Into the darkness: the ecologies of novel 'microbial dark matter' phyla in an Antarctic lake.

Uncultivated microbial clades ('microbial dark matter') are inferred to play important but uncharacterized roles in nutrient cycling. Using Antarctic lake (Ace Lake, Vestfold Hills) metagenomes, 12 metagenome-assembled genomes (MAGs; 88%-100% complete) were generated for four 'dark matter' phyla: six MAGs from Candidatus Auribacterota (=Aureabacteria, SURF-CP-2), inferred to be hydrogen- and sulfide-producing fermentative heterotrophs, with individual MAGs encoding bacterial microcompartments (BMCs), gas vesicles, and type IV pili; one MAG (100% complete) from Candidatus Hinthialibacterota (=OLB16), inferred to be a facultative anaerobe capable of dissimilatory nitrate reduction to ammonia, specialized for mineralization of complex organic matter (e.g. sulfated polysaccharides), and encoding BMCs, flagella, and Tad pili; three MAGs from Candidatus Electryoneota (=AABM5-125-24), previously reported to include facultative anaerobes capable of dissimilatory sulfate reduction, and here inferred to perform sulfite oxidation, reverse tricarboxylic acid cycle for autotrophy, and possess numerous proteolytic enzymes; two MAGs from Candidatus Lernaellota (=FEN-1099), inferred to be capable of formate oxidation, amino acid fermentation, and possess numerous enzymes for protein and polysaccharide degradation. The presence of 16S rRNA gene sequences in public metagenome datasets (88%-100% identity) suggests these 'dark matter' phyla contribute to sulfur cycling, degradation of complex organic matter, ammonification and/or chemolithoautotrophic CO2 fixation in diverse global environments.

Introduction

Uncultivated microbial clades, collectively referred to as ‘microbial dark matter’, include lineages that are inferred to play key roles in ecosystem formation and nutrient cycling (Rinke et al., 2013; Parks et al., 2017, 2020; Nayfach et al., 2020; Zamkovaya et al., 2021), including in Antarctica (Cavicchioli, 2015; Panwar et al., 2020; Ortiz et al., 2021; Williams et al., 2021a). The Antarctic lake Ace Lake is home to an extensive diversity of bacterial taxa, many of which represent higher‐rank clades (class‐ and phylum‐level) that have no cultivated representatives, and for which only a relatively small number of metagenome‐assembled genomes (MAGs) are available (Ng et al., 2010; Lauro et al., 2011; Panwar et al., 2020; Williams et al., 2021a). Ace Lake is a marine‐derived, 25 m deep, meromictic (permanently stratified) system located in the Vestfold Hills of East Antarctica. The water column comprises an upper oxic mixolimnion, an oxic–anoxic interface (at ~13 m) that is defined by a strong halocline and oxycline, and a stable anoxic monimolimnion (Franzmann et al., 1991; Rankin et al., 1999; Lauro et al., 2011; Panwar et al., 2020). Historically, the limnology and ecology of Ace Lake have been well studied, including being the first Antarctic system to be examined using a combination of both metagenomics and metaproteomics (Ng et al., 2010; Cavicchioli, 2015). Subsequent studies characterized the lake microbial community composition and function by depth (Lauro et al., 2011), as well as seasonal and annual dynamics (Panwar et al., 2020). Specific abundant members of the community, green‐sulfur bacteria (GSB; Candidatus Chlorobium antarcticum) and ‘dark matter’ taxa, have also been the targeted for metagenome‐based characterization (Ng et al., 2010; Panwar et al., 2021; Williams et al., 2021a).

In the current study we examined deeply divergent (phylum‐level) uncultivated clades that were represented by Ace Lake MAGs with an estimated completeness of 88%–100%. The ‘dark matter’ taxa belonged to Genome Taxonomy Database (GTDB) (Chaumeil et al., 2019; Parks et al., 2020) candidate phyla SURF‐CP‐2 (Aureabacteria; Momper et al., 2017), OLB16, AABM5‐125‐24, and FEN‐1099. The MAGs exhibited <5% relative abundance (the majority <1%) in any of the 120 available Ace Lake metagenomes (Panwar et al., 2020), and were primarily from oxic–anoxic interface and anoxic monimolimnion metagenomes. The availability of previous Ace Lake physicochemical data and metagenomic assessments of the microbial community enabled us to infer ecological roles for these uncultivated bacterial phyla in Ace Lake, and interrogate public metagenome datasets to consider their possible global ecological relevance.

Results and discussion

From 12 high‐ and medium‐quality MAGs, 10 new bacterial genera and species representing four candidate phyla were identified: six for Candidatus Auribacterota (SURF‐CP‐2; ‘Aureabacteria’), one for Candidatus Hinthialibacterota phylum nov. (OLB16), three for Candidatus Electryoneota (AABM5‐125‐24) phylum nov. and two for Candidatus Lernaellota phylum nov. (FEN‐1099) (Fig. 1). The MAGs were determined to represent new candidate taxa based on their average nucleotide identity (ANI) (≤77%) and amino acid identity (AAI) (≤66%) to reference strains from GTDB (Konstantinidis et al., 2017) and their 16S rRNA gene identity to sequences in integrated microbial genomes (IMG) isolates (<92%) and NCBI nt (<95.7%) databases (Table S2). For each MAG given below, the IMG identifier is provided, along with the MAG that we have designated as the type for the new genus and species. The metabolic capacities of each of the four phyla predicted from the MAGs of the new candidate species are described below with a view to inferring their possible ecological roles in Ace Lake.

Fig. 1

Phylogeny of four ‘dark matter’ phyla from Ace Lake.

A. GTDB bacterial phylogeny overview showing the placement of taxa described in this study along with major bacterial phyla. The tree was constructed from 120 single copy bacterial marker genes using the GTDB‐tk de novo workflow with default parameters, and is rooted with the candidate phylum Patescibacteria (top left).

B–E. Zoomed‐in regions of the tree showing Ace Lake MAGs from this study (given in bold, with MAG accession number preceding proposed Candidatus names), and neighbouring reference GTDB strains labelled with their GTDB_accession, assigned phylum and NCBI_WGS_project; (B) phylum Candidatus Auribacterota; (C) phylum Candidatus Hinthialibacterota; (D) phylum Candidatus Electryoneota; (E) phylum Candidatus Lernaellota. An additional MAG belonging to the Candidatus Electryoneota (D) was detected in Ace Lake in this study (3300035698_578), but due to its low completion (63%) a genus or species name was not proposed. Three sequences labelled AABM5‐125‐24 without a GTDB accession in (D) are single‐cell microbial genomes from Youssef et al. (2019a) and Youssef et al. (2019b). Non‐parametric bootstrap values are shown at nodes. T, type strain. * GTDB classes and orders in key.

Phylum Candidatus Auribacterota (=SURF‐CP‐2)

Candidate phylum SURF‐CP‐2 was originally based on a single MAG (SURF_26) recovered from deep terrestrial subsurface fluid of a former gold mine (Deep Mine Microbial Observatory, South Dakota, USA) (Momper et al., 2017). SURF‐CP‐2 was previously named Candidatus Aureabacteria, here emended to Candidatus Auribacterota for orthography (Trüper and Euzéby, 2009), without changing the meaning (‘gold bacteria’) or authorship (Momper et al., 2017); MAG SURF_26 is named Ca. Auribacter fodinae gen. et sp. nov. (Table S2). Candidate phylum SURF‐CP‐2 was previously described as having a ‘cryptic lifestyle’ on account of the absence of a suite of specific metabolic genes searched for and found to be absent on the SURF_26 MAG (Momper et al., 2017). Based on our interrogation of six Ace Lake MAGs (88%–98% complete; Table 1) plus the Ca. Auribacter fodinae MAG, we infer that Ca. Auribacterota are anaerobes that rely mainly on simple sugars and amino acids as heterotrophic substrates, and produce hydrogen or sulfide as fermentative byproducts.

Table 1

Summary of the physiological and metabolic traits inferred from six Ace Lake MAGs of phylum Candidatus Auribacterota (=Aureabacteria).

Candidatus Tritonobacter lacicola gen. et sp. nov.

Genus named after the Greek demi‐god Triton, a son of the sea god Poseidon who inhabited a salt lake also called Triton; species name means ‘lake‐dweller’

3300035698_867 (type): 92% complete, 2.2% contamination

Highest abundance: 3.9% (23 m, anoxic zone)

Glycolysis and gluconeogenesis (includes both ATP‐dependent and PPi‐dependent 6‐phosphofructokinase; both NAD‐ and ferredoxin‐dependent glyceraldehyde 3‐phosphate dehydrogenase) TCA cycle (oxidative) [includes citrate (Si)‐synthase] Pentose phosphate pathway (non‐oxidative only) F‐type ATP synthase Ion‐translocating ferredoxin:NAD+ oxidoreductase Rnf complex Membrane‐bound [NiFe] hydrogenase (Mbh) (Group 4 g) + Mrp‐type Na+/H+ antiporter module Ech‐type membrane‐bound [NiFe] hydrogenase (Group 4e) Carbon monoxide oxidation (coupled to Ech‐type hydrogenase) and reduction Partial Wood–Ljungdahl pathway Sulfhydrogenase ([NiFe] hydrogenase [Group 3b]/sulfur reductase), cytoplasmic Glycoside hydrolases: 3 (two with signal peptide) Trehalose and glycogen synthesis Sulfatases: 4 (three with signal peptide) Proteases/peptidases: seven (none with signal peptide) Fermentation of BCAA BMC (involved in DNA degradation/aldehyde oxidation) Degradation of diaminopropionate, glycolate alcohols Synthesis of all 20 amino acids required for protein synthesis ABC transporters: sugars, peptides, osmoprotectants, tungstate Other transporters: amino acids, nucleosides, nucleobases, phosphate, Fe(II), Zn, Mg Desulfoferrodoxin, hydroxylamine reductase

Candidatus Erginobacter occultus gen. et sp. nov.

Genus named after Erginus the Argonaut, a son of the sea god Poseidon, in reference to both maritime exploration and to the marine origin of Ace Lake; species name means ‘hidden’

3300035698_653 (type): 95% complete, 2.2% contamination

Highest abundance: 0.16% (24 m, anoxic zone)

Glycolysis and gluconeogenesis (includes both ATP‐dependent and PPi‐dependent 6‐phosphofructokinase) Incomplete TCA cycle [includes citrate (Re)‐synthase] Pentose phosphate pathway (oxidative, non‐oxidative) V‐type ATP synthase Ion‐translocating ferredoxin:NAD+ oxidoreductase Rnf complex Sulfhydrogenase ([NiFe] hydrogenase [Group 3b]/sulfur reductase), cytoplasmic [FeFe] hydrogenase, bifurcating (Group A3), cytoplasmic Partial Wood–Ljungdahl pathway, CO reduction to acetyl‐CoA. Glycoside hydrolases: 7 (two with signal peptide) Trehalose and glycogen synthesis Sulfatases: 8 (seven with signal peptide) Proteases/peptidases: 10 (nine with signal peptide) Fermentation of BCAA and aromatic amino acids Degradation of lactate, diaminopropionate BMC (involved in DNA degradation/aldehyde oxidation) Acetyl‐CoA to acetate, with ATP generation by substrate‐level phosphorylation Synthesis of almost all 20 amino acids required for protein synthesis ABC transporters: sugars, peptides, phosphate, cobalamin, tungstate Other transporters: amino acids, carboxylic acids, nucleosides, nucleobases, phosphate, Fe(II), Mg Type IV pili

Candidatus Euphemobacter frigidus gen. et sp. nov.

Genus named after Euphemus the Argonaut, a son of the sea god Poseidon; species name means ‘cold’

3300031227_11 (type): 95% complete, 4.3% contamination 3300035698_1748: 96.8% complete, 5.4% contamination

Highest abundance: 2.6% (23 m, anoxic zone)

Glycolysis and gluconeogenesis (includes both ATP‐dependent and PPi‐dependent 6‐phosphofructokinase; both NAD‐ and ferredoxin‐dependent glyceraldehyde 3‐phosphate dehydrogenase) Incomplete TCA cycle [includes citrate (Re)‐synthase] Pentose phosphate pathway (oxidative, non‐oxidative) V‐type ATP synthase Ion‐translocating ferredoxin:NAD+ oxidoreductase Rnf complex Ech‐type membrane‐bound [NiFe] hydrogenase (Group 4e) Sulfhydrogenase ([NiFe] hydrogenase [Group 3b]/sulfur reductase), cytoplasmic Partial Wood–Ljungdahl pathway Glycoside hydrolases: 6 (three with signal peptide) Trehalose and glycogen synthesis Sulfatase: 1 (has signal peptide) Proteases/peptidases: 910 (five with signal peptide) Fermentation of BCAA and aromatic amino acids BMC (involved in DNA degradation/aldehyde oxidation) Degradation of lactate, diaminopropionate Acetyl‐CoA to acetate, with ATP generation by substrate‐level phosphorylation Synthesis of almost all 20 amino acids required for protein synthesis ABC transporters: sugars, peptides, BCAA, cobalamin, molybdate, tungstate Other transporters: sugars, amino acids, nucleosides, nucleobases, phosphate, Fe(II), Zn, Mg, Co Desulfoferrodoxin Gas vesicles Type IV pili

Candidatus Ancaeobacter aquaticus gen. et sp. nov.

Genus named after Ancaeus the Argonaut, a son of the sea god Poseidon; species name means ‘found in the water’

3300035698_1732 (type): 98% complete, 2.2% contamination

Highest abundance: 1.6% (18 m, anoxic zone)

Glycolysis and gluconeogenesis (includes both ATP‐dependent and PPi‐dependent 6‐phosphofructokinase) Incomplete TCA cycle [includes citrate (Re)‐synthase] Pentose phosphate pathway (oxidative, non‐oxidative) NADH‐quinone oxidoreductase (complex I) V‐type ATP synthase Ech‐type membrane‐bound [NiFe] hydrogenase (Group 4e) Sulfhydrogenase ([NiFe] hydrogenase [Group 3b]/sulfur reductase), cytoplasmic Glycoside hydrolases: 9 (none with signal peptide) Trehalose and glycogen synthesis Proteases/peptidases: 7 (none with signal peptide) Degradation of glycerol Fermentation of BCAA Synthesis of almost all 20 amino acids required for protein synthesis ABC transporters: sugars, peptides, cobalamin, Ni, Mn Other transporters: sugars, ammonia, amino acids, nucleosides, nucleobases, phosphate, Fe(II), Zn, Mg, Co Desulfoferrodoxin, hydroxylamine reductase

Candidatus Theseobacter exili gen. et sp. nov.

Genus named after Theseus the Argonaut, a son of the sea god Poseidon, as for above; species name means ‘exile' or 'isolation’

3300035698_1968 (type): 88% complete, 3.2% contamination

Highest abundance: 0.15% (23 m, anoxic zone)

Glycolysis and gluconeogenesis (includes PPi‐dependent 6‐phosphofructokinase) Incomplete TCA cycle Pentose phosphate pathway (oxidative, non‐oxidative) F‐type and V‐type ATP synthases Ion‐translocating ferredoxin:NAD+ oxidoreductase Rnf complex Ech‐type, membrane‐bound [NiFe] hydrogenase (Group 4e) Sulfhydrogenase ([NiFe] hydrogenase [Group 3b]/sulfur reductase), cytoplasmic Glycoside hydrolases: 8 (one with signal peptide) Trehalose and glycogen synthesis Sulfatases: 4 (all with signal peptide) Proteases/peptidases: 11 (six with signal peptide) Degradation of dihydroxyacetone, alcohols Synthesis of almost all 20 amino acids required for protein synthesis Polyphosphate synthesis ABC transporters: sugars, peptides, BCAA, cobalamin, Zn/Mn, molybdate Other transporters: sugars, nucleobases, phosphate, sulfate, Fe(II), Zn, Mg Desulfoferrodoxin, hydroxylamine reductase

The complete etymologies of genera and species are given in Table S2. A summary of inferred traits for the deep terrestrial subsurface fluid Candidatus Auribacter fodinae gen. et sp. nov. is also provided in Table S2.

The Ace Lake Ca. Auribacterota MAGs represent five different genera and species: Ca. Tritonobacter lacicola gen. et sp. nov. [3300035698_867 (type)]; Ca. Erginobacter occultus gen. et sp. nov. [3300035698_653 (type)]; Ca. Euphemobacter frigidus gen. et sp. nov. [3300031227_11 (type) and 3300035698_1748]; Ca. Ancaeobacter aquaticus gen. et sp. nov. [3300035698_1732 (type)]; Ca. Theseobacter exili gen. et sp. nov. [3300035698_1968 (type)] (Table S2). Collectively, these MAGs were most abundant in the Ace Lake oxic–anoxic interface and anoxic metagenomes. Ca. Tritonobacter lacicola and Ca. Euphemobacter frigidus MAGs showed up to 3.9% and 2.6% relative abundance respectively, at 23 m depth, whereas the Ca. Ancaeobacter aquaticus MAG showed very low abundance in metagenomes derived from the bottom waters of the lake (Table S3). The Ca. Euphemobacter frigidus MAGs have genes for gas vesicle production; vesicles function by providing buoyancy, enabling cells to adjust and maintain position in the water column to occupy zones with favourable growth conditions (Walsby, 1994; Youssef et al., 2015). The Ca. Erginobacter occultus MAG encodes type IV pili for attachment to surfaces, possibly cell debris and other detritus (Craig et al., 2019). However, whereas the Ca. Auribacter fodinae MAG (SURF_26) encodes flagella, none of the Ace Lake Ca. Auribacterota MAGs do (Table S2).

Ace Lake Ca. Auribacterota MAGs encode enzymes for glycolysis and gluconeogenesis pathways, including a pyrophosphate‐dependent phosphofructokinase as well as ATP‐dependent 6‐phosphofructokinase, the former of which can reversibly function in both glycolysis and gluconeogenesis (Mertens, 1991; Kemp and Tripathi, 1993). Both enzymes are also encoded in Ace Lake MAGs from the other three candidate phyla (Ca. Hinthialibacterota, Ca. Electryoneota, Ca. Lernaellota), and may increase the energetic efficiency of glycolysis, especially during fermentative growth (Mertens, 1991). Among the Ace Lake Ca. Auribacterota, the MAG for Ca. Tritonobacter lacicola is exceptional in encoding a possible complete TCA cycle (Fig. 2); although there is no succinate dehydrogenase in the MAG, fumarate reductase is encoded, which is a reversible enzyme and may therefore catalyse this step (Lu and Imlay, 2017; Foo et al., 2020). MAGs of the other four Ace Lake species encode an incomplete TCA cycle in both the oxidative and reductive directions: genes for 2‐oxoglutarate:ferredoxin oxidoreductase, succinyl‐CoA synthetase, succinate dehydrogenase and fumarate reductase are all absent. Given the estimated completeness (88%–98%) of these MAGs, we regard the incomplete TCA cycle as a likely metabolic trait, and consistent with these species having a ‘horse‐shoe’‐type TCA pathway, as found in certain other anaerobic bacteria, where this pathway functions solely in biosynthesis (Wood et al., 2004; Herlemann et al., 2009; Marco‐Urrea et al., 2011; Williams et al., 2021a). Thus, the right branch terminates at 2‐oxoglutarate, the amino acceptor for ammonia assimilation and transamination reactions (see below), and the left branch allows the interconversion of oxaloacetate, malate, and fumarate (Herlemann et al., 2009).

Fig. 2

Metabolic capacities of Candidatus Tritonobacter lacicola (candidate phylum Auribacterota) as inferred from MAG sequences. Elemental sulfur (S0) is shown entering the cell as soluble polysulfide (S‐Sn‐S), but the mechanism for transport of S‐Sn‐S into the cytoplasm is not known. 3b, [NiFe] hydrogenase Group 3b; 4e, [NiFe] hydrogenase Group 4e; 4g, [NiFe] hydrogenase Group 4g; Acs, acetyl‐CoA synthase; BMC, bacterial microcompartment; CoA, coenzyme A; Coo, carbon monoxide dehydrogenase; Ech, energy‐converting hydrogenase; Mbh, membrane‐bound [NiFe] hydrogenase; Mrp, Mrp‐type Na+/H+ antiporter; OM, outer membrane; Rnf, ion‐translocating ferredoxin:NAD+ oxidoreductase complex; Shy, sulfhydrogenase; TCA, tricarboxylic acid.

Across the Ace Lake Ca. Auribacterota MAGs there are gene clusters that encode an array of redox‐driven ion translocation systems (Complex I, Rnf, energy‐converting hydrogenases, Mrp) that generate a gradient that can be used by ATP synthase to generate ATP. The majority of Ca. Auribacterota species encode an Rnf complex, which couples electron transfer from reduced ferredoxin to NAD+ to generate NADH, with concomitant translocation of ions across the membrane (Tremblay et al., 2013). The Ca. Tritonobacter lacicola MAG additionally encodes a membrane‐bound [NiFe] hydrogenase (Mbh) (Group 4g) (Søndergaard et al., 2016) and an Mrp‐type Na+/H+ antiporter in the same gene cluster (Fig. 2). It has been proposed that Mbh transfers electrons from reduced ferredoxin to protons, thereby producing H2 gas; doing so is likely to generate a Na+ gradient across the cell membrane via the Mrp‐type Na+/H+ antiporter module (Mayer and Müller, 2014; Søndergaard et al., 2016; Yu et al., 2018).

MAGs of four Ace Lake Ca. Auribacterota species (Ca. Tritonobacter lacicola, Ca. Euphemobacter frigidus, Ca. Ancaeobacter aquaticus, Ca. Theseobacter exili) encode a reversible, membrane‐bound, energy‐converting (Ech‐type) hydrogenase (Group 4e). This can couple oxidation of ferredoxin (such as generated from the catabolism of sugars and amino acids) to reduction of protons (which generates H2) and to ion translocation across the cell membrane (Sapra et al., 2003; Søndergaard et al., 2016). Ca. Tritonobacter lacicola additionally encodes carbon monoxide dehydrogenase (CODH), comprising the catalytic subunit (CooS) and electron transfer protein (CooF), as in Rhodospirillum rubrum (Kerby et al., 1992; Singer et al., 2006). We propose that CODH in Ca. Tritonobacter lacicola works in concert with the Ech‐type hydrogenase to generate a chemiosmotic gradient by coupling CO oxidation to H2 evolution, as reported for certain CO‐oxidizing bacteria and archaea (Kerby et al., 1992; Schoelmerich and Müller, 2019). In the same gene cluster as the CODH genes of Ca. Tritonobacter lacicola are genes for acetyl‐CoA synthase (ACS): the subunit (AcsB) responsible for acetyl‐CoA formation via condensation of CO with a methyl group and CoA, and a corrinoid iron–sulfur complex (AcsC/AcsD) that provides the methyl group (Adam et al., 2018). Given that CooS (=AcsA) is reversible, one possibility is that this enzyme can also reduce CO2 to CO for acetyl‐CoA synthesis, as part of a CODH–ACS complex. This raises the possibility that CO2 could be fixed to generate acetyl‐CoA, which could be further carboxylated to pyruvate using pyruvate:ferredoxin oxidoreductase (POR) (Ragsdale and Pierce, 2008). Thus, Ca. Tritonobacter lacicola may have a reversible CODH that has distinct metabolic functions according to the identity of the enzyme it associates with: CO oxidized to drive proton translocation when CODH acts in concert with Ech hydrogenase; or CO2 reduced to CO for acetyl‐CoA synthesis when CODH acts as part of a CODH–ACS complex.

Ca. Erginobacter occultus and Ca. Euphemobacter frigidus MAGs also encode components of ACS (AcsB, AcsC/AcsD), but no CooS/AcsA homologue; thus, in both species the ACS could be used to cleave acetyl‐CoA to generate a methyl group for methionine synthesis, as in Dehalococcoides mccartyi (Zhuang et al., 2014). No Ca. Auribacterota MAG encodes a complete methyl branch of the Wood–Ljungdahl pathway, although a truncated version is encoded in the Ca. Tritonobacter lacicola MAG, comprising formate‐tetrahydrofolate [THF] synthase [Fhs] and bifunctional methylene‐THF dehydrogenase/methenyl‐THF cyclohydrolase [FolD]), as in D. mccartyi. In D. mccartyi, this truncated Wood–Ljungdahl pathway is used for the conversion of glycine to serine, by sourcing the methylene group from exogenous formate (Zhuang et al., 2014). Overall, Ace Lake Ca. Auribacterota MAGs possess genes that indicate diverse approaches to one‐carbon metabolism.

The Ca. Erginobacter occultus MAG encodes a [FeFe] hydrogenase (Group A3) that might act during fermentation as an H2‐evolving, confurcating hydrogenase from reduced ferredoxin and NADH; although it could also serve in the opposite direction to oxidize H2 as an energy source through the bifurcation of electrons to ferredoxin and NAD+ (Poudel et al., 2016; Søndergaard et al., 2016). A bifunctional sulfhydrogenase ([NiFe] hydrogenase/sulfur reductase) is encoded in MAGs of all Ace Lake Ca. Auribacterota species. This tetrameric sulfhydrogenase allows excess reductant generated during fermentation to be disposed of as H2 and sulfide respectively (Ma et al., 1993; Silva et al., 1999; Ma et al., 2000).

The Ace Lake Ca. Auribacterota MAGs encode relatively few glycoside hydrolases for the depolymerization of polysaccharides and glycoconjugates. Substrates such as simple sugars, peptides and free amino acids would be imported directly from the environment by specific transporters. Amino acids could be procured through digestion of peptides; numerous proteases and peptidases are encoded in these MAGs. A pathway for the fermentation of branched‐chain amino acids (BCAA) is encoded in MAGs of four Ace Lake species, and three encode a pathway for aromatic amino acid fermentation. For the former, BCAA aminotransferase converts BCAA to branched‐chain oxoacids, using 2‐oxoglutarate as the amino group acceptor, with this transamination also producing glutamate; branched‐chain oxoacids can then be oxidatively decarboxylated by ketoisovalerate oxidoreductase (VOR) to generate reduced ferredoxin (Schut et al., 2001). Similarly, aromatic amino acid aminotransferase converts aromatic amino acids to arylpyruvates; arylpyruvates are oxidatively decarboxylated by indolepyruvate oxidoreductase (IOR) to generate reduced ferredoxin (Mai and Adams, 1994). The glutamate generated from the initial transaminations can be recycled back to 2‐oxoglutarate using glutamate dehydrogenase with the concomitant reduction of NAD+ or NADP+ (Daebeler et al., 2018).

MAGs of two species (Ca. Tritonobacter lacicola, Ca. Erginobacter occultus) encode shell and vertex proteins for the construction of BMCs. BMCs have diverse metabolic functions across the domain Bacteria (Kennedy et al., 2021; Sutter et al., 2021); we predict that those encoded in these Ca. Auribacterota are catabolic BMCs used for DNA degradation. Deoxyribose‐phosphate aldolase converts 2‐deoxy‐d‐ribose to glyceraldehyde‐3‐phosphate (a glycolytic intermediate) and acetaldehyde, the latter of which is toxic. Multiple aldehyde:ferredoxin oxidoreductases are encoded in both species and may function in the oxidation of toxic aldehydes, generating reduced ferredoxin and acetate, and acetaldehyde may be processed by acetaldehyde dehydrogenase to acetyl‐CoA (Sutter et al., 2021).

Phylum Candidatus Hinthialibacterota (=OLB16)

Formerly included in phylum Ca. Omnitrophota within the ‘Planctomycetes–Verrucomicrobia–Chlamydiae’ superphylum, OLB16 is a recently recognized bacterial phylum named after OLB16, a MAG recovered from a partial‐nitritation anammox reactor (Liu et al., 2020). Another MAG from phylum OLB16 (SURF_12) was recovered from same deep terrestrial subsurface fluid as SURF_26 (Momper et al., 2017). SURF_12 was predicted to be capable of nitrate reduction, sulfite oxidation, methane oxidation, and carbon fixation via the Wood–Ljungdahl pathway (Momper et al., 2017). Two further OLB16 MAGs from the same subsurface fluids were predicted to be fermentative heterotrophs capable of sulfur oxidation/reduction, metal reduction, and hydrogen oxidation/generation (Momper et al., 2021).

From Ace Lake, a single OLB16 MAG (3300035698_1186; 100% complete) was identified and is here designated the type MAG for Ca. Hinthialibacter antarcticus gen. et sp. nov., with the OLB16 candidate phylum named Ca. Hinthialibacterota phylum nov. (Table 2; Table S2). Very few of the metabolic abilities inferred from the deep terrestrial subsurface fluid OLB16 MAGs were encoded in the Ca. Hinthialibacter antarcticus MAG, suggesting that the metabolism of the Ace Lake representative is distinct.

Table 2

Summary of the physiological and metabolic traits inferred from the Ace Lake MAG of phylum Candidatus Hinthialibacterota phylum nov.

Candidatus Hinthialibacter antarcticus gen. et sp. nov.

Genus named after the Etruscan word hinthial, meaning ‘shade’ or ‘of the underworld’; the species name means ‘southern’

3300035698_1186 (type): 100% complete, 0% contamination

Highest abundance: 0.9% (13 m, oxic–anoxic interface)

Glycolysis and gluconeogenesis (includes both ATP‐dependent and PPi‐dependent 6‐phosphofructokinase) TCA cycle (oxidative) [includes citrate (Si)‐synthase] Pentose phosphate pathway (non‐oxidative) Aerobic respiration: cytochrome c oxidase complex Anaerobic respiration: dissimilatory nitrate reductase and dissimilatory nitrite reductase (cytochrome c‐552); nitrate reduced to ammonia via nitrite NADH‐quinone oxidoreductase (complex I) F‐type ATP synthase Cytoplasmic [NiFe] hydrogenases: NAD‐coupled (Group 3d); bifurcating, heterodisulfide reductase‐linked (Group 3c): used for H2 oxidation and/or fermentation Assimilatory sulfate reduction NAD(P) transhydrogenase Glycoside hydrolases and pectinolytic enzymes: 42 (34 with signal peptide) Sulfatases: 39 (34 with signal peptide) Degradation of fucose, rhamnose, xylose, gluconate/glucuronate, phosphonate, alcohols Glycogen synthesis Fermentation of aromatic amino acids BMC (involved in fucose, rhamnose and phosphonate degradation/aldehyde oxidation) Acetyl‐CoA to acetate, with ATP generation by substrate‐level phosphorylation Proteases/peptidases: 14 (five with signal peptide) Polyphosphate synthesis ABC transporters: sugars, peptides, BCAA, osmoprotectants, nucleosides, cobalamin, Fe(III), Mn, Zn, Co/Ni, molybdate Other transporters: sugars, ammonia, amino acids, nucleosides, carboxylates, gluconates, phosphate, sulfate, Fe(II), Zn, Mg Catalase, catalase‐peroxidase, superoxide dismutase, hydroxylamine reductase Flagella Tight adherence (Tad) pili

The complete etymologies of the genus and species are given in Table S2.

The Ca. Hinthialibacter antarcticus MAG had the highest relative abundance (0.9%) in the oxic–anoxic interface of Ace Lake (where Ca. Chlorobium antarcticum dominates; Panwar et al., 2020; Panwar et al., 2021), as well as being detected in the oxic zone above and anoxic zone below the interface (Table S3). Genes for cytochrome c oxidase complex are consistent with the capacity for aerobic respiration. We predict Ca. Hinthialibacter antarcticus to be a facultative anaerobe capable of dissimilatory nitrate reduction to ammonia (DNRA), based on genes for respiratory nitrate reductase (NarGH) and nitrite reductase (cytochrome c‐552) (NrfAH) (Einsle et al., 1999) (Fig. 3). Nitrate and nitrite levels are at low concentrations in the oxygenated layers of Ace Lake (<0.4 mM) (Rankin et al., 1999), and not detected in the anoxic zone, which likely limits the capacity for DNRA below the interface. Under anoxic conditions, Ca. Hinthialibacter antarcticus is inferred to also use fermentation: two bidirectional cytoplasmic [NiFe] hydrogenases are encoded in this MAG (Table 2), which would facilitate fermentative growth, as well as allow H2 to be used as an energy source (Heim et al., 1998; Kaster et al., 2011; Greening et al., 2016).

Fig. 3

Metabolic capacities of Candidatus Hinthialibacter antarcticus (candidate phylum Hinthialibacterota) as inferred from MAG sequences. Oligo/polysaccharides include both sulfated and non‐sulfated oligo/polysaccharides. The transporter by which phosphonate enters the cytoplasm is not known. 3c, [NiFe] hydrogenase Group 3c; 3d, [NiFe] hydrogenase Group 3d; ASR, assimilatory sulfate reduction; BMC, bacterial microcompartment; Cta, cytochrome c oxidase; DNRA, dissimilatory nitrate reduction to ammonia; GH, glycoside hydrolases; Nar, respiratory nitrate reductase; Nrf, nitrite reductase (cytochrome c‐552); OM, outer membrane; Tad, tight adherence; TCA, tricarboxylic acid.

Based on the numerous GH genes from many GH families, Ca. Hinthialibacter antarcticus is predicted to have the capacity to degrade a diverse array of polysaccharides, including those derived from the cell walls of algae. The MAG also encodes abundant proteases and peptidases. Thus, we infer that this bacterium is capable of obtaining substrates from the degradation of biopolymers. As well as flagella for swimming motility, Ca. Hinthialibacter antarcticus encodes Tad (tight adherence) pili (Giltner et al., 2012), which could possibly be utilized for attachment to cell debris and other detritus. Potential pectinolytic enzymes (e.g. glucuronyl hydrolase, polygalacturonan lyase, polygalacturonase), all of which include N‐terminal signal peptides (suggesting they act extracytoplasmically), may generate oligosaccharides and hexuronates from pectin for further catabolism. Other polysaccharides that Ca. Hinthialibacter antarcticus could degrade include sulfated polysaccharides such as fucoidans. Fucoidans are predominantly composed of fucose with a variable composition of other monosaccharides such as glucose, rhamnose, mannose, and xylose (Ale et al., 2011). Fucosidases, rhamnosidases, and xylosidases are also encoded (the majority with signal peptides) for the release of fucose, rhamnose, and xylose, respectively, from polysaccharides, as are enzymes for their subsequent catabolism (Bunesova et al., 2016). Fucoidan digestion is facilitated by sulfatases, which catalyse the hydrolytic cleavage of sulfate esters to remove the sulfate groups (Wegner et al., 2013). Ca. Hinthialibacter antarcticus encodes 39 sulfatases; the majority of these (34) include N‐terminal signal peptides, which suggest a periplasmic or extracellular location, as predicted for sulfatases involved in sulfated polysaccharide degradation by other bacteria (e.g. Reisky et al., 2019; van Vliet et al., 2019). Collectively, these sulfatases from Ca. Hinthialibacter antarcticus appear to be involved in the degradation of disparate substrates, with certain sulfatases showing the highest sequence identities to sulfatases that target sulfated algal polysaccharides (Reisky et al., 2019), whereas others show the highest sequence identities to glycosaminoglycan sulfatases (Ulmer et al., 2014).

The Ca. Hinthialibacter antarcticus MAG encodes shell and vertex proteins for BMCs, which could be used to contain lactaldehyde, a toxic product of fucose and rhamnose degradation (Orellana et al., 2021) that can be converted to lactate by lactaldehyde dehydrogenase. BMCs may also be associated with phosphonate degradation, with the MAG encoding the two enzymes (2‐aminoethylphosphonate‐pyruvate transaminase, PhnW; phosphonoacetaldehyde hydrolase, PhnX) for the degradation of the most common biogenic phosphonate, 2‐aminoethylphosphonate, to alanine, phosphate, and acetaldehyde, with the latter converted by acetaldehyde dehydrogenase to acetyl‐CoA in BMCs (Zangelmi et al., 2021). Also encoded is the ammonia‐lyase for conversion of another naturally occurring phosphonate, (R)‐1‐hydroxy‐2‐aminoethylphosphonate, to phosphonoacetaldehyde, which can then be catabolized by PhnX (Zangelmi et al., 2021). Thus, phosphonates have the potential to be used by Ca. Hinthialibacter antarcticus as a source of phosphate, nitrogen, and perhaps energy, if acetyl‐CoA is used to generate ATP by substrate‐level phosphorylation by the sequential action of acetate kinase and phosphate acetyltransferase.

Phylum Candidatus Electryoneota (=AABM5‐125‐24)

Candidate phylum AABM5‐125‐24 belongs to the FCB (‘Fibrobacteres‐Chlorobi‐Bacteroidetes’) superphylum, and is here named Candidatus Electryoneota phylum nov. Three Ace Lake MAGs (90%–97% complete) represented two new genera and species of this phylum: Ca. Electryonea clarkiae gen. et sp. nov. [3300035698_1675 (type) and 3300025642_6]; Ca. Hatepunaea meridiana gen. et sp. nov. [3300035698_979 (type)] (Table 3). A third genus represented by a fourth MAG had low completion (3300035698_578; 63%), and revealed comparatively little regarding its potential metabolic abilities (Table S1). We interpret Ca. Electryonea clarkiae and Ca. Hatepunaea meridiana to be facultative anaerobes, capable of aerobic respiration using high O2 affinity enzymes and anaerobic respiration using dissimilatory sulfate reduction (DSR), as previously inferred for other MAGs (from aquatic sediment or oceanic samples) assigned to candidate phylum Ca. Electryoneota (Youssef et al., 2019a). The high O2 affinity enzymes cytochrome bd respiratory O2 reductase (Borisov et al., 2011) and cytochrome c oxidase cbb3‐type (Buschmann et al., 2010) are encoded in the Ca. Electryonea clarkiae MAGs (Fig. 4), whereas only the former was detected in the Ca. Hatepunaea meridiana MAG. Both have a full suite of genes required for DSR: sulfate adenylyltransferase (Sat), adenylylsulfate reductase (AprAB), dissimilatory‐type sulfite reductase (DsvAB), and quinone‐modifying oxidoreductase complex (QmoABC) (Duarte et al., 2016; Youssef et al., 2019a). Ca. Electryonea clarkiae and Ca. Hatepunaea meridiana MAGs had the highest relative abundances in the anoxic zone, at 1.2% (14 m depth) and 0.6% (18 m depth), respectively (Table S3). These facultative aerobes may use high O2 affinity respiratory enzymes within the steep oxycline of the oxic–anoxic interface (Rankin et al., 1999; Lauro et al., 2011); however, unlike Ca. Electryonea clarkiae, Ca. Hatepunaea meridiana was barely detected in the oxic–anoxic interface metagenomes (Table S3). Sulfate, required for respiration by DSR, is abundant in the oxic–anoxic interface of Ace Lake, but increasingly limiting with depth in the anoxic zone (Burton and Barker, 1979; Franzmann et al., 1991; Rankin et al., 1999) (Fig. S1).

Table 3

Summary of the physiological and metabolic traits inferred from three Ace Lake MAGs of phylum Candidatus Electryoneota phylum nov.

Candidatus Electryonea clarkiae gen. et sp. nov.

Genus named after the Greek demi‐goddess Electryone, who presides over awakening from sleep; species named in honour of marine biologist Eugenie Clark

3300035698_1675 (type): 97% complete, 1.1% contamination 3300025642_6: 90% complete, 6.1% contamination

Highest abundance: 1.2% (14 m, anoxic zone)

Glycolysis, gluconeogenesis (includes both ATP‐dependent and PPi‐dependent 6‐phosphofructokinase) TCA cycle (oxidative) [includes citrate (Si)‐synthase] TCA cycle (reductive) (includes fumarate reductase, reversible citrate synthase) Pentose phosphate pathway (non‐oxidative) Aerobic respiration: cytochrome bd respiratory O2 reductase, and cytochrome c oxidase cbb3‐type; both high O2 affinity Anaerobic respiration: dissimilatory sulfate reduction Sulfite oxidation Thiosulfate reduction NADH‐quinone oxidoreductase (complex I) F‐type ATP synthase Ferredoxin:NAD+ oxidoreductase Rnf complex NAD(P) transhydrogenase Cytoplasmic [FeFe] hydrogenase, bifurcating (Group A3): used for H2 oxidation (such as to drive reductive TCA cycle) Glycoside hydrolases: 3 (none with signal peptide) Glycogen synthesis Acetyl‐CoA to acetate, with ATP generation by substrate‐level phosphorylation Proteases/peptidases: 38 (25 with signal peptide) Fermentation of BCAA and aromatic amino acids Degradation of aspartate, proline, trans‐4‐hydroxy‐l‐proline, methionine, histidine, tryptophan, diaminopropionate, glycerol, alcohols Poly‐γ‐glutamate synthesis Synthesis of glutamate, glutamine, aspartate, asparagine, alanine, serine, glycine, methionine, cysteine, lysine, BCAA ABC transporters: sugars, peptides, phosphate, molybdate, Zn, Mn Other transporters: sugars, peptides, amino acids, nucleosides, carboxylates, phosphate, sulfate, Fe (II), Zn, Mn, Mg Bacillithiol synthesis Catalase, desulfoferrodoxin Flagella

Candidatus Hatepunaea meridiana gen. et sp. nov.

Genus named after Hatepuna, a Hattian water deity and daughter of a sea god; species name means ‘of the south’

3300035698_979 (type): 95% complete, 0% contamination

Highest abundance: 0.6% (18 m, anoxic zone)

Glycolysis, gluconeogenesis (includes PPi‐dependent 6‐phosphofructokinase) TCA cycle (oxidative) [includes citrate (Si)‐synthase] TCA cycle (reductive) (includes fumarate reductase, reversible citrate synthase) Pentose phosphate pathway (non‐oxidative) Aerobic respiration: cytochrome bd respiratory O2 reductase; high O2 affinity Anaerobic respiration: dissimilatory sulfate reduction Sulfite oxidation NADH‐quinone oxidoreductase (complex I) F‐type ATP synthase Ferredoxin:NAD+ oxidoreductase Rnf complex Glycoside hydrolases: 2 (one with signal peptide) Glycogen synthesis Acetyl‐CoA to acetate, with ATP generation by substrate‐level phosphorylation Proteases/peptidases: 27 (16 with signal peptide) Fermentation of BCAA and aromatic amino acids Degradation of aspartate, methionine, histidine, tryptophan Poly‐γ‐glutamate synthesis Synthesis of glutamate, glutamine, aspartate, asparagine, alanine, serine, glycine, methionine, cysteine, lysine, BCAA ABC transporters: peptides, Zn Other transporters: peptides, amino acids, carboxylates, phosphate, sulfate, Fe (II), Zn, Mg Bacillithiol synthesis Catalase, desulfoferrodoxin Flagella? (flagellar genes are encoded, but not flagellin genes)

The complete etymologies of genera and species are given in Table S2.

Fig. 4

Metabolic capacities of Candidatus Electryonea clarkiae (candidate phylum Electryoneota) as inferred from MAG sequences. The transporter by which glycerol enters the cytoplasm is not known. The identity of the sulfur species transported by the sulfur compound transporter (Pmp) is not known, and the function of the sulfurtransferase (ST) is also not known; both are encoded in the same gene cluster as thiosulfate reductase (Phs). One possibility, presented here, is that an unknown sulfur species (S*) that is a product of thiosulfate reduction is taken up by Pmp. A3, [FeFe] hydrogenase Group A3; Cyd, cytochrome bd ubiquinol oxidase; Cco, cbb3‐type cytochrome c oxidase; DHAP, dihydroxyacetone phosphate; DSR, dissimilatory sulfate reduction; Fdh, formate dehydrogenase; OM, outer membrane; PGA, poly‐ γ‐glutamate; Qmo, quinone‐modifying oxidoreductase complex; Rnf, ion‐translocating ferredoxin:NAD+ oxidoreductase complex; Soe, sulfite dehydrogenase.

A number of previously unreported metabolic traits for phylum Ca. Electryoneota were apparent from examination of the Ace Lake MAGs. The Ca. Electryonea clarkiae and Ca. Hatepunaea meridiana MAGs possess gene clusters for membrane‐bound Complex Iron–Sulfur Molybdoenzyme (CISM) complexes (Schoepp‐Cothenet et al., 2012). Typical of these complexes, each comprises three subunits: a molybdopterin‐binding catalytic subunit, an electron transfer subunit, and a membrane anchor subunit involved in the transfer of electrons from the quinone pool (Kadnikov et al., 2013) that has also been proposed to carry out proton translocation (Jormakka et al., 2008; Marreiros et al., 2016; Calisto and Pereira, 2021). Ca. Electryonea clarkiae encodes two CISM clusters, whereas Ca. Hatepunaea meridiana encodes one. Phylogenetic analysis of the catalytic subunits of each of these revealed that both Ca. Electryonea clarkiae and Ca. Hatepunaea meridiana encode a sulfite dehydrogenase (SoeABC; SoeA corresponds to the catalytic subunit), and that Ca. Electryonea clarkiae additionally encodes a possible thiosulfate reductase (see below) (Fig. 5). Sulfite dehydrogenase is responsible for the oxidation of sulfite in the cytoplasm, as in Allochromatium vinosum (Dahl et al., 2013). Thus, we posit that Ca. Electryoneota possess two possible routes for sulfite detoxification: reduction of sulfite by DsvAB, as the terminal step of DSR; or oxidation by SoeABC, which as well as being energetically favourable (Simon and Kroneck, 2013), allows sulfate to be regenerated.

Fig. 5

Unrooted maximum‐likelihood tree of Complex Iron–Sulfur Molybdoenzyme catalytic subunit homologues from select Ace Lake MAGs and reference sequences. Ace Lake sequences: IMG gene ID, taxonomy, red font. Reference sequences (accession ID, taxonomy, function; experimental evidence shown in bold font): P06131.1 (Shuber et al., 1986), P07658.2 (Axley et al., 1990), D7AF63.1 (Kaufmann and Lovley, 2001; Coppi et al., 2007), P42434.3 (Ogawa et al., 1995), ABB51928.1 (Kashyap et al., 2006), AAR05656.1 (Santini and vanden Hoven, 2004), Q9HR74.1 (Muller and DasSarma, 2005), P18775.2 (Weiner et al., 1988), P45004.1 (Loosmore et al., 1996), Q47CW6.1 (Bender et al., 2005), I3R9M9.1 (Lledo et al., 2004), Q9S1H0.1 (Schröder et al., 1997), P60068.1 (Thorell et al., 2003), Q71EW5.1 (Rosner and Schink, 1995), A0A369NIV7.1 (Maini Rekdal et al., 2019), D3RNN8.1 (Dahl et al., 2013), Q9Z4S6.1 (Winter et al., 2010), AAU11839.1 (Malasarn et al., 2004), AAQ01672.1 (Saltikov and Newman, 2003), Q72E84.1 (Venceslau et al., 2010), WP_010937484 (Aketagawa et al., 1985), P31075.1 (Krafft et al., 1992), P37600 (Clark and Barrett, 1987), WP_011073774.1 (Burns and DiChristina, 2009). PSRLC, polysulfide reductase‐like complex; this notation for uncharacterized CISM superfamily proteins from green sulfur bacteria (Chlorobi) follows Frigaard and Bryant (2008). PSLRC3 is inferred to be sulfite oxidase (Frigaard and Bryant, 2008; Gregersen et al., 2011). Dots located on and to the right of nodes indicate the SH‐aLRT branching support and ultrafast bootstrap support respectively. Black dot = 90%–100% support; grey dot = 50%–89% support; no dot indicates <50% support.

Ca. Electryonea clarkiae MAGs encode a second CISM complex, for which the catalytic subunit has a signal peptide (Tat) for a periplasmic orientation. Phylogenetic analysis of this catalytic subunit recovered it within a cluster that includes the catalytic subunits of known thiosulfate reductases (PhsA) from Salmonella enterica (Clark and Barrett, 1987; Hinsley and Berks, 2002), Shewanella oneidensis (Burns and DiChristina, 2009), and Desulfovibrio vulgaris (Aketagawa et al., 1985), and polysulfide reductase (PsrA) from Wolinella succinogenes (Krafft et al., 1992) (Fig. 5). Within this PhsA/PsrA cluster, the homologue in the Ca. Electryonea clarkiae MAG showed the highest sequence identity (57%) and the closest phylogenetic relationship (Fig. 5) to the PhsA/PsrA subunit of a putative thiosulfate/polysulfide reductase (Phs/Psr) in Caldithrix abyssi (Kublanov et al., 2017). Previous work found that thiosulfate stimulated fermentative growth of C. abyssi when peptone was supplied as a growth substrate, and led to sulfide being produced; however, thiosulfate did not support C. abyssi growth by respiration using acetate or hydrogen as substrates (Kublanov et al., 2017) (see Supplementary Information: Other sulfur metabolism enzymes). A function of Phs in C. abyssi fermentation is consistent with the proposed function of Phs in S. enterica, in which thiosulfate reduction is mainly an adjunct to a fermentative metabolism, rather than used for respiration (Heinzinger et al., 1995; Hinsley and Berks, 2002; Stoffels et al., 2012). Thus, we propose that Ca. Electryonea clarkiae has a Phs that uses thiosulfate as an electron acceptor during fermentation, with thiosulfate reduction taking place in the periplasm. Immediately downstream of the three Phs genes (phsABC) in Ca. Electryonea clarkiae are two genes for sulfur compound transporters (PmpA/B) (Gristwood et al., 2011) and a sulfurtransferase gene (Table S1). A homologous gene cluster is encoded in C. abyssi (GenBank accession CP018099.1). These PmpA/B transporters contain a single sulfur compound transport domain, and are homologous to YeeE thiosulfate transporters, which contains two such domains (Tanaka et al., 2020). The identity of the sulfur species transported by PmpA/B and the function of the sulfurtransferase are unclear; but the gene cluster suggests that thiosulfate reduction catalysed by the Ca. Electryonea clarkiae Phs complex may have a metabolic function outside of fermentation.

The heterotrophic abilities of Ace Lake Ca. Electryoneota appear to be directed toward the breakdown of proteinaceous matter, with amino acids utilized by further catabolism. In MAGs of both Ca. Electryonea clarkiae and Ca. Hatepunaea meridiana very few genes for poly‐ or oligosaccharide degradation enzymes were detected, in contrast to numerous genes for proteases and peptidases, many of which are predicted to be serine proteases of the subtilisin family (Peptidase S8) (Table S1). Most of the latter have N‐terminal signal peptides, which suggest an extracytoplasmic location. Ca. Electryonea clarkiae possesses genes for the degradation of amino acids including the catabolism of trans‐4‐hydroxy‐l‐proline to glutamate (Levin et al., 2017). Ca. Electryonea clarkiae MAGs encode VOR and IOR genes for BCAA and aromatic amino acid fermentation respectively (Schut et al., 2001; Daebeler et al., 2018); in addition to VOR and IOR, a branched‐chain α‐keto acid dehydrogenase complex is also encoded in Ca. Electryonea clarkiae, which like VOR uses branched‐chain oxoacids as substrates (Venugopal et al., 2011).

The MAGs of both Ca. Electryonea clarkiae and Ca. Hatepunaea meridiana each encode a protein that is homologous to both trimethylamine:corrinoid methyltransferase (MttB) and glycine betaine:corrinoid methyltransferase (MtgB) (Ticak et al., 2014). In general across bacteria and archaea, MttB includes a genetically encoded pyrrolysine residue, which is not present in MtgB. There is no evidence for pyrrolysine in the sequences of the Ca. Electryoneota MttB/MtgB homologues. Phylogenetic analysis recovered the Ca. Electryoneota proteins as belonging to a larger clade of uncharacterized proteins separate from both MtgB and MttB, but more closely related to MttB (Fig. S2). In MAGs of both Ca. Electryonea clarkiae and Ca. Hatepunaea meridiana, each MttB/MtgB homologue is encoded in the same gene cluster as three genes that encode homologues of three of the four modules of cobalamin‐dependent methionine synthase (MetH) (Evans et al., 2004): homocysteine‐binding domain, cobalamin‐binding domain (=corrinoid protein), and S‐adenosylmethionine‐binding activation domain (Fig. S3). This gene cluster (non‐pyrrolysine MtgB/MttB homologue, homocysteine S‐methyltransferase, corrinoid protein, activation protein) is found in diverse bacteria (Table S4), and we hypothesize that this MttB/MtgB homologue catalyses a novel corrinoid‐dependent mechanism of methyl transfer for methionine synthesis, using an unknown methyl donor. A more typical multi‐domain cobalamin‐ and THF‐dependent methionine synthase (Evans et al., 2004) is also encoded in MAGs of both Ca. Electryonea clarkiae (1115 amino acids) and Ca. Hatepunaea meridiana (1223 amino acids), suggesting versatility in one‐carbon metabolism.

Ace Lake Ca. Electryoneota MAGs also encode possible autotrophic abilities. The Ca. Electryonea clarkiae MAG encodes a citrate synthase that, among experimentally characterized citrate synthases, shows the highest sequence identity (47%) to the reversible citrate synthase of Thermosulfidibacter takaii, which employs this citrate synthase as part of a reverse TCA cycle (Nunoura et al., 2018). In addition, Ca. Electryonea clarkiae encodes a cytoplasmic, bifurcating [FeFe] hydrogenase (Group A3), and fumarate reductase. We propose that Ca. Electryonea clarkiae has the potential for chemolithoautotrophy that involves carbon fixation via a reverse TCA cycle driven by hydrogen oxidation catalysed by a bifurcating hydrogenase, as described for T. takaii (Nunoura et al., 2018). The same proteins as Ca. Electryonea clarkiae are encoded in the Ca. Hatepunaea meridiana MAG (including citrate synthase 49% match to T. takaii citrate synthase), with the exception of a hydrogenase gene, which is lacking in the Ca. Hatepunaea meridiana MAG.

The Ca. Electryonea clarkiae MAGs possess a complete set of genes for flagellar motility, whereas the Ca. Hatepunaea meridiana MAG encodes a complete set except for flagellin genes. Flagellin genes were reported previously in members of this phylum (Youssef et al., 2019a). MAGs of both Ca. Electryonea clarkiae and Ca. Hatepunaea meridiana encode poly‐γ‐glutamate synthetase (CapBC) and other proteins required for synthesis and transport of poly‐γ‐glutamate, a biopolymer involved in capsule formation, or released extracellularly as a water‐binding component of a biofilm matrix (Rehm, 2010).

Phylum Candidatus Lernaellota (=FEN‐1099)

Two Ace Lake MAGs (both 98% complete) represent candidate phylum FEN‐1099, which is here named phylum Candidatus Lernaellota phylum nov. Each MAG represents a separate genus and species: Ca. Lernaella stagnicola gen. et sp. nov. [3300035698_1648 (type)]; Ca. Alcyoniella australis gen. et sp. nov. [3300035698_1390 (type)] (Table 4). The Ca. Lernaella stagnicola MAG encodes a cytochrome bd ubiquinol oxidase for aerobic respiration under low‐O2 conditions (Fig. 6); no aerobic respiration genes were found in the Ca. Alcyoniella australis MAG. Ca. Alcyoniella australis encodes citrate (Re)‐synthase, which [unlike citrate (Si)‐synthase, encoded in Ca. Lernaella stagnicola] is oxygen‐sensitive and only found in strict anaerobes (Li et al., 2007; Marco‐Urrea et al., 2011). Ca. Lernaella stagnicola and Ca. Alcyoniella australis MAGs were most abundant in the anoxic zone (1.5% and 0.64% respectively), although the former had much higher relative abundance in the oxic–anoxic interface than the latter did (Table S3).

Table 4

Summary of the physiological and metabolic traits inferred from two Ace Lake MAGs of phylum Candidatus Lernaellota phylum nov.

Candidatus Lernaella stagnicola gen. et sp. nov.

Genus named after Lerna, a mythical lake and portal to the underworld; species name means ‘stagnant lake dweller’

3300035698_1648 (type): 98% complete, 0.9% contamination

Highest abundance: 1.5% (19 m, anoxic zone)

Glycolysis, gluconeogenesis (includes both ATP‐dependent and PPi‐dependent 6‐phosphofructokinase) TCA cycle (oxidative) [includes citrate (Si)‐synthase] Glyoxylate bypass Pentose phosphate pathway (non‐oxidative) Aerobic respiration: cytochrome bd respiratory O2 reductase: high O2 affinity NADH‐quinone oxidoreductase (complex I) F‐type ATP synthase Ion‐translocating ferredoxin:NAD+ oxidoreductase Rnf complex Na+‐translocating NADH‐quinone reductase Nqr complex Anaerobic glycerol‐3‐phosphate oxidation Formate oxidation linked to menaquinone reductase complex Thiosulfate reduction Cytoplasmic [NiFe] hydrogenase, NAD‐coupled (Group 3d); cytoplasmic [FeFe] hydrogenase, bifurcating (NAD, ferredoxin) (Group A3): used for H2 oxidation and/or fermentation Cytoplasmic [FeFe] hydrogenase (Group C3): sensory? Assimilatory sulfate reduction Glycoside hydrolases: 27 (14 with signal peptide) Glycogen synthesis Sulfatases: 14 (11 with signal peptide) Proteases/peptidases: 33 (18 with signal peptide) Degradation of rhamnose, xylose, aldehydes, aspartate, methionine, histidine, tyrosine, tryptophan, glycerol, formate, lactate, alcohols, aldehydes, sarcosine, phosphoglycolate Fermentation of aromatic amino acids, glutamate, lysine Synthesis of almost all 20 amino acids required for protein synthesis Poly‐γ‐glutamate synthesis Polyhydroxyalkanoate synthesis ABC transporters: sugars, peptides, cobalamin, riboflavin, phosphate, Zn Other transporters: sugars, peptides, amino acids, carboxylates, formate, lactate, nucleobases, phosphate, sulfate, Fe(II), Zn, Mg Catalase, desulfoferrodoxin, hydroxylamine reductase Flagella

Candidatus Alcyoniella australis gen. et sp. nov.

Genus named after Alcyonia, a mythical lake and portal to the underworld; species name means ‘southern’

3300035698_1390 (type): 98% complete, 1.9% contamination

Highest abundance: 0.64% (19 m, anoxic zone)

Glycolysis, gluconeogenesis (includes both ATP‐dependent and PPi‐dependent 6‐phosphofructokinase) TCA cycle (oxidative) [includes citrate (Re)‐synthase] Pentose phosphate pathway (non‐oxidative) NADH‐quinone oxidoreductase (complex I) F‐type ATP synthase Ion‐translocating ferredoxin:NAD+ oxidoreductase Rnf complex Na+‐translocating NADH‐quinone reductase Nqr complex Cytoplasmic [NiFe] hydrogenase, NAD‐coupled (Group 3d): used for H2 oxidation and/or fermentation Formate oxidation linked to menaquinone reductase complex Wood–Ljungdahl pathway Glycoside hydrolases: 25 (12 with signal peptide) Trehalose and glycogen synthesis Proteases/peptidases: 25 (15 with signal peptide) Degradation of rhamnose, xylose, aspartate, histidine, tryptophan, glycerol, alcohols, aldehydes, sarcosine, phosphoglycolate Fermentation of BCAA, aromatic amino acids, glutamate Synthesis of almost all 20 amino acids required for protein synthesis Polyhydroxyalkanoate synthesis ABC transporters: sugars, peptides, cobalamin, Zn, Mn, Co/Ni, tungstate Other transporters: sugars, peptides, carboxylates, nucleobases, phosphate, Fe(II), Zn, Mg, Co Catalase, desulfoferrodoxin, hydroxylamine reductase

The complete etymologies of genera and species are given in Table S2.

Fig. 6

Metabolic capacities of Candidatus Lernaella stagnicola (candidate phylum Lernaellota) as inferred from MAG sequences. Oligo/polysaccharides include both sulfated and non‐sulfated oligo/polysaccharides. The transporter by which glycerol enters the cytoplasm is not known. For the possible functions of the sulfur compound transporter (Pmp) and sulfurtransferase (ST), see Fig. 4. 3d, [NiFe] hydrogenase Group 3d; A3, [FeFe] hydrogenase Group A3; ASR, assimilatory sulfate reduction; C3, [FeFe] hydrogenase Group C3 (sensory); Cyd, cytochrome bd ubiquinol oxidase; DHAP, dihydroxyacetone phosphate; Fdh, formate dehydrogenase; G3P, glycerol‐3‐phosphate; GH, glycoside hydrolases; Glp, anaerobic glycerol‐3‐phosphate dehydrogenase; OM, outer membrane; PHA, polyhydroxyalkanoate; Phs, thiosulfate reductase; Qrc, menaquinone reductase; Qmo, quinone‐modifying oxidoreductase complex; PGA, poly‐γ‐glutamate; Rnf, ion‐translocating ferredoxin:NAD+ oxidoreductase complex; S*, unknown sulfur species; TCA, tricarboxylic acid.

The Ca. Lernaella stagnicola MAG encodes glycerol kinase and an anaerobic glycerol‐3‐phosphate dehydrogenase (GlpABC) (Varga and Weiner, 1995), suggesting that oxidation of glycerol‐3‐phosphate could be used to directly reduce the menaquinone pool. Both Ca. Lernaella stagnicola and Ca. Alcyoniella australis MAGs encode a four‐subunit complex that we interpret as menaquinone reductase (Qrc) (Fig. 5). In the Ca. Lernaella stagnicola MAG, the entire membrane‐bound complex is encoded (QrcABCD), comprising the multiheme cytochrome c protein (QrcA) and three CISM complex subunits (the gene cluster is incomplete in the Ca. Alcyoniella australis MAG). The membrane‐bound Qrc of Desulfovibrio vulgaris catalyses the reduction of the menaquinone pool using electrons derived from the oxidation of H2 or formate, with the QrcD subunit proposed to be capable of proton translocation (Venceslau et al., 2010). MAGs of Ca. Lernaella stagnicola and Ca. Alcyoniella australis both encode a periplasmic formate dehydrogenase (FdhAB) that lacks a membrane subunit for direct reduction of the menaquinone pool, as also found in D. vulgaris (da Silva et al., 2013); thus, as in the latter (Venceslau et al., 2010), the Qrc in Ca. Lernaellota species likely also uses formate as an electron donor. Unlike what was proposed for D. vulgaris (Venceslau et al., 2010), there is no evidence that the Qrc complex is linked to DSR in these Ca. Lernaellota MAGs; for the latter, the reduced menaquinone pool could possibly be re‐oxidized using a cytoplasmic hydrogenase or fumarate reductase, although the precise mechanism(s) could not be deduced.

The Ca. Lernaella stagnicola MAG encodes a second CISM that is a possible Phs and includes a periplasmic PhsA subunit (Fig. 5). Immediately downstream of the genes for PhsABC are PmpA/B sulfur compound transporter and sulfurtransferase genes (Table S1), similar to what we found in Ca. Electryonea clarkiae MAGs (see Phylum Candidatus Electryoneota (=AABM5‐125‐24), above), except that one of the two PmpA/B transporters in Ca. Lernaella stagnicola MAG also contains a sulfurtransferase domain. As proposed for Ca. Electryonea clarkiae, this Ca. Lernaella stagnicola Phs‐containing gene cluster may encode proteins involved in uptake and/or conversion of other sulfur compounds in addition to thiosulfate.

The Ca. Lernaella stagnicola MAG encodes three hydrogenases, all inferred to be cytoplasmic. Two are reversible: an [NiFe] NADH‐dependent hydrogenase (Group 3d) and a bifurcating [FeFe] hydrogenase (Group A3) (Søndergaard et al., 2016). Additionally, a [FeFe] hydrogenase (Group C3) is encoded by a gene adjacent to the Group A3 hydrogenase gene; Group C3 hydrogenases have been postulated to have an H2‐sensory role (Søndergaard et al., 2016), consistent with the presence of a histidine kinase domain in the Ca. Lernaella stagnicola Group C3 hydrogenase. Ca. Alcyoniella australis encodes a single [NiFe] hydrogenase (Group 3d); in this case, the hydrogenase may function to oxidize H2 to generate reductant for CO2 fixation (Søndergaard et al., 2016), given that Ca. Alcyoniella australis encodes an almost complete Wood–Ljungdahl pathway, with the carbonyl branch (CODH‐ACS) and most of the methyl branch (all enzymes except Fhs) present in the MAG. Thus, CO2 may be successively fixed by CODH‐ACS and POR to generate acetyl‐CoA and pyruvate, respectively (Ragsdale and Pierce, 2008).

Based on the many GHs and proteases/peptidases encoded in the two MAGs, both Ca. Lernaella stagnicola and Ca. Alcyoniella australis are predicted to be proficient at saccharolysis and proteolysis. Abundant sulfatase genes in the Ca. Lernaella stagnicola MAG may serve to release sulfate from sulfated polysaccharides to make the latter more amenable to degradation [see Phylum Candidatus Hinthialibacterota (=OLB16), above]. Ca. Lernaella stagnicola and Ca. Alcyoniella australis MAGs possess genes for rhamnose and xylose degradation; no BMC genes were found, but the MAGs encode multiple aldehyde:ferredoxin oxidoreductases for detoxification of lactaldehyde, a byproduct of rhamnose degradation. Both Ca. Lernaella stagnicola and Ca. Alcyoniella australis encode complete glycolysis and gluconeogenesis pathways, and a complete oxidative TCA cycle. Ca. Lernaella stagnicola additionally encodes both enzymes of the glyoxylate bypass (isocitrate lyase, malate synthase), allowing acetate and glyoxylate to be used as carbon sources. Ca. Lernaella stagnicola and Ca. Alcyoniella australis MAGs also encode enzymes for fermentation of certain amino acids: BCAA, aromatic amino acids, lysine, glutamate (Barker et al., 1982; Buckel, 2001; Daebeler et al., 2018). In addition, both MAGs are notable for encoding enzymes for the deamination of various amino acids (e.g. aspartate, methionine, histidine, tyrosine, tryptophan), potentially allowing their utilization as ammonia sources.

Ecosystem contributions of the new Ace Lake candidate phyla

Interrogation of the 12 Ace Lake MAGs enabled new metabolic traits to be inferred for four candidate phyla of Bacteria: Auribacterota (SURF‐CP‐2), Hinthialibacterota (OLB16), Electryoneota (AABM‐125‐24), and Lernaellota (FEN‐1099). By considering the metabolic potential of the new candidate phyla in combination with relative abundance calculations by depth, understanding of the physicochemical characteristics of the zones they inhabit, and knowledge of other lake taxa, we were able to infer broader ecosystem contributions. MAGs of Ca. Hinthialibacterota and Ca. Lernaellota encode numerous enzymes used for the degradation of polysaccharides (including fucoidan, in the case of the former), and MAGs of Ca. Electryoneota and Ca. Lernaellota encode numerous enzymes required for the degradation of polypeptides. This suggests that these three phyla contribute to the degradation of polymeric organic matter in Ace Lake, especially below the oxic zone. Thus, these new phyla add to the repertoire of Ace Lake bacteria inferred to degrade recalcitrant organic material and particulate matter, including Bacteroidota, Verrucomicrobiota, Planctomycetota, Gammaproteobacteria, and Ca. Cloacimonadota (Panwar et al., 2020; Williams et al., 2021a). We inferred the presence of a Wood–Ljungdahl pathway in certain MAGs belonging to Ca. Auribacterota and Ca. Lernaellota; however, in the absence of genes required for TCA cycle reversal (Schuchmann and Müller, 2016; Youssef et al., 2019b), the Wood–Ljungdahl pathway cannot support autotrophic growth in these taxa. Only in Ca. Electryonea clarkiae (Ca. Electryoneota) could we reconstruct a complete autotrophic pathway, in this case by using the reverse TCA cycle.

Members of all four new phyla are inferred to contribute to sulfur cycling in Ace Lake, including by the cycling of inorganic sulfur species and mineralization of sulfated organic compounds (Fig. 7). During the summer months, the oxic–anoxic interface of Ace Lake is dominated by Ca. Chlorobium antarcticum and sulfate‐reducing Desulfobacterota, which together constitute the centrepiece of a sulfur cycle in which Ca. Chlorobium antarcticum uses sulfide as an electron donor, oxidizing it to sulfate that is used by Desulfobacterota for DSR, which generates sulfide (Ng et al., 2010; Lauro et al., 2011; Panwar et al., 2020) (Fig. 7; Fig. S1). We interpret Ca. Electryoneota in Ace Lake as being capable of DSR, and therefore may also be metabolically linked to Ca. Chlorobium antarcticum. We posit that sulfur and thiosulfate reducers of the phyla Ca. Auribacterota and Ca. Lernaellota also benefit from sulfur species generated as intermediates in the complete oxidation of sulfide to sulfate by Ca. Chlorobium antarcticum. These taxa would therefore provide a source of sulfide to Ace Lake that augments the contribution made by dissimilatory sulfate reducers.

Fig. 7

Predicted roles in sulfur cycling played by members of four candidate phyla (Auribacterota, Hinthialibacterota, Electryoneota, Lernaellota) in the Ace Lake ecosystem. Only the predicted roles of the four aforementioned phyla and the dominant sulfur cycling taxa Candidatus Chlorobium antarcticum and Desulfobacterota are depicted here. The relative abundances of the six taxa are not equivalent. Double‐headed arrows indicate that the sulfur compound is generated and assimilated by different processes by the same taxon.

All five Ace Lake Ca. Auribacterota species encode a complete sulfhydrogenase, which to the best of our knowledge has not been previously reported for this phylum. Ca. Chlorobium antarcticum would be a major source of elemental (zero‐valence) sulfur in Ace Lake. In general, GSB oxidize sulfide to sulfur globules, which are stored and degraded extracellularly (including at a distance from the cell) before being further oxidized to sulfate; it has been proposed that the extracellular location of elemental sulfur benefits GSB by providing extracellular sulfur to sulfur‐reducing bacteria, thereby facilitating the generation of sulfide (Marnocha et al., 2016). Not all sulfur needs to be converted to sulfate for GSB growth to occur (Holkenbrink et al., 2011), and the retention of sulfur globules may be advantageous under dark conditions because sulfur (but not sulfate) can be used by GSB as an electron acceptor during fermentation of storage carbohydrate (Brune, 1989; Holkenbrink et al., 2011); thus, sulfur reduction may allow survival of Ca. Chlorobium antarcticum in Ace Lake during the prolonged darkness of the polar winter. Biogenic elemental sulfur would therefore be readily available in Ace Lake for anaerobic sulfur reducers. The abundance of sulfhydrogenase, based on catalytic subunit ShyC (K17995), was found to be comparable across multiple depths in the interface and anoxic zone (Fig. 8; Fig. S4), which is consistent with the availability of elemental sulfur in these zones (Rankin et al., 1999). These sulfhydrogenases are encoded in representatives of diverse bacterial phyla in Ace Lake (Fig. 9A), including taxa previously implicated in sulfur reduction in Ace Lake (e.g. Desulfobacterota, Ca. Cloacimonadota, Ca. Omnitrophota) (Panwar et al., 2020; Williams et al., 2021a), and phyla previously unreported from Ace Lake, including Ca. Auribacterota, as well as other phyla for which sulfhydrogenases have been reported in other environments (e.g. Elusimicrobiota, Ca. Margulisbacteria) (Matheus Carnevali et al., 2019; Méheust et al., 2020).

Fig. 8

Normalized coverage data for selected sulfur cycling enzymes across different depths and seasons in Ace Lake. Sulfhydrogenase, based on ShyC (K17995), the catalytic subunit of the sulfur reductase component of sulfhydrogenase (yellow circles); thiosulfate/polysulfide reductase, based on the catalytic subunit (PhsA/PsrA) (K08352), with the dataset including Candidatus Chlorobium antarcticum (green squares) or excluding Ca. Chlorobium antarcticum (brown squares); sulfatases (K01130; blue triangles). Metagenomes were from six lake depths: surface, 5 m (Oxic 1), 11.5–13 m (Oxic 2), 12.7–14.5 m (Interface), 14–16 m (Anoxic 1), 18–19 m (Anoxic 2) and 23–24 m (Anoxic 3). Seasons are highlighted by coloured background: red, summer (Dec, Jan, Feb); blue, winter (Jul, Aug); green, spring (Oct, Nov). Normalized coverages are for the metagenomes of the three size fractions (3.0, 0.8 and 0.1 μm) representing a single depth and sampling date (see Experimental procedures). For a logarithmic scale representation of this graph, refer to Fig. S4.

Fig. 9

Contribution of phylum‐level taxa to functional potential in Ace Lake. Pie charts showing the percentage contribution for (A) sulfhydrogenase based on ShyC (K17995), (B) thiosulfate/polysulfide reductase based on PhsA/PsrA (K08352), (C) sulfatases (K01130). Thiosulfate/polysulfide reductase and sulfatase proteins were identified through KEGG analysis, as described previously (Panwar et al., 2020). Sulfhydrogenase (ShyC) proteins were identified through matches to reference ShyC protein sequences and KEGG analysis (see Experimental procedures). Proteins were taxonomically classified using GTDB taxonomy indicated in the Ace Lake co‐assembly IMG data (IMG genome ID: 3300035698) (see Experimental procedures). Each taxon is consistently represented by a single colour in all plots. Note that, for convenience, class Chlorobia has been separated from phylum Bacteroidota in (B). Only taxonomically classified proteins are shown.

Unlike sulfhydrogenase, the distribution of Phs/Psr (based on the catalytic subunit PhsA/PsrA; both classified under K08352) in the Ace Lake water column peaked in abundance at the oxic–anoxic interface, and declined with increasing depth in the anoxic zone (Fig. 8; Fig. S4). The prevalence of Phs/Psr at the interface is partly due to the presence of PSLRC2 in Ca. Chlorobium antarcticum, previously reported in Chlorobium phaeobacteroides and of unknown function (Frigaard and Bryant, 2008); our phylogenetic analysis recovered PSLRC2 in the Phs/Psr group (Fig. 3). However, this decline in abundance of Phs/Psr with increased depth in Ace Lake is apparent even after Ca. Chlorobium antarcticum is removed from the dataset (Fig. 8; Fig. S4). This distribution is consistent with Phs/Prs acting on thiosulfate. Thiosulfate forms abiotically in the environment by chemical reaction between dissolved sulfide and oxygen (Gregersen et al., 2011), the latter available only in oxygenated waters. Thiosulfate is also generated by GSB during sulfide oxidation to sulfate, possibly as an adventitious byproduct of the reaction of the transient intermediate sulfite with other sulfur species (Gregersen et al., 2011; Holkenbrink et al., 2011). Unlike many GSB, Ca. Chlorobium antarcticum lacks the Sox system for thiosulfate oxidation (Ng et al., 2010; Panwar et al., 2021). Thus, thiosulfate generated and exported by Ca. Chlorobium antarcticum could be utilized by other bacteria, including Gammaproteobacteria in the oxic zone that encode Sox systems (Panwar et al., 2021), and by diverse bacteria in the anoxic zone that encode Phs, including Ca. Electryonea clarkiae (Ca. Electryoneota) and Ca. Lernaella stagnicola (Ca. Lernaellota) (Fig. 9B). Moreover, reduction of thiosulfate generates both sulfide and sulfite as products, and it has been proposed that sulfite released from cells can react with elemental sulfur in the environment to generate further thiosulfate, thereby sustaining an intracellular sulfur cycle (Hinsley and Berks, 2002; Burns and DiChristina, 2009), as well as consuming elemental sulfur in the ecosystem. Sulfite could also be used as an energy source in Ca. Electryoneota using sulfite dehydrogenase. Furthermore, the fact that the Phs complex is encoded in the same gene cluster as PmpA/B transporters and sulfurtransferases in Ca. Electryonea clarkiae and Ca. Lernaella stagnicola suggests that thiosulfate (and/or the products of thiosulfate reduction) could be utilized for other purposes in these bacteria.

In addition to sulfate being produced by Ca. Chlorobium antarcticum and other sulfide‐oxidizing bacteria, it can also be generated biogenically by the action of sulfatases, which release sulfate from a broad range of sulfated compounds (Berteau et al., 2006). Sulfatase genes were detected throughout the Ace Lake water column (Fig. 8; Fig. S4). Sulfated organic compounds may be degraded to provide sulfate under conditions of sulfur limitation, or ‘pruned’ of sulfate groups to facilitate access to the carbon skeleton for utilization as nutrient sources (Kertesz, 1999). The latter was previously inferred to be important for aerobic members of the phyla Verrucomicrobiota, Planctomycetota, and Bacteroidota (Panwar et al., 2020), which dominate the Ace Lake taxa that encode sulfatases (Fig. 9C). Ca. Hinthialibacter antarcticus encodes numerous GH and sulfatase genes, and we infer that it uses sulfated organic compounds as sources of both carbon and sulfur (Table 2), including polysaccharides and glycosaminoglycans from algae in Ace Lake and terrestrial fauna in the vicinity of the lake.

To the best of our knowledge, we have ascribed ecological functions to the new Ace Lake species that have not previously been reported for the four ‘dark matter’ phyla: degradation of complex organic matter (Ca. Hinthialibacter antarcticus, Ca. Electryonea clarkiae, Ca. Hatepunaea meridiana, Ca. Lernaella stagnicola, Ca. Alcyoniella australis); ammonification by DNRA (Ca. Hinthialibacter antarcticus); and chemolithoautotrophy (Ca. Electryonea clarkiae, Ca. Hatepunaea meridiana). We also implicate these phyla in processes associated with sulfur cycling (sulfur reduction, thiosulfate reduction, sulfate ester hydrolysis), in addition to DSR, which was previously deduced for Ca. Electryoneota from other environments (Youssef et al., 2019a; Youssef et al., 2019b). The presence of some of the new species in non‐Antarctic environments was indicated by finding matches (88%–100% identity) to sequences in the IMG 16S rRNA gene public assembled metagenomes database; diverse environments were represented, including marine sediments, meromictic ponds, salt marshes, the deep subsurface, a solar saltern, and uranium‐contaminated floodplain (Table S2). In such environments it is likely the Ca. Auribacterota species would be restricted to anoxic conditions, as it is in Ace Lake. By contrast, we interpret species of the phyla Ca. Hinthialibacterota, Ca. Electryoneota, and Ca. Lernaellota (with the exception of Ca. Alcyoniella australis) as facultative anaerobes, and certain species (Ca. Electryonea clarkiae, Ca. Hatepunaea meridiana, Ca. Lernaella stagnicola) capable of respiration under low‐O2 conditions, which may allow growth under a range of oxygen concentrations in aqueous environments. Knowledge of the metabolic potential of the four Ace Lake ‘dark matter’ phyla combined with availability of their MAGs provides considerable scope for performing ecophysiological assessments of a range of environments for which metagenome data are becoming increasingly available (Nayfach et al., 2020).

Experimental procedures

Sample collection and sequencing

Microbial biomass was sampled from Ace Lake in the austral summers of 2006/2007 and 2008/2009, and a full Antarctic seasonal cycle of summer 2013/2014 to summer 2014/2015. Biomass was collected by sequential size fractionation through a 20 μm prefilter onto 3.0, 0.8‐ and 0.1‐μm pore‐sized, large format (293‐mm polyethersulfone membrane) filters, and DNA was extracted from the biomass as described previously (Ng et al., 2010). Six depths were sampled: surface, 5, 11.5–13, 12.7–14.5 (oxic–anoxic interface), 14–16, 18–19 and 23–24 m with the precise depths varying depending on the water level in the lake (Panwar et al., 2020). In winter 2014, samples were not taken below the oxic–anoxic interface (Panwar et al., 2020). DNA sequences for 120 individual metagenomes (Panwar et al., 2020) were uploaded to IMG (Huntemann et al., 2015). High‐ and medium‐quality MAGs were auto‐generated from individual metagenomes during the IMG pipeline process. QC_filtered raw reads from the individual Ace Lake metagenomes were co‐assembled using Megahit v1.1.1 (Li et al., 2016) with a setting of meta‐large, and MAGs were generated from the co‐assembly using MetaBAT v2.12.1 with minContig length 2500 bp (Kang et al., 2019). MAGs from the co‐assembly (available in IMG as Metagenome ID 3300035698) were assessed for completeness and contamination using CheckM v1.0.7 (Parks et al., 2015), for taxonomic identity using RefineM v 0.0.23 (Parks et al., 2017), and for phylogenetic placement using GTDB Toolkit (GTDB‐Tk) v.1.4.0 with GTDB release R95 (Chaumeil et al., 2019; Parks et al., 2020) and dependencies as described previously (Williams et al., 2021a).

Relationship to other taxa, and biogeographic distribution

Phylogenetic trees were constructed using the classify and de novo workflows in GTDB‐Tk and visualized using Dendroscope v3.5.7 (Huson and Scornavacca, 2012). Relationships between Ace Lake ‘dark matter’ MAGs and their nearest relatives in the GTDB‐Tk reference tree were investigated using CompareM v0.1.1 (https://github.com/dparks1134/CompareM) to calculate average AAI and fastANI v 1.32 (Jain et al., 2018) to calculate ANI. To better understand the uniqueness and biogeographic distribution of these taxa, 16S rRNA genes present in the Ace Lake MAGs were used to query the IMG ‘All Isolates’ dataset (includes all isolates, MAGs and SAGs), the IMG ‘16S rRNA public assembled metagenomes 2021‐09‐24’ dataset and the NCBI nt database (accessed 17‐12‐2021). MAG 3300035698_578 did not contain a 16S rRNA gene so for this MAG the 23S rRNA gene was used to query the IMG ‘All Isolates’ dataset, the IMG ‘23S rRNA public assembled metagenomes 2021‐10‐05’ dataset and NCBI nt.

Genome annotation

The genomic functional potential of the MAGs was assessed by considering cellular and metabolic traits based upon manual examination of proteins and pathways that was performed in a similar way to previous assessments of the validity of gene functional assignments (Allen et al., 2019; Panwar et al., 2020; Williams et al., 2021a, 2021b). Instead of querying individual genes or pathways as the basis for predicting metabolic reconstructions, we used automated annotations as a starting point for a more rigorous approach: individual proteins annotated by IMG were vetted manually, and metabolic reconstructions were inferred by organizing these proteins into known pathways. Individual metabolic pathways inferred for each MAG were then connected to provide a comprehensive assessment of the predicted metabolic capacities of these taxa. The manual vetting process for each protein was as follows. Protein sequences were submitted to ExPASy BLAST or NCBI BLASTP (both using the ‘UniProtKB/Swiss‐Prot only’ option); proteins needed to show ≥30% sequence identity over the length of the protein and/or an expectation (E) value <10−10 (Pearson, 2013) to an experimentally verified protein in the ExPASy BLAST database for the functional annotation to be considered valid. If this threshold was not reached, protein sequences were submitted to InterProScan (Blum et al., 2020) to identify functional domains and potential subcellular locations, and in certain cases phylogenetic analysis was performed to help resolve the function of a protein [e.g. CISM superfamily enzymes, corrinoid:methyltransferases]. GH families were identified using CAZy (Carbohydrate‐Active enZymes) classification (Lombard et al., 2014). Protein sequences that were initially identified as hydrogenases based on catalytic domains were classified further using HydDB (Søndergaard et al., 2016). IMG annotations that could not be verified using the aforementioned process were excluded from this study. All protein annotations in this study are putative, and a full list of proteins identified in this study has been provided (Table S1). MAGs were named according to recommendations for describing novel Candidatus species (Konstantinidis et al., 2017; Chuvochina et al., 2019; Murray et al., 2020). All MAGs used in this study, including type MAGs, etymologies, and supporting metadata, are provided in Table S2.

Functional potential analysis and determination of contributing taxa

IMG‐generated KEGG orthology data were used to assess the functional potential of thiosulfate/polysulfide reductases (K08352), sulfhydrogenases (K17665), and sulfatases (K01130) in Ace Lake, using a previously described method (Panwar et al., 2020). Additional sulfhydrogenase proteins were identified through alignment of the protein sequences from 120 Ace Lake metagenomes to reference protein sequences of ShyC, the catalytic subunit of sulfhydrogenase. The alignment was performed using the blastp module of DIAMOND v0.9.31 (Buchfink et al., 2014) with E value 10−5 and 35% minimum alignment identity, and only alignments covering ≥50% of reference sequence lengths were considered. To determine contributing taxa, the proteins identified as thiosulfate/polysulfide reductases, sulfhydrogenases, and sulfatases were aligned to reference proteins in Ace Lake co‐assembly MetaBAT data (IMG genome ID: 3300035698), and the MAG taxonomies were used to classify the query proteins. The alignment was performed using the blastp module of DIAMOND v0.9.31 with e‐value 10−5, 35% minimum alignment identity, 50% minimum query cover and 50% minimum subject cover. To show the distribution of contributing taxa, the taxonomically classified proteins were plotted as pie charts.

Where specific protein families of interest were identified (CISM catalytic subunits; corrinoid:methyltransferase), multiple sequence alignments were created with the proteins of interest and selected reference sequences from Swiss‐Prot or NCBI nr, and trees constructed using MEGA (Kumar et al., 2018) or IQ‐TREE. For phylogenetic analysis of CISM catalytic subunit homologues, multiple sequence alignment was created using Muscle (Edgar, 2004), manually curated, and submitted to IQ‐TREE webserver (Nguyen et al., 2015) for maximum‐likelihood model selection (Kalyaanamoorthy et al., 2017), tree reconstruction, and ultrafast bootstrapping (1000 replicates; Hoang et al., 2017). Maximum‐likelihood tree of corrinoid:methyltransferases was prepared with MEGA X v10.1.7 using MUSCLE algorithm for protein alignment and 1000 bootstrap replications.

Calculation of taxon abundances

The abundances of candidate phyla Auribacterota, Hinthialibacterota, Electryoneota, and Lernaellota MAGs were calculated using the contigs in the Ace Lake co‐assembly MetaBAT data, similar to a previously described approach used for calculating operational taxonomic unit abundances (Panwar et al., 2020; Panwar et al., 2021). Here, the read depths of co‐assembly MAG contigs and unbinned contigs in 119 Ace Lake metagenomes and contig lengths were used to calculate contig‐based abundances (Panwar et al., 2020). The Dec. 2014 3–20 μm filter metagenome from the oxic–anoxic interface of Ace Lake was not available when the Ace Lake co‐assembly MetaBAT data were generated, so this metagenome was not included in MAG abundance analyses. MAG relative abundance was calculated by dividing the MAG abundance in a metagenome by the total metagenome abundance (Panwar et al., 2020). Highest relative abundance of each MAG refers to peak relative abundance in a single metagenome (Panwar et al., 2020). The total metagenome abundance refers to the sum of abundances (average read depth × length) of all MAG contigs and unbinned contigs in a metagenome (Panwar et al., 2020). To generate Ace Lake depth/season abundance profiles of MAGs, the relative abundances of MAGs were plotted as bar charts.

Authors' Contributions

T. J. Williams conceived the study and performed the primary analyses and writing of the manuscript. M. A. Allen and P. Panwar performed computational analyses. All authors participated in the interpretation of the findings and contributed to the writing of the manuscript. All authors read and approved the final manuscript.

Availability of Data

All metagenomes and medium and high‐quality MAGs used in this study are available in IMG: Metagenome ID 3300035698, 3300031227 and 3300025642.

Appendix S1. Other sulfur metabolism enzymes.

Fig, S1. Sulfate and sulfide concentrations in Ace Lake.

Fig. S2. Maximum‐likelihood tree of corrinoid:methyltransferases from select Ace Lake metagenome‐assembled genomes and reference sequences.

Fig. S3. Methionine synthase gene cluster arrangement.

Fig. S4. Normalized coverage data for selected sulfur cycling enzymes across different depths and seasons in Ace Lake, shown in logarithmic scale.

Table S4. Proteins homologous to methionine synthase domains, and a novel methyltransferase encoded in the same gene cluster.

Click here for additional data file.

Table S1. Proteins and Gene IDs for Ace Lake MAGs and SURF_26 chosen for analysis in this study. (Excel spread sheet)

Click here for additional data file.

Table S2. Etymologies and metadata for proposed new Candidatus genus and species. (Excel spread sheet)

Click here for additional data file.

Table S3. MAG abundance data. (Excel spread sheet)

Click here for additional data file.