupled with its promoter sequence and chaperon-like proteins (rbcLX). Reaction mix was carried out following O’Brien et al. [47], completing a 25-μl final volume, consisting of: dNTPs (0.2 mM of each), 1.5 mM of MgCl2, 0.625 units of TaqPolimerase (1 unit·μl−1 BioTools, Madrid, Spain), 25 μg of BSA, 0.5 μM primers (forward and reverse) and 1× PCR buffer. The cyanobacterial specific primer pair CX-CW [48] was used for the amplification of the rbcLX, using the following thermocycle conditions: a first 4-min step at 94 °C followed by 36 cycles of three steps: 94 °C for 30 s, 55 °C for 30 s, 72 °C for 2 min, and a final step of 72 °C for 7 min. The 16S and ITS sequences were amplified jointly using the primer pair 359F [49] and 373R [50], following a ‘Touchdown PCR’ protocol, whose conditions followed Janse et al. [51]: an initial step at 94 °C for 5 min followed by 20 cycles of three steps: 94 °C for 1 min, 62–52 °C for 1 min (descending 0.5 °C per cycle, the first being performed at 62 °C and the last at 52 °C) and 72 °C for 1.5 min; 10 cycles of other three steps (94 °C for 1 min, 52 °C for 1 min and 72 °C for 1.5 min) and a final step of 72 °C for 30 min. All PCR amplifications were carried out in either a MJ Mini Personal Thermal Cycler (BIO-RAD) or a GeneAmp PCR System 2400 (Applied Biosystems). PCR products were purified using the UltraClean PCR Clean-Up kit (MO BIO Laboratories Inc.). Both DNA directions (5′-3′/3′-5′) were sequenced, with the same primer pairs used in the amplification step, by Macrogen Inc. Laboratories (South Korea) through its automatic standard sequencing service, using a 3730XL DNA sequencer under required conditions by BigDye™ terminator cycle sequencing kit. DNA extracted from non-cyanobacteria-infected plant rhizome tissue could not be amplified with the primer pairs used.

Sequence Alignment, Genetic Diversity and Phylogenetic Analyses

Complementary sequences from the same specimen and DNA region were collapsed into contigs using SeqMan software (Lasergene v 7.00, DNASTAR). Approximate identifications were obtained comparing contigs against the GenBank database by means of the BLAST algorithm [52] in order to check for contaminations, using 97 % of sequence coverage and E value of 0.001 parameters as threshold in the searches. Alignments were made using the software ClustalW [53], implemented within BioEdit v.7.0.9 [54], being carried out for each of the three genomic regions. Furthermore, rbcLX sequences obtained from the Gunnera specimens were aligned with sequences from Gunnera chloroplasts retrieved from GenBank database in order to discard previously undetected plant organelle contamination. Bellerophon [55] software was used to look for possible chimeras in all the alignments. Datasets with ambiguously aligned regions (rbcLX gene) were treated with the software Gblocks v.0.91b [56] prior to phylogenetic analyses.

Alignments were collapsed into haplotypes with the software Collapse 1.2 (David Posada, available at http://darwin.uvigo.es/software/collapse.html). Genetic diversity measurements were computed in DnaSP v.5 [57]. The following parameters were calculated: number of polymorphic or segregating sites (S); total number of mutations (Eta); haplotype diversity (Hd) [58]; and nucleotide diversity (π) [58]. Statistical parsimony [59] was used for later generation of genealogies (haplotype networks) using TCS v.1.21 software [60]. MOTHUR v.1.21.1 [61] was employed in order to cluster the sequences into operational taxonomic unit (OTUs) for posterior phylogenetic analyses and for study of their relative abundance by the furthest-neighbour method. Mantel tests were carried out by using the Microsoft Excel software complement GenAlEx [62] to test the role of geographic distance in the genetic structure of the endosymbionts. The genetic structure of the populations of Nostoc associated with G. magellanica was checked using the analysis of molecular variance (AMOVA) [63] for 16S and rbcLX genes using the software Arlequin v.3.11 [64]. In these analyses, different groups were structured, in order to find out which factors—environmental components of the sites, approximate time of exposition after ice retreat [38, 40] or position concerning Cordillera Darwin, i.e. North vs. South slope—better explain the genetic variance in the dataset.

Phylogenetic analyses were carried out for the 16S and rbcLX genes. Collapsed haplotypes were aligned by using of ClustalW [53] with the most similar sequences found in BLAST searches [52] in the GenBank database. Most likely nucleotidic substitution models for the alignments were searched by means of the software jModelTest [65] using the Akaike information criterium [66]. The general time reversible [67] +I +G was chosen for the 16S and the rbcLX alignments. Maximum likelihood phylogenetic analyses [68] were then performed, estimating support for each node using bootstrapping (10,000 repetitions), by the nearest-neighbour interchange method implemented in the software MEGA 5.05 [69]. Phylogenetic analyses based on Bayesian inference [70] were carried out with Mr. Bayes software [71], performing 30 mill generations for 16S and 17 mill for rbcLX.

Low-Temperature Scanning Electron Microscopy

Small pieces of vegetal tissue hosting cyanobacterial colonies were observed by low-temperature scanning electron microscopy (LTSEM) method by being mechanically fixed onto the specimen holder of a cryotransfer system (Oxford CT1500), plunged into subcooled liquid nitrogen, and then transferred to the microscope’s preparation unit via an air-lock transfer device following the protocol described by de los Ríos et al. [72]. The frozen plant tissue was cryofractured in the preparation unit and transferred directly via a second air lock to the microscope cold stage, where it was etched for 2 min at −90 °C. After ice sublimation, the etched surfaces were sputter coated with gold in the preparation unit and the tissue then placed on the cold stage of the SEM chamber. Fractured surfaces were observed under a DSM 960 Zeiss SEM microscope at −135 °C.

Results

Fifteen plants from different sampling sites were used for the intra-specimen diversity study. Sixty-three sequences for each of the three genomic regions were obtained by the amplification of DNA extracted from the colonies (with at least three sequences from different colonies per plant) for the intra-specimen diversity study. The sequences obtained from different colonies within the same host specimen were identical. Further, analyses of tissue from 133 plants generated 113 sequences each for the 16S and ITS regions and 110 for the rbcLX from those same specimens of G. magellanica that were subsequently used for the population study level. All sequences were identified as belonging to Nostoc through BLAST searches in GenBank. Moreover, no chimeras were found in any alignment by the Bellerophon software analysis.

Polymorphism analyses for each genomic region and for each sampling site showed different ranges of genetic diversity (Table 2). All polymorphic parameters for 16S showed their lowest values in sites 1 and 9. ITS region presented its lowest values for every parameter in site 1. However, rbcLX gene region reached its lowest values for most of the polymorphic parameters in sites 2 and 9, except for haplotype diversity, which was reached in site 1. In 16S gene analyses, highest levels for each parameter were reached in site 3. Parameters for ITS chromosomal region reached these maximum values in site 2 and for rbcLX gene in site 3, but for haplotype diversity, which reached its maximum value in site 5 for both regions.

Table 2 Analysis of symbiotic Nostoc sequences diversity based on three genomic regions
Full size table

Haplotypes found in each site and their relative abundance for 16S, ITS and rbcLX regions are shown in Figs. 1, 2 and 3, respectively. For the 16S gene, the lowest diversity was found in sites 1 and 9 with only one haplotype, followed by sites 6 and 7 with two. The greatest diversity was found in site 5 (five haplotypes). Except for sites 7 and 3, the rest of sites showed a predominant haplotype (Fig. 1). Most of the haplotypes seem to be fixed; eight of them occurred in only at a single locality. On the other hand, haplotype 12 (present in two sites) and haplotypes 1, 2 and 5 (present in four sites) showed a broader distribution.

Fig. 1
figure 1

Map of sampling area showing haplotypic diversity in different sites and relative abundance of each haplotype for 16S gene data

Full size image
Fig. 2
figure 2

Map of sampling area showing haplotypic diversity in different sites and relative abundance of each haplotype for ITS genomic region data

Full size image
Fig. 3
figure 3

Map of sampling area showing haplotypic diversity in different sites and relative abundance of each haplotype for rbcLX gene data

Full size image

For the ITS region, only site 1 showed one unique haplotype, whereas the rest showed at least two (Fig. 2). Site 5 showed the highest diversity with eight haplotypes. All sampling sites showed a predominant haplotype except for sites 7 and 3, represented by two equally abundant haplotypes. The broadest distribution corresponded to haplotypes number 1 and 6 (occurring in four sites), followed by 2 and 14 (three sites) and 4, 10 and 18 (two sites).

Regarding the rbcLX region, all sites showed more than one haplotype (Fig. 3). Site 9 showed the lowest number of haplotypes (two), while site 5 showed the highest genetic diversity with 12 haplotypes. Every site had an haplotype with more sequences than the others within it, except for sites 1 and 6, which showed two and three respectively with the same number. Haplotypes number 2 and 11 showed the broadest distribution, being present in four sites; haplotype 31 occurred in three sites, and the rest occurred only in one site.

Figure 4 shows how unique haplotypes are related by statistical parsimony, and their distribution among sampling sites, for the sequences of genomic 16S gene (A) and ITS (B) regions. Two unconnected groups under a 95 % parsimony criterion [59] were found in the haplotypes of the 16S ribosomal gene (Fig. 4a). The first includes 59 sequences in four haplotypes (I, II, III and IV), while the second includes the other 44 sequences in 8 haplotypes (V, VI, VII, VIII, IX, X, XI and XII). On the other hand, haplotypes from the ITS region were included in five different groups in the network analysis (Fig. 4b). Three of them (xvii, xviii and xix with 5, 1 and 2 sequences, respectively) were shown to be completely separated from the rest, two (xv and xvi with 11 sequences) were only related to each other, while the fourteen remaining haplotypes (67 sequences) showed connection under this probability criterion.

Fig. 4
figure 4

Statistical parsimony networks depicting relationships among analysed haplotypes. Each circle represents one haplotype, and its surface area is directly related with the number of sequences integrated in it. Roman numerals refer to posterior comments. Red dots along branch lines represent the number of nucleotide substitutions between two haplotypes. a Haplotypes found in the sequences of 16S gene dataset. b Haplotypes found in the sequences of ITS genomic region dataset

Full size image

For the 16S gene, at the level of 3 % dissimilarity (indicating putatively distinct species) 2 OTUs were found, while at 5 % dissimilarity (indicating congeneric taxa) a single grouping was observed (Table 3). The ITS region showed 12 OTUs while the rbcLX gene presented 23 OTUs (6 after Gblocks trimming), using 3 % sequence dissimilarity. For the 5 % cut-off value, the number of OTUs was reduced to 8 for ITS and to 15 (4 after trimming) for rbcLX. Only at 20 % dissimilarity for ITS and at 18 % (8 % if Gblocks trimming is applied) for the rbcLX gene could 2 OTUs be recognised. These two OTUs included the same specimens independently of the analysed genomic region.

Table 3 Number of OTUs and the number of sequences in the most common OTU obtained under different dissimilarity cut-off levels
Full size table

The haplotype distribution through the sampling area could reflect different environmental gradients. Figure 5 shows the response of Hd values with respect to three environmental factors. Figure 5a shows a clear trend of Hd increasing with longitude rise of the sampling areas for the three genomic regions, although low r 2′s and no significant p values were found in all the cases. Evolution of Hd values for the three genomic regions in relation to time since ice retreat (Fig. 5b) and slope in the Cordillera Darwin (Fig. 5c) are also represented. Figure 5b shows there is a trend of higher Hd values in areas of longer times since deglaciation, while in Fig. 5c higher Hd values correspond to northern sites. A Mantel test showed relatively low correlation between geographic and genetic distances (0.125, p value < 0.01). Position concerning Cordillera Darwin was revealed as the most determining within the factors analysed, with percentages of 92.5 and of 63.05 of variance explained (p value < 0.01) through AMOVA analyses for 16S and for rbcLX gene regions, respectively (Table 4, North vs. South location). Other factors considered, such as the location of sampling site in retreating glacier areas and the time since ice retreated, were less explicative through independent factor analyses (Table 4).

Fig. 5
figure 5

Hd values for sampling sites concerning the different environmental features taken into account in present study. a Variation in Hd values in response to Longitude coordinate of sampling sites. Hd values are represented for the three genomic regions: diamonds for 16S values and solid line for 16S tendency; squares for rbcLX values and dotted line for rbcLX tendency; and triangles for ITS values and discontinuous line for ITS tendency. b Hd values for sampling sites, split into two groups showing sites in recently deglaciated areas (left) and previously deglaciated areas (right). Hd values are represented for the three genomic regions: light grey for 16S, grey for ITS and dark grey for rbcLX. c Hd values for sampling sites, split into two groups showing northern (left) and southern sites (right). Hd values are represented for the three genomic regions: light grey for 16S, grey for ITS and dark grey for rbcLX

Full size image
Table 4 Results from AMOVA testing for 16S and rbcLX genomic regions
Full size table

The 16S phylogenetic tree showed haplotypes clustered into four different groups (Fig. 6). All these groups were relatively close to sequences of different known species retrieved from the GenBank database in BLAST searches. Thus, haplotypes V, VI, VII and VIII (group I) and haplotypes XI and XII (group II) were closely related to Nostoc edaphicum (strain X); haplotypes I, II, III and IV (group III), together with other sequences obtained from GenBank, resulted in a group sister to Nostoc sphaeroides (strain HBHF0604) and finally, haplotypes IX and X (group IV) were clustered in a supported clade that includes Nostoc flagelliforme (strain IMGA0408). In majority of cases, the most closely related strains corresponded to sequences from uncultured Nostoc found in symbiosis with hepatics or lichen-forming cyanobacteria (bi- and tripartite lichens).

Fig. 6
figure 6

Phylogenetic 50 % majority rule tree for 16S gene dataset. Lines in bold show branches that are supported in Bayesian analysis (PP > 0.95); those showing ML are supported only in maximum likelihood analysis (BP > 0.70); those marked with an asterisk are supported in both analyses. Line under the tree represents substitutions per site scale. Roman numerals refer to posterior comments

Full size image

Phylogenetic analyses for rbcLX gene haplotypes (Fig. 7) showed that all the haplotypes were also clustered in four major groups, with 14, 10, 6 and 2 haplotypes, respectively. The four groups are closely related to Nostoc sequences obtained from other symbiotic associations (mainly of the cyanolichen genus Peltigera, especially for the second and third groups). These groups could not be easily assimilated within known Nostoc taxa. The first and second groups seem to be close to Nostoc known from lichen symbioses; the third group is also close to a symbiotic Nostoc from a lichen, but in this case to the photobiont of Massalongia carnosa, which appears far from the others in the analysis. Finally, the fourth group is likely close to these Nostoc not only from different lichens but also to N. flagelliforme.

Fig. 7
figure 7

Phylogenetic 50 % majority rule tree for rbcLX gene with collapsed haplotypes/OTUs after Gblocks trimming. Lines in bold show branches that are supported in Bayesian analysis (PP > 0.95); those showing ML are supported only in maximum likelihood analysis (BP > 0.70); those marked with an asterisk are supported in both analyses. Line under the tree represents substitutions per site scale. ‘Haplotype 1’ also comprises haplotypes/OTUs 33, 34, 35, 49 and 50; ‘Haplotype 2’ comprises also haplotypes/OTUs 41, 52 and 53; ‘Haplotype 11’ comprises also haplotypes/OTUs 36, 7, 38, 46, 47, 48, 54, 55 and 56; ‘Haplotype 12’ also comprises haplotype/OTU 39; ‘Haplotype 26’ also comprises haplotype/OTU 57; ‘Haplotype 27’ also comprises haplotypes/OTUs 40 and 42; ‘Haplotype 31’ also comprises haplotypes/OTUs 43, 44, 45 and 51. Roman numerals are only for easier interpretation of posterior comments

Full size image

The biotrophic interface between symbiotic Nostoc and G. magellanica cells was analysed by LTSEM (Fig. 8). Infected Gunnera cells were numerous with larger size and more rounded shape (Fig. 8a) than uninfected ones (0 in Fig. 8b). This was especially noteworthy in later stages of infection (III in Fig. 8b). Earliest stages (I in Fig. 8b) are characterised by the intracellular location of a few Nostoc cells in some host cells. Penetration of small Nostoc cells through host cell wall was detected at these stages (arrow in Fig. 8c), which may represent first steps of infection. Microsymbiont cells were generally observed positioned close to plant cell wall (arrowheads in Fig. 8c). Nostoc cells appeared surrounded by host plasma membrane and embedded in a matrix of extracellular polymeric substances (EPS) (asterisk in Fig. 8c). In later stages of infection (II and III in Fig. 8b), host cells appeared colonised by filaments of Nostoc vegetative cells with numerous heterocysts (Fig. 8d, e). These latter cells were distinguishable by their larger size and a slightly thickened cell wall. Filaments containing heterocysts appeared embedded in an EPS matrix (asterisks in Fig. 8f).

Fig. 8
figure 8

LTSEM images of G. magellanica tissue infected with Nostoc. a Infected root tissue of G. magellanica; Gunnera cells containing Nostoc cells (asterisks) are larger than uninfected ones, also shown; b Area of root showing plant cells in different stages of infection by Nostoc cells; 0 uninfected plant cell; I plant cells at earliest stages of infection, with few Nostoc cells inside them; II intermediate stage of infection, with larger size and higher number of symbionts; III plant cell at mature infection stage, showing largest more numerous symbionts within it and with well-developed EPS matrix. c Infected Gunnera cell at an early stage of infection. Asterisk marks a filament of vegetative Nostoc cells embedded in EPS matrix. Arrowheads point to Nostoc symbiotic cells close to plant cell wall while arrow indicates Nostoc cell penetrating through host cell wall (first steps of infective processes). d Plant cell at a later infection stage (III), containing numerous filaments of Nostoc vegetative cells (V) with numerous heterocysts (H) distinguishable by larger size and thickened cell wall. Arrow points to the heterocysts cell wall showing thickness. e Infected host cell showing numerous Nostoc filaments in the proximity of its cell wall. Filaments of small vegetative cells are observed closely associated to the host cell wall (arrowhead). Filaments containing H were located inwardly. GCW Gunnera cell wall. f Detail of an infected host cell at an intermediate stage of infection (II), showing Nostoc cells bounded by host cell plasma membrane and associated with EPS fibers (asterisks). Arrows point to the three layers sourrounding Nostoc cell inside the host cell: Gunnera plasma membrane (GM), Nostoc cell wall (NW) and Nostoc membrane (NM)

Full size image

Discussion

Results showed that every G. magellanica endosymbiont analysed belongs to the genus Nostoc, but different species are involved in the Gunnera–Nostoc symbioses studied. This is consistent with previous studies that identified Nostoc as symbiont of Gunnera [30, 36, 73–75], including the South Chilean host species used in the present study, G. magellanica, and the nearby Gunnera tinctoria [37, 76]. However, previous studies included only a few specimens of G. magellanica. The extensive sampling of different colonies in the same Gunnera specimen and in different specimens from diverse sites provided a better perspective on symbiont diversity at different levels and also some clues about the factors influencing these diversity values. The phylogenetic position of the different Nostoc symbionts found in G. magellanica species has been analysed for the first time.

No genetic variability was detected in Nostoc sampled from the same host specimen of G. magellanica (sequences from 15 plants). Following these results, it is possible to state that each plant was likely infected by a single strain. These results agree with those obtained in different species of Gunnera using fingerprinting technique [37]. The lack of intraspecimen variability has also been reported for other Nostoc symbiotic associations, especially those formed with bryophytes or lichen-forming fungi [77, 78]. Bonnett and Silvester [76] found similar conclusions by means of Nostoc inoculation assays in Gunnera plants, suggesting further that not every strain is able to infect every plant. Nevertheless, Nilsson et al. [79] isolated more than one strain from one single infection in Gunnera, identifying them by using STRR-PCR fingerprinting on cultures. Thus, present results and those obtained in cited studies point at a process of selection of symbiont strains by the plant that could occur at a very early stage of the infection process. This early selection might inhibit a second infection by other strains of cyanobacteria [80].

By contrast, genetic diversity analysis showed high intra- and intersite variability of G. magellanica Nostoc symbionts in Tierra del Fuego, in spite of the short distance between the easternmost and the westernmost sites (c. 350 km). Our results showed a genetic diversity gradient from the most internal sampling areas (central part of Isla Grande de Tierra del Fuego) to the most external ones (areas near oceanic coast) (Figs. 1, 2 and 3). Sites with lower haplotype diversity values (coincident results for the three studied genomic regions) were those where glacier retreat occurred most recently (since the later Late Glacial period, 10 ka) [38, 81]. On the other hand, higher haplotype diversity values were found in sites close to areas where ice disappeared in earlier times (Last Glacial Maximum and first Late Glacial periods, 20–15 ka) [38, 81]. This tendency is also shown in Fig. 5b, despite uneven Hd values for different sites. Similar results were found by Nemergut et al. [82] in a glacier foreland of the Peruvian Andes. Easternmost sites 8 and 9 (located in very close forest and tundra areas) did not show high haplotype diversity values as expected from the length of time since ice retreat [81] and the presence of stable vegetation in the area [83]. The lower diversity observed at these sites might be more influenced by the low precipitation values for both sites, and the tundra environment conditions in site 9 [40]. High precipitation values may benefit the development of both Nostoc and G. magellanica because of their physiological requirements [34], and this could favour the Nostoc symbiotic diversity as well as hormogonia (infective filaments) development [84, 85]. This development could drive to a more intense infection processes of all competent Nostoc strains of the area. In fact, higher degrees of Nostoc diversity was indeed found in sampling sites 3, 4, 5 and 7, areas where precipitation is more abundant. This precipitation regime is due to proximity of the Pacific Ocean [38, 40] and southern exposure on a slope of Cordillera Darwin [40, 41], as well as more stable surrounding vegetation [83]. These results agree with AMOVA testing for independent factors, which showed that location with respect to the Cordillera Darwin was the factor that most explained the genetic variance (Table 4). Moreover, low correlation between geographic and genetic distances, showed by means of a Mantel test, suggests that diversity of Nostoc strains associated to G. magellanica at each site is influenced not only by the geographic position but also by a combination of several factors.

Regarding the presence or absence of different symbiotic Nostoc strains in each sampling site, it is noteworthy that while some of haplotypes were present only in sampling sites furthest from the ocean (where ice-retreatment happened later), others were only shared among sites close to the ocean. This could mean that specific Nostoc strains might be predominant in the early stages after glacial ice retreat, as was found in analyses of NifH gene sequences in a retreating alpine glacier [7]. Further investigations of Nostoc haplotypes from soil and in symbiosis with G. magellanica, including Fuegian samples as close as possible to glacier ice might help to test this fact.

The finding of two OTUs by the analysis of the 16S region of ribosomal RNA gene using a dissimilarity cut-off value of 0.03 [86–90] may indicate the presence of two distinct species. Moreover, following Hart and Sunday [91], the existence of two isolated networks of haplotypes with a 95 % parsimony probability criterion, might point to the presence of two species as well. However, four different taxa would be distinguished following the assumptions of Stackebrand and Ebers [92]: these latter authors, based on data correlation between sequence similarity and DNA-DNA reassociation experiments, considered that for prokaryotes a cut-off value of 99 % is required to discriminate between two species. In fact, phylogenetic analysis (for 16S and also for rbcLX) showed that haplotypes analysed were clustered in four groups (Figs. 6 and 7). At a higher taxonomic level (using a cut-off value of 0.05, following [86, 87, 89, 90]) all the 16S gene sequences were included in one group, indicating the presence of only the genus Nostoc.

The clusterings obtained by rbcLX sequences (without any trimming of ambiguously aligned regions) and ITS ones were equivalent to that obtained for 16S one only when lowest cut-off values were used. The two groups resulted from 16S sequences using a 3 % dissimilarity cut-off could also be obtained for both ITS and rbcLX genomic regions using cut-off values of approximately 20 % dissimilarity. Indeed, these groups for 16S, ITS and rbcLX were composed of sequences obtained from the same specimens. Rudi et al. [48] found that, depending on the Nostoc lineage of interest, rbcLX region sequences divergence displays a 2- to 35-fold-higher difference compared with 16S. With this taken into account, our results may point to the possibility of using ITS and rbcLX genomic regions not only for population studies [93, 94] but also for taxonomic purposes. Further analyses using a broad range of cyanobacterial groups are necessary to determine the optimum cut-off values for delimiting taxonomic groups by using sequence similarity in ITS and rbcLX regions, extending the criteria beyond 16S results.

Nostoc haplotypes obtained in the present study have been shown to be closely related to N. edaphicum, N. sphaeroides, and N. flagelliforme by analysis of 16S sequences. A correspondence to N. flagelliforme was also found in rbcLX analysis; this Nostoc species has been found before in symbiosis with other Gunnera species [75]. It is remarkable that many of our haplotypes showed close relationship to sequences from symbiotic Nostoc of different groups of organisms found in the northernmost areas of Europe. For instance, the sequences clustering with N. edaphicum were related to the Nostoc strains SKS2, KVJ2 or KVJ4, found in hepatics (retrieved from the GenBank database, not published), while those clustering with N. flagelliforme were related to cyanolichen symbionts from China [75].

The intracellular position of symbiotic Nostoc cells within G. magellanica has been confirmed using LTSEM. This observation was in agreement with what other authors have found using other microscopy techniques, such as TEM [95, 96]. As far as we are aware, the Nostoc–Gunnera symbiosis has not been previously examined with LTSEM. This technique allowed us to visualise the EPS matrix in which filaments of Nostoc cells are immersed through the infection process. Penetration of Nostoc cells through host cell walls and location outside the host plasma membrane could also be observed and different stages of infection characterised. The infection started with the penetration of small vegetative cells through the host cell wall (I); in later stages (II and III) filaments of Nostoc containing vegetative and heterocysts cells progressively occupied the entire host cell, as a host cell size increasing. All morphological and ultrastructural features were in agreement with those described in previous TEM studies [94, 95]. Further observations with this technique at different stages of host colonisation by Nostoc cells could increase our understanding of infection and symbiotic Nostoc differentiation processes.

From the results obtained in this study we conclude that an individual G. magellanica specimen takes up and hosts only one Nostoc strain. Although the Nostoc genetic diversity values can vary when using different markers (16S, ITS or rbcLX regions), we found a maximum of four species of Nostoc occurring in symbiosis with G. magellanica in the populations analysed. The span of time for which soils have been exposed to environmental conditions and the precipitation regime were recognised as the two prime factors influencing the values of diversity of symbiotic Nostoc in G. magellanica specimens from distinct sites in Tierra del Fuego. Through symbiosis, Nostoc may be extending its ecological niche sheltered from external predators. In this way, pioneer strains might persist in certain localities while new colonisations increase the diversity of Nostoc recognisable species by means of a dynamic process. Further studies comparing the diversity of Nostoc strains present in soils to those found in Gunnera plants are necessary to assess the degree of selectivity in this cyanobacteria–angiosperm symbiosis.