|Home | About | Journals | Submit | Contact Us | Français|
The use of carbon monoxide (CO) as a biological energy source is widespread in microbes. In recent years, the role of CO oxidation in superficial ocean waters has been shown to be an important energy supplement for heterotrophs (carboxydovores). The key enzyme CO dehydrogenase was found in both isolates and metagenomes from the ocean's photic zone, where CO is continuously generated by organic matter photolysis. We have also found genes that code for both forms I (low affinity) and II (high affinity) in fosmids from a metagenomic library generated from a 3,000-m depth in the Mediterranean Sea. Analysis of other metagenomic databases indicates that similar genes are also found in the mesopelagic and bathypelagic North Pacific and on the surfaces of this and other oceanic locations (in lower proportions and similarities). The frequency with which this gene was found indicates that this energy-generating metabolism would be at least as important in the bathypelagic habitat as it is in the photic zone. Although there are no data about CO concentrations or origins deep in the ocean, it could have a geothermal origin or be associated with anaerobic metabolism of organic matter. The identities of the microbes that carry out these processes were not established, but they seem to be representatives of either Bacteroidetes or Chloroflexi.
Carbon monoxide (CO) oxidation is a source of energy for a wide diversity of prokaryotes and is an important process within the global carbon cycle. There is a wide diversity of CO oxidation pathways among both archaea and bacteria (27, 28), and their wide distribution attests to both the ecological importance and ancient origin of CO oxidation. Most of these pathways are anaerobic (31, 40) and have been reported in both archaea and bacteria. However, aerobic CO oxidation is found only in a few groups of bacteria, specifically in many Actinobacteria and Proteobacteria spp. and in at least one Firmicutes sp. (for examples, see references 16, 17, 26, 35, 46, and 47). Classically, aerobic oxidation of CO has been known to be carried out in soils where, in addition to geological or anthropogenic emissions, there are local biological sources connected to plant roots and animals (15, 18, 19). However, more recently, the relevance of CO oxidation processes in the marine environment has also become clear, mostly from evidence from the fields of genomics and metagenomics (26, 42, 43).
The aerobic oxidation of CO is very amenable to genomic analysis, since the genes involved are very characteristic, and their presence in marine bacterial genomes and in metagenomic databases can be considered diagnostic. The genes required for aerobic CO oxidation were first described in detail in chemolithoautotrophic Oligotropha carboxidovorans OM5 (10, 35, 36). The enzyme CO dehydrogenase (CODH) catalyzes the oxidation of CO and water to produce carbon dioxide, two electrons, and two protons (8, 11). The electrons are transferred to an electron transfer chain and used to generate a proton gradient across the membrane. Three genes, coxL, coxM, and coxS (for large, medium, and small subunits, respectively), encode the polypeptides for the CODH enzyme. Two heterotrimers, each composed of one CoxL, CoxM, and CoxS subunit, combine to form a functional aerobic CODH enzyme. The large subunit contains the molybdenum cofactor, the medium subunit binds flavin adenine dinucleotide, and the small subunit has two iron-sulfur clusters (13). In addition to these three genes, a number of other accessory genes have also been identified (CoxB, CoxC, CoxH, CoxD, CoxE, CoxF, CoxG, CoxI, and CoxK) that are believed to be required in the processes of regulation, posttranslational modification, and anchorage of the CODH complex to the cytoplasmic membrane. A number of these accessory genes are membrane-bound proteins themselves (CoxB, CoxC, CoxH, and CoxK), containing several transmembrane helices, and indeed, in O. carboxidovorans OM5, the CODH enzyme itself has been observed to associate with the inner cytoplasmic membrane.
Based on sequence differences, genome organization, and catalytic properties, there are two types of aerobic molybdenum-based CODH (the anaerobic enzymes are a different class of genes) (20). Both forms can be readily differentiated from other molybdenum hydroxylases by phylogenetic analysis. Form I CODH (also called OMS, named after Oligotropha, Mycobacterium, and Pseudomonas) has been conclusively proven by mutagenesis experiments and X-ray crystallography (8, 32, 35) to be the key enzyme in aerobic CO oxidation by carboxydotrophic bacteria, i.e., those that can grow on CO as the sole carbon and energy source (at >10% CO concentration). The reaction mechanism has also been clearly defined. Form I CODH large-subunit CoxL can be readily diagnosed by its characteristic catalytic site motif AYXCSFR. Moreover, in all the organisms in which form I CODH genes have been identified so far, the genomic organization of the three subunits is always M→S→L. The organization of the accessory genes, however, may vary from organism to organism.
There is much less known about the other form, form II CODH (or BMS, after Bradyrhizobium, Mesorhizobium, and Sinorhizobium), which was first described in Bradyrhizobium japonicum USDA 110 (23), a gram-negative bacterial strain and a nitrogen-fixing symbiont of soybeans. Form II CODH enables these bacteria to grow, albeit slowly, in the presence of CO as the sole carbon and energy source, but the rate of CO oxidation by form II CODH of B. japonicum USDA 110 is 10 to 1,000 times lower than that for form I CODH in O. carboxidovorans OM5 and Pseudomonas carboxydohydrogena OM5. The catalytic site of the form II CoxL large subunit is AYRGAGR. The genome organization of the form II subunits is S→L→M, different from that of form I. The number of accessory genes present along with these genes is also variable (20). Form II is often found as a paralogous copy of three subunits of form I, but without the accompanying set of CODH-related genes. This is not surprising, since in most cases, the genes appear to be already associated with the form I cluster elsewhere in the genome, like in Rhodothermus marinus DSM 4252, Dinoroseobacter shibae DFL 12, and Bradyrhizobium sp. strain BTAi1.
The discovery of the role played by CODH in marine waters is relatively recent. First, it was found in the genome of Silicibacter pomeroyi DSS-3, a marine alphaproteobacterium of the Roseobacter cluster (26), which was concomitantly proven to be able to oxidize CO at low concentrations (as should be expected in marine waters). Later on the process, it was also found in metagenomic studies of surface waters for the Sargasso Sea metagenome project (26, 45). It has been proposed that many heterotrophic bacteria in surface waters are lithoheterotrophs and take advantage of the CO released by organic matter photolysis as an alternative energy source to supplement the scarce dissolved organic matter in a way akin to the photoheterotrophy mediated by proteorhodopsin or anoxygenic photosynthesis.
We recently found evidence of a CODH presence deep in the Mediterranean Sea by the end sequencing of fosmids from a metagenomic library from a 3,000-m depth in the Ionian Sea (southeast of Sicily, Italy) (24). Here we present the analysis of nine fully sequenced fosmids that were chosen on the basis of the presence of CODH cluster genes at their ends. The results confirm the presence of complete CODH clusters, including one that has the gene sequence and cluster structure of a form I CODH. Although the source of CO deep in the ocean is unclear, the frequency in which these genes were found and the retrieval of similar sequences from the deep-ocean metagenomic database of the Hawaii Ocean Time-Series (HOT) station (7, 21) point toward an important contribution of this lithotrophic metabolism deep in the ocean, similarly relevant to that found in the surface waters.
Planktonic samples were recovered at the Ionian Km3 station (Mediterranean Sea), and a metagenomic library of fosmids was constructed as described previously (24). Metagenomic DNA fragments (35 to 40 kb) were cloned in the pCC1Fos vector and replicated in Escherichia coli EPI300. This collection has a total of 20,757 clones, and the terminal sequences of ca. 5,000 inserts were sequenced, generating approximately 7.2 Mb of DNA sequence (i.e., roughly two prokaryotic genome equivalents). From these fosmid ends, those clones with a significant BLASTX hit over 1e−25 (against the nonredundant database) with any of the CODH cluster genes were chosen (at least 15 sequences gave a good hit with CoxL proteins, and 2 with CoxM). Nine of these clones were chosen, and fosmid DNA was individually isolated using the QIAprep spin miniprep kit (Qiagen). For the sequencing process, the concentration of each DNA fosmid was measured using Quant-iT PicoGreen double-stranded DNA reagent (Invitrogen) and pyrosequenced (Roche 454 GS FLX system; GATC, Constance, Germany), tagging each one individually using a multiplex identifier adaptor that contains a unique 10-base sequence that is recognized by the sequencing analysis software and allowing for automated sorting of multiplex identifier adaptor-containing reads. The average read length was 230 bp, and the average number of reads was ~4,500 per fosmid (about 20% of them belonged to E. coli EPI300 and the cloning vector). Assembly was performed with the program SeqMan (DNASTAR) using the following parameters: 20 bp of sequence overlapping and 95% similarity. All the fosmids, with the exception of KM3-45-H11, were assembled in one single contig (Table (Table11).
Protein-coding genes were predicted using GLIMMER (6) and the SEED server (29) and were further manually curated, especially the ends of the fosmids. Spacers were subsequently searched against the nonredundant database (http://www.ncbi.nlm.nih.gov/) using BLAST (1, 2) to ensure that no open reading frame (ORF) was missed. Identified ORFs were compared to known proteins in the nonredundant database using BLASTX, and all hits of >1e−5 were considered nonsignificant. COGNITOR was used for COG (clusters of orthologous groups) assignments and COG functional categories (39). Putative protein transmembrane domains were predicted using TMHMM 2.0 (22). For comparative analyses, reciprocal BLASTN and TBLASTX searches among the different fosmids were carried out, leading to the identification of regions of similarity. To allow for the interactive visualization of genomic fragment comparisons, we used Artemis Comparison Tool version 8 (4) and Perl software developed in our laboratory.
To detect the coxL, coxM, and coxS genes in environmental sequences, a first screening was done using BLASTX comparisons in the different metagenomic collections. For the HOT/A Long-Term Oligotrophic Habitat Assessment (ALOHA) (7, 21) and Global Ocean Sampling (GOS) (34, 45) environmental collections, all the sequences of >1e−5 were recovered, and their coding sequences were extracted to confirm the presence of at least one of the two domains of the CoxL (Ald-Xan-dh-C and Ald-Xan-dh-C2), CoxM (flavin adenine dinucleotide-Binding5 and CO-deh-flav-CO), and CoxS (Fer2 and Fer2-2) proteins. This was done using the hmmpfam program of the HMMER package (9). The hidden Markov models for the protein domains were obtained from the Pfam database (http://pfam.sanger.ac.uk). Positive sequences were further examined for the presence of the form I and form II catalytic site motifs AYXCSFR and AYRGAGR, respectively. Fosmid recruitments were done using TBLASTX comparisons of the metagenomic libraries against the genomic fragment. A cutoff of 30% similarity in at least 50% of the environmental sequence was used. In the case of the GOS collections, GS033 (hypersaline lagoon, Punta Cormorant, Floreana Island, Ecuador) and GS020 (freshwater, Gatun Lake, Panama Canal, Panama) samples were not used in the analysis. For the metatranscriptome analysis (where sequence size was only about 107 bp) (37), we considered only the sequences with over 50% similarity in more than 70% of their lengths. Due to the different sizes of the databases, the number of sequences was normalized by dividing by the number of megabases sequenced in each collection.
Sequences obtained and annotated in this study have been deposited in GenBank under the accession numbers GU058051 to GU058057.
We fully sequenced nine fosmids that were selected from a database of fosmid ends from the Km3 sample (see Materials and Methods). The criterium followed was to have a significant BLASTX hit (over 1e−25) to any of the CODH cluster genes. This does not guarantee that a bona fide form I or II CODH is found within the fosmid, since some of the subunits are shared with other protein clusters of different function, and in any case, being at the end of the fosmid, the relevant genes might be in the other direction and not present in the fosmid. Even with these caveats, seven fosmids that show evidence of coding either a form I or form II CODH large subunit were found. The remaining two fosmid clones had other similar proteins not related to CODH but that belonged to the same family (data not shown). The genomic fragments cloned in the fosmids had sizes between 44.1 and 36.7 kb, and with the exception of KM3-45-H11, all could be assembled in one single contig (Table (Table11).
The most interesting genome fragment found was the one in fosmid KM3-41-E12, in which we identified two CODH clusters, one belonging to form I and the other to form II (Fig. (Fig.1).1). The catalytic subunits coxL contained here were complete, as this fosmid was chosen by the partial coxE subunit that appeared at one of its ends. Also, in clone KM3-60-B01, it was possible to identify other complete coxL genes. In the other fosmids, the coxL genes were placed at the ends and were truncated (Fig. (Fig.1).1). However, except in KM3-26-C03, KM3-28-H12, and KM3-29-C02, other subunit of the CODH cluster were identified, providing more reliable evidence of the presence of a functional CODH pathway. When the diagnostic catalytic site was available, the sequences of CoxL permitted us to identify putative form I or II CODHs. When the active site was not present in the sequenced stretch (KM3-41-E12, KM3-60-B01, KM3-26-C03, and KM3-28-H12), we studied the relative positions of the other subunit genes, coxS and coxM, and the similarities with other proteins to classify them in either CODH form I or II.
Form I of the large subunit of CODH is the best-studied and the most reliable indicator for the process of CO oxidation. It has been described in several carboxydotrophs (3, 12, 30, 35). The only representative of this type of CoxL subunit was found in fosmid KM3-41-E12 (Fig. (Fig.1).1). It shows high overall similarity to other known form I CoxL proteins and has the cysteine-containing catalytic sequence motif AYXCSFR, which differentiates the CO-oxidizing form I from the larger family of molybdenum hydroxylases (including form II). In addition, the organization of the coxS and coxM genes associated with this coxL gene is M→S→L, typical of form I CODH clusters. Four other accessory genes, coxE, coxD, coxG, and coxF, are also present in the same cluster and provide good evidence of the presence of a complete and functional aerobic CODH complex in the organism to which this fosmid belongs. KM3-41-E12 CoxL has the highest similarity with form I CoxL of Rhodothermus marinus DSM 4252 (86%), a thermophilic Bacteroidetes strain from a shallow marine hot spring in Iceland (unpublished, draft genome), and with the CoxL plasmidic protein of Thermomicrobium roseum DSM 5159 (82%), an extreme thermophile isolated from a Yellowstone National Park hot spring, which has been proven experimentally to oxidize CO (47). The presence of this form I coxL, along with its accessory genes, could be a strong indicator that CO oxidation takes place at a 3,000-m depth in the Mediterranean Sea. Interestingly, the same fosmid contains another cluster of CODH genes with a form II large subunit (see below).
Among the other fosmids, three CoxL proteins could be diagnosed as form II from the identification of the conserved catalytic site AYRGAGR. The other CoxL protein encoded by the fosmid KM3-41-E12 is 60% similar to form I CoxL also contained in this fragment, but it was confirmed to belong to form II. This subunit is also very similar (69%) to form II CoxL found in the R. marinus DSM 4252. The subunits CoxS and CoxM have their best similarities with “Thermobaculum terrenum” ATCC BAA-798 (84%), an unclassified bacterium, and again with R. marinus DSM 4252 (67%), respectively. Clone KM3-60-B01 contains the subunits coxS (truncated), coxL, and coxM. These subunits have their best hits within Chloroflexi and Bacteroidetes/Chlorobi bacteria, CoxL has its best hit with Roseiflexus castenholzii DSM 1394167 (79%), and CoxM has its best hit again with R. marinus DSM 4252 (65%) (Table (Table1).1). Also, the CoxL protein encoded in KM3-28-H12 could be clearly assigned to form II CODH by its catalytic site. This subunit is not accompanied by any other subunit related to the CODH cluster and has its best similarity with form II CoxL of T. terrenum ATCC BAA-798 (63%). It has been suggested that the lack of all other subunits, such as those involved in the posttranslational modification, may indicate that this cluster could act over other substrates rather than CO (20), but it could also happen that the presence in the genome of other complete form I or form II clusters could provide the required subunits to assemble a functional CODH.
The remaining fosmids had truncated large subunits that did not contain the catalytic site. The cluster of KM3-45-H11 contained the genes coxL (truncated), coxM, coxD, and coxE (Fig. (Fig.1).1). Although the CoxL subunit lacks the active site, the gene sequence L→M corresponds to the order found in form II gene clusters. Also, its high similarity (73%) with form II CoxL of Sphaerobacter thermophilus DSM 20745 (Chloroflexi) would support this hypothesis. The other subunits had also the highest similarity hits to genes belonging to Chloroflexi genomes (Table (Table1).1). Fosmid KM3-54-A05 carried the subunits coxL (truncated), coxM, and coxG. Between the last two genes, there is a short gene for a transcriptional regulator of the MerR family. This CoxL subunit is the most divergent CoxL sequence found in this work and has its best similarity (60%) with a CoxL-like protein of the alphaproteobacterium B. japonicum USDA 110 (Fig. (Fig.1),1), the only known microbe lacking form I CoxL that has been proven experimentally to grow with CO as the sole carbon and energy source (23). The coxM gene in this fosmid appears upstream of coxL, a gene order that has not been found yet in any CODH cluster sequenced. KM3-26-C03 and KM3-29-C02 coxL genes (both truncated) are not accompanied by any other subunit related to the CODH cluster. The first one, KM3-26-C03, had a part of the catalytic site (GAGR) and therefore could be tentatively assigned to form II CoxL. Also, in this case there was a high similarity, 70%, with form II CoxL of R. marinus DSM 4252. In the case of CoxL (truncated) of KM3-29-C02, it is more similar (66%) to form II CoxL of S. thermophilus DSM 20745.
In fosmid KM3-45-H11, we found molybdopterin biosynthesis genes close to the CODH cluster. The CODH cluster is frequently located near genes related to molybdopterin biosynthesis, i.e., in Alkalilimnicola ehrlichii MLHE-1 or Jannaschia sp. strain CCS1. In some cases, the biosynthesis of the molybdopterin cofactor seems to be coupled to the transcription of the genes of the CODH cluster, so that as soon as molybdopterin is made available, it can be inserted into the CODH enzyme immediately.
One of the fosmids, KM3-29-C02, was found to contain one of the key enzymes of the reductive tricarboxylic acid (TCA) cycle, the ATP-dependent citrate lyase (Fig. (Fig.1),1), one of the three key enzymes essential for fixing carbon through this pathway. At least in Epsilonproteobacteria in deep-sea hydrothermal vents, it has been shown that functional reductive TCA cycle enzymes are present, and it is believed that they sustain the predominant primary production in these habitats (38). This raises the possibility that CO2 may be channeled into the reductive TCA cycle, thus making CO oxidation deep in the ocean an important component in fixing carbon.
The phylogenetic relations shown by the CoxL subunits of different microbes have been shown to be largely consistent with the 16S rRNA phylogeny (20). Our sequences appear to be associated to others from the Bacteroidetes/Chlorobi or Chloroflexi group. Actually, the only housekeeping gene found in our fosmids that allows an easy taxonomic placement was the ribosomal protein S1 of KM3-60-B01, which had the highest BLAST hit to the Deltaproteobacteria Stigmatella aurantiaca DW4/3-1, but with a relatively low (66%) similarity. However, the genes of the fosmids gave overall highest similarities to the Chloroflexi and Bacteroidetes genomes (Fig. (Fig.1;1; see also Table S1 in the supplemental material), supporting the classification indicated by the coxL genes. This also suggests that the CODH cluster genes found in our Km3 fosmids may belong to related microbes or are transferred horizontally so often that their sequences appear independently from the phylogenetic affiliation of the rest of the genome.
Although the presence of coxL-related genes in superficial metagenomic collections has been already established (26), the presence of similar genes deep in the ocean has not been previously investigated.
Besides the sequences deep in the Mediterranean Sea, the only deep-water column DNA sequence databases available are the HOT database in the subtropical Pacific (HF10, HF70, HF130, HF200, HF500, HF770, and HF4000; 63.95 Mb sequenced) (7) and the recently published whole-genome shotgun (WGS) library from the microbial community at a 4,000-m depth from the same sample and the same DNA preparation as the 4,000-m-depth fosmid library (HF4000) (77.43 Mb) (21). First, we have searched for the presence of the CoxL, CoxM, and CoxS proteins in these databases by using BLAST and corroborated these results, looking for their characteristic domains (see Materials and Methods). Surprisingly, the results along the water column showed that putative coxL genes were more abundant below 200 m than in superficial waters (Fig. 2a and b). The presence of one or even two of the domains only is not enough to confirm that the sequences found code for true CoxL subunits, but at least 4 sequences have the form I catalytic center below 200 m depth (3 sequences in the fosmid libraries and 1 in the 4,000-m-depth WGS library), and another 26 sequences have form II (9 in the fosmid libraries and 17 in the 4,000-m-depth WGS library) (Fig. 2a and b). For comparison, a similar search detected three form II CoxL proteins at Km3 (3,000 m), the sample from which our sequences are derived. The relative abundance (normalized to database size; see Materials and Methods) of the coxL-like genes in these samples and Km3 is shown in Fig. Fig.2a.2a. The data show that, contrary to what we expected, these genes were relatively more frequent in deeper waters. If we assume an average genome size of 3.5 Mb and consider the size of the metagenomic libraries used, the data obtained would imply that, at the least, there must be one coxL gene per genome below 200 m. This is a very high number compared with the results of the previous work with the Sargasso database, in which a ratio of one coxL gene per 14 genomes was estimated (26). We searched for CODH genes in the more recent GOS (25) surface collections (16.96 Gbp) (34), and we could find 103 form I genes that could be confirmed by their catalytic domain and 847 form II genes. The distribution in the different collections was not homogeneous, with a higher concentration in coastal waters (see Fig. S1 in the supplemental material). But still, normalizing to the size of the database, the frequency of CODH genes appears smaller than that in the deep-water sequence collections. The deep-water CoxL from the Pacific Ocean appeared related to those from the Mediterranean Sea in the case of form I, with three of the four Pacific CoxL proteins also having their best similarities with R. marinus DSM 4252 (similarities over 89%) and KM3-41-E12 CoxL (similarities near 60%). However, regarding form II, things were not so clear, as only one Pacific fosmid end from a 4,000-m depth had 80% similarity to KM3-60-B01, and the taxonomic affiliation of the remaining Pacific form II sequences seemed different, being more related to alphaproteobacterial genes.
Also, since several complete fosmids from HOT metagenomic samples have been fully sequenced, we searched for the presence of CODH gene subunits. We detected up to four coxL-like genes in fosmids from the 4,000-m collection, but only one of them, HF4000-APKG5H11, contains a CoxL form II protein with the catalytic site conserved together with the CoxM subunit. In this case, its highest similarity was also to T. roseum DSM 5159 (CoxL, 75%; CoxM, 56%). On the other hand, aside from the CODH subunits, none of these genomic clones from the HOT station have conserved synteny or any other similarities with the fosmids of the Mediterranean Sea.
We have also analyzed the presence of coxL, coxM, or coxS in the metatranscriptomic collection from biomass recovered at 25, 75, 125, and 500 m of depth in the central North Pacific Gyre (48 Mbp) (37). Presumably due to the difficulties inherent to obtaining fresh deep-ocean samples, there are no metatranscriptomic data sets from deep in the ocean. However, we found that the number of transcripts similar to coxL increased with depth from 75 m, and at 500 m, the number of cDNA sequences obtained was the same as that obtained at 25 m (Fig. (Fig.2c).2c). This fact is a clear indicator that the CODH genes found in the metagenomic collection at depths down to at least 500 m are transcribed and might be functional.
Finally, we have analyzed the recruitment of the fosmids in the HOT and GOS metagenomic collections to assess the presence of similar microbes there (Fig. (Fig.3;3; see also Fig. S2 in the supplemental material). Focusing on the fosmid containing form I coxL, KM3-41-E12, we found 1,473 sequences in the GOS collection with over 60% similarity, not a very high number considering the database size (1 sequence per 11.5 Mb). On the other hand, the aphotic HOT database contained relatively more sequences with higher similarity (19 sequences, 1 per 3.36 Mb); despite the much smaller size of the database, most of them belonged to the deeper-ocean collections (Fig. (Fig.3).3). This suggests that the genes found deep in the Mediterranean Sea are more frequently found in the aphotic zone, although they might also be present at the surface. In this fosmid, it was also observable that form I coxL recruits at higher similarity than form II, not surprising considering that form II is known to have a much wider diversity of sequences even among cultivated microbes. Similar analysis was done with the rest of the fosmids with similar results (see Fig. S2 in the supplemental material).
The finding of many genes with best hits to putative CODH coding sequences in a relatively small fosmid library from deep in the Mediterranean Sea was unexpected, since CO oxidation is known to be a common process in superficial waters but not in the abyssal region. In the photic zone, CO is known to be generated by organic matter photolysis (20, 44, 49), which is assumed to be the major source for this reduced compound. Previous work on the same deep Mediterranean Sea metagenome already hinted at the existence of CO oxidation in the bathypelagic habitat, but the evidence for a widespread occurrence of this metabolic strategy was much weaker, providing only short sequences that are not enough to diagnose this complex metabolic activity that requires the contributions from a large gene cluster. The evidence presented here indicates that bona fide CODH genes with a high probability of corresponding to CO-oxidizing microbes are at least found deep in the Mediterranean Sea but also probably in other deep oceanic environments. The fact that the form I cluster has been found in a screening of only about 7 Mb shows that indeed the presence of CODH genes is very common among organisms found deep in the Mediterranean Sea. Moreover, the same probability applies at least to the deep North Pacific, as shown by our analysis of the HOT database. That the potential for this metabolism is very widespread was confirmed by the fact of finding cDNA of sequences very similar to those of some of its components at 500 m. This is particularly remarkable, considering that form I CODHs are low-affinity enzymes that require high partial pressure of CO to be active. These findings raise the question of what could be the origin of CO in the bathypelagic habitat. In general, CO concentration decreases at deeper waters, at least for the 200-m upper layer (14). Unfortunately, there is very little information about CO concentrations and biogeochemistry in the mesopelagic and bathypelagic zones. Considering that both Km3 and HOT stations are located in tectonically active areas rich in submarine volcanoes, the geothermal origins appear to be the most likely. However, we have not found evidence for other chemolithotrophic energy generation mechanisms in the same screening, while CO is considered a rather minor product of geothermal emissions (33, 41). More data from other oceanic regions with more stable geologic settings might help prove this hypothesis. An alternative might be that the anaerobic metabolism takes place in the water column within large particles or aggregates or even in the sediment (although the mesopelagic samples from the HOT database also had potential for CODH presence). There have been observations of a strong “dark production,” which contributed a significant fraction of the diurnal CO source within the top 17 m of the ocean (5, 50). The dark CO source correlated strongly with biological oxidation rates and organic carbon, suggesting that it is the result of incomplete respiration of biologically labile organic matter. There is always the possibility that fosmid cloning might bias the results, enriching the microbes that carry CODH genes by some molecular selection. However, the fact that they have been found in independent libraries and origins and, in the case of form II, that they seem to have different biological origins (taxonomy) support the presence of real carboxydovores deep in the ocean. How relevant they are for the ecosystem functioning and its biogeochemistry remains an open question that more extensive sampling from deep-ocean waters might help to answer.
As for the identities of the microbes carrying out the process, we cannot advance a definitive answer either. However, the repetitive finding of very high similarity hits to Bacteroidetes and, more specifically, to R. marinus DSM4252 might indicate that the microbes involved belong to a psychrophilic relative of this microbe and belong to this phylum, at least the ones possessing the form I CODH homologs found in KM3-41-E12. The next phylum represented in the hits for this fosmid is the Chloroflexi; both phyla have been found to have other converging features such as the possession of similar carotenoids (44), and both groups have been found in significant numbers in deep oceanic samples by PCR 16S rRNA amplification and metagenomic studies (7, 24, 48). The repeated finding of highest similarities with genes from cultivated thermophiles in this rather cold environment is intriguing. It is even more remarkable considering the relatively high GC content characteristic of most thermophiles not found in our fosmids, and that would tend to decrease nonfunctional similarity. In any case, the taxonomic affiliation of the microbes carrying out this process deep in the ocean remains elusive, and the evidence presented is significant only to point out that the microbes that carry out the CO oxidation in the bathypelagic compartment are probably different from those possessing this metabolism in the photic zone.
This work was supported by projects GEN2007-30014-E and BIO2008-02444, and A.-B.M.-C. was supported by a Juan de la Cierva scholarship, all from the Spanish Ministerio de Ciencia e Innovación.
Published ahead of print on 2 October 2009.
†Supplemental material for this article may be found at http://aem.asm.org/.