Dissolved organic sulfur in the ocean: Biogeochemistry of a petagram inventory

See allHide authors and affiliations

Science  28 Oct 2016:
Vol. 354, Issue 6311, pp. 456-459
DOI: 10.1126/science.aaf7796

Inventory of an essential marine element

Sulfur is necessary for marine primary production and has a large impact on climate processes. Because it is difficult to detect accurately, the amount of dissolved organic sulfur in the ocean is poorly defined. Ksionzek et al. measured dissolved organic sulfur in the Atlantic to estimate its distribution and infer its quantity in the world's oceans (see the Perspective by Levine). The findings suggest that dissolved organic sulfur exceeds all other forms of organic sulfur by a factor of 10.

Science, this issue p. 456; see also p. 418


Although sulfur is an essential element for marine primary production and critical for climate processes, little is known about the oceanic pool of nonvolatile dissolved organic sulfur (DOS). We present a basin-scale distribution of solid-phase extractable DOS in the East Atlantic Ocean and the Atlantic sector of the Southern Ocean. Although molar DOS versus dissolved organic nitrogen (DON) ratios of 0.11 ± 0.024 in Atlantic surface water resembled phytoplankton stoichiometry (sulfur/nitrogen ~ 0.08), increasing dissolved organic carbon (DOC) versus DOS ratios and decreasing methionine-S yield demonstrated selective DOS removal and active involvement in marine biogeochemical cycles. Based on stoichiometric estimates, the minimum global inventory of marine DOS is 6.7 petagrams of sulfur, exceeding all other marine organic sulfur reservoirs by an order of magnitude.

In the early 1930s, Alfred Redfield noted that the ratio of carbon, nitrogen, and phosphorus in algal phyla remains surprisingly consistent across marine biomes. The canonical 106:16:1 Redfield ratio (1) originated from these observations and has since become a cornerstone of ocean biogeochemistry. Subsequent stoichiometric studies quantified the cellular quota of organic sulfur (OS) and found it to be similar to that of organic phosphorus (C124N16P1S1.3) (2). The magnitude of S acquisition, assimilation, and metabolism is not trivial given an average molar elemental ratio of C124N16P1S1.3 for marine algae (2). Based on this C/S ratio of ~95, the global phytoplankton biomass (~1 Pg C) (3) contains 0.028 Pg S, and the annual net marine primary production (48.5 Pg C year−1) (4) requires a sulfur assimilation of 1.36 Pg S year−1. Whereas regional marine dissolved organic sulfur (DOS) budgets have been constructed (5), quantification of the global inventory and its ties to other elemental biogeochemical cycles (C, N, P, and Fe) has been analytically hampered by the background concentration of sulfate (29 mmol S L−1), which exceeds the concentration of DOS by five orders of magnitude.

The discovery of OS coupling to climate processes (6) generated a surge of interest in the OS cycle and dimethylsulfoniopropionate (DMSP) specifically. DMSP is the precursor of dimethylsulfide (DMS) (7), a gas that is assumed to contribute to aerosol formation and climate regulation (6). The estimated annual production of DMSP by phytoplankton of 3.8 Pg C year−1 or 2.0 Pg S year−1 (8) represents an important sulfur assimilation pathway with rapid turnover rates and provides a substantial source of reduced carbon and sulfur for heterotrophic bacteria (9, 10). At the cellular level, the organic S and N cycles are intimately coupled through algal biosynthesis of the amino acids methionine and cysteine (11). Sulfur-rich peptides can also form metal-organic complexes and thus influence the speciation and mobility of trace metals in the ocean (12), with cascading effects on phytoplankton production, community composition, and carbon storage. Nonvolatile DOS is tightly linked to other major mineral assimilation pathways because it also comprises amino acids, vitamins, osmolytes, and primary metabolites (13, 14). The major sinks for these marine biogenic sulfur compounds are (i) remineralization to sulfate, (ii) incorporation into microbial biomass, (iii) efflux to the atmosphere (15), and (iv) transformation into the sizeable pool of nonvolatile marine dissolved organic matter (DOM) (662 Pg C) (16). Despite the relevance of marine DOS for ocean biogeochemistry, its quantitative depiction and connections and feedbacks to the C and N cycle remain elusive.

This study is based on water samples from the East Atlantic (EA) and the Southern Ocean (SO) collected in November and December 2008 between 50.2°N and 70.5°S (Fig. 1A) (17, 18). The concentrations of solid-phase extractable DOS (DOSSPE in μmol L−1 seawater) were analyzed by inductively coupled plasma sector field mass spectrometry (ICP-MS). Similar to the ambient dissolved organic carbon (DOC) concentration (17, 18), DOSSPE in the EA decreased significantly from 0.14 ± 0.02 μmol L−1 at surface depths of 0 to 105 m to ≤ 0.08 ± 0.01 μmol L−1 in deeper water ≥200 m (P < 0.001) (Fig. 1B and Table 1). DOSSPE correlated linearly with both extractable dissolved organic nitrogen (DONSPE) and DOCSPE (P < 0.001, RDOC = 0.86, RDON = 0.75) (Fig. 1C and fig. S1A), whereas the slopes differed significantly (P < 0.001). The molar DOSSPE/DONSPE ratios of 0.11 ± 0.024 were almost constant (slope of 5.3) throughout the water column and comparable to phytoplankton stoichiometry (S/N ~ 0.08, C:N:S = 124:16:1.3) (2), suggesting a predominantly biogenic DOS imprint (19) rather than abiotic incorporation of S into DOM as found in oxygen-limiting conditions (20). In contrast, molar DOCSPE/DOSSPE ratios in the EA increased with depth from 213 ± 25 in the surface to 268 ± 39 in deeper water (slope of 99.7; P < 0.001), suggesting higher biological reactivity of DOS relative to DOC. This is supported by earlier studies showing that microbial growth can be limited by the availability of reduced sulfur sources such as DMSP (9, 10).

Fig. 1 Cruise track and distribution of DOSSPE and DONSPE and molar DOCSPE/DOSSPE ratios in the surface ocean.

(A) Surface DOSSPE concentrations (μmol L−1) (colors) along the cruise track of research vessel Polarstern expeditions ANT XXV/1+2. (B) DOCSPE/DOSSPE ratios (contours) and DOSSPE concentrations (μmol L−1) (colors). (C) Potential density anomaly σ0 (kg m−3) (contours) and DONSPE concentrations (μmol L−1) (colors). For data below 200 m water depth, refer to Table 1.

Table 1 Average values and root mean square deviations of DOC and sulfur concentrations in the EA and SO and calculated global DOSMIN inventory.

View this table:

DOSSPE concentrations in the SO were pervasively low, whereas primary production was relatively high (see fig. S2 for chlorophyll concentrations). Depth-related changes in DOSSPE concentrations of 0.08 ± 0.01 μmol L−1 in the surface and 0.07 ± 0.01 μmol L−1 at ≥200 m depth and changes in molar DOCSPE/DOSSPE ratios of 262 ± 28 in the surface and 254 ± 26 at ≥200 m were insignificant (P > 0.05) (Table 1). Molar DOSSPE/DONSPE ratios of 0.10 ± 0.027 were similar to those found in the EA. A correlation of chlorophyll a with DOC or DOS was not observed. We speculate that the biogenic signature of DOS production was not detected due to short residence times in the mixed surface water and upwelling of old (5226 ± 64 years), nonlabile DOS from the deep SO (16) with low DOSSPE concentrations (0.07 ± 0.001 μmol S L−1) (Table 1).

To provide an estimate of nonlabile DOS removal, we correlated measured and reconstructed DOCSPE radiocarbon ages (17, 18) with DOSSPE concentrations (fig. S1B). Based on first-order kinetics, we found a strong correlation (R = 0.75, P < 0.01) of DOSSPE concentration with age, similar to that previously determined for DOCSPE (17, 18) (R = 0.61, P < 0.01). The long-term degradation rate coefficients for DOSSPE of kDOS = 2.54 × 10−4 year−1 and DOCSPE of kDOC = 1.53 × 10−4 year−1 differed significantly (P < 0.001) and reflected a higher reactivity (lability) of DOSSPE compared with DOCSPE. The long-term net removal rate of 2.7 × 10−5 μmol S L−1 year−1 for this nonlabile DOSSPE pool (see the supplementary materials for definition) results in stoichiometric changes in DOM over time and depth, similar to the preferential remineralization of N (and P) relative to C (21). In contrast, degradation rate coefficients for DOSSPE and DONSPE were similar and, consequently, molar DOSSPE/DONSPE ratios did not change significantly with age. Differences between DOSSPE and DOCSPE degradation kinetics are also reflected in DOSSPE and DOCSPE lifetimes (time at which the DOM concentration decreases to 1/e of its initial value): We calculated the average lifetime of DOSSPE of τ = 3937 years, which is lower than the lifetime for DOC of τ = 4500 years (18) and DOCSPE of τ = 6536 years (see the supplementary materials for details). As the molecular composition of the DOC and DOS pools differs, a direct comparison of DOSSPE degradation kinetics with commonly applied DOC fractions (labile, semilabile, or refractory), which are based on the DOC removal rate and lifetime (22), cannot be applied. Our results also indicate that DOS degradation kinetics, similar to previous studies on DOC (17), are determined by a continuum of reactivities of the contributing sulfur compounds rather than discrete degradation stages.

Relative changes in the contribution of labile DOS derived from biogenic production to the total DOS pool were assessed from two depth profiles analyzed for total hydrolysable methionine-sulfur yield [i.e., mole % of methionine-S versus total DOSSPE]. In the EA, we found a higher molar methionine-sulfur yield of 1.02 ± 0.14% in the surface water compared with 0.21 ± 0.10% in deeper water (≥200 m). Accordingly, we observed a considerable decrease of the methionine-sulfur yield with age (fig. S3). In the SO, the methionine-sulfur yield of 0.18 ± 0.04% was consistently low throughout the water column. Assuming a methionine-S:cysteine-S ratio of 1.7 (11), less than 2 mol % of the DOSSPE was protein-derived. This low value is consistent with previous data on amino acid carbon yield (23) suggesting that labile DOS in the form of sulfur-containing amino acids is efficiently remineralized or transformed, even in the surface ocean.

For the molecular characterization of DOS, we used Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR-MS) and identified 803 unique molecular formulas containing predominantly one sulfur atom, 81 of which were exclusively identified in surface water ≤ 105 m (total number of S formulas in the data set, 81,037). None of the formulas we detected occurred uniquely at depth or matched the composition of a peptide. However, it is likely that other sulfur-containing compounds were also present and not covered by our analytical window. The diversity of sulfur-containing compounds identified by FT-ICR-MS and the average molecular S/C ratio in the EA decreased significantly from 0.06 ± 0.001 in surface water to 0.05 ± 0.001 in deeper water (≥200 m; P < 0.001) (Fig. 2), whereas comparable trends in the SO were not observed. Similar to previous molecular studies on DOC (17, 18), the most persistent S formulas at depth showed higher unsaturation (lower molecular hydrogen/carbon ratio) (Fig. 2) and slightly larger molecular size (427 ± 5.6 Da in surface water and 441 ± 10.9 Da at ≥200 m).

Fig. 2 Molecular changes of sulfur-containing compounds in the EA.

Every dot represents a specific sulfur-containing molecular formula. Each formula is represented by its molecular H/C and O/C ratio (van Krevelen plot). The size of the data points represents the molecular S/C ratio. Higher S/C ratios indicate a higher amount of sulfur in the formula. Colors represent two depth intervals: 0 to 105 m (red dots) and >1000 m (blue circles). In the surface, the number of different formulas (chemical diversity) was higher. Most unique sulfur compounds in the surface showed a higher content of hydrogen (saturation) and oxygen (oxidation). The average molecular formula for each depth interval is displayed.

The SPE method applied (24) yields lower extraction efficiencies for highly polar organic compounds (e.g., 22% for marine DON) as compared with DOC (42%) (17, 18). Changes of the DOC and DON extraction efficiencies with depth, however, were insignificant (PDOC = 0.85, PDON = 0.45). Therefore, we can assume that the extraction yield for polar OS compounds is also lower than for DOC and independent of water depth. Using the average measured molar DOCSPE/DOSSPE ratio (Table 1) and the DOC concentrations in original seawater, we can reconstruct a conservative minimum for the original DOS concentration in seawater ([DOS]MIN) (Table 1 and Eq. 1).[DOS]MIN = [DOC] ̸(DOCSPE/DOSSPE) (1)where [DOC] is the molar DOC concentration in original seawater and DOCSPE/DOSSPE is the measured molar elemental ratio in the extracts. The calculated [DOS]MIN concentrations were 0.34 ± 0.08 and 0.19 ± 0.04 μmol L−1 in EA and SO surface waters, respectively (Table 1). This concentration range was consistent with previous data from the Sargasso Sea (0.04 to 0.4 μmol DOS L−1) (5). For comparison, the mean concentrations of dissolved DMS and DMSP in the surface of the EA during our cruise were 0.0036 and 0.0032 μmol L−1, respectively (25), representing ~2% of [DOS]MIN in the EA. The global average concentration for dissolved DMS and DMSP were previously estimated at 0.001 to 0.007 μmol L−1 (26) and 0.003 μmol L−1 (27), respectively, contributing only ~2.3% of the total [DOS]MIN.

Based on the global oceanic DOC inventory of 662 Pg C (16) and depth-integrated molar DOCSPE/DOSSPE ratios, the minimum global oceanic DOS inventory (DOSMIN) is 6.7 Pg S (6700 Tg), ~600 Tg of which are present in the upper 200 m of the water column (Table 1 and Fig. 3). If we assume that the molar C/S ratio of ~95 in phytoplankton is the lowest possible ratio for DOM, the maximum size of the global DOS inventory is 18.6 Pg S. Hence, DOS represents the major reservoir of organic S in the ocean, larger than OS in biomass, particles, or volatile compounds combined (Fig. 3). More important, these numbers raise new questions on the marine sulfur budgets: Only 13 to 37 Tg S year−1 of the total DOS pool (red frame in Fig. 3) are released to the atmosphere as DMS (28) and DOS degradation products such as carbonyl sulfide (COS) (0.4 Tg S year−1) and carbon disulfide (CS2) (0.3 Tg S year−1) (Fig. 3) (28). In total, these fluxes represent less than 3% of the annual sulfur assimilation of 1.36 Pg S year−1 by primary production, suggesting that rapid biogeochemical cycles of labile sulfur compounds (red cycle in Fig. 3) are superimposed on the large background of nonlabile DOS (red frame in Fig. 3), which we consider to be derived from the microbial carbon pump (29). Seasonal variation of C/S ratios by changes in production and microbial or photodegradation has an important effect on the DOSMIN estimates in the surface (5). However, the value for our global DOSMIN estimate is dominated by the relatively invariant C/S ratios of 266 ± 41 in the large water body below the photic zone (>200 m) and therefore only marginally affected by seasonal effects. Many previous studies focused on the labile (and partly volatile) proportion of the DOS cycle (Fig. 3). This study enables important insights into the biogeochemistry of the vast pool of nonlabile DOS. So far, the organic sulfur budgets cannot be closed, particularly because the connection between the rapid cycling of labile DOS and the nonlabile proportion of the organic sulfur cycle remain unquantified.

Fig. 3 Simplified marine organic sulfur cycle.

Schematic overview of organic sulfur reservoirs and fluxes. All numbers refer to organic sulfur, except for the oceanic sulfate inventory and the land-atmosphere flux (total S). Known and calculated organic sulfur fluxes are shown as solid lines and unknown fluxes as dotted lines. The red circle indicates the rapid and important cycling of labile DOS compounds such as DMSP (depicted in the small white box). For corresponding data and references, see table S3.

Supplementary Materials

Materials and Methods

Figs. S1 to S3

Tables S1 to S3

References (3036)

References and Notes

  1. Acknowledgments: This work was supported by the DFG-Research Centre/Cluster of Excellence “The Ocean in the Earth System” and a Ph.D. grant by the Deutsche Forschungsgemeinschaft (DFG) in the framework of the priority program “Antarctic Research with comparative investigations in Arctic ice areas” (grant KO 2164/8-1+2). We are grateful to research vessel Polarstern captain, crew, and chief scientists G. Kattner (ANTXXV-1) and O. Böbel (ANTXXV-2); I. Stimac is acknowledged for technical support with ICP-MS analysis and K.-U. Ludwichowski for support with methionine analysis; we thank C. Marandino for DMS and DMSP data; S. Frickenhaus is acknowledged for support with statistical analysis, and B. Kanavati, M. Harir, and J. Uhl for support with FT-ICR-MS analyses. G. Kattner and R. Alheit are acknowledged for helpful discussions and proofreading. The data presented in this paper are available at the PANGEA data library (doi:10.1594/PANGAEA.858568). B.P.K. designed the research. O.J.L and B.P.K. collected and processed the samples. K.B.K. and W.G. performed ICP-MS analysis and data evaluation. J.G. carried out methionine analyses, S.L.M. performed 14C analysis, and P.S.-K. performed FT-ICR-MS analysis. The paper was written by K.B.K. and B.P.K, with input from all coauthors.
View Abstract

Navigate This Article