Metagenomic Comparison of Two Thiomicrospira Lineages Inhabiting Contrasting Deep-Sea Hydrothermal Environments

Background The most widespread bacteria in oxic zones of carbonate chimneys at the serpentinite-hosted Lost City hydrothermal field, Mid-Atlantic Ridge, belong to the Thiomicrospira group of sulfur-oxidizing chemolithoautotrophs. It is unclear why Thiomicrospira-like organisms thrive in these chimneys considering that Lost City hydrothermal fluids are notably lacking in hydrogen sulfide and carbon dioxide. Methodology/Principal Findings Here we describe metagenomic sequences obtained from a Lost City carbonate chimney that are highly similar to the genome of Thiomicrospira crunogena XCL-2, an isolate from a basalt-hosted hydrothermal vent in the Pacific Ocean. Even though T. crunogena and Lost City Thiomicrospira inhabit different types of hydrothermal systems in different oceans, their genomic contents are highly similar. For example, sequences encoding the sulfur oxidation and carbon fixation pathways (including a carbon concentration mechanism) of T. crunogena are also present in the Lost City metagenome. Comparative genomic analyses also revealed substantial genomic changes that must have occurred since the divergence of the two lineages, including large genomic rearrangements, gene fusion events, a prophage insertion, and transposase activity. Conclusions/Significance Our results show significant genomic similarity between Thiomicrospira organisms inhabiting different kinds of hydrothermal systems in different oceans, suggesting that these organisms are widespread and highly adaptable. These data also indicate genomic processes potentially associated with the adaptation of these lineages into strikingly different habitats.


Introduction
Microbial oxidation of sulfur is the basis of most ecosystems at seafloor hydrothermal environments. In basalt-hosted hydrothermal vents, hydrogen sulfide (H 2 S) is the most abundant electron donor driving primary production [1]. Concentrations of H 2 S are ,8 mmol/kg in typical basalt-hosted systems and can reach .40 mmol/kg at some sites [2]. Much of the animal biomass in these systems is directly supported by symbiotic H 2 S-oxidizing bacteria. At the serpentinite-hosted Lost City hydrothermal field in the Atlantic Ocean, H 2 S concentrations are lower [2], ranging between 0.05-2.8 mmol/kg in end-member hydrothermal fluids [3]. Nevertheless, mussels collected from Lost City chimneys harbor endosymbionts with close phylogenetic relationships to H 2 S-oxidizing as well as methane-oxidizing bacteria [4]. Megafaunal biomass is much lower at Lost City compared to most basalt-hosted systems, however, and the relative lack of H 2 S is the most likely cause [2]. Because fluid chemistry at Lost City is dominated by subsurface serpentinization reactions, hydrogen (H 2 , 1-15 mmol/kg) and methane (CH 4 , 1-2 mmol/kg) are much more abundant in chimney fluids than H 2 S [2]. Accordingly, archaea related to methanogens and methanotrophs comprise .80% of all detectable cells in biofilms associated with the hot, anoxic interiors of actively venting chimneys [5].
Bacteria are more abundant in biofilms attached to the outer walls of Lost City chimneys where hydrothermal fluids mix with cold, oxygenated seawater [5]. The most commonly detected bacteria in Lost City chimneys and fluids belong to the Thiomicrospira genus of Gammaproteobacteria [6,7]. Thiomicrospira species frequently inhabit zones of hydrothermal chimneys and sediments where H 2 S and oxygen are both present [8,9]. The basalt-hosted hydrothermal systems in which Thiomicrospira species are typically found are characterized by acidic fluids that contain abundant H 2 S and carbon dioxide (CO 2 ). In contrast, the fluids exiting from Lost City chimneys are alkaline (pH 9-11), contain only moderate amounts of H 2 S, and are nearly devoid of CO 2 [2,10]. Nevertheless, the most prevalent bacterial 16S rRNA sequences in these fluids are affiliated with genus Thiomicrospira [6]. No Lost City Thiomicrospira have yet been cultivated, and it is unknown how they have adapted to these extreme environmental conditions. The relatively low H 2 S concentrations in Lost City fluids may not present serious difficulties for Thiomicrospira organisms because cultivated strains are known to grow optimally at H 2 S concentrations ,1 mM [11]. The Thiomicrospira representatives at Lost City must harbor adaptations to the extremely low CO 2 concentrations and high pH of Lost City fluids, however.
Here we compare metagenomic data from a Lost City carbonate chimney containing a large number of Thiomicrospirarelated sequences to the only completed genome sequence of a Thiomicrospira organism: Thiomicrospira crunogena XCL-2 [12], which was isolated from diffuse fluids at the Galapagos Rift [13], a basalthosted hydrothermal system in the Pacific Ocean. H 2 S concentrations (,0.2 mmol/kg) in fluids venting from surface sediments at the Galapagos Rift are diluted due to extensive mixing with seawater, and H 2 S levels in subsurface sediments are estimated to be much greater [14]. Additional strains with nearly identical 16S rRNA sequences to that of T. crunogena XCL-2 [15] have been isolated from basalt-hosted systems with H 2 S concentrations around 3-7 mmol/kg [16]. In contrast, H 2 S in Lost City endmember fluids never exceeds 2.8 mmol/kg and is much lower within carbonate chimneys where end-member fluids mix with ambient seawater [3]. Lost City fluids also have higher pH and lower CO 2 concentrations than these basalt-hosted systems [16,2]. Thus the metagenomic dataset described here provides a revealing snapshot of genomic changes associated with the divergence of two lineages into geochemically distinct habitats.

DNA extraction
The data described here are a subset of the dataset first reported in [17]. The carbonate chimney sample (H03_072705_R0424) was collected from the central 'Poseidon' edifice of the Lost City Hydrothermal Field (depth, 735 m; latitude, 30.12; longitude, 242.12) on 27 July 2005 by the DSV Hercules during the 2005 Lost City Expedition aboard the R/V Ronald H. Brown. Chimney material was frozen at 280uC immediately after collection and remained frozen until onshore analysis. DNA was extracted according to a protocol modified from previous reports [6,18] and summarized here. After crushing a frozen carbonate sample with a sterile mortar and pestle, approximately 0.25-0.5 g of chimney material were placed in a 2 mL microcentrifuge tube containing 250 mL of 2x buffer AE (200 mM Tris, 50 mM EDTA, 300 mM EGTA, 200 mM NaCl, pH 8) and 2 mg of poly-dIdC (Sigma-Aldrich) and incubated at 4uC overnight to allow chelation of salts and binding of DNA to poly-dIdC. Between 36-72 replicate tubes were processed in parallel, and a total of ,1 kg of carbonate minerals were processed. This protocol involved no size fractionation between sample collection and DNA extraction. Proteinase K (final concentration 1.2 mg/mL) and 10 mL of 20% SDS were added to each tube before incubation at 37uC for at most 30 min. A further 150 mL of 20% SDS and 500 mL of phenol:chloroform:isoamyl alcohol (25:24:1 ratio by volume) were added to each tube before centrifugation at 12,000 g for 10 min. Supernatants were transferred to clean tubes for a second phenol:choloroform:isoamyl alcohol extraction. After centrifugation, supernatants were pooled into SnakeSkin dialysis tubing (Pierce Protein Research Products; Rockford, IL) and dialyzed against 20 mM EGTA overnight at 4uC. This large scale dialysis step proved to be very efficient in removing minerals and organic inhibitors. After dialysis, DNA was precipitated by adding 0.1 vol 3 M sodium acetate and 1 vol isopropanol and stored at 220uC for 2-4 hours. Pellets were collected by centrifugation at 16,000 g for 20 min at 8uC, washed once in 70% ethanol, dried in a vacuum centrifuge, and resuspended in TE (10 mM Tris, 1 mM EDTA, pH 8). Typical yield was ,35 ng of DNA per g of carbonate chimney material.

Metagenomic sequencing and annotation
Library construction and Sanger end-sequencing of pUC18 inserts was conducted according to the standard protocols at the DOE Joint Genome Institute in 2007. Two libraries were constructed from two subsamples of the same carbonate chimney sample. Reads from both libraries were combined for assembly and for analyses described here. Mean read length for the 46 361 reads from both libraries was 755 bp, and the mean length of all 6324 contigs was 1583 bp. All sequencing reads are deposited under GenBank accession numbers ACQI01006325-ACQI01026573, and assembled contigs are deposited under accession numbers ACQI01000001-ACQI01006324. End-paired sequences are those with identical names and different suffix: eg. FNHG1000.b1 and FNHG1000.g1; FOSS3464.x1 and FOSS3464.y1. Open reading frames were assigned with Glimmer [19] and compared with T. crunogena genes by blastp [20]. Potential homologs for all T. crunogena proteins were identified by searching against a database of all Lost City metagenomic contigs or all unassembled sequencing reads with tblastn [20]. Visualization of BLAST results was facilitated by use of Artemis [21]. An Artemis file containing annotated open reading frames for Lost City metagenomic contigs is available at http:// www.staff.washington.edu/braz. Files enabling identification of paired end sequences and membership of sequences in contigs are also available at the above website.

Comparison of genomic content
As previously reported [17], we obtained 35 Mb of metagenomic sequence from 46 361 shotgun reads of two pUC18 libraries constructed by the DOE Joint Genome Institute with DNA extracted from ,1 kg of a single Lost City carbonate chimney sample. A large proportion of the total shotgun reads (14.6%) had BLASTN alignments .500 bp with the Thiomicrospira crunogena XCL-2 genome [12]. Most of these reads exhibited 67-71% nucleotide similarity with T. crunogena, and very few reads contained sequences with .84% similarity ( Figure 1). These data indicate that the Lost City metagenomic dataset is dominated by a population that is genotypically uniform and contains moderate sequence similarity with T. crunogena.
Approximately half of the shotgun reads assembled into 6324 contigs, including 49 contigs .7 kb in length. Almost all of the large contigs contained open reading frames with significant sequence similarity to a Thiomicrospira crunogena XCL-2 protein, indicating that Thiomicrospira-related sequences comprise a high proportion of the metagenomic assembly ( Figure 2). The similar sequencing coverage of the largest contigs is consistent with a single population dominating the dataset (Figure 2A). These large contigs are ,38%GC ( Figure 2B), and the T. crunogena genome is 43%GC.
Our previous work has shown that multiple Thiomicrospira taxonomic units are present in Lost City chimneys, but the local conditions at each chimney determine which sequences are the most abundant [6]. The 16S rRNA sequence which dominates the youngest, warmest chimneys matches that in Lost City metagenomic contig C2148 ( Figure 3). Nearly half of all bacterial 16S rRNA clones in a library constructed from the same DNA preparation used for metagenomic sequencing showed high similarity to T. crunogena (data reported in [17]), and half of these clones were nearly identical to the sequence in contig C2148. Therefore, the largest metagenomic contigs in this study are likely to represent a Thiomicrospira population that is abundant in the youngest, warmest Lost City carbonate chimneys. Although some Lost City chimneys vent fluids containing up to 2.8 mmol/kg H 2 S, the sample used in this metagenomic study was collected from a chimney venting ,0.3 mmol/kg H 2 S (D. Butterfield, manuscript in prep.).
The large contigs contained surprisingly few archaeal sequences considering the dominance of archaea in actively venting Lost City chimneys [5]. Bacteria are known to be more abundant in the exterior, oxygenated zones of the chimneys [5], so the relative lack of archaeal sequences can be attributed to the sample containing little material from interior, anoxic zones of the chimney. These micro-scale redox zones are difficult to identify by bulk mineralogy due to pervasive mixing of ambient seawater through the highly porous carbonate chimneys [22].
Many genes in the T. crunogena XCL-2 genome [12] have putative orthologs in the Lost City metagenome; a complete list is available in the Supplementary Information as Table S1. Of the 2200 T. crunogena protein-coding genes (obtained from the Joint Genome Institute IMG database), 652 had tblastn [20] hits with E values better than 10 2100 to a Lost City contig or unassembled sequence (1217 hits better than 10 250 ; 1842 hits better than 10 210 ). Of the 358 T. crunogena proteins lacking hits with E values better than 10 210 , 214 were annotated as hypothetical proteins. Gene order is also highly conserved between the Lost City contigs and the T. crunogena genome, as exemplified by the largest contig ( Figure 4). It is clear that large genomic rearrangements have occurred since the divergence of the two lineages, however, as the Lost City contig is syntenic with two distinct T. crunogena regions separated by 161 kb (Figure 4). Interestingly, three of the open reading frames (ORFs) in this contig encode proteins that are more similar to other bacteria (Methylophaga, Marinobacter) that have been identified in Lost City chimneys and fluids [6] and are in reverse orientation with respect to surrounding ORFs. Assembly error is an unlikely explanation for this result because in each case 3-6 sequencing reads contained at least one of these three ORFs as well as a nearby ORF with high sequence similarity to T. crunogena. These observations are consistent with an origin of these ORFs by lateral gene transfer, but further sequencing and phylogenetic analyses are required to test this hypothesis.
Another striking genomic difference between T. crunogena and Lost City metagenomic sequences is the presence of a prophage genome in T. crunogena but not in the corresponding Lost City contig ( Figure 5). Genes flanking both ends of the T. crunogena prophage are present in the same Lost City contig, but no prophage sequences are present. Furthermore, a possible direct repeat caused by the insertion of the prophage genome into T. crunogena is also absent from the Lost City contig ( Figure 5). If the prophage was previously present in the Lost City sequence and subsequently lost, the direct repeat is expected to have remained. Therefore, the absence of the prophage and direct repeat in the Lost City contig indicates that the prophage was inserted into T. crunogena after its divergence from the Lost City lineage.

Comparison of sulfur utilization genes
The data presented thus far indicate that the Lost City Thiomicrospira population and T. crunogena represent two moderately closely related, but clearly distinct, lineages. We further examined the Lost City metagenomic data to examine whether these two lineages inhabiting contrasting hydrothermal systems contain similar sulfur utilization genes. T. crunogena utilizes the Sox pathway for complete oxidation of various sulfur compounds to sulfate [12]. We identified apparent homologs for all Sox genes required for sulfite-, thiosulfate-, elemental sulfur (S 0 ), and H 2 Sdependent cytochrome c reduction in Lost City metagenomic contigs ( Figure 6). Amino acid similarities between putative homologs range between 61% and 89%, and gene order appears to be conserved ( Figure 6). Interestingly, soxB and soxCD are not contiguous with soxXYZA in both the T. crunogena genome and the Lost City metagenomic contigs. It is unclear whether this genomic arrangement has been maintained in both lineages by selection or if it is a result of recent evolutionary divergence.
In most organisms utilizing the Sox pathway, all genes are organized into a single cluster, and the few exceptions may have arisen as a result of lateral gene transfer [23]. Phylogenetic discrepancies between T. crunogena SoxCD (and the highly similar Lost City SoxCD) compared to the other Sox proteins are consistent with one or more lateral gene transfer events [23]. Considering this phylogenetic interpretation, the presence of a transposase near the Lost City soxCD sequences ( Figure 6) is intriguing, and further work should investigate whether transfer of soxCD is mediated by transposases in Lost City chimneys. Scott et al. [12] have also suggested that the fragmented arrangement of Sox genes in T. crunogena could be maintained by the lack of selection pressure for regulation of a constitutively expressed pathway. Thus the Sox genes in the Lost City metagenome may be constitutively expressed, and at least some of them were likely acquired by an ancestor common to T. crunogena via lateral transfer.
Additional details of sulfur utilization by Lost City Thiomicrospira organisms can be inferred from the metagenomic data. One Lost City unassembled read (GenBank ACQI01023028) contains an open reading frame with 81% amino acid identities to a putative sulfide:quinone reductase (SQR) encoded by T. crunogena gene Tcr_1170. In Rhodobacter capsulatus, SQR is known to catalyze the reduction of H 2 S to S 0 , which accumulates as S 0 granules outside the cells, and S 0 is also deposited extracellularly by T. crunogena under certain conditions [24]. Alternatively, sulfur deposition in T. crunogena may be the result of ineffectual interactions among SoxCD, SoxYZ and SoxB due to their differing phylogenies [23]. Thiomicrospira cultivated isolates and clones collected from Lost City chimneys. Sequences with 100% identity to the V6 hypervariable region of clone LC1537B-12 and contig C2148 were found to be most abundant in a carbonate chimney aged 34 yrs venting fluids with a maximum temperature of 88uC [6]. Contig C2148 is 9 kb in length, has 42%GC, and has 8.7x coverage. The tree was constructed with Tree-Puzzle 5.2 [32] from 1278 characters aligned with MUSCLE [33]. Quartet puzzling support values are shown; nodes with less than 50% support are collapsed. Genbank accession numbers for each sequence from top to bottom: AF064545, AF069959, AF013972, AF064544, L40811, AB166731, AF013971, DQ270608, ACQI01002148, DQ270607, AF013973, AF013974, AJ237758, FJ792484, AF013975, AJ404732, AY575776, AF013978, FJ792098, DQ270609, AF016046, AF329082. doi:10.1371/journal.pone.0013530.g003 Intracellular S 0 granules have been observed in unidentified filamentous microorganisms collected from Lost City chimneys [25], but extracellular S 0 granules have not yet been reported.
The T. crunogena genome is also notable for its lack of genes encoding sulfate assimilation enzymes, indicating that it depends entirely on reduced sulfur species. This also appears to be true for the Lost City Thiomicrospira representative, as none of the metagenomic contigs with Thiomicrospira-related sequences contain ATP sulfurylase, APS kinase, or PAPS reductase. Apparent homologs for the latter two sequences were identified in very short contigs, but the taxonomic affiliation of these contigs is not readily apparent.

Comparison of carbon fixation genes
Considering the very low CO 2 concentrations in Lost City chimney fluids [2,10], it is expected that Lost City Thiomicrospira organisms harbor adaptations for living in a low CO 2 environment. Indeed, we identified metagenomic sequences encoding a partial carboxysome operon including genes for RubisCO, carboxysome shell proteins, and carbonic anhydrase (Figure 7). Carboxysomes are protein microcompartments in which CO 2 is concentrated to optimize carbon fixation by RubisCO. The carbonic anhydrase gene present in the carboxysome operon of T. crunogena is transcribed more frequently under low CO 2 conditions, consistent with its role in carbon concentration [26]. Phylogenetic analysis confirmed that the Lost City carbonic anhydrase is most closely related to this T. crunogena gene (Tcr_0841; data not shown), suggesting that it may be associated with adaptation to the low CO 2 levels at Lost City. Also present in the Lost City metagenomic dataset is a SulP-type sulfate transporter (88% identities to Tcr_1533), which shares some sequence similarity with proteins involved in bicarbonate transport into the cell [27], but it is unknown whether this protein is responsible for generating elevated concentrations of intracellular bicarbonate in Thiomicrospira.
In the genomes of all obligate autotrophs including T. crunogena, RubisCO genes are not located near other enzymes of the Calvin-Benson-Bassham cycle [12]. In the Lost City metagenomic dataset, none of the contigs containing putative RubisCO genes also contain other sequences likely involved in the Calvin-Benson-Bassham cycle. Also, an open reading frame with high similarity to phosphoribulokinase of T. crunogena (Tcr_0013) is present in a Lost City contig without any RubisCO genes. This fragmented arrangement of carbon fixation genes is consistent with the Lost City Thiomicrospira representative being an obligate autotroph.
The genome of T. crunogena encodes three different RubisCO enzymes, two form I RubisCOs and one form II Rubisco [12]. The form II RubisCO is only expressed under high CO 2 concentrations [27]. The Lost City metagenomic data includes apparent homologs for the two form I RubisCOs (Figure 7 and Table S1), but none of the assembled contigs contain a form II RubisCO. (Two unassembled sequences encode form II Rubis-COs, but their closest relatives do not include T. crunogena). Although no firm conclusions can be drawn from the absence of genes in an incomplete metagenomic dataset, it appears that the low CO 2 concentrations in Lost City fluids has rendered the high-CO 2 form II RubisCO unnecessary for Lost City Thiomicrospira.

Comparison of transposase sequences
We have previously observed that the Lost City metagenome contains a surprisingly high abundance and diversity of sequences Figure 6. The Sox sulfur oxidation system of T. crunogena [12], which encodes all enzymes necessary for complete oxidation of H 2 S to sulfate, is also present in metagenomic sequences from Lost City. In most organisms utilizing the Sox pathway, all genes are organized into a single cluster, but in both T. crunogena and the Lost City metagenome, soxB and soxCD are not contiguous with soxXYZA. Amino acid identities between putative orthologs are shown; note that the Lost City metagenome contains an additional copy of soxA not associated with other sox genes, and both copies are only 61% identical to T. crunogena and only 90% identical to each other. The Lost City soxX and soxB sequences are incomplete. The presence of a transposase downstream of Lost City soxCD is of potential interest considering strong evidence indicating lateral transfer of sulfur oxidation genes among bacteria [23]. Accession numbers for contigs C170, C4405, C2106, and C5579 are ACQI01000170, ACQI01004405, ACQI01002106, and ACQI01005579, respectively. doi:10.1371/journal.pone.0013530.g006  that encode transposases [17]. The transposase sequences were found in all of the smallest, highest coverage contigs and were rarely found in large contigs, suggesting an origin from viruses or extragenomic molecules. Another possibility is that these contigs represent genomic regions that are not amenable to assembly into larger contigs. Genomic regions containing transposase sequences can be expected to be highly variable, and such non-consensus sequences could not be assembled into large contigs with the sequencing effort of this study. Therefore, it is plausible that many of the transposase-containing contigs identified in our earlier report [17] represent variable regions of the Lost City Thiomicrospira pangenome, which differ among closely related strains within the Lost City Thiomicrospira phylotype.
To test this hypothesis, we examined whether the largest Thiomicrospira-like contigs are physically linked with unassembled transposase sequences; i.e. we searched for paired end sequences of the cloned metagenomic fragments where one member of the pair was assembled into a large contig and one member contained a transposase sequence. Figure 8 shows that 16 of the 17 contigs .10 kb contain at least one sequence paired with a transposasecontaining sequence. Of the 1294 sequences that comprise these contigs, 49 (3.8%) were paired with a transposase sequence. If we conservatively estimate that a single transposase sequence is associated with each of the 16 contigs (244 kb), then a completed genomic assembly should contain one transposase per 15 kb. In comparison, T. crunogena contains only 20 transposases: 0.8% of the genome, or one per 120 kb (from the annotation at http://img.jgi. doe.gov). Although these results must be considered preliminary until a finished genome sequence is obtained, they are suggestive that the Lost City Thiomicrospira pangenome has a high transposase content.
This analysis also sheds some light on the nature of the extremely high-coverage, transposase-containing contigs (the topleft cluster in Figure 8). The arrows in Figure 8 indicate that three of these contigs are paired with sequences on three of the large, Thiomicrospira-like contigs. Therefore, these transposases are probably part of the Lost City Thiomicrospira genome, and their extremely high coverage (eg. .30x for contig C5672) indicates that they must be present in multiple copies. These examples are rare, however. Most of the high-coverage, transposase-containing contigs were not physically associated with the large, Thiomicrospira-like contigs, and their origin remains unclear. Future genomic sequencing may prove that these contigs, too, represent genomic regions present as multiple copies, but the present data are also consistent with the previous suggestion that they are derived from viruses or small extragenomic molecules [17].

Ecological and evolutionary implications
Among the 358 T. crunogena genes without close relatives in the Lost City metagenome, a few have apparent ecological implications. For example, the phosphonate operon (encodes the ability to utilize organic phosphate) in T. crunogena is not syntenic with any Lost City contigs, and only some of the genes in the phosphonate operon appear to have homologs in Lost City sequences, In contrast, Lost City contigs contain sequences encoding both the low-affinity and high-affinity (inorganic) phosphate uptake systems present in T. crunogena. Three putative sulfonate transporters in T. crunogena are also missing in the Lost City sequences. Therefore, it is possible that Lost City Thiomicrospira are dependent on inorganic forms of phosphorus and sulfur, although no firm conclusions can be drawn due to the incomplete nature of the metagenome.
Previous studies have noted the inability of T. crunogena to utlize hydrogen gas (H 2 ) as a sole electron donor despite the presence of a Ni/Fe hydrogenase operon in its genome [12,28]. Considering the abundance of H 2 in Lost City fluids [2], the ability of Lost City Thiomicrospira to utilize H 2 with this hydrogenase operon is of interest. One contig in our dataset (C4437) appears to contain a partial Ni/Fe hydrogenase operon, but these sequences share little similarity with those in T. crunogena and most likely derive from other lineages. Due to the incomplete nature of the metagenomic dataset, however, we cannot conclude with certainty whether Lost City Thiomicrospira harbor hydrogenases.
Considering their apparent inability to utilize the most abundant electron donor (H 2 ) and their dependence on two substrates (H 2 S and CO 2 ) that are notably lacking in Lost City fluids, it is remarkable that Thiomicrospira are the most widespread and abundant bacteria in Lost City carbonate chimneys [6,7]. The surprising dominance of these organisms is most likely related to patterns of oxygen and CO 2 availability in carbonate chimneys. Those organisms expected to utilize H 2 at Lost City (Methanosarcinalesand Desulfotomaculum-related organisms) are anaerobic and more prevalent in the anoxic, interior zones of carbonate chimneys where the aerobic Thiomicrospira are unlikely to survive [5,6]. Pervasive mixing of oxygenated seawater throughout the highly porous structure of the carbonate chimneys [22] could explain the relative lack of Epsilonproteobacteria, as these organisms are typically restricted to areas with low oxygen concentrations [29]. Thiomicrospira, in contrast, can thrive in fully oxic environments as long as reduced sulfur species are present [8]. Furthermore, Thiomicrospira may out-compete Epsilonproteobacteria as well as all other bacteria in Lost City fluids because they harbor a carbon-concentrating mechanism that allows them to remain autotrophic at very low CO 2 concentrations (Figure 7). Isotopic analyses indicate that bacteria in Lost City carbonate chimneys are extremely carbonlimited [30], so a carbon-concentration mechanism could be highly advantageous.

Conclusions
In summary, metagenomic analyses of a Lost City carbonate chimney reveal a dominant Thiomicrospira population with similar Figure 8. Contigs that encode transposases (blue points) are generally small and high-coverage. Yellow points represent contigs that do not encode transposases but contain at least one sequencing read from the end of a clone that contains a transposase on the opposite end. Arrows indicate examples of contigs linked by such paired end sequences. One of these examples involves a large contig that encodes a transposase and is also linked to a small transposasecontaining contig (blue and yellow point). These links among contigs suggest that the Lost City Thiomicrospira pangenome contains a large number of transposase sequences. doi:10.1371/journal.pone.0013530.g008 genomic content to a similar but distinct species, Thiomicrospira crunogena XCL-2. Despite inhabiting different types of hydrothermal systems in different oceans, both Thiomicrospira representatives share genes encoding functions that appear to be crucial for thriving in Lost City carbonate chimneys: the ability to aerobically oxidize reduced sulfur species and to concentrate CO 2 intracellularly. This genomic similarity likely reflects a recent evolutionary divergence and that both lineages inhabit niches where H 2 Scontaining hydrothermal fluids mix with oxygenated seawater.
Further sequencing and physiological experiments will be necessary to identify particular genomic differences associated with living in different environments, but the data presented here indicate that substantial genomic evolution has occurred since the divergence of these two lineages. For example, genetic recombination appears to have been a major factor, evidenced by the large break in synteny and two gene fusion events illustrated in Figure 4. Furthermore, the evolution of the Thiomicrospira lineage has been strongly influenced by lateral gene transfer, as there is strong evidence for lateral transfer events both before (SoxCD phylogeny; [23]) and after (prophage insertion; Figure 3) the divergence of the Lost City Thiomicrospira population and T. crunogena. The extremely high transposase content of Lost City Thiomicrospira genomic regions ( Figure 8) is also strongly suggestive that lateral gene transfer has played an important role in its evolution. Future experiments should investigate whether transposase activity has promoted diversification of the dominant Lost City Thiomicrospira phylotype into multiple strains or ecotypes, each harboring unique genomic rearrangements caused by transposases.

Supporting Information
Table S1 Each of 2200 T. crunogena protein-coding genes were compared with Lost City metagenomic contigs and unassembled sequencing reads (singlets) with tblastn [20]. The Lost City contig and singlet with the best tblastn scores are shown for each T. crunogena protein.