Microbial Community Composition in Geothermal Outflow Channels
Conch and Octopus Springs are located within several km of each other in the Lower Geyser Basin of YNP. Although both springs are alkaline chloride (siliceous) geothermal channels (Figure 1) with similar velocity, temperature, pH and baseline geochemistry (Supplemental Table 1), Conch Spring is highly sulfidic (> 120 µM dissolved sulfide (DS) at discharge) while Octopus Spring contains near detectable levels of DS (< 1-2 µM) (Supplemental Figure 1). Conversely, Octopus Spring contains low levels of dissolved oxygen (DO ~ 10-20 µM) at discharge and 20-60 µM DO at 82-84 oC, where filamentous biofilms were sampled. Numerous alkaline chloride springs are found in YNP but the amount of sulfide at spring discharge can vary based on the extent of steam separation and concentration of gases via subsurface boiling36,37. This geophysical and geochemical circumstance, which results in a direct comparison of sulfidic versus oxic conditions under otherwise nearly identical habitats, provides an important case study for understanding the impacts of sulfide versus oxygen on microbial community structure, genomics, and physiology.
Filamentous biofilms from Conch and Octopus Springs both contain highly abundant populations of Thermocrinis (Aquificota), Pyrobaculum (Thermoproteota) and Caldipriscus (Pyropristinus) genera (Figure 1). Together, these three populations comprise from ~ 50 to over 80 percent of the Octopus and Conch microbial communities, respectively. Abundance data has been consistent over multiple samples taken from 2008 to 2017 (see replicates, Supplemental Figure 2, 3). However, each of these extreme communities contains a balance of microbial populations that are specific to each habitat: only 3-4 additional populations are present in sulfidic Conch communities while Octopus biofilms contain considerably greater diversity for a total of 15 different ‘species’ (Table 1, Supplemental Table 2). T-stochastic neighbor embedding (tSNE) plots of tetranucleotide frequencies from assembled metagenomes obtained in 2011 and 2012 reveal consistent compositional differences between these two microbial communities (Supplemental Figure 4). Moreover, additional sequence coverage obtained in 2012 (full lane Illumina HiSeq) versus 2011 (10 % lane) resolved several low-abundance populations (e.g., a second patescibacterial population in Octopus 2012, and several low-abundance MAGs in Conch 2012). A direct comparison of tSNE plots from both habitats shows the increased resolution and sequence coverage of all microbial populations when full-lane metagenomes were compared to samples that received 10% of a full-lane (Supplemental Figure 5ab). Both sets of MAGs are highly complete and either set (2011 or 2012) revealed essentially identical phylogenetic and functional results. The predicted temperature optima of each MAG fall within ~ 10 oC of the habitat temperature; moreover, the sum of abundance-weighted temperature optima was used to predict site temperatures and these were remarkably close to the 82-84 oC temperatures measured at the time of sampling (Supplemental Table 3).
Microbial populations present in Conch Spring but not in Octopus included Thermodesulfobacteria and several members of the Desulfurococcales (Table 1, Figure 1). Conversely, Octopus Spring contains 10-12 microbial populations not seen in Conch Spring; these include primarily additional bacteria (Armatimonadota(2), Calescibacterium, Thermoproauctor, Thermoflexus, Patescibacteria (2)) as well as additional archaea within the Aigarchaeota (Calditenuis aerorheumensis38) and Geoarchaeota39. As will be shown, these additional populations are aerobic in the Octopus Spring habitat. Both habitats also contain low amounts (< 1 % abundance) of very similar populations identified here as Thermus aquaticus, Acidilobaceae and Nanopusillus (Table 1, Supplemental Table 2).
I-tag data (~ 300 base pair 16S rDNA) was also obtained for these biofilm communities from the same sample used for the October 2012 metagenomes (Supplemental Table 4). Comparison of the i-tag dataset against metagenome abundance data shows that the single universal primer set used for 16S rRNA gene amplification adequately represents some of the predominant populations but fails to amplify all populations detected using random sequencing. Several microbial populations are missed completely using a single primer, even the highly abundant Pyrobaculum populations. This is due primarily to several introns in the 16S rRNA gene of Pyrobaculum40, which impede primer efficacy. These results are a reminder of the deficiencies of short-fragment 16S i-tag analysis for assessing microbial community structure.
The microbial community composition in Octopus Spring was considerably more diverse than in Conch Spring, which can be traced primarily to the higher levels of dissolved oxygen. Nearly all populations in Octopus Spring contained terminal oxidase complexes with a HCO (subunit 1), and most of these aerobes are heterotrophic, as indicated by the absence of chemolithotrophic markers (e.g., Figure 4). The presence of numerous additional heterotrophs in Octopus Spring explains 13C-biomass values obtained from these same biofilm communities in 2011-201341. Stable-isotope (13C) mixing models revealed that Conch Spring biomass was nearly entirely (> 90%) comprised of organic C originating from inorganic C, whereas Octopus Spring biomass could be explained by greater uptake of ‘photosynthetic’ organic C (up to 40%). Higher dissolved oxygen in the flow channels of Octopus Spring (20-60 µM) were directly correlated with the presence of Type 1 HCOs, and other associated metabolisms that require oxygen such as the oxidation of arsenite). Consequently, higher oxygen levels in Octopus Spring promote a more diverse, aerobic, and heterotrophic microbial community than present in the highly sulfidic Conch Spring, where DO levels were below detection ( < 1 µM).
Phylogenetics of Deeply Rooted Thermophiles
The Pyropristinus lineage contains two major groups of early evolved thermophiles that have not been fully recognized, in part because some of the previously described MAGs (i.e., Pyropristinus T242) were not sufficiently complete. However, the substantial number of highly complete Pyropristinus entries included here are derived from Illumina sequencing of metagenomes from Conch and Octopus Springs (2011-2012) as well as several other habitats in YNP that we have characterized with similar geochemical conditions (pH ~ 7-8, temperature = 75-85 oC). A bacterial phylogeny based on a standard group of 16 ribosomal proteins (Figure 2a) shows that Pyropristinus is the most deeply rooted bacterial lineage independent of the Candidate Phyla Radiation (CPR)43. The Pyropristinus group also includes the WOR-like populations described in the Genome Taxonomy Database (GTDB44) (Figure 2a). A bacterial phylogeny based on the 16S rRNA sequence (> 1,000 bp, Figure 2b) also shows that Caldipriscus and Thermoproauctor represent two different groups within the Pyropristinus lineage42, which is earlier evolved than either the Aquificota and/or Thermotogota, often considered to be among the earliest bacterial lineages33 (Figure 2). A phylogenetic tree including the archaeal entries from Conch and Octopus Spring is not included here since the phylogenies of the predominant archaea present in these habitats have been discussed previously (i.e., Desulfurococcales45; Pyrobaculum46; Calditenuis aerorheumensis38).
Highly Redundant ‘Keystone Species’
The Thermocrinis and Pyrobaculum MAGs found in the flow-channels (82-84 oC) of Conch and Octopus Springs were highly redundant and contained several highly related (> 90% ANI) species and/or ecotypes based on the detection of multiple copies of closely related single copy genes (Table 1, Supplemental Table 2). Thermocrinis and Pyrobaculum MAGs ranging from 5 - 8 Mb may contain at least three sub-populations as estimated from the average number of non-redundant single copy genes (CheckM47), which yielded estimates of redundancy near 200% (Table 1). Thermocrinis MAGs obtained from Conch and Octopus metagenomes (years 2011 and 2012) share core sequence content with other Thermocrinis spp. (Figure 3a). However, there is significant genome sequence within these MAGs that is unique to YNP as well as sequence that is unique to each spring. Thermocrinis ruber (genome size = 1.52 Mb) was isolated from Octopus Spring48 but it only represents a subset (~ 30 %) of the total Thermocrinis-like sequence present in the MAGs obtained from this same community. Likewise, the total Thermocrinis sequence (MAG) can be subdivided based on a mix of coverage and tetranucleotide frequency (a’nvio49); one possible solution results in at least three sub-populations, each with smaller genome size, considerably lower redundancy and a range of completeness estimates (Figure 3b). Lower completeness estimates in subdivided bins indicate the difficulty of separating highly related species and/or ecotypes using metagenome data. Although subject to numerous uncertainties including the importance of coverage versus tetranucleotide frequency, this process consistently suggests the presence of several less-redundant Thermocrinis MAGs (ecotypes) in all metagenomes from Octopus and Conch (Supplemental Figure 7). Similar analyses of Pyrobaculum sequence content in Conch and Octopus Springs also show that several species and/or ecotypes may be present in these communities, which contributes to large MAG sizes circa 4 Mb (Table 1). The complete genome sequence of Pyrobaculum yellowstonensis WP30, which was isolated from YNP46, is only 2.1 Mb and represents just a subset of the Yellowstone Pyrobaculum pangenome sequence26,50. Definitive separation of highly similar sequence content (ANI > 95%) into different ecotypes (or type strains specific to variation in key environmental parameters) is inherently problematic in natural communities using random metagenome sequence because there is no guarantee that assembled sequence originates from a singular clonal population. Numerous microbial isolates and/or single amplified genomes (SAGs) from the same environment help to reveal the variation of individual cell types within a larger population of highly related strains50,51.
Functional Genomics and Biogeochemical Cycling
A summary of KEGG pathway analyses focused on electron transfer reactions involving arsenic, sulfur, nitrogen, and oxygen for MAGs from Conch and Octopus Springs (Figure 4) revealed key similarities and differences between the two communities as well as between the three highly abundant genera that were present in both habitats. A broader look at energy related genes using hierarchal clustering (Supplemental Figure 8) shows similar results, and high reproducibility among phylotypes found in both habitats (i.e., Thermocrinis, Pyrobaculum, Caldipriscus). Both communities contained genes for utilizing sulfur species via several enzymes and/or pathways including the sox oxidation pathway, a sulfide:quinone oxidoreductase (sqr), tetrathionate reductase (ttr), polysulfide/sulfur reductase (psr/sre), and the dissimilatory reduction of sulfate (dsr). Arsenite oxidase genes (aio) were present in Thermocrinis MAGs from both habitats, but for Pyrobaculum, aio genes were only present in Octopus Spring, the more oxic habitat. Importantly, communities from both Conch and Octopus Springs contain phylotypes with genes required for respiration on oxygen using either Type 1 HCO complexes, bd ubiquinol oxidases, or cytochrome AA’ (present only in archaea). The presence and absence of different oxygen reductases from Conch versus Octopus suggests that higher affinity oxygen reductases are more prevalent in Conch Spring. Protein modeling of the subunit I HCOs (not shown) present in Caldipriscus and Thermoproauctor (Pyropristinus lineage) as well as for Thermocrinis and Pyrobaculum indicate that they each possess Type 1 HCOs, which are expected to reduce oxygen to create proton motive force52. The complete absence of Thermoproauctor under sulfidic conditions (Conch Spring) is likely due to the lack of a high-affinity cytochrome bd ubiquinol oxidase (Figure 4) and/or a direct toxicity due to high sulfide.
Both habitats contain organisms with the metabolic potential for the fixation of carbon dioxide via either the reverse TCA cycle, e.g., ccl/ccs = citryl-CoA lyase/citryl-coA synthetase33) or other carboxylases (e.g., acc = acetyl coA carboxylase; por = pyruvate oxidoreductase). Most of the bacteria from these habitats contain either the genes necessary to synthesize biotin or possess the biotin specific transporter (BioY); biotin is a necessary cofactor for carboxylation, especially in bacteria53,54. Genes for different steps of nitrogen reduction (e.g., narG, nirK, norB, nosZ) were also present in both habitats but distributed across several different phylotypes (Figure 4). Consequently, it is apparent that no single population contains genes necessary for complete denitrification, but each community appears capable of this metabolism.
Evidence of functional diversity among members of these communities was also examined using hierarchal clustering of carbohydrate-active enzymes (CAZy) as well as specific transporters (Supplemental Figures 9-10). The CAZy database55 contains enzymes that either build or breakdown complex carbohydrates and glycoconjugates, including glycoside hydrolases, glycosyltransferases, polysaccharide lyases, carbohydrate esterases and carbohydrate-binding modules. The presence of diverse CAZy-related genes within each community reveals broad metabolic potential for carbohydrate metabolism distributed across different populations. Glycosyl transferases (family GT4, Family GT5) and glycoside hydrolases (Family GH109) were especially important in the two Armatimonadota populations as well as the Calescibacterium in Octopus Spring (Supplemental Figure 9). A broad distribution of different transporter genes across the MAGs also supports extensive diversification with respect to metabolic processes and/or specific auxotrophic requirements, including vitamins, nucleotides, carbohydrates, amino acids, lipids and various inorganic elements such as phosphate and Mo (Supplemental Figure 10). Finally, the acquisition of sufficient Fe can be problematic in high pH environments due to the low solubility of ferric iron solid phases; concentrations of soluble Fe in Conch and Octopus Springs were indeed below detection using ICP-OES (~ 1 µM), and so it may be expected that numerous phylotypes in these habitats showed genomic capabilities for enhanced Fe transport, siderophore production and transport, as well as Fe gene regulation (Supplemental Figure 11). No evidence of Fe reduction and/or Fe oxidation existed in these communities.
Virus Sequence and CRISPR-Cas Systems
Candidate virus sequence was identified using both VirSorter256 and geNomad57, then compared with sequence databases to determine the importance of possible viruses on members of each microbial community (Table 2). The amount of candidate virus sequence and/or the number of ‘hallmark’ viral genes found within each metagenome assembly (2012 reported in Table 2) was not consistent using these two different algorithms. The number of viral contigs and total viral sequence within a MAG predicted using VirSorter2 was highly correlated (r > 0.8) with the number of CRISPR arrays identified using independent approaches (Minced). Conversely, geNomad output did not yield significant correlations, possibly due to poor recognition of archaeal viruses. The Thermocrinis and Pyrobaculum MAGs, which were highly redundant and exhibited large genome sizes (e.g., Table 1) contained the most viral contigs and ‘hallmark’ viral genes predicted by VirSorter2. Moreover, these MAGs also contained the highest numbers of CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) arrays compared to other MAGs from these hot springs.
The large majority of viral contigs were not related (ANI < 50%) to sequences found in public databases except for contigs related to two Aquificales viruses58 and Pyrobaculum Spherical Virus (PSV59). Contigs found in the Thermocrinis MAGs and nearby tSNE clusters (OS3173 virus, Supplemental Table 2) were successfully mapped back to the known58 circular virus OS3173 (Figure 5a). The phylogeny (DNA PolA; Figure 5b) of OS3173-like viruses indicated two similar entries from Octopus Spring also consistent with the two copies mapped to OS3173-like viruses. A similar OS3173-like virus was also found in Conch Spring but a separate ‘Conch37’ virus was not found in Octopus Spring (Figure 5b). This family of viruses was also noted in other habitats that support different genera of the Aquificales, such as the Sulfurihydrogenibium populations from Yellowstone Lake34 as well as the Hydrogenobaculum populations important in acidic habitats (e.g., JC39 from Joseph’s Coat Springs, Palmer et al. 2020). The other OS3173-like sequences from Perpetual Spouter and Great Boiling Spring (GBS) track with Thermocrinis populations.
The number of CRISPR-associated proteins (Cas) (Supplemental Figure 12) was also highly correlated with the number of CRISPR arrays and VirSorter2 viral sequence (Table 1, 2). CRISPR-Cas proteins of different types were especially important in Thermocrinis and Pyrobaculum populations. Specifically, Thermocrinis exhibited Csm and Cmr proteins indicative of Type III CRISPR systems60,61. In fact, Thermocrinis transcripts reveal expression of complete Csm and Cmr systems in both 2011 and 2016, which indicates that these CRISPR-associated proteins were actively being used in viral defense. CRISPR arrays, specifically their non-repetitive spacers, contain a record of previous viral encounters in the form of short viral sequences. To identify the hosts for Conch and Octopus viruses, we tested the match between the CRISPR spacers and predicted viral sequences. We found 189 contigs with CRISPR spacers that matched the OS3173-like MAG from Octopus Spring. Of those, 176 contigs were < 2000 bp and were not binned, but the remaining 13 contigs > 2000 bp contained 33 matches to OS3173-like contigs, and were all binned in the Thermocrinis MAG. Notably, one CRISPR array with 29 spacers had 9 hits to OS3173-like contigs. Consequently, Thermocrinis is the most likely host for this virus, which is consistent with previous results58. Spacers from 4 different contigs had matches to ‘Conch37’ virus sequences. Only one of those contigs was larger than 2000 bp and was also binned with the Thermocrinis MAG. This CRISPR array had 31 spacers, of which 6 matched ‘Conch37’; these results also suggest that the candidate virus ‘Conch37’ from Conch Spring likely infects Thermocrinis organisms. This clade of Aquificales-related viruses is clearly important in the ecology and dynamics of naturally occurring Aquificales populations in YNP (Figure 5b).
CRISPR-associated proteins were also observed for Calescibacterium, Thermoproauctor, Thermodesulfobacteria, Thermoflexus, Acidilobaceae and the Armatimonadota populations (Type 1, Type 2). The Caldipriscus population, which was common to both Conch and Octopus Springs, exhibited very low redundancy and small estimated genome sizes (1.5-1.8 Mbases); notably, this population contained few CRISPR arrays and/or Cas proteins (Table 1, Supplemental Table 2, Supplemental Figure 12). These results consistently show that viruses play a significant role in the genomic characteristics of many community members62, but not all. There was no evidence for CRISPRs in Nanopusillus and Patescibacteria MAGs, organisms with reduced genome sizes.
Metatranscriptomics: Metabolic Activity
Thermocrinis, Pyrobaculum and Caldipriscus MAGs accounted for ~ 80-90 % of the total transcript sequence mapped to community populations from Conch and Octopus Springs (Figure 6a). The large number of transcripts mapped to these highly abundant and active populations provided sufficient coverage to evaluate the activity of numerous cellular processes (Supplemental Tables 5-7 provide annotation of all mapped transcripts for each MAG). Although transcripts were observed for most community members identified in metagenomes (e.g., Table 1), insufficient coverage precluded a thorough characterization of metabolic activity in the remaining phylotypes, which only comprised ~ 10-20% of the total transcripts sequenced (Figure 6a). Plots of metagenome abundance versus transcript abundance show that the recovery of RNA for numerous populations was not representative of abundances observed from DNA extraction (Supplemental Figure 12) and reflects inherent biases in the recoveries of RNA versus DNA. Replicate transcriptomes from Octopus Spring collected in 2011, 2014 and 2016 were highly correlated (p < 0.01) with one another (Supplemental Figure 13, 14), which is further evidence of the overall stability of these microbial biofilms; replicate transcriptomes from 2014 also confirmed reproducibility of RNA extraction and sequencing steps; transcriptomes from 2011 and 2016 were highly similar (p <0.001), and the fraction of transcripts mapped to Thermocrinis, Pyrobaculum and Caldipriscus was remarkably consistent (Figure 6a).
Transcript abundances for specific energy and electron transport genes for Thermocrinis, Pyrobaculum and Caldipriscus populations in Conch versus Octopus Springs reveal several major shifts in metabolism between microaerobic and sulfidic conditions (Figure 6b). Nearly 2-3 % of transcripts mapped to Thermocrinis corresponded to arsenite (Aio) and sulfur oxidases (Sox) in Octopus Spring but these same genes were not transcribed in Conch Spring. Higher transcription levels of sulfide: quinone oxidoreductase (Sqr), high-affinity bd ubiquinol oxidases, rhodanese sulfur transferase domains and a hemoglobin gene (oxygen binding) were observed in Conch Spring, indicating a major shift in energy metabolism yet still dependent on oxygen. Arsenite oxidases were also highly transcribed in Pyrobaculum populations in Octopus Spring, yet this activity was completely absent in Conch Spring despite nearly identical levels of total soluble arsenic. Moreover, Pyrobaculum HCOs (and associated cytochrome bc complexes) were only transcribed in Octopus Spring, while the high affinity cytochrome CydAA’ were highly transcribed in Conch Spring. The CydAA’ cytochromes have recently been shown29 to serve as bona fide oxygen reductases and are important in numerous archaea26,63. The Caldipriscus population may employ a slightly different strategy for dealing with low oxygen levels in Conch Spring. In addition to high expression levels of Sqr and high-affinity bd ubiquinol oxidases, Caldipriscus also expressed high levels of an HCO and cytochrome bc complex in the sulfidic Conch system. The major shifts in respiration pathways observed for these three common populations reveal a tight linkage between the availability of oxygen and the metabolic strategies employed for energy conservation in these two contrasting environments.
Arsenite serves as an important electron donor in two predominant organisms (Thermocrinis and Pyrobaculum) from Octopus Spring, yet this metabolism was not active in Conch Spring at the same concentrations of soluble arsenic (~ 20 µM). The ratio of dissolved oxygen: sulfide ranges by a factor of 1000 from ~ 0.02 in Conch Spring to > 20 in Octopus Spring. Consequently, the low concentrations of dissolved oxygen coupled with high sulfide in Conch Spring precludes the oxidation of significant amounts of arsenite using oxygen as a terminal electron acceptor. Arsenite oxidases were previously shown to be expressed in several Aquificales-dominated filamentous communities using reverse transcriptase-PCR64, especially in oxic geothermal channels; consequently, the large fraction of transcripts mapped to aioAB genes observed here is not only consistent with prior observations but re-enforces the importance of arsenite as an energy source for chemolithoautotrophs, especially under microaerobic conditions. Thermocrinis ruber was cultivated from Octopus Spring48 and has been shown to oxidize arsenite aerobically in pure culture65.
The Sox sulfur oxidation pathway66 was also highly expressed in Thermocrinis , but again, only in Octopus Spring where > 1% of Thermocrinis transcripts mapped to the Sox genes (Figure 6). Although dissolved sulfide levels were very low and near detection in Octopus Spring (1-3 µM DS at 82-84 oC), the likely substrate for Sox oxidation is thiosulfate (SoxB/C/Y66, which appears to be more readily oxidized in the presence of oxygen versus 100 µM sulfide in Conch Spring. The Thermocrinis Sox complex was not expressed in Conch Spring under high levels of sulfide, again likely due to insufficient oxygen. Conversely, sulfide: quinone oxidoreductases were only expressed significantly in Conch Spring, and this was observed consistently for each of the three microbial populations common to both springs.
Perhaps the most striking shift in metabolism between the two springs occurred with HCOs, cytochrome bd ubiquinol oxidases and the archaeal cytochrome CydAA’, which is like the high affinity bd ubiquinol oxidases but only present in archaea29. A major metabolic shift from Type 1 HCOs to higher affinity oxidases occurred in the presence of high sulfide. The cytochrome bd ubiquinol oxidases (Thermocrinis, Caldipriscus) or the cytochrome CydAA’ oxidase (Pyrobaculum) were highly expressed in Conch Spring, where dissolved oxygen levels were below detection (< 1 µM) and likely in the nanomolar range. Interestingly, only the HCO of Caldipriscus remained highly expressed in Conch Spring, along with high expression of the bd ubiquinol oxidase. Otherwise, there was a notable absence of low-affinity Type 1 HCOs in Conch Spring.
Type IV Filament Systems: A Community Anchor
One of the most striking features of high-temperature filamentous biofilms is the high abundance of extracellular structures resembling pili with diameters of ~ 10 nm (Figure 7, Figure 1, Supplemental Figure 15). Although Thermocrinis MAGs also contained genes required for production of flagella (Supplemental Figure 16), numerous genes important in the synthesis of Type IV filament machinery67–69 were highly expressed in Thermocrinis populations in both Conch and Octopus Spring (Figure 7). Genes known to be involved in Type IV filament production and activity were found on two different contigs and were observed in numerous metagenomes as well as Thermocrinis ruber, which was isolated from Octopus Spring48. Highly conserved topology included genes pilA, pilW, pilY, pulG, and FimT, which are important components of Type IV filament (Tff) systems69–71 and encode various proteins important in the synthesis, secretion and function of pili. Very high transcript levels (%) were observed for various components of these Tff systems including from 1 to over 5 % of transcripts for PilA and from 0.6 to 2.5% of transcripts for NfuA. PilA is the protein comprising actual major pilin structure69 and NfuA is an Fe/S assembly protein used in maturation and repair of Fe-S proteins, which is often associated with Fe stress72,73. These Thermocrinis transcripts were among the most abundant from either Conch or Octopus Springs and confirm the visual evidence that large amounts of cellular resources are being directed to the formation and maintenance of an extensive Type IV filament network. Specifically, the fixation of inorganic C via citryl co-A lyase and citryl-CoA synthetase (succinyl-CoA synthetase)33 was also highly expressed in Thermocrinis (up to 0.1 % of transcripts in Conch Springs). Moreover, it has been shown using 13C isotope analyses that autotrophically fixed inorganic C is a very significant fraction (i.e., 50- > 90 %) of the total biomass C in both communities41. Ultimately, the energy required to reduce significant amounts of inorganic C comes from the oxidation of arsenite and reduced sulfur (several possible choices including elemental S, polysulfide(s), and thiosulfate).
Piliation can serve multiple functions including natural transformation through DNA uptake, export of filamentous phages, protein secretion, adhesion and electron transport67–71,74,75. PilY, which was transcribed by the ‘Pil’ operon in Thermocrinis (Figure 7) is a known adhesin protein shown to be important in attachment in Pseudomonas spp.76. No other microbial population within these communities exhibited the genes required for Type IV filament production, consequently, Thermocrinis is the predominant chemoautotroph responsible for piliation in these biofilms, where strong adhesion to solid substrates is an absolute prerequisite for colonization in these high-velocity ( ~ 0.2 m s-1) streams (Supplemental videos 1,2). The attachment and growth of (Thermocrinis) contributes to the formation of a filamentous biofilm that provides autotrophically fixed inorganic C to other heterotrophic community members41.
FimT has also been shown to mediate DNA uptake in Legionalla77 by directly binding with DNA, which is then transformed after uptake. The actual secretin (PilQ)78 in Thermocrinis was found on a second, highly expressed contig, also conserved across numerous metagenome entries and present in Thermocrinis ruber (Figure 7). The presence of a RecJ immediately adjacent to PilQ78 suggests that the Type IV filament system in Thermocrinis is likely involved in natural transformation, mediated by the activity of FimT and RecJ, for DNA binding and modification, respectively. High rates of natural transformation through an extensive pilin network would also contribute to high redundancies and large genome sizes observed in the Thermocrinis populations from these sites (e.g., Figure 3). This was recently suggested as a primary mechanism for redundant genomic content in Sulfurihydrogenibium, another member of the Aquificales that occupies circumneutral sulfidic systems in YNP34.