Serpentinite-associated waters were collected for geochemical analysis from 7 sites in the Zambales Ophiolite, ranging from artesian wells to travertine-depositing springs, and 2 springs in the Palawan Ophiolite. Sampling sites are identified in Figure 1; additional detail is provided in Figure 2. A YSI556 Multiparameter System was used to measure simultaneously in the field dissolved oxygen, pH, conductivity, temperature, and oxidation-reduction potential. A portable field spectrophotometer (HACH DR 2800) with commercially available reagent ampules enabled quantification of redox-sensitive chemical species (e.g., nitrate, nitrite, sulfide, ammonia). Water samples for laboratory analysis were collected with triple-flushed syringes (60-ml volumes with luer-lok tips; BD Medical No. 309653), and filtered through syringe filters or through 0.22 μm pore size Millipore Sterivex PVDF filters, assisted by peristaltic pump. Filtered solutions dedicated for ion chromatography analyses were stored cold in clean, triple-rinsed Nalgene bottles, while filtered waters dedicated for elemental analysis were stored in clean, acid-washed sampling bottles (caps were parafilmed in the field to minimize evaporation), and acidified with nitric acid for a final concentration of ~2% nitric acid.

For the 2012 field season, ions were characterized at Arizona State University (ASU). Samples for analysis of cations and anions were filtered into 60 ml Nalgene bottles prepared by soaking overnight in an acid bath. Bottles were kept frozen until analysis. Samples for major ions were filtered in the field through a series of 0.8 and 0.2 μm pore size filters (such as Supor filters made of hydrophilic polyether-sulfone, Pall Scientific). As quickly as possible after sampling, major cation (Na⁺, K⁺, Ca⁺², Mg⁺²), and anion (SO⁻²₄, Cl⁻, Br⁻, F⁻) concentrations were measured in the laboratory using ion chromatography (anions: Dionex IonPac AS11 analytical and IonPac AG11 guard columns; cations: Dionex IonPac CS12A analytical and IonPac SG11 guard columns; conductivity detection). Analytical uncertainties are on the order of 5%.

For the 2013 field season, concentrations of some major anions (chloride, bromide, nitrite, nitrate, phosphate, sulfate) were measured with a Dionex DX-120 ion chromatograph (University of Rhode Island, Kingston RI) outfitted with an IonPac AS22 Analytical Column (4 × 250 mm) and IonPac AG22 Guard Column (4 × 50 mm). A 5 point calibration was based on analysis of gravimetrically prepared calibration standards; calibration proved linear with r² ≥ 0.999 for all anions tested. Accuracy of data for Cl⁻ is ±3%, for NO⁻₂ is ±5%, Br⁻ is ±20%, NO⁻₃ is ±1%, PO³⁻₄ is ±1%, SO²⁻₄ is ±10%, based on multiple analyses of the Dionex Seven Anion Standard (PN 056933; Lot Number 20-25VY).

Elemental concentration data were collected with a Thermo Scientific iCAP Q Inductively Coupled Plasma Mass Spectrometer (housed in the Rhode Island IDeA Network for Excellence in Biomedical Research Lab). Calibration was made using the Ultra Scientific ICP-MS calibration standard #2 (Cat.#: IMS102, Ultra Scientific, North Kingstown, RI), presented to the instrument at concentrations of 100, 500, and 1000 ppb in 2% nitric acid. Linear calibration across this concentration range was observed with r² ≥ 0.999 for all analytes except Fe(57), for which r² ≥ 0.9875. An external reference solution (ICP-SS-50, High Purity Standards, Charleston, SC) was run as an unknown to assess accuracy, which differed by analyte. Uncertainty was observed as ±10% for Na(23), Mg(24), Al(27), K(39), V(51), Cr(52), Mn(55), Fe(57), Co(59), Ni(60), Cu(63), As(75), Se(82), Sr(88); Ca(44) data carry a larger uncertainty of ± 30%. Precision was typically <1%, as shown by very low relative standard deviations for triplicate analyses.

Two milli liter volumes of filtered water were stored in vials with PTFE/silicone screw tops (Fisherbrand cat. no. 03-391-15) and analyzed with a Picarro Water Isotope Analyzer L1102i (Brown University Environmental Chemistry Facilities, Providence, RI). In practice, 2 microliters of liquid water were injected into the vaporizer module, which prepared and sent the sample to the laser cavity for measurement. Each sample was measured in triplicate and corrected for memory and drift (using the protocols of Vaughan and Claymore, INSTARR, Univ. of Colorado). All samples were measured with respect to internal lab standards, which were calibrated directly to VSMOW, SLAP and GISP2. Data are reported here with respect to VSMOW; overall precision and accuracy is better than 0.1‰ and 1.0‰ for δ¹⁸O and δD, respectively.

Sampling vials were amber I-CHEM vials. DIC septa were butyl/PTFE; DOC septa were silicone/PTFE. Prior to sampling, DIC bottles and septa were acid-washed (overnight soak in 10% HCl by volume solution, rinsed with high purity water of ~18 MΩ resistivity), then dried and assembled. DOC bottles were combusted at 500°C overnight, septa were rinsed, then bottles were spiked with 100 ul of ASC grade 85% phosphoric acid. In the field, all fluid was filtered. Collection bottle [not sampling bottle] was flushed with sample 3x, as was tubing and syringe. DOC bottles were filled using a filter that had been conditioned with several liters of sample. Both DIC and DOC bottles were filled to the top with no air bubbles and kept at ~20°C until analysis. DOC and DIC concentrations were measured with an OI Analytical Model 1010 Wet Oxidation Total Organic Carbon (TOC) Analyzer at ASU, as in Meyer-Dombard et al. (2015). Fluids were reacted with phosphoric acid (DIC) or sodium persulfate (DOC), and CO₂ was analyzed by continuous flow into a Thermo Delta^Plus Advantage mass spectrometer. Three glycine working standards characterized with USGS40 and USGS41 isotopic reference materials were used that encompass expected isotopic variations (low: δ¹³C = −39.64‰, δ¹⁵N = 1.35‰; mid: δ¹³C = −8.36‰, δ¹⁵N = 27.9‰; and high: δ¹³C = 15.67‰, δ¹⁵N = 51.8‰).

10 mL serum vials and blue butyl stoppers were acid-washed (overnight soak in 10% by volume HCl and rinsed in ~18 MΩ Milli-Q distilled water). Vials and stoppers were allowed to air-dry, assembled, and crimped shut prior to flushing (Wheaton, Millville, NJ, USA). Gas sampling vials were flushed with Ar gas at a rate of 8–10 psi pressure: briefly, a 23 gauge needle was inserted into the sealed vial, connected to a Luer-Lock-equipped tubing (BioRad Laboratories, Inc., Hercules, CA, USA) to the Ar tank. To ensure vials were not over-pressured, a venting needle attached to an open-ended 60 mL syringe filled with 10–15 mL of distilled water suspended above the gassing station to allow excess gas to vent without allowing atmospheric gas back into the vial. Flushed sampling vials were transported to the field, and filled with 5 mL of sample water using the following protocol: samples were collected with a syringe, and air bubbles were forced out of sample volume by depressing the syringe stopper. The sampling syringe was flushed 3X, and samples were collected slowly to avoid the draw of vacuum. Samples were poisoned in the field with an 0.5 mL injection of Hg(NO₃)₂ (1.71 g/50 mL) and stored upside-down until analysis. Upon return to the USA, samples were immediately analyzed on a Shimadzu GC-2014 gas chromatograph/mass-spectrometer (GC-MS) equipped with flame ionization and thermal conductivity detectors (FID and TCD, respectively), a Carboxen® 1000 60/80 molecular sieve (Sigma-Aldrich Co., LLC, St. Louis, MO, USA), with a 1/8″ × 15′ column (Shimadzu Corporation, Kyoto, Japan). C1 gases of interest were first methanized (converted to CH₄), combusted, and detected with the FID. H₂ and Ar were detected on the TCD. Depending on sample concentration, 200–400 μl of the gas in the sampling vial headspace were injected into the GC-MS column. Calibration curves were constructed using H₂, CO, CH₄, and CO₂ standard gases (Sigma-Aldrich Co., LLC, St. Louis, MO, USA), both during and 6 months after analysis; the calibration curves for all gases analyzed suggested that the GC-MS was stable and results were highly reproducible.

Gibbs energy (ΔG_r) calculations were used to determine thermodynamically favorable reactions that can serve as microbial metabolisms supported by the environment under study. Gibbs energy calculations can be used to identify points in temperature-pressure-composition space where metabolisms are most favorable.

Values of ΔG_r at the temperature, pressure, and chemical composition at the sampled sites were computed with relation (3):

where ΔG°_r denotes the standard state Gibbs energy values at the temperature and pressure of interest (ΔG°_r for chemical species pertinent to this work are readily available in Amend and Shock, 2001), R represents the gas constant, T represents temperature in Kelvin, and Q_r is the activity product.

Q_r, the activity product, can be computed from environmental data as shown in relation (4):

where a_i represents the activity of the ith species, and Υ_i, _r represents the stoichiometric reaction coefficient. Values of a_i are generated from concentration data (Table 1) and are modified by normalization with reasonable activity coefficients, using the geochemical speciation code EQ3 (Wolery, 1992). In this code, activity coefficients are calculated using a variant (B-dot equation) of the Extended Debye-Hückel activity coefficient formalism (Helgeson, 1969), with reference to the SUPCRT92 (Johnson et al., 1992) thermodynamic database. Total dissolved solids (TDS) data, which must be known to quantify the ionic strength of the solution, were estimated from field conductivity data and included in EQ3 input files (see Supplementary Material). Note that EQ3 provides the equilibrium speciation of co-occurring geochemical components based on the data inputs. If data inputs lack a particular chemical species (e.g., acetate), EQ3 provides a rigorous estimate of its activity in the modeled environment, depending on thermodynamic constraints.

It is worth noting that ΔG°_r can be calculated at the appropriate temperature and pressure for the aqueous species and minerals using established equations of state (Helgeson et al., 1978, 1981; Shock and Helgeson, 1988, 1990; Tanger and Helgeson, 1988; Shock et al., 1989, 1992), or computed from data for ΔG°_r and ΔG°_i at temperatures up to 200°C at P_sat, available in Amend and Shock (2001). Conventionally, ΔG°_r can also be calculated from Gibbs energy of formation values, using relation (5):

In which ΔG°_r is the standard state Gibbs energy of reaction r, Υ_i,r is the stoichiometric reaction coefficient of the ith species in reaction r, which is negative for reactants and positive for products, and ΔG°_i is the standard Gibbs free energy of formation of the ith species at the temperature and pressure of interest in reaction r.

Waters can be grouped into four types based on dissolved gas constituents. ML1 and PB3 bear CO, CH₄, and H₂. ML2, PB2, and SI1 bear CO₂, CH₄, and H₂. BB1 is rich in CO₂ only. MF1 had overall low gas contents, although CH₄ was observed. Taken together, these springs cover a range in gas contents (Table 3, Figure 6).

Given that (a) gas contributions from a subsurface zone undergoing active serpentinization is expected to yield H₂ and CH₄ and (b) extremely high pH, Ca²⁺-OH⁻ waters are inherently poor in aqueous CO₂ since they have been isolated from the atmosphere, and are thus poor in DIC, first order conclusions can be drawn from Figure 6. Dissolved gases at Bigbiga (BB1) consist essentially of CO₂. This spring stands aparts from the others as a CO₂-effervescing spring with relatively oxidized gas phase carbon, and no detected reduced gases. It is likely open to the atmosphere at some point in its recharge path. H₂ and CH₄ are observed at low concentrations at site PB2, with the gas balance as CO₂. At PB2, the data can be interpreted as reflecting mixing between a CO₂-bearing, surface-derived, shallow groundwater regime and a separate, deeply sourced water carrying reduced gases. ML1, ML2 are nearly 50%—50% mixtures of H₂ and CH₄, with a small proportion of CO₂, if present. Gas contents at ML1 and ML2 indicate that these waters provide a plausible, deeply sourced end-member composition.

Of these newly characterized sites in the Philippines Ophiolites, the Poon Bato springs PB1 and PB3 are typical Ca²⁺-OH⁻ type waters (c.f., Barnes et al., 1967), and can be interpreted as sourced in actively serpentinizing host rock; PB2 has relatively higher Mg²⁺ content, indicative of some mixing with open system Mg²⁺-HCO⁻₃ waters. The Manleluag springs ML1–ML3 express a muted serpentinization signal in the aqueous geochemistry, with less Ca²⁺ and lower pH. Given that all country rock observed in the Manleluag area was gabbroic (deep ocean crustal rocks rather than peridotitic mantle units), waters may be escaping along a subsurface mafic-ultramafic contact (i.e., the interface between the gabbros and altering peridotites, or the interface between the crust and the mantle rocks). The gas data for PB and ML sites support this interpretation, indicating incorporation of H₂ and CH₄ likely derived from regional serpentinization.

BB1 well water samples, known to be sourced in altered pillow basalts may have some hydrological communication with extensive, adjacent bentonite clay sedimentary deposits. These waters have a lower pH signal, higher salinity, and are the only waters in which sulfate was detected in both sampling years, perhaps due to perennial leakage from sedimentary formations. Note the changing scale in Figure 4, with increasing salinity for BB1; SI1 and MF1 waters also have enhanced salinity, dramatically so for MF1, with range up to 12 meq/kg in Figure 4, along with relatively reducing oxidation-reduction potential. The relatively high salinity coupled with the weak Ca signal suggests that these waters (MF1, SI1) are groundwaters isolated from serpentinization reactions in the subsurface.

Variation in DIC and DOC between sites and at the same sites in different years is unpredictable (Table 2). These field locations are some of the first described in tropical climates, and are likely heavily influenced not only by seasonal variations in precipitation, but also vegetative input to the springs. As some site locations were sampled in different years, and under different climate conditions, these unpredictable variations can be observed (Table 4). DIC is typically low in sites with corresponding low ORP, and high in locations that have more opportunity to equilibrate with atmospheric CO₂ (e.g., BB1, PBR, MF1). Comparing 2012 vs. 2013 values, which represent dryer vs. wetter climate conditions, respectively, concentrations of DIC increase at some sites and decrease at others. Presumably, the underlying hydrology in combination with surface influences cause these variations. The isotopic composition of DIC in these fluids ranges from ~ −11 to −25‰. Values that approach ~ −7‰ likely indicate interactions with atmospheric CO₂, in other words, an open system with respect to CO₂, while more depleted values may reflect microbial waste product build up in springs that are somewhat isolated from surface processes. In comparison to other terrestrial serpentinizing seeps (few, where data are available), DIC at most Philippines sites have enriched isotopic compositions. For example, The Cedars (CA, USA) and Tablelands (Canada) locations host fluids with DIC that show isotopic compositions between ~ −30 to −32‰ (Morrill et al., 2013; Szponar et al., 2013). Again, DIC that is isotopically enriched may indicate mixing with fluids that are more associated with surface processes. Regardless, low DIC concentrations in the Philippines systems mean that subsurface microbial ecosystems (1) may be carbon poor, and (2) any microbial fractionation during carbon fixation will be starting from carbon that is less enriched in ¹³C than surface fluids. This latter will have implications for any additional studies of carbon isotope compositions of biomass.

DOC concentrations are similarly unpredictable, between sample locations and sample years. The degree of additional organic carbon input from the surrounding heavily vegetated surface environment is a variable that cannot be controlled. Thus, DOC may play a large role in microbial metabolic function in these particular springs that other terrestrial serpentinizing springs lack (such as those in more arid climates, for example), even though the overall concentration of DOC in these systems appears to be relatively low. It might be expected that DOC concentrations would increase with increased precipitation. However, with only a few exceptions, this is not the case in the Philippines field sites. In contrast, increased precipitation may have diluted the DOC, or contributed to otherwise flushing vegetative build up from the pools. Isotopic composition of the DOC is, predictably, depleted relative to that of DIC at all sites except for the 2012 PB1 sample. DOC in other terrestrial serpentinizing systems has also been reported to be low (<2 ppm—Morrill et al., 2013; Szponar et al., 2013). However, the Philippines systems may have benefit of a more consistently supplied influx of surface derived DOC, due to the dense vegetation and frequent flushing (and refreshing) of the system during precipitation events. DOC isotopic compositions at these sites are more depleted in comparison with DOC reported from the Tablelands and the Cedars locations (Morrill et al., 2013; Szponar et al., 2013)—this may indicate less recycling of carbon in the system compared to other terrestrial sites. Further, heterotrophic metabolism in the subsurface should produce isotopically heavier CO₂ as a waste product, which would be added to any native DIC.

We modeled the activities of chemical species important for microbial metabolisms based on geochemical data for 9 ophiolite-associated groundwaters (several sampled in both field seasons) and 1 regional river water, to determine if selected metabolic reactions are supported in these aqueous environments. Metabolisms (reactions provided in caption to Figure 7) were chosen to mirror those chemosynthetic metabolic strategies known to be utilized at sites of active serpentinization (such as those profiled in Table 1). For reference, Schrenk et al. (2013) group continental and marine serpentinites conceptually as a biome, detailing that low cell counts and a relatively low diversity of microbial types have been documented in a number of field studies in serpentinites. In the Archaea, representatives of the Crenarchaeota (Desulfurococcales) and Euryarchaeota, (Archaeoglobales, Thermococcales, Methanococcales, Methanobacteriales, Methanopyrales, Methanosarcinales, ANME-2, and ANME-1 groups) have been detected. In the Bacteria, representatives of the Aquificae, Bacteroidetes, Betaproteobacteria (including Hydrogenophaga), Gammaproteobacteria, Deltaproteobacteria, Epsilonproteobacteria, Thermodesulfobacteria, Actinobacteria, and Firmicutes have been observed (Schrenk et al., 2013). Accordingly, metabolisms that are tied to methane cycling and hydrogen oxidation are considered here, as well as other reactions that make use of ferric iron, sulfur, and/or N-compounds in the ultramafic environment.

Hydrogen oxidation via the knallgas reaction has the greatest bioenergetic yield in the modeled system. In general, given the nature of this study, environmental samples were collected at the interface between a relatively reducing aqueous system (groundwaters linked to serpentinization in most cases) and a relatively oxidizing system (open to the atmosphere). Although energy yields are not as great for sulfate reduction coupled to hydrogen oxidation, where sulfate is of meaningful concentration (BB1, ML sites, MF1, SI1), this metabolism is also feasible.

Several aspects of methane cycling were treated in this model. Because spring waters and well samples were all collected at the interface between surface and subsurface biogeochemical systems, it is to be expected that CO₂ is dissolving into the system, providing the reactants for methanogenesis, and it is to be expected that organic acids produced abiotically by serpentinization would also be present, possibly fueling acetic acid disproportionation. Both of these metabolisms were evaluated in the model. Methanogenesis is feasible but with the lowest calculated energy yield. Acetic acid fermentation is not feasible, likely due to the environmental abundance of methane and very low expected concentrations of acetate. Methane is abiotically produced in serpentinization, and so would be a common reactant for metabolism: both methane oxidation and methanotrophy coupled with sulfate reduction are feasible based on the model results.

Given the possible mixing of Type I and II waters in these subsurface hydrological flow regimes, there may be aqueous ferrous iron inputs tied to the surface-related water; at the same time, the mineral-hosted iron in serpentinized ultramafics is likely to be ferric, with some ferrous iron locked in nearly insoluble oxides and oxyhydroxides. In the absence of thermodynamic data at elevated temperatures for many solid phases of interest in microbial iron cycling (such as the minerals goethite and ferrihydrite), we relied on the aqueous phase Fe redox transformations, coupled with H₂ oxidation: both iron oxidation and reduction are feasible under the modeled environmental conditions.

Pyrite biodegradation, through the oxidation of either iron or sulfur, is feasible under the modeled conditions also, with relatively high energy yields. Indeed, serpentinization has been considered an important sulfur sink in the seabed (Alt and Shanks, 1998) and pyrite may be pervasive in the subsurface at ML and PB sites. In the case of BB1, we sampled an artesian well plumbing groundwaters interacting with altered pillow basalts and clay-rich sedimentary formations. The adjacent sedimentary unit is composed primarily of smectite-group clays and zeolites (x-ray diffraction data are not shown, but confirm mineral identifications), and pyrite is a reasonable minor component in this system, as a common diagenetic mineral.

Nitrification and ammonia oxidation reactions were also assessed for bioenergetics yield. Nitrification was thermodynamically favored in the modeled environments, while ammonia oxidation was not. Ammonia oxidation, as written here (Figure 7 caption) requires protons as a reactant, and in the high pH environments, the dearth of protons likely impedes this reaction. In fact, the energy yields are near zero for all sites considered. However, the inset box in Figure 7 shows the ammonia oxidation results hover near ΔG_r = 0; note that the Gibbs energy values for this metabolic strategy are slightly negative (thus the metabolism is favored) only for sites BB1 and PB2 in 2013. The swing from ΔG_r > 0 to ΔG_r < 0 between sampling years indicates that the metabolic strategy may be feasible under some field conditions, and does in fact shift from environmentally supported to environmentally inhibited over time, as is shown by our year-to-year shift.

In sum, calculated Gibbs energy of reaction values (calculated at environmental conditions) at all sites indicate that most selected metabolisms are indeed feasible, with Gibbs energy <0, indicating a thermodynamic potential for the reaction to proceed to the right, as written, under the specified conditions (Figure 7). Two exceptions occur: acetic acid disproportionation [i.e., acetic acid_(aq) = CH_4(aq) + CO_2(aq)[ and ammonia oxidation in the presence of nitrite [i.e., NH_3(aq) + NO⁻₂ + H⁺ = N_2(aq) + 2H₂O_(l)] are not shown to be feasible in the geochemical environment. No site supports acetic acid disproportionation (taken as a proxy for fermentation of acetate).

Interestingly, working with environmental samples collected at ML sites in tandem with those used for geochemical analysis and thus linked directly to the model results, Woycheese et al. (2015) report several putative metabolisms (based on nearest neighbor taxonomic affiliations via 16S rRNA gene amplicon sequencing on the Illumina MiSeq platform) that resonate with the geochemical modeling presented here.

Hydrogen oxidation is likely a major metabolism in the ML series springs: 16S data show that there are two hydrogen-oxidizing genera in the environment. Hydrogenophaga comprised 1.3% of sequence reads at the upstream edge of site ML2, 7.1% of reads at the downstream edge of ML2, and 3.4% of reads at ML3. Thiobacillus (family Hydrogenophilace) accounted for 0.02, 4.4, and 9.2% of sequence reads, at ML1–ML3, respectively.

16S data also inform the case for methane cycling in these ophiolite-associated springs. Methanogenesis is likely one of the dominant metabolisms at ML1: the archaeal family Methanobacteriaceae makes up about 7% of sequence reads at the upstream edge of site ML2, and 3.4% of reads at the downstream edge of ML2. At ML3, percentages are much lower (0.06% of reads); suggesting communities shift as the physical-chemical conditions evolve in the running stream water. We noted the methanotrophic bacterial genera Methylobacteriaceae, Methylocystaceae, and Methylobacillus in the rare taxa, comprising less than 0.03% of sequence reads. Acetic acid fermentation is not prominent in the 16S data, but there are traces of the family Acetobacteraceae at the upstream edge of site ML2 (0.01% sequence abundance), downstream edge of site ML2 (0.4% sequence abundance), and ML3 (1.2% sequence abundance). 16S data do not speak to sulfur cycling, and have very little evidence for iron cycling in the ML sites. Here, less than 0.2% of sequence reads aligned with the iron-oxidizing bacterial family Thermodesulfovibrionaceae. Lastly, ammonia-related sequence reads are not significant; the archaeal genus Candidatus Nitrososphaera comprised less than 0.099% of sequence reads for ML sites.

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.