Borehole 3P is located on the west bank of the River Ob' (N58°50′, E81°30′) in the Tomsk Region, Western Siberia, Russia. According to different sources, the depth of the borehole is 2609 or 2775 m. It was drilled as an oil exploration well in 1957–1958 and then abandoned and harnessed by the local population as a source of thermal water. The borehole penetrated almost 30 m of Quaternary sediments, followed by 182 m of Palaeogene sediments. This was succeeded by a Cretaceous sedimentary sequence down to 2250 m and a Jurassic sequence to 2600 m. At 2600 m, the borehole reached the Palaeozoic basement, thus comprising the monzonite and granite (Ulmishek, 2003; Banks et al., 2011, 2014). Water discharges at approximately 9 m³ day⁻¹, has a moderate mineralization (total dissolved solids 13–14 g l⁻¹) and contains dissolved H₂S and CH₄. Methane exsolves as bubbles on the surface. The stratigraphic location from which the water is derived is not known with certainty. Given a water temperature of around 50°C and a typical geothermal gradient in tectonically stable areas of 20–30°C per km, and allowing for the fact that the water may cool during its ascent along the borehole, it seems likely that the water is derived from near the base of the borehole.

Samples of water were collected seven times: in August 2009, February 2010, March 2010, June 2011, July 2012, September 2012, and August 2013. Temperature, pH, and redox potential of water were measured on-site (HI 8314 Hanna pH/redox-meter and corresponding electrodes). Samples for chemical, molecular, and microbiological analyses were taken from the wellhead into sterile 500 ml serum bottles sealed with rubber stoppers. Stoppers of sample containers for analysis of organic compounds were covered with Teflon films. All samples were kept at +4°C for 24 h before analysis. Gas samples were collected four times: in August 2009, March 2010, June 2011, and July 2012. For the samples with the highest hydrocarbon content (2011), the carbon isotopic composition of the gas phase was determined. Gas bubbles exsolving from thermal water were collected by a method modified from those described by Giggenbach and Goguel (1989). We used a glass gas sampler (Savannah River National Laboratory) connected with a sterile 60 ml bottle by a Norprene tubing and submerged into the tank with thermal water inflow. The gas accumulating in the sampler was collected into a bottle by displacement of 50 ml thermal water from it, and 10 ml of water were left as a hydroseal. The filled bottles, being held under water, were sealed with butyl rubber stoppers and then capped and transferred to the laboratory upside down to prevent post-sampling atmospheric contamination. Samples of microbial mats growing along the water flow were collected three times (in August 2009, February 2010, and March 2010) in 50 ml Falcon tubes and were stored at 4°C until used.

Water samples were filtered (0.22 μm) using Millex-GS filter units (Merck Millipore) before analysis. Organic compounds were extracted with chloroform from the water samples. Liquid chromatography–mass spectrometry (LC-MS) was performed in the Institute of Petroleum Chemistry, Siberian branch of Russian Academy of Sciences (Tomsk, Russia). Determination of selected organic compounds was carried out using a Thermo Scientific DFS mass spectrometer. Major ions and chemical oxygen demand (COD) were determined by the Scientific Educational Production Center Voda of Tomsk Polytechnic University by titration (bicarbonate and chloride) or spectrophotometry (sulfate), and H₂S was measured colorimetrically with N,N-dimethyl-p-phenylenediamine (dihydrochloride salt) as the chromophore (Cline, 1969). For COD measurements, the sample was refluxed with potassium dichromate in the sulfuric acid medium and the excess potassium dichromate was determined by titration against ferrous ammonium sulfate using ferroin as an indicator. Other ions were determined at the Chemical Analytical Center Plasma (Tomsk) using an ICP mass-spectrometer ELAN model DRC-e (PerkinElmer Instruments).

In 2010, parallel samples (0.45 μm filtration) were analyzed by ICP-AES and ion chromatography at the Geological Survey of Norway (Banks et al., 2011). In 2013, parallel samples (0.45 μm filtration) were analyzed by ICP-MS and ion chromatography at the British Geological Survey.

Gas samples were analyzed using the Crystal 5000.1 gas chromatograph (Chromatek) equipped with a HayeSep N 80–100 mesh Supelco column (Sigma-Aldrich), methanator, FID, and TCD detectors, which were conditioned at 40°, 350°, 200°, and 150°C, respectively. The carrier gas was argon (25 ml min⁻¹). Volatile fatty acids in the water samples were analyzed using the same equipment with a FID detector and Porapak column conditioned at 180°C. The isotopic composition of methane and ethane (δ¹³C) was determined on a Thermo Electron gas chromatograph coupled with a Delta plus mass spectrometer (Thermo Fisher Scientific).

DNA was extracted from water and mat samples, enrichment cultures, and isolates. Cells from 1.5 l aliquots of borehole groundwater were retained on cellulose nitrate filters (0.2 μm) using a Sartorius filtration unit. Filters were pestled in liquid nitrogen at −70°C and then used for DNA extraction with a Power Soil DNA isolation kit (MO BIO Laboratories) according to the manufacturer's instructions. Cells from enrichment culture broths and pure cultures were collected by centrifugation (9000 g for 15 min). Extraction of DNA from approximately 5 ml of enrichments or pure cultures was performed as described by Tsai and Olson (1991), with minor modifications (i.e., proteinase K (10 μg ml⁻¹) was added after three cycles of freezing the sample in liquid nitrogen at −70°C and thawing in a 65°C water bath).

Nested PCR (Vissers et al., 2009) was used for the amplification of bacterial and archaeal 16S rRNA genes with the primers listed in Table S2. Domain-specific primers were used separately for Bacteria and Archaea for both the first and second rounds of PCR. The products obtained in the reaction with outer primers were diluted with nuclease-free water (Fermentas) to a concentration of ca. 100 mg l⁻¹ and then used as a template for the reaction with inner primers. 16S rRNA gene fragments of Bacteria were first amplified using outer primer pair 27F–1492R (DeLong, 1992) and then with Bacteria-specific inner primer pair GC-Bacv3f and 907r (Lane, 1991; Muyzer et al., 1996). The first round of amplification of the 16S rRNA gene fragments of Archaea was conducted with the primer pair 21F-958R (Weisburg et al., 1991; DeLong, 1992) and the second round with the Archaea-specific pair GC-Arch915r and Parch519f (Coolen et al., 2004). We have successfully used these primer sets in our previous studies to amplify bacterial and archaeal 16S rRNA gene fragments separately (Karnachuk et al., 2009; Frank et al., 2016).

Table S2 shows the details of the PCR conditions used in this study. The amounts of MgCl₂, dNTPs, primers, and Taq DNA-polymerase were altered depending on the primer set. The reagents used were 25 mM MgCl₂, 2 M dNTPs, 5 U μl⁻¹ Taq DNA polymerase, 10x Taq buffer, and nuclease-free water (Fermentas). 100 μM stock solutions of oligonucleotide primers synthesized by Syntol (Moscow) were applied. 400 mg l⁻¹ of BSA (Fermentas) were added to the reaction mixture for archaeal 16S rRNA gene amplification. Reactions with outer primers were conducted in a volume of 50 μl. PCR with inner primers was performed in a volume of 100 μl. Each mixture contained 50–100 mg l⁻¹ of the DNA matrix. All PCRs were conducted in the MyCycler thermal cycler (Bio-Rad Laboratories).

The DGGE used for the separation of amplified fragments (Muyzer et al., 1993) was performed with the DCode System (Bio-Rad). Polyacrylamide gel (8%) with a 30–70% denaturing gradient was used for 16S rRNA fragments separation (100% denaturing solution contained 7 M urea and 40% formamide). Electrophoresis was performed for 17 h at 60°C and 120 V. For separation of archaeal 16S rRNA gene fragments, 8% polyacrylamide gel with 20 to 80% denaturing gradient was used. Electrophoresis was performed for 19.5 h at 60°C and 100 V, as described by Vissers et al. (2009). DGGE gel was stained with 0.5 mg l⁻¹ ethidium bromide (Bio-Rad) in a TAE buffer for 15 min and then washed in TAE for 20 min. The stained gel was visualized in UV light (320 nm) using a GelDoc-It imaging system (UVP, UK). Separate bands were cut from the gel in UV (320 nm) using transilluminator ECX-26MX (Vilber Lourmat). Then DNA was extracted to 20 μl of nuclease-free water (Fermentas) for 12 h at 4°C. Amplification of 16S rRNA gene fragments was performed as described above, but the primers did not contain GC-clamps. DGGE analysis of each PCR product was performed in duplicate.

Commercial sequencing of 16S rRNA (585 bp and ca. 400 bp) gene fragments was performed by Syntol. The sequences were analyzed against the GenBank nucleotide collection database (http://www.ncbi.nlm.nih.gov/blast) by the BLASTN algorithm with standard parameters (Altschul et al., 1997). The sequences were aligned using BioEdit sequence alignment editor (Hall, 1999), Version 7.2.5. Maximum likelihood and neighbor-joining phylogenetic trees based on the comparison of 16S rRNA gene sequences were constructed using MEGA 6 (Tamura et al., 2013). All the sequences were uploaded to the GenBank nucleotide collection database under the accession numbers KT897597 – 651, KY010808, and KY010809.

Pyrosequencing analyses were performed for water samples obtained from the borehole in June 2011. Cells from 30 L of water were collected on 0.2 μm nitrate cellulose filters. The filters were frozen in liquid nitrogen and then ground and melted with TE buffer in a water bath at 37°C. The total DNA was extracted by the CTAB/NaCl method (Wilson, 2001).

Universal primers were used for amplification of the V3–V4 variable regions of the 16S rRNA gene, U341F (5′-CCT ACG GGR SGC AGC AG-3′), and PRK806R (5′-GGA CTA CYV GGG TAT CTA AT-3′). The primer check with SILVA TestPrime confirmed that they target most of Archaea and Bacteria except only a few lineages, notably Planctomycetes and some Crenarchaeota. The PCR fragment was pyrosequenced on GS FLX (Roche) using titanium chemistry. A total of 39,276 sequence reads were obtained. Most of the reads covered the full length of the PCR fragment. The reads with mismatches to primer sequences, those containing ambiguous nucleotides, and those shorter than 400 bp were excluded from the analysis. Then 16S rRNA data were analyzed by the RDP Classifier program package (Cole et al., 2009). Initially, the sequences obtained with universal primers were distributed between Bacteria and Archaea using the online RDP Naive Bayesian rRNA Classifier Version 2.0 (http://rdp.cme.msu.edu/classifier/classifier.jsp). Subsequently, bacterial and archaeal 16S rRNA gene datasets were analyzed separately and subjected to additional filters. First, AmpliconNoise (Quince et al., 2011) was used to account for homopolymer-derived and PCR errors. Then all remaining singletons (unique sequences occurring only once) were removed, as suggested by Behnke et al. (2011). The final datasets consisted of 21,608 bacterial and 6965 archaeal 16S reads. Complete linkage clustering and selection of representative sequences for operational taxonomic units (OTUs) were performed using the RDP Classifier (Cole et al., 2009). OTUs were assigned to taxonomic groups (i.e., bacterial and archaeal divisions) on the basis of BLASTN sequence similarity searches against the NCBI database. Taxonomic assignments were refined following the construction of phylogenetic trees consisting of representative sequences of the clusters and a set of 16S rRNA gene sequences from the representatives of different archaeal and bacterial lineages. The sequences were aligned using CLUSTALX (Thompson et al., 1997), and the neighbor-joining tree was computed by TREECON (Van de Peer and De Wachter, 1994). Shannon and equitability indices based on 16S rRNA gene sequence data were calculated at the 97 and 95% cutoff levels using the RDP Pipeline (Cole et al., 2009). Pyrosequencing read data has been deposited in the NCBI SRA database under accession SRR4450631.

Microorganisms of different metabolic groups were enriched using aerobic and anaerobic conditions as previously described (Podosokorskaya et al., 2013a). Microcrystalline cellulose (Avicel; 2 g l⁻¹), cellobiose (2 g l⁻¹), lactate (Na salt; 20 mM), peptone (Sigma; 10 g l⁻¹), gelatin (Sigma; 20 g l⁻¹), and H₂ (80 or 100%) were used as the energy substrates. Sulfate (Na salt, 10 mM), arsenate (Na salt, 5 mM), and glauconite (final Fe(III) concentration ca. 20 mM) were employed as the electron acceptors. Yeast extract (0.05–0.1 g l⁻¹) was added as the source of growth factors. For batch enrichments, 10 ml portions of the media were dispensed in 18 ml Hungate tubes, and the head space was filled with an oxygen-free N₂(100%) or N₂/CO₂ mixture (80:20%). When the medium was prepared aerobically or H₂ was used as the substrate, the volume of the medium was 5 ml, and the head space was filled with either air or H₂ (100%) or H₂/CO₂ mixture (80:20%), respectively. Incubation temperatures were 47, 50, 54, or 70°C. The growth of all cultures was monitored by direct cell counting under a phase contrast microscope at 1000 × magnification.

Continuous enrichments were initiated in a 1.3 l BioFlo 110 bioreactor (New Brunswick Scientific) under microaerobic conditions (4.5% O₂ in the gas phase) with Na-acetate (10 mM) as the substrate. The mineral salts solution was modified according to the chemical analysis results of the borehole water as follows. The mineral salts solution contained (per liter) 7.0 g NaCl, 0.19 g MgCl₂, 0.25 g NH₄Cl, 0.5 g KCl, 0.2 g KH₂PO₄, 0.15 g NaHCO₃, 0.013 g Na₂S, and 1.0 g CaCl₂. The medium was prepared as described previously (Podosokorskaya et al., 2013a). Trace element solutions A (10 ml l⁻¹) and B (1 ml l⁻¹) (Table S3) were used to provide for minor and trace elements corresponding to the elemental composition of the borehole water. Spherical glass beads (av. diameter 2 mm) were spread at the bottom of the reactor before autoclaving to model the mineral surface of the wallrock in the aquifer. The medium was autoclaved in the bioreactor and feed bottles at 121°C for 20 min. The pH (7.0–7.5), temperature (50–70°C), and dissolved O₂ of the medium were adjusted before inoculation and automatically sustained during the incubation. The bioreactor was inoculated with 0.5 l of borehole water sampled in February 2010 and started in a batch mode. At the commencement of the exponential growth phase, the bioreactor was switched to continuous mode at a gradually increasing dilution rate of 0.5–1.17 ml min⁻¹. Mixing was adjusted with Rushton impellers to 160 ml min⁻¹ (coefficient of turbulent diffusion 178.2 cm² s⁻¹), which ensured a perfect mixing model. The composition of enrichment cultures was studied using PCR-DGGE analyses as described above.

The chemical analysis results of the borehole water, sampled at three different time points in 2009–2010, have been reviewed by Banks et al. (2011), and the sampling in August 2013 has also been summarized (Banks et al., 2014). Briefly, the water is of Na-(Ca)-Cl composition. The molar ratio of Na/Cl is between 0.78 and 0.86, which is almost identical to modern sea water (0.86). Bromide occurs at 28–33 mg l⁻¹ in the water, while the Cl⁻/Br⁻ mass ratio of around 262–275 is very close to 288 of standard seawater. The water contains ca. 1.1 g l⁻¹ of Ca but only about 5 mg l⁻¹ of Mg. It is supersaturated with respect to calcite and saturated with respect to dolomite; the Sr concentration is high at ca. 86 mg l⁻¹.

A significant variation in temperature was observed, from 50.2°C in February 2010 to 45.2°C in August 2010 (Table 1), while the temperature was around 50°C in August 2013. The water pH was circumneutral and also varied somewhat from 7.52 in August 2009 to 6.91 in August 2010. The redox potential of about −250 to −300 mV (vs. standard hydrogen electrode [SHE]) was indicative of strongly reducing conditions, also evidenced by the presence of dissolved H₂S, CH₄, near absence of sulfate, and elevated barium (17 mg l⁻¹ in 2013) due to the lack of a barite solubility ceiling. The H₂S concentrations varied from 2.46 mg l⁻¹ in August 2009 to 7.42 mg l⁻¹ in August 2010. Concentrations of most heavy metals (except Ba) measured in 2013 were low and were presumed to be suppressed by the high pH and sulfide concentrations. Arsenic concentrations were ca. 130 μg l⁻¹ (Table S1).

The concentration of total carbon, including organics (as judged by COD measurements), varied significantly from ca. 10³ mg l⁻¹ in August 2009 to ca. 10² mg l⁻¹ in 2010 sampling rounds. Modest concentrations of alkanes and non-volatile C₆–C₁₆ carboxylic acids were detected in all samples (Table 1). Acetate (160 mg l⁻¹) and propionate (7.5 mg l⁻¹) were present in significant concentrations, which underwent minor changes in the scope of 18% from August 2009 to March 2010. No other volatile fatty acids were detected in water samples.

The composition of the exsolved gas bubbles collected from the water varied with time. CH₄ was the main constituent, varying from 57.5% (v/v) (June 2011) to 86.3% (March 2010). The N₂ content reached up to 21.6% of the gas in June 2011 (Table 2). O₂ was detected in almost all gas samples, probably originating from nonequilibrium degassing of admixed meteoric recharge upon contact with deep hot sedimentary water. The minor components detected in almost all the samples were CO₂ and H₂ with concentrations not exceeding 0.3%. An ethane concentration of 16.3% was recorded in the June 2011 sample when methane was at its lowest fraction, but it decreased to 0.8% in the July 2012 sample (Table 2). The isotopic composition (δ¹³C) of CH₄ and CH₃CH₃ in the June 2011 sample was −53%0 (indicative of thermogenic with minor admixture of biogenic CH₄ as described by Schoell, 1984) and −40%0, respectively.

Cell numbers assessed by direct cell counting varied significantly from one sample to another, being maximal in the first water sample taken in August 2009 (ca. 10⁸ cells ml⁻¹) and minimal in the February 2010 sample (3 × 10⁴ cells ml⁻¹). In other water samples, the cell content analyzed on-site varied from 10⁵ to 10⁷ cells ml⁻¹. In the sample, taken in June 2011 for 16S rRNA-profiling with the pyrosequencing technique, the cell number comprised ca. 2 × 10⁵ cells ml⁻¹. In all the samples, various short rod-shaped morphotypes (less than 1 μm in length) prevailed and two different coccoid morphotypes were also detected.

Some members of the prokaryotic community matched with bacterial genera (95–100% similarity), namely Delftia, Desulfovirgula, Thermoacetogenium, Desulfotomaculum, Symbiobacterium, Ignavibacterium, and archaea in the genus Methanothermobacter. Many phylotypes in the borehole water were uncultured Firmicutes (88–89% 16S rRNA gene similarity with closest cultured relatives). The majority of phylotypes were detected only once or, occasionally, twice. Among these, there were phylotypes matched with Delftia tsuruhatensis (only August 2009) and Ignavibacterium album (March 2010 and August 2013). Three phylotypes from the August 2009 sample were consistent with Thermoacetogenium phaeum (99% similarity), Symbiobacterium turbinis (97%), and Desulfotomaculum salinum (99%). Two Firmicutes were present in four out of the five DNA samples: a sulfate-reducing thermophile Desulfovirgula thermocuniculi (98–99%) and an uncultured bacterium that had 86% similarity with Thermanaerovibrio acidaminovorans (Table 3, Figure 1).

Pyrosequencing analysis of 16S rRNA gene fragments was performed for the total DNA isolated from the borehole water sample of June 2011. Bacteria accounted for 76% of all 16S rRNA reads and were mostly represented by Firmicutes (Figure 2). Bacteria closely related to the sulfate-reducing thermophile Desulfovirgula thermocuniculi were the dominant component of the community (46.9% of the total). The second most abundant bacterial group (16.7%) comprised several lineages related to the genus Thermoacetogenium. Sulfate-reducer Desulfotomaculum kuznetsovii (Nazina et al., 1988) accounted for 1.3% of prokaryotes. Other Firmicutes of uncultured genera comprised 6.9% of the whole community. Members of the candidate division OP9 (Hugenholtz et al., 1998) accounted for 2.4% of the community. This group includes organotrophic bacteria proposed to ferment polysaccharides (Dodsworth et al., 2013). Other bacteria were about 1.3% of the total. Among them were Bellilinea caldifistulae from the phylum Chloroflexi (0.15%) and Ignavibacterium album (0.13%) from the phylum Ignavibacteriae.

Archaeal sequences comprised about 24% of all 16S rRNA gene fragments. Most of them could be assigned to the hydrogenotrophic Methanothermobacter thermautotrophicus frequently found in aquifers. Among the other archaeal 16S rRNA sequences, members of the genera Methanobacterium and Methanosaeta were identified in minor amounts (<0.3%).

Overall, the microbial community diversity and richness, characterized by Shannon and equitability indices, is rather low (Table 4) in comparison with values reported for subsurface sediments and oil reservoirs, where Shannon indexes from 2 to 4 are typically reported (for example, Biddle et al., 2011; Gao et al., 2016). This is especially evident for the archaeal part of the community characterized by the presence of just a few species. Thus, in the pyrosequenced DNA sample, the two groups represented by the highest number of sequences—Desulfovirgula and Methanothermobacter spp.—were those consistently detected by DGGE as the components of the deep subsurface prokaryotic community.

Microbial mats on the wooden conduit construction on the outflow pathway of the borehole water were about 1–2 mm thick and gelatinous and had a colorless, light pink, or pink-grayish hue (Figure S1). Based on the PCR-DGGE results (Table 3), completely different bacteria dominated mat communities in the two samplings in August 2009 and February 2010. Various

Bacteroidetes (Flavobacterium and Hydrogenophaga spp.) were present in August 2009 (colorless and light pink mats), while Hydrogenophilus spp. and uncultured Deferribacteres were detected in February 2010 (mainly grayish mats).

Batch enrichment cultures were obtained on the media, containing diverse organic substrates in either the presence or absence of external electron acceptors, which were inoculated with borehole water samples (Table 5). PCR-DGGE analyses of the enrichment cultures showed the presence of phylogenetically diverse prokaryotes, most of them belonging to Bacteria.

Phylum Firmicutes was represented by numerous phylotypes, including both cultured and uncultured members of this group. Among other phylotypes, Bellilinea caldifistulae (99% similarity) and Melioribacter roseus (100% similarity with the type strain; Podosokorskaya et al., 2013a), representing Chloroflexi and Ignavibacteriae phyla, respectively, should be mentioned as well as Hydrogenophilus thermoluteolous (99% identity), a member of Betaproteobacteria and a representative of the candidate division OP9 (98% similarity) (Hugenholtz et al., 1998).

Several enrichment cultures were obtained in the presence of insoluble Fe(III) forms and As(V) as the electron acceptors.

The only enrichment culture that contained prevailing phylotypes of the consortium, which has also been identified directly by PCR-DGGE and pyrosequencing of the borehole water, was that incubated at 50°C with gelatin and sulfate. It comprised of Desulfovirgula thermocuniculi (99% similarity) and Methanothermobacter thermautotrophicus (100% similarity), accompanied by Delftia tsuruhatensis (99% similarity) and two uncultured Deferribacteres.

The enrichment culture obtained from a 3P microbial mat sample on the medium with microcrystalline cellulose contained three novel genera of hydrolytic bacteria that were subsequently isolated and described as Melioribacter roseus gen. nov., sp. nov., Ornatilinea apprima gen. nov., sp. nov., and Mobilitalea sibirica gen. nov., sp. nov. (Podosokorskaya et al., 2013a,b, 2014). They were able to grow on various polysaccharides, including cellulose, as well as on sugars and peptides.

Two continuous enrichments from the borehole water developed in the bioreactor under microaerobic conditions (4% O₂). The enrichments were obtained with an incremental temperature increase from 50 to 70°C, with acetate as the sole carbon source and electron donor. The cell density in the bioreactor started at 1 × 10⁴ cells ml⁻¹ at 50°C and reached 3 × 10⁷ cells ml⁻¹ after 144 h of batch cultivation, thus indicating the beginning of the exponential growth phase. The exponential phase was sustained for 50 h of continuous cultivation under a perfect mixing regimen. Subsequently, the cultivation mode was switched back to batch culture, and the cell density reached its maximum of 7 × 10⁹ cells ml⁻¹ within 96 h. The high cell density was accompanied by the formation of biofilms on the interior glass walls, rotating impellers of the bioreactor, and the glass beads used to model the wallrock. When the cells started to lyse, the temperature was increased to 70°C, and the bioreactor was switched again to continuous mode. The growth ensued for about 100 h but to a lesser cell density (2·10⁸ cells ml⁻¹). Acetate consumption correlated well with the growth curve of the culture (Figure S2). DGGE analysis throughout the enrichments at 50 and 70°C showed only one phylotype, which is a close relative of Hydrogenophylus thermoluteolus (99% similarity).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.