Genomic comparisons of a bacterial lineage that inhabits both marine and terrestrial deep subsurface systems

Borehole fluid sampling

The methods used to collect samples during R/V Atlantis cruise ATL18_07 (28 June 2011–14 July 2011) are described elsewhere (Jungbluth et al., 2016). Briefly, basement crustal fluids were collected from CORK observatories located in 3.5 million-year-old ocean crust east of the Juan de Fuca spreading center in the Northeast Pacific Ocean. Basement fluids were collected from the polytetrafluoroethylene (PTFE) lined fluid delivery lines associated with the lateral CORKs (L-CORKs) at boreholes U1362A (47°45.6628′N, 127°45.6720′W) and U1362B (47°45.4997′N, 127°45.7312′W). These lines extend to 200 m and 30 m below the sediment-basement interface, respectively. Fluids were filtered in situ via a mobile pumping system (Cowen et al., 2012) through Steripak-GP20 filter cartridges (Millipore, Billerica, MA, USA) containing 0.22 µm pore-sized polyethersulfone membranes. A filtration rate of 1 L min⁻¹ was calculated from laboratory tests, indicating that ∼124 L (U1362A) and ∼70 L (U1362B) of deep subsurface crustal fluids were filtered. Based on average cell abundances in whole water samples collected on the same dive/sampling sequence (Jungbluth et al., 2016), ∼2.6 × 10⁹ and ∼0.18 × 10⁹ cells were collected from U1362A and U1362B, respectively.

DNA extraction and metagenome sequencing

Nucleic acids were extracted from borehole fluids using a modified phenol/chloroform lysis and purification method, and is described in detail elsewhere (Jungbluth et al., 2016; samples SSF21-22, SSF23-24). Library preparation, DNA sequencing, read quality-control, metagenome assembly, and gene prediction and annotation were conducted by the Department of Energy Joint Genome Institute as part of their Community Science Program using previously described informatics workflows (Huntemann et al., 2016), which are described in detail elsewhere (Jungbluth, Amend & Rappé, 2017).

Genomic bin identification and reconstruction

All metagenomic scaffolds greater than 200 basepairs (bp) from U1362A (n = 137,672 contigs) and U1362B (n = 212,542 contigs) were binned separately with MaxBin v1.4 (Wu et al., 2014) using the 40 marker gene set universal among bacteria and archaea (Wu, Jospin & Eisen, 2013), minimum contig length of 1,000 bp, and default parameters. Contig coverage from each metagenome was estimated using the quality control-filtered raw reads as input for mapping using Bowtie2 v2.1.0 (Langmead & Salzberg, 2012) via MaxBin. The genomic bins were screened and analyzed for completeness, contamination, and assigned taxonomic identifications using CheckM v1.0.5 (Parks et al., 2015) with default parameters.

Raw quality control-filtered sequence reads from the U1362A and U1362B metagenomes related to “Ca. D. audaxviator” were identified by mapping to three sources: (1) a single genomic bin from U1362A related to “Ca. D. audaxviator” identified via CheckM (bin A32), (2) the “Ca. D. audaxviator” genome, (3) and all “Ca. D. audaxviator”-related contigs >200 bp from the U1362A and U1362B metagenome assemblies generated by the Joint Genome Institute. Mapping was performed independently for the U1362A and U1362B metagenomes using both the bbmap v34.25 (http://sourceforge.net/projects/bbmap/) and Bowtie2 v2.1.0 (Langmead & Salzberg, 2012) software packages with default parameters and the paired-end read-mapping feature (Table S1). All reads from the U1362A metagenome mapping to any of the three sources (1,785,284 sequences) were assembled using SPAdes v3.5.0 (Bankevich et al., 2012) with options –k: 21,33,55,77, --careful –pe1-12 and default parameters. Contaminating contigs in the assembly were screened and removed using the JGI ProDeGe web portal v2.0 (https://prodege.jgi-psf.org/) on April 10, 2015, using default parameters with the following taxonomy specified: “Bacteria; Firmicutes; Clostridia” (Tennessen et al., 2016). Contigs remaining following the use of ProDeGe comprise the genome bin henceforth named “Ca. Desulfopertinax cowenii” modA32 and were screened using CheckM as described above.

Genome annotation and analysis

The modified genome bin resulting from the pipeline described above (“Ca. D. cowenii” modA32) was annotated via the Joint Genome Institute’s Integrated Microbial Genomes-Expert Review (IMG-ER) web portal (Markowitz et al., 2014; Huntemann et al., 2015). Annotations in the IMG-ER web portal served as the source of reported genome characteristics and reported genes and their assignment to COGs. Phylogenetically informative marker genes from “Ca. D. cowenii” were identified and extracted using the ‘tree’ command in CheckM. In CheckM, open reading frames were called using prodigal v2.6.1 (Hyatt et al., 2012) and a set of 43 lineage-specific marker genes, similar to the universal set used by PhyloSift (Darling et al., 2014), were identified and aligned using HMMER v3.1b1 (Eddy, 2011). Initial phylogenetic analysis used pplacer (v1.1.alpha16-1-gf748c91) (Matsen, Kodner & Armbrust, 2010) to place sequences into a CheckM tree/database (version 0.9.7) composed of 2,052 finished and 3,604 draft genomes (Markowitz et al., 2012).

An alignment 6,988 amino acids in length corresponding to the 43 concatenated marker genes from “Ca. D. cowenii,” “Ca. D. audaxviator,” other Firmicutes, and Actinobacteria were used for additional phylogenetic analysis. The concatenated amino acid alignment was used to generate a phylogeny using FastTree v2.1.9 (Price, Dehal & Arkin, 2010) with the WAG amino acid substitution model. The dendogram was visualized using iTOL v3 (Letunic & Bork, 2016).

Average nucleotide identity (ANI) was computed in IMG-ER using pairwise bidirectional best nSimScan hits of genes having 70% or more identity and at least 70% coverage of the shorter gene. The “Ca. D. cowenii” → [other genome] values are reported. Protein-coding genes in “Ca. D. cowenii” with homologs in “Ca. D. audaxviator,” and vice versa, were identified and percent similarity estimated using the “Phylogenetic Profiler” tool in IMG-ER with default parameters (max e-value: 10e⁻⁵; minimum identity: 30%). Average amino acid identity (AAI) was computed for pairs of genomes closely related to “Ca. D. cowenii” with an online web tool (http://enve-omics.ce.gatech.edu/aai/) using default parameters. All non-RNA genes at least 100 amino acids in length were used in this analysis. Two-way average amino acid identity scores are reported and the percent shared genes were calculated as follows: 100 × (2 × (number of proteins used for two-way AAI analysis))/((total number of amino acids ≥ 100 from genome A) + (total number of amino acids ≥ 100 from genome B)). Estimates of transposase and integrase abundance were derived in IMG using a functional profile of 100 pfams and COG functions selected searching for keywords “transposase” and “integrase.”

Genome and scaffold visualizations

Global genome comparisons were visualized in Circos v0.67-5 (Krzywinski et al., 2009). Links between genomic regions of “Ca. D. cowenii” and “Ca. D. audaxviator” represent best reciprocal BLAST hits, which were generated using the blast_rbh.py script (https://github.com/peterjc/galaxy_blast/tree/master/tools/blast_rbh) with blastn v2.2.29 (Altschul et al., 1990) and default parameters. Links between genomic regions from the single amplified genomes of Labonté et al. (2015) represent BLAST hits that were generated using blastn with default parameters and using “Ca. D. cowenii” and “Ca. D. audaxviator” as reference databases.

Selected scaffold regions were visualized with Easyfig v2.2.2 (Sullivan, Petty & Beatson, 2011). Similarity between regions was assessed using BLAST wrapped within Easyfig using default parameters and task: blastn; minimum hit length: 50; max e-value: 0.001; minimum identity value: 50. In all instances of blast, contigs from “Ca. D. cowenii” were used as the query and “Ca. D. audaxviator” was used as the reference, with the exception of the single three-scaffold comparison where “Ca. D. audaxviator” was used as the query and “Ca. D. cowenii” Ga007115_16 used as the reference.

Metagenome fragment recruitment

Quality-filtered raw reads from the U1362A metagenome were mapped to the six scaffolds that make up the “Ca. D. cowenii” genome bin and the “Ca. D. audaxviator” genome. Recruitment was performed using FR-HIT v0.7.1 (Niu et al., 2011) with default parameters (minimum sequence similarity 75%) and reporting a single best top hit for each read (-r 1).

Analysis of metagenome-derived SSU rRNA genes

Full length SSU rRNA genes from the raw quality-filtered U1362A and U1362B metagenome reads were assembled using EMIRGE (Miller et al., 2011) with default parameters and -a 20, -i 270, -s 100, -l 150, -j 1.0, –phred33, and using the SILVA SSURef_Nr99 version 119 database that was prepared using the fix_nonstandard_chars.py script supplied on the EMIRGE website (https://github.com/csmiller/EMIRGE). Out of 1951 (U1362A) and 1434 (U1362B) near full-length SSU rRNA sequences constructed after 66 (U1362A) and 100 (U1362B) iterations of EMIRGE, a single sequence from U1362A related to the “Ca. D. audaxviator” lineage was identified through the SILVA online portal (Pruesse, Peplies & Glöckner, 2012). The sequence was aligned using the SINA online aligner and manually curated in ARB (Ludwig et al., 2004). Ambiguous and mis-aligned positions were excluded from further analysis.

A base SSU rRNA gene phylogenetic tree was reconstructed in ARB from 36 sequences and an alignment of 797 nucleotide positions using RAxML v7.72 (Stamatakis, 2006) with default parameters, the GTR+G+I nucleotide substitution model identified via JModelTest v2.1.1 (Darriba et al., 2012), and selecting the best tree from 100 iterations. Bootstrapping was performed in ARB using the RAxML tool with 2,000 replicates (Stamatakis, Hoover & Rougemont, 2008). Sequences of short length, including a masked version of the “Ca. D. audaxviator”-related SSU rRNA gene found here, were added to the phylogeny using the parsimony insertion tool in ARB and a filter containing 363 nucleotide positions.

Phylogenetic analysis of dsrAB gene sequences

DNA sequences corresponding to dissimilatory sulfite reductase subunits alpha and beta (dsrAB) were aligned in ARB using the ‘integrated aligners’ tool and a previously published database of aligned dsrAB sequences (Loy et al., 2009). Additional sequences were identified and included via BLAST search of the non-redundant NCBI database using megablast and blastn with default parameters. Phylogenetic analyses were performed individually for dsrA and dsrB using RAxML with the GTR model of nucleotide substitution under the gamma- and invariable-models of rate heterogeneity, identified via jModelTest. The tree with the highest negative log-likelihood score was selected from performing 100 iterations using RAxML with default parameters. Phylogenies for the base trees were derived from partial length dsrA and dsrB alignments (545 and 303 nucleotides, respectively) and bootstrapping was performed in ARB using the RAxML rapid bootstrap analysis algorithm with 2,000 bootstraps.

Analysis of global distribution patterns

All protein-coding genes corresponding to the genomes of “Ca. D. cowenii” (1,782 genes) and “Ca. D. audaxviator” (2,239 genes) were used to generate a profile against 489 globally-distributed metagenomes from marine subsurface fluids, the terrestrial subsurface, terrestrial hot springs, marine sediments, and seawater (Table S2). In IMG-ER, the “Profile & Alignment” tool was used to query assembled metagenomes using genes corresponding to the two genomes, a maximum e-value of 10⁻⁵, and a minimum similarity of 70%. The number of gene hits was converted to a relative frequency and the location of hits was visualized in R v3.1.2 (R Core Team, 2015) using latitude and longitude information provided as metadata and the R maps package (version 2.3-10).

Fragment recruitment was subsequently used in effort to discriminate between the distribution of the marine (“Ca. D. cowenii’ modA32A) and terrestrial (“Ca. D. audaxviator”) genomes of this Firmicutes lineage. Raw reads corresponding to IMG-ER metagenomes with the highest hit frequencies in the profiles generated in IMG, and additional unamplified metagenomes from the marine and terrestrial subsurface available only via NCBI sequence read archive and MG-RAST, were used as references for mapping to the genomes of “Ca. D. cowenii” and “Ca. D. audaxviator” (Table S3). In order to determine a % similarity cutoff that can discriminate between the two targets, the two genomes were cut into non-overlapping 150 bp fragments to simulate the most common sequence read length in current metagenome projects, and mapped back to the intact “Ca. D. cowenii” and “Ca. D. audaxviator” genomes using FR-HIT with default parameters, restricting matches to the single top best hit. Percent similarities ranging from 70–100% were tested in one percent increments in order to quantify the frequency that the fragmented genomes map to their source genome. A 96% similarity level was ultimately used because it restricted spurious matches (i.e., reads mapping from one genome to the other) to a frequency of ∼1% (Fig. S1). The ratio of reads mapping to “Ca. D. cowenii” or “Ca. D. audaxviator” was calculated and visualized using Circos.

Sample access and affiliated information

The annotated draft genome of “Ca. D. cowenii” modA32 is available via the IMG web portal under Taxon ID number 2615840622 (Gold Analysis Project ID: Ga0071115) and NCBI whole genome shotgun (WGS) project MPOA00000000. The U1362A and U1362B metagenomes are available via the IMG-M web portal under Taxon ID numbers 330002481 and 3300002532, respectively. Gold Analysis Project ID numbers are Ga0004278 (U1362A) and Ga0004277 (U1362B). Sample metadata can be accessed using the BioProject identifier PRJNA269163. The NCBI BioSamples used here are SAMN03166137 (U1362A) and SAMN03166138 (U1362B). Raw sequence data can be accessed using NCBI SRA identifiers SRR3723048 (U1362A) and SRR3732688 (U1362B). A FASTA file containing all EMIRGE-reconstructed SSU rRNA genes from the two borehole fluid metagenomes can be accessed at https://doi.org/10.6084/m9.figshare.4539149.v1.

Bin identification and refinement

Of 60 and 41 genome bins representing diverse groups of uncultivated bacteria and archaea reconstructed from the U1362A and U1362B metagenomes, respectively, one that comprised a nearly complete genome from U1362A (bin A32) was identified as related to “Ca. D. audaxviator” by phylogenetic analyses of a set of concatenated single copy marker genes. In order to maximize genome recovery while minimizing potential contamination, contigs within genome bin A32, the “Ca. D. audaxviator” genome, and scaffolds related to “Ca. D. audaxviator” that were assembled directly from the U1362A and U1362B metagenomes were used as references for mapping raw sequence reads from the U1362A and U1362B metagenomes via several read mapping methods. Because of the relatively high abundance of reads in the U1362A library compared to U1362B, sequence mate pairs from the U1362A metagenome that mapped to these templates were pooled and reassembled (Table S1). Following subsequent screening and removal of contaminating sequences (Table S4), six genomic scaffolds from U1362A totaling 1,778,734 base pairs (bp) in length and originating from only the U1362A metagenome were identified that correspond to the draft “Ca. D. cowenii” modA32 genome described here (Table 1). Average read coverage of “Ca. D. cowenii” modA32 was 55.0. The purity of the modified genomic bin was supported by results generated using CheckM (Parks et al., 2015) (Table 2), congruent phylogenetic analyses of concatenated marker genes (Fig. 1A) and dsrB (Fig. 2A) and dsrA genes (Fig. S2), and a high percent of shared genes and gene synteny between the six genomic scaffolds of “Ca. D. cowenii” and the “Ca. D. audaxviator” genome (Figs. 1B and 3A).

The 1.78 Mbp “Ca. D. cowenii” modA32 genome is 98–99% complete based on separate analyses of tRNA and other marker gene content specific to the phylum Firmicutes (Table 1). A phylogenomic analysis of 43 conserved marker genes confirmed a monophyletic relationship between “Ca. D. cowenii” and “Ca. D. audaxviator” within the Firmicutes (Fig. 1A), a relationship that was also supported by analyses of both dsrA (Fig. S2) and dsrB genes (Fig. 2A). While no small-subunit (SSU) rRNA genes were identified in the “Ca. D. cowenii” genome bin, a single full-length SSU rRNA gene related to “Ca. D. audaxviator” was reconstructed from raw U1362A metagenome reads. Phylogenetic analyses revealed this gene to form a tight cluster with SSU rRNA genes recovered previously from the deep subseafloor along the Juan de Fuca Ridge flank and, more broadly, a monophyletic lineage with “Ca. D. audaxviator” within the phylum Firmicutes (Fig. 2B). Consistent with previous studies (Jungbluth et al., 2014; Jungbluth et al., 2016), oceanic crustal fluid SSU rRNA gene clones formed at least two independent sub-lineages within this clade (Fig. 2B). Overall, the topology of the 16S rRNA, dsrA, and dsrB gene phylogenies reveal multiple distinct lineages related to the Ammonifex, Ca. D. audaxviator, Ca. D. cowenii, and several additional uncharacterized lineages containing members from the marine and terrestrial deep subsurface. Additional genomic information from these Firmicute lineages will help reveal the functional and evolutionary characteristics shared among deep subsurface microbial life.

Comparative genomics

The genomes of “Ca. D. cowenii” and “Ca. D. audaxviator” share an average nucleotide identity of 76.9%. This similarity value is 7% higher than “Ca. D. cowenii” shares with the next most closely related Firmicute genomes and demonstrates that “Ca. D. cowenii” and “Ca. D. audaxviator” originate from, at least, different species (Konstantinidis & Tiedje, 2005a). Similarly, the genomes of “Ca. D. cowenii” and “Ca. D. audaxviator” share an average amino acid identity of 74.2%, nearly 18% higher than “Ca. D. cowenii” shares with its next most similar genome, the Firmicute Desulfotomaculum kuznetsovii DSM 6115 (Fig. 1B). In addition to reinforcing the species-level evolutionary divergence observed with ANI, an AAI value of 74.2% indicates that the genomes of “Ca. D. cowenii” and “Ca. D. audaxviator” lie at the boundary demarcating genus-level divergence (Konstantinidis & Tiedje, 2005b; Konstantinidis & Tiedje, 2007). A similar result was obtained by quantifying the proportion of genes shared between “Ca. D. cowenii” and “Ca. D. audaxviator” (73.2%) (Fig. 1B).

Compared to the genomes of its closest relatives, the 1.78 Mbp genome harbored by “Ca. D. cowenii” is small (Fig. 1B). Despite the smaller size of the “Ca. D. cowenii” genome compared to the 2.35 Mbp genome of “Ca. D. audaxviator,” the two share similar coding density (89.8% vs. 87.6%), resulting in 451 fewer genes in “Ca. D. cowenii” (1,842 vs. 2,293) (Table 1). Compared to other firmicutes, the predicted genome size of “Ca. D. cowenii” is among the smallest for members of the Class Clostridia with an elevated %GC (Fig. S3); this relatively small genome size might be expected given the low flux of energy and nutrients in the deep subseafloor environment . The smaller genome of “Ca. D. cowenii” shares 1,514 of its 1,782 (85.0%) protein coding genes with “Ca. D. audaxviator.” Despite the lower gene content overall, “Ca. D. cowenii” harbors a similar number of protein coding genes with a predicted function as the genome of “Ca. D. audaxviator” (1518 vs. 1587) (Table 1). In addition to a smaller genome and fewer genes, “Ca. D. cowenii” also contained fewer pseudogenes (0 vs. 82) and paralogs (137 vs. 265) in comparison to “Ca. D. audaxviator” (Table 1), which together suggest some form of streamlining of the “Ca. D. cowenii” genome. Compared to “Ca. D. audaxviator,” the genome of “Ca. D. cowenii” contains fewer CRISPR elements, integrases and transposases, and phage-related genes, which suggests lower viral infection and less horizontal gene transfer in the marine lineage.

Extensive gene synteny between “Ca. D. cowenii” and “Ca. D. audaxviator” was revealed by comparing locations of homologs (Figs. 3A and 3B). Aligning the genome of “Ca. D. cowenii” with five incomplete (3.6–7.8% complete) single amplified genomes (SAGs) isolated from the terrestrial South Africa subsurface and related to “Ca. D. audaxviator” (Labonté et al., 2015) revealed that all of the SAGs were more similar to “Ca. D. audaxviator” than “Ca. D. cowenii” (Fig. 4).

Similarities in functional gene complement

Comparisons of predicted proteins assigned to clusters of orthologous groups (COGs) revealed a markedly similar distribution within the “Ca. D. cowenii” and “Ca. D. audaxviator” genomes (Fig. 3C). A detailed description of these shared features is included in Table S5.

The genome of “Ca. D. cowenii” reveals a microorganism that is functionally similar to “Ca. D. audaxviator”: an independent lifestyle consisting of a motile, sporulating, thermophilic, anaerobic chemolithoautroptroph genetically capable of dissimilatory sulfate reduction, hydrogenotrophy, carbon fixation via the reductive acetyl-coenzyme A (Wood-Ljungdahl) pathway, and synthesis of all amino acids. The genome of “Ca. D. cowenii” also indicates a chemoorganotroph that possesses abundant sugar transporters and is capable of glycolysis, which is somewhat surprising given the low dissolved organic carbon concentrations in this system (Lin et al., 2012). Similar to “Ca. D audaxviator,” hydrogenases were abundant in “Ca. D. cowenii,” which is consistent with the availability of hydrogen in basement fluids of the Juan de Fuca Ridge flank (Lin et al., 2014). Altogether, the shared features between “Ca. D. cowenii” and “Ca. D. audaxviator” help to explain the wide distribution of this lineage in the global deep subsurface.

Differences in functional gene complement

Despite highly similar genomes overall, comparisons of predicted proteins assigned to clusters of orthologous groups (COGs) revealed unique genes in “Ca. D cowenii” that were not found in “Ca. D. audaxviator” (Fig. 3D; also see Tables S6 and S7). These genes are likely locations to uncover features that differentiate the marine versus terrestrial members of this lineage. While most unique genes in the “Ca. D. cowenii” genome have general functional characterizations only (COG category R), the largest fraction of unique genes in the “Ca. D. cowenii” versus “Ca. D. audaxviator” genome are found within COG category M (Cell wall/membrane/envelop biogenesis) and include nucleoside-diphosphate-sugar epimerases (e.g., galE) and glycosyltransferases (e.g., treT) involved in cell wall biosynthesis, and possibly in the production of exopolysaccharides involved with biofilm formation. Defense mechanisms (COG category V) contained the highest ratio of unique genes in the “Ca. D. cowenii” genome compared to “Ca D. audaxviator” and includes genes related to ABC-type multidrug transport systems, multidrug resistance efflux pumps (hylD), and a class-A beta-lactamase. The marine genome has numerous monosaccharide transporters not present in the terrestrial genome, including those encoding for components of ribose/xylose, arabinose, methyl-galactoside, xylose, allose, and rhamnose transport. Thus, potential differences in organic carbon substrate specificity are evident, which might be expected given the different ages and reactivity of organic material in the marine and terrestrial deep subsurface (e.g., Lang et al., 2006; Simkus et al., 2016).

Though the genome of “Ca. D. cowenii” is incomplete, within assembled contigs there are a small number of large indels that are also potential sources of functional differentiation between “Ca. D cowenii” and “Ca. D. audaxviator.” An indel present in “Ca. D. audaxviator” but lacking in “Ca. D. cowenii” includes a nitrogenase operon as well as genes for ammonium transport and nitrogen regulation (Fig. 5). While the genes for glutamine synthetase and glutamate synthase within the genome of “Ca. D. cowenii” suggest that it obtains its nitrogen from the abundant ammonia in Juan de Fuca Ridge flank crustal fluids (Lin et al., 2012), it appears to be unable to fix inorganic dinitrogen. Another indel suggests that “Ca. D. cowenii” lacks the capacity to produce cobalamin (Fig. 5). Moreover, a large cassette of genes present in the “Ca. D. audaxviator” genome that is related to gas vesicle production (and flanked by an integrase and two transposases) is missing in “Ca. D. cowenii.” Finally, CRISPR-CAS gene arrays and CRISPR elements were distinct between the two genomes (Fig. 5), with the genome of “Ca. D. cowenii” encoding 14 CRISPR-associated proteins versus 25 in “Ca. D. audaxviator.”

Distribution

The Desulfopertinax/Desulforudis lineage was detected in metagenomic data generated from the terrestrial subsurface of Mt. Terri, Switzerland and the Coast Range Ophiolite, California, USA (Fig. 6A; see also Table S2). It was also found within marine sediments from the coastal Atlantic and Pacific, a Yellowstone National Park hot spring, and the terrestrial subsurface in Ontario, Canada, but never identified in seawater worldwide. Mapping raw metagenome reads in a lineage-specific manner that discriminated between reads mapping to “Ca. D. audaxviator” and “Ca. D. cowenii” revealed partitioning of these genomes between terrestrial and marine environments, respectively (Fig. 6B; see also Table S3). Surprisingly, the ratio of mapped reads from “Ca. D. cowenii” to “Ca. D. audaxviator” was, highest (18.9) in a sample from the terrestrial subsurface. The next largest ratios were from the U1362A metagenome (7.3), three serpentinite groundwater metagenomes (1.7–1.6), and the U1362B metagenome (1.4). The ratio of “Ca. D. audaxviator” to “Ca. D. cowenii” reads was highest (up to ∼165) in samples collected from the terrestrial subsurface of Witwatersrand Basin, South Africa, although this lineage also appears present in serpentinite fluids from the terrestrial subsurface. Thus, it appears that the Desulfopertinax/Desulforudis lineage has a cosmopolitan distribution throughout the global subsurface environment, as indicated by mapping reads from 489 metagenomes from the terrestrial and marine subsurface to the genomes of “Ca. D. cowenii” and “Ca. D. audaxviator,” as well as gene clones identified in published SSU rRNA surveys (Fig. 6; see also Fig. 2B and Tables S2 and S3).