Study site

Around the MIU construction site, Tertiary sedimentary rocks of the Akeyo, Hongo, and Toki lignite-bearing formations (∼100–200 mbgl) overlie Cretaceous Toki granite [11], [12] (Fig. 1). Near the MIU site, Tono uranium ore deposits disseminated within the Toki lignite-bearing formation have been investigated for the long-term migration of uranium [13]. Geologic, hydrologic and hydrochemical properties of the MIU site have been previously characterized ([11] and reference therein). By using the vertical boreholes, the groundwater chemistry around the MIU site was determined to be Na-(Ca)-Cl type in the lower sedimentary formation and the granitic basement [11]. The boreholes were sectioned by a multi-packer system into six intervals with interval #6 closest to the borehole orifice. The use of cement to seal the borehole orifices inserted with the multi-packer system caused high pH groundwater enriched with Ca. However, hydraulic connectivity between the interval #6 and the other adjacent intervals was low, and geochemical data showed no evidence for groundwater mixing between the intervals. The Japan Atomic Energy Agency granted field permission

Sampling

Groundwater sampling began immediately after the packer installation: 07MI07 borehole in October 2007, 09MI20 borehole in December 2009 and 10MI26 borehole in January 2011. Groundwater data (December 2012) were compiled for this study. It should be noted that the hydraulic pressures of all the investigated borehole intervals sharply increased following to Great East Japan Earthquake (March 11, 2011) [14]. Because all boreholes were artesian, groundwater was directly collected without air exposure via plastic tubing connected to each interval. Before sampling, groundwater was drained extensively (at least three interval volumes) to ensure that groundwater from the geological formation around the borehole was sampled. The second interval from 10MI26 was not investigated because of a lack of groundwater outflow.

For the analysis of dissolved oxygen (DO), groundwater was collected directly into an evacuated glass ampule pre-treated with an indigo carmine solution (Low Range Dissolved Oxygen AccuVac Ampuls; HACH, Loveland, CO, USA). For the analysis of groundwater isotopic composition (δD_H2O and δ¹⁸O_H2O), groundwater was collected into a 1L polycarbonate bottle after rinsing with sample groundwater three times. For the analysis of ion concentrations, including Na⁺, K⁺, Ca²⁺, Mg²⁺, Cl⁻, NO₃⁻, NO₂⁻, SO₄²⁻, NH₄⁺, Fe²⁺ and HS⁻, groundwater was flushed for>8 h through a double-ended stainless bottle (1L) equipped with needle-type stop valves (Swagelok Co., Ltd., Solon, OH, USA). After flushing, the valves were closed, groundwater collected and immediately analysed. For organic acid analysis, including acetate, fumarate, formate, lactate, succinate, malate, pyruvate and citrate, a subsample was filtered through a filter (0.22 µm pore size, type GVWP; Millipore, Billerica, MA, USA) with polypropylene housing, collected into a 50-ml polypropylene bottle (BD Falcon, Becton, Dickinson company, Franklin Lakes, NJ, USA) that was rinsed three times with groundwater, and then stored at 4°C until analysis. For the analysis of dissolved organic carbon (DOC) and humic substances, subsamples were filtered through a pre-combusted (450°C for 4 h) 0.45-µm glass fibre filter (GF/F; Whatman, Buckinghamshire, UK), collected in pre-combusted 30-ml glass vials with screw caps and frozen until analysis.

For the analysis of dissolved H₂, groundwater was collected directly into a pre-evacuated glass vial (65 ml volume, Nichiden-Rika Glass Co., Ltd., Hyogo, Japan) sealed with a butyl grey rubber stopper through a disposable micro-syringe needle and then immediately analysed on site. For the analysis of CH₄ and C₂H₆, and the carbon isotopic composition of CH₄, sampling was the same as for H₂ and followed by the addition of degassed HgCl₂-saturated solution. For analysis of sulfur isotopic compositions of SO₄²⁻ and HS⁻, groundwater was collected through a membrane filter (pore size 0.22 µm, type GVWP; Millipore, Billerica, MA, USA) into a 1L polycarbonate bottle after rinsing three times. Then, 5 ml of zinc acetate (2N) and 0.5 ml of NaOH (5N) were added and the solution was stored at room temperature. Dissolved noble gases were directly collected without air exposure in annealed copper tubes.

Geochemical analyses

The pH and temperature were measured in a flow-through cell by multi-parameter electrodes and probes (Ocean Seven 305; Idronaut, Brugherio, Italy). Measurement precision for pH and temperature was ±0.1 and ±0.1°C, respectively. The concentration of dissolved oxygen (DO) was measured by the Indigo Carmine method with an absorptiometer (DR2800, HACH, Loveland, CO, USA) at a precision of ±0.006 mg/L. The isotopic composition of the groundwater (δD_H2O and δ¹⁸O_H2O) was determined by isotope ratio mass spectrometry (IsoPrime; Isoprime Ltd, Cheadle, UK), following the reduction of H₂O to H₂ and equilibration of H₂O and CO₂. The precision was ±1‰ for δD_H2O and ±0.1‰ for δ¹⁸O_H2O.

For the following analyses, a subsample was filtered through a 0.45-µm membrane filter with polypropylene housing. The concentrations of ionic species Na⁺, K⁺, Ca²⁺, Mg²⁺, NH₄⁺, Cl⁻, NO₃⁻, NO₂⁻, SO₄²⁻ and PO₄³⁻ were analysed with ion chromatography (ICS-1000; Dionex Corp., Sunnyvale, CA, USA). The analytical precision was ±10%. By using ferrous iron reagent powder pillows and hydrogen sulfide reagents (HACH, Loveland, CO, USA), the Fe²⁺ and HS⁻ concentrations were measured colourimetrically with a spectrophotometer (UVmini-1240; Shimadzu Co., Kyoto, Japan) with a precision of ±10%. Concentrations of dissolved organic carbon (DOC) and dissolved inorganic carbon (DIC) were measured by a combustion carbon analyser (TOC-_VCSH; Shimadzu Corp.). The low level of DOC (<70 µM) in 2011–2012 was quantified by thermocatalytic oxidation and MC-NDIR detection (multi N/C 3100; Analytik Jena Japan, Kanagawa, Japan). The organic acid concentrations (acetate, fumarate, formate, lactate, succinate, malate, pyruvate and citrate) were analysed by a high performance liquid chromatograph coupled to a conductivity detector (Prominence, Shimadzu Co., Kyoto, Japan). The precision of the organic acid analysis was ±10%.

The concentration of H₂ was measured with a trace reduction gas detector (TRA-1; Round Science, Kyoto, Japan) based on the mercury oxide to mercury vapour conversion [15]. The TRA-1 consists of a pre-cut column (5 cm long, 1 mm i.d.) and a main column (50 cm long, 2.18 mm i.d.) that is packed with a molecular sieve (5 Å) to separate H₂ from CO and other gases. The eluted H₂ was transferred by H₂-free air into an analytical cell, where the concentration was determined by an ultraviolet spectrometer. The total error, incorporating the sampling and analysis, was ±20%.

Humic substances were separated from the other organic matter and analysed by the high performance size exclusion chromatography (HPSEC) system previously described [16]. The groundwater sample (100 µl) was injected into the HPSEC column with a 20 sample loop at a carrier flow rate of 1 ml/min. The mobile phase was a 0.01 M Tris-HCl buffer solution with 0.01 M NaCl and an adjusted pH of 8.0. Fluorescence was monitored at an excitation wavelength of 320 nm and an emission wavelength of 430 nm as a function of elution volume. Similar to granitic groundwater from the Tono uranium mine [16], groundwater from the MIU site was associated with three fluorescence peaks. Total areas of the fluorescence peaks obtained at a range of elution volume from 8 to 13 ml were subdivided by a factor of 10000. This method does not determine the absolute concentrations of humic substances, but the quantitative comparison among groundwater samples allowed temporal and spatial monitoring of humic substance changes based on the fluorescence peaks.

The concentrations of CH₄ and C₂H₆ and their carbon isotopic compositions (δ¹³C_CH4 and δ¹³C_C2H6) were determined using a Finnigan MAT 252 isotope ratio mass spectrometer (Thermo Finnigan, Bremen, Germany) as described previously [17]. To determine CH₄ and C₂H₆ concentrations and δ¹³C_CH4, an analytical system consisting of a CO₂-trapping port with Ascarite II and a liquid O₂ temperature trap was used. The concentration and carbon isotope precision were ±5% and ±0.2‰, respectively. The hydrogen isotopic composition of CH₄ (δD_CH4) was determined using a CF-IRMS system (Trace GC and Delta-V, Thermo Fisher Scientific, Bremen, Germany). To determine the δD_CH4 values, purified CH₄ was used with a ceramic tube unit (35 cm long, 1 mm i.d.) held at 1350°C to thermally decompose CH₄ to H₂ and C. The analytical precision of the δD of CH₄ was better than 10‰.

DIC, including HCO₃⁻ and CO₃²⁻, in the groundwater was precipitated as SrCO₃ by adding ammonia-concentrated SrCl₂ solution to water samples in the field. The precipitated SrCO₃ was converted to CO₂ through a reaction with phosphoric acid in a vacuum line. The carbon isotopic composition of DIC (δ¹³C_DIC) was determined using an isotope ratio mass spectrometer (Micromass Dual inlet OPTIMA, Waters Corporation, Milford, MA, USA). The sulfur isotopic compositions of SO₄²⁻ and HS⁻ (δ³⁴S_SO4²⁻ and δ³⁴S_HS⁻, respectively) were analysed by an elemental analyser-isotope ratio mass spectrometer (DELTA^plus Advantage ConFlo III System, Thermo Fisher Scientific, Bremen, Germany). The stored water sample was filtered to collect ZnS. While the filter, intended for sulfide analysis, was dried in a freeze dryer overnight, the filtrate, which was for sulfate analysis, was combined with 1N BaCl₂ to precipitate BaSO₄, then dried overnight in a freeze dryer. The filter samples were packed into tin containers and introduced into a Flash EA1110 (Thermo Fisher Scientific, Bremen, Germany) where they were transformed into SO₂. The precision of the sulfur isotope analysis was ±0.3‰.

The concentrations of He and Ne and He isotopic composition were measured with a noble gas mass spectrometer (Micromass MM5400, Waters Corporation, Milford, MA, USA) located at the Geological Survey of Japan. Technical details of the dissolved noble gas extraction and mass spectrometry, including the purification procedures, were described previously [18]. The reproducibility of the ³He/⁴He ratio as well as the He and Ne concentrations obtained from replicated measurements of air saturated water (ASW) were about 2%, 3% and 3%, respectively.

Electron microscopy

In 2011, suspended particles in groundwater were collected on 0.22-µm pore size, 25-mm diameter polycarbonate filter (Advantec, Tokyo, Japan). The filter was dehydrated with 100% ethanol and dried in a desiccator. The portion of the filter was coated with Pt-Pd and observed by SEM by using a Hitachi S4500 equipped with energy dispersive X-ray (EDX) spectroscopy at 5-kV and 15-kV accelerating voltages for secondary electron imaging and EDX analysis, respectively.

Thermodynamic calculations

Saturation indices of calcite (CaCO₃) and gypsum (CaSO₄·2H₂O) in groundwater samples were calculated by PHREEQ C [19]: where Ksp is the solubility product of a mineral and IAP is the ion activity product. When SI  =  0, the groundwater is in equilibrium with a mineral. When SI>0 and SI <0, a mineral precipitates and dissolves, respectively.

In situ Gibbs reaction energy (ΔG) of hydrogen-oxidizing sulfate reduction: was calculated using the following equation:where ΔG⁰(T, P) is the Gibbs standard reaction energy at in situ temperature and pressure, R is the ideal gas constant, T is groundwater temperature, and Q denotes the activity of the reactants and reaction products. The ΔG⁰(T, P) was calculated using the SUPCRT92 software package [20]. The activities of reactants and products were calculated using the Geochemist's Workbench software package [21].

Mass balance equation for groundwater mixing and microbial consumption of sulfate

To calculate the amount of microbially reduced sulfate in the granitic aquifer, a mass balance equation was derived that incorporates the sulfate dilution by mixing of shallow and deep groundwater and the removal of sulfate by microbial reduction: (1)where δ³⁴S_ini^SO₄2− and [SO₄²⁻ini] denote the δ³⁴S value and concentration of sulfate initially recharged to the granitic aquifer, [SO₄²⁻mix] is the sulfate concentration diluted by groundwater mixing, δ³⁴S_SRB^SO₄2 and [SO₄²⁻SBR] are the δ³⁴S value and concentration of sulfate removed by microbial sulfate reduction, and δ³⁴SGW^SO₄2− and [SO₄²⁻GW] are the δ³⁴S value and concentration of sulfate measured in groundwater samples. For the calculation, it was assumed that deep groundwater is completely depleted in sulfate, based on data from Iwatsuki et al. [11], and that groundwater mixing is not accompanied by sulfur isotopic fractionation. Because groundwater flow is inferred to be slow from the residence time of groundwater [22], the mass balance equation was developed as a closed system.

Because [SO₄²⁻mix] can be expressed as: (2)

Replacing [SO₄²⁻mix] in Eq. (1) with [SO₄²⁻ini] − SO₄²⁻GW] − [SO₄²⁻SRB] gives:(3)

The expansion of Eq. (3) is:(4)

Elimination of δ³⁴S_ini SO₄²⁻ × [SO₄²⁻ini] results in: (5)

Integrating common factors gives: (6)

Thus, [SO₄²⁻SRB] is calculated as:(7)

δ³⁴S_SRB seems to be calculated:(8)

However, the shift in Δ³⁴SSO₄²⁻−HS⁻ from the initial state to the present state was taken into account. Δ³⁴SSO₄²⁻−HS⁻ was linearly correlated with sulfate concentration (R²  =  0.703): (9)

To derive δ³⁴S_SRB SO₄²⁻, Δ³⁴S_ini→GWSO₄²⁻−HS⁻ was calculated: (10)where a Δ³⁴S_ini SO₄²⁻−HS⁻ value and [SO₄²⁻ini] were set to be 52.29‰ and 180 µM, respectively.

δ³⁴S_SRB SO₄²⁻ was finally calculated: (11)

δ³⁴S_ini SO₄²⁻ was assumed to be 10‰, based on data from Iwatsuki et al. [12]

Hydrogeochemcal disturbance of the granitic aquifer

As evident from δD_H2O and δ¹⁸O_H2O values (S1 Table in S1 File), seawater has been flushed by meteoric water at great depth since 5 Ma [11]. Based on vertical borehole investigations prior to the shaft construction [11], the original depth profiles of dissolved chemicals have been determined. Chloride linearly increases with depth (R² = 0.950, Fig. 2A), whereas sulfate and dissolved inorganic carbon (DIC) linearly decrease with depth (R² = 0.664 and 0.843, respectively). The concentrations of sulfate and DIC are similarly correlated with chloride concentration (R² = 0.659 and 0.802, respectively). From the linear depth profiles of chloride, sulfate and DIC, it is inferred that shallow groundwater with low chloride and high sulfate and DIC infiltrated deep groundwater with high chloride and low sulfate and DIC. After the shaft excavation, the original depth profile of chloride (Fig. 2B) was dramatically changed by the upward movement of deeper groundwater [10].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. The correlation of depth with chloride in the groundwater before and during shaft construction.

(A) Pre-excavation correlation from DH-2 borehole, which was drilled ∼100 m south of the ventilation shaft before the shaft construction. (B) Post-excavation correlation from underground boreholes. Each different circle colour represents annual differences in sampling date. A straight line was fitted through the data by least squares and regression analysis. The deviation from linearity is expressed as a correlation coefficient (R²).

https://doi.org/10.1371/journal.pone.0113063.g002

Previously, extensive groundwater mixing has been recognized during the construction of underground shafts in granitic aquifers in Canadian and Fennoscandian Shields [23], [24]. Sampling of deep groundwater originally distributed in the granitic aquifer seems to be technically impossible where hydraulic conductivity is high due to high facture connectivity. In contrast, sampling of deep granitic groundwater from a borehole drilled from the surface is not suitable for accurate determination of the concentrations of redox sensitive species, because of the extensive intrusion of oxygenated drilling fluid into the aquifer. Because horizontal boreholes 07MI07, 09MI20 and 10MI26 were drilled mostly with outflowing groundwater instead of artificial drilling fluid, drilling disturbance was minimized. Additionally, groundwater sampling for gaseous species was conducted at substages under in situ hydraulic pressure conditions, which could avoid degassing during pumping groundwater from the surface.

By taking the advantages and disadvantages of groundwater sampling from the underground facility into account, it was attempted to reconstruct the original biogeochemical profiles by long-term monitoring of the concentrations of aqueous and gaseous species. Despite the hydrologic disturbance (S1 Figure in S1 File), linear correlations with chloride were maintained for sulfate (R² = 0.801, Fig. 3A) and DIC (R² = 0.662, Fig. 3B) over the monitoring periods. Methane (CH₄) also linearly increased with chloride (R² = 0.779, Fig. 3C). Because chloride is well known as a conservative tracer in groundwater [25] and had clear correlation with depth before shaft excavation, the depth profiles of sulfate, DIC and CH₄ were reconstructed based on the corresponding chloride concentrations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Correlations of chloride concentrations with biologically metabolized compounds and their isotopic compositions: (A) sulfate concentrations, (B) concentrations of dissolved inorganic carbon (DIC), (C) methane concentrations, (D) δ³⁴SSO₄²⁻ values, (E) δ¹³C_DIC values and (F) δ¹³C_CH4 values.

Each different circle colour represents annual differences in sampling date. A straight line was fitted for all obtained data by least squares and regression analysis. The deviation from linearity is expressed as a correlation coefficient (R²). Isotope standards for δ¹³C and δ³⁴S values were Vienna Pee Dee Belemnite and Cañon Diablo meteorite, respectively.

https://doi.org/10.1371/journal.pone.0113063.g003

Biogeochemical processes controlling sulfate, DIC and CH₄ profiles

All groundwater samples were near equilibrium with calcite (CaCO₃). Because near-equilibrium of calcite has been maintained prior to the shaft construction [11], it is not likely that calcite precipitation was promoted by groundwater mixing. Additionally, all groundwater were highly undersaturated with respect to gypsum (CaSO₄·2H₂O) (S1 Table in S1 File). Hence, the decreases in the sulfate and DIC concentrations with depth were not attributed to mineral precipitation reactions.

To clarify the influences of microbiologically mediated processes on the depth trends of sulfate, DIC and CH₄, carbon and sulfur isotopic compositions were analysed (Figs. 3D–F, S3 Table in S1 File). By plotting the δ³⁴S values of sulfate, ³⁴S enrichment with increasing chloride concentration was found (Fig. 3D) that indicates the involvement of microbial sulfate reduction [26]. In contrast, the δ¹³C values of DIC and CH₄ were relatively constant over a wide range of chloride concentrations (Figs. 3E–F), which indicates mixing of shallow and deep groundwater with contrasting DIC and CH₄ concentrations. Because large negative and positive δ¹³C shifts of CH₄ were not evident, it is likely that microbial production and consumption of CH₄ were not substantial enough to overprint the simple mixing profile [27]. Because acetate was not detected in the groundwater (S2 Table in S1 File), microbial reduction of CO₂ to acetate might be negligible.

It is well established that in low temperature aqueous systems, microbial reduction is a dominant process for the removal of sulfate, accompanied by the preferential consumption of isotopically light sulfate and the production of isotopically light hydrogen sulfide [28]. In contrast to sulfate, hydrogen sulfide concentrations were not linearly correlated with chloride (Fig. 4A). Mineral particles suspended in groundwater were collected by filtration and observed by SEM equipped with EDX analysis. Groundwater from #4 interval of 09MI20 borehole. This interval was selected because sulfide concentrations were relatively low over the monitoring period. As a result, mineral particles composed of Zn, Fe and S were detected (Fig. 4E), which suggests that the lack of linearity was caused by the precipitation of metal sulfides in the aquifer. Microbial mediation of sulfate reduction coupled to sulfide production is strongly suggested by ³⁴S depletion in sulfide relative to sulfate (Figs. 4B–C, S3 Table in S1 File), although the δ³⁴S fractionation was significantly reduced after sampling in 2011. This abrupt shift in sulfur isotopic fractionation might have been caused by microbial sulfate reduction influenced by recent groundwater mixing and the following stabilization of groundwater (S1 Figure in S1 File) [10]. Two reaction pathways for large sulfur isotopic fractionation are known: microbial sulfate reduction and sulfur disproportionation [26]. Although sulfur disproportionation may have proceeded under micro-aerobic conditions caused by drilling disturbance, the sulfur disproportionation fractionation results in ³⁴S-enriched sulfide and ³⁴S-depleted sulfate [26]. Because sulfate reduction is associated with larger sulfur isotope fractionation between sulfide and sulfate than sulfur disproportionation, the decreases in fractionation after 2011 are counterintuitive because sulfur disproportionation should diminish and be succeeded by sulfate reduction.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Correlations of chloride concentrations with (A) hydrogen sulfide concentrations, (B) δ³⁴S_S²⁻ values, (C) Δ³⁴S_{SO4²⁻-HS⁻} values, and (D) estimated amounts of microbially reduced sulfate and (E) secondary electron image and EDX spectrum of suspended particles in groundwater from the #4 borehole of 09MI20 borehole.

Estimated amounts from 2011–2012 are not included because of the abrupt shifts in δ³⁴S_S²⁻ and Δ³⁴S_{SO4²⁻-HS⁻} values between 2008–2010 and 2011–2012. The shifts might be caused by groundwater mixing after shaft construction. Each different circle colour represents annual differences in sampling data.

https://doi.org/10.1371/journal.pone.0113063.g004

H₂ is a central electron donor for microbial energy generation and not easily transported in groundwater because of continuous cycling by microorganisms [29]. Thermodynamic calculations and laboratory- and field-based measurements consistently report that terminal electron accepting processes (TEAPs) have different affinities to H₂ uptake: <0.1 nM for O₂ and nitrate reduction, 0.1–1 nM for Fe(III) reduction, 1–5 nM for sulfate reduction, and 5–30 nM for CO₂ reduction [30]. H₂ is therefore considered a sensitive indicator for an in situ TEAP occurring in groundwater. Along with the sulfur isotopic profiles in the aquifer, most H₂ concentrations were within a range characteristic for sulfate reduction (Fig. 5A, S2 Table in S1 File). Although an increase in H₂ concentration was temporarily observed in 2011, H₂ concentrations dropped to previous levels in the next year.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Correlations of chloride concentrations with potential energy sources and available energy from H₂-oxidizing sulfate reduction: (A) H₂ concentrations, (B) the concentrations of dissolved organic carbon, (C) relative peak areas from humic substances semi-quantified by high-performance size exclusion chromatography, (D) calculated Gibbs energy yields from H₂-oxidizing sulfate reduction at in situ temperature and pressure at H₂ concentrations found in the granitic aquifer.

Each different circle colour represents annual differences in sampling date. A straight line was fitted for all obtained data by least squares and regression analysis. The deviation from linearity is expressed as a correlation coefficient (R²).

https://doi.org/10.1371/journal.pone.0113063.g005

Potential energy sources for sulfate reduction

Organic carbon typically serves as an energy source for microbial sulfate reduction. The concentration of dissolved organic carbon (DOC) was barely detectable and not correlated with chloride (Fig. 5B). In contrast, highly sensitive measurements, established for humic substances derived from surface vegetation and overlying sedimentary rocks in the study area, revealed a clear negative correlation with chloride (Fig. 5C; R² = 0.571). As humic substances are less available for metabolic activities, it is reasonable that humic substances decrease with depth by groundwater mixing similarly to sulfate and DIC. If sulfate reducers use organic carbon for energy and carbon sources, the concentration of DOC is supposed to decrease with depth. The lack of correlation with chloride might be explained by authigenic production of DOC. The negligible δ¹³C shift of DIC with chloride might be explained by the small amount of authigenic organic carbon isotopically depleted in ¹³C compared with the DIC pool of groundwater. Because it is also known that fermentation of organic matter produces acetate in the deep terrestrial subsurface [6], there is the possibility that the low acetate levels could be minor energy and carbon sources for sulfate-reducing microorganisms.

Alternative electron donors available in the sulfate-reducing aquifer are CH₄ and H₂, It is known that anaerobic oxidation of methane is suppressed at sulfate concentrations below 0.5 mM [31] and/or at H₂ concentrations above 0.1 nM [32]. Therefore, H₂ best serves as an alternative electron donor in the granitic aquifer. In situ Gibbs energy of H₂–oxidizing sulfate reduction was calculated from the activities of products and reactants. The Gibbs energy yield at 1 nM H₂ was −13.6±2.7 kJ mol⁻¹ (Fig. 5D), which is substantially lower than −23±1.2 kJ mol⁻¹ reported as the minimum Gibbs energy required for H₂–oxidizing sulfate reduction in anoxic sediments [30]. These lower values might reflect an optimization of energy-conserving efficiency by indigenous microorganisms in the H₂–limited granitic aquifer. The Gibbs energy yield of −29.5±2.7 kJ mol⁻¹ at 5 nM H₂ appears to be sufficient to conserve energy by H₂–dependent sulfate reduction.

Interestingly, an intensive H₂ flux has been detected in a granitic cataclasite outcrop on an active fault 33 km northeast of the site [33]. Because a vertical fault intersects the underground MIU facility (Fig. 1), the fractures associated with the fault might serve as a fast conduit for deeply produced H₂. As previously reported from the surrounding area [34], isotopic analyses of noble gases revealed high ratios of mantle-derived ³He relative to radiogenic ⁴He in groundwater from the monitoring boreholes (S4 Table and S2 Figure in S1 File). Thus, H₂ generated in the mantle by high temperature reactions among dissolved gases in the C–H–O–S system [35] might be supplied along with ³He to the granitic aquifer.

Hydrogeochemical estimation of in situ rate of sulfate reduction

The amount of microbially reduced sulfate in the granitic aquifer was quantified by solving a mass balance equation. For the calculations, it was assumed that initial sulfate (δ³⁴S_iniSO4²⁻) with a value of 10‰ is mixed with deep sulfate-depleted groundwater without sulfur isotopic fractionation. It was also assumed that microbial reduction results in the removal of sulfate so the extent of ³⁴S fractionation in groundwater was defined as Δ³⁴S_{ini→GWSO4²⁻−HS⁻. Data after sampling in 2011 were excluded because Δ³⁴SSO4²⁻−HS⁻ values were sustantially shifted (Fig. 4D), possibly by the effects of shaft excavation and groundwater mixing on metabolic activities. The amount of microbially reduced sulfate ranged from ∼5–25 µM (S5 Table in S1 File and Fig. 4E), which is roughly consistent with the range of HS⁻ concentrations (Fig 4A). The sulfur isotopic signatures of sulfate and hydrogen sulfide along the groundwater flow path are often modelled by Rayliegh fractionation [4], because the deep aquifer can be treated as a closed system due to mass transport limitation. As Δ³⁴SSO4²⁻−HS⁻ was varied with sulfate concentration, Rayliegh fractionation model that requires the constant fractionation was not applicable to this investigated aquifer.}

Groundwater from the lower unit of the overlying sedimentary formation has been previously estimated by ¹⁴C content to have a residence time of 9.3 ka [11]. ⁴He age of granitic groundwater was measured to be ∼10 ka between recharge and middle areas in the study area [22]. ⁴He age of discharge area where the MIU is located was difficult to determine, because groundwater is associated with high ³He/⁴He ratio and enriched with ⁴He (S2 Figure in S1 File). From the age estimates based on ¹⁴C and ⁴He, it is therefore assumed that the residence time of groundwater in the discharge area is, at least, ∼10 ka. If the amount of microbially reduced sulfate (5–25 µM) was consumed for ∼10 ka, the in situ rate of microbial sulfate reduction is estimated to be 1–5 nM yr⁻¹. This rate range is slightly higher than those from well-characterized deep biospheres, such as the diverse microbial community in sub-seafloor sediments (0.66 nM yr⁻¹) [36] and the microbial ecosystem at the South African Gold Mine (0.22−1.45 nM yr⁻¹) [4].

The existence of low-energy crustal biosphere

Temperature and the concentrations of electron donors and sulfate are known to affect the kinetics of sulfur isotope fractionation in microbial sulfate reduction because sulfate transport across cell membranes and enzymatic transformations of sulfate to sulfide are both rate-limiting [26]. The sulfur isotopic fractionation is generally considered to be less than 10‰ at the low sulfate levels (<0.2 mM) based on a lack of fractionation demonstrated in a variety of laboratory culture conditions [37]. It is also well known that microbial respiration rates in nature are overestimated by 3–6 orders of magnitude because in situ conditions are difficult to reproduce in laboratory-based activity measurements [9], [38]. When a laboratory culture is limited by H₂ supply at sulfate concentrations of ∼0.2 mM, a large fractionation is observed and correlated with a reduced rate of sulfate reduction [39]. Because the in situ rate of sulfate reduction and the concentration of H₂ appear to be substantially lower in the granitic aquifer than those manipulated in the laboratory, large fractionation values appear reasonable under extreme energy limitation. Indeed, very large sulfur isotopic fractionation in low sulfate groundwater affected by sulfate reduction has been known for a long time in deep sedimentary aquifers [40], [41]. The mechanism of this large fractionation remains to be shown; the kinetics of enzymatic activities from sulfate to sulfide might be slower than that of sulfate transport across cell membrane due to low-energy fluxes [26].

The existence of a H₂-based deep biosphere has been previously reported from groundwater in the Columbia River Basalt. However, the production of H₂ by low-temperature interactions between basalt and water is controversial with respect to the long-term sustainability of subsurface lithotrophic microbial ecosystems (SLiMEs) [42]. Terrestrial and deep-sea hydrothermal fluids are colonized by thermophilic and hyperthermophilic methanogenic hydrogenotrophs, suggesting that hydrothermal activity is generally essential for the production and supply of H₂ [4]–[6]. The deep biosphere unveiled in this study is unique in that the crustal habitat is supplied with low fluxes of energy sources without hydrothermal circulation. Our findings have significant implications for the disposal of nuclear wastes because microbes can establish reducing conditions under low-energy fluxes, which would substantially decrease the mobility of many radionuclides [43].