Subduction zone forearc serpentinites as incubators for deep microbial life

The IBM subduction zone is a convergent plate margin ranging over ∼2,800 km from near Tokyo (Japan) to south of Guam (Mariana Islands; Fig. 1). The IBM is located along the eastern margin of the Philippine Sea Plate in the Western Pacific Ocean and formed due to the subduction of the Pacific Plate under the Philippine Sea Plate (11). The southern boundary is marked by the intersection of the IBM trench with the Palau–Kyushu Ridge at 11°N. The northern boundary is at 35°20′N close to southern Honshu, Japan (12). The eastern boundary extends along a deep-sea trench and ranges in depth from 3 km at the Ogasawara Plateau (trench entrance) to ∼11-km depth within the Challenger Deep—the deepest site in the world. The Mariana forearc is pervasively faulted by tectonic activity and only minor sediment accretion occurs along the margin (9, 13). As the Pacific Plate descends, oceanic upper mantle, oceanic crust, overlaying sediment, and water are transported into the forearc mantle. Some of this material is transferred from the subducting slab into the overlying mantle and oceanic plate, where large quantities of fluids rise through faults and fractures, carrying dissolved constituents from the subducting slab. These fluids can either vent as cold springs onto the seafloor (14) or hydrate and serpentinize the mantle wedge. The latter is supported by deep-sea drilling and geophysical measurements showing that at least part of the Mariana forearc mantle wedge is hydrated (15, 16). Within the deep-reaching forearc faults, serpentinite fault gouges mix with the rising slab-derived fluids. This mud–rock mixture buoyantly rises in conduits along fault planes until it extrudes onto the seafloor to form various kilometer-scale seamounts, that is, mud volcanoes, predominantly composed of serpentinite, situated on the outer forearc of the Mariana margin (e.g., ref. 17) (Fig. 1B). The mud volcanoes are located in a trench-parallel zone ∼30–100 km arcward of the trench axis and reach up to 50 km in diameter and over 2 km in height (17, 18). Unconsolidated flows of clay- to silt-sized serpentinite mud enclose up to boulder-sized rock clasts of variably serpentinized mantle peridotite and subordinately blueschist-facies fragments (18). The samples studied here are recovered from drill cores taken from the South Chamorro serpentine mud volcano (13°47′N, 146°00′E; Fig. 1B) drilled during ODP Leg 195 (19). The seamount is a partly collapsed, roughly conical structure ∼2-km high and ∼20-km wide with active serpentine/blueschist mud volcanism. The subducting slab beneath the serpentinite mud volcano is at ∼20-km depth (14, 18).

We studied 46 clasts recovered at depths of 14.80 (1200E-004-02WR-130-140), 28.70 (1200E-007H-02WR-130-140), and 110.07 m below seafloor (mbsf) (1200A-013R-02W-40877). All clasts show a serpentinite mesh texture, where lizardite veins assemble as a mesh framework encompassing central mesh cores (Fig. 2 A and B). Mesh cores are typically composed of chrysotile, or occasionally a mixture of lizardite and chrysotile. In some instances, lizardite fully overgrows the mesh core region (SI Raman Spectroscopy). Time-of-flight secondary ion mass spectrometry (ToF-SIMS) detected aliphatic ion species and ionic units within the mesh cores from a clast retrieved from 14.80 mbsf (Fig. 2 C–E). These ion clusters have previously been used as indicators for the presence of the amino acids methionine, C₅H₁₁NO₂S, and threonine, C₄H₉NO₃ (20), but could arise from other organic molecules with a similar mass. Confocal micro-Raman spectroscopy corroborates the ToF-SIMS findings showing the presence of amide functional groups within the organic material as well as aliphatic and aromatic functional groups (Fig. 2F). Hyperspectral Raman imaging (Methods and SI Raman Spectroscopy) reveals the almost exclusive accumulation of such structural components of organic compounds within the mesh cores (Fig. 2G), in agreement with the ToF-SIMS results (Fig. 2 C–E). Clasts retrieved from greater depth within the ODP Leg 195 core (110.07 mbsf) display nearly identical Raman spectra indicative of organic material (Fig. 2F). However, the additional N(C–C) stretching mode is absent, likely reflecting varying degrees of maturity.

Brucite and magnetite associated with lizardite within the clasts constrains the serpentinization temperature to below 300 °C (21). We also identified nanosized (50–350 nm) awaruite (Ni_2–3Fe) grains embedded within and in the vicinity of the mesh cores (Fig. 3 A–C; SI Opaque Mineral Grain Analysis). The ToF-SIMS measurements further reveal elevated concentrations of various redox active metal ions (Mn, Ni, Cr, Fe; Fig. 3D) encompassing the organic-rich mesh cores shown in Fig. 2. Nanotomography, acquired by focused ion beam scanning electron microscopy (FIB-SEM), shows that the mesh cores are areas of elevated porosity (Fig. 4A). Based on the segmented nanotomography volume shown in Fig. 4A, the visible median pore size is ∼25 nm. However, considerably larger pore structures, exceeding several micrometers, can be found in other mesh cores (Fig. 3A). Pore formation is likely the result of several factors such as imperfect packing of randomly orientated serpentine grains as typically observed in serpentinite mesh cores (22), chrysotile nanopores (22, 23), prolonged exposure to upwelling subduction zone fluids inducing mineral dissolution, and/or decompression cracking during upward migration along faults or within the mud volcano conduit. Mesh rims are typically characterized by fewer, micrometer-sized pores (Fig. 4 B and C) as well as mesopores (i.e., 2–50 nm) in the vicinity of mesh cores and around awaruite grain clusters (Fig. 3C). The anisotropic pore alignment (Fig. 4A) and complex pore contrast (Fig. 4B) indicate that the mesh cores contain a partially connected pore network. The intricate relationship between the serpentine and organic matter (Fig. 2), however, suggests that the detected pore sizes are substantially underestimated as the organic material within the pores obscures the true pore size that is likely orders of magnitude larger.

Although we cannot unambiguously assign the source of the organic matter, several steps can be taken toward better constraining its potential origin. Upon contact with seawater, the high pH (∼12.6) pore fluids with high alkalinity (100 mmol/kg) within the muds found at the South Chamorro mud volcano should trigger immediate carbonate precipitation (24). The absence of any carbonate within the clast-bearing mud layers from 28.70 and 110.07 mbsf thus suggests that, although the clasts were erupted onto the seafloor, they were never exposed to seawater and thus seawater-derived organic matter. Additionally, organic matter was only identified in the internal regions of the serpentinite clasts after cross-sectional opening of the clasts (Fig. 2A). Therefore, it seems unlikely that the organic matter was introduced during shipboard sampling, storage, or sample preparation (Methods). Other potential sources would be organic matter derived from subducted pelagic sediments or in situ production within the forearc system via abiotic or biological processes. Isotope studies suggest influx of fluids derived from a mixture of subducted pelagic sediments and altered oceanic crust into the forearc mantle facilitating serpentinization (25). However, their sources remain controversial (26–28), and the nature of the organic material that is potentially carried with these fluids is largely unknown. Abiotic routes have been demonstrated to produce hydrocarbons and amino acids under hydrothermal conditions akin to serpentinization (29), but it is unclear whether these processes can produce the organic complexity observed in the serpentinite clasts. Intriguingly, similar Raman spectra to those reported here have been documented in other serpentinites and correlate with spectra taken from bacteria (6) and the presence of bacterial lipid biomarkers (7). Moreover, the acquired Raman spectra (Fig. 2F) show bands in similar positions to those attributed to functional modes typically found in bacteria-derived biopolymers, including proteins, lipids, and nucleic acids (30, 31), indicating that the organics may have a biological source.

Raman spectra of organic matter were collected using a Kaiser HoloLab Series 5000 equipped with a diode-pumped solid-state laser (785 nm) (Utrecht University). Hyperspectral mapping was performed with a Horiba Scientific LabRam HR800 (Steinmann Institute, University of Bonn). Raman scattering was excited with a diode-pumped solid-state laser (784 nm) with less than 40 mW at the sample surface. A 100× objective with a numerical aperture of 0.9 was used resulting in a diffraction-limited lateral resolution of ∼1 µm. The confocal hole was set to 1,000 μm, resulting in a depth resolution of a few micrometers. The scattered Raman light was collected in a 180° backscattering geometry by an electron-multiplier charged coupled device detector after passing through a 100-μm (single measurements) or 200-µm (mapping) spectrometer entrance slit and being dispersed by a grating of 600 grooves per millimeter, yielding a spectral resolution of 1.7 and 2.3 cm⁻¹, respectively, as determined from the width of Ne lines. The spectrometer was calibrated with the first-order Si Raman band at 520.7 cm⁻¹ and Ne lines. The total acquisition time varied between 900 and 1,200 s for single measurements and 25 × 0.5 s for mapping. Normalization of Raman spectrum to the most intense epoxy band at 815 cm⁻¹ shows no contribution of bands from the epoxy resin (Fig. S1). This is evident by the absence of epoxy bands (arrow in Fig. S1) between 950 and 1,000 cm⁻¹ in the sample spectra. The false-color Raman image shown in Fig. 2G is generated from the spectra recorded for each pixel of the image by color-coding the ratio between the integrated intensity of an organic vibrational modes near 639 cm⁻¹ and a serpentine mode at 690 cm⁻¹ (Fig. S2). Warm colors reflect a high content of organic material within the analyzed volume, whereas blue colors mark regions with low or undetectable organic material. Spectra were collected over an area of 60 × 60 µm² with a step size of 0.5 µm, resulting in 7,200 spectra. The resulting map is shown in Fig. S4. To obtain reduced or R(ω) intensities for future comparison that are independent from the instrument, the laboratory temperature, and the excitation wavenumber, ν_e, and thus directly proportional to relative scattering activity, the Raman spectra from the mesh core were also corrected for (i) the instrument response function (white light correction); (ii) the excitation frequency dependence, that is, by the scattering factor, (ν_e − ν)³, with ν being the wavenumber of the scattered light (intensities were measured in photons per second); and (iii) the temperature effect, that is, by the Bose–Einstein temperature factor, 1 − exp(−hνc/kT), with h, c, k, and T being the Planck constant, the speed of light, the Boltzmann constant, and the temperature, respectively. The corrected spectra were deconvoluted by least-squares fitting Gauss–Lorentz functions along with a linear background that was subtracted after the least-squares fit (Fig. S5 and Table S1). The error of the reported band positions is smaller than 0.1 cm⁻¹. Raman evaluation of the serpentine polymorphs present in the rim and core structures was conducted by examining the water vibration region at ∼3,600 cm⁻¹. Raman spectra taken, using the 785-nm laser, showed no evidence for the OH bands due to the low scattering efficiency of these bands with this laser. Therefore, the 532-nm coherent compass sapphire laser of a WITec alpha 300R was used for the OH band analysis of serpentine group minerals. The analysis was conducted with a 50× long working distance lens with a 0.55 numerical aperture in backscattering geometry to the sample. After the scattered Raman light passed through a pinhole of 20 µm, the light was dispersed on a grating of 1,800 grooves per millimeter, resulting in a spectral resolution of 1.1 cm⁻¹ in the spectral region of interest, measured using a built-in calibration light source.

Fig. S6 shows the distribution of opaque mineral grains (high backscattering intensity in Fig. S6A) within the serpentinites clasts. Several FIB-SEM sections were prepared across the mesh rim–core interfaces, where Fig. S6B shows a representative high angle annular dark field (HAADF) image highlighting three nanoparticles with high intensity (overview image in main-text Fig. 3C). The intensity in an HAADF image scales with atomic number, implying that the nanoparticles are atomically heavier than the surrounding serpentine grains. Energy-dispersive X-ray analysis (Fig. S6C) reveals that the nanoparticle exist exclusively of Ni and Fe in a ∼2:1 ratio. Together with chemical analysis conducted on the micrometer-scale in a scanning electron microscope, the microstructural association of the particles within the serpentinites, and the limited possibilities of potential mineral phases with Ni and Fe in a ∼2:1 ratio (65), we conclude that the nanoparticles are awaruite (Ni₂Fe–Ni₃Fe).

To test whether microbial life is a feasible source of the organic matter observed, we need to establish an estimate for the depth limit for life within the Mariana forearc. As microbial organisms can survive temperatures as high as 122 °C (32) and pressures into the gigapascal range (33), we estimated the potential depth limit for microbial life in this region using a one-dimensional steady-state heat conduction model (34) (SI Estimation of the Maximum Depth for the Current Temperature Limit for Life). In this calculation, heat is transferred in one direction without consideration of minor advective heat flow through the ascending mud or heat generated through the exothermic serpentinization reaction. Both of these processes are expected to only play a minor role. Particularly heat through serpentinization has been shown to be nearly negligible (35). Assuming no variations in temperature or heat flow, the basic equation of conductive heat transfer theory is a statement of conservation of energy and can be written as follows:where T and T₀ are the temperature at depth and at the ocean floor, y is the depth in meters below seafloor, q₀ is the surface heat flow (0.03 W⋅m⁻²), k is the thermal rock conductivity (2.9 W⋅m⁻¹⋅K⁻¹), ρ is the density (2,900 kg⋅m⁻³, partially serpentinized peridotite), and H is the current mean mantle heat generation rate due to radioactive decay (7.42 × 10⁻¹² W⋅kg⁻¹) (34). At relatively shallow forearc depths, H will only play a minor role. Rearranging the equation above to solve for y at a given temperature (i.e., the known temperature limit for life at 122 °C) results in the following:where y₊ is a nonphysical solution and thus disregarded. Surface heat flow values of the Mariana forearc were taken from measurements acquired during the Deep Sea Drilling Project Leg 60 (36). Thermal conductivity values of serpentinites are based on measurements taken during ODP 209 (37) and an average value of 2.9 W⋅m⁻¹⋅K⁻¹ is used. Using these values, the maximum depth for the 122 °C isotherm varies between ∼8,000 and 15,000 mbsf (Fig. 5C). These depth estimates are based on surface heat flow values of 0.03–0.04 W⋅m⁻² that agree with the observed depression of isotherms in most forearc mantle wedges, even those of relatively hot origin such as the Cascadia subduction zone (38, 39). Moreover, our thermal calculations are in agreement with more complex geodynamic models (40, 41), confirming that the 122 °C isotherm is reached at ∼10,000 mbsf in forearcs. Hence, current serpentinization-fueled microbial life within subduction zone forearcs could be supported down to these depths and corresponding pressure (∼0.34 GPa; Fig. 5). In contrast, habitable zones in the vicinity of oceanic spreading centers are limited to the first hundred meters to few kilometers below the seafloor. The exact location of the 122 °C isotherm will likely vary in depth at and around the spreading center as a result of the ridge architecture, heat flux, and hydrothermal circulation (42, 43). The downhole temperature within the International Ocean Drilling Program (IODP) Hole U1309D at Atlantis Massif oceanic core complex (Mid-Atlantic Ridge) places the 122 °C upper temperature limit for microbial life at ∼1,000 mbsf (44). This is one order of magnitude less than compared with our estimated limit of life in the Mariana subduction zone forearc. As the serpentinite mud originates directly from the forearc mantle wedge (>20-km depth), the model indicates the potential for a biosphere deep within the forearc. This makes the Mariana serpentinite clasts a natural laboratory of prime interest when searching for habitable zones of life deep within the lithosphere.

A sketch of the one-dimensional steady-state heat conduction model to estimate the maximum depth for the current temperature limit for life in subduction zone forearcs is found in Fig. S7.

To sustain deep microbial life within a solid rock framework requires energy resources that can either migrate or be produced close to areas suitable for colonization. There is little to no chemical benefit for microbes to interact directly with serpentine minerals; thus, other life-supporting energy-generating pathways need to be present. Microgenomic studies show evidence for microbial H₂ and CH₄ utilization in serpentinizing systems with known microbial colonization. Indeed, several studies (e.g., ref. 5) indicate that Archaea found up to 20 mbsf within the South Chamorro mud volcano are fueled by deeply derived CH₄-enriched fluids. Experiments suggest that H₂ produced during low-temperature serpentinization (<200 °C) (45) could react with a carbon source to form CH₄ on catalytic mineral surfaces. Investigations of naturally occurring abiogenic CH₄ indicate that abiotic hydrocarbon synthesis can potentially take place at temperatures as low as ∼120 °C (46). However, there are contrasting experimental results concerning the formation and synthesis kinetics of CH₄ production at (very) low temperatures (47–49). In hydrothermal experiments, awaruite has been identified as a possible CH₄ production catalyst (50). The awaruite grains observed here are nanosized (Fig. 3C) and, therefore, have a high surface area-to-volume ratio, which should enhance their catalytic activity (51). Thus, nanosized alloys could have facilitated CH₄ production over geologically relevant timescales below the upper temperature limit for life (122 °C). In near-surface serpentinizing systems, hydrothermal fluids can mix with, for example, seawater, resulting in disequilibria that may provide the energy and substrates needed to support chemolithoautotrophic life (7). In contrast, a slowly ascending serpentinite mud along deep-reaching forearc faults may allow the system to remain much closer to equilibrium, regulating the activities of H₂, CH₄, CO₂, and the Fe(II)/Fe(III) ratio in the solids and fluids limiting energy sources. Nielsen et al. (25), however, documented the occurrence of rodingite within the Mariana serpentinite mud volcanoes derived from in situ alteration within the forearc mantle. These rocks suggest hydrothermal interactions between mafic and ultramafic units and together with simultaneous serpentinization may be the source of fluids that can produce disequilibrium environments.

An additional source of externally derived fluids could come directly from the subducting slab and would thus be in disequilibrium with the overriding forearc wedge. These fluids are most likely different in composition compared with serpentinizing systems at midocean ridges or passive margins. Kelemen and Manning (52) recently reevaluated the global carbon flux through subduction zones and estimated that several million tons of carbon per year could be released from subducting slabs into the overriding forearc wedge. Therefore, these fluids could provide the carbon source needed for abiotic hydrocarbon synthesis or even directly contribute organic molecules to the forearc and thus the serpentinite clasts. If the organic matter reported here was sourced from the subduction slab alone, we would expect the same level of maturation and thus evidence for the same functional groups. However, we observe the absence of specific Raman bands [N(C–C)] in clasts from different depths (Fig. 2F). Various studies have shown that subduction zone dehydration reactions across a range of temperatures (200–600 °C) and pressures release nitrogen that is able to enter the overriding forearc directly or travel with the expelled fluids upward through the subduction zone channel (53–55). Moreover, recent thermodynamic calculations of the nitrogen speciation in aqueous fluids under upper mantle conditions suggest that the oxidized mantle wedge of subduction zones favors nitrogen over ammonium (NH₄⁺), promoting outgassing rather than mineral trapping of nitrogen (56). Thus, the actions of mud volcanoes as conduits for slab-derived fluids may provide missing resources that would otherwise be limiting factors for life in the forearc mantle or contribute nitrogen to the abiogenic synthesis of more complex organic compounds. To rigorously evaluate the energy sources for microbial life within the subduction zone forearc, better constraints are needed for the fluid influxes from the subduction zone and how these slab-derived fluids interact with the forearc. Increasing sophisticated fluid speciation models coupled to fluid–rock interaction simulations and experiments at high pressures and temperatures (56, 57) will provide critical insights into this problem.

Although the origin of the organic matter cannot be unequivocally identified, we suggest, based on the similarities with molecular signatures of bacteria-derived biopolymers, that the organic matter may represent remnants of microbial life within or even below the mud volcanoes. Our simple model supports this hypothesis, showing that the temperature window for life could extend deep into the forearc. Hence, the identification of complex organic matter recovered from depths of up to 110.07 mbsf may be evidence for life in an oceanic serpentinite-hosted rock formation from depths likely far exceeding the drill core depth, where serpentinite-supported life has not yet been documented. Thus far, evidence for microbial communities within the Mariana mud volcanoes has only been indirectly detected in fluid samples no deeper than 20 mbsf (5, 24). There are a variety of examples indicating that microbial life can colonize shallow serpentinization-fueled environments and use abiogenically produced H₂ and CH₄ (1, 3–5), but microbial life within the deep subsurface, for example, deep within the Mariana forearc, with no connection to the Earth’s surface, may have little resemblance with presently known serpentinization-fueled ecosystems.

In any case, if life is present in the subduction zone forearc, it has far-reaching implications as recent studies suggest that environments resembling those both within and below the Mariana serpentinite mud volcanoes were already present on the early Earth (58, 59). Thus, even if modern-type subduction was not fully established in the Hadean and Archean, Mariana forearc-like deep subsurface environments may have allowed early forms of life to thrive, despite violent phases such as the so-called Late Heavy Bombardment, a period of intensive meteorite bombardment around 3.9 Ga (60). Even if only a small amount of the global forearc mantle hosts microbial life, fluctuations in the total subduction zone length (61) could have significant consequences for the deep carbon budget. During these fluctuations, fluid flow through subduction zone forearc regions, visible in the form of serpentinite mud volcanism, are a crucial connection between the deep biosphere and surface world, influencing geochemical fluxes throughout Earth’s history. Only if we keep exploring the windows into the deep subsurface, such as the serpentinite clasts presented here, will we be able to establish a full budget of Earth’s deep carbon and the potential for a subsurface biosphere.

We thank T. Ludwig for discussion. O.P. was supported by Netherlands Organization for Scientific Research Veni Grant 863.13.006, H.E.K. by the European Union Fellowship PIOF-GA-2012-328731, T.Z. by the German Science Foundation (ZA285/5), and Y.L. by the Utrecht University Sustainability Program. For I.P.S., Ocean Drilling Program (ODP) Leg 195 participation and post-cruise research were funded by Joint Oceanographic Institutes/US Science Advisory Committee (Schlanger Fellowship) and the UK International Ocean Discovery Program (IODP) Program via the Natural Environment Research Council (NE/M007782/1).