Hydrothermal chimneys studied and reported herein were recovered in 2011 during the multidisciplinary HYDROPRONY cruise (R/V Alis; Pelletier et al., 2011) in the Prony Bay (Figure 1). Sample sites were distributed on the seafloor at depths between 15 and 50 meters below sea level (mbsl) except for two sites located in the intertidal zone (Bain des Japonais and Rivière des Kaoris). The sites lie on serpentinized peridotites, gabbros, and dunites that form the main geological substratum of South New Caledonia (Supplementary Figure S1). The underwater structures, where high-pH hydrothermal fluids discharge and mix with seawater, were mainly made of brucite along with carbonates (Pisapia et al., 2013). They were organized as massive consolidated concretions that may take the form of needles up to 35 m high with, if active, small centimetric whitish and unconsolidated nascent protochimneys on their top (Figure 1; Supplementary Figure S2).

Twenty-five active and fossil vents (from eight different sites spanning the Prony Bay) were sampled for molecular ecology, geochemistry, and mineralogy during the HYDROPRONY expedition. We focus here on two types of active chimneys sampled at ST07 and ST09 sites, a few kilometers apart in the bay (Figure 1; Supplementary Figure S1). They are representative of two distinct development stages of the venting edifices. The youngest chimney came from ST09, a 50 m wide and 3 m high dome in the center of a 150 m wide depression at 50 mbsl, overlain by 3–4 m high chimneys with smaller active vents. The external part of the actively venting and poorly consolidated sampled chimney was not colonized by marine macrofauna, evidence that it is a young structure (Figures 1b,c). Additionally, except at the coastal sites (Bain des Japonais and Rivière des Kaoris), the fluid venting through the brucite-dominated chimney of ST09 appeared to be the most pristine among all the submerged sites, hence the most representative of the serpentinization-derived hydrothermal fluid (Monnin et al., 2014). Site ST07, the Roc Aiguille of Prony Bay, is a shallower and more massive structure with a main pinnacle ∼38 m high and culminating up to 2 mbsl. This is the tallest structure in the bay and probably the oldest. The sampled chimney was covered by a pronounced coral crust, attesting to a long residence time at the seafloor (Figures 1e,f). Chimney samples from sites ST08, ST11, ST12 and Bain des Japonais (Supplementary Figures S1 and S2) were also examined to assess the representativeness of our microscale observations. An overview of the studied sites and of the analyses performed on each is available in Supplementary Table S1. Full site descriptions along with vent fluid geochemistry and associated molecular ecology of some of the hydrothermal chimneys at PHF can be found in Pelletier et al. (2011), Monnin et al. (2014), Quéméneur et al. (2014) and Postec et al. (2015). Elemental compositions and temperatures of the fluids collected at the main selected sites are reported in Supplementary Table S2 where they are compared to the LCHF and The Cedars (Kelley et al., 2001; Suzuki et al., 2013).

Samples dedicated to SEM microimaging were extracted or cut aseptically in the field and immediately stored at 4°C. Samples for molecular ecology were collected on board using sterilized pliers immediately after the scuba dives and each sample was divided into several fractions. The fractions dedicated to FISH experiments were fixed with 2% formaldehyde in phosphate-buffered saline solution (PBS; Gérard et al., 2013) and stored in 50% ethanol/PBS at -20°C until use (see details in Postec et al., 2015). Other fractions of the same samples were immersed in 50% ethanol and kept at -20°C for further molecular characterizations or CLSM observations.

Scanning Electron Microscopy observations of freshly cut chimney fragments were performed at the Service Commun de Microscopie Electronique à Balayage (UPMC, Paris, France) using a Zeiss SUPRA 55 VP Field Emission Scanning Electron Microscope. Samples were Au-coated and no organic resin was employed. Images were collected using secondary electron detectors (Everhart-Thornley for high voltage mode, VPSE for variable pressure mode and InLens for low voltage mode) and a backscattered electron detector (AsB). Accelerating voltage ranged from 7 to 15 kV at variable pressures and high current (up to 1 nA) or was fixed at 3 kV under high vacuum and low current (down to 10 pA). Elemental microanalyses were also performed using an Energy Dispersive X-ray (EDX) spectrometer (PGT Sahara).

In order to reliably determine the nature of the minerals that encrust microbial filaments (see Section “Organomineralization Processes Drive the Early Stage of Chimney Construction at the PHF”), we extracted individual mineralized biofilaments by laser microdissection using a Zeiss PALM MicroBeam apparatus. Filaments were microdissected from the most juvenile ST09 chimney samples and were recovered in sterile Eppendorf^TM tubes. Individual mineralized filaments were stuck on a borosilicate glass rod and were analyzed by XRD phi-scan at the IMPMC (UPMC, Paris, France). Diffractograms were obtained, after a 10 h acquisition time (step of 0.039° 2𝜃), with an Agilent XCalibur S diffractometer equipped with a four-circle goniometer, a Sapphire 3 CCD detector and a monochromatic Mo X-ray source (0.71073 Å).

Epifluorescence microscopy was carried out on an Olympus BX51 microscope equipped with an Olympus UPlanApo 40×/1.00 Oil Iris objective and a U-MNIB3 filter after Syto^®9 staining of the samples (Invitrogen). CLSM was carried out using an Olympus FluoView FV1000 confocal microscope with an oil immersion objective Olympus UPSLAPO 60xO (Gérard et al., 2013). Fluorescence image stacks were obtained with concomitant excitation at wavelengths of 405, 488, and 543 nm by collecting the emitted fluorescence between 425–475, 500–530, and 560–660 nm, respectively. Three-dimensional images were acquired, visualized, and processed using the F10-ASW FLUOVIEW software (Olympus). Microbial cells were localized and imaged thanks to non-specific DNA staining using green-fluorescent Syto^®9 dye (excitation at 488 nm and emission detection range 500–530 nm). A set of at least four stained subsamples coming from the same conduit were examined each time. The autofluorescence properties of the F₄₂₀ factor involved in the energy metabolism of anaerobic methane-cycling archaea were also collected after excitation at 405 nm (detection range 420–480 nm). The associated fluorescence spectra (data not shown) were recorded following the protocol described in Postec et al. (2015). Complementary FISH experiments were conducted on mature chimneys samples in order to document the presence of Methanosarcinales-related cells. The oligonucleotide probes (Thermo Fisher Scientific Inc., Waltham, MA, USA) used were specific for Euryarchaeota (EURY498, 5′-CTTGCCCRGCCCTT-3′) and Crenarchaeota (CREN499, 5′-CCAGRCTTGCCCCCCGCT-3′; Burggraf et al., 1994). More details are available at probeBase (Loy et al., 2007). The hybridization experiments were conducted on multitest slides (10 wells) at 46°C for 2–4 h on formaldehyde-fixed chimney fragments from site ST07 using 5 ng⋅μL^-1 of each probe in 0.9 M NaCl, 20 mM Tris-HCl pH 8, 0.01% SDS containing 35% (vol/vol) formamide. Samples were then washed for 15 min at 48°C in 0.9 M NaCl, 20 mM Tris pH 8.5 mM EDTA, 0.01% SDS. The slides were then soaked in cold water for a few seconds and air-dried. Finally, samples were stained with DAPI (excitation at 405 nm and detection range between 450 and 500 nm) at a concentration of 1 μg⋅mL^-1 (Sigma) for 1 min, then washed for a few seconds in cold water, and left to dry. For CLSM examination, hybridized cells were covered by Citifluor AF3 (glycerol mounting solution from Emgrid).

When bacterial filamentous microorganisms were abundantly detected using CLSM and epifluorescence microscopy (see section “Organomineralization Processes Drive the Early Stage of Chimney Construction at the PHF”), DNA extractions were performed with the PowerSoil^®DNA Isolation Kit (Mo Bio, USA). For this purpose, a total volume of sample of about 1 cm³ was dissected from the filament-rich areas of the most juvenile ST09 chimney samples. The extracted sample was divided into two subsamples. The first one was stained with Syto^®9 to ensure using epifluorescence microscopy the abundance of filamentous microorganisms. The second aliquot was kept unstained. DNA extractions were performed on both. Due to the small amount of DNA recovered with this procedure, DNA was also extracted from bulk rock bathed by the hydrothermal fluids in ST09 and ST11 chimney samples (hereafter referred to as bulk chimney samples). This provided a broader overview of the microbial population hosted in the active conduits of ST09 and ST11 juvenile chimneys. In both cases, bacterial 16S rRNA genes were amplified by PCR using (i) the bacterial specific primer 27F (5′-AGAGTTTGATCCTGGCTCAG-3′) with the prokaryote specific reverse primer 1492R (5′-GGTTACCTTGTTACGACTT-3′) and (ii) the bacterial specific primer 63F (5′-CAGGCCTAACACATGCAAGTC-3′) with the prokaryote specific reverse primer 1387R (5′-GGGCGGWGTGTACAAGGC-3′). All primers were provided by GATC (France). Five microliters of purified DNA were used in a reaction buffer volume of 25–30 μL containing 1.5 mM MgCl₂, dNTPs (10 nmol each), 20 pmol of each primer and 1 U GoTaq polymerase (Promega, France). PCR reactions were performed under the following conditions: 35 cycles (denaturation at 94°C for 15 s, annealing at 55°C for 30 s, extension at 72°C for 2 min) preceded by 2 min denaturation at 94°C, and followed by 7 min extension at 72°C. Direct PCR amplifications were carried out with different primer combinations. For the dissected filament-rich samples, we performed in addition nested PCR. 1 μL of the products of the first PCR carried out with the most external primers (i.e., 27F and 1492R) was used in nested PCRs with primers 63F and 1387R and using the same conditions (35 cycles – denaturation at 94°C for 15 s, annealing at 55°C for 30 s, extension at 72°C for 2 min – preceded by 2 min denaturation at 94°C, and followed by 7 min extension at 72°C). The controls for all PCRs were negative, including the controls from the first PCR used as template for the nested PCR. Cloning was done using the Topo TA Cloning system (Invitrogen) following the instructions provided by the manufacturer. After plating, positive transformants were screened by PCR amplification of inserts using flanking vector primers and the PCR products were partially sequenced using either 1387R or 1492R. We first examined the 16S rRNA gene sequence of 41 clones from four different cloning experiments, two derived from the DNA extracted from the Syto^®9-stained sample, which was examined with CLSM and two derived from the unstained second-half of the sample. We obtained a larger set of clones from DNA bulk extractions (88 and 87 for ST09 and ST11, respectively) and then examined these larger sets of sequences. PCR amplifications of archaeal 16S rRNA genes were also attempted from DNA extracted from ST09 and ST11 chimney samples. While the primers were the same as in Postec et al. (2015) where Archaea sequences were retrieved, they gave no results in this study. Following the protocol described in Gérard et al. (2009), all sample manipulations were done using material dedicated to samples with high contamination risks. PCR protocols were realized using a Biocap^TM RNA/DNA hood (Captair, Erlab, France) equipped with a HEPA filter. Chamber and micropipettes were UV-sterilized before use and aerosol resistant pipette tips were used to reduce external contamination.

The sequences reported in this paper have been deposited in the GenBank database. Sequences retrieved from the unstained and Syto^®9-stained ST09 samples on which CLSM observations were carried out correspond to the accession no. KM207235 for HPst091-1-1 and KR911715, KR911716, and KR911717 for St09-1-17, St09-2-3, and ST09S-2, respectively. The sequence HPst091-1-1 was retrieved after direct PCRs (primers 63F and 1387R) and nested PCRs (primers 27F-1492R then 63F-1387R) from both Syto^®9-stained and unstained samples, St09-1-17 and St09-2-3 only after nested PCR from the unstained sample, and ST09S-2 after direct PCR from the Syto^®9-stained sample. Sequences obtained from bulk chimney samples have accession no. ranging from KX344725 to KX344743 and KX349203 for ST09 and from KX344744 to KX344775 for ST11.

Taxonomic affiliations at the phylum level were first obtained by comparing several portions of partial 16S rRNA gene sequences with sequences of the GenBank database using BLAST (Basic Local Alignment Search Tool; Altschul et al., 1997). Representative clones of the dominant phyla were then fully sequenced and analyzed with the ARB software (Ludwig et al., 2004) by using the SILVA database SSU ref LSU 123 – full release (Pruesse et al., 2007; Quast et al., 2013; Yilmaz et al., 2014). The sequences were first aligned with the SINA online aligner (Pruesse et al., 2012) and then added in the ARB guide tree using the ARB parsimony tool. The phylogenetic tree was then constructed by adding to the aligned sequences, sequences of the closest cultivated bacteria and environmental clones in the RAxML (Randomized Axelerated Maximum Likelihood) program (Stamatakis et al., 2008) by using the GTRCAT substitution model. The bootstrap values were calculated from 1,000 replicates.

Organic acid concentrations were analyzed in duplicate using the method of Albert and Martens (1997). Error of individual measurements was ± 0.5 μM and corresponded to the larger of either the standard deviation between the two analyses, or the individual measurement error. Samples were analyzed for formate, acetate, propanoate, butyrate. Only formate and acetate were detected at concentrations greater than 0.5 μM.

Both mineralogical and microbiological investigations were conducted on the chimney internal conduits (Figures 1c,f; Supplementary Figure S2d) where the reduced and alkaline fluids, derived from serpentinization at depth, circulated. They hosted brown microbial mats on which our investigations focused. Similar to what has been described for the LCHF (Schrenk et al., 2004; Ludwig et al., 2006) the actively venting chimneys were highly porous. The interior of the chimneys was dominated by brucite mixed with Mg-carbonates and aragonite, with an increasing proportion of Ca-carbonates closer to the exterior of the chimney, which precipitated when seawater started to mix with the hyperalkaline hydrothermal fluids (Pisapia et al., 2013; Monnin et al., 2014).

As shown in Figures 2a–e, SEM observations carried out on the young chimney collected at ST09 revealed that the interior of the active conduits was formed by the interweaving of numerous mineralized filaments up to hundreds of micrometer long. Similar encrusted filaments were observed in carbonated hydrothermal chimneys deriving from serpentinization at the LCHF (Kelley et al., 2005) and at the Shinkai Seep Field (Southern Mariana forearc; Okumura et al., 2016). More specifically, Okumura et al. (2016) noted the large presence of mineralized filamentous microorganisms in brucite-dominated chimneys submitted to high fluid discharge conditions that are comparable to the ST09 ones. But, at Shinkai Seep Field, as well as at LCHF, although filaments were assumed to be of bacterial origin, no direct or indirect evidences are provided. At PHF, SEM-EDX analyses showed that these filaments corresponded to C-rich structures (Figure 3a) of few micrometers in diameter on which the mineralization process appeared to initiate (Figure 2a). These structures were progressively encrusted by round-shaped and flaky or plate-like crystals that increased the filament diameter up to a few tens of μm (Figures 2b–d). SEM-EDX spectra collected on numerous minerals encrusting the C-rich filaments were indicative of the presence of Mg and O (Figure 3b). This observation was supported by XRD analyses of individual encrusted filaments extracted by laser microdissection that highlighted that the mineral crust was mainly made of brucite (Figure 3d). Brucite was sometimes associated with minor hydrotalcite, being either a brucite transformation by-product or a co-precipitated phase (Figure 3d). Whereas the hydrothermal fluid was supersaturated with respect to Ca- and Mg-carbonates along with Mg-hydroxides (Monnin et al., 2014), the dominance of brucite, a Mg-hydroxide [Mg(OH)₂], with minor hydrotalcite, a layered double hydroxide [Mg₆Al₂CO₃(OH)₁₆⋅4(H₂O)], jointly with the absence of carbonates in the most juvenile chimneys bathed by the hydrothermal fluids, suggested that these fluids were depleted in carbonate ions. It also attests to very low CO₂ partial pressure (pCO₂) at site ST09, as confirmed by fluid analysis (Supplementary Table S2), even if local variations of the pCO₂ may have occurred allowing minor hydrotalcite precipitation. Upon brucite precipitation, the filamentous structures became more dense and aggregated, leading to the formation of compact walls formed by cemented, mineralized filaments (Figure 2e). Similar tangled filaments were found in the inner conduits of the more mature chimney samples collected at site ST07. These samples were covered by or embedded with secondary carbonates (e.g., aragonite needles), as indicated by SEM-EDX analyses displaying a high Ca and O signal (Figure 3c) and XRD analyses, attesting to a greater degree of mixing of hydrothermal fluids with seawater in mature chimneys (Figure 2f). Indeed, as stated in Monnin et al. (2014) who emphasized the role of fluid mixing in mineral formation, seawater input is mandatory for carbonate precipitation.

These C-rich filaments, upon which chimney-forming minerals appeared to nucleate and grow, were ubiquitous, found in all the chimneys sampled in the bay, at the deep sites (Supplementary Figures S3a–d,h–k) as well as at the coastal ones (Supplementary Figures S3e–g). Thus, it seems to represent a general pattern for the construction and the early consolidation of the PHF hydrothermal concretions. Mineralized filaments were also detected from the very early stages of the vent history, notably in the whitish protochimneys that typified nascent overgrowth where the most pristine fluids are discharged (Figure 4). This probably indicates that the filaments are likely of subseafloor origin and are brought with the hydrothermal fluids to the seafloor where they begin to mix with seawater, as shown by Monnin et al. (2014) who located the mixing zone not in the subsurface but at the point of discharge. Epifluorescent and CLSM images carried out on ST09, ST11, and ST07 samples stained with Syto^®9 suggest a biological origin to the C-rich filaments (Figures 5a–c). They corresponded to microbial sheaths in which individual rod-shaped cells of few micrometers are visible (Figure 5b).

Thermodynamic calculations showed that the fluids at ST09 and ST07 are supersaturated with respect to brucite at pH values higher than 9.5 (Monnin et al., 2014) suggesting that brucite may precipitate abiotically from the fluid. However, the filamentous organisms appear to predominantly impact the concretion formation through organomineralization processes during the earliest stages of the hydrothermal discharge. Organomineralization refers to biologically induced mineralization through the local modification of the chemical environment by microbial activity and/or to passive mineral precipitation in the presence of extracellular polymeric substances (EPS) or biological membranes as nucleation sites (Perry et al., 2007; Defarge et al., 2009). Brucite precipitation has been shown to be induced by the alkalinizing activity of corals, diatoms, algae, or cyanobacteria when coupled to a lowering of CO₂ partial pressure and an increase of Mg²⁺ activity (Schmalz, 1965; Nothdurft et al., 2005; Buster and Holmes, 2006). However, at PHF, considering the high pH reported in Monnin et al. (2014) for the hydrothermal fluids at ST09 (up to 10.62; Supplementary Table S2), such an alkaline engine does not appear to be the missing motive force in the conduits of the chimney. Conversely, brucite precipitation was experimentally shown to co-occur with organic polymer excretion. Brucite nuclei are stabilized by carboxylate moiety (R-COO^-) from microbial cell walls at the surface of diatoms, also accelerating the polymerization of larger brucite crystals from Mg(OH)₂ molecules in water (Sato et al., 2003; Tesson et al., 2008). Accordingly, at PHF, the filamentous microorganisms and their sheaths likely served as nucleation sites for brucite precipitation.

The filamentous organisms thriving in the chimney conduits were identified on several young chimney samples by targeted phylogenetic analyses. PCR amplification, cloning and sequencing of 16S rRNA gene sequences from DNA extracted from filament-rich areas dissected from the inner conduit of ST09 young chimney (Figures 5a,b) revealed that the brown microbial mat was mainly dominated by a single bacterial phylotype belonging to the Firmicutes phylum (33/41 clones; Table 1). The 33 Firmicutes clones corresponded to a single OTU (more than 97% identity at the level of their 16S rRNA gene sequences) represented by the clone KM207235 HPst091-1-1. It belonged to a group of uncultivated bacteria that were mainly identified in alkaline and/or subsurface environments (Supplementary Figure S4). The closest environmental sequence (98% identity) of HPst091-1-1 was retrieved from the Donana’s suboxic aquifer (Spain; Lopez-Achilla et al., 2007). The closest cultivated species of HPst091-1-1, Dethiobacter alkaliphilus (88% identity), is a facultative autotrophic strain growing with H₂ as electron donor and S-compounds, but not sulfate, as electron acceptors. It can also use acetate as a carbon source and diverse C-compounds as both electron donors and carbon sources (Sorokin et al., 2008) and is related to sequences commonly obtained from continental serpentinites (Crespo-Medina et al., 2014).

CLSM observations carried out on the dense biofilm covering the minerals forming the internal conduit of the mature chimney from ST07 revealed a higher diversity of microorganisms highlighted by the green-fluorescence of Syto^®9-stained filamentous and cocci cells (Figures 5c,d). Among the cocci, irregular cells were commonly aggregated into clusters of 1 μm and presented various stages of cell division (Figures 5c–e). Notably, they displayed the typical blue-green autofluorescence of the F₄₂₀ cofactor (Figure 5e), a key co-enzyme involved in anaerobic methane-cycling (Thauer, 1998). Their affiliation to Euryarchaeota was confirmed by FISH (Figure 5f). This was in agreement with the phylogenetic analyses of the archaeal diversity carried out at PHF that showed the dominance of the uncultured LCMS in ST07 chimneys (Quéméneur et al., 2014; Postec et al., 2015). Note that these archaeal cocci were never found in the conduits of the young chimney collected at sites ST09 and ST11 (Figure 5a). The staining and FISH experiments carried out on ST09, ST07, and ST11 chimneys, combined with 16S rRNA gene sequence analyses on filament-rich areas of the juvenile ST09 chimney, indicated that the dominant OTU KM207235 HPst091-1-1 was part of the early colonizers of the PHF hydrothermal edifices, prior to archaea. The subseafloor origin of these filaments is suggested by their presence in the nascent protochimneys, their role in early chimney construction, and by the affiliation of the OTU KM207235 HPst091-1-1 to sequences retrieved from other subsurface environments.

In order to assess the representativeness of the bacterial filaments retrieved in the dissected conduit samples, bacterial diversity was characterized from DNA extracted from bulk rock samples bathed by the hydrothermal fluid in both ST09 and ST11 chimneys (Figure 6; Supplementary Table S3). Congruent with the lack of detection of methanogenic archaea by microscopic examination, we did not detect any archaeal 16S rRNA gene sequences by PCR amplification. As shown in Figure 6, the community of the inner conduits depicted by cloning and Sanger sequencing confirmed that Firmicutes were important inhabitants of ST09 and ST11 bulk chimney conduits (representing 20.7 and 17.2% of all retrieved sequences, respectively). It is worth noting the apparent discrepancy between the two sequencing approaches (i.e., targeted filament-rich areas vs. blindly bulk sampling), notably the fact that KM207235 HPst091-1-1 was not the dominant OTU when DNA was extracted from bulk chimney samples. This may be due to its thick, mineralized sheath that may hamper cell lysis. Nevertheless, the Firmicutes phylum was interestingly dominated by the KM207235 HPST091-1-1 sequence, representing 14.8 and 11.5% of all sequences retrieved for ST09 and ST11, respectively, (i.e., 72 and 67% of the total Firmicutes). Note in addition that this Firmicutes species was not restricted to these two juvenile chimneys and similar sequences were also reported under the accession numbers KF886161 (clone PHFST07_B16) and KF886087 (clone PHFST08_B19) in ST07 and ST08 sites (Supplementary Figure S1; Quéméneur et al., 2014) and under the accession number KJ159205 (clone PHF_15-B34_D22) in other chimneys of site ST09 (Postec et al., 2015). As shown in Figure 7 and Supplementary Table S3 and with the sole exception of ST11d-17, the other sequences affiliated to the Firmicutes phylum in ST09 and ST11 related to sequences retrieved from other chimneys from PHF ST09 site (Postec et al., 2015), in the serpentinization-driven subterrestrial alkaline aquifer of Cabeço de Vide aquifer in Portugal (Tiago and Verissimo, 2013), in a deep H₂-rich gold mine in the South Africa Precambrian craton (Lin et al., 2006) and in a terrestrial hydrocarbon seep in Taiwan (Cheng et al., 2014).

In the bulk chimney samples collected in ST09 conduits, we detected numerous sequences of Acetothermia (candidate division OP1) and Omnitrophica (candidate division OP3) (Hugenholtz et al., 1998; Rinke et al., 2013), respectively, representing 11.5 and 13.8% of all retrieved sequences. Similar to Firmicutes, these bacteria may also constitute key components of the ST09 community. The sequences affiliated to Acetothermia were related to taxa retrieved from a deep aquifer showing high methane production (Kimura et al., 2010). The closest relatives of the sequences affiliated to Omnitrophica were retrieved from a subseafloor environment along the Juan de Fuca Ridge flank (Huber et al., 2006), which are also closely related to sequences from the Cedars (Suzuki et al., 2013) and Lost City (Brazelton et al., 2006) hydrothermal fields. Additionally, as shown in Figure 7, a large number of unclassified sequences (up to 35.6% of the total sequences; Figure 6) were also related to Acetothermia (St09-52c, St09-50c). Nonetheless, they only shared 87% identity with their closest environmental sequence retrieved in a hydrothermal system associated with a back-arc seated seamount (Stott et al., 2008). Interestingly, Postec et al. (2015) has reported the occurrence of similar but shorter sequences forming a distinct clade during the 6-year monitoring of ST09 site. Both Acetothermia and Omnitrophica affiliated sequences occurred at a level comparable to the present study (each representing 13.8% of the total clones). The sequences retrieved in our study from bulk samples collected in ST11 juvenile chimneys (Figure 6; Supplementary Table S3) indicated that Alphaproteobacteria dominated the bacterial community (i.e., 59.8% of all retrieved sequences; Figure 6), thus providing evidence for more important mixing with seawater compared to ST09 chimney. Members of Acetothermia were still detectable but to a much lesser extent (i.e., 2.3% of all sequences) and no members of Omnitrophica were highlighted. Although Proteobacteria were less abundant in ST09 chimney conduits (10.3% of all sequences compared to 77.0% for ST11), the Deltaproteobacteria constituted the most important class (6.9% of all sequences or 66.7% of the Proteobacteria sequences). Phylogenetic affiliations of the Proteobacteria showed that they were related to sequences retrieved in alkaline environments including serpentinization associated ones (Supplementary Figure S5).

These results obtained on bulk chimney samples confirmed the presence of OTU KM207235 HPst091-1-1 retrieved from the filament-rich areas collected in ST09 chimney samples. The predominance of representatives of the Firmicutes phylum dominated at ∼70% by one species (KM207235 HPst091-1-1), combined with the absence of seawater species (Figures 6 and 7), reinforced the idea that it likely represents one of the microorganisms that first colonized these chimneys and could be representative of community associated with the most pristine hydrothermal fluids. Considering that the bacterial mineralizing filaments were involved in the early stages of chimney construction and that their phylogenetic affiliations suggest a subsurface origin, we conclude that this Firmicutes species was discharged with hydrothermal fluids from the subseafloor. The hydrothermal origin of these bacteria may also be assumed based on the fact that the closest environmental sequences of some of the sequences retrieved, including those affiliated to Gammaproteobacteria, Deltaproteobacteria (Supplementary Figure S5), Omnitrophica, and Acetothermia (Figure 7) were found in alkaline and/or hot crustal environments. All these taxa may be representative of endolithic ecosystems relying on the serpentinization reaction at depth and thus of SLIMEs. Furthermore, the predominance, during the early stages of chimney construction, of Bacteria rather than archaeal Methanosarcinales as observed in Lost City (Schrenk et al., 2004; Brazelton et al., 2006), raises questions about the functioning of this serpentinization-related ecosystem.

Our micrometric scale imaging of the chimney conduits coupled to 16S rRNA gene analyses highlight the importance of bacteria with a subseafloor hydrothermal origin in the serpentinization-related ecosystem of the PHF. In accordance, their metabolisms may document the functioning of a SLIME. However, inferring metabolisms or carbon substrates is particularly challenging if not impossible in the present case as the retrieved taxa are lacking cultivable relatives and are far from any environmental clones (Figure 7; Table 1; Supplementary Table S3). Nonetheless, due to the potential that they represent a relevant model for the study of serpentinization-based ecosystems, they deserve further attention and efforts.

Members of Acetothermia and Omnitrophica have been shown to occur across a geochemically diverse range of environments including hydrothermal fields from oceanic ridges (e.g., Brazelton et al., 2006; Perner et al., 2007) to organic-rich environments (Dojka et al., 1998; Teske et al., 2002; Dhillon et al., 2003; Alain et al., 2006; Inagaki et al., 2006; Duncan et al., 2009; Harrison et al., 2009; Kobayashi et al., 2012; Zhou et al., 2012) and also in fluids associated with actively serpentinizing rocks in the Oman ophiolite (Miller et al., 2016). Within the past few years, two nearly complete genomes of Acetothermia were established based on cultivation-independent genome-resolved metagenomic analysis of a subsurface thermophilic microbial mat community (Takami et al., 2012) and of an oil reservoir community (Hu et al., 2016). Whereas the first pointed to its capability of implementing acetogenesis through the ancient reductive acetyl-CoA pathway by utilizing CO₂ and H₂, the second suggested significant differences in metabolic capacities and was more in favor of heterotrophy with consequences on the biogeochemical transformations occurring in the oil reservoir. In either case, both are very distantly related to the sequences retrieved in ST09 and ST11 chimneys, and the metabolic capacities of taxa from Acetothermia and Omnitrophica remain entirely unknown at PHF, although their relative proportions and ubiquity may imply a prominent role in the functioning of the ecosystem.

A large diversity of Firmicutes was observed at Lost City (Brazelton et al., 2006). When compared to subterrestrial sites of active serpentinization, also rich in electron donors such as H₂ and CH₄, bacterial phylotypes belonging to the Clostridia class were also found to be dominant and closely related to D. alkaliphilus (Figure 7; Supplementary Figure S4). This was the case for the deep terrestrial groundwaters at The Cedars formation (Suzuki et al., 2013), Cabeço de Vide aquifer (Tiago and Verissimo, 2013), and Tablelands ophiolitic complex (Brazelton et al., 2012). Molecular microbial surveys of the PHF chimneys have also demonstrated that Firmicutes were particularly abundant among the diverse bacterial community (Quéméneur et al., 2014; Postec et al., 2015). The systematic presence of these phylotypes may thus be a common feature of subsurface ecosystems in serpentinization-based environments, as proposed by Suzuki et al. (2013).

It has been highlighted at LCHF and in the mature chimneys of PHF that Methanosarcinales were abundant and associated with sulfate-reducing bacteria (SRB). At LCHF, dsrB genes, a molecular marker for SRB, affiliated to Firmicutes were found to be abundant (Brazelton et al., 2006; Gerasimchuk et al., 2010). In contrast, at PHF, the dsrB gene abundance estimated using quantitative PCR indicated that SRB represented less than 6% of the total bacterial community in various ST09 chimneys (Postec et al., 2015) and only 2–8% in other chimneys of the PHF (Quéméneur et al., 2014). Furthermore, dsrB genes sequences were affiliated to Desulfovibrionales and Desulfobacterales orders and not to Firmicutes (Quéméneur et al., 2014; Postec et al., 2015). The observation that Firmicutes may be one of the first inhabitants of the conduits of the PHF nascent vents, dominated by hydrothermal fluids circulation over seawater, raises the questions of their metabolic capabilities, the nature of the available electron donors/acceptors and carbon sources, as well as their role in the PHF ecosystem, particularly with respect to Methanosarcinales that are only observed in more mature chimney conduits. Archaea cycling methane are acknowledged to be highly pervasive across diverse ecosystems, showing a tolerance to wide ranges of extreme physical and chemical parameters, and large ranges of pH (from 3 to 10), temperatures (from 0 to 122°C), and salinities (i.e., 1–5 M NaCl; Hoehler et al., 2010). On the contrary, compared to members of Firmicutes that can tolerate oxygen at a relatively high level (Whitham et al., 2015), Methanosarcinales are highly sensitive to oxygen (Zinder, 1993; Hoehler et al., 2010). Even if the fluids circulating within juvenile conduits were highly dominated by anoxic hydrothermal fluids, we cannot exclude slight oxygen contamination of the deep fluids when they reached the seafloor, hence affecting Archaea. However, DIC concentrations are quite low in the most juvenile ST09 chimney (Supplementary Table S2). It may limit the presence of autotrophic microorganisms. Deprivation in inorganic carbon in the pristine fluids is also indicated by the near absence of solid carbonates in the chimney conduits bathed with the reduced hydrothermal fluids where brucite, a Mg-hydroxide, and minor hydrotalcite, a double-layered Mg-Al-hydroxide, precipitated on the Firmicutes filaments (Figure 3). As we have no information on the metabolism of the bacteria represented by OTU HPSt091-1-1, we cannot exclude that this strain could be an autotrophic Firmicutes consuming the scarce inorganic carbon and preventing the formation of carbonate minerals. Knowing that in the subsurface serpentinizing environments, organic compounds are likely not as limited as the inorganic compounds (Schrenk et al., 2013), we propose an alternative hypothesis that could explain the predominance of Firmicutes over Methanosarcinales in juvenile chimney conduits. Organic carbon compounds can indeed have multiple origins including in situ microbial activities at depth (Ménez et al., 2012), hydrothermal degradation of biological remnants (Pasini et al., 2013) and abiogenic synthesis whose amplitude and pathways still need to be assessed in natural systems (Sephton and Hazen, 2013). Consistently, heterotrophic and fermentative species belonging to the Firmicutes phylum have been retrieved from active serpentinization sites (Brazelton et al., 2012; Schrenk et al., 2013; Tiago and Verissimo, 2013). Notably, the two obligate alkaliphilic strains recently isolated from the interiors of PHF chimneys (Alkaliphilus hydrothermalis and Acetoanaerobium pronyense; Ben Aissa et al., 2014; Bes et al., 2015) can produce H₂, CO₂, and acetate from the fermentation of various carbon sources or the disproportionation of crotonate to acetate and butyrate (Ben Aissa et al., 2014). Hence, as proposed by Schrenk et al. (2013) heterotrophy and fermentation may be important metabolic strategies in these serpentinizing environments where autotrophy is severely challenged by the limited availability of inorganic C-compounds, those being instead very rapidly trapped into solid carbonates at these hyperalkaline pH (Alt and Teagle, 1999; Bradley et al., 2009). Lever (2012) has shown that under conditions similar to those encountered at the PHF, the versatility of the biochemical pathways of heterotrophic microorganisms such as acetogens may confer to them an ecological advantage compared to the more substrate specialized Methanosarcinales. This might be supported by a preliminary study carried out on the PHF fluids and showing elevated acetate concentrations (around 70 μM) and minor formate (3–4 μM) whereas inverse proportions were found at LCHF (8 and 140 μM for acetate and formate, respectively; Lang et al., 2010). We thus hypothesize that the dominant bacteria associated with the serpentinizing fluids in the early stages of chimney formation at PHF could have versatile metabolisms that would allow them overcoming the scarcity of inorganic carbon compounds.

Considering the potential of the present taxa to be representative members of serpentinization-based ecosystems and in order to infer their metabolic capabilities, genomic analysis from single cell isolates along with metagenomic analyses could prove highly enlightening. Similarly, a better understanding and consideration of the diversity of carbon sources in the serpentinizing subsurface could shed more light on the way they shape microbial ecosystems at depth. Finally, considering that serpentinization-based ecosystems are considered as analogs of primitive microbial ecosystems, and that some authors suggest that Acetothermia may be deeply branched on the tree of life (Takami et al., 2012), elucidating the metabolic capabilities of the dominant bacteria at PHF may also give clues about ancestral metabolisms.