Report

Sulfate was a trace constituent of Archean seawater

See allHide authors and affiliations

Science  07 Nov 2014:
Vol. 346, Issue 6210, pp. 735-739
DOI: 10.1126/science.1258966

Dissecting ancient microbial sulfur cycling

Before the rise of oxygen, life on Earth depended on the marine sulfur cycle. The fractionation of different sulfur isotopes provides clues to which biogeochemical cycles were active long ago (see the Perspective by Ueno). Zhelezinskaia et al. found negative isotope anomalies in Archean rocks from Brazil and posit that metabolic fluxes from sulfate-reducing microorganisms influenced the global sulfur cycle, including sulfur in the atmosphere. In contrast, Paris et al. found positive isotope anomalies in Archean sediments from South Africa, implying that the marine sulfate pool was more disconnected from atmospheric sulfur. As an analog for the Archean ocean, Crowe et al. measured sulfur isotope signatures in modern Lake Matano, Indonesia, and suggest that low seawater sulfate concentrations restricted early microbial activity.

Science, this issue p. 703, p. 742, p. 739; see also p. 735

Abstract

In the low-oxygen Archean world (>2400 million years ago), seawater sulfate concentrations were much lower than today, yet open questions frustrate the translation of modern measurements of sulfur isotope fractionations into estimates of Archean seawater sulfate concentrations. In the water column of Lake Matano, Indonesia, a low-sulfate analog for the Archean ocean, we find large (>20 per mil) sulfur isotope fractionations between sulfate and sulfide, but the underlying sediment sulfides preserve a muted range of δ34S values. Using models informed by sulfur cycling in Lake Matano, we infer Archean seawater sulfate concentrations of less than 2.5 micromolar. At these low concentrations, marine sulfate residence times were likely 103 to 104 years, and sulfate scarcity would have shaped early global biogeochemical cycles, possibly restricting biological productivity in Archean oceans.

Sulfur interacts with carbon and oxygen in global biogeochemical cycles that regulate Earth’s surface chemistry and biology (1). At 28 mM, sulfate is abundant in modern seawater, fueling extensive sedimentary microbial sulfate reduction (MSR) (2). At these concentrations, MSR typically imparts large sulfur isotope fractionations (3), allowing the use of sulfur isotopes to reconstruct past global change (4, 5). Small sulfur isotope fractionations preserved in bulk pyrite from Archean (>2400 million years ago) rocks led to the original conclusion that the Archean oceans contained <200 μM sulfate, or ~1% of modern seawater (5). The distribution of mass-independent sulfur isotopes in Archean sediments (4, 68) and box models of global sulfur cycling (9) imply even lower Archean seawater sulfate of <60 to 80 μM. Paradoxically, microscale sulfur isotope data from Archean pyrites (1012) reveal large sulfur isotope fractionations of up to 40 per mil (‰) in the Archean—fractionations only seen at hundreds to thousands of micromolar sulfate in modern environments (1315). Electron-donor availability (3, 16) and MSR rate (3, 16, 17), however, also exert influence, with larger fractionation typically imparted when electron donors limit sulfate-reduction rates, allowing the unexplored possibility for large fractionations in low-sulfate (<100 μM) environments.

We explored sulfur isotope fractionation in Lake Matano, Indonesia, an extremely low-sulfate analog for the Archean oceans (18). Lake Matano is a persistently stratified ferruginous (Fe2+-rich) lake, where dissolved ferrous iron (Fe2+) accumulates below a chemocline located at ~115 m depth (Fig. 1A). The upper waters of Lake Matano have sulfate concentrations of less than 30 μM (Fig. 1B), far lower than that of the natural environments studied for sulfur isotope fractionation thus far (13, 14) and lower than previous upper limits for Archean seawater (5, 8, 9). MSR is active within Lake Matano’s water column, and peak rates of 30 to 40 nM day−1 are reached at sulfate concentrations of between 5 and 10 μM (11) (Fig. 1C).

Fig. 1 Vertical chemical profiles in Lake Matano.

(A) Dissolved O2 and Fe2+. (B) SO42– and HS (solid line shows modeled SO42– concentrations). (C) Sulfate-reduction rates (SRRs) (solid line depicts SRRs imposed in the 1D reaction-transport model). d, day. (D) δ34S value of SO42–. ICP-MS, inductively coupled plasma mass spectrometry; IR-MS, isotope-ratio mass spectrometry. Data in (A) to (C) come from (32), except model results.

We measured the δ34S values of sulfate and sulfide in Lake Matano’s water column (Fig. 1D) and in the underlying sediments (19, 20). The δ34S values of sulfate ranged from 8.1‰ within the surface waters to 39.1‰ in the lower reaches of the chemocline, where sulfate is present at barely detectable concentrations (Fig. 1D). Sulfate reduction in the chemocline thus leads to strong isotopic fractionation, despite extremely low sulfate concentrations, favoring the incorporation of 32S into the sulfide produced. Measurements of water-column sulfides reveal δ34S values from –13.2 to 5.4‰ (table S1), demonstrating that they record large fractionations with a range in δ34S of up to 18.6‰. Depending on the depths considered, the δ34S value of water-column sulfides translates to an appreciable isotopic difference Embedded Image between water-column sulfide and the surface-water sulfate pool of up to 23‰.

We used both a Rayleigh distillation model and a one-dimensional (1D) reaction-diffusion model to calculate sulfur isotope fractionation factors. Rayleigh models underestimate fractionation in open systems (21, 22) like Lake Matano and therefore provide minimum estimates of the true fractionation. Using the Rayleigh model, we obtain a fractionation factor Embedded Image of 21 ± 1‰ (16) (Fig. 2A). To further assess the true magnitude of fractionation during MSR, we constructed an open-system, reaction-diffusion model. As expected, applying the fractionation factor obtained from the Rayleigh model to the reaction-diffusion model underestimates the fractionation observed (Fig. 2B), whereas fractionations ranging between 20 and 70‰ encompass our entire sulfur isotope data set (Fig. 2B). The best fit with constant fractionation, independent of sulfate concentrations, comes from a fractionation factor of 35‰. As MSR proceeds, sulfate concentrations decrease (Fig. 1B), probably shifting MSR from organic matter limitation, which allows expression of large isotope fractionation, to sulfate limitation, which progressively mutes fractionation as sulfate concentrations decrease (3, 16). We therefore also tested a model with fractionation of 70‰ at sulfate concentrations of >6 μM, with a linear decrease to 0‰ when sulfate is exhausted, obtaining an equally good fit. Regardless of the model used, these trends in the δ34S values of sulfate illustrate large isotope fractionations down to sulfate concentrations below 6 μM (Fig. 2, A and B), confirming that MSR can produce large isotope fractionations at sulfate concentrations more than one order of magnitude lower than previously demonstrated (5, 14, 15).

Fig. 2 Sulfur isotope fractionation in Lake Matano.

(A) The slope of the line taken from the natural log of sulfate 34S/32S (R) normalized to that of sulfate at the top of the chemocline 34S/32S (R0, 0.0446) versus the natural log of the sulfate concentration (C) normalized to the sulfate concentration at the top of the chemocline (C0, 16 μM). The slope of this relation (m = –0.2075) is used to calculate the Rayleigh fractionation factor [α = (m + 1)−1 = 1.0211]. (B) Results from our 1D reaction-diffusion modeling with constant fractionation factors (α) of 1.020, 1.035, 1.045, and 1.070, in addition to a variable fractionation factor with α of 1.070 at sulfate concentrations >6 μM and a linear decrease of α from 1.070 to 1.000 below 6 μM. (C) Ranges in δ34S observed (solid lines and circles) or computed (hollow lines and circles) for different sulfur pools in Lake Matano. The shaded box outlines 1 SD from the anoxic sediment mean. AVS, acid volatile sulfide; CRS, chromium reducible sulfide.

Sediments under the chemocline record the integrated δ34S values of sulfide exported from the water column and exhibit a range of δ34S values from –4.2 to 6.6‰, with a mean of 2.5 ± 2.5‰ (Fig. 2C). These isotopic compositions are consistent with the range of δ34S values observed in the water column and predicted by our fractionation models (Fig. 2, A and B) and are up to 14.9‰ lower than that of sulfate in the surface waters of the lake (table S1). Reaction-diffusion models with either a constant fractionation of 35‰ or a variable fractionation of 70‰ that decreases below 6 μM sulfate yield integrated sulfide export with δ34S values of 3.8 and 2.7‰ (Fig. 2C), respectively. These δ34S values are similar to the mean of measured sediment sulfides, in contrast to the model with a 20‰ fractionation, which yields a sulfide export flux at the far maximum range of the sediment δ34S values (6.5‰) and outside of the standard deviation of the sediment mean.

The lack of a full expression of water-column sulfur isotope fractionation in bulk δ34S measurements of sedimentary sulfides is due to the depletion of sulfate to low concentrations in the chemocline and the development of a strong water-column gradient in sulfate concentration and isotopic composition. As a result, δ34S values of sulfate increase with decreasing sulfate concentrations (Fig. 1, B and C), leading to a reservoir effect and the production of correspondingly 34S-enriched sulfide, despite the strong fractionation imparted during sulfate reduction. In the end, the net isotopic fractionation between lake-surface sulfate and sedimentary sulfide is only ~7.5‰. A similar effect has been observed in other lakes, albeit at much higher sulfate concentrations (14, 15).

Compilations of the S-isotope composition of Archean sulfides from bulk sediment analyses suggest that the expression of S-isotope fractionation in the Archean was typically less than ~10‰ (Fig. 3). Though most Archean sulfides display little fractionation at the scale of bulk sediment analyses, up to 30‰ (Fig. 3) variability can be observed in the δ34S of some Archean bulk sediment analyses. Microscale analyses also reveal a broader range in δ34S of more than 40‰, implying that much larger isotope fractionations were possible (1012). The scatter of sulfur isotope data around the Archean MIF-S array (23) supports the idea that microbial sulfate reduction in the Archean was accompanied by large sulfur isotope fractionations, possibly more than 40‰, but that the expression of this isotope fractionation at the scale of bulk sedimentary sulfides was muted, similar to our observations in Lake Matano.

Fig. 3 Compilation of nearly 3000 individual measurements of the δ34S values of bulk Archean sedimentary sulfides.

The black line shows the normal distribution, and the general agreement between the data and the normal distribution suggests a single population. The vertical hatched band delineates the likely range of δ34S for surface ocean seawater sulfate (2628). The vertical gray band shows a 10‰ difference from seawater sulfate. Very few measurements extend beyond this 10‰ difference.

To test possible upper limits on Archean seawater sulfate concentrations, we have adapted our reaction-diffusion model to simulate a stratified Archean ocean water column. Like in Lake Matano, we expect that pelagic MSR would have ensued under the ferruginous ocean conditions dominating marine chemistry throughout much of the Archean eon (24). In an approach similar to previous models for sedimentary S-isotope fractionation (5), we varied seawater sulfate concentrations and computed the integrated δ34S of sulfide exported from the water column to underlying sediments (Fig. 4). We also assume that MSR would take place in sediments, so we modeled the δ34S of diagenetic sulfides formed under a range of overlying seawater sulfate concentrations.

Fig. 4 Models of marine sulfur cycling and isotope fractionation in the Archean eon.

(A) Modeled rates of microbial sulfate reduction in a stratified Archean ocean water column with different surface seawater sulfate concentrations. (B) Resulting sulfate concentration profiles. (C) Sulfate (solid lines) and sulfide (dashed lines) δ34S profiles generated using a variable ε of 30‰ at sulfate concentrations >6 μM that decreases to 0‰ when sulfate is exhausted. (D) Mean integrated δ34S for sulfide produced and exported from the water column to sediments (blue) and diagenetic sulfide (red) at various surface seawater sulfate concentrations. The solid black line indicates the imposed sulfur isotope fractionation factor at different sulfate concentrations (ε ≤ 30‰, justified from microscale analyses of Archean pyrites), and the gray dashed line symbolizes a constant fractionation factor (ε = 30‰). The golden diamond distribution plot at left illustrates sedimentary Δ34Ssulfate-sulfide calculated from bulk Archean sulfides (from Fig. 3) and using 5‰ as a conservative value for δ34S of seawater sulfate. The horizontal gray band delineates values between the 25th and 75th percentiles, whereas the horizontal dashed lines delineate the 5th and 95th percentiles, encompassing 90% of the Archean data. The arrow demarcates the previous 200 μM threshold for the full expression of sulfur isotope fractionation (5).

Our water-column model shows that at modest MSR rates, comparable to those measured in the Chilean oxygen minimum zone (25), appreciable sulfate drawdown occurs with surface seawater sulfate concentrations in the low micromolar range (Fig. 4, A and B). Fractionation factors typical for marine environments (30‰), and justified as a conservative estimate based on microscale measurements of δ34S in Archean pyrites, translate to a large range in the δ34S of pelagic sulfate and sulfide (Fig. 4C), showing that reservoir effects similar to those in Lake Matano develop under conditions typical for stratified marine environments. Due to a combination of sulfate drawdown, reservoir effects, and decreased isotope fractionation at low sulfate concentrations, the integrated sulfide exported from the modeled Archean water column has δ34S values closer to seawater sulfate than would be expected due to the isotope fractionation imparted. The imparted fractionation is best reflected by the sulfide produced in the upper regions of the water column.

Application of a constant fractionation factor of 30‰ in our models results in large differences between the δ34S values of seawater sulfate and bulk pyrites of 15 to 23‰ (Fig. 4D). Comparing these modeled δ34S values for sulfide with the distribution of δ34S in bulk Archean sulfides (yellow diamond distribution plot in Fig. 4D) and assuming seawater δ34S of 5‰ [the range reported is 3 to 8‰ (2628)] shows that these large differences are not supported by the bulk pyrite record. It therefore implies that other processes, such as sulfate limitation of MSR, were at play in the Archean. Conservatively applying a fractionation factor that decreases below 6 μM sulfate, a sulfate concentration that imparts large fractionation in Lake Matano, brings modeled differences between the δ34S of seawater sulfate and bulk pyrites into a range supported by the Archean bulk pyrite record (Fig. 4D). Under this scenario, both water-column and sediment models predict that MSR would impart bulk sediment sulfide δ34S values of more than 10‰ lighter than seawater sulfate at sulfate concentrations more than ~5 μM (Fig. 4D). Comparison under this scenario suggests that more than half of the measured sulfide δ34S values could be described by deposition at seawater sulfate concentrations between 1 and 2.5 μM, and more than 90% deposited at seawater sulfate concentrations <5 μM (Fig. 4D). Taking into consideration that the δ34S value of seawater sulfate may have reached up to 15‰ in the Neoarchean, and that δ34S values in pyrite may also include contributions from 34S-enriched photochemical sources (29), the concentration window may have extended as high as 10 to 15 μM. Higher seawater sulfate concentrations would have left bulk sulfide δ34S values much lighter and are therefore not supported by the sedimentary δ34S sulfide record. The ~30‰ variability observed at the scale of some bulk sediments could be imparted by dynamic processes that cause changes in the concentration of sulfate, the rate of sulfate transport into the sulfate-reduction zone, the rate of MSR (17), electron-donor availability (16), or the δ34S of seawater sulfate. These could include variability in depositional depth of sediments analyzed, fluctuations in the depth of mixing, shifting organic matter availability, or variable contributions of atmospheric versus riverine sulfur fluxes to the oceans. Overall, both the limited sulfur isotope fractionation measured at the bulk sediment scale and the large isotopic fractionations at the microscale point strongly to sulfate concentrations less than 2.5 μM.

Our results suggest that Archean ocean sulfate concentrations were <0.01% modern seawater, implying very different global sulfur dynamics. With surface seawater sulfate concentrations in the low micromolar range, sulfate residence times (16) would have been on the order of 103 to 104 years, and sulfate could have been poorly mixed in the Archean oceans. Though homogeneity of S isotopes in some Archean barites has been taken as evidence for conservative sulfate behavior (29), such conservative behavior at these low seawater sulfate concentrations would imply smaller-than-estimated volcanic and weathering sulfate fluxes to the Archean oceans (30). Regardless, the short residence times would have rendered seawater sulfate and its isotopic composition extremely sensitive to perturbations in the global sulfur cycle.

At seawater sulfate concentrations up to 2.5 μM, sediment sulfate reduction would have contributed less than ~10% to sedimentary organic carbon degradation (16), leaving the balance to fuel other microbial processes. Organisms also require sulfur as a nutrient, using it for protein synthesis at a typical cellular ratio of 48C:1S:0.45P (31); cellular sulfur quotas are thus higher than those of phosphorus. At low concentrations, nutrients such as phosphorus tend to limit biological production, and by analogy, sulfur may have played a more important role as a biologically scarce nutrient.

Supplementary Materials

www.sciencemag.org/content/346/6210/735/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 and S2

Tables S1 to S5

References (3368)

Data S1

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank A. Sturm and C. Henny for help with fieldwork. S. Poulton provided sediment δ34S data. A. Hefford helped compile Archean S-isotope data. Funding to S.A.C. was provided by an Agouron Institute Geobiology Fellowship and a Natural Sciences and Engineering Research Council of Canada Postdoctoral Fellowship. Additional funding was provided by the Danish National Research Foundation (grant no. DNRF53) and the European Research Council. All data are available in the supplementary materials.
View Abstract

Navigate This Article