Correspondence: Pierre E. Galand, Observatoire Océanologique de Banyuls – LECOB, Av. du Fontaulé, Banyuls sur Mer 66650, France. Tel.: +33 430 192 451; fax: +33 468 887 395; e-mail: firstname.lastname@example.org
The Clipperton lagoon in the North Pacific Ocean has been isolated from the surrounding sea for c. 160 years. It has a stratified water column that comprises an oxic and brackish upper water layer (mixolimnion) and a deep sulfuric anoxic saline layer (monimolimnion), separated by a steep pycnocline. Here, we test whether the Clipperton lagoon with its distinctive physico-chemical features, geographic isolation, recent water column stratification, and large nutrient input harbors original microbial communities. The combination of capillary electrophoresis single-strand polymorphism (CE-SSCP) fingerprinting and sequencing of cloned bacterial and archaeal 16S rRNA genes, and functional genes for methanogenesis (mcrA), methanotrophy (pmoA), and sulfate reduction (dsrAB), revealed that microbial communities and pathways were highly stratified down the water column. The mixolimnion contained ubiquitous freshwater clades of Alpha- and Betaproteobacteria, while the pycnocline contained mostly green sulfur bacteria (phylum Chlorobi). Sequences of the upper layers were closely related to sequences found in other aquatic ecosystems, suggesting that they have a strong potential for dispersal and colonization. In contrast, the monimolimnion contained new deeply branching bacterial divisions within the OP11 cluster and the Bacteroidetes, and was the most diverse of the layers. The unique environmental conditions characterizing the deep layers of the lagoon may explain the novelty of the microbial communities found at the Clipperton atoll.
The lagoon of the Clipperton atoll in the North Pacific Ocean has been enclosed and isolated from the sea for c. 160 years (Charpy et al., 2010), resulting in unique characteristics because of its lack of exchange with the open ocean. As a consequence of its tropical oceanic climate, seasonal heavy rainfalls have progressively covered the lagoon with a cap of freshwater, creating a strong vertical salinity gradient or halocline. Because the halocline has hampered mixing, the water column consists of a fresh oxic surface layer covering anoxic, saline, and sulfidic waters. The lagoon is now meromictic or rarely mixed, and its two layers are separated by a pycnocline, where steep gradients of oxygen, salinity, pH, and nutrients meet (Charpy et al., 2010). Such a meromictic and lagoonal system, described only at the Clipperton atoll, differs from stratified continental water bodies. In contrast to meromictic lakes, the lagoon has no catchment area and freshwater is only provided by direct rainfall. Microorganisms can thus colonize the isolated fresh water only by atmospheric deposition or runoff from the atoll. The atoll has other noteworthy features. It is inhabited by 110 000 birds, including the biggest colony of Sula dactylatra in the world (Weimerskirch et al., 2009), providing large nutrient inputs of about 13 tons nitrogen and 3 tons phosphorus per year (Charpy et al., 2010). The Clipperton atoll is also far away from the main land (1280 km). Geographically isolated and without anthropogenic activity, it remains a pristine environment. Finally, the salty waters of the lagoon have been isolated from the ocean for a relatively short period of time compared with meromictic saline lakes originating from sea water trapped thousands of years ago. All of these unique characteristics suggest that waters of the Clipperton lagoon may harbor microbial communities different from the ones previously described in aquatic ecosystems.
The upper mixed layers (mixolimnion) of meromictic ecosystems often consist of aerobic, fresh, or brackish water, while the deeper layers (monimolimnion) contain saltier and anoxic waters. These heterogeneous physico-chemical properties provide different habitats for the development of different microbial communities within short distances. The oxic mixolimnion favors communities typically found in lakes and rivers (Zwart et al., 2002), while the physicochemical conditions at the pycnocline favor chemolithotrophs or photolithotrophs. Green or purple photosynthetic sulfide-oxidizing bacteria are often abundant in meromictic lakes (Overmann et al., 1991; Koizumi et al., 2004; Musat et al., 2008) and in the Black Sea (Overmann et al., 1992; Manske et al., 2005), and ammonia-oxidizing archaea have been reported (Coolen et al., 2007; Pouliot et al., 2009). In the monimolimnion, the anoxia and darkness can promote chemoheterotrophic organisms such as sulfate-reducing bacteria (Casamayor et al., 2001; Koizumi et al., 2004). Anaerobic degradation of organic matter through fermentation and methanogenesis can also take place as illustrated by the presence of methanogens in anoxic waters of meromictic lakes (Ovreas et al., 1997; Lehours et al., 2005; Karr et al., 2006). Bacteria and Archaea can then oxidize methane aerobically or anaerobically, but to the best of our knowledge, anaerobic methanotrophs have never been detected in meromictic lakes.
The main goal of this study was to explore bacterial and archaeal diversities of the Clipperton lagoon. More precisely, we wanted to test whether the stratified physico-chemical conditions of the water column selected distinct microbial communities. We also wanted to examine whether microbial groups found in other meromictic systems were present in the geographically isolated Clipperton lagoon. We examined bacterial and archaeal 16S rRNA genes by capillary electrophoresis single-strand conformation polymorphism (CE-SSCP) and sequence analyses, to describe the microbial diversity along the depth profile of the lagoon. We also targeted functional genes to explore microbial pathways involved in carbon and sulfur cycling. The pathways and genes included methanogenesis [the α subunit of methyl coenzyme M reductase (mcrA)], methanotrophy [the α subunit of the particulate methane monooxygenase gene (pmoA)], and sulfate reduction [the α and β subunits of the dissimilatory sulfite reductase (dsrAB)].
Materials and methods
Sampling site, collection, DNA extraction, and flow cytometry
The Clipperton Island (10°17′N, 109°12′W) is a small atoll, with an area of 8.9 km2, located off the west coast of Mexico in the North Pacific Ocean. The lagoon is mostly 2–5-m deep and has three deep meromictic basins (33–45 m deep). Samples were collected from the ‘Oriental Fosse’ that reaches a maximum depth of 45 m. Samples were gathered from 11 different depths: three depths from the mixolimnion (5, 10, 12 m), four from the pycnocline (14, 15, 16, 17 m), and four from the monimolimnion (18, 20, 30, 40 m). Water was collected in March 2005 with a Niskin bottle from a small boat and brought immediately to the atoll for filtration or fixation. Depth profiles of the water temperature, oxygen concentration, pH, salinity, and redox potential were determined in situ using a YSI 600 probe. A volume of 4 mL of water was fixed with formaldehyde (2% final concentration) and 0.5 L was filtered successively through a 3-μm pore size 47-mm polycarbonate filter and through a 0.2-μm pore size Durapore filter. The whole cells concentrated on the filters were preserved in sucrose lysis buffer and frozen at −20 °C in sterile Whirlpak bags. DNA was extracted from the 0.2-μm filters back in the laboratory by digesting the cells with sarkosyl 1.5% and proteinase K (final concentration 0.1 mg mL−1) (Obernosterer et al., 2008). Nucleic acids were then separated in three steps with (1) phenol (2) phenol–chloroform–isoamyl alcohol mixture (25 : 24 : 1, v/v), and (3) chloroform, and precipitated with ethanol and sodium acetate solution pH 5.2 (final concentration 0.3 M). Heterotrophic and photosynthetic prokaryotes were enumerated by flow cytometry, as described earlier (Obernosterer et al., 2008). Putative Chlorobi (green sulfur bacteria) were discriminated as small cells containing chlorophyll (Casamayor et al., 2007).
CE-SSCP fingerprinting of bacterial and archaeal communities
Short fragments (~200 bp) of the V3 region of the 16S rRNA gene were amplified from all 11 depths using the universal reverse primer w049 (Delbès et al., 2000) and the TET-labeled bacterial-specific w034 forward primer, or archaea-specific w036F primer (Leclerc et al., 2001) following earlier conditions (West et al., 2008). Fragments were separated by capillary electrophoresis (CE) using an ABI310 Genetic Analyzer (Applied Biosystems). Mobility of DNA fragments between different runs was normalized using the internal standard Genescan 400-ROX (Applied Biosystems), a common baseline was set up for all the CE-SSCP profiles, and the total area under the profiles was normalized to one of the total areas. Similarity between CE-SSCP profiles was compared with the StatFingerprints software (Michelland et al., 2009). Peak areas were used to calculate Bray–Curtis distance between profiles and construct dendrograms with the Ward linkage method.
Cloning and sequencing of PCR-amplified 16S rRNA and functional genes
One sample from the mixolimnion, one from the pycnocline, and one from the monimolimnion were further analyzed by cloning and sequencing. Archaeal 16S rRNA genes were amplified with primers ARC-21F (DeLong, 1992) and ARC-1492R (Lane, 1991), and bacteria with EUB-27F (Lane, 1991) and EUB-1492R (Weisburg et al., 1991). A portion of the mcrA gene was amplified using the MCR primers, the pmoA gene with the PMO primers (pmoA189/mb661), and the dsrAB genes with the DSR primers (DSR1F/DSR4R) according to conditions described earlier (Nercessian et al., 2005). We constructed a clone library for every sample that produced a PCR product: three bacterial clone libraries (mixolimnion, pycnocline, and monimolimnion), two archaeal libraries (pycnocline and monimolimnion), two pmoA (pycnocline and monimolimnion), one mcrA (monimolimnion), and one dsrAB library (monimolimnion). PCR products were cloned with TA cloning kit (Invitrogen) and randomly chosen functional genes clones were sequenced using the M13 pair of primers and the 16S rRNA gene clones using the domain-specific forward primer. Sequences (c. 700-bp-long 16S rRNA gene and full-length functional sequences) were aligned with muscle (Edgar, 2004), checked manually, and grouped in operational taxonomic units (OTU) with dotur (Schloss & Handelsman, 2005) using a 97% similarity threshold, and a representative 16S rRNA gene sequence from each OTU was then sequenced entirely from both ends. Sequences were screened for putative chimera using CHIMERA_CHECK of the Ribosomal Database Project (Cole et al., 2003). Sequence data have been archived in the GenBank database under accession numbers HQ691895-HQ692057.
OTU definition, diversity calculations, and phylogenetic analysis
All sequences were grouped into OTUs at a 97% identity threshold, and the Shannon–Wiener diversity index and the Chao1 non-parametric richness estimator were calculated with the program dotur (Schloss & Handelsman, 2005) through Kimura-2 parameters distance matrix obtained using the dnadist program from phylip (Felsenstein, 2008). Sequences were compared with those in the GenBank database using the blast server at the National Center for Biotechnology Information (NCBI). The full-length 16S rRNA gene sequences or the approximately 500-base-pair long functional gene DNA sequences were aligned using muscle (Edgar, 2004) implemented within mega v. 5.05 (Tamura et al., 2011), taking into account the coding structure of the sequences when needed, and checked visually. Phylogenetic analyses were completed with the program phylip. dnadist was used to calculate genetic distances using Kimura-2 model. The distance tree was estimated with the phylogenetic algorithm FITCH and tested with 1000 bootstrap replicates. Phylogenetic reconstructions were checked using maximum likelihood and Bayesian analysis, using, respectively, PhyML (Guindon & Gascuel, 2003) and MrBayes 3.1.2 (Ronquist & Huelsenbeck, 2003), and similar trees were obtained.
Physico-chemical and biological characteristics of the Clipperton lagoon
Vertical profiles of physico-chemical properties illustrate the two distinct water layers separated by a steep gradient (Fig. 1). The mixolimnion (upper water layer) reached down to 13 m, the pycnocline went from 13 to 18 m depth, and the monimolimnion (lower water layer) from 18 m to the bottom (Fig. 1). Salinity was low in mixolimnion with values of 6‰, increased drastically between 13 and 18 m and reached maximum values of 35‰, slightly higher than the surrounding seawater (31‰), from 18-m depths to the bottom (Fig. 1). Oxygen saturation was highest in the mixolimnion (75%) and decreased to close to 0 below 13 m. Waters were slightly alkaline (pH = 8.8) in the mixolimnion and became acidic in the monimolimnion (pH = 6.4). The redox potential changed from −165 mV in the upper layer to −371 mV in the monimolimnion with a steep transition gradient in the pycnocline. Temperature decreased from 27.7 °C in the mixolimnion to 25.5 °C in the monimolimnion, but there was an inversion between 15 and 17 m, where temperature reached 27.4 °C (Fig. 1). Prokaryotic cells were four times more abundant in the pycnocline than in the rest of the water column, and putative green sulfur bacteria (Chlorobi) represented half of the pycnocline cells (Fig. 1). During our expedition, Charpy et al. (2010) found that the monimolimnion had very high concentrations of hydrogen sulfide (3785 μM), dissolved inorganic nitrogen (88.4 μM), total organic carbon (10643 μM), and iron (76 μg L−1).
CE-SSCP fingerprinting of bacterial and archaeal communities through the water column
The bacterial community composition through the water column of the Clipperton lagoon was examined by SSCP profiles of 16S rRNA genes amplified from 11 depths (Fig. 2). The CE-SSCP fingerprinting separated the bacterial communities into three main clusters corresponding to the three well-defined layers of the water column. One cluster contained all samples from the mixolimnion, the second had only samples from the pycnocline, and the third cluster consisted of the monimolimnion samples. Pycnocline samples formed a separate cluster, indicating that they differed the most from both mixolimnion and monimolimnion communities (Fig. 2).
Archaeal 16S rRNA genes were successfully amplified from the pycnocline and monimolimnion, but not from the mixolimnion. CE-SSCP fingerprinting separated six archaeal communities into two main clusters that corresponded to their water layer of origin. All pycnocline samples grouped together were separated from the monimolimnion samples (Fig. 2).
Diversity and novelty of microbial communities
Bacterial 16S rRNA gene clone libraries were constructed from samples taken at 5 m (mixolimnion), 16 m (pycnocline), and 20 m (monimolimnion) and yielded 139, 188, and 166 sequences for each depth, respectively. The monimolimnion community had the highest diversity, while the mixolimnion and pycnocline had similar diversity (Table 1; Supporting Information, Fig. S1). Rarefaction did not level off, indicating that the true richness of the natural communities is probably higher than the one we measured with our sequencing effort (Fig. S1).
Table 1. Diversity of archaeal and bacterial 16S rRNA gene clone libraries from the mixolimnion (5 m), the pycnocline (16 m), and the monimolimnion (20 m) of the Clipperton lagoon. Archaeal 16S rRNA genes from the mixolimnion could not be detected. Phylotypes and diversity were based on OTUs defined by > 97% similarity between sequences
No. of clones
No. of phylotypes
Libraries of archaeal 16S rRNA gene were constructed from the pycnocline and monimolimnion and yielded 115 and 99 sequences, respectively. Rarefaction curves leveled off (Fig. S1), indicating that the natural communities were well covered by our sequencing effort. Archaea were more diverse in the monimolimnion than in the pycnocline (Table 1). Archaea communities were less diverse than bacteria (Fig. S1).
We compared the similarity of 16S rRNA gene sequences from Clipperton with known 16S rRNA gene sequences deposited in GenBank (Fig. 3). Most of the sequences belonging to the mixolimnion and pycnocline communities were > 97% similar to known sequences, while 42% of the monimolimnion sequences were < 90% similar to GenBank sequences (Fig. 3). These results suggest that several members of the bacterial communities in the monimolimnion waters of the Clipperton lagoon were quite novel and possibly represent new bacterial divisions.
Composition of bacterial communities
Proteobacteria was overall the most abundant phylum in the clone libraries of 16S rRNA genes, but proportions varied between depths. With 56% of the clones, it accounted for most of the sequences in the surface library, but only 20–25% of the clones in the deepest libraries. Within Proteobacteria, the class composition varied clearly between water layers. Alpha- and Betaproteobacteria dominated the surface library, but were absent or very rare in the deeper layers (Fig. 4). The proteobacterial 16S rRNA genes were related to lake or estuary sequences (Fig. 5a), and Betaproteobacteria fell within the newly named freshwater clade XZNMC77 (Xing et al., 2009). As many as 88% were ≥ 97% similarity to their best-matching sequences from GenBank. Most were related to uncultured lineages with the exception of few sequences, such as Clip 67 identified as Roseobacter and Clip 68 as Erythrobacter (Fig. 5a). Deltaproteobacteria sequences were only detected in the two deeper layers (Fig. 4), and the majority belonged to the family Desulfobacteraceae. They had similar proportions at both depths (16% of the clones), but their phylogenetic composition differed (Data S1). Overall, only 9% of the Deltaproteobacteria sequences were ≥ 97% similarity to their best hits from GenBank. Gammaproteobacteria varied with depths and were more abundant in the mixolimnion and pycnocline (Fig. 4). At 16 m, the most abundant sequences belonged to sulfur oxidizers (Fig. 5a) (Data S1).
Bacteroidetes were the second most abundant phylum in the 16S rRNA gene clone libraries. They were more abundant in the surface waters (25% of the sequences) than in the pycnocline (9%) and monominolimnion (19%) (Fig. 4). In deeper layers, there were no clear difference in community composition between pycnocline and monimolimnion, as 75% of the deep sequences were found at both depths, and < 6% of the sequences had > 97% similarity to database sequences. There were, however, more than twice as many Bacteroidetes sequences in the monimolimnion than in the pycnocline. The two most abundant deep phylotypes (Clip 83 and Clip 84) had < 92% similarity to known 16S rRNA gene sequences and could represent a new family within Cytophagales (Fig. 5b) (Data S1).
OP11 candidate division sequences were the third most abundant overall, but were mainly present in the deeper layers (Fig. 4). They were especially abundant in the monimolimnion, where one single OP11 OTU represented 25% of all clones and dominated the community (Fig. 5c). That OTU was only 84% similar to the closest hits from the database, and thus probably represents a novel order or class of bacteria. In contrast, the pycnocline was dominated by OD1 candidate division sequences that are related to the OP11 division (Harris et al., 2004) (Fig. 4).
Chlorobi, also called green sulfur bacteria, were the fourth most abundant phylum overall and dominated the pycnocline, where it represented 37% of the sequences in the library (Fig. 4) and its abundance reached 5.76 × 106 cells mL−1 (Fig. 1). But they were rare in the monimolimnion and not detected at surface. Chlorobi diversity was low with only six different OTUs detected. These OTUs were all > 97% similar to sequences from anaerobic photosynthetic bacteria detected in pycnoclines of meromictic water bodies. Two different genera of green sulfur bacterium dominated the libraries: Prosthecochloris and Chlorobium (Fig. 5c).
Finally, Cyanobacteria and Actinobacteria were found exclusively in the surface library. Actinobacteria belonged to classical freshwater clades (Newton et al., 2011), such as Luna1-A1 (Clip72) and AC-III (Clip74).
Composition of archaeal communities
All archaeal sequences belonged to Euryarchaeota, but pycnocline communities were different from monimolimnion communities. Overall, the Marine Benthic Group D (MBGD) (Vetriani et al., 1999), which overlaps with the Deep-Sea Hydrothermal Vent Group I (DHVE 1) (Takai & Horikoshi, 1999), was the most abundant. It was mostly present in the monimolimnion, where it represented 68% of the sequences. The most abundant single MBGD OTU (OTU 3) was related to environmental sequences from hypersaline microbial mat and from marine sediments (Fig. 6). The second most abundant archaeal clade was the GN-4n1 (Jahnke et al., 2008), related to MBGD and detected in a hypersaline mat (Fig. 6). It dominated the pycnocline and was related to sequences from marine sediments. The third most abundant group belonged to methanogens that were more present in the pycnocline, belonging mostly to the class Methanosarcinales and related to deep sediments sequences. The monimolimnion also contained Methanosarcinales sequences, but the most abundant OTU belonged to Methanomicrobiales (OTU20, Fig. 6) and was related to an endosymbiont of the anaerobic free-living ciliated protozoon Metopus contortus. Archaeal 16S rRNA genes could not be amplified from the surface waters of the lagoon.
Diversity of the mcrA, pmoA, and dsrAB genes
The gene diagnostic of methanogens, mcrA, could be amplified only from the monimolimnion. All 50 sequences belonged to the order Methanomicrobiales and separated into two main clusters (Fig. 7a). The most abundant cluster was distantly related (88% sequence similarity) to sequences retrieved from brackish lake sediments. The other cluster contained fewer sequences and was related to Methanogenium boonei, originally isolated from marine sediments (Kendall et al., 2007). Methanosarcinales were present in the pycnocline, as indicated by the detection of their 16S rRNA genes, but related mcrA genes could not be amplified by the primers we used (Luton et al., 2002). This lack of amplification is consistent with a previous study reporting a low-amplification yield of Methanosarcinales with mcr primers (Banning et al., 2005).
A gene diagnostic of methanotrophic bacteria, pmoA, was amplified and cloned from all depths (5, 16, and 20 m) and yielded 65, 41, and 75 sequences, respectively. All pmoA sequences belonged to type I methanotroph family Methylococcaceae, but there was a clear difference between the pmoA composition of the mixolimnion and the monimolimnion. More than 96% of the sequences at 5 m were related to the genus Methylomicrobium, while in the monimolimnion, 92% belonged to the family Methylomonas. The pycnocline community was a mix of sequences from the upper and deeper communities, but was characterized by sequences related to the genus Methylococcus that was rarely present in the other layers (Fig. 7b).
The gene indicative of sulfate reduction, dsrAB, could only be amplified from the pycnocline. All sequences belonged to Deltaproteobacteria. Similarly to results obtained with the 16S rRNA gene, most of the sequences (92%) were related to the mXyS1 strain (Harms et al., 1999), which forms a distinct branch within the Deltaproteobacteria. The other sequences were related to Desulfobacterium aniline belonging to the same branch as mXyS1 (Fig. 7c). dsrAB genes were not detected in the monimolimnion, even though 16S rRNA sequences from typical sulfate-reducing Deltaproteobacteria were present.
The Clipperton lagoon has a stratified water column with probable meromictic properties that originated when the lagoon was isolated from the sea about 160 years ago, a short period of time in a geological perspective. Over this time, abiotic factors changed dramatically, dividing the water column into surface brackish waters and deep anoxic and sulfuric waters, separated by a short transition layer (Charpy et al., 2010). In this study, we showed that the microbial community of the Clipperton lagoon was totally different from classical marine communities, suggesting that the stratification process has transformed the original community and selected microorganisms with specific carbon- and sulfur-related metabolisms that differ dramatically between depths and from other aquatic ecosystems.
The surface layer of the Clipperton lagoon harbored bacteria mostly related to those found in lakes and estuaries belonging to Betaproteobacteria, Alphaproteobacteria, and Bacteroidetes (Zwart et al., 2002). Most remain uncultured and their metabolisms undescribed. Some of these surface bacteria may possibly use light energy for growth like the few phototrophic Cyanobacteria that were detected. Other could use methane as suggested by the presence of genes associated to methanotrophic bacteria. Most of the pmoA genes detected in the surface waters were from type I methanotrophs (Gammaproteobacteria) related to Methylomicrobium buryatense, an alkaliphilic organism with optimal growth at pH 8.5–9.5 (Kaluzhnaya et al., 2001), consistent with the alkaline conditions found at the lagoon surface. They probably oxidize the methane produced in the deeper anaerobic layers of the lagoon, using methane as sole carbon and energy source (Hanson & Hanson, 1996).
The Clipperton atoll is far from other freshwaters, with the closest land being 1280 km away in Mexico. The presence of typical freshwater clades in the lagoon, thus, indicates that these microorganisms disperse easily over large distances. Mixolimnion communities also were not very novel in terms of 16S rRNA genes, an additional indication for broad dispersion and colonization. Atmospheric dispersal through airborne particles is an efficient mechanism for the transport of viable bacteria over long distances (Hervàs et al., 2009) and could be one vector for the inoculation of freshwater strains. Birds nesting on the island could also transport microorganisms during migrations.
The pycnocline layer characterized by steep physico-chemical gradients contained high proportion of sequences and high abundance of cells identified as green sulfur bacteria (phylum Chlorobi), obligate anaerobic photoautotrophs that can oxidize sulfur compounds. The two genera detected in the lagoon (Prosthecochloris and Chlorobium) are non-motile (Overmann, 2001) and their abundance in the water column thus reflects light availability, absence of oxygen, and presence of sulfide. This balance is seen at the pycnocline of the Clipperton lagoon, where the density gradient separates the upper aerobic layer from the deep sulfuric layer. Green sulfur bacteria have been detected in the pycnoclines of stratified lakes (Gregersen et al., 2009; Taipale et al., 2009), estuaries (Schmidtova et al., 2009), and in the Black Sea (Manske et al., 2005). Interestingly, while one single cluster often dominates these systems, two clades with comparable 16S rRNA gene clone abundance were detected in the Clipperton lagoon. Prosthecochloris and Chlorobium differ in terms of shape, gas vesicles, and bacteriochlorophylls (Overmann, 2001). These phenotypic differences may illustrate adaptation to different environmental conditions and suggest that the Clipperton pycnocline harbor diverse ecological niches for different green sulfur bacteria. Interestingly, even though the pycnocline of Clipperton is strongly isolated from other similar environments, Chlorobi sequences were closely associated to known sequences (> 97% similarity). This low level of novelty may reflect the low level of diversity, generally characterizing the phylum (Overmann, 2001).
Additional bacteria involved in cycling sulfur in the Clipperton pycnocline not only included the sulfur-oxidizing bacterium Thioalkalimicrobium microaerophilum but also sulfate-reducing Deltaproteobacteria as suggested by the detection of dsrAB genes belonging to Desulfobacteraceae. Interestingly, Desulfobacterium anilini related to the mXyS1 strain, isolated on m-xylene and crude oil (Wilkes et al., 2000), was also present. They could speculatively grow on hydrocarbons formed during the anaerobic degradation of sinking organic matter trapped in the pycnocline.
Bacteroidetes may also be involved in degrading particles at the pycnocline, as they were related to the AN-BI4 strain that ferments sugars and biopolymers for energy and carbon source (Daffonchio et al., 2006). The occurrence of methanogenic archaeal 16S rRNA gene sequences further indicates that anaerobic pathways associated to organic matter degradation are present in the pycnocline. Methanogens conduct the last step of the degradation using substrates produced by fermenting bacteria. Methanogens in the Clipperton lagoon were mainly from the Methanosarcinales order related to Methanolobus zinderi, a methylotrophic methanogens isolated from saline waters (Doerfert et al., 2009), as well as a few Methanomicrobiales related to Methanocorpusculum labreanum. The finding of methanogen genes is surprising, because these archaea are not common in saline meromictic systems (Scholten et al., 2005; Pouliot et al., 2009), and when present, some have been identified as ciliate endosymbionts (Casamayor et al., 2001). One hypothesis for the coexistence of sulfate reducers and methanogens is that methanogens could use non-competitive substrate, such as methylamines and methanol (King, 1984). The presence of methylotrophic methanogen sequences in the pycnocline supports that hypothesis.
As cloning results gave clues to some microbial pathways possibly present in the lagoon's pycnocline, many pathways remain undefined. In particular, the role of the organisms containing sequences affiliated with the OD1 division (Harris et al., 2004) is still not understood. OD1, a very diverse clade, is related to the OP11 division. OD1 sequences are often found in reduced environments with sulfur compounds, but OD1 organisms have not been cultivated, and their metabolisms never described. The same holds true for the most abundant archaeal OTU [GN-4n1 (Jahnke et al., 2008)], from which sequences have been detected in saline mats and in marine sediments. No cultured representative has been obtained to date.
The monimolimnion was anaerobic, anoxic, and saline and had higher bacterial diversity than the transition layer. The most abundant bacterial OTU belonged to the candidate division OP11 (Hugenholtz et al., 1998) that is encountered in a variety of anoxic environments (Harris et al., 2004), but without a cultivated representative and unknown physiological properties. The Clipperton cluster was distantly related to all other environmental sequences (84% to closest match) and may be a new order or class of bacteria. The most abundant archaeal OTU belonged to the MBGD/DHVE1 cluster whose metabolism is also unknown. The MBGD cluster was until recently considered as benthic only, as it was always recovered from marine sediments (Teske & Sorensen, 2007). MBGD members were, however, detected in the water column of a saline, alkaline Tibetan lake (Jiang et al., 2008), suggesting a broader occurrence. Our finding extends the records of MBGD archaea and confirms that the cluster is not intrinsically benthic.
Even though the majority of the monimolimnion sequences belonged to uncultured groups, functional and 16S rRNA gene sequences gave some clues about the metabolisms possibly present in the deep layers. The detection of sequences related to Cytophagales strain AN-BI4, which grows on sugar and biopolymers (Daffonchio et al., 2006), may indicate the presence of fermenters like in the pycnocline. The detection of bacterial pmoA genes is more intriguing, as methane-oxidizing bacteria containing pMMO genes normally use oxygen as an electron acceptor. Their presence implies that micro-aerobic niches are present in the monimolimnion or that alternative electron acceptors could be used by the Clipperton methanotrophs. Anaerobic methane oxidation is documented, but is thought to be restricted to consortia of sulfate-reducing bacteria and archaea, in which the archaea oxidize methane using the mcr gene in a reverse methanogenesis reaction (Conrad, 2009). A new research on enrichment cultures indicates, however, that bacteria may be capable of anaerobic methanotrophy coupled to denitrification (Ettwig et al., 2008). Nitrate or nitrite could theoretically be electron acceptors, replacing oxygen for methane oxidation in the monimolimnion that has very high dissolved inorganic nitrogen concentrations [88.4 μM (Charpy et al., 2010)]. These are more than 200 times surface values and four times surrounding oceanic deep water values and could be associated to the large number of birds colonizing the atoll (Weimerskirch et al., 2009). Other compounds such as iron are also abundant and could also act as oxidizing agent.
Sulfur may be cycled in the monimolimnion by sulfate-reducing Desulfobacteraceae (Deltaproteobacteria) that represented the third most abundant group of sequences and must have been metabolically active, as very high concentration of sulfide was measured in the deep layers of the lagoon [3785 μM (Charpy et al., 2010)].
The bacterial community in the monimolimnion was novel, as illustrated by the new OP11-related bacterial division (Clip 15). That novelty could reflect not only the unique environmental conditions prevailing in the deep layers of the lagoon but also the geographical isolation of the atoll. However, the very short lifetime of the lagoon (c. 160 years) compared with other meromictic systems suggests that microbial novelty is not created by long-term evolutionary processes because of geographical isolation or island evolution. While surface organisms may easily disperse and colonize new habitats through atmospheric or bird transport, deep water or marine sediment organisms may have an alternative way for dispersing. These microorganisms could be present as rare dormant cells in marine waters and develop when conditions become adequate. The unique environmental conditions of the monimolimnion include anoxic waters containing high concentrations of organic carbon, fuelled by high primary surface production (five to ten times higher than in the surrounding sea), inorganic nitrogen compounds, and iron. These exceptional environmental conditions may explain the microbial novelty that characterized the Clipperton lagoon.
We thank Jean-Louis Etienne for organizing the 2005 expedition to the Clipperton Island. This work and P.L. and H.E. participation to the expedition were financially supported by the Pierre Fabre company in France and by the joint research team Pierre Fabre-UPMC-CNRS (EMR) at the Observatoire Océanologique de Banyuls-sur-mer. D.L.K. was supported by NSF grant MCB-0453993. We thank Alain Couté and Loïc Charpy for their helpful contribution to sampling operations and for their efficient help in microscopic observations and CTD measurements.