Isolation of genome-predicted Caldatribacterium (Atribacterota) reveals pervasive microbial cultivation problem due to folate precipitation

Abstract

Most bacterial phyla have few or no pure cultures, including Atribacterota, comprised of ubiquitous anaerobes. Here, we report genome-guided enrichment and isolation of two Atribacterota species representing a new family, Caldatribacterium saccharofermentans from a hot spring, and Caldatribacterium inferamans from a deep aquifer. Both were co-enriched with sulfate-reducing bacteria and initially resisted isolation, which we link to inadvertent removal of precipitated folic acid by filter-sterilization of unbuffered Wolin’s vitamin solution. We then predict folate auxotrophy across the Atribacterota and ~29% of all bacteria, with extensive auxotrophy in 27% of phyla. Since ≥604 of 791 ( ≥ 76%) media with folic acid additions in the MediaDive database use unbuffered vitamin solutions in which folic acid is likely removed during filter-sterilization, we propose that folate auxotrophy limits culturability in defined media en masse. We also uncover unusual features of Caldatribacterium, including three lipid membrane-like layers (LMLs), with the inner LML surrounding the nucleoid, and a high percentage of secreted proteins, supporting a unique cell biology of Atribacterota.

Most bacterial phyla have few or no representatives in pure culture. Here, Hedlund et al. isolate two new species within the phylum Atribacterota with unique cell biological features. In addition, they reveal a common mistake with media preparation leading to folate removal, which might explain some cultivation challenges.

Introduction

Despite ~150 years of cultivation efforts¹, most prokaryotic phyla have few or no representatives in pure culture². One such poorly cultivated phylum is Atribacterota, formerly OP9/JS1, which was identified as a key cultivation target due to its wide distribution and abundance in anoxic environments and proposed roles in carbon and hydrogen metabolism^3–7. Atribacterota are divided into two major classes that are physiologically and ecologically distinct. Candidatus Phoenicimicrobiia (JS1) is abundant in hydrocarbon-containing terrestrial and marine sediments, aquifers, and oil reservoirs, where they proliferate over thousands to millions of years of sedimentation^8–10 and catabolize alkanes^7,11 and other organic matter^9,11. However, no Ca. Phoenicimicrobiia isolates have been described. By comparison, Atribacteria (OP9) is more common in terrestrial environments, including hot springs, the subsurface, and bioreactors^12–14. In an early study, Ca. Caldatribacterium saccharofermentans was enriched on lignocellulose in Great Boiling Spring (GBS), NV, USA¹⁵ and predicted to ferment polysaccharides based on glycoside hydrolases, a core acetogenesis pathway, hydrogenases, alcohol dehydrogenase, and a Rhodobacter nitrogen fixation (RNF) complex^12,14. Since then, other Atribacteria were also predicted to be sugar⁸ or propionate fermenters¹⁴, or play a role in anaerobic acetate degradation⁷, and the reductive glycine pathway (RGP) was proposed as a near-universal core of Atribacterota metabolism enabling CO₂ or formate assimilation in some members⁷.

The first Atribacterota pure culture, Atribacter laminatus RT761^T, was isolated from a deep, methane-bearing aquifer¹⁶. It is a thermophilic sugar fermenter that requires yeast extract and produces H₂, acetate, carbon dioxide, and ethanol. Its growth is stimulated by the hydrogenotrophic methanogen Methanothermobacter thermoautotrophicus or H₂ removal from the headspace, indicating interspecies electron transfer via H₂. A. laminatus has an unusual cell ultrastructure, described as three lipid membrane-like layers (LMLs), with the innermost LML surrounding the nucleoid¹⁶. These layers were interpreted as an outer membrane (outer LML), cytoplasmic membrane (middle LML), and intracytoplasmic membrane (inner LML), though alternative interpretations were suggested^16,17. More recently, Jiao et al.⁷ and Kawamoto et al.¹⁸ isolated two additional Atribacteria, described as Thermatribacter velox B11^T and Atrimonas thermophila M15, from a terrestrial oil well and gas reservoir. However, they belong to the same species, and the name Thermatribacter velox has priority and is used here. Both strains ferment sugars and require yeast extract, producing H₂ and acetate, and have three LMLs like A. laminatus. Together, these three isolates represent two families of Atribacteria, with Ca. Caldatribacterium represents the third, yet-uncultivated family.

Here we describe enrichment and isolation of genome-predicted Caldatribacterium saccharofermentans from GBS and Caldatribacterium inferamans from a deep, hot, fractured-rock aquifer. Both were co-enriched with sulfate-reducing bacteria (SRB), yet they initially resisted isolation. We describe their isolation and attribute cultivation challenges to folate auxotrophy combined with inadvertent removal of precipitated folic acid (pteroyl-L-glutamic acid) by filter-sterilization of Wolin’s vitamin solution. We predict folate auxotrophy in ~29% of bacteria, including all Atribacterota, with extensive auxotrophy ( > 33% of genomes) in 27% of phyla, including both symbionts and free-living bacteria. Given the prevalence of predicted folate auxotrophy in bacteria, widespread pterin requirements in archaea, and problematic media preparation protocols (e.g., ≥604 media in MediaDive¹⁹), we believe folic acid precipitation is a significant problem for microbial cultivation at scale. We also show that Caldatribacterium cells contain three LMLs similar to A. laminatus and T. velox, but ribosomes are partially or fully separated from the nucleoid by the inner LML in Caldatribacterium, indicating an unusual cell biology in Atribacterota.

Results

Enrichment of two Caldatribacterium strains with distinct SRB

An in-situ corn stover enrichment from GBS (Supplementary Fig. 1) was transferred to a defined medium containing xyloglucan as the sole carbon source, incubated at 73 °C, and transferred biweekly. A stable community developed and was maintained for over two years, with Caldatribacterium comprising 8–11% of 16S rRNA gene reads, along with the fermenters Dictyoglomus, Fervidobacterium, and Pseudothermotoga, and the SRB Thermodesulfobacterium (phylum Desulfobacterota) (Fig. 1a). Shotgun metagenomics and binning into metagenome-assembled genomes (MAGs) as well as fluorescence in situ hybridization (FISH) yielded similar results (Supplementary Fig. 2).

Fig. 1. Co-enrichment of two Caldatribacterium species with distinct sulfate-reducing bacteria.

a Relative abundance of Caldatribacterium in a xyloglucan enrichment culture in 2015 and 2016 and a derived fucose enrichment culture based on 16S rRNA gene amplicons. GBS sediments were the inoculum. b Plot of all contigs from a shotgun metagenome derived from the same fucose enrichment culture. c Relative abundance of Caldatribacterium in borehole Inyo-BLM 1; a community enriched in situ on polyurethane foam plug; and a xylitol enrichment culture inoculated with sponge material. All data are based on 16S rRNA gene amplicons. d Plot of all contigs from a shotgun metagenome derived from the same xylitol enrichment culture. Color legends for taxa are coordinated for (a, b) and for (c, d)). Metagenomic binning was performed using MetaWatt.

To further enrich Ca. C. saccharofermentans, the xyloglucan culture was inoculated into media containing mono- or disaccharides as the sole carbon source. After two transfers, Caldatribacterium abundance increased (Supplementary Fig. 3), and fucose cultures were selected for dilution-to-extinction. In each dilution series, bottles receiving ≥18 cells showed growth, while those receiving ≤14 cells did not (Supplementary Fig. 4). After multiple rounds of dilution-to-extinction, Caldatribacterium dominated, and both 16S rRNA gene sequencing and shotgun metagenomics revealed a co-culture of Caldatribacterium and Thermodesulfobacterium, with >99.9% of 16S rRNA genes or metagenomic contigs assigned to these two genera (Fig. 1a, b). Additional dilution-to-extinction did not remove Thermodesulfobacterium, underscoring its importance for Caldatribacterium growth.

A distinct Caldatribacterium enrichment was obtained from borehole Inyo-BLM 1, which intersects the Lower Carbonate Aquifer near Death Valley National Park^20–22. Polyurethane foam plugs were suspended at 751 m depth for three months (Supplementary Fig. 5) and used to inoculate a defined medium²³ amended with xylitol (10 mM) and yeast extract (0.01% mass vol⁻¹). Bottles were incubated at 60 °C and transferred weekly. Although Caldatribacterium was not detected in situ, the plugs enriched for Caldatribacterium, and the xylitol enrichment further enriched Caldatribacterium to 97.6% of 16S rRNA genes (Fig. 1c). The most abundant non-Caldatribacterium 16S rRNA gene was the SRB Thermodesulfovibrio (1.1%), and a shotgun metagenome confirmed the dominance of Caldatribacterium and Thermodesulfovibrio in the culture (Fig. 1d). Thermodesulfovibrio remained in low abundance after multiple rounds of dilution-to-extinction in defined medium with xylitol as the sole carbon and energy source, suggesting the Inyo-BLM 1 Caldatribacterium strain also depended on SRB, albeit one from a distinct phylum (Nitrospirota).

Attempts to isolate either Caldatribacterium strain on solid defined medium with fucose or xylitol as the sole carbon source were unsuccessful, yet pure cultures of the SRB were readily obtained. The Thermodesulfobacterium strain grew chemolithoautotrophically with H₂ as the electron donor or heterotrophically using lactate (Supplementary Note 1). It represents a new species herein proposed as Thermodesulfobacterium auxiliatoris TA1^T sp. nov. for its ability to help Caldatribacterium grow (Table 1; Supplementary Data 1). The Thermodesulfovibrio strain was identified as Thermodesulfovibrio yellowstonii, herein strain SIUC14, and it grew using lactate or H₂ and acetate (Supplementary Note 2). Neither SRB grew on its own under the conditions of the Caldatribacterium enrichment cultures nor with other sugars.

Table 1.

Taxon names proposed under the ICNP

Proposed taxon	Etymology	Description
Genus Caldatribacterium	Cald.atri.bac.te’ri.um. L. adj. kal -di, hot; L. adj. ater -tra -trum, black; L. neut. n. bacterium, rod or staff. N.L. neut. n., Caldatribacterium, ‘black’ or ‘dark’ references both ‘microbial dark matter,’ referring to the poor culturability of the genus and phylum, and the dark, anoxic environments where it is found.	Cells are rod-shaped or ovoid cells with pointed ends. Cells stain Gram-variable, but genomic data and cryo-EM reveal three lipid membrane-like structures, including an outer membrane and a nucleoid that appears to be membrane-bound. No endospore formation is observed. Thermophilic. Obligately anaerobic chemoorganoheterotroph. Fermentation end products are acetate and H2. The major fatty acids ( > 10%) were C13:0 iso 3OH, C15:0 iso, C16:0, and C18:0. The type species is Caldatribacterium saccharofermentansT.
Species Caldatribacterium saccharofermentans	sac.cha.ro.fer.men′tans. Gr. n. sakchâr sugar; L. part. adj. fermentans fermenting; N.L. neut. n. saccharofermentans sugar-fermenting). Caldatribacterium saccharofermentans. The mysterious heat-loving sugar-fermenting rod.	Cells are rod-shaped or ovoid cells with somewhat pointed ends 1.5–3.0 µm long and 0.5–0.9 µm wide, occurring singly or in pairs. Flagellar motility is rarely observed, but flagella are observed by cryo-EM. Flagellar biosynthesis gene clusters are complete. Optimal growth temperature is 70 °C, with a maximum of 77.5 °C and a minimum of 60 °C. Optimal pH is 7.0 with a range of 5.5–9.0. Obligately anaerobic chemoorganoheterotroph that ferments sugars and sugar alcohols as sole carbon sources and fermentation substrates: adonitol, cellobiose, fructose, fucose, galactose, gluconate, glucose, inositol, lyxose, maltose, mannose, N-acetylglucosamine, raffinose, rhamnose, ribose, sorbitol, sucrose, tagatose, and xylitol. Fermentation end products are acetate and H2. The major fatty acids ( > 10%) were C13:0 iso 3OH, C16:0, and C18:0. The type strain is GBS T (DSM 110181 T and JCM C190371T), which was isolated from Great Boiling Spring, NV., USA.
Species Caldatribacterium inferamans	in.fer’a.mans. L. pl. n. inferi, the underworld; L. pres. part. amans, loving; N.L. part. adj. inferamans, loving the underworld, referring to the origin of the strain). Caldatribacterium inferamans, The mysterious heat-lover from the underworld” or “The mysterious heat-loving rod that loves the hot underworld.	Cells are rod or ovoid cells with somewhat pointed ends 1.0–2.6 µm long and 0.5–0.8 µm wide, typically occurring singly. Flagellar motility rarely observed, but flagella are observed by cryo-EM. Flagellar biosynthesis gene clusters complete. Optimal growth temperature is 60–70 °C, with a maximum of 75 °C and minimum of 55 °C. Optimal pH is 7.4–8.7 with a range of 5.9–9.5. Obligately anaerobic chemoorganoheterotroph that ferments sugars, sugar alcohols, and proteins as a sole carbon source and electron donor/acceptor for growth: arabinose, cellobiose, fructose, fucose, fumarate, galactose, gluconate, glucose, lactose, maltose, mannitol, mannose, peptone, raffinose, sorbitol, starch, succinate, sucrose, trehalose, xylose, xylitol, xylan, and yeast extract. Fermentation end product is acetate and H2. The estimated genome size of SIUC1T is 1,967,957 bp, with 56.5 mol% DNA G + C content. The major fatty acids ( > 10%) were C13:0 iso 3OH, C15:0 iso, C16:0, and C18:0. The type strain is SIUC1T (DSM 110249 T and JCM 39079 T), which was isolated from the Deep Carbonate Aquifer located in the southern hydrographic Great Basin, NV, USA.
Family Caldatribacteriaceae	Cald.atri.bac.te.ri.a’ce.ae. L. adj. kal -di, hot; L. adj. ater -tra -trum, black; L. neut. n. bacterium, rod or staff. N.L. masc. n. Caldatribacterium type genus of the family; suff. -aceae, ending to denote a family; N.L. fem. pl. n. Caldatribacteriaceae, the family of the genus Caldatribacterium.	Thermophilic or hyperthermophilic inhabitants of freshwater thermal environments. Strictly anaerobic sugar fermenters. The family is a distinct phylogenetic lineage in the order Atribacterales, class Atribacteria, and phylum Atribacterota. The type genus is Caldatribacterium.
Species Thermodesulfobacterium auxiliatoris	aux.il.i.a.tor’is. L. gen. masc. n. auxiliatoris, one who gives aid). Thermodesulfobacterium auxiliatoris, the thermophilic sulfate-reducing rod that provides aid, based on its ability to facilitate the growth of Caldatribacterium in co-culture.	Cells are regular rods averaging 1.5 μm long and 0.5 μm wide, usually occurring singly. Diderm cell envelope structure. Optimal growth temperature is 70–75 °C with a maximum of 80 °C and a minimum of 50 °C. Optimal pH is 7.0 with a range of 5.5–8.0. Facultative chemolithoautotroph capable of using H2 as the electron donor and CO2 and the carbon source, with sulfate as the terminal electron acceptor. Acetate stimulates growth under autotrophic conditions, and formate and lactate can serve as electron donors and carbon sources for heterotrophic growth. The type strain is TA1T (DSM 107971 T and JCM 32932 T), isolated from Great Boiling Spring, NV, USA.

Caldatribacterium in GBS enrichments assimilates sugars and amino acids

In the absence of Caldatribacterium pure cultures, we used nanometer-scale secondary-ion mass spectrometry (nanoSIMS) coupled with FISH to assess incorporation of ¹³C-labeled substrates (Fig. 2). In xyloglucan enrichments, most cells assimilated ¹³C atoms from labeled glucose, xylose, ribose, and amino acids, yet only ¹³C xylose incorporation was higher in Caldatribacterium, suggesting direct uptake. In fucose enrichments, uptake of glucose, xylose, and ribose was higher in Caldatribacterium compared to Thermodesulfobacterium, again suggesting direct uptake, yet acetate and amino acid assimilation were higher in Thermodesulfobacterium. In both cultures, little or no uptake of ¹³C-bicarbonate, formate, or acetate was detected in Caldatribacterium (atom percent excess <0.5%). These taxon-specific carbon uptake patterns indicate metabolic interactions between Caldatribacterium and Thermodesulfobacterium in the GBS enrichments.

Fig. 2. ¹³C incorporation of sugars by Caldatribacterium in enrichment cultures by FISH-nanoSIMS.

a Isotopic enrichments in Caldatribacterium saccharofermentans cells identified by FISH (orange bars) and other cells (gray bars) in a xyloglucan enrichment culture. b Parallel experiment with a fucose co-culture containing Caldatribacterium (orange bars) and T. auxiliatoris (blue bars). Images shown are representative of ~20 Caldatribacterium cells imaged after labeling with xylose for 2 hours. Each point reflects the percent of ¹³C atom fraction excess in a single cell following stable isotope probing with ¹³C-labeled substrates. Boxes represent 25th and 75th percentiles and central mark is the median. Vertical black lines show standard deviations. Red asterisks show significant differences versus unlabeled controls (two-sided Wilcoxon rank-sum, *p < 0.05). Brackets with asterisks show significant differences between Caldatribacterium saccharofermentans cells and other cells (two-sided Wilcoxon rank-sum, *p < 0.05). The numbers (n) of cells analyzed for a single biological replicate for Caldatribacterium (left) and other cells (right) are indicated after the substrates along the x axis. Inset, FISH showing Caldatribacterium cells (left, small rod with strong labeling) and nanoSIMS ion ratio images reflecting ¹³C assimilation in a Caldatribacterium cell (right).

Isolation of Caldatribacterium strains

Parallel co-enrichment of the Caldatribacterium cultures with unrelated hydrogenotrophic SRB initially suggested interspecies electron transfer via H₂, similar to A. laminatus RT761^T and Methanothermobacter thermoautotrophicus¹⁶. Consortial fermentation of sugars was also supported by nanoSIMS results. For example, the uptake of acetate by Thermodesulfobacterium indicates Caldatribacterium and Thermodesulfobacterium co-mineralize sugars, with both acetate and H₂ serving as metabolic handoffs.

However, primary fermentations of sugars producing acetate and H₂ are highly exergonic and should not require co-cultivation with a hydrogenotroph. Thus, we considered an alternative hypothesis that Caldatribacterium strains rely on SRB for resources, since both co-cultures were maintained in defined media. To address this alternative and try to obtain pure cultures, dilution-to-extinction was repeated with GBS fucose cultures, and the lowest dilution without growth was inoculated into media with fucose alone or with fucose plus yeast extract and casamino acids (0.1% mass vol⁻¹ each). While cultures with fucose alone showed no growth, those supplemented with yeast extract and casamino acids grew after four days. Subsequently, solid media with yeast extract and casamino acids, in addition to fucose or fucose and xylitol, enabled the isolation of both Caldatribacterium strains via repeated colony picking and streak plating. Culture purity was confirmed by FISH (Supplementary Fig. 6), genome sequencing, and plate counts. Both Caldatribacterium pure cultures grew in liquid medium supplemented with yeast extract, casamino acids, and either fucose or xylitol, and subsequent experiments determined that yeast extract alone plus fucose or xylitol was necessary and sufficient for growth (Supplementary Fig. 7).

The isolated Caldatribacterium strains, herein named Caldatribacterium saccharofermentans GBS^T gen. nov. and sp. nov. and Caldatribacterium inferamans SIUC1^T sp. nov. (Table 1), fermented sugars and sugar alcohols, producing acetate and H₂. Both strains grew at circumneutral pH at high temperatures (Supplementary Fig. 8), with GBS^T growing optimally at 70 °C and pH 7 and SIUC1^T growing optimally at 60–70 °C and pH 7.4–8.7. Additional phenotypic data and comparisons between Atribacterota isolates are presented in Supplementary Data 2–4 (see Supplementary Note 3).

Complete, circular genomes were 2,255,665 bp for C. saccharofermentans and 1,967,957 bp for C. inferamans. Phylogenetic analysis of GTDB representative genomes (09-RS220) plus Atribacteria isolate genomes confirmed they form a monophyletic clade with the Ca. C. saccharofermentans MAG and Ca. C. californiense single-amplified genome (SAG)¹², along with MAGs from the Shengli Oilfield and hot springs in Tengchong, China; Tibet; Sungkai, Malaysia; and Yellowstone, USA (Fig. 3; Supplementary Data 5). Based on relative evolutionary distance and average amino acid identity (AAI) ≥ 60%²⁴, this clade is the genus Caldatribacterium. A larger monophyletic clade including MAGs from the deep subsurface in Russia and South Africa and anaerobic digesters globally represents Caldatribacteriaceae fam. nov. (see Supplementary Note 4 and Supplementary Data 5). Marker proteins compiled in GTDB (10-RS226) using Sandpiper²⁵ confirmed the primary habitats ( > 90% of marker proteins) of Caldatribacterium to be terrestrial hot springs and terrestrial subsurface and of Caldatribacteriaceae to be terrestrial hot springs, the subsurface (including oil fields), and various anaerobic waste treatment systems (e.g., anaerobic digestors, biogas fermenters, landfills, sludge, wastewater), both with global distributions.

Fig. 3. Phylogeny and genomic relatedness in the Caldatribacteriaceae.

a A maximum-likelihood phylogeny of GTDB species-representative genomes (09-RS220) plus isolate genomes (Supplementary Data 5) based on the concatenated alignment of 120 conserved bacterial marker proteins (bac120) using the optimal evolutionary model (LG + F + I + R3) and rooted at the midpoint. Filled circles represent supported nodes based on ≥80% SH-aLRT support and >95% support from ultrafast bootstrapping (1000 pseudoreplicates). The two new strains are indicated in color and bold. b Average nucleotide identity (ANI) and average amino acid identity (AAI) among Caldatribacteriaceae genomes. The same numbering and color scheme is used. See Supplementary Note 4 for additional discussion about taxonomy.

Using our Caldatribacterium isolates, we confirmed growth on defined media with fucose but only in co-culture with the SRB T. auxiliatoris TA1^T, T. yellowstonii SIUC14, Thermodesulfobacterium commune DSM 2178 ^T, or Thermodesulfobacterium hveragerdense DSM 12571 ^T (Fig. 4A). Under these conditions, Caldatribacterium dominated (Supplementary Fig. 9). Yet, both Caldatribacterium pure cultures were only weakly inhibited by fermentation products, including H₂ (Supplementary Fig. 10), contrasting with the strong H₂-based inhibition of A. laminatus¹⁶ (see Supplementary Note 5). Thus, semi-syntrophic growth of Caldatribacterium with hydrogenotrophic SRB was not driven by interspecies H₂ transfer or the removal of acetate or other metabolic products.

Fig. 4. Semi-syntrophic growth of Caldatribacterium is due to folate dependency.

A Neither Caldatribacterium pure culture grew with filter-sterilized Wolin’s vitamins (Filt. Vit.), but growth could be restored by adding yeast extract (YE, 0.01% mass/vol) or by co-cultivation with a variety of SRB. All experiments contained filter-sterilized Wolin’s vitamins. B Caldatribacterium pure cultures grew with autoclaved Wolin’s vitamins but not filter-sterilized Wolin’s vitamins. Growth could be restored by adding folic acid (strain GBS^T and SIUC1^T) or biotin or riboflavin (strain SIUC1^T). C Growth of Caldatribacterium pure cultures could be restored by adding filtered and autoclaved supernatant (supt.) from SRB cultures to GBS salts medium (1:3 vol vol⁻¹) but not SRB medium itself (SRB med.). SRB abbreviations: T.hv, Thermodesulfobacterium hveragerdense DSM 12571 ^T; T.com, Thermodesulfobacterium commune DSM 2178 ^T; T. aux, Thermodesulfobacterium auxiliatoris; and Tvib, Thermodesulfovibrio yellowstonii. For all panels, bars show the mean and standard deviation. Bars with different letters are significantly different (p < 0.05) based on ANOVA and Tukey’s post-hoc tests. n = 3 for biological replicates. Source data including exact p values are available as a Source Data file.

Folic acid permits Caldatribacteriumgrowth in defined media

Since semi-syntrophic growth was not driven by interspecies H₂ transfer, we hypothesized that one or more vitamins required by Caldatribacterium might be provided by the SRB or yeast extract. We previously predicted Caldatribacterium to be auxotrophic for folate, riboflavin, and biotin¹², yet these vitamins were added to our defined medium using standard protocols. Specifically, we used Wolin’s vitamin solution (1x)^26,27 that was prepared in oxygenated >18 MΩ water, sparged with N₂, filter-sterilized, and added to the base medium after autoclaving. However, we observed fine precipitates in the vitamin solution that were removed during filter-sterilization.

To test whether filter-sterilization inadvertently removed one or more insoluble vitamin, we autoclaved unfiltered Wolin’s vitamin solution (1x) prepared as above and added it to media prior to inoculation, and it restored growth (Fig. 4B). Since autoclaving did not destroy the necessary vitamin(s), we tested single autoclaved vitamins in combination with filter-sterilized Wolin’s vitamin solution (1x). Although autoclaved biotin or riboflavin failed to enable growth of C. saccharofermentans and enabled only poor growth of C. inferamans, autoclaved folic acid fully restored growth of C. saccharofermentans and C. inferamans (Fig. 4B). Additional experiments with C. saccharofermentans determined that folic acid, riboflavin, and pantothenate were necessary and sufficient for growth on fucose (Supplementary Fig. 7) and that ≥0.5 ng mL⁻¹ folic acid supported growth, although ≤20 ng mL⁻¹ resulted in slow growth (Supplementary Fig. 11).

Folic acid is nearly insoluble at neutral pH and insoluble below pH ~6^28,29. Our Wolin’s vitamin solutions (1x and 10x) prepared with ≥18 MΩ water in two different labs in different U.S. states, were pH 3.7–3.9, explaining the precipitates. We note that Wolin’s vitamin solution is also called Wolfe’s vitamin solution or many other names with many minor variations in MediaDive and in the primary literature (Supplementary Data 6-7). Additionally, Vitamin Supplement MD-VS™, sold by the American Type Culture Collection (ATCC), is phosphate-buffered but otherwise identical to Wolin’s vitamin solution (1x). We measured a pH of 6.0 in Vitamin Supplement MD-VS™. We note that only a few vitamin solutions in MediaDive are pH-adjusted, accounting for ~4.2% (33/791) of media with folic acid additions. The pH of these vitamin solutions is important to consider, but >99.99% of media in MediaDive use vitamin solutions that are oversaturated with folic acid even at pH 7 (777/791 media), and some are >1000-fold oversaturated (65/791 media; ~8.2%) (Supplementary Data 6-7); thus, even if these vitamin solutions are neutralized in some labs, some folic acid precipitation is very likely.

To directly test the hypothesis that folic acid was removed during media preparation by filter-sterilization, we added unbuffered filter-sterilized Wolin’s vitamin solution (1x) to GBS salts medium^19,30, and then quantified folic acid and other vitamins by liquid chromatography coupled with mass spectrometry (LC-MS) (Supplementary Data 8). This showed that folic acid, but not other vitamins, was removed by filter-sterilization. In contrast, when unfiltered Wolin’s vitamin solution (1x) was autoclaved and added to media after autoclaving (with no filter-sterilization step), soluble folic acid was low but measurable. We speculate that the low folic acid concentration was due to folic acid precipitation in the medium, although this required additional experimentation (see below).

To explore the possibility that SRB secrete folates required by Caldatribacterium, filter-sterilized SRB culture supernatants were added to GBS medium in the presence of filtered vitamins. SRB supernatants, but not uninoculated SRB medium, supported growth of both strains (Fig. 4C). LC-MS of SRB supernatants did not detect folic acid in the supernatants (Supplementary Data 8), but physiologically relevant forms of folate^31,32 were not targeted in our analysis.

Since these experiments did not fully resolve the behavior of folic acid in the vitamin solutions themselves, we conducted additional experiments using the most-used vitamin solution prescribed in MediaDive, Solution 5980 (Wolin’s vitamin solution (10x)), while varying pH/buffering, the N₂-sparging protocol, and the sterilization method. We analyzed soluble folic acid concentrations in each of 12 vitamin solutions by LC-MS and their ability to support growth of C. saccharofermentans along with folic-acid-free negative controls, autoclaved folic acid spike-in positive controls, and vitamin solutions that were not sterilized (Fig. 5). We also took photos to document visible precipitates in the solutions (Supplementary Fig. 12). These experiments showed that folic acid was removed to below LC-MS detection limits ( < 0.082 mg L⁻¹) by filter-sterilization of the unbuffered Wolin’s Vitamin Solution (10x), which severely limited growth of C. saccharofermentans, consistent with visible precipitates in the unbuffered vitamin solution that were collected on filters during filter-sterilization. All other vitamin preparations supported strong C. saccharofermentans growth, with only minor delays attributed to the sterilization method or whether the vitamin solution was prepared in anoxic water under a stream of N₂ or whether it was prepared in oxic water and then sparged with N₂ after vitamin addition. LC-MS data were consistent with the growth experiments, but were more difficult to interpret due to our inability to detect insoluble folic acid. Yet the LC-MS experiments with the vitamin solutions confirmed that autoclaving did not have a major effect on folic acid. In fact, autoclaving promoted LC-MS-detection of folic acid by increasing its solubility, but folic acid eventually precipitated during storage at 4 °C. A similar analysis of newly purchased ATCC Vitamin Supplement MD-VS™ showed a severe delay of C. saccharofermentans growth and an  ~ 75% lower concentration of folic acid in LC-MS experiments than expected, indicating a problem during preparation or storage impacting folic acid. The low concentration of folic acid in ATCC Vitamin Supplement MD-VS™ may be due to the relatively low pH of the solution (6.0) combined with filter-sterilization at the ATCC.

Fig. 5. Folic acid is best preserved in buffered and autoclaved vitamin solutions.

A Interaction plot of observed folic acid in a 1:20 dilution of Wolin’s vitamin solution (10×) (ppm) measured by LC/MS, under different buffering, sparging, and sterilization conditions. Differences among treatments were assessed using ANOVA on the full factorial design (buffering, sterilization, and sparging), followed by Tukey’s HSD for multiple comparisons. The inset shows the proportion of variance explained by each main factor based on sums of squares. B Growth of Caldatribacterium saccharofermentans GBS^T in media supplemented with the corresponding vitamin solutions as measured by a Petroff-Hausser counting chamber. Line color represents buffering, line style represents sterilization method, and marker shape indicates sparging treatment. Statistical comparisons were performed using Tukey’s HSD on the full factorial design (buffering, sterilization, sparging, and time) after fitting to a mixed linear model. Letters indicate selected Tukey’s HSD comparisons at Day 14; the bracketed group “a” comprises conditions that were not significantly different (P < 0.05), while “b–d” denote significant contrasts relative to this group. Symbols represent means. Error bars (A) and shaded regions (B) indicating ± s.d for replicates (n = 3) are shown on both plots. A, B data represent technical replicates and biological replicates, respectively. C Wolin’s vitamin solution (10×) prepared with N₂ sparged H₂O and mixed under a stream of N₂. D Wolin’s vitamin solution (10×) prepared with N₂ sparged H₂O buffered to pH 7 and mixed under a stream of N₂. Aliquots of both preparations are pictured not sterilized, filter-sterilized, and autoclaved. Opaque precipitate is seen in the unbuffered/not sterilized solution, but removed in the filter-sterilized solution (C) while no visible precipitate is seen in any buffered solution (D). Source data including exact p values are available as a Source Data file.

Folate auxotrophy and salvage are conserved in Atribacterota

To further assess the implications of inadvertent folic acid removal from vitamin solutions, we examined folate biosynthesis, salvage, and folate-dependent enzymes in the Caldatribacterium genomes and all other high-quality Atribacterota genomes (Fig. 6; Supplementary Data 9). Folates are universal one-carbon donors and acceptors that can be synthesized de novo from GTP, chorismate, and glutamate by many bacteria via a pathway consisting of a pterin branch, a para-aminobenzoic acid (pABA) branch, and C1 ligations, C1 reductions, and glutamylation³³. All Atribacterota genomes encode methionyl-tRNA formyltransferase and other folate-dependent enzymes, indicating folate dependency for translation, purine synthesis, S-adenosylmethionine synthesis, amino acid synthesis and catabolism, and iron-sulfur cluster maturation (Fig. 6B). Folate would also be required for CO₂ or formate fixation via the RGP, but Caldatribacterium genomes lack formate dehydrogenase, aminomethyltransferase, and glycine dehydrogenase, and therefore do not encode the RGP⁷ (Supplementary Data 9).

Fig. 6. Folate metabolism in Atribacterota.

A Canonical folate biosynthetic pathway highlighting the pterin branch, pABA branch, glutamylation/reduction steps, folate salvage, and folate ligation, C1 reduction, and polyglutamylation steps. B Folate-dependent biochemistry (grey) and reduction steps (yellow). A, B enzymes in black are encoded by the Caldatribacterium isolates, whereas those in red are absent. C Structure of dihydrofolate, with parts of the molecule colored according to branches of the biosynthetic pathway in (A). D The largest gene neighborhood encoding folate-related genes. HDR, heterodisulfide reductase. E The presence (filled circle) or absence (open circle) of folate-related genes in all Atribacterota species with a high-quality genome, along with SRB used in experiments; genes are color-coded by role in pathway. Presence/absence was determined using the COGs for each folate-related gene. FolK/FolP are color-coded purple as the absence of either gene results in predicted folate auxotrophy.

However, despite folate dependency, all Atribacterota genomes lack any known genes for pterin biosynthesis (Fig. 6A), including hydroxymethyldihydropterin pyrophosphokinase (HPPK, FolK), a key pterin branch enzyme, and dihydropteroate synthase (DHPS, FolP), which condenses 6-hydroxymethyl-7,8-dihydropterin pyrophosphate from the pterin branch and pABA. The absence of HPPK, DHPS, and other pterin branch enzymes also suggests pABA biosynthesis or salvage would not satisfy folate needs, consistent with the lack of Caldatribacterium growth on defined media with filter-sterilized Wolin’s vitamin solution, which contains pABA. pABA is highly soluble in water and would not be removed by filter-sterilization³⁴.

Intact folate salvage has been partially characterized in some bacteria³⁵ and is the basis of its inclusion in media vitamin solutions. We found that energy-coupling factor (ECF) transporters with substrate-binding domains for folate (EcfS) were encoded in almost all Atribacteria genomes, with the remaining Atribacteria and nearly all Ca. Phoenicimicrobiia genomes encoding folate-biopterin transporters (Fig. 6E). Consistent with intact folate scavenging, most Atribacterota genomes encode enzymes necessary for reduction, C1 ligation, C1 reduction, and polyglutamylation: dihydropterin reductase (FolM), formyltetrahydrofolate synthase (Fhs), bifunctional methylenetetrahydrofolate dehydrogenase/methenyltetrahydrofolate cyclohydrolase (FolD), folylpolyglutamyl synthase (FPGS, FolC), and 5,10-methylenetetrahydrofolate reductase (MetF) (Fig. 6A, D, E). Several of these genes are clustered on Caldatribacterium genomes, but we could not identify candidate alternative enzymes for HPPK, DHPS, and other missing pterin biosynthetic enzymes within these gene clusters. In contrast, SRB that enabled Caldatribacterium growth encoded complete de novo folate synthesis pathways (Fig. 6E).

We also used computed structure models (CSMs) and molecular docking to compare the C. saccharofermentans FolC to the bifunctional Escherichia coli FolC and the monofunctional Lactobacillus casei FolC, which only encodes FPGS activity (Supplementary Fig. 13). The CSM was similar to both enzymes, including distinct N- and C-terminal domains and a conserved ATP-binding site in the interdomain cleft; however, the C. saccharofermentans FolC resembled the L. casei enzyme in the dihydropteroate binding region that differentiates dihydropteroate from 5,10-methylene-THF (E. coli FolC residues 28–32), suggesting a monofunctional role for the C. saccharofermentans FolC during folate salvage.

Predicted folate auxotrophy is pervasive in bacteria

To assess folate biosynthesis in other bacteria, we searched for an extensive list of folate-related Clusters of Orthologous Genes (COGs) within the Web of Life 2 (WoL2) reference genome collection and required FolK (COG0801) and FolP (COG0294) to predict pterin or folate prototrophy. This revealed ~29% predicted auxotrophs out of 15,953 total genomes (Supplementary Data 10). Although some of these may be false hits due to genome incompleteness or novel enzymes, 83% of these predicted auxotrophs lack both FolK and FolP, along with additional enzymes, making them strong candidates for true auxotrophy.

When we expanded our analysis of rare phyla by adding all GTDB reference genomes for phyla with <50 WoL2 genomes (13,326 additional genomes), our analysis revealed 46 phyla (27% of GTDB phyla) in which ≥33% of genomes are predicted folate auxotrophs (i.e., “extensive folate auxotrophy”) (Fig. 7; Supplementary Data 10−11). Several of these phyla are notoriously difficult to cultivate, such as Atribacterota, Caldisericota, Chloroflexota, Gemmatimonadota, Spirochaetota, and Thermotogota, and additional phyla with extensive predicted folate auxotrophy have no isolates, including Ca. Aerophobota, Bipolaricaulota, Muiribacteriota, Sysuimicrobiota, and many others. Others include symbionts with known genome reductions and host dependencies, such as Babelota, Bdellovibrionota, Chlamydiota, and Patescibacterota. These symbionts likely obtain folates from their hosts. Bacillota was also on this list, with ~46% of WoL2 Bacillota genomes predicted to be folate auxotrophs. Many of these Bacillota represent orders with host dependencies, such as cell-wall-less Acholeplasmatales, Izemoplasmatales, Mycoplasmatales, RF39, and RFN20, but others are not known symbionts, such as Caldicoprobacterales, Christensenellales, Clostridiales, DTU010, DTU014, Erysipelotrichales, Halanaerobiales, Lachnospirales, Lactobacillales, Oscillospirales, Natranaerobiales, Peptostreptococcales, Proteinivoracales, Saccharofermentanales, Selenomonadales, TANB77, and Tissierellales.

Fig. 7. Bacterial phyla with extensive predicted folate auxotrophy.

Pie plots of bacterial phyla where ≥33% of all genomes were missing either FolK (required for pterin prototrophy), FolP (required for folate prototrophy), or both FolK and FolP. Presence/absence for FolK and FolP was determined by their respective COGs (COG0801 and COG0294). Each phylum is named per GTDB (10-RS226), and the number of high-quality genomes analyzed follows in parentheses. 30 phyla are shown; 16 phyla with extensive auxotrophy were excluded from the figure due to having ≤5 high-quality genomes available in GTDB.

Including all phyla, we predict extensive folate auxotrophy in many “most-wanted” taxa, including seven of the 10 most abundant orders in the human gut microbiome³⁶ - Lactobacillales, Lachnospirales, Oscillospirales, Peptostreptococcales, Erysipelotrichales, Bifidobacteriales, and Coriobacteriales, consistent with a previous report of predicted folate auxotrophy in 87% of Human Microbiome Project (HMP) genomes and pterin auxotrophy with pABA prototrophy predicted in 61% of the remaining HMP genomes³⁷. Folate auxotrophy was also predicted in “most-wanted” groups in other biomes. For example, among abundant marine bacterioplankton, we predict extensive folate auxotrophy in HIMB59³⁸, SAR86³⁹, Pelagibacterales, and UBA11654⁴⁰. Given that the same vitamin solutions are also used as a source of folates for archaea to synthesize archaeal pterins such as methanopterin, halopterins, or sulfopterin, this problem also extends to archaea. Given the large number of vitamin and media preparation protocols that filter-sterilize unbuffered vitamin concentrates and the prevalence of predicted folate auxotrophy, we propose that inadvertent folic acid removal is a major problem that limits the cultivation of medically, ecologically, and industrially important prokaryotes inhabiting diverse biomes in defined media.

Caldatribacterium shares an unusual cell structure with other Atribacterota

Given the unusual cell ultrastructure of the two previously isolated Atribacterota species^7,16,18, we examined the Caldatribacterium isolates. Both are rod-shaped or ovoid with somewhat pointed ends and flagella (Fig. 8; Supplementary Figs. 14 and 15). Similar to other Atribacterota^7,16,18, dense cytosolic regions were observed, along with three LMLs comprised of an internal membrane and a double layer at the cell boundary (Fig. 8; Supplementary Figs. 14 and 15; Supplementary Data 12–13). We noted previously¹² that draft Caldatribacterium genomes encode genes for synthesis and export of lipid A (lpxABCD; msbA), but lack genes for production and attachment of keto-deoxyoctulosonate and the liposaccharide inner core. Analysis of all Atribacterota genomes confirmed these observations (Supplementary Data 9), suggesting an outer membrane with unusual lipopolysaccharides.

Fig. 8. Caldatribacterium cell structure.

a, Cryo-electron micrograph of SIUC1^T cell with white arrow indicating denser cytosolic region. b, Closer cryo-electron micrograph view of strain SIUC1^T with dark arrows indicating 1. outer, 2. middle, and 3. inner lipid membrane-like layers. c, Epifluorescence micrograph of GBS cells hybridized with Cy3-labeled Caldatribacterium FISH probe and counterstained with DAPI. The white arrow indicates the cell shown in higher detail in d, with a profile of Cy3 (rRNA probe, red) and DAPI (DNA, blue) fluorescence along the axis of the cell. a–c images are representative of ~20 fields. Measurements and additional images are shown in Supplementary Data File 13 and Supplementary Figs. 14–16.

Epifluorescence microscopy showed that rRNA and DNA are not colocalized. DAPI fluorescence was visible only within compartmentalized nucleoids, yet rRNA probe fluorescence was distinct and concentrated near cell poles (Fig.  8c, d; Supplementary Fig. 16). This contrasts with imaging of A. laminatus showing co-localization of DNA, total RNA, and ribosomes in the inner and middle LML-bounded spaces, and rRNA probe fluorescence distributed throughout the middle of cells¹⁶. The localization of Caldatribacterium rRNA outside of compartmentalized nucleoids is inconsistent with interpretations^16,17 that the three LMLs are an S-layer, an outer membrane, and separated cytoplasmic membrane, because that model would place rRNA in the periplasm. Thus, our data suggest Atribacterota nucleoids are either bound by a third membrane (inner LML) or by invaginations of the cytoplasmic membrane, as observed in some Planctomycetota⁴¹. However, invaginations of the second LML are not apparent in any images of Atribacterota^7,16,18.

Given the unusual ultrastructure of Caldatribacterium and since a “membrane-centric” metabolism was suggested previously for A. laminatus^16,42, we examined the proteomes of Caldatribacterium and other Atribacterota. We observed a high proportion of proteins with transmembrane helices for C. saccharofermentans (29.7% of proteins) and C. inferamans (30.2%) and a high abundance of unusual signal peptides recognized by SignalP-5.0⁴³ but not SignalP-4.1 (Supplementary Fig. 17). In contrast to a previous report¹⁶, we found that Atribacterota SignalP-5.0 improved detection of signal peptides recognized that are less hydrophobic, have lower isoelectric points, and are longer than other Atribacterota signal peptides (Supplementary Figs. 18–20).

Discussion

Atribacterota was designated a candidate phylum >25 years ago, but it continues to resist cultivation. Our genome-guided predictions led us to enrich and isolate two species of Caldatribacterium. Both are sugar and sugar alcohol fermenters that produce H₂ and acetate, extending this core metabolism throughout the Atribacteria. This metabolism is highly exergonic⁴⁴, and such bacteria were isolated >80 years ago⁴⁵. So why do Atribacteria resist cultivation? A. laminatus is sensitive to H₂ and co-metabolizes sugars via interspecies H₂ transfer with M. thermoautotrophicus. Caldatribacterium cohabitates with hydrogenotrophs and probably also benefits from interspecies electron transfer. C. saccharofermentans cohabitates with various hydrogenotrophs and was enriched with T. auxiliatoris and a novel hydrogenotrophic methanogen^15,46. C. inferamans cohabitates with the hydrogenotrophic methanogens Methanobacterium and Methanothermobacter, along with abundant and active SRB⁴⁷. The co-enrichment of both Caldatribacterium species with hydrogenotrophs along with our nanoSIMS experiments, supports this co-metabolism model. Yet primary sugar fermentation is highly exergonic, and both Caldatribacterium species were relatively insensitive to H₂, so interspecies H₂ transfer could not explain our cultivation challenges.

We used genome-based predictions to enrich C. saccharofermentans on a defined medium and then isolated both strains by adding yeast extract to the solid medium. Through careful review of predicted auxotrophies and experimentation, we showed that folate auxotrophy hindered their isolation due to a decades-long problem with media preparation. Many researchers seeking to cultivate new lineages use Wolin’s vitamin solution^26,27 or its derivatives, which contain ten different vitamins and vitamin precursors and were developed >60 years ago for this purpose. For example, MediaDive¹⁹ lists ≥791 media using Wolin’s vitamins or similar vitamin solutions and ~76.4% of protocols (604/791) specify filter-sterilization of unbuffered vitamin solutions to avoid autoclaving. Only ~17.2% (136/791) of media specify that vitamins should be autoclaved, nearly all of which are linked to protocols in the Japan Collection of Microorganisms (JCM), rather than MediaDive directly. ~6.4% (51/791) of media in MediaDive do not specify how vitamins should be sterilized. Since all Atribacterota lack folate synthesis pathways, and predicted folate auxotrophy is extensive, this mistake is an insidious problem limiting our ability to cultivate prokaryotes in defined media. We note that neither Wolin^26,27 nor Balch^26,27 described any special protocol to sterilize these vitamins. Rather, vitamins were listed along with other medium ingredients that were autoclaved. We confirmed with long-standing members of the Wolfe lab (W. Whitman and B. Mukhopadhyay) that vitamin concentrates were added without filtration prior to autoclaving the medium. By surveying the early literature, we traced filter-sterilization of unbuffered vitamin solutions containing folic acid back at least 49 years to 1977^48,49, and prominent labs in Europe and the U.S. documented its proliferation in the late 1970s and early 1980s^50–53 to the point that this procedure is currently used in most microbiology labs globally⁵⁴, although details of media preparation are often not clearly documented in the literature.

The stability and solubility of folic acid solutions are also affected by other vitamins, with degradation catalyzed by riboflavin, ascorbic acid, and degradation products of thiamine, and solubility promoted by nicotinamide⁵⁵. We also note that protocols for mammalian cell culture use folinic acid (5-formyl-5,6,7,8-tetrahydrofolic acid), which is ~30x more soluble than folic acid and physiologically active because the pterin group is reduced, bypassing a requirement for dihydrofolate reductase such as FolA or FolM, and the C1 group is present, bypassing a requirement for formyltetrahydrofolate synthetase (Fhs). However, folinic acid may not be effective for Atribacteria or other prokaryotes lacking homologs of 5-formyltetrahydrofolate cyclo-ligase (YgfA; Fig. 6E). Thus, the behavior of vitamin mixtures prepared and stored under different conditions is complex. Yet we are not aware of any systematic studies of their behaviors in Wolin’s vitamin solution or other concentrates used by microbiologists, and this should be a research priority. However, our results clearly show that folic acid persists in neutralized and buffered Wolin’s vitamin solution (10x) for at least three months when stored under N₂ in the dark at 4 °C regardless of whether the solution is autoclaved or filter-sterilized or the details of the procedure to make the solution anoxic. We also show that folic acid persists in acidic and unbuffered Wolin’s vitamin solution (1x or 10x) that is autoclaved. However, we strongly advise against filter-sterilization of acidic and unbuffered vitamin solutions containing folic acid due to the removal of precipitated folic acid by filter-sterilization.

The relevance of vitamin solutions for microbiological media could be questioned because yeast extract and other undefined organic substrates are also sources of vitamins; yet we note that yeast extract is not a source of cobalamin (vitamin B12)⁵⁶. We argue that it would be regressive for the research community to rely solely on complex media. Regarding the cultivation of poorly understood microorganisms—the general aim of this study—media with defined sources of energy and elements give researchers more precise control over enrichment cultures than complex media. In our hands, the use of defined media has helped us to target specific taxa without overgrowth of non-target heterotrophs that are often faster-growing. This case is especially clear for enrichment or early stages of cultivation of autotrophs or diazotrophs that can be outcompeted or even inhibited in the presence of yeast extract and other complex organic compounds⁵⁷. Once axenic cultures or long-term mixed cultures are obtained, defined media give researchers more precise control for experiments to probe central catabolic and anabolic pathways, gene expression networks, metabolomics, and other physiological traits. Thus, while we acknowledge the merits and successes offered by complex media, including isolation of Atribacteria^7,16,18, we also emphasize the importance of defined media and advocate for broader use of defined media over complex media for better experimental control of both enrichment cultures and axenic cultures.

The Caldatribacterium isolates are similar in ultrastructure to other Atribacteria, with what appears to be an intracellular membrane containing a nucleoid. However, Caldatribacterium may separate the nucleoid from ribosomes by the inner LML. Atribacterota also possess a high percentage of transmembrane proteins, congruent with a “membrane-centric metabolism”. The unique ultrastructure of Atribacterota should be a high priority to understand its origin and roles in cell biology and ecophysiology.

Methods

GBS in situ enrichment and isolation of strain GBST

C. saccharofermentans GBS^T was obtained from GBS, a circumneutral pH, NaCl spring in northwestern Nevada (40°39’41.3”N, 119°21’58.1”W)⁴⁹. In-situ enrichments consisting of 1 g of AFEX-treated corn stover sealed in 100 µm nylon mesh bags were performed essentially as described¹⁶, because previous in situ enrichments contained Ca. Caldatribacterium saccharofermentans¹³. The bags were incubated at the water/sediment interface at the “C site” in the outflow of GBS from 26 October 2013 to 28 March 2014. Immediately after removal, the corn stover was transferred to a 25 mL serum tube containing 10 mL of sterile, anaerobic GBS spring water (previously collected, sparged with N₂, autoclaved, and reduced by addition of cysteine to 0.05%) under a stream of N₂. The tube was sealed and shaken, and then 0.1 mL of the contents was transferred using an N₂-flushed, sterile needle and syringe to 50 mL of anaerobic medium in a 150 mL serum vial sealed with a butyl rubber stopper. The medium was GBS salts medium⁵⁰ with 5 mM sodium phosphate pH 7 buffer and a filtered vitamin solution, but with xyloglucan (0.02% mass vol^-1) as the sole carbon source, reduced with 0.01% sodium sulfide, with an N₂ headspace. The concentrated Wolin’s vitamin solution contained 2 mg biotin, 2 mg folic acid, 10 mg pyridoxine HCl, 5 mg thiamine HCl, 5 mg riboflavin, 5 mg nicotinic acid, 5 mg DL-calcium pantothenate, 0.1 mg vitamin B12, 5 mg p-aminobenzoic acid, and 5 mg lipoic acid per liter^31–33. Unless otherwise noted, the concentrated Wolin’s vitamin solution was prepared in >18 MΩ-water, filtered-sterilized (0.2 μm nylon filters; 09-719 C, Fisher Scientific, Waltham, MA, USA), sparged with 99.999% N₂, and stored at 4 °C in a foil-covered borosilicate glass bottle in the dark until it was added at 1:100 (vol vol⁻¹) to the base medium after the medium was autoclaved. After the importance of folate and its problem with solubility was recognized, the concentrated Wolin’s vitamin solution was sterilized instead by autoclaving for 30 min at 121 °C and added to the base medium at 1:100 (vol vol⁻¹) medium after the medium was autoclaved.

The inoculated medium was brought back to the lab and incubated at 73 °C without shaking within 24 hours of sampling. The enrichment culture was initially transferred to fresh medium (1/1000 dilution) after a 1-week incubation for the first 2 weeks, and then it was transferred every two weeks thereafter. Subsequent experiments replaced xyloglucan with various individual sugars or other substrates. Dilution-to-extinction was performed using fucose (0.02% mass vol⁻¹) using 10-fold serial dilutions up to 10⁻¹⁰ dilution. Isolation was performed using media solidified with Gelrite (0.8% mass vol⁻¹) and magnesium chloride hexahydrate (0.4% mass vol⁻¹). Plates were prepared in an anaerobic chamber and incubated in anaerobic incubation vessels for 7–10 days with an N₂ headspace⁵¹. Single colonies were streaked for isolation three times.

Well Inyo-BLM 1 and in situ enrichment and isolation of strain SIUC1T

C. inferamans SIUC1^T was isolated from a sterile polyurethane foam plug that was suspended in borehole/well Inyo-BLM 1 (36° 24’04.19 N / 116° 28’06.58 W) for three months. This borehole was completed in 2007 reaching a depth of 883.5 m below land surface (mbls) by The Hydrodynamics Group, LLC (Edmonds, WA, USA) on behalf of Inyo CO, CA, the US Department of Energy, and the US National Park Service as part of a suite of holes used to test hydrologic connectivity across the Funeral Mountains adjacent to Death Valley, CA, USA⁵⁸. It transects lithologies ranging from volcanic tuff and valley-fill alluvium to lake sediments, ultimately intersecting Paleozoic dolomite that hosts a low-transmissivity portion⁵⁹ of the Lower Carbonate Aquifer of the Death Valley Regional Flow System at 748 mbls^58,60. The plug was recovered from an inferred fracture zone roughly corresponding to a partially collapsed void²⁰ immediately below the casing terminus ~750 – 753 mbls. After retrieval, the foam plug was placed into pre-reduced artificial groundwater medium (AGM)²³. For initial enrichment and isolation attempts, vitamins were supplied as Vitamin Supplement MD-VS™ from the ATCC before autoclaving. An additional sample of the aquifer water without enrichment was obtained from Inyo-BLM 1 at ~752 m via a gas-tight discrete sampler (“bailers,” Comprobe, Inc.; Fort Worth, TX)²⁰. 250 ml of the bailed water was filtered using a 0.22 μm PES filter (UX-06730-43, Thermo Scientific; Waltham, MA). 250 μl of Inyo-BLM 1 water was also added to the PowerSoil lysis tubes to prevent detection bias against small cells²⁰.

Lab cultivation was conducted by removing 1 mL of liquid from the stored sponge in AGM and injecting it into a 160 mL serum bottle containing 50 mL (2% vol vol⁻¹ inoculum) of AGM amended with 10 mM xylitol and 0.01% mass vol⁻¹ yeast extract. The final pH of the medium was 8.0 after reduction of the solution. Enrichments were incubated at 60 °C without shaking, and transferred every 7 days, using a 0.5% vol vol⁻¹ inoculum into sterile anaerobic AGM media in 160 mL serum bottles. Cell morphology and density were monitored using a Zeiss Axioskop2 Plus microscope and a Petroff-Hauser counter under phase contrast. Dilution-to-extinction was conducted three times, which successfully enriched the culture to >90% C. inferamans. A clonal isolate was generated from the third dilution-to-extinction culture by picking a colony and restreaking it two additional times on anaerobic medium solidified with Gelrite (0.8% mass vol⁻¹) and magnesium chloride hexahydrate (0.4% mass vol⁻¹), with 10 mM xylitol, 0.05% (mass vol⁻¹) yeast extract, and 0.05% (mass vol⁻¹) casamino acids.

16S rRNA gene qPCR and amplicon analysis of enrichment cultures

For GBS enrichments, DNA extraction using the FastDNA Spin Kit for Soil (MP Biomedicals, Solon, OH, USA) and 16S rRNA gene amplicon sequencing was done with an Illumina MiSeq (2×250) using 806r reverse primer and a modified 515f-mod forward primer (5′ GTGYCAGCMGCCGCGGTAA) to enhance coverage of archaea⁴⁶. Quantitative PCR (qPCR) was done using primers OP9_16S_F (5’ AGGAAAGCTGGCCTCTGC) and OP9_16S_R (5’ ACCGTCACAGGAAGGAGC) specifically targeting Caldatribacterium (designed based on an alignment of Atribacterota 16S rRNA genes) or with primers 515f-mod and 806r targeting total bacteria and archaea⁴⁶. The standard was a plasmid containing a fragment of the Caldatribacterium 16S rRNA gene (SSW_L1_H02⁶¹).

For the Inyo-BLM 1 aquifer bailed sample and in-situ enrichment (foam plug), DNA was extracted using the MoBio PowerSoil DNA Isolation Kit (Carlsbad, CA, United States). Lab enrichment cultures were extracted using the MoBio UltraClean Microbial DNA Isolation Kit (Carlsbad, CA, United States), both according to the manufacturer’s protocol amended with one freeze/thaw cycle (−80 °C/65 °C) at the beginning of DNA extraction procedure. DNA concentration was determined using a NanoDrop® ND-1000 UV-Vis Spectrophotometer (Thermo Scientific, Waltham, MA, USA) with wavelength settings of 260 nm and 280 nm. Extracted DNA was sent to the Environmental Sample Preparation and Sequencing Facility at Argonne National Laboratory (Lemont, IL, USA) and sequenced using the Illumina MiSeq platform (2 × 151 bp). Universal bacterial primers targeting the V4 region of the 16S rRNA gene were used for diversity analysis⁶². Paired-end reads were merged, denoised, and demultiplexed using deML⁶³. For both datasets, amplicon sequence variants (ASVs) were processed using the DADA2 pipeline⁶⁴ implemented in R using default parameters (https://benjjneb.github.io/dada2/tutorial.html).

Stable isotope labeling

Stable isotope labeling was performed with xyloglucan mixed cultures and fucose co-cultures derived from GBS. Universally ¹³C-labeled substrates (Cambridge Isotope Laboratories) were added to freshly transferred enrichments grown for 2 days from concentrated, anaerobic stocks to the microcosms at 1 mM final concentration unless otherwise noted: bicarbonate, formate, acetate, glucose, xylose, ribose, or algal amino acids (0.01% mass vol⁻¹). Negative controls with no compound added were incubated in parallel. All incubations were done at 73 °C without shaking for 2 hours. After incubation, cells were pelleted by centrifugation for 5 min at 16,100 × g, pooled in 0.5 mL of 1× phosphate-buffered saline (PBS), and 0.25 mL of freshly prepared 3% paraformaldehyde (PFA) was added and mixed. Samples were fixed for 1 hour on ice, cells were pelleted by centrifugation for 5 min at 9200 × g, and pellets were washed twice with 1×PBS. Washed cells were resuspended in 200 µL of 50% ethanol, and stored at –20 °C.

FISH

FISH was performed using protocol v 2.2⁶⁵ (https://www.arb-silva.de/fish-probes/fish-protocols/) using the Caldatribacterium-specific 16S rRNA probe OP9-480 (5’-AGCTRTTCACCCCTYCCCTC-3’) labeled with Cy3 and the Bacteria-specific 16S rRNA probe Bact927⁶⁶ (5’-ACCGSTTGTGCGGGCCC-3’) labeled with 6-FAM. Cells from cultures were pelleted by centrifugation (10,000 × g for 5 minutes), washed in 1× PBS, and fixed with 1% paraformaldehyde in 1× PBS on ice for 1 hour. After fixation, the cells were pelleted, washed three times in 1× PBS, resuspended in 50% ethanol and stored at −20 °C. Hybridization was performed on slides at 46 °C with 30% vol vol⁻¹ formamide. Hybridization conditions were optimized for the OP9-480 probe using the Clone-FISH technique⁶⁷ via expression of the near-full-length 16S rRNA gene sequence of a Caldatribacterium relative obtained from Mud Hot Springs⁶¹ (SSW_L1_H02) in E. coli strain JM109 (DE3) from the plasmid pGEM-T. After hybridization, cells were counterstained with DAPI (1 µg mL⁻¹). Cells were visualized by epifluorescence microscopy using an Eclipse Ti-U inverted microscope (Nikon, Melville, NY, USA) equipped for epifluorescence with Nikon filter sets compatible with Cy3 (96312 G-2E/C), 6-FAM (96343 EN GFP), and DAPI (96310 UV-2E/C), with image capture using a Retiga-SRV camera (QImaging, Surrey, BC, Canada) and Nikon Elements v4.13 software. Isotopically labeled cells were deposited onto ITO-coated slides and visualized with a Leica DM5500B microscope using MetaMorph software with a ×100 magnification dry immersion objective. Fluorescence and brightfield images were collected, and the X-Y and fiducial locations were noted to enable navigation in the nanoSIMS.

NanoSIMS

NanoSIMS was conducted on a CAMECA NanoSIMS 50 at Lawrence Livermore National Laboratory. Fiducial locations and locations of FISH-positive cells were found with a charge-coupled device camera using X-Y coordinates and a real-time imaging unit. The primary Cs⁺ ion beam was set to 1.5 pA, corresponding to an ~150 nm beam diameter at 16 keV. Rastering was performed over 20 × 20 μm areas with a dwell time of 1 ms pixel^–1 for 19–30 scans and generated images containing 256 × 256 pixels, yielding data for 10–114 Caldatribacterium cells per experiment. Sputtering equilibrium at each area was achieved with an initial beam current of 90 pA to a depth of ~10 nm. After tuning the SIMS for mass resolving power of ~7000, secondary electron images and quantitative secondary ion images were simultaneously collected for ¹²C₂^– and ¹³C¹²C⁻ on individual electron multipliers in pulse counting mode. NanoSIMS data were initially processed using L’Image (https://www.limagesoftware.net/) to perform deadtime and image shift correction of ion image data before creating ¹³C¹²C/¹²C₂ ratio images, which reflected the level of ¹³C incorporation into biomass. Regions of interest for isotopic ratio quantification were drawn manually around each cell.

Pure culture experiments

Characterization of strain GBS^T was performed in 10 mL of GBS salts medium in 20 mL serum vials. Substrates were tested at either 0.05% mass vol^-1 for sugars and complex organic substrates, 1 mM for organic acids, 0.1 atm methane, or 0.4/0.1 atm H₂/CO₂. Cultures were incubated in the dark without shaking at 73 °C, and growth was assessed by phase-contrast microscopy using a Petroff-Hauser cell counter after seven days of incubation. Growth rate tests for pH (using 5 mM sodium phosphate buffer) and temperature optima were performed using 0.05% fucose as a growth substrate, with samples taken every 12-24 hours after growth was initially observed. Growth was tested at 50–80 °C in 5 °C increments, at 77.5 °C, and at pH values from 4.5 to 9.5.

Strain SIUC1^T cultivation experiments were conducted in 160 mL serum bottles containing 50 mL volume of AGM and 10 mM xylitol with a 2 atm headspace of 99.9% pure N₂. Temperature and pH culturing assays were averaged across quadruplicate replicates. Cultivation experiments to define temperature optima were incubated at 55, 60, 65, 70, and 75 ^oC in the dark without shaking. Cultivation experiments to determine optimal pH utilized alternative buffers. A final concentration of 10 mM for each buffer was used to achieve the desired pH values ranging from 6.0–9.5. The buffers used were: 2-ethanesulfonic acid (MES) for pH 6.0, 1,4-piperazinediethanesulfonic acid (PIPES) for pH 6.75, 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) for pH 7.5, tris(hydroxymethyl)aminomethane (TRIS) for pH 8.0/8.75 and N-Cyclohexyl-2-aminoethanesulfonic acid (CHES) for pH 9.5. Cell densities for temperature and pH assay were calculated from 24-hour intervals by microscopy using a Petroff-Hausser counting chamber.

Growth of strain SIUC1^T on different substrates was determined by quantifying growth in AGM media amended with 10 mM of the selected carbon/energy source at 65 °C, or 2 atm for H₂/CO₂ (80%:20%) or methane (99.9%). Undefined substrates tested were adjusted to a concentration of 0.1% mass vol⁻¹, which included casamino acids, peptone, starch, yeast extract, and xylan. Growth was monitored by microscopy using a Petroff-Hausser counting chamber. If a culture reached a cell density of 1 × 10⁶ cells mL⁻¹ or greater within 7 days, it was transferred to a new culture. If a culture reached 1 × 10⁶ cells mL⁻¹ or greater within 7 days in the 3rd transfer, it was considered a stable culture growing on the specific substrate. Experiments with 10 mM xylitol as the carbon/energy source were combined with sulfur-containing potential electron acceptors at concentrations of 10 mM (sulfate, sulfite, thiosulfate, and elemental sulfur). Sulfide production was assessed qualitatively using lead-acetate strips to indicate reduction of the sulfur compound.

For both strains, acetate was quantified by high-performance liquid chromatography (HPLC) using authentic standards and a Shimadzu Prominence-i LC-2030 HPLC equipped with a Resex ROA-Organic Acid H + 8%, LC-column (300 × 7.8 mm) and Security Guard KJ0-4282 (Phenomenex). 1 mL samples were taken from serum bottles by syringe and centrifuged for 10 minutes at 16,873 × g. Supernatant was transferred to a new tube and acidified by adding 25 µL of 200 mM sulfuric acid. The acidified sample was then filtered through 0.2 µm filter into an HPLC vial. Samples were run on the HPLC under the following conditions: column temperature 30 °C, isocratic, 5 mM sulfuric acid mobile phase at a flow rate of 0.5 mL min⁻¹, UV detector set to 254 nm, and run time 30 minutes. Peaks were integrated using LabSolutions LC/GC release 5.87. H₂ was detected qualitatively in headspace samples collected by syringe using a portable hydrogen detector (Forensics Detectors, Model: FD-90A-H2, Rolling Hills Estates, CA, USA).

SRB pure cultures were grown in GBS salts medium as prepared for C. saccharofermentans GBS^T (5 mM sodium phosphate pH 7 buffer, 0.01% mass vol⁻¹ sodium sulfide) except that vitamins and xyloglucan/sugars were excluded, and instead 1 mM sodium bicarbonate, 1 mM sodium acetate, 1 mM sodium thiosulfate, 5 mM lactate, and 0.66 atm H₂ were added. For co-cultivation experiments, late-exponential- or early-stationary-phase pure cultures of Caldatribacterium and SRB were diluted 1/100 in medium without added vitamins, and then 0.1 mL of each pure culture was used to inoculate 10 mL cultures in Balch tubes. The extra dilution ( ~ 10,000-fold total) was used to dilute out folic acid in the original pure cultures of Caldatribacterium. The medium was GBS salts medium with 0.05% mass vol⁻¹ fucose and autoclaved vitamins without added folic acid.

SRB culture supernatant preparation and initial analysis of water-soluble vitamins

Cultures of SRB were grown in the absence of added vitamins with 1 mM sodium bicarbonate, 1 mM sodium acetate, 1 mM sodium thiosulfate, 5 mM lactate, and 0.66 atm H₂ in 50 mL GBS salts medium in 150 mL serum vials until late exponential phase. Aliquots of culture were centrifuged at 800 × g in 1.5 mL microcentrifuge tubes in an anaerobic chamber, and pooled supernatants were filtered with a 0.2 μm nylon filter into sealed serum vials and autoclaved for 30 min. For Caldatribacterium growth experiments, 2.5 mL of sterile supernatant was mixed with 7.5 mL of GBS salts medium previously amended with sulfide, and then fucose and autoclaved vitamins without folic acid were added to yield their typical 0.05% mass vol⁻¹ and 1x concentrations, respectively. Dilution of the Caldatribacterium inocula was performed as above for the co-culture experiments to minimize carryover of folic acid. Samples of supernatant for vitamin analysis were frozen at −80 °C and sent on dry ice for analysis at The Metabolomics Innovation Center (TMIC; Edmonton, AB, Canada) for quantification of water-soluble vitamins using LC-MS/MS⁶⁸. Seven-point calibration curves were generated by adding 10 μL of isotopically labeled internal standard mixture to 50 μL of calibration solutions in Eppendorf tubes. Samples were also prepared by adding the isotopically labeled internal standard mixture to 50 μL of sample. A total of 60 μL of an aqueous TCA solution (50 mg/mL) was pipetted to each Eppendorf tube containing the calibrants or samples. Each tube was vortexed for 30 s for thorough mixing and then left on ice for 1 h. After cooling, each tube was centrifuged at 13,000 rpm for 20 min, and 100 μL of the supernatant was transferred to a new HPLC vial. A volume of 10 μL was injected for LC − MS/MS analysis. An Agilent 1260 series UHPLC system (Agilent Technologies, Palo Alto, CA) coupled with an AB Sciex QTRAP 4000 mass spectrometer (Sciex Canada, Concord, Canada) was used to analyze water-soluble vitamins. An Agilent reversed-phase Zorbax Eclipse XDB C18 column (3.0 mm × 100 mm, 3.5 μm particle size, 80 Å pore size) coupled to a Phenomenex (Torrance, CA) SecurityGuard C18 precolumn (4.0 mm × 3.0 mm) was used for the separation of all water-soluble vitamins in the samples. Data analysis was performed with Sciex Analyst 1.6.2⁶⁸.

Vitamin solution preparation, soluble folic acid quantification, and growth assay

Wolin’s vitamin solution (10x)²⁶ was prepared as written in MediaDive (Solution 5980)¹⁹. This is 10-fold higher than the concentration of the vitamin solutions used for cultivation in this study and was chosen because it is the most-used vitamin solution in the database. Four solutions were made, varying pH/buffering and the N₂-sparging protocol. Buffered solutions were prepared by first adding 900 mg/L of monopotassium phosphate to 18 MΩ water and adjusting the pH to 7 using NaOH, the same buffer used in ATCC Vitamin Supplement MD-VS™, while unbuffered solutions were not pH-adjusted. N₂-sparging was done in two ways: by first sparging 18 MΩ water with stirring until anoxic (45 minutes) and then preparing the solution under a stream of N₂ (“before addition”), or by preparing the solution with oxic 18 MΩ water and then sparging with N₂ with stirring (45 minutes) (“after addition”). All four variants were then separated into three aliquots, one sterilized by autoclaving, one sterilized by filtering with a 0.22 μm PES filter (UX-06730-43, Thermo Scientific; Waltham, MA), and one left not sterilized, for a total of 12 variants of Wolin’s vitamin solution (10x). All solutions were then protected from light and stored at 4 °C.

Aliquots of all 12 variants, as well as fresh ATCC Vitamin Supplement MD-VS™, were sent to the Mass Spectrometry Core at Southern Illinois University, Carbondale, for LC-MS analysis of soluble folic acid. LC-MS was conducted using a Shimadzu LCMS-2020 with SPD-20A UV-vis detector, and HPLC conducted with an Acclaim C30 column (2.1 mm × 150 mm, 3 μm). The folic acid standard was diluted using 10 mM ammonium formate, and the peak area determined to be 280 nm. The 12 vitamin solutions were diluted 20× while ATCC Vitamin Supplement MD-VS™ was diluted 2×. Samples were first run as a single replicate after 1–2 weeks of storage at 4 °C in the dark and later in triplicate after storage at 4 °C in the dark for three months. For analysis, vitamin solutions were removed from the refrigerator, warmed to room temperature for ~3 hours, and mixed by inverting ten times before sampling for the LC-MS run. Values were converted to observed folic acid in parts per million (ppm) for all values above the level of detection or the level of quantification recommended by Shimadzu (0.082 ppm/0.249 ppm), with standard deviation and % relative standard deviation calculated. Observed folic acid concentrations were analyzed using a three-way factorial ANOVA with buffering, sterilization method, and sparging treatment as fixed effects. Post-hoc pairwise comparisons were performed using Tukey’s HSD. The proportion of variance explained by each main effect was calculated from the sums of squares derived from the fitted model.

Aliquots of these Wolin’s 10× solutions were also tested to determine how well they would support growth of C. saccharofermentans. They were warmed to room temperature for ~3 hours, mixed by shaking for 10 seconds, and then diluted ten-fold in sterile, N₂-sparged water to achieve 1× solutions. The ATCC MD-VS™ vitamins were either used as purchased (sterile) or sparged with N₂ gas for 1 hour and then autoclaved for 30 minutes to resterilize. These vitamin solutions were then used in place of the typical (1×), autoclaved vitamin solution for growth of C. saccharofermentans with 0.05% fucose as described for pure culture experiments, with the inoculum diluted 10,000-fold as used for the SRB supernatant experiments to minimize folic acid carryover. Cultures were incubated at 70 °C in the dark, sampled at four, seven, and 14 days after inoculation, and cell counts were determined with a Petroff-Hausser counting chamber. Growth data were analyzed using a linear mixed-effects model with buffering, sterilization, sparging, and time as fixed effects and replicate as a random effect. Post-hoc pairwise comparisons were performed using Tukey’s HSD on the full factorial design.

Electron microscopy and sample preparation

For SEM, late-exponential phase cells were immediately fixed with 2% vol vol⁻¹ glutaraldehyde in 0.1 M sodium cacodylate buffer for 1 hour at 4 °C. Cells were pelleted by centrifugation (10,000 × g for 10 min), washed three times with 0.1 M sodium cacodylate buffer, and resuspended in 1% OsO₄ in 0.1 M sodium cacodylate buffer. Cells were then incubated for 1 hour at 4 °C and then rinsed twice with distilled water. The cells were dehydrated by washing for 15 minutes at a time with increasing concentrations of ethanol (25, 50, 75, and 95%) followed by three washes of 100% ethanol. Samples were then embedded on a 0.2 μm Millipore filter (13 mm diameter) mounted in a Swinney filter holder. Filters were rinsed with 100% ethanol and kept in ethanol until critical point drying on a Tousimis SAMDRI-790 (Rockville, MD, USA). Sputter coating was carried out using a Denton Vacuum Desk II, and the samples were coated with a 600 Å layer of Au-Pd. Imaging was performed using a FEI Quanta FEG 450 SEM, using an acceleration voltage of 20 kV.

For cryo-EM, 1 mL of cell culture was centrifuged at 1000 × g for 2 min to remove insoluble components from the media. The supernatant was centrifuged at 3000 × g for 5 min to pellet the cells, and the pellet was resuspended in 20 μL of media. Samples were prepared with an automated Leica EM GP plunge freezer set at 21 °C and 95% humidity in the sample chamber. A 3 μL sample of the cell suspension was applied onto glow-discharged copper R2/2 200 grids (Quantifoil), pre-blotted for 60 seconds, blotted for 2 seconds, plunged into liquid ethane, and stored in liquid nitrogen. The samples were imaged on a 120 kV Talos L120C transmission electron microscope at the Netherlands Center for Electron Nanoscopy (NeCEN).

Lipid analysis

Cultures of GBS^T (0.05% mass vol⁻¹ fucose, 73 °C) and SIUC1^T (0.05% mass vol⁻¹ xylitol, 67.5 °C) were grown to late-exponential phase, and cells were harvested by centrifugation (10,000 × g for 10 min). Cell pellets were stored at −80 °C, and fatty acid methyl ester analysis was performed by Microbial ID (Newark, DE, USA) with the MIDI Sherlock Microbial Identification System and SMOORE6 database.

DNA extraction and PCR amplification of the 16S rRNA gene

DNA was extracted from cells of strain GBS^T grown for seven days and SRB isolates grown for 3 days using the FastDNA Spin Kit for Soil. Strain SIUC1^T cells were grown to a density of 1 × 10⁸ cells mL⁻¹ in a cumulative volume of 1 L to 500 mL (10 × 50 mL bottles), and cells were pelleted by centrifugation at 15,000 × g for 30 min in 50 mL conical tubes (Sorvall RC6 Plus centrifuge, rotor F13S14x50cy) at 4 °C. The genomic DNA was extracted by cetyltrimethylammonium bromide (CTAB) treatment followed by mild homogenization, buffer/chloroform, and finally purified/cleaned with QIAquick PCR (Qiagen) purification columns⁶⁹. The DNA was then assessed for quality and quantity by Qbit and Nanodrop. DNA was aliquoted and stored at −80 °C, until needed.

For strain GBS^T and SRB isolates, the primers 9bF (5’-GRG TTT GAT CCT GGC TCA G-3’) and 1512uR (5’ ACGGHTACCTTGTTACGACTT) were used for amplification of 16S rRNA genes⁶¹, and Sanger sequencing with these primers was performed by Retrogen, Inc. (San Diego, CA, USA). For strain SIUC1^T, universal bacterial primers 27F-YM (5’-AGA GTT TGA TYM TGG CTC AG-3’) and 1492 R (5’-TAC CTT GTT ACG ACT T-3’) were used to amplify the 16S rRNA gene, and the resulting amplicons were sequenced by MCLAB (South San Francisco, CA, USA).

Genome sequencing, assembly, and annotation

For strain GBS^T, short read data (2 × 150) was obtained using an Illumina NextSeq 2000 with libraries prepared using the Illumina DNA Prep kit at SeqCenter (Pittsburgh, PA, USA). Long read sequencing was performed with a MinION Mk1B device and FLO-MIN106 flow cell using libraries prepared with the EXP-NBD104 Native Barcoding Expansion Kit (Oxford Nanopore Technologies, Oxford, UK) and the SQK-LSK109 Ligation Sequencing Kit. Short reads were trimmed using Trimmomatic⁷⁰, and hybrid assemblies were performed using Unicycler⁷¹. Initial annotation was performed using RAST⁷².

Genome assembly for strain SIUC1^T was carried out using data from multiple short read runs (Illumina) GeneWiz (San Diego, CA, USA) and UIC Research Resource Center (Chicago, IL, USA), as well as long reads (Oxford Nanopore) using Unicycler⁷¹. Assemblies were analyzed and compared using QUAST⁷³. The final assemblies were circular. Both genomes were annotated with the NCBI Prokaryotic Genome Annotation Pipeline (PGAP). Alternative annotations were obtained using eggnog-mapper v.2 and RAST.

Phylogenomic and genome distance analysis

Atribacteria species-representative genomes from the GTDB 09-RS220² plus those from recent publications^7,18 were used to construct a species tree. The phylogenomic relationships of genomes were inferred based on a concatenation of the bac120 protein marker set⁷⁴ identified by GTDB-Tk v. 2.3.2⁷⁵. The concatenated data matrix was analyzed using maximum-likelihood inference in IQ-TREE v.2.2.6⁷⁶, with 1000 ultrafast bootstraps (UFBoot) and 1000 SH-like approximate likelihood ratio test (SH-aLRT) replicates (-bb 1000 -alrt 1000), using the best-fit model identified by ModelFinder⁷⁷ (LG + F + I + R3). The phylogeny was rooted at the midpoint. Pairwise average nucleotide identity (ANI) and average amino acid identity (AAI) between the MAGs were calculated using FastANI v.1.1⁷⁸ and FastAAI v.1⁷⁹, respectively.

Presence/absence analysis of folate biosynthesis pathway genes across bacteria

A total of 37 COGs from the 2024 COG dataset⁸⁰, compiled from enzymes described in this work as well as from literature review^33,81,82, were used to study the presence or absence of different components of the folate biosynthesis pathway across a diverse population of bacterial phyla. The Web of Life 2 (WoL2) dataset consists of 15,953 bacterial and archaeal genomes, chosen specifically to be representative of microbial phylogeny^83,84. Archaeal genomes (n = 810) and genomes suppressed by NCBI (n = 1817) were removed after downloading the GTDB taxonomy (09-RS226)² for each genome, resulting in a final dataset of 13,326 bacterial genomes representing 97 phyla. Annotation of all genes by COG functions was completed using anvi’o v8, specifically the ‘contigs’ Snakemate workflow^85,86. FASTA files for each genome were first converted into contig-db files by anvi’o before having their open reading frames predicted by Prodigal v2.6.3⁸⁷. The resultant predicted proteins were then aligned against the 2024 update COG functions by DIAMOND v2.1.13⁸⁸. The resulting COG functions were sorted by presence/absence using the command ‘anvi-script-gen-function-matrix-across-genome,’ filtered for only the 37 COGs of interest, and the resulting data were organized by phyla.

To improve genomic coverage of rare phyla, any phylum ≤50 genomes in the WoL2 was supplemented by downloading each species-representative genome for that phylum in GTDB (10-RS226). This included 86 phyla with no genomes in the WoL2 dataset, as well as 77 other poorly represented phyla in WoL2. Additionally, four phyla with ≥50 genomes in the WoL2 were included in this secondary examination, as members of these phyla are known to have host-associated lifestyles (Chlamydiota, Bdellovibrionota, Myxococcota, and Omnitrophota). All analyses were otherwise performed as described above, resulting in an additional 10,929 genomes studied.

More careful annotations for folate-related proteins were done with the Caldatribacterium isolate genomes by manually examining gene neighborhoods around key genes and identifying key domains and residues using NCBI’s Conserved Domain Database (CDD) and InterProScan^89,90. To generate high-quality predictions of folate-related genes across the Atribacterota, reviewed proteins were used for a local blastP analysis, using genomes from the two Caldatribacterium strains and high-quality GTDB reference genomes for each genus of Atribacterota per GTDB (10-RS226)².

CSM and molecular docking

CSMs of C. saccharofermentans GBS^T FolC were generated using AlphaFold 3 via the AlphaFold Server and visualized using ChimeraX v1.6rc202304050652⁹¹. The native FolC CSM was generated on July 11, 2026 using seed 1776415026 with a pTM value of 0.94. This structure was aligned to FolC from E. coli (PDB ID 1w78), and the bound ligands from 1w78, including dihydropteroate, were docked into the CSM for visualization. The FPGS from Lacticaseibacillus casei (PDB ID 1jbw) was also aligned to model 1w78, and the bound ligands from E. coli FolC were docked to L. casei FPGS. Residues from all models that structurally align to the dihydropteroate binding residues from E. coli FolC were colored using the RasMol amino acid color scheme; ligands were colored by the Corey-Pauling-Kultun color scheme.

Annotation of secreted and membrane proteins

Predicted secreted and transmembrane genome content was analyzed within high-quality ( ≥ 90% completeness, <5% contamination)¹² Atribacterota, Thermotogota, Dictyoglomota, Synergistota, and Verrucomicrobiota genomes obtained from the Genome Taxonomy Database¹³ release 202, as well as other genomes from bacteria included in the TEMPURA database⁹². Open reading frames were predicted and translated into protein sequences using Prodigal⁸⁷ version v2.6.3. Secretion signal cleavage sites were predicted using SignalP-4.1⁴² and SignalP-5⁴³. Amino acid sequences three residues upstream or two downstream of cleavage sites were extracted and rendered using the R package ggseqlogo⁹³ version 0.2. Transmembrane helices were predicted using TMHMM 2.0⁹⁴. Hydrophobicity was calculated for each signal peptide sequence by averaging the sum of individual per-residue hydrophobicities as estimated using the Kyte and Doolittle scale⁹⁵. Isoelectric point was estimated for each signal peptide using the isoelectric point calculator web interface²⁰.

Inclusion and ethics

This research included local researchers as full authors, when possible, to recognize both logistical and intellectual contributions. No potential or listed authors were discriminated against on the basis of gender, race, ethnicity, or any other factors not related to scientific contributions.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

J.A.D. and S.H.B. conceived of the study. B.P.H., J.A.D., and D.P.M. led field campaigns, managed sampling permits, and obtained and managed major grants supporting the work. S.H.B., J.A.D., T.A., J.L., J.J., A.M., A.T., A.V., A.R.M., A.C., J.L., Z.O., A.M.B., E.L., J.T.P., and K.M. conducted culturing enrichment, isolation, and strain characterization. T.A. and J.A.D. performed FISH and ¹³C-labeling experiments. X.M., J.J., P.K.W., J.P.-R., and D.L. analyzed samples with nanoSIMS and nanoSIMS data analysis/interpretation. N.S.S., A.C., A.V., J.A.D., S.H.B., and B.P.H. conducted LC-MS and growth experiments with different vitamin preparations. T.S. prepared cells for SEM imaging. A.M. and A.B. conducted cryo-EM and led structural interpretations. T.R.M., R.X., and J.A.D. conducted genome sequencing and assembly. D.L. led on bioinformatic analysis related to taxonomy, and B.P.H. led on interpretations of the taxonomy. N.S.S., D.L, C.O.S., and R.K.D. led bioinformatic analyses focused on vitamin biosynthetic pathways and N-terminal signal sequences. B.P.H., J.A.D., and S.H.B. wrote the manuscript, and all authors contributed to the final version of the paper.

Peer review

Peer review information

Nature Communications thanks Michaela Salcher and the other, anonymous, reviewer for their contribution to the peer review of this work. A peer review file is available.

Funding

Part of this work was carried out at Lawrence Livermore National Laboratory (LLNL) under Contract AC52-07NA27344 (X.M., P.W., J.P.-R.). Funding was also provided by the U.S. National Science Foundation (DEB 1557042, B.P.H., D.L., C.O.S., J.A.D., X.M., J.P.-R.), NASA (80NNSC17KO548, B.P.H., D.L., C.O.S., J.A.D., X.M., J.P.-R.), NASA Astrobiology Institute (NNA13AA92A, D.P.M., S.H.B.), Nevada NASA Space Grant (80NSSC20M0043, B.P.H., D.L., D.P.M.), and NASA EPSCoR (80NSSC25M0046, B.P.H., D.P.M.). This work was also supported by startup funds from Southern Illinois University Carbondale (S.H.-B.) and the Research-Enriched Academic Challenge (REACH) undergraduate grant (A.M.B.). The scanning electron microscope used in this work was purchased through a grant from National Science Foundation (CHE 0959568). We also acknowledge the National Science Foundation through the grant CHE 0959568 that facilitated the purchase of the FEI Quanta 450 scanning electron microscope. We thank Valerie Jimenez, Sandy Macias, Katelyn Holt, Maidy Ramos, Jorge Torres, Alejandra Moreno, Joseph Mansuri, and Matthew David for assistance with characterization of GBS and SRB strains.

Data availability

The complete C. saccharofermentans GBS^T and C. inferamans SIUC1^T genomes are available in GenBank under accession numbers CP187957 and CP189974. Raw reads for all metagenomes and 16S rRNA gene amplicons are available from the Sequence Read Archive within BioProjects PRJNA1250235 and BioProject PRJNA956542. GenBank accession numbers for GBS^T, SIUC1^T, and T. auxiliatoris 16S rRNA genes are MN114535, MT023787, and MH513620. Source data are provided with this paper.

Code availability

Scripts used to account for vitamin biosynthetic pathways and compare signal peptides are available on GitHub: https://github.com/Hedlund-Lab/atribacterota-2025.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-73575-2.