Exploring deep microbial life in coal-bearing sediment down to ~2.5 km below the ocean floor

See allHide authors and affiliations

Science  24 Jul 2015:
Vol. 349, Issue 6246, pp. 420-424
DOI: 10.1126/science.aaa6882

A deep sleep in coal beds

Deep below the ocean floor, microorganisms from forest soils continue to thrive. Inagaki et al. analyzed the microbial communities in several drill cores off the coast of Japan, some sampling more than 2 km below the seafloor (see the Perspective by Huber). Although cell counts decreased with depth, deep coal beds harbored active communities of methanogenic bacteria. These communities were more similar to those found in forest soils than in other deep marine sediments.

Science, this issue p. 420; see also p. 376


Microbial life inhabits deeply buried marine sediments, but the extent of this vast ecosystem remains poorly constrained. Here we provide evidence for the existence of microbial communities in ~40° to 60°C sediment associated with lignite coal beds at ~1.5 to 2.5 km below the seafloor in the Pacific Ocean off Japan. Microbial methanogenesis was indicated by the isotopic compositions of methane and carbon dioxide, biomarkers, cultivation data, and gas compositions. Concentrations of indigenous microbial cells below 1.5 km ranged from <10 to ~104 cells cm−3. Peak concentrations occurred in lignite layers, where communities differed markedly from shallower subseafloor communities and instead resembled organotrophic communities in forest soils. This suggests that terrigenous sediments retain indigenous community members tens of millions of years after burial in the seabed.

Microbial life has been found in marine sediments as deep as 1922 m below the seafloor (mbsf) (1), with cell numbers decreasing logarithmically as burial depth increases (2). However, the extent of the deep subseafloor biosphere and the factors limiting or stimulating life at its lower boundaries remain largely unknown, partly because of the technological challenges associated with obtaining pristine scientific samples from great depths. The deepest microbial signatures measured thus far come from the biodegradation of heavy oils in fossil fuel reservoirs down to subsurface depths of ~4 km (3, 4).

We studied the microbial communities associated with deep subseafloor coal beds, which are widespread along the Western Pacific continental margins (5). During Integrated Ocean Drilling Program (IODP) Expedition 337 in 2012, we drilled and recovered sediments down to 2466 mbsf using the riser-drilling vessel Chikyu at Site C0020, located along the northwestern Pacific margin (41°10.5983′N, 142°12.0328′E, 1180 m water depth) (fig. S1) (6). The top 365 mbsf of the borehole had been drilled in 2006 during the Chikyu cruise CK06-06 (6, 7). The site is located in the forearc basin offshore from the Shimokita Peninsula, Japan, where terrigenous sediments rich in organic matter were deposited in coastal environments during the late Paleogene to early Neogene periods, before the depositional environment became marine in the course of subsidence (5). The series of buried lignite coal beds at depths of ~1.5 to 2.5 km are 0.3 to 7.3 m thick (6). Because of the relatively low geothermal temperature gradient of 24°C km−1 (6), the deepest horizons, with in situ temperatures of <60°C, are well within the thermal limits of microbial life.

We detected intact microbial cells throughout the entire drilled sediment column, down to the deepest sample at 2458 mbsf. Cell concentrations decreased with depth but in an unexpected fashion. In the “shallow” subseafloor (above 365 mbsf), concentrations decreased steadily from ~109 to ~107 cells cm−3 (8) and exceeded predictions based on the global regression line from previous surveys of cell concentrations at ocean margins (2) (Fig. 1A and table S1). In contrast, raw cellular concentrations were substantially lower in the “deep” subseafloor (below ~1.5 km), ranging from ~102 to 103 cells cm−3, but they exceeded this range in coal-bearing horizons and adjacent sediment layers (Fig. 1A).

Fig. 1 Depth profiles of microbial cell counts and geochemical data at Site C0020.

(A) Microbial cell concentrations, (B) δ13C and δD of methane, and (C) C1/C2 ratios and δ13C of CO2. (A) For cell concentrations in the deep subseafloor, raw data of fluorescence image-based cell counts (8), the most likely indigenous cell concentrations based on the probability-relationship set analysis (7), and the most conservative indigenous cell concentrations estimated based on the taxonomic classification (7) are shown (figs. S6 and S7 and table S1). The correction factor is the proportion of sequences estimated to be indigenous (table S1). The minimal quantification limit for raw cell counts was 1.43 × 102 cells cm−3 (i.e., the upper 95% confidence interval of the negative background). All δ13C and δD in (B) and (C) are in per mil versus the Vienna Pee Dee Belemnite (VPDB) and Standard Mean Ocean Water (SMOW) standards, respectively. The Δ13CH3D-T values in (B) designate the apparent equilibrium temperatures derived from measurements of 13CH3D, a clumped isotopologue of methane (table S2) in discrete formation fluid (FF) samples (6, 7). Temperature is based on the temperature gradient of 24°C km−1 determined by downhole logging (6).

The low cell concentrations required the implementation of rigorous contamination tests to characterize indigenous microbial communities in the deep coal-bed biosphere. To minimize and quantify contamination potentially introduced during drilling, we (i) investigated whole round core sections by x-ray computed tomography before selecting sediment samples, (ii) took microbiological samples only from the center part of undisturbed core intervals, and (iii) quantified the intrusion of drill fluids into the cores by perfluorocarbon tracer assays (6, 7). As additional validation, we sequenced the V1 to V3 region of 16S ribosomal RNA genes in all sediment samples (7) obtained by riser drilling. We compared these to control samples consisting of drill mud and lab experimental blanks to differentiate indigenous microbial communities from contaminant cells (fig. S2) (7). We applied a probabilistic approach, incorporating taxon variability across samples to identify the likelihood that each taxon would be consistently sampled, either exclusively from the control or sediment sample sets or from both sample sets (7). In this way, we identified likely indigenous taxa, which were either (i) exclusive to sediment samples (“most conservative”) (figs. S2 to S4 and table S1) or (ii) consistently found in sediment samples in significant abundance and only occasionally found in contamination controls in low abundance (“most likely”) (figs. S5 and S6 and table S1). From this approach, we derived correction factors (table S1) to the raw cell concentrations to estimate the corresponding population sizes. We used the results of taxonomy-based sequence filtration and the probability-based relationship analysis for 16S sequences to estimate the “most conservative” and “most likely” indigenous cell concentrations, respectively (Fig. 1A, figs. S5 to S7, and table S1) (7). These estimated cell concentrations are much lower than predicted by the slope of the global regression line (fig. S7) (2) and are even lower than previously reported values from the most oligotrophic subseafloor setting on Earth, the South Pacific Gyre (9).

Despite the very low cell numbers, geochemical data indicated microbial activity even in the deepest horizons sampled. Carbon isotopic compositions of methane (δ13C-CH4) and ratios of methane over ethane (C1/C2) (Fig. 1, B and C), both continuously monitored in circulating mud gas during riser drilling (6), point to microbial methanogenesis as the predominant source of methane (10) throughout the entire drilled sequence. Positive inflections of C1/C2 ratios between 1700 and 2000 mbsf suggest that biological methanogenesis is stimulated in coal-bearing horizons (Fig. 1C), where contamination-corrected cell concentrations reached ~102 to 104 cells cm−3 (Fig. 1A), as well as in the overlying 200 m of sediment. Hydrogen isotopic compositions of methane (δD-CH4) ranged from –200 to –150 per mil (‰) (Fig. 1B), consistent with its production by hydrogenotrophic carbon dioxide (CO2) reduction (10). Locally increased δ13C-CO2 values in coal-bearing horizons provide further evidence that the CO2 pool was isotopically fractionated by microbial methanogenesis (Fig. 1C). Moreover, in situ production of methane is indicated by the abundance of 13CH3D (a rare “clumped” isotopologue of methane with two heavy isotopes) in formation fluids sampled in two discrete coal-bed horizons (Fig. 1B and table S2) (6, 7). These analyses returned low Δ13CH3D-based temperatures of Embedded Image and Embedded Image (table S2), providing evidence against major contributions of more deeply sourced thermogenic methane, which would be expected to carry clumped-isotope temperatures >150°C (11). In addition, we detected coenzyme F430 in core samples, providing biomarker evidence for methanogenesis in coal beds ~2 km below the seafloor (Fig. 2A and table S3) (12, 13). Coenzyme F430 is a key prosthetic group of methyl-coenzyme M reductase that catalyzes the last step of methanogenesis. Its concentrations in deep sediments were about two orders of magnitude lower than in shallow sediments (table S3), suggesting the presence of a small but persistent community of methanogens in deep coal-bearing layers.

Fig. 2 Geochemical and microbiological indications for methanogenic microbial communities in ~2-km-deep subseafloor coal beds at Site C0020.

(A) A representative chromatogram of the diagnostic methanogen biomarker F430 (as methyl ester) and its epimers from a coal sample (core 18R-2, 1946 mbsf) (table S3). (B) to (D) Photomicrographs of microbial cells in an enrichment culture from ~2-km-deep coal-bed samples using a continuous-flow bioreactor with powdered coal as the major energy source (7). (B) Phase-contrast micrograph of microbial cells attached to mineral particles. (C) Fluorescent micrograph of the same field shown in (B) showing the growth of methanogens that produce autofluorescence derived from coenzyme F420. (D) Phase-contrast micrograph showing spherical spore-like particles, indicated by arrows. (E) to (G) NanoSIMS analysis of cells in the reactor enrichment culture incubated with 13C-labeled bicarbonate (7). (E) Fluorescent micrograph of SYBR Green I–stained cells. (F) and (G) NanoSIMS ion images of 13C/12C (F) and 12C (G). The color gradient indicates the relative abundance of 13C expressed as 13C/12C. The length of the bars is 10 μm.

In a continuous-flow bioreactor (14) at near in situ temperature (40°C), we enriched methanogenic communities from ~2-km-deep coal-bed samples amended with powdered coal as the major energy source (7) (Fig. 2, B to G). Analysis of methyl-coenzyme M reductase gene (mcrA) sequences in this enrichment culture indicated the growth of hydrogenotrophic methanogens closely related to Methanobacterium subterraneum and M. formicicum (fig. S8). These species have previously been detected in terrestrial coal beds (15) and in shallower, methane hydrate–bearing sediments at Site C0020 (14). Using nanoscale secondary ion mass spectrometry (NanoSIMS) (7, 16), we detected the incorporation of 13C-labeled bicarbonate into cellular biomass (Fig. 2, E to G). Collectively, these microbiological and geochemical findings indicate that microbial communities are stimulated within coal-bed environments and that hydrogenotrophic methanogens act as terminal remineralizers.

Despite these multiple lines of evidence for the activity of methanogenic archaea, we could only amplify one archaeal 16S sequence related to Methanosarcina barkeri (from a ~2-km-deep coal sample) and one mcrA sequence related to Methanococcus maripaludis (from ~1-km-deep cuttings samples) (fig. S8) (7). Archaeal 16S genes were neither quantifiable by digital polymerase chain reaction (dPCR) (17) nor stably amplifiable using multiple primer sets, indicating that this deep subseafloor microbial ecosystem harbors substantially lower proportions of archaea than shallower sediments at this and other ocean margin sites (18). The difficulty in detecting methanogenic archaea using multiple molecular assays is not unexpected, given their generally low relative abundance of <1%, even in methane-laden subseafloor sediments (19).

The taxonomic distribution based on 16S sequences of the most conservative indigenous community members showed a marked difference between bacterial communities in deeper layers (1279.1 to 2458.8 mbsf) and those in shallower layers (9.5 to 364.0 mbsf) (Fig. 3, figs. S3 and S4, and table S4). For example, in deeper layers, sequences affiliated with the phyla Chloroflexi or “Atribacteria” [candidate division JS1 (20, 21)], which are both globally abundant groups in subseafloor sediments at ocean margins (22, 23), were detected in lower proportions than in shallower layers (Fig. 3A, fig. S3, and table S4). The indigenous sequence assemblage in deeper layers was dominated instead by bacterial groups known to dominate a wide range of terrestrial soil environments (Actinobacteria, Proteobacteria, Firmicutes, Bacteroidetes, and Acidobacteria) (24). In fact, many operational taxonomic units (OTUs) from deeper layers showed high sequence similarity to those from forest soils (figs. S3 and S4). Deeper layers also included microbial phyla that were not or were only barely detectable in shallower layers (such as Gemmatimonadetes and Synergistetes) (fig. S9) (7). Clustering and Bray-Curtis dissimilarity analyses (Fig. 3B and fig. S10), as well as multidimensional scaling and permutational analysis of variance based on genus-level classification (fig. S11), further support the notion that deep communities in coal-associated sedimentary environments on the former forearc basin off Shimokita differ compositionally from those found in shallow diatom-rich marine sediments.

Fig. 3 Taxonomic distribution of the most conservative indigenous bacterial communities in sediments at Site C0020.

(A) Phylum-level taxonomic composition of bacterial 16S gene-tagged sequences (i.e., V1–V3 region) in shallow (sample numbers 1 to 5, 9.5 to 364 mbsf, from Chikyu cruise CK06-06) and deep (sample numbers 6 to 32, 1279.1 to 2458.8 mbsf, from IODP Expedition 337) subseafloor sediment samples (7). (B) Cluster and Bray-Curtis dissimilarity analyses of bacterial community structure based on the genus-level classification of the same sequence assemblages used in (A). Colored dots on the cluster tree represent the sedimentological characteristics of each sample horizon (6). The OTU-based analysis for the most likely indigenous bacterial communities in deep subseafloor sediment samples is shown in fig. S10.

Our combined microbiological and geochemical data set provides an opportunity to examine the potential energy sources of life at depths greater than ~1.5 km below the seafloor. The concentration of dissolved hydrogen (H2), a key intermediate in the anaerobic degradation of organic matter, is an important gauge of the bioenergetic status of anaerobic microbial ecosystems (25, 26). At Site C0020, high H2 concentrations of ~1 to ~500 μM in sediments below 1.5 km (fig. S12 and table S5) resulted in Gibbs free energy yields of hydrogenotrophic methanogenesis that were much more negative than those previously documented from energy-rich surface environments (fig. S13 and tables S6 and S7) (25). These high H2 concentrations suggest very low H2 turnover rates, possibly due to low concentrations of microorganisms with low viabilities and consequently low cell-specific energy turnover. Under these circumstances, the coupling between substrate production and substrate uptake may be severely delayed, resulting in long residence times and the accumulation of H2 to high concentrations in sediment porewater. These substrates are probably generated from lignite (15), microbial necromass (27), and/or adsorbed organic matter. In addition to enzymatic hydrolysis, chemical degradation of recalcitrant organic matter (28, 29) in situ and in deeper, warmer strata bearing Cretaceous coal beds (30) may supply monomeric substrates and contribute to the accumulation of H2. Despite the high H2 concentrations, a range of organotrophic reactions involving the breakdown of lignin phenols, cellular building blocks, and anaerobic degradation intermediates are likely to be thermodynamically favorable and thus could explain the presence of alive and active microbial populations (fig. S13).

In addition to energy availability from organotrophic and hydrogenotrophic reactions, an important factor controlling the viability and size of microbial communities buried deeper than 1.5 km could be the increase in energy expended on the repair of biomolecules as a function of depth. The rates of abiotic amino acid racemization and DNA depurination increase exponentially with temperature (31), and substantial increases in their modeled rates coincided with a dramatic drop in cell numbers at Site C0020 (fig. S14). The relatively increased energetic cost of biomolecule repair could result in a higher cell-specific energy demand in the deeper layers at Site C0020 and may explain why microbial abundance was only a small fraction of the size predicted by the global regression line (Fig. 1A and figs. S7 and S14).

Our findings provide a comprehensive view of the deep subseafloor biosphere associated with coal beds. Despite energetic challenges, this environment appears to have maintained some of the taxonomic groups that populated the original shallow depositional setting and have contributed to carbon cycling ever since.

Supplementary Materials

Materials and Methods

Figs. S1 to S14

Tables S1 to S7

References (3283)


  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: The authors are grateful to IODP and the Ministry of Education, Culture, Sports, Science and Technology of Japan for providing an opportunity to explore the deep coal-bed biosphere off Shimokita during Expedition 337. We thank all crews, drilling team members, lab technicians, and scientists on the drilling vessel Chikyu for supporting core sampling and onboard measurements during Chikyu shakedown cruise CK06-06 and IODP Expedition 337. The authors thank J. A. McKenzie and K. H. Nealson for useful discussions during project design; S. Fukunaga, S. Hashimoto, A. Imajo, Y. Saito, S. Tanaka, K. Uematsu, and N. Xiao for assistance in microbiological analyses; and D. Gruen for technical assistance during clumped isotope analysis. This work was supported in part by the Japan Society for the Promotion of Science (JSPS) Strategic Fund for Strengthening Leading-Edge Research and Development (to F.I. and JAMSTEC), the JSPS Funding Program for Next Generation World-Leading Researchers (grant no. GR102 to F.I.), the JSPS Grants-in-Aid for Science Research (no. 26251041 to F.I., no. 24770033 to T.H., no. 24687011 to H.I., no. 26287142 to M.I., no. 25610166 to M.K., and nos. 24651018 and 24687004 to Y.M.), the European Research Council (Advanced Grant no. 247153 to K.-U.H.), the Deutsche Forschungsgemeinschaft trough project HI 616/16 (to K.-U.H.) through MARUM–Cluster of Excellence 309, and NSF (no. EAR-1250394 to S.O). All shipboard and shore-based data presented in this manuscript are archived and publicly available online in the IODP Expedition 337 Proceedings (6) through the J-CORES database ( and the PANGAEA database ( This is a contribution to the Deep Carbon Observatory.
View Abstract

Stay Connected to Science

Navigate This Article