Introduction

Antarctic freshwater lake ecosystems are generally classified as oligotrophic or ultra-oligotrophic depending on the phytoplankton biomass in lake water (Padisák 2004; Lizotte 2008). These oligotrophic lakes are covered with ice for the majority of the year, leading to variation in the light environment depending on the ice thickness and snow cover (Vincent et al. 1998). Moreover, biologically harmful levels of solar radiation may penetrate through the ice cover and shallow water column (Vincent et al. 1998) because of low concentrations of UV-absorbing material (Morris et al. 1995). Benthic microbial communities in the lake produce carotenoids, xanthophyll, scytonemin, and mycosporine-like amino acids (Vincent et al. 1993a; Quesada et al. 1999; Hodgson et al. 2004; Tanabe et al. 2010), presumably to protect against intense light conditions (Kudoh et al. 2009). Despite this extreme environment, large pillar communities of aquatic mosses have been found in on the lake-bed near the Syowa Station located in East Antarctica (Imura et al. 1999). These pillars were distributed at depths greater than about 2 m (Imura et al. 1999). Analysis of lake environmental data obtained during a continuous year-long observation from January 2004 to February 2005 revealed that the chlorophyll a concentration, an indicator of primary production, was between 0.1 and 0.8 μg L−1 in the lake where moss pillars were observed (Tanabe et al. 2008), thereby classifying the environment as ultra-oligotrophic.

The exterior of the moss pillars is green with adherent oxygen bubbles resulting from photosynthesis, while the interior is packed with degraded moss tissues and gives off a strong odor of sulfur compounds (Imura et al. 2000). It is thought that an oxygen gradient exists between the exterior and interior (Kudoh et al. 2003a). Thus, the pillars have distinct redox-affected sections: oxidative exteriors and reductive interiors. Nakai et al. (2012a) reported high bacterial diversity and unique genotypes, or phylotypes, in moss pillars. Some phylotypes were distributed pillar-wide, while others were section-specific. Bacterial communities differ between the exterior and interior of the pillar and are hypothesized to participate in different biogeochemical processes. Moreover, 18S rRNA genetic diversity analysis revealed that a wide range of eukaryotic phylotypes, related to fungi, ciliates, tardigrades, rotifers, and nematodes, were present in the moss pillar ecosystem formed in ultra-oligotrophic lakes (Nakai et al. 2012b). However, SSU rRNA genes do not encode a physiological function that is important in biogeochemistry, and this sequence alone does not provide information on the metabolism of the cell. Thus, the biota involved in primary production and the formation and maintenance of this ecosystem remains unknown in ultra-oligotrophic lakes.

We investigated microorganisms with the potential for CO2 fixation in moss pillars by using genetic diversity analysis of ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO, E.C. 4.1.1.39)—an enzyme involved in CO2 fixation in the Calvin–Benson–Bassham (CBB) pathway. Since it constitutes approximately 30 % of the total protein in a plant leaf, RuBisCO is probably the most abundant protein (Chase 1993; Jensen 2000). The three known forms of RuBisCO consist of a hetero-16-unit polymer form I, a homodimer form II, and a homo-10-unit polymer form III (Badger and Bek 2008). Form I RuBisCO is generally divided into two types, green-like and red-like RuBisCOs. Green-like RuBisCO is classified into forms IA and IB, whereas red-like RuBisCO is classified into forms IC and ID based on homology of amino acid sequences, resulting in a total of four forms. Form IA RuBisCO is found primarily in proteobacteria, form IB in cyanobacteria and higher plants, form IC in proteobacteria, and form ID in heterokonts and haptophytes. Form II RuBisCO is found in proteobacteria and dinoflagellate algae. To date, form III RuBisCO has only been found in archaea (Berg et al. 2010).

The results of molecular phylogenetic analysis have been reported for a diverse range of RuBisCO genes in extreme environments, including the deep sea (Elsaied and Naganuma 2001; Elsaied et al. 2002, 2006, 2007), groundwater (Alfreider et al. 2009), caves (Chen et al. 2009), sediment from Lake Natron (Kovaleva et al. 2011), and salt lakes (Giri et al. 2004). Although the RuBisCO gene sequences from known isolated strains of Antarctic microorganisms have been reported (e.g., Liu et al. 2006), there is limited information on the genetic diversity in Antarctic moss pillars. Our data thus provide further insights into RuBisCO gene diversity among Antarctic microorganisms and the carbon cycle in Antarctic lakes.

Materials and methods

Study area

An Antarctic moss pillar specimen (diameter 22 cm; height 30 cm) was harvested on January 19, 2000, during the 42nd Japanese Antarctic Research Expedition at Lake Hotoke-Ike (69°28′S, 39°34′E), Skarvsnes, ~50 km south of Syowa Station in East Antarctica, as described in Nakai et al. (2012a). The water quality of the lake was described by Kudoh et al. (2003b).

DNA extraction

The moss pillar specimen was thawed partially and sectioned into 14 samples (seven exterior and seven interior samples) by separating the interior (blackish brown) from the exterior (green) layer and longitudinally dividing each specimen into seven horizontal sections. Exterior sections were labeled O1–O7 and interior sections were labeled I1–I7, as described previously (Nakai et al. 2012a). Sections were prepared by freeze-drying and then grinding with a sterilized mortar and pestle. Mixed microbial genomic DNA was extracted by applying a partially modified version of the bead-beating method reported by Miller et al. (1999) to 500 mg (dry weight) of each sample. Briefly, a sub-sample (100 mg dry weight) was placed in a 2-mL screw-cap microtube, with 1.2 g of sterilized 0.1-mm diameter zirconium/silica beads (BioSpec Products Inc., Bartlesville, OK, USA), 0.3 mL of phosphate buffer (100 mM NaH2PO4, pH 8.0), 0.3 mL of lysis buffer (10 % SDS, 100 mM NaCl, 500 mM Tris–HCl, pH 8.0), and 0.3 mL of chloroform-isoamyl alcohol (24:1). The microtube was set in a Mini Bead-Beater 8 (BioSpec Products) and shaken (3,200 rpm, 3 min). Beads and broken cell fragments in the tube were then removed by centrifugation (15,000 rpm, 5 min). The supernatant was purified using a Mag Extractor-Genome kit (Toyobo, Osaka, Japan) following the manufacturer’s instructions.

PCR amplification

A PCR clone library was generated from the RuBisCO gene in the purified genomic DNA. The cbbL gene of form I RuBisCO was amplified with primers IAB-595f and IAB-1385r, IC-537f and IC1212r, and ID-537f and ID-1212r (Table 1). The cbbM gene of form II RuBisCO was amplified with primers II-537f and II-1113r. The archaeal cbbL gene of form III RuBisCO was amplified with primers Arch-357f and Arch-891r. We used a previously reported primer specific to the form IB RuBisCO in Bryophyta (Tsubota et al. 1999) to obtain the near-full-length RuBisCO gene sequences derived from moss, the dominant component of moss pillars. The PCR conditions are described by Tsubota et al. (1999), Elsaied and Naganuma (2001), and Elsaid et al. (2007).

Table 1 Primers used for PCR amplification of different RuBisCO forms
Full size table

Construction of clone libraries and sequence analyses

The PCR-amplified products were cloned using a TOPO TA cloning kit (Invitrogen, Carlsbad, CA, USA) before transformation into Escherichia coli TOP10 (Invitrogen). The sequence of the inserted RuBisCO gene was determined using an ABI 3730XL automatic DNA sequencer (Applied Biosystems, Foster, USA). Similar sequences were grouped into operational RuBisCO units (ORUs) with >90 and 100 % identity for nucleotide and amino acid sequences, respectively, according to Elsaied and Naganuma (2001) and van der Wielen (2006). Inferred amino acid sequences for the ORUs were obtained using transeq (European Bioinformatics, http://www.ebi.ac.uk/emboss/transeq/). When the amino acid sequence contained the stop codon, it was excluded from further analysis. Nucleotide–nucleotide and protein–protein BLAST homology searches were conducted using the NCBI nt-database to compare the ORUs with known RuBisCOs (Altschul et al. 1997). Sequences that did not match with RuBisCO genes were excluded. We used CLUSTALX (Larkin et al. 2007) to perform alignments with ORU sequences and the closest sequences obtained from both nucleotide–nucleotide and protein–protein BLAST searches. The post-alignment sequences were used to create a phylogenetic tree by the neighbor-joining (NJ) method (Saitou and Nei 1987) using the MEGA 5.0 program (Tamura et al. 2011). The diversity of the ORUs was analyzed using the Rarefaction Calculator (http://www.biology.ualberta.ca/jbrzusto/rarefact.php). The ORU sequences were deposited in the DDBJ/EMBL/GenBank database under accession numbers AB695360 to AB695363 for bacterial form IA ORUs; AB695364 to AB695372 for bacterial form IB ORUs; AB695373 for a bryophytic form IB ORU; and AB695374 to AB695375 for bacterial form II ORUs.

Results and discussion

Distribution of RuBisCO genes obtained from the moss pillar

The form IA/IB RuBisCO gene was amplified using the primer set IAB-595f and IAB-1385r from all 14 sections (seven exterior and seven interior) of the studied moss pillar. The near-full-length form IB RuBisCO in Bryophyta was also amplified using the primer set rbcL7 and rbcL1346hR from all 14 sections. Conversely, form II was only amplified from I2 of the lower interior layer. No form IC, ID, and III RuBisCO genes were detected by PCR in any of the sections despite several PCR runs. In addition, the archaeal 16S rRNA gene was not detected in the microfloral analysis reported by Nakai et al. (2012a) for the moss pillar ecosystem. The number of archaea with the potential for CO2 fixation is low to non-existent, almost to the point of being undetectable by PCR.

A total of 29 PCR clone libraries were constructed for each section. We then randomly selected and sequenced 65 clones from each of the form IA/IB libraries, 9–10 clones from each of the bryophytic form IB libraries, and 45 clones from a form II library (total N = 1,092 clones). After determining the 1,092 clone sequences constructed from the PCR products, 884 clones were obtained for final analysis that excluded sequences containing the stop codon and sequences that did not hit with the RuBisCO gene. These 884 clones were grouped into 16 ORUs based on their ORU classification (matching >90 % of nucleotide sequences and 100 % of amino acid sequences) following to Elsaied and Naganuma (2001) and van der Wielen (2006). Four, ten, and two ORUs (out of 16) were assigned to form IA, IB, and II RuBisCO genes, respectively, by molecular phylogenetic analysis (Table 2; the distribution of each ORU in the 14 sections, results of the BLAST analysis for nucleotide identity, and amino acid identity are shown in Table S1).

Table 2 Clone number distributions in the RuBisCO gene clone libraries and phylogenetic affiliations of ORUs
Full size table

There is a high probability that the 14 form I ORUs could function as CO2 fixation enzymes because they contained unique functional motifs (DFTKDDE) (Shinozaki et al. 1983; Newman and Gutteridge 1993) specific to form I RuBisCO. The previously reported functional motifs (GGDFIKNDE) (Schneider et al. 1990) were also found in the two form II ORUs obtained in this study. The rarefaction curves for bacterial form I RuBisCO gene obtained from sections O2, O3, I3, and I5 did not reach clear saturation, indicating that further sampling of these clone libraries would have revealed little additional diversity (Fig. 1).

Fig. 1
figure 1

Rarefaction curves for the expected number of ORUs for bacterial form I RuBisCO gene based on grouping of clones that have >90 and 100 % identity for nucleotide and amino acid sequences, respectively

Full size image

Diversity of form IA RuBisCO genes from proteobacteria

We obtained 342 sequences belonging to the form IA RuBisCO gene from proteobacteria (approximately 49 % of the 702 total clones) among the clone libraries constructed with the primer set IAB-595f and IAB-1385r specific to the form IA/IB, and they were classified into four ORUs (Table 2). These were detected at roughly the same frequency as the form IB RuBisCO gene from cyanobacteria. ORU1 was most dominant (288/342 clones) of the four ORUs and was found in all 14 exterior and interior moss pillar sections (Table 2). ORU1 was most closely related to an environmental clone obtained from a tar oil-contaminated aquifer (EU926498), with 82 %ND (nucleotide identity) and 93 %AA (amino acid identity). The difference in identity between the nucleotide and amino acid sequence is influenced by the variability of the third base of the codon. A BLAST search for the RuBisCO genes of known isolates revealed that ORU1 was most closely related in nucleotide sequence to the RuBisCO gene of genus Thioalkalivibrio (CP001905) with 79 %ND, and in amino acid sequence to the RuBisCO gene of genus Bradyrhizobium (YP_001238690) with 89 %AA. ORU1 was also closely aligned on the phylogenetic tree with the genus Bradyrhizobium (Fig. 2). In addition, a previously reported bacterial 16S rRNA genetic diversity analysis (Nakai et al. 2012a) found phylotypes closely related to the genus Bradyrhizobium in all 14 exterior and interior sections. A member of the genus Bradyrhizobium can grow both heterotrophically, in the presence of oxygen, and anaerobically, as denitrifiers (van Berkum and Keyser 1985; Delgado et al. 2003). Moreover, some species fix carbon dioxide under microaerobic conditions using hydrogen as an electron donor (Hanus et al. 1979; Lepo et al. 1980), whereas others grow chemolithoautotrophically using thiosulfate as an electron donor (Masuda et al. 2010). The wide distribution of Bradyrhizobium-related sequences may reflected by the ability to grow under aerobic, microaerobic, and anaerobic conditions.

Fig. 2
figure 2

Phylogenetic tree of ORUs of the forms IA and IB RuBisCO amino acid sequences obtained from aquatic moss pillars. Evolutionary history was inferred by using the neighbor-joining method. The bootstrap consensus tree inferred from 1,000 replicates is taken to represent the evolutionary history of the taxa analyzed. Branches corresponding to partitions reproduced in <50 % bootstrap replicates were collapsed. The RuBisCO amino acid sequence of Olisthodiscus luteus (BAF80663) was used as outgroup. Accession numbers are showed in parentheses

Full size image

Another surprising finding was that ORU2 (49/342 clones), found in all sections of the moss pillar, was closely related (81–85 %ND and 95 %AA) to the RuBisCO gene sequences from hydrothermal plumes and endosymbiotic bacteria from the deep-sea mussel Bathymodiolus azoricus. The phylogenetic tree indicated that it was closely aligned with a RuBisCO sequence found in endosymbiotic bacteria from deep-sea mussels (AAX48774) (Spiridonova et al. 2006) (Fig. 2). Deep-sea mussels are thought to rely on the primary production of chemoautotrophic bacteria that live symbiotically in the branchial cells of the mussels (McKiness and Cavanaugh 2005). There are three known types of symbiotic, chemoautotrophic bacteria: sulfur-oxidizing bacteria, methane-oxidizing bacteria, and both sulfur- and methane-oxidizing bacteria (McKiness and Cavanaugh 2005). A BLAST search for the RuBisCO genes of known isolates revealed that ORU2 found in this study was most closely related in nucleotide sequence to the sulfur-oxidizing bacteria genus Thiohalophilus RuBisCO gene (GQ888587) with 82 %ND, and in amino acid sequence to the sulfur-oxidizing bacteria genus Acidithiobacillus RuBisCO gene (P0C917) with 89 %AA. Hydrogen sulfide gas was undetected in the water column and at the surface of the phytobenthos in Lake Hotoke-Ike, but ≥1 μg g−1 (wet weight) of hydrogen sulfide gas was detected from deeper than 5 mm within the phytobenthos mat samples (Kudoh et al. 2009). The low concentration at the surface is assumed due to the rapid oxidation of hydrogen sulfide by sulfur-oxidizing bacteria. Thus, sulfur-oxidizing bacteria may contribute to primary production by synthesizing organic matter via the oxidation of hydrogen sulfide in the pillar.

The two remaining ORUs shared 83 %ND and 90 %AA, and 81 %ND and 89 %AA identity, respectively, with an environmental clone obtained from tropical muddy sediments (AY773062) and an environmental clone obtained from an aquifer (EU926498).

Our results suggest that proteobacteria are distributed throughout the moss pillar, where they may play a role in primary production via chemosynthesis in the moss pillar ecosystem by fixing carbon dioxide without relying on light energy. In addition, RuBisCO genes with extremely high identity were found at the deep-sea environments and in Antarctic lakes, two environments with widely differing conditions. Bacteria with this type of RuBisCO gene may be present in other Antarctic lakes since they are thought to be widely distributed.

Diversity in form IB RuBisCO genes from mosses and cyanobacteria

We obtained 137 sequences from the clone libraries that were constructed with the primer set rbcL7 and rbcL1346hR specific to the form IB RuBisCO gene in Bryophyta, and they were classified into a single ORU (Table 2). ORU14, considered to be the major component species of the pillar, shared 100 % nucleotide sequence identity with the RuBisCO gene sequence of Leptobryum sp. isolated from Lake Hotoke-Ike (AB690288) where the moss pillars were found. This agrees with the report by Imura et al. (1999) that Leptobryum sp. comprises the main structural component of moss pillars. However, the sequences related to Bryum, a minor structural species in moss pillars, were not obtained. Previous reports have shown that Leptobryum species are the primary mosses in moss pillars, whereas Bryum pseudotriquetrum is a minor or absent (Ohtani et al. 2001). Moreover, 18S rRNA gene-based diversity analysis (Nakai et al. 2012b) revealed clones closely related to Leptobryum 18S rRNA genes in all moss pillar sections, whereas Bryum-related clones were not found. Thus, Bryum sp. may have been absent in the moss pillar samples used in this study.

We obtained 360 sequences belonging to the form IB RuBisCO gene from cyanobacteria (approximately 51 % of total 702 clones) among the clone libraries constructed with the primer set IAB-595f and IAB-1385r specific to the form IA/IB, and they were classified into nine ORUs (Table 2). Excluding ORU10, these ORUs shared comparatively high identity in the range of 81–99 %ND and 91–98 %AA with the RuBisCO gene sequences of cyanobacteria such as Anabaena, Arthrospira, Fischerella, Leptolyngbya, Microcoleus, Nostoc, Raphidiopsis, Synechococcus, and Thermosynechococcus. Cyanobacteria form mats at the bottom of Antarctic lakes and they are thought to be the dominant primary producers on the lake-bottom (Vincent 1988, 2000; Quesada et al. 2008). ORU5 was the most dominant (209/360 clones) of these nine ORUs and was found in all 14 exterior and interior sections of the moss pillar (Table 2). ORU5 was highly homologous with the RuBisCO gene sequence of known species Leptolyngbya laminosa ETS-08, with 90 %ND and 97 %AA. ORU5 was closely aligned on the phylogenetic tree with an environmental clone obtained from an Antarctic lake, Lake Bonney, in the McMurdo Dry Valleys (ACY74728) (Fig. 2). A number of species of cyanobacteria, including Lyngbya, Oscillatoria, Nostoc, and Phormidium spp., were detected by using microscopy in samples of the lake-bed from Lake Bonney, the surface of which is covered with ice year-round (Vincent et al. 1993b; Wing and Priscu 1993; Paerl and Pinckney 1996). Our data demonstrate that RuBisCO genes from geographically distant Antarctic lakes, with Lake Bonney and Hotoke-Ike located approximately 3,000 km apart, are phylogenetically related. ORU13 was close on the phylogenetic tree to an environmental clone found in Lake Bonney (ACY74730), and it formed one independent cluster.

With 87 %ND and 96 %AA identity with the RuBisCO gene sequence for the genus Leptolyngbya, ORU11 was found exclusively in the exterior layer and it clustered with the widely distributed ORU5 on the phylogenetic tree. Thus, the distribution characteristics differed even among phylogenetically related ORUs. The phylotype related to the genus Leptolyngbya were also found exclusively in the exterior layer of moss pillars in a previous bacterial 16S rRNA genetic diversity analysis (Nakai et al. 2012a), indicating a high probability that some Leptolyngbya bacterial species are specific to the exterior layer of the pillar. ORU10 differed from the other eight ORUs as they shared homology with all RuBisCO gene sequences in the database. It was most closely related to the Nostoc RuBisCO gene with 84 %ND and 83 %AA. It was found only in the deepest part of the exterior layer (O1) of the moss pillar and formed an independent cluster on the phylogenetic tree (Fig. 2).

Diversity and phylogeny of form II RuBisCO genes from proteobacteria

We obtained 45 sequences from the clone libraries constructed with the primer set II-537f and II-1113r specific to the form II RuBisCO gene, and they were classified into two ORUs (Table 2). Form II was only detected from I2 of the lower interior layer. The first, ORU15 (35/45 clones), was most closely related in nucleotide sequence to an environmental clone from Movile Cave in Romania (FJ604806) with 92 %ND, and in amino acid sequence to an environmental clone from a tar oil-contaminated aquifer (ACH70392) with 93 %AA. A BLAST search of known RuBisCO gene isolates revealed that this ORU was most closely related to the RuBisCO gene sequence of the obligately chemolithoautotrophic, sulfur-oxidizing bacterium Halothiobacillus sp. RA13 (AY099399), with 90 %ND and 95 %AA. Salt is not essential for the growth of Halothiobacillus sp., although it is known to tolerate a high salt concentration (4 M NaCl and 0.25 M sodium thiosulfate) (Kelly and Wood 2000). The freshwater lakes inhabited by Leptobryum sp., the dominant component of moss pillars, have electronic conductivity in the range of 15.0–390.0 ms/mm (Imura et al. 2003). This highest value is close to one-tenth of the value in seawater.

The second, ORU16 (10/45 clones), was related to the RuBisCO gene sequence of the Chromatiales Thioalkalicoccus limnaeus B7-1 (AEN02453), with 86 %ND and 95 %AA. This result was supported by the phylogenetic tree (Fig. 3). T. limnaeus is an alkalophilic bacterium isolated from Lake Natron in southeast Siberia, Russia with a pH range of 9.5–10.1 (Bryantseva et al. 2000), which differs significantly from the 7–8-pH range of Lake Hotoke-Ike (Kimura et al. 2010). Purple sulfur bacteria, of the order Chromatiales, generally engage in photoautotrophic metabolism using reduced inorganic sulfur compounds (e.g., hydrogen sulfide) as electron donors. Thioalkalicoccus sp. contains bacteriochlorophyll b that functions intracellularly as a photosynthetic pigment (Bryantseva et al. 2000). Thus, this group can only grow in limited areas in anaerobic environments that receive light and contain a stable supply of hydrogen sulfide. The detection of the form II ORU related to Chromatiales may indicate that light is available to phototrophs even in the lower interior part of the pillar.

Fig. 3
figure 3

Phylogenetic tree of ORUs of the form II RuBisCO amino acid sequences obtained from aquatic moss pillars. Evolutionary history was inferred by using the neighbor-joining method. The bootstrap consensus tree inferred from 1,000 replicates is taken to represent the evolutionary history of the taxa analyzed. Branches corresponding to partitions reproduced in <50 % bootstrap replicates were collapsed. The RuBisCO amino acid sequence of Nostoc punctiforme PCC73102 (ACC82570) was used as outgroup. Accession numbers are showed in parentheses

Full size image

Conclusion

We studied the distribution and diversity of all forms (forms IA, IB, IC, ID, II, and III) of the RuBisCO large subunit-encoding gene in a whole moss pillar. Proteobacterial form IA ORUs were present at the same frequency as cyanobacterial form IB ORUs. Surprisingly, the form IA ORU, which is closely related to the sequences from deep-sea environments such as hydrothermal plumes and endosymbiotic bacteria from deep-sea mussels, was detected in all moss pillar sections. The form IB ORU belonging to Bryophyta, considered to be derived from moss, was 100 % identical to the RuBisCO gene sequence from Leptobryum sp. In addition, some cyanobacterial ORUs were found exclusively in the exterior layer of the pillar, whereas form II ORUs related to obligately chemolithoautotrophic sulfur oxidizers, and purple sulfur bacteria were detected exclusively in the interior layer. This layer-specific distribution of ORUs reflects the double-layered structure of an oxidative exterior and a reductive interior layer in the pillars. This is the first report demonstrating that bacteria with the potential for CO2 fixation and chemoautotrophy are present in the Antarctic moss pillar ecosystem. CO2 fixation by chemoautotrophic bacteria is likely to be especially important during the polar night months in Antarctica. This could be verified by investigating the types and amounts of RuBisCO genes expressed in the field using a RNA-based PCR clone library analysis, or metatranscriptome and metaproteome analysis.