Tor Caldara is a natural reserve located approximately 60 km south of Rome and 40 km south-west of Colli Albani, Italy (Patwardhan et al., 2018). The study site(41°29′ 9″ N 12°35′ 23″ E) is a coastal submarine gas vent located at a depth of approximately 3 meters. The site is characterized by vigorous venting of gases that escape from the sandy seabed. The sediment in the venting area is distinctly darker in color than the control sediment because of possible sulfide deposition. Conspicuous growth of white microbial filaments is seen on the rocks directly exposed to the gas venting.

Sampling of filaments and collection and analysis of gases were performed as described previously (Patwardhan et al., 2018). Briefly, established filaments (EF from here on) growing on rocks exposed to the gas venting were collected by a SCUBA diver in August 2016. To study the early stages of colonization, young biofilms (YF from here on) were collected on sterile glass slides mounted on an aluminum rod and exposed to the venting area for four days. Biofilms samples were then stored in RNA Later at –80°C for further nucleic acid and protein extraction. Representative samples of the two communities (one EF and one YF) were further used for downstream applications.

Using data from previous analyses of the composition of the Tor Caldara microbial communities as a guideline (Patwardhan et al., 2018), we selected one YF and one EF biofilm sample for metaproteogenomic analyses. DNA was extracted from YF and EF biofilm biomass stored in RNA Later following a phenol:chloroform extraction protocol. Briefly, 0.8 g of sample were added to 850 μl of extraction buffer (50 mM Tris–HCl, 20 mM EDTA, 100 mM NaCl; pH 8.0) supplemented with100 μl of lysozyme (100 mg/ml) and incubated at 37°C for 30 min. This mix was then supplemented with 5 μl of proteinase K (20 mg/ml), incubated at 37°C for 30 min, and subsequently supplemented with 50 μl SDS (20%) and further incubated at 65°C for 1 hr. Nucleic acids were extracted by performing a series of phenol:chloroform:isoamylalcohol (25:24:1) and chloroform:isoamyl alcohol (24:1) extractions. Multiple samples were co–extracted to reduce potential bias. The final supernatant was precipitated in 3 M sodium–acetate and isopropanol, washed twice with 70% ice–cold ethanol and re–suspended in ultra–pure sterile water. The library was prepared at Molecular Research LP (Shallowater, TX), using the Nextera DNA Sample preparation kit (Illumina) following the manufacturer’s user guide. The initial concentration of DNA was evaluated using the Qubit^® dsDNA HS Assay Kit (Life Technologies). Initial library preparation resulted in very large size library (>2,500 bp), therefore, library preparation was repeated after DNA inhibitor removal using the DNEasy PowerClean Pro Cleanup Kit (Qiagen). Fifty nanogram DNA was used to prepare the library. The sample underwent the simultaneous fragmentation and addition of adapter sequences. These adapters are utilized during a limited-cycle (5 cycles) PCR in which unique indices were added to the sample. Following the library preparation, the final concentration of the library was measured using the Qubit^® dsDNA HS Assay Kit (Life Technologies), and the average library size was determined using the Agilent 2100 Bioanalyzer (Agilent Technologies). The library was diluted (to 10 pM) and sequenced paired end (2 × 2 50 bp, 10 million paired sequences per sample) for 500 cycles on a shared multiplexed run using an Illumina Novaseq 6000 instrument. Following the removal of the adapters, at the sequencing facility, the sequences were quality checked using FastQC v.0.11.5¹. A total of 6,698,798 and 10,125,188 paired end sequences were used for downstream analysis for the EF and YF samples respectively. Sequences encoding rRNA genes were extracted using Metaxa2 v2.1 using default parameters (Bengtsson-Palme et al., 2015) and annotated using Silva ngs v.1.3.9 (Quast et al., 2013). The paired end sequences were then assembled using Megahit v1.1.3 (Li et al., 2015) with default parameters, and assembly was checked using QUAST v3.4 (Gurevich et al., 2013). The EF assembly had a total of 99,137 contigs with the largest contig being 212,020 bp long, whereas the YF assembly had a total of 87,836 contigs with the largest contig being 261,286 bp long. Assembled metagenomes were submitted to DOE Joint Genome Institute’s Integrated Microbial Genome Metagenomic Expert Review (IMG/ER) pipeline for Open Reading Frame (ORF) identification as well as functional and taxonomic annotation (Markowitz et al., 2012). The EF and YF metagenomic assemblies had a total of 308,874 and 362,063 sequences, respectively. The sequences of both metagenomes had COG (51%), KO (41%) and KEGG (25%) annotations. Post annotation, using default parameters for bowtie2 v.2.3.3.1, the reads were mapped back to the assembly (Langmead et al., 2009). Using HTSeq v0.9.1, the number of reads mapping to each gene was counted (Anders et al., 2015). Absolute read counts mapped to each gene were normalized for varying gene lengths and difference in sequencing depth between samples. These normalized gene abundances were expressed as TPM (Transcripts Per Million; Wagner et al., 2012). Sequences are available through the NCBI Short Read Archive database with accession number PRJNA498803.

The total protein fraction was extracted from approximately 1 g of the YF and EF biofilms. Hundred microliter of 2X Laemmli buffer and 50 mM DTT was added to each sample, which was then sonicated, heated at 95°C for 10 min, and centrifuged, with supernatant saved. The precipitate was further extracted by addition of 100 μL of 8M urea, sonicated and centrifuged. Both supernatant were combined and protein concentration was measured using the Pierce 660nm Protein Assay Reagent (ThermoFisher). Thirty microliter out of Two hundred microliter of each sample were run into NuPAGE 10% Bis-Tris Gel (1.5 mmX10 well, Invitrogen). The gel was stained with Coomassie brilliant Blue R250. Standard in-gel tryptic digest was performed as described in Xie et al. (2015). Label-free liquid chromatography-tandem mass spectrometry (LC-MS/MS) was carried out on the two samples, which were analyzed by nano LC-MS/MS using a Q-Exactive HF mass spectrometer interfaced with a Ultimate 3000 RSLCnano chromatography system (Thermofisher). Samples were loaded onto a fused silica trap column (Acclaim PepMap 100, 75 μm × 2 cm, ThermoFisher). After washing for 5 min at 5 μl/min with 0.1% TFA, the trap column was brought in line with an analytical column (Nanoease MZ peptide BEH C18, 130A, 1.7 μm, 75 μm × 250 mm, Waters) for LC-MS/MS. Peptides were fractionated at 300 nL/min using a segmented linear gradient 4–15% B in 30min (A: 0.2% formic acid; B: 0.16% formic acid, 80% acetonitrile), 15–25%B in 40 min, 25–50%B in 44 min, and 50–90%B in 11min. Mass spectrometry data were acquired using a data-dependent acquisition procedure with an MS1 scan (resolution 120,000) followed by MS/MS (resolution 30,000; HCD relative collision energy 27%) on the 20 most intense ions with a dynamic exclusion duration of 20 s. The open-source X!Tandem software (Craig and Beavis, 2004) was used to match spectral data to peptide sequences of the corresponding metagenomes for functional and taxonomical annotation. The minimum acceptable peptide expectation scores were set at 10^–2; the overall peptide false positive rate (Gupta et al., 2011) was 0.07%. A total of 138,387 and 97,607 peptides were obtained in the metaproteomes of EF and YF, respectively. Protein counts were normalized to the total proteins obtained from each sample. Normalized counts for proteins of interest were manually compared between the two samples. These normalized counts were expressed as a percent of total proteins observed in each sample. Raw metaproteomic data were deposited into the ProteomeXchange database with accession number PXD023357².

Quality checked reads from both samples were co-assembled using Megahit v1.1.3 with default parameters and QUAST v3.4 was used to check the quality of the co-assembly. The co-assembled contigs from the two metagenomes were then binned using MaxBin2.0 (Wu et al., 2014) to recover individual genomes. Completeness, taxonomic affiliation and contamination of recovered bins was checked using CheckM v1.0.7 (Parks et al., 2015). The resulting three metagenome-assembled genomes (MAGs TCMF1, TCMF2, and TCMF9) were annotated in RAST (Rapid Annotation using Subsystem Technology; Aziz et al., 2008) and are available with IDs: TCMF1: 6666666.318663; TCMF2: 6666666.318627; TCMF9: 6666666.318829).

Sulfide quinone oxidoreductase (SQR) amino acid sequences obtained in this study and from GenBank (via Blast) were aligned with ClustalX v 2.0 (Thompson et al., 1997) and manually adjusted using Seaview (Galtier et al., 1996). Phylogenetic distances were calculated using the LG amino acid replacement matrix (Le and Gascuel, 2008) and the maximum likelihood method was used to evaluate tree topologies in PhyML 3.0 (Guindon et al., 2010). Branch support was estimated using the approximate likelihood ratio test (aLRT; Anisimova and Gascuel, 2006) and by bootstrap analysis with 1,000 resamplings. Both estimates gave comparable results.

Previous 16S rRNA gene and transcript analyses of the Tor Caldara biofilms showed that there was very little taxonomic variation within replicates of the YF and EF communities, respectively, and that the observed shift in community composition between the YF and EF biofilms was controlled predominantly by the biofilm age (Patwardhan et al., 2018). Based in these findings, two biofilms (one YF and one EF) were selected for further metaproteogenomic analyses. The bacterial diversity of EF and YF was evaluated following in silico extraction of the 16S rRNA gene sequences from the two metagenomes. The EF biofilm was dominated by Gammaproteobacteria (54%), followed by Epsilonproteobacteria (29%) and small proportions of Bacteroidia (4%) and Alphaproteobacteria (3%; Figure 1A). Within the Gammaproteobacteria, the genus Thiomicrospira (40%; now reclassified as Thiomicrorhabdus; Boden et al., 2017) and uncultured members belonging to the Thiotrichales (11%) were dominant (Figure 1B). A reverse trend was observed in the YF biofilm, with Epsilonproteobacteria (67%) dominating the community, followed by Gammaproteobacteria (23%), Bacteroidia (2%), and Alphaproteobacteria (1%; Figure 1A). In the YF biofilm, Sulfurovum (66%) was the most abundant genus within the Epsilonproteobacteria (Figure 1B).

We compared the normalized abundances of key genes and proteins involved in central metabolic pathways, including carbon fixation, nitrogen, sulfur metabolism and (micro)aerobic respiration (Supplementary Tables 1, 2). Since shallow-water vents are enriched in heavy metals, genes and proteins involved in their detoxification were also identified and compared (Supplementary Tables 1, 2). In line with the 16S rRNA gene-inferred community composition, the majority of functional genes and proteins were affiliated with Gamma- and Epsilonproteobacteria (Figures 2–4). The metagenomes provided a snapshot of the metabolic potential of the microbial communities, and were also used to annotate the metaproteomes, which revealed the in situ expression of central metabolic pathways of the communities. The metagenomes were also used to obtain metagenome-assembled genomes (MAGs) of the most abundant members of the filamentous biofilms.

Despite the fundamental role as primary producers of Gamma- and Epsilonproteobacteria in marine geothermal habitats, there is a dearth of representative cultures and their genomes. A total of 31 MAG bins were obtained from the co-assembly of the two metagenomes. Out of these, three (TCMF1, TCMF2, and TCMF9) were ≥85% complete with contamination of ≤8% and complied with the proposed criteria that define medium to high quality draft MAGs (Bowers et al., 2017). The 16S rRNA genes were recovered from bins TCMF2 and TCMF9 (Table 1).

The 16S rRNA gene of MAG TCMF2 showed >97% sequence similarity to 16S rRNA genes related to the Thiotrix-related CF-26 group recovered from microbial mats at shallow-water vents located on the South Tonga Arc (HQ153886) (Murdock et al., 2010) and at the Kuieshan Island (JQ611107). In our previous study of the Tor Caldara vents, dominant OTUs annotated as Thiotrix-related CF-26 from EF showed sequence similarity to the OTUs recovered from the same microbial mats, indicating that the TCMF2 MAG likely belongs to this group of uncultured Thiotrichales. Most of the genes involved in the CBB cycle, including RuBisCo (rbcSL), were present in MAG TCMF2. Further, the potential for thiosulfate oxidation was evidenced by the presence of sox genes, in addition to the sqr and fccAB genes involved in sulfide oxidation. This MAG also contained the genes for cbb3-type cytochrome oxidase and nirBD, involved in microaerobic and nitrite respiration, respectively (Table 1).

MAG TCMF9 showed 93% 16S rRNA gene similarity to a gammaproteobacterial endosymbiont clone of the gutless oligochaete worm, Olavius ilvae. O. ilvae is known to harbor sulfur-oxidizing gammaproteobacterial endosymbionts that fix CO₂ via the CBB cycle (Ruehland et al., 2008). This MAG contained most of the genes involved in the CBB cycle, including RuBisCo (rbcSL) and phosphoribulokinase (prk). A whole suite of genes required for thiosulfate and sulfide oxidation such as the sox system (except soxC), sqr, and fccAB were present. This MAG also contained all three genes (dsrAB, aprAB, sat) involved in the reverse dsr pathway, indicating versatility for the oxidation of sulfur compounds. Three out of the four genes encoding the aa3-type cytochrome oxidase (cox) and all genes encoding the microaerobic cbb3-type cytochrome oxidase (cco) were present (Table 1). The majority of the genes involved in bacterial chemotaxis and flagellar motility were also present (data not shown).

Since the 16S rRNA gene was not recovered in MAG TCMF1, other single-copy ribosomal proteins were used to constrain its lineage. Both the 30S ribosomal protein S8 and S9 were greater than 90% identical to a Sulfurovum species. Additionally, the average amino acid identity (AAI) shared between TCMF1 and Sulfurovum strain NBC37 was 58%, which is well within the range that defines the genus boundary (Rodriguez-R and Konstantinidis, 2014). Based on this data, TCMF1 belonged to the Sulfurovum lineage within the Epsilonproteobacteria, which dominated the YF community (Figure 1B). Sulfurovum spp. are sulfur/hydrogen oxidizers that fix CO₂ using the rTCA cycle (Inagaki, 2004; Mino et al., 2014; Giovannelli et al., 2016). Out of the three diagnostic genes for the rTCA cycle, kor and frd were present. Sox and sqr genes for sulfide/sulfur oxidation were also present. Interestingly, this MAG contained the nifDHK gene essential for nitrogen fixation, while the periplasmic nitrate reductase encoding gene, napAB, which is widespread in the Epsilonproteobacteria (Vetriani et al., 2014), was missing. Three out of the four genes encoding the aa3-type cytochrome oxidase (cox) and all genes encoding the microaerobic cbb3-type (cco) and bd cytochrome oxidase (cyd) were present, indicating the ability to respire oxygen within a broad range of concentrations. The luxS gene, involved in quorum sensing and conserved across all members of the Epsilonproteobacteria (Pérez-Rodríguez et al., 2015), was also present in TCMF1.

Given the presence of genes coding for sqr both in the metagenomes and in the recovered MAGs, we carried out phylogenetic analyses of representative SQR enzymes from the Tor Caldara biofilms and close relatives obtained from GenBank. This analysis placed the sequences into three discrete clusters separated from the Chlorobium lineage, from which SQR was originally characterized: one cluster was related to Sulfurovum spp., one was related to the Thiotricales (Thiomicrospira and Thiothrix spp.) and a third cluster included two groups of sequences: one related to Sulfurovum spp. and one related to the Thiotricales (Figure 5). In line with the taxonomic annotation derived from other genes, the SQR from MAG TCMF1 was related to Sulfurovum sequences, while the enzymes from MAGs TCMF2 and TCMF9 were related to Thiotricales-derived sequences (Figure 5).

Chemoautotrophs are the primary producers at hydrothermal vents, forming the basis of the food web by fixing inorganic CO₂ into organic material (Jannasch and Wirsen, 1979; Karl et al., 1980; Sievert and Vetriani, 2012). Till date, six pathways for carbon fixation are known: the Calvin-Benson-Bassham (CBB) reductive pentose phosphate cycle, the reductive tricarboxylic acid cycle (rTCA), the reductive acetyl coenzyme A pathway (Wood-Ljungdahl), the 3-hydroxypropionate bi-cycle, the 3-hydroxypropionate/4-hydroxybutyrate cycle, and the dicarboxylate/4-hydroxybutyrate cycle (Nakagawa and Takai, 2008; Berg, 2011). The rTCA and CBB cycles are most prevalent at diffuse-flow hydrothermal vents and, in these habitats, they are generally diagnostic of Epsilon- and Gammaproteobacteria, respectively (Nakagawa and Takai, 2008). Enrichment of proteins of the rTCA cycle affiliated with Epsilonproteobacteria, in addition to a similar signature at the gene level, confirmed that autotrophy using the rTCA cycle in Epsilonproteobacteria was the predominant mechanism for carbon fixation in the YF communities (Figures 2, 6 and Supplementary Tables 1, 2). The majority of the rTCA genes and proteins were classified as Sulfurovum spp. (Figure 2), which is known to be ubiquitous in vent environments (Inagaki, 2004; Dahle et al., 2013; Mino et al., 2014; Giovannelli et al., 2016). These findings underline the importance of Sulfurovum spp. in the YF community and are in accordance with data from the 16S rRNA diversity analysis (Patwardhan et al., 2018; Figure 1). While the EF community also encoded genes and expressed enzymes of the rTCA cycle, the CBB cycle was overrepresented in the EF biofilm at both the gene and protein levels, with contributions from various genera within the Gammaproteobacteria, in particular Thiomicrospira, Thioalkovibrio and Thiothrix-related spp. (Figures 2, 6 and Supplementary Tables 1, 2). This is also in line with our 16S rRNA survey, which showed these genera as dominant in the EF biofilm (Patwardhan et al., 2018; Figure 1). Thus, we observed a shift from rTCA-based carbon fixation (Epsilonproteobacteria) to CBB-based carbon fixation (Gammaproteobacteria) during biofilm maturation. The lower energy demand of the rTCA cycle, compared to the CBB cycle (Nakagawa and Takai, 2008; Hügler and Sievert, 2011), might provide an advantage to the Epsilonproteobacteria during early substrate colonization.

In shallow-water geothermal systems, the concentration of reduced sulfur compounds and other electron donors for microbial oxidations, along with the availability of oxidants, vary considerably over time (Yücel et al., 2013; Price and Giovannelli, 2017). Hence, chemolithoautotrophic microorganisms have adapted to such fluctuating redox regimes. In particular, reduced sulfur compounds represent an important energy source that drives energy-yielding reactions in geothermal environments (Nakagawa and Takai, 2008). The abundance and range of various genes and proteins involved in sulfur oxidation pathways presented here validated our initial findings about the importance of sulfur-oxidizing bacteria at Tor Caldara (Patwardhan et al., 2018). For instance, the Sox pathway for sulfur oxidation was one of the most abundant energy-yielding pathway recovered both in the metagenomes and metaproteomes (Figures 3, 4, 6 and Supplementary Tables 1, 2).

Thiosulfate oxidation catalyzed by the multienzyme complex Sox is widespread in sulfur-oxidizing chemoautotrophs (Friedrich et al., 2001, 2005; Yamamoto and Takai, 2011). This enzyme complex, consisting of four protein components, soxYZ, soxXA, soxB, and soxCD, catalyzes the oxidation of thiosulfate and sulfide to sulfate. The abundance of the Sox complex in both filament communities at the gene as well as the protein (YF:1.8%; EF:1.6%) level indicates that this is the predominant pathway used by the biofilm bacteria to conserve energy (Figures 3A, 4A and Supplementary Tables 1, 2). This is consistent with a metaproteogenomic study of chimney-associated microbial communities from deep-sea hydrothermal vents, which revealed that Sox represented one of the most highly expressed energy-yielding pathways (Pjevac et al., 2018). However, the taxonomic affiliation of the sox reads (Figure 3B) and Sox enzymes (Figure 4B) from the Tor Caldara biofilms shows a finer resolution, indicating that Epsilonproteobacteria are mostly expressing the Sox complex in the YF biofilm, while Gammaproteobacteria take over that same role in the EF biofilms (Figure 6).

A similar trend is evident in the profile of the sulfide:quinone oxidoreductase enzyme complex (SQR) involved in oxidation of sulfide to elemental sulfur, which is predominantly expressed by Epsilonproteobacteria in the YF filaments, while its contribution by Gammaproteobacteria increases in the EF filament community (Figures 4B, 6). In addition to SQR, in some chemolithotrophic bacteria oxidation of hydrogen sulfide to elemental sulfur is catalyzed by the flavocytochrome c-sulfide dehydrogenase (FccAB; Visser et al., 1997; Muβmann et al., 2007). The metaproteomic data presented here show that the FccAB enzyme is affiliated only with Gammaproteobacteria, and that its expression is higher in the EF biofilm than in the YF community (Figures 4A, 6). The bioenergetics of microbial sulfide oxidation might provide a framework to interpret the SQR and Fcc expression profiles in the Tor Caldara biofilms. Since the midpoint potential of the NAD⁺/NADH couple at pH 7 is approximately 50 mV more negative than the midpoint potential of the S⁰/H₂S couple, chemolithotrophic SOBs require energy to transport electrons from hydrogen sulfide upwards to NAD⁺ by reverse electron flow. In SQR-catalyzed sulfide oxidation, the electrons from sulfide enter the transport chain at the level of quinones, generating the necessary electrochemical proton potential across the membrane for reverse electron transfer (Griesbeck et al., 2000). However, when Fcc is involved, the electrons enter the transport chain at the level of c-type cytochromes, whose midpoint potential is more electropositive than that of quinones. That increases the energetic burden for the reverse transfer of electrons upwards to NAD⁺ and implies that, in chemolithotrophic SOBs, sulfide oxidation by SQR provides more energy than sulfide oxidation by Fcc (Griesbeck et al., 2000). However, due to its high affinity for sulfide (Brune, 1995), Fcc is hypothesized to be expressed at low sulfide concentrations, supplementing the energetically more efficient SQR. Previous work on the Tor Caldara biofilms demonstrated that Epsilonproteobacteria are adapted to higher sulfide concentrations than Gammaproteobacteria and that they are the pioneer colonizers at this site (Patwardhan et al., 2018). We hypothesize that, early in the colonization process (YF biofilm), Epsilonproteobacteria might be driving down sulfide levels within the biofilms and that the increased expression of the high affinity FccAB in EF might provide an advantage to the Gammaproteobacteria in the subsequent stages of colonization (EF biofilm; Figure 6). The hypothesis that Fcc is more prevalent under low sulfide and more oxidized conditions (Brune, 1995; Griesbeck et al., 2000) supports our model. Overall, the co-expression of the energy efficient SQR and the high affinity Fcc may provide metabolic flexibility in response to fluctuating sulfide concentrations within the biofilm communities.

The reverse dissimilatory sulfate reductase pathway (rDSR) oxidizes stored intracellular sulfur to sulfate via sulfite (Friedrich et al., 2005). Consistent with findings in other filamentous Gammaproteobacteria (Muβmann et al., 2007; Chernousova et al., 2009; Sharrar et al., 2017), all three proteins involved in this pathway, namely, DsrAB, AprAB, and SAT, were affiliated with Gammaproteobacteria and were enriched in the EF biofilm (Figures 4, 6 and Supplementary Table 2). The expression of the rDSR pathway is in accordance with the stored intracellular elemental sulfur observed in electron micrographs of EF (Patwardhan et al., 2018), while its higher expression in the EF biofilm suggests that this community may experience transient depletions of hydrogen sulfide. This stored sulfur might also be disproportionated into sulfite, thiosulfate and sulfide as evidenced by the expression of gammaproteobacterial Sor (Figure 4) in the EF biofilm, and might provide a competitive advantage to the secondary colonizers (Kletzin, 1989; Veith et al., 2012). Since the Dsr pathway for S-oxidation is not encoded in the currently available genomes of Epsilonproteobacteria (Yamamoto and Takai, 2011), the epsiloproteobacterial SAT found in the metaproteome is probably involved in sulfate assimilation.

Sulfur oxidation was complemented with sulfur reduction evidenced by high expression of the a putative Sulfurovum-affiliated NADH-dependent sulfur reductase (NSR; Schut et al., 2007) in the YF (Figures 4, 6 and Supplementary Table 2). Despite sulfur respiration is known to be coupled to hydrogen oxidation in two Sulfurovum spp. (Yamamoto et al., 2010; Mino et al., 2014), we did not detect hydrogenases in the metaproteomes. Since hydrogen concentrations are low at Tor Caldara, the Sulfurovum spp. found there may be adapted to oxidize alternative energy sources, such as malate and/or formate (Campbell et al., 2006; Yamamoto et al., 2010; Keller et al., 2015).

Microbial nitrate respiration occurs via two main pathways, dissimilatory nitrate reduction to ammonia (DNRA) and denitrification. Both processes have been investigated in marine geothermal environments (Bourbonnais et al., 2012a, b; Pérez-Rodríguez et al., 2013; Vetriani et al., 2014). The first step in nitrate reduction is carried out by the enzyme nitrate reductase. Currently, two different types of respiratory nitrate reductases, the membrane-bound Nar and the periplasmic Nap, are known to occur in bacteria (Richardson et al., 2001). Nar is a low-affinity enzyme expressed under nitrate rich conditions, whereas Nap is a high-affinity enzyme whose expression is favored under low nitrate conditions (Potter et al., 1999; Mintmier et al., 2018). The genes encoding both Nar and Nap were present in the EF and YF communities, as well as the genes encoding various nitrite reductases (NirBD, NirK, NirS and NrfA), the nitric oxide reductase (NorB) and the nitrous oxide reductase (NosZ; Figure 3A and Supplementary Table 1). However, metaproteomic analyses showed that only the Gammaproteobacteria-affiliated Nar and Nir enzymes were expressed, while Nap and the enzymes involved in the denitrification pathway could not detected (Figure 4B). The coastal area at Tor Caldara is likely enriched in nitrate (possibly from runoff of lawn fertilizers), which might explain the expression of NarGHI, but not of NapAB (Figure 4A and Supplementary Table 2). The NirBD nitrite reductase, often involved in the reduction of nitrite to ammonia in DNRA (Reyes et al., 2017), was also found to be abundant. Both the Nar and Nir enzymes were enriched in the EF biofilm, indicating that nitrate respiration is prevalent in the established community.

Apart from nitrate respiration, nitrogen fixation also plays an important role in nitrogen cycling at vents (Rau, 1981; Mehta et al., 2003). Expression of epsilonproteobacterial NifDHK was observed in both filament communities but was highly enriched in the EF (Figures 4, 6 and Supplementary Table 2). While nitrogen fixation is not common in Epsilonproteobacteria, recent studies have shown that some members of this class have the genomic potential for it (Keller et al., 2015; Waite et al., 2017), and the ability to fix nitrogen was demonstrated experimentally in Lebetimonas spp. (Meyer and Huber, 2013). Nitrogen fixation is an energy expensive reaction and nitrogenases are usually expressed in nitrogen limiting conditions (Olivares et al., 2013). We hypothesize that, when the filamentous community transitions from a young to an established state, nitrate respiration carried out by Gammaproteobacteria might result in a depletion in nitrogen within the biofilm microenvironment. Hence, the ability to fix nitrogen might give the Epsilonproteobacteria an advantage while competing for a fixed nitrogen source against the secondary colonizers in the established biofilm.

Most of the well-characterized bacterial oxidases belong to the heme-copper superoxidase family. There are two types of oxidases that use cytochrome c as a substrate: cytochrome c oxidase aa3-type and cytochrome c oxidase cbb3-type. The former is expressed in aerobic (high oxygen) conditions while the latter is expressed in microaerobic (low oxygen) conditions. Additionally, cytochrome bd, a ubiquinol oxidase, is also expressed in microaerobic conditions (García-Horsman et al., 1994; Pitcher and Watmough, 2004). At Tor Caldara, the microaerobic, high affinity cbb3-type and bd-type cytochrome oxidase were enriched in EF biofilm and were mostly classified as Gammaproteobacteria (Figures 4, 6 and Supplementary Table 2), suggesting that the biofilms experience microaeobic/anoxic conditions as they mature. On the other hand, primary colonizers in the YF might experience oxic to microaerobic conditions as suggested by equal abundance of aerobic aa3 type and microaerobic cbb3 type cytochrome c oxidase classified dominantly as Epsilonproteobacteria.

Based on the metaproteomic data, we reconstructed the central metabolic pathways expressed by the Tor Caldara young and established biofilm communities (Figure 6). According to this model, substrates exposed to the gas emissions are initially colonized by a population dominated by sulfide-tolerant Epsilonproteobacteria (Figure 1; Patwardhan et al., 2018). Members of this class oxidize various reduced sulfur compounds via the Sox and SQR pathways coupled to oxygen reduction (both aerobic and microaerobic), while fixation of carbon dioxide occurs via the rTCA cycle. In the early stages of colonization, Gammaproteobacteria are present, but underrepresented (Figures 1, 6). Over time, sulfide oxidation by Epsilonproteobacteria may lower the sulfide concentration within the biofilm community, conditioning the environment for the growth of the less sulfide-tolerant Gammaproteobacteria (Patwardhan et al., 2018), which dominate the established filaments. Gammaproteobacteria also oxidize various sulfur compounds using the Sox and SQR pathways, but this reaction is now coupled to either microaerobic and/or anaerobic respiration using oxygen and nitrate as terminal electron acceptors, respectively. The high affinity, periplasmic FccAB enzyme is co-expressed with SQR in response to fluctuating concentrations of sulfide, while elemental sulfur is stored in the cytoplasm. Energy thus conserved is used to fix carbon dioxide via the CBB cycle (Figure 6). When sulfide availability becomes limiting within the biofilm community, certain Gammaproteobacteria further oxidize the stored sulfur to sulfate via the reverse dissimilatory sulfate reductase pathway, thus making full use of their metabolic repertoire. As the filamentous community transitions from a young to a more established stage, anaerobic respiration by the Gammaproteobacteria may lead to loss of nitrate from the system. This triggers increased nitrogen fixation by the Epsilonpreoteobacteria as a potential strategy to compete with the Gammaproteobacteria.