Olivine and other igneous minerals and glasses were incubated in 3.5-mya oceanic crust for 4years to investigate the properties of surface-colonizing microbial communities from the JdFR aquifer. These ~2-mm-sized mineral grains were emplaced in flow cells at 275–287m below seafloor (mbsf) in International Ocean Drilling Program Hole 1301A (47° 45.210' N, 127° 45.833' W; 2,667m below sea level) on the eastern flank of the JdFR (Fisher et al., 2005; Smith et al., 2011, 2016). Details of incubation and sample retrieval have been reported previously and are only briefly summarized here (Smith et al., 2011, 2016). After the 4-year incubation in Hole 1301A, recovered minerals were frozen at −40°C.

Extraction and amplification methods were published previously (Smith et al., 2011, 2016). Briefly, genomic DNA was extracted from ~500mg olivine using a modified protocol for the FastDNA Spin Kit for Soil (MP Biomedicals Catalog #116560200) as described previously (Wang and Edwards, 2009; Smith et al., 2016). Genomic DNA extracted from the two most common olivine mineral phases (forsterite and Fo₉₀ olivine) was pooled to produce the olivine metagenome (a total of 50–70ng of DNA). The olivine genomic DNA extract was sent to the Census of Deep Life sequencing facility at Marine Biological Laboratory in Woods Hole, MA, United States. Metagenomic sequencing was carried out on an Illumina Hi-Seq1000 instrument with 2×101bp paired-end sequencing and dedicated-read indexing. Sequence reads were de-multiplexed using the Consensus Assessment of Sequence and Variance (CASAVA) 1.8.2 (Illumina). A total of 84 million reads with average length of 108 bases were produced for the olivine metagenome, totaling 9.8 gigabases of sequence data.

Sequence quality check, assembly, and binning protocols were reported previously (Smith et al., 2019) and are summarized briefly again here. Raw sequence files were quality-filtered using String Graph Assembler (SGA)‘s preprocessing pipeline (Simpson and Durbin, 2012; Smith et al., 2019). High-quality mate pair sequence reads were assembled using the Iterative deBruin Graph Assembler—Uneven Depth (IDBA-UD) program (Peng et al., 2012) with the following specifications: minimum contig length of 450 bases and iterative kmer values from 45 to 69, stepwise by 4. The assembly with kmer length of 65 was chosen for use in analytical subsequent steps based on optimal values of n50 and maximum contig size.

VizBin (Laczny et al., 2014), a Java-based genomic binning program, which uses nonlinear dimension reduction of genomic signatures (i.e., nucleotide frequency) to assign contigs to taxonomic bins, was used to separate contig groups from the full assembly into reduced genome bins based on sequence similarity (see Smith et al., 2019). A total of 12 bins were detected in the olivine metagenome assembly, each of which was exported for manual bin reconstruction. The Ca. A. pyornia metagenome-assembled genome (MAG JdFRolivine-5) is described here. All genome MAGs from the olivine community were uploaded to the Metagenomics Analysis Server (MG-RAST) for initial analyses and comparison of genome features. CheckM (Parks et al., 2015) was used to verify taxonomic purity (3.96% contamination) and completeness (91.58%) of MAGs using a total of 43 universal marker genes present in complete bacterial genomes. The genome is publicly available on NCBI’s website with accession #SESX00000000, biosample SAMN10863484.

A number of 16S rRNA genes were included within the JdFRolivine-5 MAG (subsequently defined as the Ca. A. pyornia MAG); therefore, we used several methods to determine phylogenetic relatedness of this genome’s sequences to other references. First, we employed whole genome phylogenetic tree construction in The Department of Energy Systems Biology Knowledgebase (KBase; Arkin et al., 2018) with Species Tree v.1.1.2. We used a set of 49 concatenated core universal marker genes to construct the genome tree. Input genomes are combined with closely related genomes from the RefSeq database from KBase, and a curated multiple sequence alignment of a subset of 49 COG domains is used to determine relatedness. The alignments are trimmed, concatenated, and then used to a construct phylogenetic tree with maximum likelihood phylogeny. We set a threshold of 20 nearby genomes to construct the tree and did not remove any genomes. We performed FastANI v.0.1.3 (Goris et al., 2007; Jain et al., 2018), or whole-genome Average Nucleotide Identity (ANI), in KBase to determine relatedness to these nearby genomes.

We also compiled phylogenetic results from PhyloPythia, CheckM, MG-RAST, Kyoto Encyclopedia of Genes and Genomes (KEGG), and the PyroTag sequencing of olivine community DNA (Smith et al., 2016, 2019) to identify the relatedness of known species. In order to include more environmental sequences in our phylogenetic analysis, we chose the ffh gene sequence, another highly conserved, single copy phylogenetic marker gene (Sunagawa et al., 2013) that can be used in lieu of the 16S rRNA gene, to construct a maximum likelihood tree based on the Tamura-Nei model in MEGA5 (Tamura et al., 2011; Supplementary Figure 1). This tree also includes related sequences obtained from the same sample: JdFRolivine-7, JdFRolivine-9, and the organism reported here, Ca. Acetocimmeria pyornia (identified as MAG JdFRolivine-5; Smith et al., 2019). The initial ffh trees were obtained through the Neighbor-Join (Saitou and Nei, 1987) and BioNJ (Gascuel, 1997) algorithms. These algorithms were applied to a matrix of pairwise distances that were estimated by maximum composite likelihood. Positions that represented gaps or missing data were eliminated from the analysis. A total of 264 positions were used to construct the phylogenetic tree.

Gene sequences of MAGs obtained using Prodigal v2.6.0 were uploaded to the KEGG and annotated using BlastKOALA (Kanehisa et al., 2016). The reconstructed metabolism for the Ca. A. pyornia MAG was used to assess the potential for carbon fixation (Table 1), determine likely energy sources, and explore mineral or surface-based metabolic properties that may yield clues as to how this organism survives in JdFR aquifer biofilms. The resulting metabolic pathway reconstruction was compared against KEGG genome annotations of closely related acetogens Ca. Desulforudis audaxviator (Chivian et al., 2008), Moorella thermoacetica (Pierce et al., 2008), and Desulfitobacterium hafniense (Nonaka et al., 2006; Tables 2–4; Supplementary Tables 1–3).

Genes not covered by pathway analysis were searched using the KEGG-annotated genome data obtained through BlastKOALA for all contigs. Genes necessary for completion of near-complete pathways or those that may be involved in alternative pathways were searched and identified. Information regarding other genes relating to biofilm formation, chemotaxis and flagella, pilus formation, metal transport or utilization, and others was gathered from the annotated genome.

The evolutionary relationships of Ca. A. pyornia indicate this organism represents a novel acetogen within the bacterial Class Clostridia (Figure 1; Supplementary Figure 1). Candidatus A. pyornia is placed within an “acetogen” cluster which includes M. thermoacetica (Pierce et al., 2008), Ca. D. audaxviator (Chivian et al., 2008), and D. hafniense (Nonaka et al., 2006; Supplementary Figure 1). FastANI analysis did not return results as there was <80% whole genome nucleotide similarity between Ca. A. pyornia and all nearby available genomes. An ANI percentage of ≥95% is considered the same species as the reference genome (Jain et al., 2018); thus, Ca. A. pyornia is at least a new species and most likely a separate genus of Clostridia.

The genome of Ca. A. pyornia was predicted to be 91.5% complete, with 3.96% contamination detected (Smith et al., 2019). Of 2,725 predicted coding sequences, 1,446 genes were annotated to a given KEGG orthologous functional group (KO number). Candidatus A. pyornia’s genome summary statistics (i.e., genome size and number of protein-encoding genes) were more similar to those of the closely related acetogens Ca. D. audaxviator and M. thermoacetica than to D. hafniense. Desulfopertinax hafniense’s genome is roughly twice the size with double the amount of protein-encoding genes (Nonaka et al., 2006). The G+C content, which can reflect environmental factors, was similar among all acetogens in this study (~50%) with the exception of Ca. D. audaxviator, whose G+C content is much higher (60.9%; Chivian et al., 2008). When Ca. A. pyornia’s genome attributes from BlastKOALA were compared to the seven other high-quality bacterial genomes (>90% complete and <5% contamination; Bowers et al., 2017) from the same community, Ca. A. pyornia had the highest percentage of genes for amino acid metabolism and respiration and lowest percentage of carbohydrate metabolism genes. There were 21 protease genes and 15 peptidase genes identified in the Ca. A. pyornia genome.

Candidatus A. pyornia contained the complete Wood–Ljungdahl pathway for acetogenesis (Figure 2). This included genes that code for enzymes and electron transfer agents needed to complete both branches of the pathway. This organism’s genome also contained the complete phosphate acetyltransferase–acetate kinase pathway to transform acetyl-CoA to acetate and recover the energy expended earlier in the pathway (Figure 2). Genes for the carbonyl branch of the Wood–Ljungdahl pathway are contained within an acs gene cluster, which is commonly found in acetogens (Figure 3; Table 1). The synteny of the Ca. A. pyornia acs gene cluster most closely resembled that of Ca. D. audaxviator, with only slight differences in the annotations of a 4Fe-4S ferredoxin gene, which is specifically denoted as heterodisulfide reductase A (hdrA) in Ca. A. pyornia. The acs gene clusters of M. thermoacetica and D. hafniense were less similar to the acs gene cluster of Ca. A. pyornia due to the addition or loss of genes within the cluster.

Candidatus A. pyornia encodes for an incomplete TCA cycle; the genes coding for the enzymes malate dehydrogenase and citrate synthase are missing from this MAG (Figure 4). Malate dehydrogenase transforms malate into oxaloacetate, and citrate synthase uses oxaloacetate and acetyl-CoA to produce citrate and begin the cycle again. However, if portions of the TCA cycle are run in reverse, key biosynthetic intermediates can still be produced from anaplerotic reactions. This genome bin does, however, contain the oxaloacetate-decarboxylating malate dehydrogenase, which is situated in a TCA cluster on the same contig immediately following the acs cluster from the Wood–Ljungdahl pathway (Table 1). This portion of the contig also encodes genes for fumarate hydratase, all four succinate dehydrogenase/fumarate reductase subunits, and succinyl-CoA synthetase; however, the oxaloacetate-decarboxylating malate dehydrogenase is likely making pyruvate as a part of anaplerosis, converting it to phosphoenolpyruvate (PEP) and on up the gluconeogenesis/glycolysis pathway to make carbohydrates in order to supplement what it cannot get from the environment. At least one other pathway that results in oxaloacetate formation was identified. The enzyme aspartate aminotransferase can produce oxaloacetate and L-glutamate from L-aspartate and α-ketoglutarate (Figure 4). Oxaloacetate can also be converted to pyruvate using oxaloacetate decarboxylase. The gene for citrate lyase beta subunit was also present; therefore it is possible that other genes not identified could be present in the genome, allowing Ca. A. pyornia to produce oxaloacetate from citrate in the reverse direction (Ragsdale and Pierce, 2008). Candidatus A. pyornia contains at least two complete pathways for the production of acetate from acetyl-CoA (Figure 4) and can convert pyruvate to PEP and vice versa.

Candidatus A. pyornia contained all but one gene to construct a complete anaerobic electron transport chain (ETC) for oxidative phosphorylation. We found the near-complete gene set for Complex I, a complete Complex II, and a complete prokaryotic F-type ATPase (Supplementary Table 1). Complex I, or NADH: quinone oxidoreductase (NADH dehydrogenase), is a multi-subunit complex consisting of 14 gene products, 13 of which are present in the Ca. A. pyornia genome [subunits A–F and H–N (no G), Supplementary Table 1]. We also found the complete pathway for succinate dehydrogenase, which consists of four genes that code for each of the four subunits of this enzyme, sdhABCD (Supplementary Table 1). Moorella thermoacetica, Ca. D. audaxviator, and D. hafniense are all missing the complete gene set to make succinate dehydrogenase (Supplementary Table 1), but all have at least one subunit in their genome. All four acetogens discussed in this manuscript have the F-type ATPase.

The Ca. A. pyornia genome does not contain genes that allow it to perform dissimilatory sulfate reduction (Table 2). Dissimilatory sulfate reductase genes (dsrA and dsrB) were not present in the genome, although sat and aprAB genes, whose enzyme products catalyze the reversible transformation of sulfate to sulfite, were detected (Table 2).

Genes encoding transporters for proteinaceous compounds (e.g., oligopeptide, peptide/nickel, and branched chain amino acid transporters), minerals, and metals were exclusively found within the Ca. A. pyornia genome (Supplementary Tables 2, 3; Figure 5). When coupled with a cytosolic aminotransferase, the branched chain amino acid transport system allows for the uptake and utilization of leucine, isoleucine, and valine. Candidatus A. pyornia also encodes one of three enzymes that make up the branched-chain α-keto acid dehydrogenase complex, involved in the degradation of branched-chain amino acids to produce ATP. There were also complete pathways for iron, nickel, molybdate, and tungstate transporters, some of which are cofactors required for Wood–Ljungdahl pathway and respiratory enzymes.

Candidatus A. pyornia appears unable to import lipids or carbohydrate molecules from the environment, since genes encoding complete transporters were missing from its genome. These include the saccharide, polyol, and lipid transport system and sugar-importing phosphotransferase (PTS) systems (Supplementary Table 2).

Genes involved in export of lipooligosaccharides, heme (used in membrane-bound succinate dehydrogenase), and other components related to the cell membrane or extracellular structures were detected (Supplementary Table 2). The Sec secretion system, involved in the secretion of proteases, toxins, and pilus proteins like adhesin, was present in all acetogens compared in this study. Pathways for export of some lipooligosaccharides, which are likely membrane components, and heme, which is an integral part of the membrane-bound succinate dehydrogenase complex, were also present in the genome.

Candidatus A. pyornia may be capable of glycolysis via an alternative route to the canonical pathway. The glycolysis pathway is missing the step performed by fructose bisphosphate aldolase (Table 3), which converts α-D fructose 1,6-bisphosphate to D-glyceraldehyde 3-phosphate; however, the reverse of this step is not reported missing in the gluconeogenesis pathway, as Ca. A. pyornia contains the fructose 1, 6-bisphosphate aldolase, phosphatase. The gluconeogenesis pathway was deemed incomplete since the conversion of oxaloacetate to phosphoenolpyruvate via PEP carboxykinase is missing, but this organism encodes the oxaloacetate decarboxylase and pyruvate orthophosphate–dikinase enzymes, which provide an alternative route to gluconeogenesis (Figure 4). These enzymes enable the conversion of oxaloacetate to pyruvate and then to PEP, from which gluconeogenesis can then proceed. There are complete pathways for the glycolysis 3-C compound core module and the non-oxidative pentose phosphate pathway. The pentose phosphate pathway allows this organism to produce 5-carbon compounds for biosynthesis, and the gluconeogenesis pathway allows it to produce 6-carbon compounds. Moorella thermoacetica, Ca. D. audaxviator, and D. hafniense all have complete pathways for glycolysis and the pentose phosphate pathway (Table 3).

Novel genomic evidence from this study suggests that Ca. A. pyornia is a thermophilic acetogen from the bacterial order Clostridiales. Previously we found evidence that Ca. A. pyornia is a dominant member from a mineral biofilm collected from a deep subsurface oceanic crust sample (estimated to be ~15% of the community; Smith et al., 2016). This organism has the genetic potential to act as the dominant primary producer of this simple microbial community (Smith et al., 2019) and appears to be capable of chemosynthesis and energy generation using molecular hydrogen as a reductant to fix inorganic carbon in seawater. It may be supplementing its energy by utilizing amino acids or peptides commonly found in the mineral-attached biofilms, but it does not appear capable of sulfate reduction or importing carbohydrates, as has been observed for other acetogenic Clostridiales.

The full annotated MAG of Ca. A. pyornia is most similar to terrestrial acetogens M. thermoacetica and Ca. D. audaxviator; however, the genome of Ca. A. pyornia is not complete and can complicate pathway reconstruction efforts and hinder our understanding of its role in the subsurface biosphere. Further, this MAG was obtained from a flow cell loaded with crystals incubated in a subseafloor borehole rather than directly obtained from the pristine basaltic aquifer. The potential for contamination from surrounding fluids or via drilling processes adds to the uncertainty in this study. Nonetheless, the evolutionary relationships among acetogens in this clade provide strong evidence that this MAG represents a thermophilic organism adapted to a deep marine subsurface environment and is capable of performing acetogenesis using molecular hydrogen and CO₂.

The phylogenetic relationships of Ca. A. pyornia to other acetogens suggest that this is a novel organism within a clade of related acetogens (Figure 1; Supplementary Figure 1). This group now contains both marine and terrestrial organisms; however, the marine acetogens from oceanic crust have only recently been described (Jungbluth et al., 2017; Smith et al., 2019). The terrestrial acetogen Ca. D. audaxviator was first described from deep continental crust (Chivian et al., 2008), D. hafniense from contaminated soil (Nonaka et al., 2006), and M. thermoacetica from horse manure (Fontaine et al., 1942). A new marine acetogen related to Ca. D. audaxviator, Ca. Desulfopertinax cowenii, was recently described from the JdFR flank aquifer fluid community (Jungbluth et al., 2017), and seven other putative acetogens were discovered from the JdFR olivine community, one of which is the Ca. A. pyornia MAG described here. Candidatus D. audaxviator and M. thermoacetica appear to be the closest genomic relatives to Ca. A. pyornia that are known acetogens; however, their genome similarity to Ca. A. pyornia does not exceed 80%. We are proposing Ca. A. pyornia represents at the very least a new genus within the Clostridiales.

The presence of this acetogen could impact carbon cycling and the health and function of the suboceanic aquifer ecosystem and the deep ocean by fixing carbon that makes its way back to the seafloor (Mason et al., 2009; McCarthy et al., 2010). Candidatus A. pyornia may be a prominent member of the attached community in Hole 1301A (and perhaps 1026B) as evidenced by its ubiquity in colonized rock and mineral biofilms (Orcutt et al., 2011; Smith et al., 2016) and because it encodes the full Wood–Ljungdahl carbon fixation pathway (Smith et al., 2019; Figure 2). Additionally, olivine-rich crust in this same region of the seafloor was previously revealed as a habitat for active methane- and sulfur-cycling microbes (Lever et al., 2013). This organism’s genome also encodes reductases and a ferredoxin that may be used to generate ATP via oxidative phosphorylation (Pierce et al., 2008). Since this organism has only been described from mineral biofilms and does not appear to be a dominant member of the aquifer fluid community, the organism this genome represents has indications that it benefits from a mineral biofilm lifestyle. A mineral biofilm can provide a wider range of substrates for metabolism either obtained from the mineral as it reacts with water or from other members of the community. Since H₂ can be produced by mineral surfaces as they react with water in the thermal aquifer (Mayhew et al., 2013), this organism may be using H₂ to power the Wood–Ljungdahl pathway. Charged mineral surfaces can also attract other charged molecules like certain amino acids from aquifer fluid (Lin et al., 2015), effectively concentrating these nutrients near the mineral surface. There is also some indication that this organism may participate in a syntrophic relationship in the biofilm community, especially since Ca. A. pyornia is missing some complete biosynthetic pathways, such as those involved in amino acid biosynthesis, and is able to import small peptides (Table 4; Supplementary Table 2).

The co-localization of acs genes within a gene clusters is a common attribute of acetogens and may allow this pathway to be horizontally transferred (Pierce et al., 2008). Previously, it was found that acs gene clusters come in several varieties (Pierce et al., 2008), and the gene arrangement of Ca. A. pyornia, along with the gene cluster that most closely resembles that of Ca. A. pyornia (Ca. D. audaxviator), may form a new class of acs gene cluster that is unlike those that were described before. The gene clusters of Ca. A. pyornia and Ca. D. audaxviator are near-identical, with the only difference being the annotation for the 4Fe-4S ferredoxin (green arrow, Figure 3). This gene was annotated as hdrA, also a 4Fe-4S ferredoxin, but these genes may represent isoforms of the same heterodisulfide reductase.

The acs gene clusters of M. thermoacetica and D. hafniense are similar to the Ca. A. pyornia and Ca. D. audaxviator gene clusters in that they share the same acs genes in the same arrangement (Figure 3); however, there are some marked differences with each genome. Moorella thermoacetica’s genome contains another heterodisulfide reductase gene (hdrC) and a hypothetical protein (hp) after the acsE gene, and D. hafniense has a simplified gene cluster with no other associated genes after acsE. Both M. thermoacetica and D. hafniense also contain an acsF gene at the beginning of the acs cluster (Figure 3). Moorella thermoacetica also contains an unannotated isoform of the acsF gene within its cluster, the cobyrinic acid a,c-diamide synthase gene. Due to the similarity of Ca. A. pyornia’s acs gene cluster with that of Ca. D. audaxviator’s, it is likely that these organisms have a shared history relating to the Wood–Ljungdahl pathway, and these gene clusters may form a new cluster arrangement apart from the one M. thermoacetica or D. hafniense occupies.

Candidatus A. pyornia contains an incomplete TCA cycle and is missing citrate synthase and malate dehydrogenase. Although incomplete TCA cycles are common in Clostridia, such as the closely related acetogens discussed here (Nonaka et al., 2006; Chivian et al., 2008; Pierce et al., 2008), the missing enzymes vary. All of the bacteria from the olivine community to which Ca. A. pyornia belongs (eight microbes – two “Bacteria,” six Clostridia) were also missing citrate synthase from their genomes, and the acetogens Ca. D. audaxviator and D. hafniense are also missing this enzyme. It is not known whether Ca. A. pyornia can make citrate using alternative pathways. Candidatus A. pyornia is also missing malate dehydrogenase, as is M. thermoacetica (Pierce et al., 2008); however, Ca. A. pyornia should still be able to make oxaloacetate through the aspartate aminotransferase reaction (Figure 4). It is possible this is a sign of genome reduction in the deep subsurface, but since it appears widespread in this biofilm community, this reduction would have had to have taken place long ago in an ancestral lineage. Alternatively, the ancestral organisms in this environment never evolved to have them and this absence of TCA cycle steps was the ancestral form.

The other closely related acetogens to Ca. A. pyornia are also missing some TCA cycle genes; however, these varied among genomes. As noted above, among those that shared the same missing enzymes as Ca. A. pyornia, Ca. D. audaxviator and D. hafniense are both missing citrate synthase (Nonaka et al., 2006; Chivian et al., 2008), and M. thermoacetica is missing malate dehydrogenase (Pierce et al., 2008). Candidatus D. audaxviator is missing the additional enzymes aconitase and α-ketoglutarate dehydrogenase, M. thermoacetica is also missing succinyl-CoA synthetase and fumarase, and D. hafniense is missing isocitrate dehydrogenase and succinate dehydrogenase. Candidatus D. audaxviator and D. hafniense do contain fumarate reductase that can act similarly to succinate dehydrogenase; thus, these organisms can still complete the reversible enzymatic step that forms fumarate from succinate, depending on the conditions present (Nonaka et al., 2006; Chivian et al., 2008). However, Ca. A. pyornia is the only one of this group that contains all four succinate dehydrogenase genes that code for different subunits (Smith et al., 2019).

Candidatus A. pyornia may use oxidative phosphorylation during anaerobic respiration with formate as a terminal electron acceptor. Candidatus A. pyornia contains the genes for succinate dehydrogenase (sdhABCD), and heme export (Complex II) has all but one gene coding for NADH dehydrogenase (Complex I) and has the complete gene set for production of the F-type ATPase (Supplementary Table 1). This genome also contains abundant hydrogenases, ferredoxins, and cytochromes that may be used to shuttle protons out of the cell to create a proton-motive force and drive oxidative phosphorylation (Vignais and Billoud, 2007; Smith et al., 2019). These electron transport proteins may be coupled to the Wood–Ljungdahl pathway to provide additional energy for the cell, as has been proposed in M. thermoacetica (Pierce et al., 2008; Schuchmann and Müller, 2014). All genomes from the olivine community from which Ca. A. pyornia originated lacked at least one gene to complete NADH dehydrogenase, but most of the 11 high-quality genomes lacked more than one gene. Candidatus A. pyornia and two other genomes lacked only one gene, and one of these was missing the gene that codes for the same subunit as Ca. A. pyornia, nuoG. This organism is not an acetogen and is unrelated as it belongs to Archaea. Although these nuo genes were typically found in clusters of two or more throughout the olivine community genomes and were often associated with Wood–Ljungdahl pathway genes, there did not appear to be a conserved operon with potential un-annotated or mis-annotated nuo genes. There is evidence that formate dehydrogenase and some NAD(P)⁺-dependent hydrogenases contain the nuoEFG module of NADH dehydrogenase (Friedrich and Scheide, 2000), so perhaps Ca. A. pyornia is able to couple these enzymes together for respiration or use an alternative module to complete the NADH dehydrogenase complex. The NADH dehydrogenase complex has also been shown to be coupled with hydrogenases such as ech, and Ca. A. pyornia has this and a myriad of other hydrogenases that may be candidates for such coupling (Smith et al., 2019). Thus, it remains unknown whether Ca. A. pyornia is able to synthesize a fully functional NADH dehydrogenase, but there may be a mechanism to allow it to complete anaerobic respiration. In contrast to the other acetogens discussed here, the deep subsurface Ca. D. audaxviator only has the nuoEFG complex, but M. thermoacetica and D. hafniense contain the complete gene set (nuoA–nuoN) to synthesize this complex.

Candidatus A. pyornia does not appear to use sulfate reduction, carbohydrates, or lipids for energy, although key genes may have been either unannotated or mis-annotated in our analysis. This may be indicated a loss of function due to competition with others for sulfate or a possible syntrophic relationship whereby less energy can be devoted to producing substrates that can be obtained from others nearby. Although the Ca. A. pyornia genome is high-quality and is nearly complete, we are missing 8.5% of its genome that may contain key metabolism genes. The closely related Ca. D. audaxviator is a sulfate reducer that uses nitrogen fixation and carbohydrate degradation to supplement its energy (Chivian et al., 2008). Moorella thermoacetica also uses carbohydrates (Pierce et al., 2008), and among other metabolic strategies, D. hafniense is known to dechlorinate compounds for energy (Nonaka et al., 2006). Since Ca. A. pyornia is unable to use these substrates and energy sources for supplemental metabolism, it may be using other substrates that are not typical of the other acetogens.

Although all acetogens considered in this study can import branched-chain amino acids, oligopeptides, and contain a peptide/nickel transporter, Ca. A. pyornia may rely on these compounds as a source of energy. Candidatus A. pyornia had more genes related to respiration and amino acid metabolism in its genome than any other organism in its community (4.86 and 9.17%, respectively), as well as a larger number of peptidases and proteases. This suggests that Ca. A. pyornia is either more dependent on amino acids and peptides as an additional source of energy than other members of its community or that it imports these compounds to spend less energy in producing them.

Dissolved amino acids are available in the JdFR aquifer, although their concentrations are less than typical submarine hydrothermal fluid and are instead roughly equivalent with concentrations in deep seawater (1–13nM for dissolved free amino acids and 43–89nM for dissolved hydrolyzable amino acids; Lin et al., 2015). The lower amino acid concentrations in the JdFR basement fluids relative to seawater could indicate interaction with adsorptive mineral surfaces as has been suggested (Lin et al., 2015), or it could be an indication of consumption by the microbial community. Candidatus A. pyornia could be one such organism contributing to this drawdown, and since it was detected here in mineral biofilms, it may be utilizing those amino acids adsorbing to the mineral surfaces as an energy or nitrogen source. For Ca. A. pyornia, it would be beneficial to use amino acid fermentation in this environment, especially since this organism does not appear to use or import other exogenous substrates, and using an uncommon metabolic strategy (Fonknechten et al., 2010) may mean less competition.

Although the pathways to import and begin the degradation of branched-chain amino acids are present in the Ca. A. pyornia genome, the degradation pathway is incomplete. It is possible that the fate of these amino acids lies in Stickland fermentation; however, we were unable to identify enzymes that could be used to ferment amino acids using this strategy. Stickland fermentation couples the oxidation and deamination of one amino acid to the reduction and deamination of another (Nisman, 1954; Fonknechten et al., 2010), producing ATP during a subsequent enzymatic reaction involving acetylphosphate. Candidatus A. pyornia contains the enzymes alanine dehydrogenase and glycine dehydrogenase (Supplementary Table 5) to complete the deamination reaction; however, we could not identify an amino acid reductase that these could be coupled with. Glycine dehydrogenase is coupled with glycine hydroxymethyltransferase in the interconversion of glycine and serine during the production of a Wood–Ljungdahl pathway intermediate 5,10-methylene THF (Fonknechten et al., 2010). Alanine dehydrogenase will deaminate alanine and convert it to pyruvate, while producing NADH and H⁺. The NADH and H⁺ are usually shuttled to an amino acid reductase to deaminate the other amino acid to acetylphosphate and then acetate. ATP is produced from each amino acid in the Stickland reaction when acetylphosphate is converted to acetate; therefore, it could be used as an alternative energy source for acetogens in environments like the suboceanic aquifer where carbohydrates and lipids are scarce.

Candidatus A. pyornia may be able to build 5- and 6-carbon sugars for biosynthesis of saccharides, nucleotides, and energy and enzymatic cofactors via a modified gluconeogenesis pathway originating from pyruvate. The enzyme pyruvate, orthophosphate kinase allows for the conversion of pyruvate to phosphoenolpyruvate that is one of the steps of the Crassulacean acid metabolism (CAM) carbon fixation pathway (Table 2; Supplementary Table 5; Figure 4). Candidatus A. pyornia is unable to complete glycolysis, which is unusual among the closely related acetogens (Table 3); however, the lack of sugar importers along with the missing genes for the glycolysis pathway suggests that Ca. A. pyornia is not utilizing exogenous carbohydrates and must therefore synthesize these biomolecules from pyruvate using the modified gluconeogenesis pathway.

Candidatus A. pyornia is able to synthesize nucleotides and some amino acids by using the non-oxidative pentose phosphate pathway (Table 3); however, the pathway for lipid synthesis is incomplete. Candidatus A. pyornia is missing the gene necessary to make acetyl-CoA carboxylase (ACC), which forms malonyl-CoA from acetyl-CoA; however, this initiation of fatty acid synthesis is also incomplete in the other closely related acetogens (Table 3). All of these acetogens are able to elongate fatty acids, so they may have an unknown pathway for malonyl-CoA synthesis.