Recent studies have suggested that some resident Antarctic biota are of ancient origin and may have been isolated for millions of years. The phylum Tardigrada, which is part of the Antarctic terrestrial meiofauna, is of particular interest due to an impressive array of biochemical abilities to withstand harsh environmental conditions. Tardigrades are one of the few widespread Antarctic terrestrial animals that have the potential to be used as a model for evolution and biogeography on the Antarctic continent. We isolated 126 individual tardigrades from four geographically isolated soil samples from two remote nunataks in the Sør Rondane Mountains, Dronning Maud Land, Antarctica. We examined genetic variation among individuals utilising three gene regions: cytochrome c oxidase subunit I gene (COI), 18S rDNA (18S), and the wingless (Wg) gene. Comparison of sequences from worldwide and Antarctic tardigrades indicated long-term survival and isolation over glacially dominated periods in ice-free habitats in the Sør Rondane Mountains.

Introduction	Go To >> Top Introduction Materials and methods Results Discussion Acknowledgements References

Antarctica moved towards the southern polar region 100–65 million years ago, when it had a southern temperate or tropical climate (Francis 1986; Clarke and Crame 1989; Francis and Poole 2002). The opening of the Drake Passage and the Tasman Gateway between 45 and 30 million years ago separated Antarctica from other land masses (Lawver and Gahagan 2003; Livermore et al. 2007), which initiated the Antarctic circumpolar current that thermally isolated the continent from the rest of the world (Pfuhl and McCave 2005; Scher and Martin 2006). The fossil record suggests that the glaciation of Antarctica began at this time (Ashworth and Thompson 2003; Ashworth and Cantrill 2004) and the last relics of cool temperate tundra communities went extinct around 14–12 million years ago (Ashworth and Cantrill 2004; Stevens et al. 2006). Since then the extent of Antarctic ice sheets has fluctuated (Whitehead et al. 2006) with only around 0.32% of land being ice-free at present (Convey et al. 2009).

The original theory that all terrestrial biota recolonised the continent after the last glacial maximum has recently been refuted (e.g. Stevens and Hogg 2006; Convey and Stevens 2007; Convey et al. 2009; Hills et al. 2010). Understanding the Antarctic ice sheet configuration at the last glacial maximum (~17 000 years ago) has evolved over the last decade (Bentley et al. 2009; Bentley 2010). However, based on biological data, some doubt about the accuracy of ice sheet models has been raised (Convey and Stevens 2007; Convey et al. 2009; Storey et al. 2010; Magalhães et al. 2012). Antarctic terrestrial biota are often characterised by high endemism (Chown and Convey 2007) and it has been hypothesised that long-term glacial habitat fragmentation and the in situ isolation of biota was a likely cause (Stevens and Hogg 2003, 2006). In comparison, the sub-Antarctic islands show a mixed evolutionary history, combining biota isolated through glacial maxima and interglacial recent recolonisation (Stevens et al. 2006; McGaughran et al. 2010, 2011; Mortimer et al. 2011), though range expansion appears limited (Stevens and Hogg 2002; Nkem et al. 2006; Adams et al. 2006, 2007). Collectively, these biological studies suggest that much of the Antarctic biota is of ancient origin (Stevens et al. 2006; Convey and Stevens 2007). Such biological signals also support a clear division, the Gressitt line, between two main biogeographic zones comprising Antarctic Peninsula and continental Antarctica (Chown and Convey 2007; see Fig. 1). Studies show that these two zones share no nematode (Andrassy 1998), mite (Pugh 1993), or springtail (McGaughran et al. 2011) species, which is indicative of ancient and different evolutionary origins. However, the Gressitt line separation is less apparent for tardigrade morphospecies (Convey and McInnes 2005), although this has not been examined using molecular markers.

Fig. 1.  Regions of Antarctica. The Gressitt Line between maritime and continental Antarctica is marked with a dashed line. Localities of Antarctic and sub-Antarctic specimens from which comparison sequences were available are: the Sør Rondane Mountains: MI, Marion Island; CR, Chapman Ridge; BH, Bunger Hills; MMS, McMurdo Sound; AI, Alexander Island; CI, Charcot Island; KGI, King George Island; SI, Signy Island; SG, South Georgia; SR, Shackleton Range. Inset: Locality of seven samples collected from the Sør Rondane Mountains, Dronning Maud Land. Organisms could be isolated from all locations, while tardigrades were found in Tanngarden in samples cdh38, cdh39, cdh40 and in Brattnipane, sample cdh63.

Limno-terrestrial Tardigrada are of cosmopolitan distribution (Nelson 2002), and an ancient phylum that show pronounced Laurasian/Gondwana separation at both generic and familial levels in the extant fauna (McInnes and Pugh 1998, 2007; Pugh and McInnes 1998). Tardigrades are found in almost all environments, from ocean sediments to the tops of Antarctic nunataks. They are small, ranging from ~50 μm to 1200 μm in body length (Nelson 2002). The subtle morphological variations in the buccal apparatus, the insertion of the stylet muscles, and the claw morphology are important taxonomic characters (Pilato and Binda 2010; Marley et al. 2011). Due to their small size and conserved morphology, tardigrade identification at generic and species level is particularly challenging. Cryptobiosis allows tardigrades to endure abiotic stresses such as freezing, drought, ionising radiation and osmotic stress (Wright et al. 1992; Wright 2001). These physiological capabilities make them ideally adapted to survive the harsh arid Antarctic conditions (Sømme and Meier 1995).

Despite being small, lightweight and cryptobiotically adapted, the capability of tardigrades to disperse over long distances has been questioned (McInnes and Pugh 1998; Pilato and Binda 2001). Tardigrade distribution patterns cannot be accounted for in terms of anemochoric dispersal, and the occurrence of most recognised tardigrade species (over 1157) reveal distinct biogeographic affinities (McInnes and Pugh 1998, 2007; Pilato and Binda 2001). Sixty-four tardigrade species have been identified from Antarctica, of which 37 (58%) are endemic (McInnes and Pugh 2007), which is the highest proportion of endemic tardigrade species for any biogeographic region (McInnes and Pugh 1998; McInnes and Pugh 2007). This suggests long-term isolation (at least a pre-Pleistocene origin) on maritime and continental Antarctica.

Here we examined genetic variation among individual tardigrades from Sør Rondane Mountains in Dronning Maud Land (Antarctica) using three genes, 18S rDNA (18S), the nuclear wingless (Wg), and cytochrome c oxidase subunit I (COI), and compare these to individuals from other Antarctic and worldwide locations. In doing so, we ask two questions: (1) Can we identify the tardigrades of the Sør Rondane mountains by utilising vouchered (morphologically identified) sequences; and (2) Are the Sør Rondane Mountains tardigrades endemic, part of an Antarctic fauna, or part of a worldwide cosmopolitan fauna.

Materials and methods	Go To >> Top Introduction Materials and methods Results Discussion Acknowledgements References

The Sør Rondane Mountains comprise several nunataks with elevations from 1200 m to 1400 m above sea level. We sampled two main nunataks, Tanngarden and Brattnipane (Fig. 1 inset) during 1–13 February 2009. At each site soil samples were obtained from north-facing sites using a stainless steel trowel which was cleaned with snow and ethanol after each use to avoid cross contamination. Samples were double-bagged in Ziploc^® bags and frozen (-80°C).

Soil samples were thawed under tepid water, still sealed in Ziploc^® bags to avoid contamination. A sugar centrifugation method adapted from Freckman and Virginia (1993) was used to extract animals from 50–100 cm³ of soil; all equipment was cleaned with distilled water between each sample. Individual tardigrades were sorted using a stereo-microscope and each individual was placed in a separate 1.5-mL Eppendorf tube with 10 μL of distilled water using an Irwin loop or glass capillary. The Sør Rondane Mountains individuals were not morphologically identified; our aim was to assess whether it was possible to use a molecular approach to identify these animals (MOTUs).

We followed Sands et al. (2008a) for DNA extraction protocols for individual specimens. We used PCR (Saiki et al. 1988) to amplify three gene regions. Analysis of rDNA (18S) was chosen due to its suitability and prevalent use in tardigrade systematics providing suitable comparison data (e.g. Sands et al. 2008a , 2008b ). The wingless gene (Wg) was employed as this is a single-copy nuclear gene with presumed Mendelian inheritance and was demonstrated as a potentially useful population genetic marker in an earlier study investigating tardigrades (Sands et al. 2008a ), and identified phylogeographic structure in springtails (Garrick et al. 2007). The cytochrome c oxidase subunit I gene (COI) was employed to detect fine-scale diversity (e.g. species and population level) due to the expected higher mutation rate (Yang 1998).

Amplification was conducted following Sands et al. (2008a) while taking precautions against contamination, as suggested by Niederhauser et al. (1994). PCR of tardigrade individuals and control reactions were visualised with agarose gel electrophoresis. All reaction products were purified with SAPEXO (USB Corp., Cleveland, OH) and sequenced using the forward and reverse primers of Sands et al. (2008a ) employing BigDye Terminator chemistry (Applied Biosystems, Foster City, CA). Sequencing was performed on a capillary ABI3730 genetic analyser (Applied Biosystems) at Brigham Young University.

Sequences were trimmed, assembled and sorted into contigs using Geneious v5.1.6 (Drummond et al. 2012). Sequence reads that could not be successfully merged into contigs were individually compared with the NCBI nucleotide collection using blastn in BLAST 2.2.21 (Altschul et al.1990) and only those of tardigrade origin were kept. Only unique sequences were used for phylogenetic analysis; identical sequences were removed from the alignments before analysis.

Additional sequences were included from vouchered (morphologically identified) tardigrade individuals collected from continental, maritime and sub-Antarctica, plus individuals from France, UK and Greenland (C. Sands and S. J. McInnes, unpubl. data). These individuals had been identified to species or the closest species group (see Sands et al. 2008a ). Additionally, relevant tardigrade sequences were obtained from GenBank for the three gene regions. These sequences combined provided the most complete dataset possible for comparing the Sør Rondane Mountains MOTUs with worldwide, Antarctic and sub-Antarctic taxa (Fig. 1). The GenBank accession number, species (where available) and origin of each sequence is used to identify individuals in all analyses. Superfamily, order, and class followed Sands et al. (2008b) and Marley et al. (2011).

For the 18S sequences, we combined 65 eutardigrades from Antarctic, sub-Antarctic and worldwide locations (see Fig. 1) with two (of 78) representative sequences from the Sør Rondane Mountains (all accessions available from GenBank). The Wg dataset included 15 eutardigrade sequences from GenBank with 22 unique Sør Rondane Mountains sequences and with representatives from Antarctic and sub-Antarctic locations. Both the 18S and Wg analyses included Milnesium sp. (Apochela: Milnesiidae) as an outgroup. In order to minimise the effects of saturation (Yang 1998) the COI data were restricted to separate superfamilies (Macrobiotoidea and Hypsibioidea). In the Macrobiotioidea analyses we included 33 macrobiotid GenBank accessions with the 40 unique sequences from the Sør Rondane Mountains, using Dactylobiotus sp. (Murrayidae, Macrobiotoidea) as the outgroup. In the Hypsibioidea analyses, a single sequence of Acutuncus antarcticus was combined with five unidentified unique sequences from Chapman Ridge, Bunger Hills and the McMurdo ice shelf, and eight unique sequences from the Sør Rondane Mountains, with Hypsibius sp. from King George Island (South Shetland Islands, maritime Antarctica) as an outgroup.

Nucleotide alignments were performed using MAFFT (Katoh et al. 2002, 2005; Katoh and Toh 2008). The 18S alignment consisted of 1483 nucleotides (after removal of all sites where homology could not be confirmed) with 67 unique sequences. The Wg alignment was converted to amino acid translations (due to saturation) and sequences missing more than 5% of information of the entire length were excluded. The final Wg alignment consisted of a total of 22 unique sequences by 94 amino acids (282 nucleotides). The COI alignments contained 74 unique sequences with 560 nucleotides (Macrobiotioidea) and 15 unique sequences with 510 nucleotides (Hypsibioidea). All COI sequences were checked for stop codons (through amino acid translation). Sequences missing more than 10% of information for the Macrobiotioidea-alignment and more than 5% of information for the Hypsibioidea-alignment were excluded.

Testing of substitution models was conducted using jModelTest 0.1.1 (Posada 2008) and with Protest3 (Darriba et al. 2011) in case of the Wg alignment. Employment of ML approaches seemed appropriate due to the considerable variation in sequence length and overlap for the 18S dataset. RAxML 7.2.6 (Stamatakis 2006) was used for inference of phylogenetic trees, using the rapid hill-climbing mechanism, while applying a GTR Model (Tavaré 1986), Γ modelling of rate heterogeneity (Yang 1993) and an ML estimate of the α Parameter. P-Invar (Yang 1993) was not estimated alongside Γ (Gu et al. 1995). In case of Wg the JTT Model (Jones et al. 1992) was applied. Trees for 18S were calculated with 2000 ML inferences, using 2000 MP starting trees. Confidence values were estimated with 2000 non-parametric bootstrap replicates. The ML analyses for COI were calculated with the same parameters, but with 1000 ML inferences and 1000 MP starting and 1000 non-parametric bootstrap replicates. The ML analyses for Wg were calculated using the mentioned parameters, but 2000 ML inferences, 2000 MP starting trees and 2000 non-parametric bootstraps were calculated. Bayesian inference (Ronquist and Huelsenbeck 2003) was employed to compare to the results of the ML analyses. The 18S and COI analyses was calculated under a GTR model, Γ rate variation and four Γ categories, four MCMC chains, a chain length of 20 000 000 generations, a chain temperature of 0.2, subsampling frequency of 4000 generations and burn-in of 500 000 generations. The analysis for the Wg phylogeny was calculated with a fixed JTT rate matrix, using a chain length of 5 000 000 generations, Γ rate variation, subsampling frequency of 1000 and burn-in length of 500 000 generations. Trees were visualised using Dendroscope 3.2.2 (Huson et al. 2007).

The Sør Rondane Mountains COI data were used for analysis using statistical parsimony as implemented in the software TCS (Clement et al. 2000). To not impede the results of the analysis due to missing data (Joly et al. 2007), the first 26 as well the last nucleotide were removed from the Macrobiotioidea-alignment, resulting in a complete alignment of 35 sequences and 605 nucleotides. For the same reason, 18 nucleotides were removed from the Hypsibioidea-alignment resulting in a complete alignment of 7 sequences and 493 nucleotides. Although we left gaps treated as missing data (default), there were no gaps present in the alignment, and a 95% connection limit was chosen, which has been suggested as a threshold to detect separation of networks corresponding to Linnean species (Hart and Sunday 2007). Sequence divergence values were calculated with the software MEGA5 (Tamura et al. 2011). Values were obtained using the Kimura 2-parameter (Kimura 1980) and uncorrected pairwise distances using all 3 codon positions.

Results	Go To >> Top Introduction Materials and methods Results Discussion Acknowledgements References

A total of 126 tardigrade individuals were isolated from four of seven soil samples collected from the two nunataks in the Sør Rondane Mountains (Fig. 1, inset – tardigrades were found in Tanngarden in samples cdh38, cdh39, cdh40 and in Brattnipane sample cdh63). Although frozen (at -80°C) for around two years, a high proportion of animals from all soil samples were moving (alive) when sampled. Overall biodiversity reflected the results of previous studies from separate mountain ranges in the Dronning Maud Land region (Sohlenius et al. 1995, 1996, 2004). There were, however, some notable abundance differences between sampling sites with samples from Tanngarden dominated by bdelloid rotifers and Brattnipane dominated by nematodes, while mites and springtails were rare across all sites.

To identify the tardigrade fauna from the Sør Rondane Mountains we compared amplified fragments of 18S, COI and wingless gene (Wg). These gene regions have been used previously in other studies that include vouchered specimens (e.g. Sands et al. 2008a , 2008b ). Using these gene regions, we demonstrate that there are two taxonomically distinct tardigrades, or MOTUs, in the Sør Rondane Mountains. There was little intrataxon diversity in the 18S sequences and so a single representative was used for each unknown taxon in the phylogeny (Fig. 2). These were compared with a broad taxonomic range of tardigrades available either as vouchered specimens (C. Sands and S. J. McInnes, unpubl. data) or downloaded from GenBank. Both Maximum Likelihood and Bayesian analyses were conducted on each phylogeny to assure that the results were due to relationships uncovered by the data and not procedural bias. Nodes corroborated by both analyses are supported by bootstrap values and posterior probabilities (Bayesian analyses included in Supplementary Figs 1–4). In the 18S phylogeny (Fig. 2) it is clear that one MOTU belongs to the Macrobiotoidea, perhaps belonging to the genus Macrobiotus. Given the limits of our sampling and the data available for identified, unidentified and misidentified tardigrades, both the wingless (Fig. 3) and COI (Fig. 4) comparisons suggest similar relationships. The second MOTU in the 18S (Fig. 2) belongs to the Hypsibioidea and appears similar to the Antarctic endemic Acutuncus antarcticus, and again both the wingless (Fig. 3) and COI (Fig. 5) comparisons suggest similar relationships.

Fig. 2.  Phylogenetic analysis for 18S using the GTR+Γ model, with 2000 independent Maximum Likelihood tree searches using 2000 distinct Maximum Parsimony starting trees. Confidence values are drawn from 2000 pseudoreplicates using the non-parametric Bootstrap. Support values below 50 are not shown. There was little intrataxon diversity in the 18S sequences from the Sør Rondane Mountains so a single representative was used within the Hypsibioidea (JX296322, representing JX296262–76, 294 from cdh40, and JX296286, 293, 304, 310, 313, 318–20, 322, 325, 330–31, 333 from cdh38), and within the Macrobiotioidea (JX296290, representing JX296277–85 from cdh39, JX296287–92, 295–303, 305–309, 311–12, 314–17, 321, 323–4, 326–29, 332 from cdh38, and JX296335–38 from cdh63). Taxa in bold indicate the closest sequence match to the Sør Rondane Mountains sequences.

Fig. 3.  Phylogenetic analysis for the wingless gene (Wg) under a JTT+Γ model (for amino acids), with 2000 independent Maximum Likelihood tree searches using 2000 distinct Maximum Parsimony starting trees. Confidence values are drawn from 2000 pseudoreplicates using the non-parametric Bootstrap. Support values below 50 are not shown. Taxa in bold indicate the closest sequence match to the Sør Rondane Mountains sequences. All accessions starting with ‘JX’ are derived from this study.

Fig. 4.  Phylogenetic analysis for the Macrobiotioidea-alignment of COI sequences, using a GTR+Γ model, with 1000 independent Maximum Likelihood tree searches using 1000 distinct Maximum Parsimony starting trees. Confidence values are drawn from 1000 pseudoreplicates using the non-parametric Bootstrap. Support values below 50 are not shown. All accesions starting with ‘JX’ are derived from this study. Taxa in bold indicate the closest sequence match to the Sør Rondane Mountains sequences. Letters correspond to groups labelled in Table 1 (mean sequence divergence) and discussed in the text.

Fig. 5.  Phylogenetic analysis for the Hypsibioidea-related alignment of COI sequences, using a GTR+Γ model, with 1000 independent Maximum Likelihood tree searches using 1000 distinct Maximum Parsimony starting trees. Confidence values are drawn from 1000 pseudoreplicates using the non-parametric Bootstrap. Support values below 50 are not shown. All accessions starting with ‘JX’ are derived from this study. Taxa in bold indicate the closest sequence match to the Sør Rondane Mountains sequences. Letters correspond to groups labelled in Table 2 (mean sequence divergence) and discussed in the text.

The two Sør Rondane Mountains MOTUs are not an exact match for any vouchered sequence for any of the genes examined but in each case they are closely associated with either Antarctic-specific species, such as A. antarcticus, or specimens that were collected from the Antarctic or sub-Antarctic (Fig. 2). The macrobiotid MOTU falls into a clade containing several sequences attributed to the genus Macrobiotus (Fig. 2), two identified as M. furciger and the others unidentified but noted as ‘furciger-species group’ (C. Sands and S. J. McInnes, unpubl. data). Again, all individuals in the clades are of Antarctic or sub-Antarctic origin (Fig. 2). The same associations for both MOTUs are supported in the analyses of the two other genes (Figs 3, 4, 5), although taxon sampling is limited in the wingless analysis.

The Macrobiotoidea alignment of COI shows that the Sør Rondane Mountains MOTUs from Tanngarden and Brattnipane (Clade A: Fig. 4) are grouped with several other Macrobiotus spp. of Antarctic and sub-Antarctic origin (Clades B, C, D, E: Fig. 4). The mean sequence divergence between the Sør Rondane Mountains MOTUs (Clade A) and the Macrobiotus sp. from South Georgia (Clade B) is 20.3% (see Table 1 and Fig. 4). The minimum mean sequence divergence (see Table 1) observed among all sequences between neighbouring clades with morphologically identified sequences (Fig. 4) was 19.7% (Clades V–P). Numerous other clades (that include morphologically identified and unidentified sequences) reveal mean sequence divergences (see Table 1 and Fig. 4) that are above 20%.

Table 1.  Estimates of mean sequence divergence between the COI haplotypes in the Macrobiotoidea-related alignment Analyses were conducted in MEGA5, using the Kimura 2-parameter model (below diagonal) and uncorrected P-distance (above diagonal). The analysis involved 74 nucleotide sequences. All three codon positions were included. There were 511 positions in the final dataset. Letters correspond to groups labelled in Fig. 3 and discussed in the text

The Hypsibioidea COI reveals that the Tanngarden MOTUs are subdivided between individuals from cdh38 (Clade A) and cdh40 (Clade B) (Fig. 5) with mean sequence divergence of 0.6% (Table 2). These two Tanngarden clades (A, B) are, however, quite divergent (23.1–24.1%: Table 2) from Clade D comprising A. antarcticus and several closely related unknown sequences (Clade C) from a wide geographic range (from Shackleton Range, McMurdo Sound, Bunger Hills and Chapman Ridge) (see Figs 1 and 5). The mean sequence divergence between both main groups (Tanngarden and other Antarctic sequences) to the outgroup is 29.05% and 31.3%, respectively.

Table 2.  Estimates of mean sequence divergence between the COI haplotypes in the Hypsibioidea-related alignment Analyses were conducted in MEGA5, using the Kimura 2-parameter model (below diagonal) and uncorrected P-distance (above diagonal). The analysis involved 15 nucleotide sequences. All three codon positions were included. There were 482 positions in the final dataset. Letters correspond to groups labelled in Fig. 4 and discussed in the text

Population structure was identified in both MOTUs (see Figs 4, 5). The putative ‘Macrobiotus’ MOTU is clearly different from other species sampled from locations around Antarctica. Interestingly, it appears to show substructure in the COI gene even within the limited sampling region, with all haplotypes from the Brattnipane Mountain chain being more closely related to each other than to any haplotype from Tanngarden Mountain chain (Fig. 6). This is despite considerable haplotype diversity. Although the putative ‘Acutuncus’ MOTU associates closely with A. antarcticus (Figs 2, 3), it is considerably different from other A. antarcticus specimens, which is particularly noticeable when comparing COI data for the Sør Rondane Mountains fauna with vouchered specimens (Sands et al. 2008a , 2008b ; C. Sands and S. J. McInnes, unpubl. data) and other A. antarcticus ‘like’ sequences from East Antarctica (Fig. 5).

Fig. 6.  Haplotype networks calculated from COI alignments of the Sør Rondane Mountains MOTUs with statistical parsimony as implemented by the program TCS. Networks are joined by a 95% connection limit. Networks (a) and (b) are derived from the Macrobiotioidea-related alignment (but could not be joined using the 95% connection limit), and (c) from the Hypsibioidea-related alignment. Accessions for the sequence present at each haplotype are shown, followed by the sample site corresponding to the map.

Discussion	Go To >> Top Introduction Materials and methods Results Discussion Acknowledgements References

In this study we utilised three gene regions to investigate tardigrades collected at two nunataks in the Sør Rondane mountains, Dronning Maud Land, Antarctica. The aims were: (1) to identify the tardigrades of the Sør Rondane mountains by utilising vouchered (morphologically identified) sequences; and (2) to determine whether the Sør Rondane Mountains tardigrades are endemic, part of an Antarctic fauna, or part of a worldwide cosmopolitan fauna.

The results from our study showed that one clade of the Sør Rondane Mountains MOTUs falls within the Macrobiotidae family (Fig. 2). The genus Macrobiotus is globally abundant (McInnes and Pugh 1998) and common in Antarctica (Davis 1981; Ottesen and Meier 1990; Convey and McInnes 2005). Many of the macrobiotids are described as ‘species groups’, which are defined by the type species, e.g. Macrobiotus hufelandi species group. The Sør Rondane Mountains MOTUs were clearly related to Macrobiotus species, and in our worldwide dataset closest to other taxa described from the Antarctic (Fig. 2). The deep divergences in Wg and COI analyses (Figs 3, 4; Table 1) between the Sør Rondane Mountains MOTUs and the other macrobiotid taxa from Antarctic and sub-Antarctic localities implies that species-level separation and percentage sequence divergence for COI was greater than 20% (Table 1). This is comparable to that observed among other morphologically distinct species in our study. Cesari et al. (2009) found similar levels of mean sequence divergence (P-distance) among morphologically identified species (16% between M. macrocalix and M. terminalis and 24.8% between M. macrocalix and P. richtersi).

Our samples appear closely related to, but genetically separate from, the Macrobiotus furciger species group. Within Dronning Maud Land Macrobiotus species have been reported from other mountain ranges including Mühlig-Hofmannfjella (Sømme and Meier 1995), Heime-Frontfjella, Vestfjella and Schirmacher Oasis (Sohlenius et al. 1996, 2004; Sohlenius and Boström 2005). Initially (tentatively) these studies identified species as Macrobiotus furciger but have since been corrected to Macrobiotus blocki and Macrobiotus krynauwi (Sohlenius et al. 2004; Sohlenius and Boström 2005, 2006) belonging to the Macrobiotus hufelandi and Macrobiotus harmsworthi species groups, respectively. Although M. blocki and M. krynauwi have not been included in molecular studies, we can say that the Sør Rondane Mountains MOTUs do not belong to either the Macrobiotus hufelandi (Fig. 2) or Macrobiotus harmsworthi species (Fig. 4) groups as example representatives of these groups do not cluster within the Macrobiotus furciger species group (Figs 2, 4). Our samples do group within the Macrobiotus furciger species group but further morphological work is required to identify these to species. It is, however, beyond the scope of the present paper to provide further taxonomic treatment of the Sør Rondane Mountains MOTUs, which will be the subject of a separate morphological and molecular analysis to determine the new species.

The individuals of the second Sør Rondane Mountains MOTU, falling within the superfamily Hypsibioidea, were all closely related to the species Acutuncus antarcticus (see Figs 2, 3, 5), with 18S and wingless analyses (Figs 2, 3) clustered in a close relationship with samples from Antarctic and sub-Antarctic localities (Fig. 2). Acutuncus antarcticus has been widely reported from East Antarctica, West Antarctica (Antarctic Peninsula and islands), and sub-Antarctic islands (South Georgia, Macquarie and Heard Islands) (McInnes 1994; Convey and McInnes 2005). In Dronning Maud Land, Sohlenius et al. (1996, 2004) reported Hypsibius antarcticus (now A. antarcticus) from Vestfjella and Schirmacher Oasis. Fig. 5 shows two divergent lineages, one consisting of the Sør Rondane Mountains MOTUs, and the second from the combined east Antarctic (McMurdo Sound, Chapman Ridge) and vouchered Shackleton Range specimens. The COI (Fig. 5) sequence divergence of ~23% (Table 2) indicated that the Sør Rondane Mountains individuals have been isolated (although found only in Tanngarden) for a considerable period (Fig. 4, Table 2).

The geographical restriction of many limno-terrestrial tardigrades had been observed by Pilato and Binda (2001), who noted that adaptations to microenvironmental niches affect the successful movement of passively dispersed propagules. As mentioned earlier, the harsh environmental conditions of the Antarctic realms require special adaptations (Convey et al. 2008, 2009). Both McInnes and Pugh (1998) and Pilato and Binda (2001) concluded that the present distribution of tardigrade species might represent ancient widespread tardigrade species that gave rise to localised speciation after the continental land masses separated. Cromer et al. (2008) and Gibson et al. (2007) are some of the few studies dealing with palaeobiogeography of tardigrades. Tardigrade remains in temperate environments may not preserve well due to bacterial and enzymatic breakdown; however, polar regions hold subfossil records of these animals and Antarctica is one of the few places with a comparatively rich record (Cromer et al. 2008). These studies reported subfossil evidence, in particular empty eggs, of Macrobiotus furciger, Macrobiotus blocki and Acutuncus antarcticus, with ages of up to 4000, 12 500 and 42 000 years, respectively. The latter age predates the last glacial maximum in Antarctica.

The tardigrade MOTUs from the Sør Rondane Mountains appear to have been isolated in continental Antarctica and have developed levels of sequence divergence in the gene regions we have analysed that are comparable to that observed among other morphologically distinct species (e.g. Cesari et al. 2009). The deep lineage divergence found in both macrobiotid (Fig. 4, Table 1) and hypsibid (Fig. 5, Table 2) Sør Rondane Mountains MOTUs and the network analysis (Fig. 6) supports the theory that tardigrades disperse at more local levels (e.g. nearby nunataks within Brattnipane), but limited dispersal between more distant nunataks (e.g. between Brattnipane and Tangarden). It also provides evidence of long-term survival and isolation of these species within the Antarctic, though the actual refugia or means of recolonisation through glacial maxima is still unknown. Evidence for refugia through the glacial maxima is possible within the Dronning Maud Land region. Analysis of isotope records have shown that the level of ice sheet elevation in Dronning Maud Land varied over the last 10 million years (Altmaier et al. 2010), with the current level of glaciation established at ~100 000 years ago. These results indicate that the ice sheet elevations during the last glacial maximum left exposed rock formations to act as potential refugia in this region.

The careful combination of morphology and molecular techniques to provide an integrative taxonomy has been proposed (Sands et al. 2008a ; Bertolani et al. 2011; Cesari et al. 2011; Stevens et al. 2011). Macrobiotus spp. (and tardigrades in general) have proven to be difficult to morphologically identify to species (Cesari et al. 2009, 2011; Bertolani et al. 2011). The tardigrades from the Sør Rondane Mountains reveal high levels of sequence divergence that indicated the presence of two tardigrade species. We found that MOTUs can be used to identify taxa, provided previous data are available, and can indicate the potential presence of new species and biogeographic associations. Our molecular data indicate that the Sør Rondane Mountains have been isolated for a considerable period. The deep levels of sequence divergence indicated that populations on relatively close nunataks within the same mountain range have been separated and isolated, with limited population exchange over a significant period.

Acknowledgements	Go To >> Top Introduction Materials and methods Results Discussion Acknowledgements References

We thank two anonymous reviewers for helpful comments on the manuscript. We are indebted to Annick Wilmotte for her invitation to join the Belgium BELDIVA (2009) expedition to the Sør Rondane Mountains and for her assistance in the field, and to Jeremy Whiting for DNA sequencing, Peter Fretwell (BAS) for assistance with Fig. 1, and Professor Dr Martin Schlegel (University of Leipzig) for assistance (to PC) throughout the study. All laboratory work was undertaken at the British Antarctic Survey within the Coastal and Terrestrial monitoring workpackage (CJS), in collaboration with Australian Antarctic Division project ASAC 2355 (MIS and BJA) and at Brigham Young University with the support of the United States National Science Foundation ANT-0840979 (BJA). Support was provided by the European Commission for Education and Training for funding from the ERASMUS Program to PC. This paper contributes to the SCAR EBA research programs.

References	Go To >>\ Top Introduction Materials and methods Results Discussion Acknowledgements References