A Contemporary Microbially Maintained Subglacial Ferrous "Ocean"

See allHide authors and affiliations

Science  17 Apr 2009:
Vol. 324, Issue 5925, pp. 397-400
DOI: 10.1126/science.1167350


An active microbial assemblage cycles sulfur in a sulfate-rich, ancient marine brine beneath Taylor Glacier, an outlet glacier of the East Antarctic Ice Sheet, with Fe(III) serving as the terminal electron acceptor. Isotopic measurements of sulfate, water, carbonate, and ferrous iron and functional gene analyses of adenosine 5′-phosphosulfate reductase imply that a microbial consortium facilitates a catalytic sulfur cycle. These metabolic pathways result from a limited organic carbon supply because of the absence of contemporary photosynthesis, yielding a subglacial ferrous brine that is anoxic but not sulfidic. Coupled biogeochemical processes below the glacier enable subglacial microbes to grow in extended isolation, demonstrating how analogous organic-starved systems, such as Neoproterozoic oceans, accumulated Fe(II) despite the presence of an active sulfur cycle.

Subglacial environments represent a largely unexplored component of Earth's biosphere (1). In the McMurdo Dry Valleys, Antarctica, an iron-rich subglacial outflow (Blood Falls) flows from the Taylor Glacier (Fig. 1A), providing unique access to a subglacial ecosystem. The likely fluid source to Blood Falls is a pool of marine brine of unknown depth trapped underneath the glacier ∼4 km from the glacier snout where the overlying ice is ∼400 m thick (2). Pliocene surface uplift of the Taylor Valley floor, and the associated recession of the Ross Sea Embayment, isolated this pocket of brine (3). Before isolation from direct contact with the atmosphere, the brine was cryoconcentrated (4, 5), resulting in hypersalinity (∼1375 mM Cl). This brine has been isolated for at least 1.5 million years (My), when the Taylor Glacier last advanced over the area (6). Although the brine at present is anoxic and highly ferrous and the pH is circumneutral (Table 1), activity and DNA sequence data reveal that it supports a metabolically active, largely marine microbial assemblage (7).

Fig. 1.

(A) Blood Falls at the snout of the Taylor Glacier (77°72′S 162°27′E). (Inset) Conceptual model for the possible modes of redox cycling of iron, sulfur, and organic material in the Blood Falls brine based on data from this study. Red arrows indicate assimilatory pathways (via the cell), and blue arrows indicate an alternate pathway via catalytic sulfur cycling mediated by APS reductase. (B) Values of δ18OSO4 relative to δH218O, given different reaction scenarios. Sulfate and water will have equilibrated by no more than 1‰ within 50 My, therefore no net change would be expected in the δ18OSO4 of a seawater brine mixed with 18O-depleted glacier water (gray diamond). Where SO42– reduction occurs, SO32– equilibrates with the in situ water, ϵ = 25 to 30.5‰, resulting in δ18OSO4 predicted by the shaded gray line (15, 24). In marine sediments (32) (black diamond), complete equilibration of SO32– with in situ water is observed, indicating quantitative reduction of SO42–. If SO42– reduction occurred to completion in Blood Falls, values of δ18OSO4 between –8 and –10‰ (25 to 30.5‰ isotopically heavier than the in situ water) would be expected. Because the data (3.3‰, black circle) plot above this line of complete equilibration, only a portion (30 to 40%) of the total SO42– pool has equilibrated.

Table 1.

Biogeochemical parameters of Blood Falls outflow during a brine discharge event. Outflow samples collected in December 2004. nd indicates none detected. The first six parameters and sulfate and chloride concentrations are from (33); total iron is from (7).

View this table:

Taylor Glacier is frozen to its bed, and surface-derived water does not penetrate to the base (8). Poorly understood hydrologic controls result in episodic release of brine. The data in this study are from one of these active discharge events. Salts and the iron mineral goethite rapidly precipitate upon contact of the outflow with the oxidizing atmosphere (5). Although regelation as the glacier slides over the bedrock and brine may add trace amounts of meteoric gases including O2 (9), multiple geochemical measurements reveal no quantitatively significant contributions. Radiocarbon data confirm that dissolved inorganic carbon (DIC) is old [Δ14CDIC = –993 ± 1 (SE) per mil (‰) (10)]. Presumably, interactions with air at the point of outflow collection are responsible for the small amount of 14CDIC in our samples (14C-free would be –1000‰). No dissolved O2 was detected in the brine [reduction potential (Eh) = 90 mV], and iron was 97% Fe(II), indicating that inputs from ice-bound atmospheric O2 are minimal.

Heterotrophic activity was measured by using 3H-thymidine incorporation (Table 1) (10). Although the most abundant electron acceptor in the brine is SO42–, isotope data suggest that SO42– is not terminally reduced in quantities sufficient to affect isotopes of sulfur. Values of δ34SSO4 and Δ33SSO4 in the brine (21.0‰ and range = 20.8 to 21.7‰ for 34SSO4; 0.08‰ and range = 0.06 to 0.09‰ for 33SSO4; n = 6) were similar to measurements of seawater SO42– from the past 15 My measured in marine barites (11). In contrast, values of δ18OSO4 (3.3‰; range = 2.7 to 4.9‰; n = 6) were up to 7‰ depleted compared with that of seawater δ18OSO4 from marine barites over the Pliocene [10.4 ± 1.6‰ (12)]. Because values of both δ34S and δ18O in SO42– are influenced independently by microbial sulfur metabolism, including sulfate reduction, disproportionation, and reoxidation reactions (13, 14), we expect that both oxygen and sulfur isotopes would be affected. For example, dissimilatory sulfate reduction to sulfide would cause preferential enrichment of 34S (SO42– → H2S, fractionation factor (ϵ) = 20 to 40‰ for natural populations) (15) because 34S-depleted S2– is sequestered in iron sulfides. This would increase the δ34SSO4 value of the remaining sulfate pool.

Our isotope data indicate that incorporation of 18O-depleted brine water oxygen into sulfate has occurred (Fig. 1B and table S3). During glacial advancement, meltwater mixing with the remaining seawater decreased the δ18O value of the brine from a marine value to its current composition (δ18H2OBrine = –39.5‰; Table 1). The depleted value of δ18OSO4 cannot be explained by abiotic oxygen isotope exchange between SO42– and water. Such equilibration would take tens of millions of years at subglacial temperatures and pH (16) (Table 1). However, oxygen isotope exchange between sulfite (SO32–) and water occurs rapidly [50% exchange in <5 min (17)], and the reduction of SO42– to SO32– is biologically mediated (7). Complete equilibration of SO42– via cycling through SO32– and exchanging all oxygen atoms with water would result in a 30.9‰ offset from brine water or a value for δ18OSO4 of –8.6‰ (Fig. 1B). This offset represents the temperature-adjusted equilibrium fractionation factor of 18O in SO32– relative to brine H2O (ϵ = 30.9‰ at –5.2°C). Given the measured values of δ18OSO4 (2.7‰ to 4.9‰), isotopic mass balance requires that 30 to 40% of the SO42– pool has exchanged its O atoms with water, likely through equilibration with SO32–.

Characterization of the Blood Falls microbial assemblage has revealed taxa that could participate in active sulfur cycling, including autotrophs and heterotrophs (table S1). Sulfate is biologically reduced to SO32– by using (phospho-)adenosine 5′-phosphosulfate–reductase by assimilatory [3′-phosphoadenosine 5′-phosphosulfate (PAPS)] or dissimilatory (APS) metabolisms, although APS also has been identified in sulfur oxidizers and certain organisms that only assimilate sulfate (18). The majority of APS genes detected in Blood Falls brine (Fig. 2) clustered with APS sequences of known dissimilatory and sulfur-disproportionating species (group 1). Sequences closely related to the APS gene from Desulfocapsa sulfexigens are consistent with the presence of a relative of this species among the 16S ribosomal RNA clones (7). Clone APS_20 (group 2) shares relations with sulfur-oxidizing isolates and Thermacetogenium phaeum, a syntrophic acetate-oxidizing bacterium that can also reduce sulfate (19). We cannot eliminate the possibility of trace production of H2S because one sequence (clone B_11) showed distant relation to groups known to mediate complete sulfate reduction (e.g., Desulfovibrio spp.). However, we were unable to detect dissimilatory sulfite reductase (dsrA) genes across several methods and attempts, and the sulfur isotope data are inconsistent with measureable quantities of sulfide formation.

Fig. 2.

Neighbor-joining phylogenetic tree of APS-reductase genes cloned from Blood Falls genomic DNA. Results of bootstrap analysis with 1000 replications are noted. The scale bar represents 10 nucleotide substitutions per sequence position. Clones from Blood Falls are in bold. GenBank accession numbers are listed in parentheses.

The presence of metabolically active cells requires a small supply of assimilated sulfur because sulfur composes ∼0.1% of cell biomass (15). We estimated a doubling time for heterotrophs of ∼300 days (Table 1) (10) equaling ∼106 generations over 1.5 My of isolation. If all cellular organic sulfur requirements were derived each generation by de novo assimilatory reduction, an upper limit of ∼35% of the SO42– pool would have been used. This result is consistent with our estimate of 30 to 40% SO42– turnover but is unlikely to provide the full explanation for sulfur cycling in the brine. Initially the system would have included reduced organic nutrients (N, S, and P), including S-containing amino acids, in stoichiometric proportion to the initial supply of organic matter. Availability of this additional pool of biological metabolites could decrease the effective assimilatory SO42– demand substantially. Additionally, the presence of diverse sulfur-cycling microbes (7) and several groups of APS genes indicate the presence of metabolic processes beyond assimilation strictly for biomass.

How then is the additional sulfate cycled? The unchanged values of δ34SSO4 indicate that there is insignificant loss of sulfur to sedimentary pyrite. This implies that the reduction of SO42– to SO32– does not proceed all the way to H2S, which is consistent with the observation of ferruginous and not euxinic conditions (no detectable H2S). To achieve these isotopic values, the net metabolism in the brine requires two remarkable properties: (i) SO42– must recycle through sulfur intermediates, and (ii) these sulfur intermediates must be quantitatively reoxidized to SO42–.

We advocate that sulfur is catalytically cycled to facilitate the oxidation of organic matter in a system in which Fe(III) is the terminal acceptor. Both assimilatory and dissimilatory reduction of SO42– proceed enzymatically through intermediate species (15). To date, no cultured dissimilatory organism has been described in which SO42– reduction does not ultimately terminate in H2S. The initial steps of SO42– → SO32– → S2O32– are endergonic, but progression to S0 when coupled to common electron donors [e.g., acetate and lactate (10)] does yield free energy (table S2). Coupling of any of the subsequent reoxidation reactions to the reduction of Fe(III) would also yield free energy (table S2). Additionally, numerous examples suggest that microbes use carrier molecules, including sulfur compounds, as electron shuttles between cells and iron oxides (2022). Cycling sulfur as an electron shuttle to catalyze iron reduction would also be a feasible route to reduced sulfur species in this system.

Such a catalytic cycle would preserve the total dissolved sulfur concentration (Fig. 1, inset) while progressively accumulating Fe(II) to a level controlled by the solubility product constant of siderite. Although the mechanism by which recycling of sulfur species mediates Fe(III) reduction is not fully understood (23), in contemporary marine pore waters at depths below O2 penetration values of δ18OSO4 show SO42– regeneration perhaps in association with Fe and Mn oxides (24). The net thermodynamics of such a system are favorable and could be described as syntrophic.

Ferric iron likely is mobilized by the scouring action of the glacier over the basement complex of metamorphic rocks intruded by granodioritic and granitic plutons (25). The redox state of these rocks is minimally Fe2.4+ in the contacts and Fe2.2+ in the basement sills and diabase sheets (26). A nonzero δ56Fe signature in Fe(II) dissolved in the brine indicates iron redox cycling, and the strong negative fractionation (–2.6‰; Table 1) implies dissimilatory iron reduction (27). Given the elemental abundance of iron in Earth's crust, Fe(III) may be an important electron acceptor for many subglacial microbial ecosystems. Most subglacial environments do not contain entrapped marine salts, but all subglacial biomes interact with their underlying bedrock (28), and many contain low concentrations of dissolved organic carbon (29). Consequently, glaciers may be important in delivery of soluble Fe(II). Below Taylor Glacier, the subglacial supply of Fe(III) appears to exceed the supply of electron donors such as reduced sulfur and organic matter. Thus, the system is paradoxically rich in electron acceptors relative to electron donors, despite being anaerobic; the result is the absence of euxinia.

During periods of diminished net photosynthesis, such as Neoproterozoic Snowball Earth episodes (30, 31), a decrease in organic flux from reduced photosynthetic production would drive the ocean away from SO42– reduction, analogous to the Blood Falls system. When organic matter became sufficiently limited, euxinia would cease and Fe(II) would accumulate. This respiratory source of Fe(II) would add to the flux of hydrothermal Fe to the deep ocean (13), and both could lead to an episodically ferruginous ocean (32). Importantly, our model is independent of the size of the marine SO42– reservoir because it depends only on the ratio of Fe(III) to labile organic matter. The brine below the Taylor Glacier provides a contemporary, natural example of an active, catalytic sulfur cycle, and it uniquely allows for the study of the long-term persistence of life and associated bioenergetics under ice.

Supporting Online Material

Materials and Methods

Fig. S1

Tables S1 to S4

References and Notes

View Abstract

Navigate This Article