The study site was located on a sand dune 300 m from the coastline of the Teshio Plain located in northern Hokkaido (44.9948∘ N, 141.6882∘ E) at an elevation of 5.2 m above sea level (Fig. S1). The geology of this site down to a depth of 1000 m b.g.s. consists of the Yuchi Formation, the Sarabetsu Formation, and alluvium, in ascending order. Hydraulic gradient and conductivity, water isotopic, and geological data indicate that clay aquitards at approximately 300 m b.g.s. prevent water in the upper part of the Sarabetsu Formation from mixing with connate water in the lower part of the Sarabetsu Formation and the underlying Yuchi Formation (Ikawa et al., 2014). Indeed, a drastic change observed in the stable hydrogen (δD) isotope ratio of the water at approximately 300 m b.g.s. (Fig. 1c) indicates that the aquifers above and below that depth are hydrologically different (Ikawa et al., 2014).

The water samples for this study were collected from three wells: D-1, D-2, and D-3. The D-2 and D-3 wells are located within 30 m of D-1. Two samples from D-2 and D-3 are made up of fresh water derived from the upper Sarabetsu Formation, and four samples from D-1 are brines derived from the Yuchi Formation. Samples from D-2 and D-3 were pumped through strainers at depths of 90–100 and 130–149 m b.g.s., respectively, while samples from D-1 were pumped with borehole packer assemblies at depths of 476, 613, 715, and 943 m b.g.s. The water samples were obtained after chemical parameters, such as the water temperature, electrical conductivity, pH, and oxidation–reduction potential, had stabilized. Before sample collection, waters of approximately 44–160 times the wellbore volume of D-2 or D-3 and 4–13 times the packer-sealed area of D-1 were pumped out.

Water samples for microbial cultures were collected in sterilized glass bottles with butyl rubber stoppers and screw caps. The bottles were purged with N2 gas before and during sample collection and then filled with the water sample to raise the internal pressure of the bottles, effectively preventing the penetration of air. For the molecular analysis, 4 L water samples were collected and filtered through a 0.2 µm pore size Millipore Express Plus membrane filter (Millipore, Billerica, MA, USA) and stored at -20 ∘C. The samples used for total cell counts were fixed with formalin at a final concentration of 2 % (v/v) immediately after sampling and stored at 4 ∘C. The gases that were associated with the formation water in aquifers (i.e., under high-pressure conditions) were naturally separated from the water when it was sampled under atmospheric pressure. These separated gases were collected using the water displacement method.

The chemical compositions and stable hydrogen isotope ratios (δD) of the water samples were measured by using ion chromatography (DIONEX ICS-5000, Thermo Fisher Scientific, Bremen, Germany) and a liquid water isotope analyzer (L2120-i, Picarro, Santa Clara, CA, USA), respectively. The standard deviation of δD for water was 1 ‰.

The gas composition was analyzed using two gas chromatographs: a Shimadzu GC-8A equipped with a thermal conductivity detector (TCD) and a Molecular Sieve 5A column and an Agilent 6890 equipped with a flame ionization detector and a PoraPLOT Q column. Stable carbon (δ13C) and hydrogen (δD) isotope ratios of methane, along with the δ13C of carbon dioxide, were measured using a Trace Ultra gas chromatograph connected to a DELTA V plus isotope ratio mass spectrometer (IRMS) through a GC IsoLink combustion–pyrolysis interface (Thermo Fisher Scientific), and the column was a PoraPLOT Q. The Natural Gas Standard NGS3 was used as an isotope reference material. The standard deviations of δD and δ13C for methane were 1.6 ‰ and 0.3 ‰, respectively, and the standard deviation of δ13C for carbon dioxide was 0.2 ‰.

The total organic carbon (TOC) content of silty or clayey sediment core samples from the Yuchi Formation, collected previously by Ikawa et al. (2014), was measured with a TruSpec CHN analyzer (LECO). Before the measurements, samples were pulverized to less than 200 mesh and treated with 1 M HCl to remove inorganic carbon.

A fixed water sample was filtered through a 0.2 µm pore size Isopore membrane filter (Millipore), stained for 10 min with SYBR Green solution (10 µgmL-1), and observed under an epifluorescence microscope (Olympus, Tokyo, Japan).

DNA was extracted from the filtered water and methanogenic culture (as described below) samples by using a PowerWater kit (MoBio Laboratories, CA, USA) according to the manufacturer's protocol. Quantitative PCR targeting bacterial and archaeal 16S rRNA genes in water samples was performed in triplicate by the quenching probe method (Tani et al., 2009) using TITANIUM Taq DNA polymerase (Takara, Otsu, Japan) in a Rotor-Gene Q real-time cycler (QIAGEN, Valencia, CA). The primers and probes used for real-time PCR and sequencing (as described below) are listed in Table S1 in the Supplement. The cycling conditions were 95 ∘C for 2 min, followed by 50 cycles of 93 ∘C for 15 s, 61 ∘C for 20 s, and 72 ∘C for 25 s. The copy numbers of the mcrA gene, which encodes a methyl-coenzyme M reductase alpha subunit, an enzyme central to methanogenesis, were quantified in triplicate by SYBR Green real-time PCR using SYBR Premix Ex-Taq II (Takara) in a LightCycler 1.0 (Roche, Basel, Switzerland). The cycling conditions were 95 ∘C for 30 s, followed by 50 cycles of 95 ∘C for 15 s, 52 ∘C for 20 s, and 72 ∘C for 25 s. Tenfold serial dilutions of the target PCR products for Escherichia coli K12 (ATCC 10798) (for the bacterial 16S rRNA gene) and Methanobacterium bryantii M.o.H. (ATCC 33272) (for the archaeal 16S rRNA and mcrA genes) were also amplified to calculate the gene copy numbers.

The 16S rRNA genes, including the V3 and V4 regions, were amplified using AmpliTaq Gold 360 DNA polymerase (Life Technologies, CA, USA) with a Univ515F primer (fused to the 454-specific adaptor A and 6-nt barcode sequences) and a Univ926R primer (fused to adaptor B). The cycling conditions were 95 ∘C for 10 min, followed by 25–27 cycles of 95 ∘C for 30 s, 50 ∘C for 40 s, and 72 ∘C for 30 s, with a final extension period of 7 min at 72 ∘C. Four replicates of PCR products for each sample were pooled and purified by using a MonoFas DNA purification kit. Pyrosequencing was performed using a 454 Life Sciences GS FLX Titanium platform (Roche, Basel, Switzerland) at Hokkaido System Science Co., Ltd. (Sapporo, Japan).

The mcrA gene was amplified from the six original water and methanogenic culture samples (as described below) by using the MLf and MLr primer pair (Luton et al., 2002) and AmpliTaq Gold 360 DNA Polymerase (Thermo Fisher Scientific). The PCR products were purified by using a MonoFas DNA purification kit (GL Sciences, Tokyo, Japan), cloned in the pCR4-TOPO vector (Thermo Fisher Scientific), and sequenced by the dideoxynucleotide chain-termination method using BigDye terminator reagents (Thermo Fisher Scientific) and an automated sequence analyzer (3730 DNA Analyzer, Thermo Fisher Scientific) according to the manufacturer's instructions.

The 454 pyrosequencing reads of the 16S rRNA genes were analyzed by using Mothur version 1.48 software (Schloss et al., 2009), as described previously (Katayama et al., 2015, 2022) with the following modifications. Quality-filtered sequences with an average length of 250 bp were classified by using a Bayesian classifier based on the Silva taxonomy SSU Ref 138.1 dataset (Pruesse et al., 2007) with a confidence threshold of 80 %. The putative methanogens in the 16S rRNA gene sequences were searched based on this taxonomic classification.

Sanger sequences of the mcrA gene were translated to amino acids and aligned by using MAFFT version 7 software (Katoh and Standley, 2013). Amino acid sequences with >93 % sequence identity were treated as operational taxonomic units (OTUs). In each OTU, the most abundant sequence was selected as the representative sequence. The species most closely related to the OTUs were searched by using BLAST (http://blast.ncbi.nlm.nih.gov/Blast.cgi, last access: 6 May 2023).

The 454-sequencing data were submitted to the DDBJ Sequence Read Archive database under accession number DRA001113. The GenBank/EMBL/DDBJ accession numbers for the mcrA gene sequences are LC214911 to LC214935.

The basal medium used for the methanogenic cultures consisted of 10 mM NH4Cl, 1 mM KH2PO4, 15 mM MgCl2⋅6H2O, 1 mM CaCl2⋅2H2O, 30 mM NaHCO3, 1 mLL-1 selenium and tungsten solution, 1 mLL-1 trace element solution, 2 mLL-1 vitamin solution, 1 mLL-1 resazurin solution (1 mgmL-1), and 0.5 mM titanium(III) nitrilotriacetate (as a reducing agent) (Katayama and Kamagata, 2018; Katayama et al., 2020). A total of 20 mL of basal mineral medium was dispensed into 67 mL serum vials. The vials were sealed with butyl rubber septa and aluminum crimps under an atmosphere of N2/CO2 (80:20, v/v). The medium was supplemented with either H2/CO2 (80:20, v/v; 0.1 MPa) or acetate (20 mM) as methanogenic substrates. The medium was further supplemented with NaCl at final concentrations of 250 and 500 mM, which approximated its in situ concentrations at 613 and 943 m b.g.s., respectively. The 1 mL aliquots of the water samples from 613 and 943 m b.g.s. were dispensed into each medium and incubated at 25 ∘C (for the 613 m b.g.s. sample) and 35 ∘C (for the 943 m b.g.s. sample) to approximate the in situ water temperature. Methane production was measured using a GC equipped with a TCD. After methane production was terminated, 4 mL of the sample cultures were harvested by filtration, and the mcrA gene was cloned and sequenced as described above.

Methane-producing cultures supplemented with H2/CO2 or acetate from the 943 m b.g.s. water sample were subsequently inoculated into fresh media to examine the methanogenic activity under different salinity and temperature conditions. Cultures with different salinities were grown in basal mineral media containing 15, 270, or 480 mM Cl- at 35 ∘C. Cultures with different temperatures were grown in basal mineral media containing 480 mM Cl- at 20, 25, 35, or 45 ∘C. Both culture sets were supplemented with H2/CO2 (80:20, v/v; 0.1 MPa) or acetate (20 mM). The time course for methane production was determined to calculate the methane production rate.

Semicontinuous cultivation supplemented with a low concentration of acetate (0.4 mM) was performed in a modified 132 mL glass vial containing sterilized pieces of nonwoven fabric as the carrier material for microbial cells (Fig. S2a in the Supplement) to culture microorganisms involved in syntrophic acetate oxidation (SAO) coupled to methanogenesis via CO2 reduction. A total of 40 mL of basal mineral medium (as described above) supplemented with NaCl (500 mM) and acetate (0.4 mM) was dispensed into the vial. Then, 10 mL of the 943 m b.g.s. water sample was used as an inoculum. The top and bottom of the vial were sealed with a butyl rubber septum and aluminum crimps, and the culture was incubated at 35 ∘C under an N2/CO2 (80:20, v/v) atmosphere for 10 months. During cultivation, the culture was manually fed acetate (0.4 mM) at 3-week intervals. Before feeding, 20 mL of culture liquid was removed from the vial through the bottom septum using a needle syringe, and 20 mL of fresh medium containing acetate (final concentration: 0.4 mM) was then added to the vial through the top septum so that the syntrophic association of microbial cells was not physically disrupted by turning the vial upside down.

After cultivation, 2 mL of culture liquid and a piece of nonwoven fabric were transferred from the semicontinuous cultivation system to 67 mL serum vials containing basal mineral medium with 0.4 mM [2-13C]-acetate or nonlabeled acetate (used as a control) to determine the presence of SAO activity. In both cultures, 0.4 mM labeled or unlabeled acetate supplement was added at 2-week intervals. The incubation was performed in duplicate. Time courses for methane production and the stable carbon isotopic ratio of dissolved inorganic carbon (DIC) in the culture liquid were determined by using GC and GC/IRMS, respectively, as described above.

The geochemical properties of the six water samples are summarized in Tables 1 and 2. The redox potentials in the freshwater samples from the upper Sarabetsu Formation aquifers (100 and 149 m b.g.s.) were higher (> -210 mV) than those in the brine samples from the Yuchi Formation aquifers (476, 613, 715, and 943 m b.g.s.) (< -290 mV). NO3- and SO42- were detected only in the upper Sarabetsu Formation samples but mostly in small amounts not exceeding 1.4 mgL-1. The water temperature increased with depth (r>0.99, p<0.001; linear regression t test) and ranged from 10.6 to 35.4 ∘C. Stiff diagrams showed a difference in water chemistry between the upper Sarabetsu and Yuchi Formation samples (Fig. S3 in the Supplement), which is consistent with hydrologic separation above and below the clay aquitards (Ikawa et al., 2014). The Cl- concentrations and δD-H2O values of all six water samples were similar to those of pore water within the sediment cores at the corresponding depths (Fig. 1); thus, they showed no sign of cross-contamination among the samples.

In all water samples, CH4 accounted for >75 % of the total dissolved gas (Table 2). The proportions of CH4 and CO2 increased with depth, whereas that of N2 decreased. The stable carbon (δ13C) and hydrogen (δD) isotopic ratios of methane ranged from -77.1 ‰ to -67.5 ‰ and from -258 ‰ to -196 ‰, respectively. Among the samples from the Yuchi Formation, changes in δD-CH4 values were small (7 ‰) compared with changes in δD-H2O values (56 ‰) (also evident in Fig. 2c). Plots of the isotopic ratios of the gases and water, i.e., δ13C-CH4 versus δD-CH4, δ13C-CO2 versus δ13C-CH4, and δD-CH4 versus δD-H2O (Whiticar, 1999) (Fig. 2a–c), indicated a microbial origin of dissolved methane via the CO2 reduction pathway in both the upper Sarabetsu and the Yuchi Formation samples. Methane dissolved in water from the Koetoi Formation, which underlies the Yuchi Formation (Fig. S1), plotted near the boundary between a biogenic and a thermogenic origin (Tamamura et al., 2014) (Fig. 2a). The isotopic difference indicates that the proportion of methane in the upper Sarabetsu and Yuchi Formations derived from the lower Koetoi Formation was minimal.

The TOC contents in sediment core samples from the Yuchi Formation ranged from less than 0.1 % to more than 0.5 % (Fig. S4 in the Supplement). Despite some dispersion associated with lithological changes, the TOC contents showed an overall increasing trend with increasing depth (r=0.77, p<0.001).

The number of microbial cells in the water samples ranged from 1.8×104 to 1.5×106 cellsmL-1 (Fig. 3a). The highest and lowest numbers were measured in the 100 and 715 m b.g.s. samples, respectively, and the microbial cell densities in the studied aquifers are comparable to those reported in other deep aquifers (102 to 106 cellsmL-1) (Pedersen, 1993).

Bacterial and archaeal populations were measured by quantitative real-time PCR (Fig. 3a). The copy numbers of the bacterial 16S rRNA gene were 103–105 mL-1, whereas those of archaea were 102–104 mL-1. In the Yuchi Formation, the copy numbers of bacterial and archaeal genes were the highest in the deepest (943 m b.g.s.) sample. The copy number of the mcrA gene, which is used to estimate methanogen populations, was also the highest in this sample at 3.5×103 genecopiesmL-1, which is 2–3 orders of magnitude higher than the number of copies of the mcrA gene in the other samples.

The 454-pyrosequencing analysis of 16S rRNA genes was performed to examine the compositions of prokaryotic and methanogenic communities and their consistency with the mcrA gene sequencing analysis results, as described below. After quality filtering, the pyrosequencing reads yielded 14 304–43 532 reads per sample. The major taxonomic groups (>5 % of the total reads in at least one sample) belonged to the following phyla or classes: Gammaproteobacteria, Bacillota (formerly Firmicutes), Euryarchaeota, Bacteroidota, Desulfobacterota, and Verrucomicrobiota (Fig. 3b). Gammaproteobacteria sequences were more abundant in the upper Sarabetsu Formation samples, whereas Bacillota and Bacteroidota were more abundant in the Yuchi Formation samples.

The sequences assigned to putative methanogens accounted for 0.2 %–30 % and 11 %–95 % of prokaryotic and archaeal 16S RNA gene sequences (Fig. 3c), respectively. The high proportion of methanogenic sequences in the 943 m b.g.s. sample is consistent with the quantitative PCR results. Sequences assigned to hydrogenotrophic Methanobacteriales were commonly detected at all depths. In the 149 m b.g.s. sample, high proportions of acetoclastic and/or methylotrophic methanogens of the genus Methanosarcina (Methanosarcinales) were detected.

Apart from the conventional methanogens mentioned above, recent studies indicate the potential of archaea outside the classification of Euryarchaeota to produce methane or break down nonmethane alkanes. Specifically, Methanomethylicales, Nezhaarchaeales, Korarchaeia, and Methanodesulfokores have been found to possess methyl-S-CoM reductase, the central enzyme of methanogenesis, while Bathyarchaeota and Asgard possess alkyl-S-CoM reductase (Acr), which activates nonmethane alkanes (Mei et al., 2023). In the archaeal 16S RNA gene sequences of our samples, however, none of these potential methanogens or hydrocarbon-oxidizing archaea were identified.

A total of 64–69 clones of the mcrA gene per sample were grouped into 14 OTUs (Table 3). Similar to the 16S rRNA gene sequencing analysis results, methanogen diversity differed between the upper Sarabetsu and Yuchi Formation samples. Sequences related to acetoclastic Methanosaeta and hydrogenotrophic Methanoregula were abundant in the upper Sarabetsu Formation samples, whereas hydrogenotrophic Methanobacterium sequences were abundant in the Yuchi Formation samples. In the upper Sarabetsu Formation samples, Candidatus Methanoperedens, which oxidizes methane by coupling to nitrate reduction (Haroon et al., 2013), was also detected. Despite its high proportion, a shift in carbon isotope values toward the methane oxidation region on the δ13C-CH4 versus δ13C-CO2 plot (Whiticar, 1999) was not observed in those samples (Fig. 2b).

The 613 and 943 m b.g.s. water samples from the Yuchi Formation aquifers were cultured with methanogenic substrates (i.e., H2/CO2 or acetate) under in situ salinity and temperature conditions to obtain culturable methanogens. Methane was produced from all samples. More than 82 % (v/v) of the maximum theoretical yield of methane was obtained, indicating that the supplied methanogenic substrates were primarily used for methanogenesis.

In the acetate-amended cultures, Methanosarcina dominated, whereas Methanoculleus and Methanobacterium were detected in large proportions in the H2/CO2-amended cultures (Table 3). The sequences of these taxa were almost identical to those obtained directly from the original water samples, indicating that the predominant methanogens inhabiting the saline aquifers were successfully cultured. The diversity of the culturable methanogens was not clearly different between the 613 and 943 m b.g.s. samples.

The methanogen cultures from the 943 m b.g.s. water sample were subsequently cultured under different salinity and temperature conditions based on their depth profiles in the Yuchi Formation (Fig. 1b and d). In the H2/CO2-supplemented cultures, the methane production rate decreased slightly under the lowest salinity condition (i.e., 15 mM Cl-), whereas a more notable decrease was observed under the highest salinity condition (480 mM Cl-) in the acetate-supplemented cultures (Fig. 4a). Temperature changes more drastically affected methanogenic activity than salinity changes (Fig. 4b). In both H2/CO2- and acetate-supplemented cultures, methane production rates were the highest at 35 ∘C. At a temperature of 45 ∘C, which approximately corresponds to that at a depth of 1275 m b.g.s. (assuming a thermal gradient of 2.92 ∘C per 100 m; Fig. 1d), methane production was observed only in the H2/CO2-amended culture.

The SAO activity coupled with hydrogenotrophic methanogenesis (Zinder and Koch, 1984) in the 943 m b.g.s. water sample was assessed by semicontinuous cultivation of SAO microorganisms fed a low concentration of acetate (Shigematsu et al., 2004) (Fig. S2a). After 10 months of cultivation, SAO activity was measured by using [2-13C]-acetate.

Methane was produced stoichiometrically from acetate in cultures supplemented with labeled and nonlabeled acetate (Fig. S2b). The values of δ13C-DIC increased from approximately -25 ‰ to 9 ‰ over time in the culture supplemented with [2-13C]-acetate, whereas no significant change was observed in the culture with nonlabeled acetate (Fig. S2c), clearly indicating SAO activity: in the acetoclastic methanogenic pathway, the methyl group of acetate is converted to methane but not to CO2 (Ferry, 1993), whereas the methyl group of acetate is converted to CO2 and subsequently to methane when SAO is coupled with hydrogenotrophic methanogenesis (Zinder and Koch, 1984).