Genome-Resolved Metagenomics Extends the Environmental Distribution of the Verrucomicrobia Phylum to the Deep Terrestrial Subsurface

The microbial ecosystem of a shale gas well in the Marcellus formation.The salinity of produced fluids collected from a Marcellus shale well (Marcellus-5 [33]) range from freshwater levels (upon injection) to a maximum salinity (as total dissolved solids; TDS) of 122 g/liter 215 days after fracturing, with an average of 104 g/liter over the duration of the sampling period (Fig. 1a). This salinity increase is the result of interactions between freshwater-based input fluids and salts in the shale formation (36). Consistent with prior reports by our team and others (34, 35, 37–39), we show that salinity increases in the shale produced fluids over time in the Marcellus-5 well.

• Open in new tab
• Download powerpoint

FIG 1

Geochemistry and microbial community attributes of the Marcellus-5 fractured shale ecosystem since the collection of produced fluids on day 34. (a) Produced fluids from 93 days after fracturing onwards are characterized by high salinity, expressed as total dissolved solids (TDS) in grams per liter. (b) Relative abundances of fractured shale members (as inferred from rpsC gene relative abundance) across the five metagenomes. (c) “Candidatus Marcellius” is a rare member of the community only detected in later metagenomes.

To better understand microbial community response to salinity changes, we selected five samples (spanning 103 to 313 days posthydraulic fracturing) where TDS nearly doubled from input fluid concentrations. Microbial relative abundance (based on rpsC gene [S3 ribosomal protein] relative abundance) revealed that Arcobacter and Marinilabilia dominated earlier samples until day 152, with Methanohalophilus and Halanaerobium dominating later time points (Fig. 1b; see also Table S1 in the supplemental material). These findings are consistent with the recent cultivation of Arcobacter and Methanohalophilus strains recovered from early and late produced fluids, respectively, from an adjacent shale gas well (40, 41). Moreover, our findings highlight salinity as an environmental filter, likely contributing to the low microbial diversity and enrichment in halotolerant taxa over time.

Notably, we recovered a genome belonging to the Verrucomicrobia from the day 313 metagenome (Fig. 1c). Based on rpsC gene (S3 ribosomal protein) relative abundance, this genome is a low-abundance member of the community, reaching 2.6% relative abundance 313 days after fracturing. Given this low abundance and that this system was operated in a closed fashion (limited perturbation after initial hydraulic fracturing event), we consider it likely that the genome was not recovered in other samples due to it being below the detection level needed for assembly. Supporting this assumption, mapping reads from the other temporal samples to the day 313 genome demonstrated read recruitment but no assembled contigs from the day 152 metagenome. This organism therefore appears to be a rare member of the fractured shale community (Table S1).

A novel genus of Verrucomicrobia recovered from fractured shale.From metagenomic data, we reconstructed a nearly complete, high-quality (42) (>99% estimated completeness with <1% contamination based on the presence of single-copy genes (34) from a genome belonging to the Verrucomicrobia phylum from the day 313 sample. The recovered genome is 4.10 Mbp in size with an average GC content of 62.3%. It has 3,400 genes coding for predicted proteins on 40 contigs (>5 kb) and includes a 1,425-bp 16S rRNA gene as well as 5S and 23S rRNA genes and 44 tRNA genes. This represents the first nearly complete genome from the Verrucomicrobia recovered from hydraulically fractured shales and therefore offers a unique opportunity to understand genome-encoded traits that may extend the depth and salinity range for members of this phylum.

16S rRNA gene and concatenated gene phylogenetic analyses revealed this genome belonged to a new genus within the family Puniceicoccaceae of the order Opitutales in the Verrucomicrobia phylum (Fig. 2 and Fig. S1). Following the naming convention of MAGs (43), we propose the name “Candidatus Marcellius,” after the Marcellus shale formation it was recovered from. The nearest cultivated representative is an obligate aerobe isolated from seawater surrounding hard coral, Coraliomargarita akajimensis strain DSM 45221, sharing 87% 16S rRNA gene sequence identity. Supporting the new genus designation (15), “Ca. Marcellius” had a maximum of 52% average amino acid identity (AAI) to all other publicly available Verrucomicrobia genomes estimated to be >75% complete (May 2019) (Table S2). Despite a bootstrap value of 44 between “Ca. Marcellius” and Coraliomargarita akajimensis in our phylogenetic tree constructed from 11 ribosomal proteins (Fig. 2), these organisms share 48% genome-wide similarity by AAI, well below the genus boundary of 55 to 60% defined in Rodriguez-R et al. (44). Furthermore, analysis using the Genome Taxonomy Database-Tool Kit (GTDB-tk [http://gtdb.ecogenomic.org/]), which assigns taxonomy to MAGs based on a phylogenetic tree of 102 concatenated single-copy marker genes, confirmed our phylogenetic inferences that “Ca. Marcellius” is the first representative of a novel genus in the order Optiutales.

• Open in new tab
• Download powerpoint

FIG 2

Phylogenetic tree of 32 nearly complete Verrucomicrobia genomes, their key attributes, and habitats of origin. “Ca. Marcellius” is highlighted in boldface. Metagenome-assembled genomes (MAGs) are denoted with asterisks. Square brackets denote orders containing cultivated representatives. According to Silva (v 132, May 2019), the phylum is comprised of five orders containing cultured representatives: 1, Optitulates; 2, Verrucomicrobiales; 3, Chthoniobacterales; 4, Methylacidiphilales; 5, Pedospherales. Order 6 represents the formerly acknowledged subdivision 5, the only cultivated member of which (Kiritimatiella glycovorans strain L21-Fru-AB) was recently assigned to the novel sister phylum Kiritimatiellaeota (27). The outgroup is Escherichia coli. The tree was constructed from a concatenated alignment of 11 conserved ribosomal genes (amino acid sequences; large subunits L5, L6, L14, L15, L16, L18, L22, and L24 and small subunits S3, S8, and S17) using 100 bootstraps. Bootstrap values equal to or greater than 85% are represented with black diamonds. Genome completeness was estimated from the presence of 31 single-copy genes (34). Genes for the nifHDKENB subunits coding for nitrogen fixation, a complete tricarboxylic acid (TCA) cycle, and complete glycolysis pathway were assessed for each genome (present, black circles; absent, white circles; 1 gene missing from complete set, gray circles).

Like the rest of the Verrucomicrobia phylum, members of the Puniceicoccaceae family are diverse in range of habitats and metabolisms. Other cultivated strains from this family inhabit seawater (14, 15), marine sand (45), the digestive tract of a marine clamworm (14), and a fermenter (46). The metabolic lifestyles of these cultivated strains range from obligately aerobic (45) to facultatively anaerobic (14), and some are known to perform dissimilatory nitrate reduction (46). To date, six Verrucomicrobia MAGs have been recovered from the terrestrial subsurface, three from shallow (5-m) aquifer waters (32) and three from deep (300- to 450-m) granitic groundwaters (26). However, phylogenetic analysis of these MAGs places half of these in the newly recognized sister phylum Kiritimatiellaeota (27) and not in the Verrucomicrobia phylum itself (Fig. S2). The recovery and analysis of “Ca. Marcellius” from the shale gas well reported here therefore extends the distribution of the Verrucomicrobia phylum to kilometers below the surface of the Earth and represents the first detailed analysis of a deep-biosphere Verrucomicrobia genome.

“Ca. Marcellius” is the only Verrucomicrobia detected in late-stage shale produced fluids.Given the recovery of the “Ca. Marcellius” genome from this shale gas well, we investigated the distribution of Verrucomicrobia 16S rRNA genes from our previously published hydraulically fractured shale wells (Table S3). Besides the Marcellus-5 well that “Ca. Marcellius” was assembled from, our 16S rRNA gene sampling included 37 time series samples collected from 4 shale wells in the Marcellus and Utica formations (33, 47). Surprisingly, our analysis yielded no evidence of members of the “Ca. Marcellius” genus in produced fluids from these other 4 wells. However, we did recover a total of fifteen other Verrucomicrobia amplicon sequence variants (ASVs) from two Utica shale gas wells (Utica-2 and Utica-3) (33, 40). These Verrucomicrobia were rare members of the community, ranging from 0.1% to 3.6% (average, 1.2%). Notably, these Verrucomicrobia were recovered from input and early produced fluids, specifically from flowback fluid collected before 17 days, when salinities were less than 40 g/liter (chloride [35]).

Like “Ca. Marcellius,” the majority of the fifteen shale Verrucomicrobia ASVs belong to the order Opitutales. These were divided equally between the families Opitutaceae (5 ASVs) and Puniceicoccaceae (5 ASVs). All Puniceicoccaceae ASVs had greatest amplicon alignment similarity to marine strain Cerasicoccus frondis YM31-066 (87 to 91% identity over 279 to 396 bp [48]) and a maximum of 84% identity to the 16S rRNA gene sequence recovered in “Ca. Marcellius.” Our results indicate that Verrucomicrobia, especially members of the Opitutales, are often present in low abundance in input fluids but rarely persist in fractured shales beyond the early stages of flowback, when salinities become elevated due to water-rock interactions. “Ca. Marcellius” therefore appears to be unique among the Verrucomicrobia in its tolerance of the extreme conditions of a fractured shale and can persist hundreds of days postfracturing.

Members of the Verrucomicrobia phylum may play roles in nitrogen cycling.Analysis of the “Ca. Marcellius” genome uncovered the capacity for aerobic respiration (complete TCA cycle, NADH dehydrogenase complex, and an electron transport chain with cytochrome c oxidase), yet we failed to detect the functional genes for dissimilatory nitrate (narGHI and napAB), nitrite (nirBD and nrfAH), sulfate (sat, aprA, and dsrA), thiosulfate (psrA), or tetrathionate (ttrA) reduction. Given the absence of detectable oxidized electron acceptors at the time of sampling, the prevalence of ferrous iron in the well (33), and the enrichment and cooccurrence with obligate anaerobes (e.g., fermenters like Halanaerobium [34, 47, 49] and methanogens [41]) (Fig. 1b), we infer this organism is likely a facultative anaerobe, with the capacity for anaerobic metabolism via fermentation. While we recognize nonobligate fermentation can be difficult to infer from genome alone, based on the surrounding metabolisms and geochemistry, absence of any anaerobic respiratory metabolisms encoded in the genome, and the capacity for fermentation by other Verrucomicrobia (27), we posit that fermentation enables “Ca. Marcellius” persistence in the shale ecosystem.

We also recovered evidence that “Ca. Marcellius” has genes to fix nitrogen to ammonia via the nifHDKENB complement, although we appreciate the presence of genes does not always confer nitrogen fixation ability (50). A number of ammonium-containing additives were included in the injected fluids of the Marcellus-5 shale gas well studied here (33 and www.fracfocus.org). Given the high cost of nitrogen fixation and the abundant fixed nitrogen in the system, we consider it unlikely the organism would be fixing nitrogen at the time sampled. However, if “Ca. Marcellius” is able to fix nitrogen, this and other fractured shale taxa may provide an important ecosystem service as these substrates become depleted over time.

We looked for putative nitrogen fixation capacity via the presence of nif genes in other nearly complete Verrucomicrobia genomes. Completion of these genomes was assessed on the presence of the same core set of single-copy genes used to assess “Ca. Marcellius” completion. Aside from the known nitrogen-fixing capacity of acidophilic methanotrophs Methylacidiphilum infernorum V4 and M. fumariolicum SolV (19, 51), “Ca. Marcellius” is the only other Verrucomicrobia genome (>75% complete) of the 32 analyzed with the complete nifHDKENB complement for nitrogen fixation (Fig. 2). Two less complete genomes not included in the phylum-wide comparison were also found to have all 6 nif genes: TH2747, isolated from the nutrient-poor Trout Bog freshwater lake in Wisconsin, USA (2), and Opitutaceae bacterium TSB47, recovered from the gastrointestinal tract of a termite. In summary, genomic potential for nitrogen fixation is patchy but encoded across clades within the Verrucomicrobia phylum. This potential for nitrogen fixation expands the metabolic repertoire of the Verrucomicrobia and indicates these taxa play roles in nitrogen cycling across habitats.

“Ca. Marcellius” has the genomic capacity to gain energy from chemicals injected during hydraulic fracturing.To analyze possible carbohydrate utilization by “Ca. Marcellius,” we inventoried substrates targeted by glycoside hydrolases (GHs), enzymes that hydrolyze glycosidic bonds between two or more carbohydrates. “Ca. Marcellius” has a total of 60 GHs inferred to degrade hemicellulose (43 GHs), starch (12), cellulose (4), and chitin (1). Many hemicellulose-specific GHs were annotated as galactosidases (18), fourteen of which belong to the GH family 2. We also detected three acetyl xylan esterase and three endoxylanase genes (GH families 8 and 210), as well as four genes encoding alpha-rhamnosidase (3 from GH family 78). Our results indicate “Ca. Marcellius” is able to use a wide range of polysaccharides as substrates but is especially optimized to degrade hemicellulosic polysaccharides.

We examined metabolic overlap between genes in the “Ca. Marcellius” genome and common polymers injected during well drilling and hydraulic fracturing processes (52). Cellulose- and starch-based additives are frequent constituents of drilling muds, used to add viscosity to the fluid (53, 54). While we were unable to confirm the composition of the drilling mud used in Marcellus-5, these muds likely contained similar organic polymers (53). Given that large volumes of drilling mud often remain in formation after drilling (54), it is possible that residual starch and cellulose from drilling muds support the growth of microorganisms with genetic machinery like “Ca. Marcellius” (53).

Polyacrylamide addition to shale formations is increasing, as this synthetic amide-based polymer is used as a friction reducer in slickwater hydraulic fracturing operations (52). This compound represents the only organic polymer we could confirm was added to the Marcellus-5 well and is known to support microbial growth (55). As such, we looked for evidence that “Ca. Marcellius” can degrade polyacrylamide using amidase enzymes that act on the carbon-nitrogen bonds in amide-based polymers. We found evidence for three amidase-encoding genes, suggesting “Ca. Marcellius,” like other taxa from fractured shales (40), degrade polyacrylamide.

Guar gum is one of the most volumetrically significant organic polymers added to input fluids and is used in traditional gel-based hydraulic fracturing operations (52). Although not used in hydraulic fracturing of the Marcellus-5 well studied here, we also found evidence that “Ca. Marcellius” can fully degrade guar gum (alpha-galactosidase and beta-mannosidase genes) and metabolize mannose, the major sugar that makes up the polymer (glucokinase and phosphomannose isomerase genes). In addition to polyacrylamide degradation, “Ca. Marcellius” has the capacity to degrade guar gum and ferment mannose to support growth, similar to other fractured shale taxa (56). Taken together, our findings highlight that degradation of exogenous organic polymers introduced in subsurface engineering operations likely facilitates microbial metabolism and growth in the deep subsurface.

Hemicellulose degradation is a common trait encoded across the Verrucomicrobia.Given the potential importance of polysaccharide degradation to the establishment of “Ca. Marcellius” in the fractured shale environment, we assessed this trait across Verrucomicrobia genomes from a diverse range of habitats (Fig. 3). Like many bacteria, all Verrucomicrobia genomes included in our analyses have the capacity for starch degradation. Notably absent from Verrucomicrobia genomes sampled were annotated genes for polyphenolic degradation, suggesting this metabolism is not a widely encoded function. GH enzymes common to the Verrucomicrobia include GH 5, GH 2, and GH 1, all common GH families with multiple functional capacities. Supporting our findings from a single genome, genes for hemicellulose degradation were by far the most enriched in genomes across the phylum. We recovered an average of 27 hemicellulytic GH genes/genome, with genes for cleavage of galactose monomers (GH 42) and xylose monomers (GH 10) being the most prevalent. Given that hemicellulose, a critical component of plant biomass, is one of the most abundant polymers on Earth (57), degradation of this compound could explain Verrucomicrobia persistence across host, soil, aquatic, and subsurface environments.

• Open in new tab
• Download powerpoint

FIG 3

Glycoside hydrolase (GH) genes in “Candidatus Marcellius” (boldface) and 31 other nearly complete Verrucomicrobia genomes. The “Ca. Marcellius” genome is particularly enriched in hemicellulose-degrading genes, a characteristic of other Opitutales representatives. Genome size (in megabase pairs, or Mbp) of each Verrucomicrobia genome is given in gray text next to each bar. Genomes are grouped by order (1, Optitulates; 2, Verrucomicrobiales; 3, Chthoniobacterales; 4, Methylacidiphilales; 5, Pedospherales; and 6, the formerly acknowledged subdivision 5, the only cultivated member of which [Kiritimatiella glycovorans strain L21-Fru-AB] was recently assigned to the novel sister phylum Kiritimatiellaeota [27]). Taxon names are color coded for their habitat of origin.

Habitat of origin does not appear to be linked to carbohydrate-degrading capacity in the Verrucomicrobia, as genomes with the most and least GHs were found in similar habitat types (Fig. 3 and Fig. S3). Generally, the smaller genomes are less enriched in GHs and the larger genomes more enriched. However, the largest genomes among those sampled here carried some of the smallest GH gene sets, notably “Candidatus Xiphinematobacter sp. strain Idaho Grape,” with only 4 GHs found in a genome 9.2 Mbp in length (Fig. 3 and Fig. S3). In contrast, we did find patterns of glycoside hydrolase number associated with taxonomy, regardless of genome size. In particular, Opitutales has the most genomes enriched in hemicellulose-degrading genes, including the novel genome “Ca. Marcellius” described here (Fig. 3). Furthermore, clusters of closely related genomes in subdivisions 1 and 7 share strikingly similar patterns of polysaccharide-degrading genes, regardless of their habitat of origin. For instance, host-associated genomes of Verrucomicrobium sp. strains BvORR106 and BvORR034 share glycoside hydrolases almost identical to that of the genome of freshwater-derived Verrucomicrobium spinosum DSM 4136 (Fig. 3). Similarly, genomes of acidophilic methanotrophs Methylacidiphilum infernorum V4 and M. fumariolicum SolV carry a set of starch- and cellulose-degrading genes similar to that of freshwater-derived Verrucomicrobium sp. strain 3C (Fig. 3). These findings suggest that the ability to degrade polysaccharides is more conserved along phylogenetic, rather than habitat, boundaries.

Despite the relatively limited genome sampling of this phylum, the capacity to degrade polysaccharides is emerging as a widely acknowledged trait among Verrucomicrobia genomes (2, 4, 5, 58). In agreement with our phylum-wide assessment (Fig. 3), various degrees of GH enrichment have been reported for less complete genomic representatives not included in our analysis. For example, Martinez-Garcia et al. used fluorescently labeled substrates, fluorescence-activated cell sorting, and single-cell genomics to identify 5 Verrucomicrobia genotypes that were active in polysaccharide degradation (59). These Verrucomicrobia genomes were found to be enriched in a wide spectrum of GHs and likely made a disproportionate contribution to active xylan degradation despite their low relative abundance in the community (59). In a similar fashion, it is possible that the low-abundance “Ca. Marcellius” organism plays a critical role in carbon processing in the fractured shale ecosystem.

These prior studies, when considered alongside our own assessment of polysaccharide degradation potential, provide compelling evidence that Verrucomicrobia are widespread contributors to the cycling of carbon. Microorganisms that degrade heterogeneous, polymeric organic matter, such as biological polysaccharides, act as gatekeepers to the microbial carbon cycle, providing downstream monomers (e.g., hexose or pentose) that serve as substrates for lower trophic levels (57). To date, our knowledge of Verrucomicrobia ecology is largely limited to freshwater and soil ecosystems. While these environments represent significant contributors to the global carbon cycle, our recovery of “Ca. Marcellius” from a deep rock-hosted, hypersaline ecosystem demonstrates that members of Verrucomicrobia extend across a greater range of habitats than previously considered. As such, the magnitude and contribution of Verrucomicrobia to the carbon cycle is only beginning to be appreciated.

Tolerance to environmental conditions may enable Verrucomicrobia to persist in the deep subsurface.The “Ca. Marcellius” genome was only detected between 250 and 313 days after fracturing, when TDS measured 104 g/liter on average over the period sampled, suggesting this organism should have mechanisms for salinity tolerance. Genome analysis for osmoprotection strategies (Table S4) revealed the presence of genes for both salt-in and osmolyte salt tolerance mechanisms (60). First, “Ca. Marcellius” contained genes for the salt-in strategy (nhaC, mnhA, phaD, and trkH), whereby ions like potassium accumulate in the cell to balance with the surrounding environment (Fig. 4). “Ca. Marcellius” also has genomic potential to employ the compatible solute method, whereby an organic molecule (e.g., amino acid or sugar derivative) accumulates by transport or de novo synthesis to maintain osmotic pressure within the cell. “Ca. Marcellius” has the capacity to take up but not degrade known osmolytes, including proline (proW and proV), suggesting this compound can be transported intracellularly for osmoprotection. We also note that it is possible that proline could be incorporated into proteins, as its sole use as an osmoprotectant is not possible to determine from genome sequence alone. The genome also carries genes encoding transporters for osmoprotectants like maltose (malK) and glycerol-3-phosphate (upgC); however, given that genes for the metabolism of these compounds are also present, use of these compounds in osmoprotection is challenging to infer. This genome also has the capacity to synthesize nitrogenous osmolytes like (i) proline from glutamate (proAB) and (ii) glycine betaine from glycine (bsmA, sarcosine/dimethylglycine N-methyltransferase).

• Open in new tab
• Download powerpoint

FIG 4

Genes encoding the compatible solute osmoprotectant strategy are common in Verrucomicrobia. Osmoprotectant genes detected in “Ca. Marcellius” (boldface) are compared to the 31 other nearly complete Verrucomicrobia genomes, grouped by order (1, Optitulates; 2, Verrucomicrobiales; 3, Chthoniobacterales; 4, Methylacidiphilales; 5, Pedospherales; and 6, formerly acknowledged subdivision 5, the only cultivated member of which [Kiritimatiella glycovorans strain L21-Fru-AB] was recently assigned to the novel sister phylum Kiritimatiellaeota [27]). Genes are listed on the x axis, and genomes are listed on the y axis. Colored boxes indicate the presence of that gene in the genome. The different strategies for osmoprotection are listed at the top and are represented by color (salt-in, blue; osmolyte transport, green; osmolyte production, red). *, SDMT, sarcosine/dimethylglycine N-methyltransferase. All other gene descriptions are given in Table S4.

Nuclear magnetic resonance analyses of metabolites from forty different shale-produced fluids support the osmoprotectant strategies identified in the “Ca. Marcellius” genome. For instance, metabolite data have demonstrated that glycine and glycine betaine were widely recovered in shale-derived produced fluids, including those from the Marcellus-5 well at the time “Ca. Marcellius” was recovered (33). Moreover, the capacity for shale microorganisms, like Methanohalophilus, to synthesize glycine betaine has been shown in laboratory microcosms (33, 41). When released into the surrounding environment (e.g., by viral lysis), these compounds can be used as growth substrates, sustaining a metabolic network composed of key shale members, including Halanaerobium (33). Based on analyses presented here, we further entertain the possibility that “Ca. Marcellius” could contribute to in situ synthesis of critical osmolytes/substrates.

Other Verrucomicrobia genomes have been recovered from hypersaline environments. The first hypersaline Verrucomicrobia genome was an isolate, Kiritimatiella glycovorans strain L21-Fru-AB (21), recovered from a photosynthetic mat (salinity of 170 g/liter at the time of sampling) (61). While it was originally posited, in part due to the lack of other saline-tolerant Verrucomicrobia, that strain L21-Fru-AB should be a new phylum (Kiritimatiellaeota) (27), we have included it in our phylum-wide analysis given its similarly hypersaline habitat of origin. In line with their recovery from environments with similar salinities, “Ca. Marcellius” (116 g/liter TDS) and K. glycovorans genomes share nearly the same osmoprotectant gene profile (10/11 genes), a gene pattern not conserved across the remaining Verrucomicrobia genomes (Fig. 4). Interestingly, both saline-derived genomes contain a glycine betaine synthesis pathway, suggesting this pathway is an important trait for Verrucomicrobia residing in hypersaline habitats.

Our comparative genome analyses revealed osmoadaptation is broadly encoded across other Verrucomicrobia subdivisions. We found no correlation between osmoprotectant gene number and habitat origin or phylogeny. For instance, the number of osmoprotectant genes in freshwater environments range from two (ME12756) to thirteen (TH4590) genes and those associated with hosts range from one (“Candidatus Xiphinematobacter sp. strain Idaho Grape”) to eleven (Verrucomicrobium sp. strain BvORR106). Osmoprotectant strategies like those we report here enable more than salinity adaptation, they also allow microorganisms to withstand osmotic stressors like those caused by desiccation in wet and dry seasons in soils (62) and along brackish coastal estuaries. Therefore, resistance to environmental perturbations may allow members of this phylum to colonize and withstand dynamic habitats like those in fractured shales, the gut, and soils.

A large number (32) of Verrucomicrobia MAGs were recently recovered from soda lake sediments (28) and a hypersaline pond (29). The salinity of these environments ranges from 42.9 to 110 g/liter TDS, similar to that from which the production fluids of “Ca. Marcellius” were recovered. These MAGs were not publicly available at the time we conducted the phylum-wide comparative genomics analyses reported above. We calculated similarity to “Ca. Marcellius” using AAI (Table S2) and constructed an additional phylogenetic tree (Fig. S2). The closest AAI identity is 52% (CSSed10_295, 88% complete [28]), placing it within the Puniceicocacceae family but not within the “Candidatus Marcellius” genus. Nine other hypersaline MAGs clade with “Ca. Marcellius” in the Opitutales order (Fig. S2); however, the majority appear to belong to the newly established sister phylum Kiritimatiellaeota and, hence, are no longer recognized as Verrucomicrobia according to SILVA taxonomy (release 132) (3). Nevertheless, the recovery of a large number of MAGs belonging to Verrucomicrobia and its sister phylum from hypersaline environments suggests a more widespread tolerance to high salinity than previously appreciated. The recovery of “Ca. Marcellius” from hydraulic fracturing production fluids therefore extends not only the environmental distribution but also the salinity tolerance of the Verrucomicrobia phylum.

Evidence for Verrucomicrobia-virus interactions.Our prior research (33, 34, 47) suggests viral immunity is an important factor to persisting in shale ecosystems. Evidence for this includes (i) the recovery of CRISPR-associated proteins in all persisting genomes (34), (ii) active spacer incorporation within CRISPR-Cas arrays from genomes sampled over time in produced fluids (34, 47), and (iii) expressed CRISPR-Cas genes in response to active viral predation (33). Here, we explore the possible relationships between viruses and members of the Verrucomicrobia.

We mined the nearly complete Verrucomicrobia genomes for the presence of CRISPR-Cas systems and found that CRISPR-Cas immunity was present in 32% of the genomes. Of the CRISPR-Cas systems that could be classified, we recovered both types I and III. Interestingly, only Verrucomicrobium sp. strain BvORR034 (from the root of a sugar beet) and “Ca. Marcellius” contained both types. Within “Ca. Marcellius” we detected six CRISPR-Cas arrays containing a total of 189 spacers. We previously created a database of viruses from hydraulically fractured shale wells consisting of 1,838 viral genomes from 40 metagenomes (47) and found that none of the 189 “Ca. Marcellius” spacers matched this viral database. This indicates that “Ca. Marcellius” experienced viral predation in a different habitat before its recovery from the hydraulically fractured well or that the viruses preying on “Ca. Marcellius” were in too low abundance to be sampled by our efforts. Further, no spacer matches from the “Ca. Marcellius” genome to viral sequences in the NCBI RefSeq database were found. We must consider, however, that like the host genomes of this phylum, Verrucomicrobia viruses are likely also undersampled. Taken together, it appears that “Ca. Marcellius” has the genomic mechanisms to withstand the viral predation known to be active in fractured shale ecosystems.

Despite the presence of a functional CRISPR-Cas-acquired immune system in the “Ca. Marcellius” genome, we also detected genes indicative of prophage or viruses integrated into the host genome (Table 1). A genome-based network analysis of shared protein content classified these as members of the double-stranded DNA viral order Caudovirales. These two prophage regions were located on scaffold_15 and scaffold_45, spanning ∼39 kbp and ∼18 kbp, respectively, when analyzed using VirSorter (63, 64). Examination of these genomic regions indicate that they contain multiple genetic defects and are remnants of degraded prophage, lacking genes required for phage integration, induction, and propagation (65). For example, the defective prophage region on scaffold_15 contains only three genes annotated as phage: two Mu-like phage proteins and a large terminase subunit (genes 32 to 34). Flanking the phage proteins are a series of short genes annotated as unknown function or hypothetical proteins (∼300 to 450 bp), with best gene hits to Verrucomicrobia genomes. The defective prophage region on scaffold_45 contains a single gene annotated as phage, a 281-bp fragment of a terminase large subunit, and six sequential transposase genes (genes 2 to 7). Collectively, we instead posit that these regions are likely rapidly evolving parts of the genome due to interactions with mobile genetic elements, including prophage.