Genomic Characterization of Candidate Division LCP-89 Reveals an Atypical Cell Wall Structure, Microcompartment Production, and Dual Respiratory and Fermentative Capacities

Overall, we obtained 87 high-quality, 196 medium-quality, and 42 low-quality draft genomes from the source (as defined in reference 22). Concurrently, 75 draft single-cell genomes were sequenced from the spring source, bringing the total number of genomes already available to this effort to 400 genomic assemblies. Initial taxonomic classification of genomic bins obtained from Zodletone Spring source sediments emphasized the high phylogenetic diversity of the spring. Collectively, representatives of 46 bacterial and 8 archaeal phyla were identified, 32 of which belong to uncultured bacterial and archaeal phyla.

Six MAGs and four SAGs were recovered from Zodletone Spring sediments as part of the effort described above. Detailed assembly statistics for these assemblies are presented in Table S3 in the supplemental material. In addition, 2 SAGs were recovered from Lake Baikal, Irkutsk, Russia, 1 SAG was recovered from CrabSpa hydrothermal vent, East Pacific Rise, and 1 SAG was recovered from sediment of Walker Lake, Nevada, with members of the phylum Calditrichota (Calorithrix insularis, Calditrhix abyssii, and Calditrhix palaeochoryensis) as their closest cultured relatives (12.7% to 17.3% 16S rRNA gene divergence and 39.5% to 54.5% average amino acid identity [AAI]) (Table 1). Detailed phylogenomic analysis (Fig. 1) grouped these 14 genomic assemblies into five distinct phylum-level lineages based on the GTDB taxonomic scheme. Group 1 (Zodletone Spring Zgenome_0241 MAG, Zodletone Spring SCGC_AG-640-A22 SAG, and Walker Lake sediment SCGC_AG-301-P11 SAG) was monophyletic with candidate phylum AAMBM5-125-24, a phylum currently defined by MAGs from Aarhus Bay sediments and estuaries of White Oak River, North Carolina, and SAGs from the oxygen-minimum zones of the Northeastern Subarctic Pacific Ocean (Table 1). Group 2 (Zodletone Spring Zgenome_0002 MAG, Zodletone Spring Zgenome_0273 MAG, and CrabSpa hydrothermal vent SCGC AD-699-J03 SAG) was monophyletic with the phylum Calditrichota, a phylum currently defined by MAGs from Guyamas Basin sediment, Guyamas Basin hydrothermal vent, and Rifle aquifer sediment, as well as the pure-culture Cadithrix abyssi LF13 genome. Group 3 (Zodletone Spring Zgenome_0027 and Zodletone Spring Zgenome_0048 MAGs) was monophyletic with 6 MAGs belonging to candidate phylum KSB1 assembled from Guyamas Basin sediment (3 MAGs), Aarhus Bay sediments (1 MAG), Suncor tailing pond (Canada) (1 MAG), and Rifle aquifer sediment (Rifle, CO) (1 MAG) (Table 1). Group 4 (Lake Baikal SCGC AG-636-I10 and SCGC AG-636-N09 SAGs) was monophyletic with one MAG from estuaries of White Oak River, North Carolina, belonging to the candidate phylum SM23-31. It is worth noting that the phylum names utilized here are based on GTDB taxonomic outlines and that prior publications have often used one phylum name interchangeably, e.g., Calditrichaeota in reference 23 or KSB1 in references 24 and 25, as a broad umbrella to describe genomes from all four phyla. Interestingly, the fifth group encompassed 3 Zodletone Spring SAGs and 1 Zodletone Spring MAG (SCGC AG-640-J10 SAG, SCGC AG-640-B15 SAG, SCGC AG-640-I23 SAG, and Zgenome_0250 MAG). These four genomes were low- to medium-quality drafts (Table 1) with a placement suggesting that they belong to a novel, distinct sister phylum to AAMBM5-125-24, Calditrichota, KSB1, and SM23-31 (Fig. 1). This distinct phylum-level placement was corroborated by high intraphylum AAI (80.3% ± 25% [mean ± standard deviation]) and shared gene content (37.9% ± 12.3%) scores (Table 2) and low interphylum AAI (38% to 42%) and shared gene content (15% to 18%) scores (Table 3). Two-way intraphylum average nucleotide identities were also calculated for members of the fifth group (using alignment options of 700-bp minimum alignment length, a minimum of 50 alignments, and 70% minimum identity with a 1,000-bp window size and 200-bp step size). Values were obtained for SCGC AG-640-J10, SCGC AG-640-B15, and SCGC AG-640-I23 SAGs (99.99 ± 0.007%). However, due to the incompleteness of the genomes, values for Zgenome_0250 MAG in comparison to those of the three SAGs were below the detection level. Using the LSU ribosomal protein L3, three additional genotypes belonging to LCP-89 were identified in the unbinned contigs in the Zodletone Spring metagenomics assembly (Fig. S1).

One of the four genomic assemblies belonging to this novel candidate phylum described above (SCGC AG-640-I23 SAG) harbored a single nearly complete (1,536-bp) 16S rRNA gene. Comparative 16S rRNA gene-based phylogenetic analysis corroborated the distinct position of this novel phylum in relationship to representatives of the Calditrichota, SM32-31, AABM5-125-24, and KSB1 (Fig. 2). In addition, multiple (n = 24) environmental 16S rRNA gene sequences with high (90% to 94%) similarity to the 16S rRNA gene from SCGC AG-640-I23 SAG were identified in the SILVA database (release 132, queried in October 2018). These highly similar and monophyletic environmental sequences all belonged to the SILVA-defined candidate phylum LCP-89 and were reported in 15 different culture-independent studies, mainly in freshwater and marine environments (Table S3). It is worth noting that not all 16S rRNA gene sequences designated as members of the phylum LCP-89 in the SILVA database clustered with this novel lineage. Several clustered with candidate phylum AAMBM5-125-24 sequences, while others show little similarity to 16S rRNA of any sister phyla examined in this study (Calditrichota, AAMBM5-125-24, KSB1, SM23-31, and LCP-89).

Zodletone Spring LCP-89 organisms are predicted to be slow growers (iRep replication index of 1.38, indicating that at the time of sampling, about 40% of the cells belonging to this lineage were actively replicating, with one replication fork) and extremely rare (0.08% of the overall number of reads in the original metagenomic data set mapped to the representative Zgenome_0250 MAG). LCP-89 genomes recovered from Zodletone Spring possess various GC contents, ranging from 43% to 54.8%. Genome size estimates for Zodletone Spring LCP-89 predict medium-sized genomes (4.34 ± 0.62 Mb) with a few clustered regularly interspaced short palindromic repeat (CRISPR) sequences (0 to 2) identified per genome (Table 4).

We examined the salient structural features of LCP-89 genomes and compared these features to those identified in the genomes of all four sister phyla (Calditrichota, candidate phyla SM32-31, AABM5-125-24, and KSB1). LCP-89 cells are predicted to be Gram negative, based on the identification of several enzymes of lipid A and core oligosaccharide biosynthesis (Table 5), and rod shaped, based on the identification of the rod shape-determining proteins MreBCD and RodA. This Gram-negative rod-shaped morphology is similar in all genomes from sister phyla (Table 5) (26–28).

Interestingly, our analysis suggests an unusual cell wall composition within members of the LCP-89 phylum. With the exception of d-alanine–d-alanine ligase and two penicillin-binding proteins, all LCP-89 genomes analyzed lacked genes encoding peptidoglycan biosynthesis [e.g., UDP-N-acetylglucosamine 1-carboxyvinyltransferase (EC 2.5.1.7), UDP-N-acetylmuramate dehydrogenase (EC 1.3.1.98), UDP-N-acetylmuramate–alanine ligase (EC 6.3.2.8), UDP-N-acetylmuramoylalanine–d-glutamate ligase (EC 6.3.2.9), UDP-N-acetylmuramoyl-l-alanyl-d-glutamate–2,6-diaminopimelate ligase (EC 6.3.2.13), UDP-N-acetylmuramoyl-tripeptide–d-alanyl-d-alanine ligase (EC 6.3.2.10), phospho-N-acetylmuramoyl-pentapeptide transferase (EC 2.7.8.13), and UDP-N-acetylglucosamine–N-acetylmuramyl-(pentapeptide) pyrophosphoryl-undecaprenol N-acetylglucosamine transferase (EC 2.4.1.227), as well as membrane-bound lytic murein transglycosylase A and l,-d-transpeptidase]. Since FtsZ (the bacterial tubulin homolog) is essential for peptidoglycan remodeling during the septum formation process in cell division, we also queried the genomes of LCP-89 for FtsZ. FtsZ homologues were identified in only two LCP-89 genomes but were of an apparent archaeal origin and fused with a C-terminal COG0643 (chemotaxis protein histidine kinase CheA) domain (IMG gene numbers Ga0186948_10031 and Ga0186948_10305), casting doubt on their functionality. No pseudomurein biosynthesis genes were identified. However, two genes encoding S-layer homology domain-containing proteins (Pfam accession number PF00395) were identified, as well as genes encoding enzymes for CMP-legionaminate biosynthesis from UDP-N,N'-diacetylbacillosamine, an unusual alpha-keto sugar known to glycosylate extracellular structures in bacteria, e.g., Legionella and Campylobacter (29, 30), arguing for the possibility of an N-glycosylated S-layer in the cell walls of LCP-89 members. Interestingly, both S-layer homology domain-containing proteins in LCP-89 genomes were present upstream from a curli biogenesis system outer membrane secretion channel gene (csgG) homologue. CsgG in curli fiber-producing bacteria is implicated in the export of the protein components of the curli fiber, a thin aggregative cell surface fiber used for adhesion to surfaces (31). A possible function for the LCP-89 CsgG homologues in the export of the S-layer protein could therefore be hypothesized. However, S-layer protein export via type I secretion system, as reported for other S-layer-containing bacterial species (32, 33), could not be ruled out. The lack of peptidoglycan biosynthesis genes and the proposal of the presence of an N-glycosylated S-layer instead has previously been suggested in members of the Dehalococcoidia class of Chloroflexi (34–36), albeit members of Dehalococcoidia seem to lack an outer lipopolysaccharide (LPS) membrane. The lack of peptidoglycan biosynthesis machinery in LCP-89 genomes is in contrast to its presence in all Calditrichota, SM32-31, AABM5-125-24, and KSB1 genomes examined (Table 5 and Fig. 3). All sister phyla except AABM5-125-24 also encode S-layer homology domain-containing proteins (Table 5 and Fig. 3).

Additionally, although LCP-89 genomes possessed a nearly complete chemotaxis machinery (methyl-accepting chemotaxis protein, two-component system, chemotaxis family, sensor kinase CheA [EC 2.7.13.3], two-component system, chemotaxis family, response regulators CheB [EC 3.1.1.61] and CheY, chemotaxis protein CheD [EC 3.5.1.44], purine-binding chemotaxis protein CheW, chemotaxis protein methyltransferase CheR [EC 2.1.1.80], and chemotaxis proteins MotAB), they lacked the majority of genes for flagellar synthesis and assembly. This argues for the utilization of alternative types of motility, e.g., type IV pili (37), for which genes were identified in LCP-89 genomes (Table 5), as shown before for Myxococcus and Synechocystis spp. (38, 39). In comparison, flagellar synthesis and assembly genes were identified in the genomes of Calditrichota, SM23-31, KSB1, and AAMBM5-125-24.

Another interesting structural feature in LCP-89 genomes is their predicted capacity to synthesize bacterial microcompartments (BMCs), as suggested by the identification of homologues of the proteins with Pfam accession numbers PF03319 (EutN_CcmL) and PF00936 (BMC domain). BMCs are most probably utilized by members of LCP-89 and other sister phyla as protective shells to contain products of rhamnose or fucose metabolism (see metabolic characterization below). Such capacity to synthesize BMCs was also identified in all genomes of LCP-89’s four sister phyla. No evidences for encapsulin nanocompartment (Pfam accession number PF04454) (40) or magnetosome biogenesis (41) were identified in any of the genomes analyzed.

Genes encoding various catabolic and anabolic abilities identified in the LCP-89 genomic assemblies are presented in Fig. 4 and Table 5. LCP-89 genomic analysis revealed a heterotrophic lifestyle, with organic compounds acting as the sole sources of carbon, electrons, and energy. The genomes encoded an extensive sugar degradation machinery (Fig. 4, Table 5), enabling the channeling of a wide range of sugars (including glucose, mannose, fructose, and xylose) and sugar alcohols (including sorbitol and xylitol) to the organisms’ central glycolytic pathways. LCP-89 genomes encoded complete Embden-Meyerhof, pentose phosphate, and Entner-Doudoroff pathways for conversion of sugars to pyruvate (Fig. 4). In addition, LCP-89 genomes encoded a complete fucose and/or rhamnose degradation machinery that breaks down these sugars into propanol and propionate. Rhamnose and/or fucose degradation produces propionaldehyde as a toxic intermediate that needs to be sequestered in the organism’s microcompartment (42).

A complete pyruvate dehydrogenase enzyme complex and a tricarboxylic acid (TCA) cycle for pyruvate oxidation to CO₂ were identified in all LCP-89 genomes. However, the absence of functional elements of an aerobic respiratory chain (Fig. 4, Table 5) casts doubt on the use of oxygen as a possible electron acceptor. Nevertheless, the identification of nrfAH (cytochrome c nitrite reductase [NH3 forming] [EC 1.7.2.2]) suggests nitrite ammonification as a possible respiratory process in LCP-89 genomes, most probably coupled to lactate oxidation via d-lactate dehydrogenase (EC 1.2.1.4). No genes for nitrate reduction to nitrite were identified in the LCP-89 genomes.

In addition to their respiratory capacity, elements of pyruvate reduction to fermentative end products were identified in the genomes, suggesting fermentative capabilities. Predicted metabolic end products from sugar degradation include the short-chain fatty acids acetate, d-lactate, and propionate, based on the identification of genes encoding phosphate acetyltransferase and acetate kinase (EC 2.3.1.8 and EC 2.7.2.1), as well as d-lactate dehydrogenase (EC 1.1.1.28) and ethanol, propanol, butanediol, and acetoin, based on the identification of genes encoding alcohol dehydrogenase, acetolactate synthase (EC 2.2.1.6), acetolactate decarboxylase (EC 4.1.1.5), and meso-butanediol dehydrogenase/(S,S)-butanediol dehydrogenase/diacetyl reductase (EC 1.1.1.-, EC 1.1.1.76, and EC 1.1.1.304) enzymes.

Several metabolic distinctions were identified between members of LCP-89 and its sister phyla Calditrichota, AABM5-125-24, SM23-31, and KSB1 (Table 5). One important distinction is the variation in respiratory chain structure and putative electron acceptors. While LCP-89 genomes lacked evidence of a functional aerobic respiratory chain, all of the sister phyla encoded complexes I, II, and III and a variety of cytochrome oxidases or reductases with different affinities to O₂ (e.g., high-affinity cytochrome bd respiratory O₂ reductase, high-affinity cbb₃-type cytochrome c oxidase, and/or low-affinity aa₃-type cytochrome c oxidase). LCP-89 and AABM5-125-24 genomes contained nrfAH (cytochrome c nitrite reductase [NH3 forming] [EC 1.7.2.2]), which could possibly suggest respiratory nitrite ammonification, but lacked evidences for nitrate reduction to nitrite (no napAB or narGHIJ genes). Calditrichota appears to be capable of dissimilatory nitrate reduction to ammonium (DNRA). Such capacity is due to the possession of complete napAB and nirBD machinery for nitrate reduction to nitrite and nitrite reduction to ammonia (43). Indeed, pure cultures of Caldithrix abyssi were shown experimentally to use nitrate as an electron acceptor (28). Partial evidence of elemental sulfur/polysulfide reduction to sulfide occurs in the genomes of some members of LCP-89, SM23-31, and Calditrichota (43). One of the AABM5-125-24 genomes (SCGC AG-640-A22 SAG) encodes a full machinery for dissimilatory sulfate reduction to sulfide, a property not encountered in any of the other genomes analyzed.

LCP-89, Calditrichota, AABM5-125-24, SM23-31, and KSB1 genomes also differed in their oxygen detoxification mechanisms. A plethora of oxidative stress enzymes were encoded by LCP-89 genomes (including superoxide dismutase, superoxide reductase, rubrerythrin, and rubredoxin), the majority of which do not produce O₂ during their catalytic cycle (44), further attesting to the lack of aerobic capacities in LCP-89 organisms. On the other hand, genomes from all sister phyla encode some combination of catalase/peroxidase, both of which were missing from LCP-89 genomes (Table 5).

The levels of amino acids and cofactor auxotrophies also differed between genomes from different phyla. While genomic analysis of LCP-89, KSB1, SM23-31, and Calditrichota suggested 0 to 2 amino acid auxotrophies, genomes of AABM5-125-24 harbored the most auxotrophies (for 7 amino acids) (Table 5). In addition, genomes from different phyla encoded different substrate degradation capacities. Genomes of LCP-89, SM23-31, KSB1, and Calditrichota harbored a wide range of carbohydrate degradation capacities, including both sugar and sugar alcohols (Table 5). On the other hand, AABM5-125-24 genomes suggest a much narrower range of sugar catabolic capacities. Conversely, while LCP-89 genomes encoded amino acid degradation machineries for only 6 amino acids, genomes of all sister phyla encoded various degrees of amino acid degradation capabilities, ranging from 11 to 14 amino acids (Table 5).

We observed differences between LCP-89 and its sister phyla in the predicted products of fermentative metabolism. On one hand, LCP-89, SM23-31, Calditrichota, and KSB1 encoded enzymes suggestive of the production of various combinations of short-chain fatty acids and alcohols, including acetate, formate, l-lactate, d-lactate, propionate, ethanol, propanol, butanediol, and acetoin. On the other hand, genomic analysis of AABM5-125-24 suggested the production of acetate and ethanol as the only two fermentation end products.

This study provides an overview of the structural features and metabolic capacities of a yet-uncultured bacterial phylum previously identified in 16S rRNA data sets and for which no prior genomes have been described. Current thrusts for gauging global microbial diversity utilize either amplicon-based diversity surveys for faster, high-throughput community characterization (2, 45) or metagenomics/single-cell genomics approaches for more in-depth, genome-based predictions of organismal properties and characteristics (15). Obtaining genome representatives of the torrent of novel bacterial lineages identified in 16S rRNA gene diversity surveys represents an important step toward the understanding of the metabolic abilities and physiological preferences of yet-uncultured microbial lineages. Moreover, such efforts help to reconcile both taxonomic outlines and facilitate the development of a unified scheme for microbial taxonomy encompassing both approaches.

Multiple interesting features were identified in the analyzed genomes of LCP-89, some of which appear to be characteristic of closely related sister phyla Calditrichota, SM32-31, AABM5-125, and KSB1 (e.g., BMC possession), while others appear to be distinct characteristics representative of this phylum, e.g., respiratory nitrite ammonification and lack of peptidoglycan biosynthetic capabilities. The latter trait, coupled with the predicted possession of an outer membrane, an LPS layer, and an S-layer, is quite unique in the bacterial world. With the exception of the intracellular Mycoplasma genus, the lack of peptidoglycan appears to be an extremely rare trait within the domain Bacteria, although quite common in the Archaea. Recent reports have conclusively demonstrated the presence of peptidoglycan in the cell wall of members of the Planctomycetes and the Chlamydia, two phyla previously reported to have a peptidoglycanless cell wall structure (46, 47). It is worth noting that the cell wall structure reported here partly resembles those speculated for members of the Dehalococcoidia class of Chloroflexi (34–36), albeit Dehalococcoidia lack an outer LPS membrane. This commonality in two divergent phyla suggests gene loss through reductive evolution, which might be responsible for the observed lack of peptidoglycan in the bacterial world. The evolutionary and ecological drivers for this process remain to be discovered.

Finally, we acknowledge the fact that, as with most studies that investigate genomes of uncultured phyla, the SAGs and MAGs analyzed were incomplete. However, we stress that the majority of our analysis highlights features and suggested capabilities that are present rather than absent from the genomes. As such, it is possible that our analysis might underestimate the breadth of structural or metabolic capabilities of the phyla studied. Also, in instances where complete pathways were not detected, we believe that the analysis of several genomes belonging to each phylum (4 LCP-89 genomes, 3 SM23-31 genomes, 8 KSB1 genomes, 6 AABM5-125-24 genomes, and 7 Calditrichota genomes), rather than just one genome, strengthens the predicted absence of certain features or capabilities in the phyla studied.

Sediment samples were obtained from the source of Zodletone Spring, an anaerobic sulfide and sulfur-rich spring in southwestern Oklahoma (34.99562°N, 98.68895°W) as previously described (48).

Sediment DNA was extracted from the sample obtained in July 2015 using the DNeasy PowerSoil kit (Qiagen, Valencia, CA, USA). Sequencing of the sediment DNA was conducted using two lanes of the Illumina HiSeq 2500 system. A total of 281.0 Gbp of raw data were obtained from the single sediment sample. Low-quality reads were filtered using iu-merge-pairs (https://github.com/merenlab/illumina-utils/blob/master/scripts/iu-merge-pairs).

Details of the sequencing output and read quality control are provided in Table S1 in the supplemental material. Sequence reads that passed quality control were assembled and binned into individual genomes as previously described (48). Briefly, reads were assembled using MegaHit (49) with a minimum contig length of 1,000 and default parameters. Contigs were binned into metagenome-assembled genomes (MAGs) using MaxBin (50) with the default parameters. Assembly details for all MAGs and SAGs analyzed are provided in Table S2. To ensure that contigs in each MAG originated from a single population genome, the sequencing coverage and GC content of each contig were compared to the median values for the whole MAG. Contigs were removed from the MAG if their sequencing coverage or their GC contents were outside 5% of the median MAG value. Contigs were also compared to the GTDB database using BLASTX, and contigs with divergent phylogeny were removed. CheckM (51) was utilized for estimation of genome completeness, strain heterogeneity, and contamination based on the lineage-specific workflow. Briefly, genome bins are first placed into a reference genome tree, and then a file of lineage-specific marker sets is created for each genome. Marker genes are then identified and used to estimate the completeness and contamination of each genome bin. The marker set for all MAGs and SAGs analyzed here was k__Bacteria (UID2495), comprising 147 single-copy marker genes. Bins with >5% contamination were cleaned by removal of the outlier contigs identified, and the percent completeness and contamination were again rechecked using CheckM to ensure that the final genomic assemblies analyzed were of high quality.

Sediments collected in November 2013 were transferred to the laboratory, and amounts of 5 g were immediately suspended in 20 ml of sterile phosphate-buffered saline (PBS). Samples were vortexed for 30 s at 2,700 rpm and centrifuged for 30 s at 2,500 × g to remove large particles. Glycerol stocks of 20% PBS sample supernatant with 80% sterile glycerol were prepared, cryopreserved in liquid nitrogen, and shipped on dry ice to the Single Cell Genomics Center (SGSC) at Bigelow Laboratory for Ocean Sciences for processing as part of the Microbial Dark Matter MDM-II project, a wider effort for SAG generation and characterization from multiple global habitats (52) and follow-up study of the Genomic Encyclopaedia of Bacteria and Archaea-MDM project (7). Cells were sorted and lysed, whole-genome amplification was performed using WGA-X, and a preliminary identification of the SAGs obtained was performed by PCR-based 16S rRNA gene sequencing at the Bigelow Laboratory SCGC as previously described (53). Illumina library preparation (at SCGC), shotgun sequencing, and de novo genome assembly were performed as previously described (53).

Raw Illumina sequences were quality filtered using BBTools (54) according to SOP 1056, which removes reads with known contamination or low quality. Normalization was performed using BBNorm (54), and error correction was performed using Tadpole (54). The following steps were then performed for assembly: (i) artifact-filtered and normalized Illumina reads were assembled using SPAdes (version 3.9.0; ––phred–offset 33 –t 16 –m 120 ––sc –k 25,55,95 ––12) (55), and (ii) 200 bp was trimmed from all contig ends and contigs discarded if the length was <2 kbp or read coverage was less than 2 (BBMap: nodisk ambig, filterbycoverage.sh: mincov). Final SAG quality was defined based on the MISAG standards (22).

In addition to SAGs and MAGs originating from Zodletone Spring, single amplified genomes from a wider range of habitats were also generated and analyzed as part of this study. These include 2 SAGs from Lake Baikal, Irkutsk, Russia, 1 SAG from CrabSpa hydrothermal vent, East Pacific Rise, and 1 SAG from Walker Lake sediment, Nevada. The sampling and sequencing procedures were conducted as described above for Zodletone Spring samples. Although detailed analysis demonstrated that these 4 SAGs do not belong to the candidate phylum LCP-89, but instead are members of closely related sister phyla (see below), their inclusion greatly strengthened comparative genomic analysis, given the extreme paucity of genomic representatives in these sister phyla.

Genome-based phylogenomic analysis followed the taxonomic scheme of the Genome Taxonomy Database using GTDB-Tk (https://github.com/Ecogenomics/GtdbTk). In addition to the SAGs and MAGs mentioned above, multiple publicly available genomic representatives reported to belong to closely related phyla (Calditrichota, SM32-31, AABM5-125-24, and KSB1) were included in the analysis (Table 1). Phylogenetic placement was conducted using a concatenated alignment of 120 single-copy markers as previously described (56). Concatenated alignments were used to construct maximum-likelihood trees in RaxML (57). Alignment of 16S rRNA gene sequences was conducted using SINA aligner (58), and trees were constructed using FastTree (59). In addition to tree-based phylogenetic analysis, putative taxonomic ranks were also deduced using average amino acid identity (AAI; calculated using AAI calculator [http://enve-omics.ce.gatech.edu/]) and shared gene content (SGC; calculated using CompareM [https://github.com/dparks1134/CompareM]). Interlineage similarities were also confirmed by average nucleotide identity (ANI) calculation (http://enve-omics.ce.gatech.edu/).

The relative abundance of LCP-89 in Zodletone Spring sediment was deduced from the number of reads belonging to this lineage as a percentage of the total reads comprising the 281 Gbp of raw data obtained from Zodletone Spring sediments. Reads were mapped to the total metagenomic assembly using Bowtie2 (60). Coverage profiles were calculated for each contig in the LCP-89 genomic bin using the “coverage” command in CheckM (51), and these coverage profiles were then used to calculate the percentage of reads that mapped to the LCP-89 genomic bin using the “profile” command in CheckM. iRep (61) was used to predict the replication rate of the LCP-89 genome at the time of sampling. iRep calculates the ratio of sequencing coverage at the origin compared to sequencing coverage at the terminus of replication to measure replication rates. Since iRep calculates average coverage values using a sliding window of 5 Kbp, it does not require sequencing coverage of Ori and Ter sites, which makes it ideal for use with less-than-complete genomic assemblies. The percentages of cells replicating with one replication fork were predicted from the iRep index value as described in the document at https://github.com/christophertbrown/iRep/blob/master/iRepValues.pdf.

The IMG platform (http://img.jgi.doe.gov) was used for gene annotation, determination of general genomic features, and metabolic reconstruction (62). For instances where an absence of a specific gene was noted (e.g., peptidoglycan biosynthesis and respiratory complexes in LCP-89 genomes), this absence was confirmed by performing a tblastn search against all genomes using gene representatives from sister phyla. Detailed analysis of relevant pathways was performed using the KEGG database (63). Proteases, peptidases, and protease inhibitors were identified using BLASTP against the MEROPS database (64). Transporters were identified using the transporter classification database (TCDB) (65).

MAGs from this effort were deposited at DDBJ/ENA/GenBank under the Whole Genome Shotgun Bioproject accession number PRJNA498893, Biosample accession numbers SAMN10336777 to SAMN10336782, and WGS Project accession numbers RQOF01, RQOG01, RQOH01, RQOI01, RQOJ01, and RQOK01. SAGs from this effort are available from the IMG website (https://img.jgi.doe.gov/) under taxon identification numbers 3300015955, 3300016572, 3300016610, 3300016590, 2713897514, 3300016611, 2634166879, and 3300016634.