Sulphur and carbon isotopes as tracers of past sub-seafloor microbial activity

Microbial life below the seafloor has changed over geological time, but these changes are often not obvious, as they are not recorded in the sediment. Sulphur (S) isotope values in pyrite extracted from a Plio- to Holocene sequence of the Peru Margin (Ocean Drilling Program, ODP, Site 1229) show a down-core pattern that correlates with the pattern of carbon (C) isotopes in diagenetic dolomite. Early formation of the pyrite is indicated by the mineralogical composition of iron, showing a high degree of pyritization throughout the sedimentary sequence. Hence, the S-record could not have been substantially overprinted by later pyrite formation. The S- and C-isotope profiles show, thus, evidence for two episodes of enhanced microbial methane production with a very shallow sulphate-methane transition zone. The events of high activity are correlated with zones of elevated organic C content in the stratigraphic sequence. Our results demonstrate how isotopic signatures preserved in diagenetic mineral phases provide information on changes of past biogeochemical activity in a dynamic sub-seafloor biosphere.

The S isotope composition (δ 34 S, a measure for the deviation of the isotope ratio of a sample from the international reference Vienna Canyon Diablo Troilite, VCDT, which for convenience is reported in ‰ values) of sulphate and sulphide is a good indicator for sulphate-reducing activity. Sulphide produced during microbial sulphate reduction is depleted in 34 S relative to the sulphate, with an S isotope discrimination, referred to as S isotope enrichment factor (ε 34 S) of up to 75‰ (e.g., Sim et al. 17 ). Accordingly, the residual sulphate is increasingly enriched in 34 S as more sulphate is consumed. The δ 34 S values in porewater sulphate thus show an increase with depth in the sulphate reduction zone, and the isotopic composition of the produced sulphide follows the same trend but is offset, approximately by the value of the apparent ε 34 S in the porewater (Jørgensen 14 ; Hartmann and Nielsen 18 ). Near the sulphate methane transition zone (SMTZ), where sulphate is almost entirely consumed, the δ 34 S of sulphide can therefore reach very high values (e.g. Borowski et al. 19 ; Pellerin et al. 20 ). As sulphate is the essential electron acceptor in anoxic sediments, and the sulphate reduction rate is stoichiometrically coupled to the organic carbon mineralization rate, the gradient at which δ 34 S increases with depth is also indicative of the overall dissimilatory activity in anoxic sediments. We propose that this increase is reflected in the δ 34 S of pyrite.
The challenge lies in the understanding of how this signal is recorded, because the δ 34 S of pyrite from a sediment sample averages all δ 34 S values of individual pyrite grains that accumulated over time within the zone of pyrite formation. The depth of the pyrite formation zone depends on the availability of reactive iron, modified by the dynamics of the sedimentary system. Moreover, existing records can be overprinted or even erased by subsequent biological or abiotic processes. Such overprinting may particularly occur where methanogenic zones persist at shallow depth for a very long time, so that under sulphide-free conditions substantial re-distribution of Fe may take place (e.g. Riedinger et al. 21 ).
We present a profile of δ 34 S-values extracted from diagenetic pyrite through a 200-m-thick sediment sequence at ODP Site 1229 to assess its validity as an archive of microbial activity through the last >2 Ma. This site shows a sulphate depletion through the top 30 m below seafloor (mbsf) with a SMTZ at 35 mbsf and a second SMTZ at 90 mbsf, below which sulphate diffuses upwards from a deep-seated brine (D'Hondt et al. 2 ; Fig. 1A; further information on the geological setting is provided as Supplementary Material). Dissolved total sulphide (here referred to as HS − ) was detected throughout the entire sediment column, with near zero concentrations around 75 to 95 mbsf, and at the bottom of the retrieved sediment (D'Hondt et al. 2 ). To test if pyrite was formed early after deposition or at a later stage, the amount of iron (Fe) bound to pyrite was quantified via the content of chromium-reducible S (CRS) and compared to the content of Fe-fractions showing different reactivity in the sediment. The Fe-fractions were determined using a sequential iron extraction procedure combined with Fe K-edge X-ray absorption near-edge structure (XANES) spectroscopy of iron minerals (three samples from different depths). Additional insights into coupled Fe-S cycling were gained by the comparison of the δ 34 S of modern-day dissolved sulphate in the porewater to the δ 34 S of pyrite. The Fe-S record was then juxtaposed with the total organic carbon (TOC) content and a previously measured δ 13 C record in diagenetic dolomite, to identify a potential co-variation that would be indicative of changes in microbial activity over time.

Results and Discussion
The S isotope record in pyrite. The downcore distribution of δ 34 S values in pyrite (CRS; Fig. 1B; Table S1) shows values between −30‰ and +5‰. In the top few metres of the sedimentary sequence, pyrite shows δ 34 S values around −20‰, and similar values are found in the middle part, within the present-day methanogenic zone. Higher values, reaching 0‰ or more occur near 30 m and 100 m below seafloor (mbsf), and one value near 180 mbsf. The isotope offset of about 70‰ between pyrite and sulphate in porewater of the uppermost few metres (data from Böttcher et al. 22 ) is at the upper limit of fractionation by microbial sulphate reduction commonly observed in marine sediments (e.g. Wortmann et al. 23 ; Jørgensen et al. 24 ; Pellerin et al. 20 ). The offset increases to 80‰ around 100 mbsf but decreases to about 30‰ near the bottom of the sequence, hence showing a strong variation in the apparent fractionation. The highly variable offset between the δ 34 S signatures of pyrite and sulphate is not consistent with pyrite precipitation from present porewater sulphide, which would expectedly have a more constant offset from porewater sulphate (cf. Jørgensen et al. 24 Table S1). These ratios approximately correspond to the degree of pyritization (DOP; ratio of pyritic Fe to pyritic Fe + acid-soluble Fe; Berner 26 ), indicating that most pyrite forms within the uppermost 2 mbsf. The increase is even higher, to ~0.8, if only fractions I through IV of the extraction scheme are considered for calculation. According to Poulton and Canfield 25 fractions I through IV essentially include all Fe-phases but the poorly reactive silicates. As shown by leaching experiments (Kasina et al. 27 ) fraction III includes reactive sheet silicates, such as smectites, from which Fe could potentially react with sulphide. Fe K-edge XANES spectroscopy showed that Fe occurs in significant proportions as pyrite (17-26%), while the rest of the Fe is present in the structure of sheet silicates, mostly from the illite and the chlorite groups (Fig. 3). Only in the uppermost sample at 2.55 mbsf, some Fe was present in smectite. As XANES spectra at the Fe K-edge are similar for different smectites ( Fig. 3B; Rennert et al. 28 ), the contribution of the standard spectra we had available (beidellite, saponite, and nontronite) may also represent montmorillonite, which commonly occurs in marine sediments. Overall, the pyritizable Fe-fractions make up a minor amount of the total Fe in the sediment (triangles in Fig. 2C).
More abundant is the Fe bound in poorly reactive sheet silicates, which are extracted with boiling HCl (extraction step V). This fraction contributes up to 30% of the total Fe as visualized by the slope of the regression line of the total extracted Fe (fractions I-V) vs. total Fe in Fig. 2C (dashed line). Canfield 29 has shown that the poorly reactive sheet silicates, such as chlorite and illite do not significantly react if exposed to sulphidic conditions over millions of years. Indeed, XANES analyses showed chlorite and illite as the most abundant Fe-mineral phases. The presence of these unreactive silicate phases explains why the ratio of CRS-Fe to total extracted Fe is only around 0.6 (blue regression line in Fig. 2A). An overestimation of CRS-Fe due to organically bound S is unlikely, as organic S has been shown to be largely non-extractable by Cr(II)-solution in Peru Margin sediments (Mossmann et al. 30 ; cf. also Fossing and Jørgensen 31 ), and would cause a perturbation of the linear trend of extracted vs. total Fe ( Fig. 2D; cf. Böttcher et al. 32 ). Because of the approximate extraction efficiency of 80% for sequential leaching of Fe (cf. Fig. 2C), the ratio of CRS-Fe to the total Fe-content, measured by X-ray fluorescence (orange regression line in Fig. 2A), is somewhat lower than the ratio of CRS-Fe to extracted Fe. Nevertheless, the same pattern is observed as for the ratio of CRS-Fe to extracted Fe (both with and without unreactive sheet silicates) and the regression lines of all three ratios in Fig. 2A show the same slope. The pool of reactive iron is exhausted in the uppermost few metres (~5 m) below the seafloor, and pyritizable Fe-fractions only make up a minor proportion of the total Fe in the rest of the sediment column.
The precipitation of Fe-sulphides is the result of reductive dissolution of reactive Fe(III) phases by free sulphide. HS − is produced by microbial sulphate reduction and reacts abiotically with different reactive Fe-oxides and oxyhydroxides (Berner 26 ; Rickard and Luther III 33 ). The actual electron transfer from HS − to the Fe(III) has been demonstrated to control rates of iron reduction, whereas binding of HS − on the Fe-mineral surface is not rate limiting (Afonso and Stumm 34 ). The reductive dissolution of Fe minerals is therefore not dependent on the HS − concentration in the porewater, as long as HS − is available. While Fe-sulphide (FeS) can be directly precipitated from the porefluid under sulphidic conditions, the formation of pyrite (FeS 2 ) requires an additional oxidation step. According to Wächtershäuser 35 and Thiel et al. 36 , the oxidation step may be coupled to the reduction of water to H 2 (eq. 1), which, however, may be readily consumed if sulphate is present.
Consistent with this reaction, Riedinger et al. 37 showed that acid volatile sulphide is rapidly converted to CRS if measurable quantities of free sulphide are present in the porewater. Within the sulphidic zone, the depth at which pyrite forms is, thus, limited and controlled by the amount and reactivity of solid-phase Fe(III). Based on the present activity at depth, where sulphide is still produced, it is reasonable to assume that sulphide was always available during the deposition of the 200-m-thick sediment sequence at ODP Site 1229, except, perhaps interrupted by episodes of expanded methanogenic zones, which however, did not cause a significant re-distribution of Fe. This is supported by the observation that the DOP increases rapidly in the uppermost sediment and remains rather constant through the sequence, suggesting that the reactive Fe-phases are rapidly pyritized in the uppermost few metres, independent of how intense the sulphate reduction rates are. Therefore, δ 34 S in diagenetic pyrite at ODP Site 1229 reflects the conditions in the near surface porefluid, unless sediment was affected by a permanently shallow SMTZ (i.e. shallower than the 2.5 m that may have been episodically reached; Contreras et al. 7 ) or the sediment was largely exposed to suboxic, i.e. sulphide-free conditions. This means that the pyrite has recorded the isotopic composition of dissolved sulphide in the top few metres of the sediment through time, and the sulphide produced at depth today is not preserved in the pyrite record. Assuming that the sedimentation rate was constant, the S isotope profile could thus be used as a record of average microbial sulphate-reducing activity in the near-surface sediment interval.
Sulphur isotope fractionation between sulphate and sulphide. A high sulphate reduction activity in the top interval of sediment leads to a steep depth-gradient in δ 34 S (i.e. sulphide is more strongly enriched in 34 S in the same depth interval), whereas a low sulphate reduction activity leads to a low depth-gradient in δ 34 S. Thus, if Fe-sulphide formation occurs over a constant depth range, one would observe less negative/more positive values preserved in the Fe-sulphides for situations in which the sulphate reduction activity was intense, i.e. higher δ 34 S values indicate higher sulphate reduction activity.  The situation is complicated by the fact that the sedimentation rate may change (Hartmann and Nielsen 38 ; Pasquier et al. 39 ). An increase in sedimentation rate would result in a larger depth range over which the Fe-sulphides form. At the same time, a higher sedimentation rate would lead to a more rapid burial of reactive organic matter, higher sulphate-reducing activity and accordingly steeper sulphate gradients (Meister et al. 40 ). Indeed, an upward shift of the SMTZ could be shown for the Peru margin ODP Site 1229, where the age model in Contreras et al. 7 suggests variations in sedimentation rate between 0.01 and 0.8 m/ka. The steeper gradient in δ 34 S would then also lead to more positive values in the Fe-sulphides.
A further uncertainty could be the isotope enrichment factor. However, large fluctuations in the ε 34 S are not expected for marine sediments that receive organic matter that -with regard to its composition and content -does not vary substantially over time. In laboratory experiments with sulphate reducing bacteria, ε 34 S increases when cell-specific sulphate reduction rates become smaller, but this effect is tied to the substrate, respectively the energy yield rather than the total availability of the substrate (e.g., Kaplan 48 ). Thus, the isotopic difference between sulphate and sulphide of ~70‰ in the top few mbsf at Site 1229 could reflect a long-term average ε 34 S between sulphate and sulphide observed in many porewaters of marine sediments (e.g. Claypool 45 ; Jørgensen et al. 24 ; Böttcher et al. 22 ). This offset should be representative for the true separation factor near the zone of pyrite formation. Thus, changes by more than 20‰ in the δ 34 S record of pyrite at Site 1229 are likely reflecting long-term shifts in the microbial activity in the top metres of the sub-seafloor.
The diagenetic history at Peru Margin Site 1229. For the reconstruction of the diagenetic history, the δ 34 S record of pyrite can be compared to several other parameters (Fig. 4). The two intervals showing higher δ 34 S-values also show somewhat elevated concentrations of total organic carbon (TOC). While the TOC data show a large scatter, and the measurements are not available at high resolution, shipboard core scans of the chromaticity value a* (red-green value; D'Hondt et al. 2 ) have been shown to correlate with the abundance of diatom ooze (Meister et al. 49 ; Aiello and Bekins 6 ). Diatom ooze contains high TOC, and a* can therefore be used as a proxy for TOC. Both TOC and a* show two maxima near 30 mbsf and 100 mbsf, respectively. A third maximum in a* occurs near 180 mbsf, where TOC data are lacking. Even though high TOC contents occur in the top few metres, one should take into account that this part of organic matter decays with a power-law with depth and its impact on the future record is not exactly known. Also, this part of the section still lies in the zone of ongoing pyrite formation. Based on the observed correlation with TOC it is conceivable that intervals of elevated δ 34 S indeed reflect past episodes of enhanced rates of sulphate-reducing microbial activity. In addition to elevated TOC it is very likely that also the sedimentation rates were larger during these two intervals. Although an age model at sufficiently high resolution is currently not available for these intervals, based on the age model of the last 100 ka at Site 1229 (Contreras et al. 7 ), higher sedimentation rates co-occur with times of enhanced upwelling on the Peru Margin and accordingly also higher primary productivity, higher organic sedimentation rate and higher TOC in the sediment. Apparently, however, the 100 ka glacial-interglacial cyclicity at Site 1229 has not resulted in elevated δ 34 S throughout the 100-m-thick Pleistocene interval. It is only the large-scale intervals (at 30 and 100 mbsf), which may reflect episodes of enhanced upwelling that lasted long enough to have left a visible imprint in the diagenetic δ 34 S record.
The intervals with elevated δ 34 S also correlate with elevated δ 13 C DIC values in diagenetic dolomite (Fig. 4). There is general consensus that dolomites form in carbonate-free sediment as a result of alkalinity production due to anaerobic metabolic respiration, in particular, sulphate reduction and anaerobic methane oxidation (e.g. Kelts and McKenzie 9 ; Baker and Burns 50 ; Moore et al. 51 , Meister et al. 10 ). While these processes produce negative δ 13 C DIC values from the decomposition of organic matter, more positive values in the inorganic carbon result from microbial production of isotopically light methane (Claypool and Kaplan 13 ). The δ 13 C values in dolomite in the two intervals at 30 and 100 mbsf (Fig. 4, solid diamonds) are far higher than δ 13 C in DIC of the porewater (line), even within the present methanogenic zones. These values can only be explained by strongly enhanced methanogenic activity in the past. This would be consistent with the interpretation of the S isotope record, which also indicates higher microbial activity during these two intervals.
In conclusion, two episodes of enhanced sub-seafloor microbial activity during early and mid Pleistocene are documented independently in the diagenetic S and C isotope records. Apparently the short-term 100 ka cycles described by Contreras et al. 7 are not recorded, but the records are useful for long-term variations in sub-seafloor biosphere activity. In continuous records, a constant ratio of CRS-bound Fe to total Fe provides a useful indicator to exclude that substantial amounts of Fe were re-distributed upon later diagenetic processes and, thus, that the δ 34 S in pyrite represents robust record of microbial activity. Our study thus demonstrates that diagenetic S and C isotope records bear the potential to trace million-year-scale variations in sub-seafloor microbial activity in the deep-time rock record.

Methods
Five differently reactive iron fractions were sequentially extracted and quantified by the method of Poulton and Canfield 25 using 5 different solvents: (I) Na-acetate for Fe-carbonates, parts of the acid volatile Fe-sulphides, adsorbed and dissolved Fe, (II) hydroxylamine-HCl for lepidocrocite and ferrihydrite, (III) Na-dithionite for goethite, hematite, and akaganéite, (IV) oxalate for magnetite, and (V) boiling HCl for Fe in sheet silicates. Eluents were analysed by a Thermo iCE 3000 Series atomic absorption mass spectrometer (AAS). The precision of the measurement was better than ±2% (standard deviation). A high extraction efficiency was confirmed by a near 1:1 ratio of total extracted Fe to total Fe from X-ray fluorescence analysis (XRF; Wien et al. 52 ).
The extraction of acid volatile and chromium reducible S (AVS and CRS) was performed after the standard method of Canfield et al. 53 and Fossing and Jørgensen 31 . CRS-bound Fe was calculated from CRS multiplied by a factor of two, assuming that CRS mainly contains iron sulphide with the stoichiometry FeS 2 (pyrite). Sulphur isotopes were analysed by the same method as described by Arning et al. 54 . Values are reported relative to the Vienna Canyon Diablo Troilite (VCDT) standard.
X-ray absorption near-edge structure (XANES) spectroscopy was carried out at the A1 beamline of the DORIS storage ring at Deutsches Elektronen-Synchrotron (DESY, Hamburg, Germany). Acquisition parameters were described in Meister et al. 55 . The samples were prepared as dry pellets or, for the sample at 2.55 mbsf, using anoxically stored wet sediment. In an anaerobic chamber, the sediment was filled into the well in the sample holder, which was tightly sealed with a kapton plate. XANES spectra were collected at the Fe K-edge from 6960 to 8000 eV with 5 eV steps up to 7082 eV and 0.25 eV between 7082 and 7152 eV. A reference foil of metallic Fe(0) was used for internal energy calibration of the monochromator (the first inflection point of the Fe K-edge was set at 7112.1 eV). XANES spectra were processed and analysed using the Horae Athena free software (Newville 56 ; Ravel and Newville 57 ). Experimental spectra were normalized and fitted to a linear combination of standard spectra of Fe minerals using a least-square minimization procedure.