Introduction

On geological timescales, the burial rate of sedimentary organic carbon (OC) exerts major control on the concentrations of atmospheric O2 and CO21 and thus substantially influences Earth’s environmental conditions. Reactive iron (FeR, defined as iron (oxyhydr)oxides that can be reductively dissolved by sodium dithionite) can promote the preservation of OC in a wide range of terrestrial and marine environments2,3,4,5,6, as reactive iron-bound OC (FeR-OC) has long been believed to be more resistant to microbial degradation under aerobic conditions7,8. It is estimated that ~20.2 ± 15.5% of the total OC (TOC) is directly bound to FeR in marine surface sediments9, which corresponds to a global reservoir of 7–54 Pg of OC (roughly 1–6% of the amount of the atmospheric CO2)9,10, demonstrating its potentially important role in preserving the sedimentary OC. A comprehensive picture regarding the fate of FeR-OC on geological timescales is still lacking but several studies of sediments of Pleistocene as well as late Paleocene to early Eocene age suggest that the fraction of FeR-OC in TOC (fFeR-OC) remains relatively stable11,12,13,14. However, the question to what degree the FeR-OC reservoir interacts with sedimentary biogeochemical processes, in particular those involving redox reactions of iron and sulfur, remains unresolved.

FeR has long been found to promote organic matter preservation in terrestrial soils as well, especially via adsorption5,15. However, recent studies indicated the interactions between FeR and OC in soils are highly influenced by the redox oscillations during water-table fluctuatioins16,17,18 in terrestrial environments. Under anoxic conditions, FeR reduction releases FeR-OC and increases the anaerobic remineralization of soil organic matter17,18. While under oxic conditions, newly-formed FeR promotes OC retention on mineral surfaces16. In anoxic marine sediments, FeR is one electron acceptor actively involved in biogeochemical processes19,20,21. Both FeR reduction by microorganisms and biogenic sulfide produced during sulfate reduction could potentially weaken the association with OC. Indeed, recent lab incubations demonstrated that FeR-OC can be remobilized during microbial iron reduction and subsequently utilized as an electron donor and/or carbon source for microbial communities22,23,24. Consequently, the remobilization of FeR-OC during early diagenesis may influence the relative size of the FeR-OC reservoir that is preserved in sediments for a long time. Additionally, the supply of FeR-OC is expected to depend on the hydrological conditions through the influence on continental weathering and FeR formation as well as on sea level through the influence on shelf topography and thus the transport of continental detritus into the deep sea25,26,27; both factors are expected to vary periodically on glacial-interglacial timescales. To disentangle the effect of FeR-OC supply and early diagenetic reworking on sedimentary FeR-OC, downcore FeR-OC records need to be established and related to both geochemical zonation and sediment chronology.

In this study, we analyzed two gravity cores (QDN-G1 and QDN-14B, Fig. 1a) from the northern South China Sea in order to determine the fate of the sedimentary FeR-OC via its quantitative and isotopic analysis. The core QDN-G1 represents typical continental slope sediments (1478 m water depth), while core QDN-14B (1370 m water depth), around 35 km southwest from core QDN-G1, was influenced by nearby cold seeps expelling methane-rich fluids28. Consequently, QDN-14B can be used to comparatively evaluate the influence of microbial activities on the potential remobilization of FeR-OC in the diagenetically active zones where sulfate reduction coupled to either organic matter remineralization or methane oxidation occurs. Meanwhile, the core QDN-G1 consisting of sediments with relatively low microbial activity is used for exploring the long-term preservation of FeR-OC on glacial-interglacial timescale with well established age model covering the past 97 kyr (see Methods and Supplementary Fig. 1 for detailed information).

Fig. 1: Map of the cores studied in the northern South China Sea and their biogeochemical zonation.
figure 1

a Map of cores QDN-G1 and QDN-14B. River systems entering the northern South China Sea are indicated by blue and red lines. b A schematic of biogeochemical zonation, including iron reduction zone (IRZ), sulfate reduction zone (SRZ), sulfate-methane transition zone (SMTZ) and methanogenic zone (MZ), in cores QDN-G1 and QDN-14B inferred by porewater geochemistry. Fe(II) concentration, sulfate concentration, dissolved inorganic carbon (DIC) concentration and carbon isotope ratio of DIC (δ13CDIC) are shown in green, red, blue and magenta lines, respectively. The location of the SMTZ is highlighted in purple. Porewater geochemistry data in QDN-14B are from Niu et al., 2017 and Hu et al., 201928,60. Detailed porewater geochemistry data are plotted in Supplementary Fig. 2 and 3. Source data are provided as a Source Data file.

Full size image

To quantify the amount of FeR-OC, the citrate-bicarbonate-dithionite (CBD) method described in detail by Lalonde et al.2 and Salvadó et al.3 was applied in this study (see Methods). This method targets only reactive iron (oxyhydr)oxides, which are presumably accessible for microorganisms, and leaves unreactive phases such as iron-containing silicates untouched29. By incorporating analyses of porewater geochemistry, including the concentrations of ferrous iron, sulfate, dissolved inorganic carbon (DIC), and the carbon isotope ratio of DIC (δ13CDIC), all of which are tied to microbially mediated processes, our study takes a critical step in assessing the stability of sedimentary FeR-OC in response to post-depositional microbial activities and sheds lights on its fate in subseafloor sediments.

Results and discussion

FeR-OC remobilization in SMTZ sediments

The core recoveries of QDN-G1 and QDN-14B were 10.8 m and 8.4 m, respectively. These two cores penetrated sediments with distinct diagenetic activities. Microbially mediated early diagenetic redox reactions have resulted in distinct geochemical zonation30, i.e., the iron reduction zone (IRZ), the sulfate reduction zone (SRZ), the sulfate-methane transition zone (SMTZ), and the methanogenic zone (MZ) (Fig. 1b). Despite similar TOC contents in sediments from the two cores (see below), the higher methane fluxes and the associated rapid depletion of sulfate via anaerobic oxidation of methane resulted in a shallower SMTZ in core QDN-14B (280–440 cmbsf) compared to core QDN-G1 (potentially around 720–960 cmbsf) (Fig. 1b). This difference in activity is supported by geochemical modeling (see Methods), which showed a more than two times higher net sulfate reduction rate in the SMTZ in core QDN-14B compared to core QDN-G1 (0.034 vs. 0.015 mol m−3 yr−1) (Supplementary Fig. 4).

The FeR contents in the two cores range from 0.7 to 5.7 mg g−1, which reaches the maximum at 200 cmbsf and 240 cmbsf in QDN-G1 and QDN-14B, respectively (Fig. 2a, f). TOC contents in the two cores range from 0.5 to 1.2% (Fig. 2b, g, black lines). The FeR-OC contents in the two cores range from 0.02 to 0.17% (Fig. 2b, g, red lines) and account for 3.2–20.5% of TOC (Fig. 2d, i). The downcore patterns of both TOC and FeR-OC content profiles are not consistent with that of FeR content. The presence of low levels of dissolved Fe(II) throughout the core QDN-G1 (3–35 μM, Fig. 1b) is consistent with some background activity of dissimilatory iron reduction, independent of the biogeochemical zone. However, there is no discernable imprint on the content of FeR-OC, except in the SMTZ (Fig. 2b). Interestingly, both the lowest FeR-OC content and the lowest fFeR-OC occurred in the SMTZ of each core. In QDN-G1 and QDN-14B, the average FeR-OC contents in non-SMTZ sediments are 0.12% and 0.10%, respectively, but drop to 0.06% and 0.03% in the SMTZ sediments, respectively (Table 1). Similarly, the average fFeR-OC in the non-SMTZ layers are 14.0% and 13.0%, repectively, but drop to 12.3% and 5.2% in the SMTZ sediments, respectively (Table 1).

Fig. 2: Content and δ13C profiles of total organic carbon (TOC) and reactive iron-bound organic carbon (FeR-OC) in cores QDN-G1 (a–e) and QDN-14B (f–j).
figure 2

a, f Content of FeR. b, g Content of TOC (black lines) and FeR-OC (red lines). Separate x-axes are used for TOC and FeR-OC, labeled in the same color as the data profile. c, h Carbon isotope ratio of TOC (black lines) and FeR-OC (red lines). Separate x-axes are used for TOC and FeR-OC, labeled in the same color as the data profile. d, i The fraction of FeR-OC in TOC (fFeR-OC). e, j The molar ratio of FeR-OC and FeR. The locations of the SMTZ are highlighted by the purple bars. Source data are provided as a Source Data file.

Full size image
Table 1 Comparisons of the contents and carbon isotope ratios of TOC and FeR-OC between non-SMTZ and SMTZ sediments in two cores
Full size table

The carbon isotope ratios of TOC (δ13CTOC) and FeR-OC (δ13CFeR-OC) provide insights into the sources of the corresponding OC pool. In QDN-G1 and QDN-14B, δ13CTOC varies from −22.5 to −19.7‰, and from −25.9 to −20.2‰, respectively (Fig. 2c, h, black lines); δ13CFeR-OC varies from −31.1 to −20.0‰, and from −25.2 to −17.2‰, respectively (Fig. 2c, h, red lines). Lower δ13CTOC and δ13CFeR-OC values are observed in SMTZ of both cores (Fig. 2c, h). The average δ13CTOC values of non-SMTZ sediments are −20.6 ± 0.7‰ and −21.5 ± 0.7‰, but 1.5‰ and 1.9‰ lower in SMTZ sediments in QDN-G1 and QDN-14B, respectively (Table 1). Similarly, the average δ13CFeR-OC values of non-SMTZ sediments are −21.8 ± 1.8‰ and −20.8 ± 1.4‰, but 3.1‰ and 2.8‰ lower in SMTZ sediments in QDN-G1 and QDN-14B, respectively (Table 1). All the δ13CTOC and δ13CFeR-OC values in non-SMTZ sediments indicate the accumulation of mainly marine-algal OC, which in the sampled area has an average δ13C value of −20.8 ± 0.4‰31. However, terrestrial OC has much higher contribution in SMTZ sediments, where particulate terrestrial OC and riverine primary production have characteristic carbon isotope signatures of −24.1 ± 1.0‰ and −30.0 ± 2.6‰32.

The cumulative evidence of lower FeR-OC content, lower fFeR-OC, and lower δ13CFeR-OC in SMTZ sediments, compared to relatively uniform records in non-SMTZ sediments (significance levels in Table 1, see Methods for more information) is suggestive of dynamic organic matter cycling in this sediment horizon. While we cannot entirely rule out that these distinct signals in SMTZs of both cores are related to the depositional history and coincidentally located in this horizon, we view this scenario as highly unlikely (see Supplementary Fig. 6 and Supplementary Discussion for details). Instead, we propose that the post-depositional microbial processes account for these distinctive characteristics of FeR-OC records.

Accordingly, two observations deserve attention: (i) the distinct depletions in 13C content and contents of FeR-OC within the SMTZ, and (ii) sediments below the SMTZ show similar features with respect to 13C content and contents of FeR-OC as sediments above the SMTZ. The relatively low content of FeR-OC strongly suggests that a substantial fraction of it has been remobilized and possibly remineralized. Two mechanisms could account for the low δ13CFeR-OC of the residual fraction: (a) remobilization/degradation of a 13C-enriched, weakly bound and/or more reactive fraction33 with the residual fraction being 13C-depleted; since the δ13C of the residual fraction is broadly consistent with an origin from terrestrial plants, the residual fraction could be tightly adsorbed terrestrial organic matter supplied together with the detrital minerals from land. (b) The alternative involves again remobilization of a large fraction of FeR-OC combined with adsorption of 13C depleted dissolved organic matter (DOM) produced in the course of anaerobic oxidation of methane34; the extremely low molar ratios of FeR-OC to FeR below 1 in both SMTZs (Fig. 2e, j) indicate abundant binding sites available for DOM molecules15,35. These two mechanisms are not mutually exclusive and could act in combination. In any case, both scenarios require remobilization and possibly degradation of a substantial fraction of FeR-OC. The fact that sediments below the SMTZ resemble those above the SMTZ is at first surprising if we assume that the more deeply buried sediments were at some point in the geologic past also situated in the SMTZ. In this case, re-adsorption of DOM produced during cycling of particulate organic matter of marine origin could be the cause for the similarity of FeR-OC related signals above and below the SMTZ. However, there is also evidence that SMTZs do not “move” continuously through the sediment column with increasing sedimentation but rather “oscillate” from one horizon to another where they then remain stagnant in position for extended periods36,37.

Remineralization of FeR-OC supports microorganisms in the SMTZ

The SMTZ as a discrete sedimentary horizon, where anaerobic oxidation of methane is coupled to sulfate reduction (S-AOM)38, ubiquitously occurs in organic-rich continental shelf and slope sediments. S-AOM is performed syntrophically by sulfate-reducing bacteria and methanotrophic archaea39. In core QDN-14B, geochemical modeling reveals higher net sulfate reduction rates in the SMTZ stimulated by the methane-rich fluid in the core (Fig. 3b), which is further supported by the higher relative abundance and cell concentration of sulfate-reducing bacteria (Fig. 3e, f) and by elevated copy number of methyl-coenzyme M reductase subunit A (mcrA) gene (Fig. 3g)28, the key marker gene for both methane-producing and methane-consuming archaea40. By contrast, in core QDN-G1, there is no evidence for elevated sulfate reduction activity in the SMTZ compared to the shallower sulfate reduction zone, based on modeled rates as well as relative and absolute abundance of sulfate-reducing bacteria (Supplementary Fig. 7b, e, f). This is consistent with the relatively moderate variation in fFeR-OC record in core QDN-G1 compared to core QDN-14B. Collectively, these lines of evidence suggest that due to the particularly high methane flux at core QDN-14B, the microbial activity in the SMTZ has stimulated the remobilization of FeR-OC.

Fig. 3: Geochemical modeling and microbial evidence show that a high sulfate reduction rate is associated with low fraction of FeR-OC in TOC (fFeR-OC) in the SMTZ of QDN-14B.
figure 3

a Measured (red dots) and fitted (black line) sulfate concentrations. A 5-point Gaussian filter was applied to the concentration with the weighting on the 5 points of: [0.06, 0.24, 0.4, 0.24, 0.06]65. The measured data are from Niu et al.28 and Hu et al.60. b Modeled net reaction rate profile of sulfate. Negative values indicate net consumption, and positive values indicate net production. The 1σ envelope is shown in dashed line (See Methods for details). c Downcore record of fFeR-OC. d Downcore record of δ13CFeR-OC. e Relative abundance of Desulfobacterota in bacteria, to which most of the sulfate-reducing bacteria belong78 (data from Niu et al.28). f The cell number of Desulfobacterota, translated from the cell number of bacteria by multiplying the relative abundance of Desulfobacterota in bacteria (data from Niu et al.28). g The copy number of mcrA gene (data from Niu et al.28). The gray bar shows the position of the SMTZ. Source data are provided as a Source Data file.

Full size image

A plausible mechanism for FeR-OC remobilization in the SMTZ is the occurrence of reductive dissolution of FeR by the biogenic hydrogen sulfide produced during sulfate reduction41,42, and subsequent release of FeR-OC. Apart from chemical reduction of FeR, sulfate-reducing bacteria may also reduce FeR directly43, with the rate of reduction being enhanced by sulfate reduction and the presence of biogenic sulfide44. In addition, the process of AOM may be directly coupled with iron reduction (Fe-AOM)45 and facilitate FeR dissolution in this zone. In summary, the reduction of FeR could be enhanced by biological and chemical processes in the SMTZ, and result in the remobilization of FeR-OC.

Assuming that both TOC as well as FeR-OC lost in the SMTZ are remineralized by heterotrophic microorganisms, the remineralization fluxes of TOC and FeR-OC in the SMTZ can be estimated by the difference of their average accumulation rates between the sediment column above and within the SMTZ (see methods). In QDN-14B, the SMTZ is well defined by the depletion of sulfate, increasing DIC, and low δ13CDIC values (Fig. 1b). The TOC and FeR-OC remineralization fluxes in the SMTZ are estimated to be 15.8 and 5.8 mol m−2 kyr−1, respectively. FeR-OC remineralization contributes 36.7% to the TOC remineralization, which is disproportionally high compared to fFeR-OC (5.2 ± 1.8%) in this zone. Therefore, FeR-OC is preferentially remineralized compared to OC that is not bound to FeR (non-FeR-OC) and results in the low fFeR-OC. Stimulated by higher microbial activity, e.g., through S-AOM, Fe-AOM, and iron reduction, FeR-OC is remobilized and potentially remineralized in the SMTZ.

The estimation of the TOC and FeR-OC remineralization fluxes for core QDN-G1 is not as straightforward as for core QDN-14B due to the higher uncertainty of assigning the vertical extension of the SMTZ and the fact that contents of TOC and FeR-OC steadily decline already above the SMTZ within this core (Fig. 2b). This decline implies that in this core the remineralization fluxes are partially unrelated to processes within the SMTZ. To address this, we used an additional approach for estimating a lower flux boundary, which is only based on the four samples directly above the SMTZ as average (see methods). The resulting range of TOC and FeR-OC remineralization fluxes for core QDN-G1 are 21.6–32.8 mol m−2 kyr−1 and 3.3–5.0 mol m−2 kyr−1, repectively. FeR-OC remineralization contributes 10.1–23.1% to the TOC remineralization, which is comparable to the fFeR-OC (12.3 ± 2.6%) in this zone.

Altogether, the remineralization of FeR-OC may substantially contribute to TOC remineralization in the SMTZ. Moreover, since FeR-OC is likely enriched in labile organic matter2, it has the potential to enhance the activity and growth of heterotrophic microbial communities in subseafloor sediments and be rapidly recycled upon remobilization. Considering that the SMTZ is ubiquitous in continental slope sediments46, and assuming a FeR-OC remineralization flux of 3.3–5.8 mol m−2 kyr−1 as observed in the SMTZ of the two cores and a continental slope area of 3.01 × 107 km2 46, the global FeR-OC remineralization flux in the SMTZ of the continental slope area is estimated to be 0.10–0.17 Tmol yr−1, which is about 18-30% of the methane consumption flux in the SMTZ of continental slope sediments (0.56 Tmol yr−1 46) and 0.61% of the global sedimentary OC burial rate (13.3 Tmol yr−1 47). Using acetate as a proxy for OC, and the Gibbs energies of OC oxidation by sulfate of −81.5 KJ (per mol acetate)48 at 5 °C and 100 bars of pressure (1000 m water depth), the remineralization of FeR-OC in the SMTZ of the continental slope can provide a power supply from 4.1 × 1015 to 6.9 × 1015 J yr−1. Assuming the cellular power demand for organisms oxidizing OC with sulfate of 7.7 × 10−14 W cell−1 48, the FeR-OC remineralization in the SMTZ of continental slope area could sustain a biomass from 1.7 × 1021 to 2.8 × 1021 microbial cells. Considering generally lower power demand of 10−20 to 10−16 W cell−1 for microorganisms in marine sediments49, this amount of energy could potentially support an even larger population from 1.3 × 1024 to 2.2 × 1028 microbial cells. The average SMTZ depth in continental slope sediments is estimated to be 12.8 ± 12.1 m below seafloor46, where the cell abundance typically ranges from 106 to 108 cells cm−3 50. Assuming the SMTZ depth interval of 1 m, the volume of global continental slope SMTZ sediments reaches 3 × 1013 m3, which harbors from 3 × 1025 to 3 × 1027 cells. Therefore, the remineralization of FeR-OC in continental slope SMTZ sediments could support a substantial fraction of subseafloor microbial life in this zone.

FeR as an important reservoir of labile OC in marine sediments

The seafloor receives FeR mainly from terrestrial input27,51, which is typically associated with terrestrial OC52,53,54. After entering the ocean, additional organic matter from the pool of marine dissolved organic carbon may be bound to FeR-bearing minerals by adsorption55,56. Finally, after deposition in sediments, the upward diffusing Fe(II) can coprecipitate with marine OC into authigenic FeR after re-oxidation at the redox interface57. The binding mechanism between FeR and OC in marine sediments has been inferred from the molar ratio of FeR-OC to FeR2, with adsorption and coprecipitation characterized by ratios of <1 and 6–10, respectively2,15. In the two cores studied here, the molar ratio of FeR-OC to FeR is always larger than 1, except in the SMTZ, suggestive of a mixed binding mechanism of adsorption and coprecipitation (Fig. 2e, j). Furthermore, δ13CFeR-OC values in both sediment cores are typical for marine-sourced OC and the high molar ratio of FeR-OC to FeR indicates coprecipitation as an important binding mechanism. All these clues suggest that a major proportion of FeR-OC in marine sediments is intrinsically labile marine OC.

The average fFeR-OC in the two cores studied, except within the SMTZ in QDN-14B, is 13.3 ± 3.2%. The relatively stable fFeR-OC in the two cores suggests a similar degradation rate of FeR-OC and non-FeR-OC on the timescales represented by these cores. The only exception is the active SMTZ in QDN-14B, where iron reduction is enhanced significantly by processes likely induced by sulfate-reducing bacteria, methanotrophic archaea and/or biogenic sulfide. The stimulated degradation of FeR-OC over non-FeR-OC leads to extremely low fFeR-OC in the SMTZ of QDN-14B. As these especially active SMTZs have limited distribution in global marine sediments and occupy narrow depth intervals in sediment columns, the overall impact on the FeR-OC reservoir in subseafloor sediments is small.

Taken together, the FeR-OC records in these two cores suggest that a stable proportion of TOC survives early diagenesis as FeR-OC and is sequestered in marine sediments on at least 100-kyr timescales. Considering previous studies showing high fFeR-OC in the samples even from late Paleocene to early Eocene12, persistence of FeR-OC on even longer timescale is expected in marine sediments. This enables a rough estimation of the global FeR-OC reservoir by multiplying the TOC reservoir in global marine sediments from the Quaternary Period (0–2.59 Ma) storage48 with published fFeR-OC data in marine surface sediments. For deriving representative fFeR-OC values, all published fFeR-OC data in marine sediments using the CBD method were compiled and combined with our measured data and grouped into eight categories based on their study area information: delta and estuary, continental shelf, continental slope, deep sea, anoxic/sulfidic regions, wetland, mangrove and tephra (Fig. 4, Supplementary Table 3). One-way ANOVA and Tukey Honest Significant Difference (HSD) test revealed that significant difference (P < 0.05) in fFeR-OC values exists only between tephra-containing sediments and other environments (Supplementary Table 4). Considering TOC storage in the Quaternary sediments is estimated in the three domains shelf (water depth < 200 m), margin (200 m <water depth < 3500 m) and abyss (water depth > 3500 m)48, delta/estuary and continental shelf were combined to represent shelf, continental slope was used to represent margin and deep sea was used to represent abyss. Combining the published TOC reservoir and the corresponding surface fFeR-OC values in these three domains48, the global FeR-OC reservoir in Quaternary marine sediments was estimated to be 28,690 ± 12,165 Pg C (Supplementary Table 5), which is approximately 19–46 times the size of the atmospheric carbon pool.

Fig. 4: Fraction of FeR-OC in TOC (fFeR-OC) values of surface sediments in different marine environments.
figure 4

All the published fFeR-OC data obtained from marine sediments using the CBD method were compiled (detailed information of data cited are listed in Supplementary Table 3). The average fFeR-OC value of each study area was calculated to avoid overrepresentation of study areas with large numbers of data points (Supplementary Table 3). The average fFeR-OC values were then grouped into eight categories according to the depositonal environments of these study areas and used for statistical analyzes and box-whisker plots. Box plots indicate mean (open square), median (middle line), 25th, 75th percentile (box) and 1.5 times interquartile range (whiskers) with data points (solid dots) overlapped on top. The outlier (n = 1) in the mangrove group is not shown. Source data are provided as a Source Data file.

Full size image

Methods

Sediment core collection and age model establishment

The Qiongdongnan (QDN) basin lies in the rift belt of the northern continental margin of the South China Sea and receives terrestrial material mainly from the Indo-China Peninsula and Southeast China (especially Hainan Island) by river runoff and eolian dust58,59. In addition, the material contribution of marine authigenic sources to the basin cannot be neglected59.

Core QDN-14B was recovered from the QDN basin at a water depth of 1370 m in 2015, ~600 m east of ROV1, which is an active seep site of “Haima” cold seeps60. Core QDN-G1 was recovered outside the area of “Haima” cold seeps around 35 km away from QDN-14B at a water depth of 1478 m in 2018 (Fig. 1a). Sediment samples in core QDN-G1 were subsampled every 40 cm and preserved at −80 °C. Known volumes of sediments were taken using tip cut-off syringes and sealed in glass vials for further measurements of density and porosity. For core QDN-G1, the age model of the upper 4.0 m was based on 7 accelerator mass spectrometry 14C ages of planktonic foraminifera (G. ruber) performed at the Beta Analytic Laboratory (Florida, USA) (Supplementary Table 1). The age model of the lower part (4.0–10.8 m) was established by aligning the benthic foraminifera (C.wuellerstorfi) δ18O record to the global benthic δ18O stack LR0461 using the software QAnalySeries (v. 1.4.2) (Supplementary Fig. 1).

Porewater sampling and analysis

Porewater samples were extracted immediately onboard using Rhizon samplers (0.22 μm filter) before the core was cut open for sediment subsampling. To prevent oxidation of Fe(II), an aliquot of porewater was added to the ferrozine solution. Porewater samples for DIC measurements were preserved in pre-vacuumed glass vials. Porewater samples for ion measurements were acidified with concentrated HNO3. The rest of the porewater samples were preserved in pre-combusted amber glass vials at −20 °C for dissolved organic carbon measurements.

The Fe(II) concentration was determined by ferrozine assay62 using a spectrophotometer (DR5000, Hach, USA). The concentrations of major cations and anions were determined by ion chromatography (Dionex ICS-5000 + , Thermo Scientific, USA). Dissolved inorganic carbon (DIC) was measured using a total carbon analyzer (Multi 3100, Analytik Jena, Germany).

Physical properties of bulk sediments

Known volumes of sediments were subsampled by tip cut-off syringes and preserved in sealed glass vials for the measurements of physical properties, such as porosity and dry bulk density. All the physical properties were only measured for QDN-G1 samples. The averaged porosity and dry bulk density were used to represent the general porosities of the two cores.

Porosity was calculated by the volume ratio of water and wet sediment, where the volume of water was calculated from the difference between the wet and dry sample weight. The dry bulk density was calculated by dividing the dry sample weight by the original wet sample volume.

FeR and FeR-OC extraction

FeR and FeR-OC were extracted using the citrate-bicarbonate-dithionite (CBD) method according to Mehra and Jackson (1960)29, as modified by Lalonde et al.2 and Salvadó et al.3. Both poorly crystalline and crystalline iron (oxyhydr)oxides can be reduced by this method, including mainly goethite and hematite, which are the major iron (oxyhydr)oxides in marine sediments63,64. Briefly, samples were freeze-dried and homogenized using an agate mortar and pestle. Samples (0.5 g) were weighed carefully into 40 mL Teflon tubes and extracted in a 30 mL solution of sodium dithionite (Cat. No. 71699-50 G, Sigma-Aldrich, Germany) and trisodium citrate (Cat. No.1009408, Sinopharm, China) buffered with sodium bicarbonate (Cat. No. S837271-500G, Macklin, China) at 80 °C for 15 min in a water bath. After the extraction, the suspensions were centrifuged for 10 min at 4000 × g and then rinsed 5 times with artificial seawater. To evaluate the amount of OC remobilized during the extraction that was not bound to FeR, another sample aliquot was extracted as a control under the same experimental conditions, but replacing sodium dithionite and trisodium citrate with sodium chloride (Cat. No. 10019308, Sinopharm, China) with equivalent ion strength. The residuals were dried overnight in a 50 °C oven, carefully weighed and manually ground.

The supernatant and rinse water were combined, acidified to pH < 2 and filtered through 0.22 μm filters. The dissolved iron was determined using a ferrozine assay62. The iron extracted in the control group were negligible; therefore, the amount of FeR was calculated based on the iron extracted in the dithionite reduction group without further correction.

FeR-OC quantification and carbon isotope analyzes

The OC content and carbon isotope ratio were determined for both untreated and treated samples using an elemental analyzer (Vario EL III, Elementar) coupled to an isotope ratio mass spectrometer (Isoprime, Elementar) at the instrumental analysis centre, Shanghai Jiao Tong University. The inorganic carbon was removed by acid fumigation before the analysis. The samples were measured in triplicate, and the standard deviation was <0.05% for TOC and <0.2‰ for δ13C. The analytical precision was <0.06% for TOC (standard deviation for repeated measurements of the low organic content soil standard; n = 3) and <0.09‰ for δ13C (standard deviation for repeated measurements of the USGS40 standard, n = 3).

Calculation of fFeR-OC and δ13CFeR-OC

The fraction of FeR-OC in TOC (fFeR-OC) was determined as the difference of TOC amount in the residuals between the CBD extraction and control extraction devided by the TOC amount in the untreated sample.

$${{{\rm{fF}}}}{{{{\rm{e}}}}}_{{{{\rm{R}}}}}\!{\mbox{-}}{{{\rm{OC}}}}=({{{{\rm{TOC}}}}}_{{{{\rm{control}}}}}{\mbox{-}}{{{{\rm{TOC}}}}}_{{{{\rm{CBD}}}}})/{{{{\rm{TOC}}}}}_{{{{\rm{bulk}}}}}$$
(1)

in which,

TOCcontrol = total organic carbon (mg) left after the control extraction,

TOCCBD = total organic carbon (mg) left after the CBD extraction,

TOCbulk = total organic carbon (mg) in the untreated sample.

A binary mixing model was used to calculate the carbon isotope ratio of FeR-OC, where the OC pool in the residual of the control group (TOCcontrol) is comprised of FeR-OC and OC remaining untouched after CBD extraction (TOCCBD). The carbon isotope ratios of FeR-OC was calculated according to the following mass-balanced equations:

$${{{{\rm{TOC}}}}}_{{{{\rm{CBD}}}}}+{{{\rm{F}}}}{{{{\rm{e}}}}}_{{{{\rm{R}}}}}{\mbox{-}}{{{\rm{OC}}}}={{{{\rm{TOC}}}}}_{{{{\rm{control}}}}}$$
(2)
$$f{{{{\rm{TOC}}}}}_{{{{\rm{CBD}}}}}\times {{{{\rm{\delta }}}}}^{13}{{{{\rm{C}}}}}_{{{{\rm{CBD}}}}}+f{{{\rm{F}}}}{{{{\rm{e}}}}}_{{{{\rm{R}}}}}{\mbox{-}}{{{\rm{OC}}}}\times {{{{\rm{\delta }}}}}^{13}{{{{\rm{C}}}}}_{{{{\rm{FeR}}}}{\mbox{-}}{{{\rm{OC}}}}}={{{{\rm{\delta }}}}}^{13}{{{{\rm{C}}}}}_{{{{\rm{control}}}}}$$
(3)

in which,

FeR-OC = the amount of organic carbon (mg) bound to FeR,

TOCCBD = total organic carbon (mg) left after the CBD extraction,

TOCcontrol = total organic carbon (mg) left after the control extraction,

fTOCCBD = TOCCBD/TOCcontrol,

fFeR-OC = FeR-OC/TOCcontrol,

δ13CCBD = the measured carbon isotope signature of TOCCBD,

δ13Ccontrol = the measured carbon isotope signature of TOCcontrol,

δ13CFeR-OC = the carbon isotope signature of FeR-OC.

Geochemical modeling

Net reaction rates of sulfate were estimated using the MATLAB (2021b) script published in Wang et al.65, considering molecular diffusion, sediment burial, fluid advection, and reaction under the assumption that the sulfate concentration profiles represent steady-state conditions.

The mass balance of sulfate is expressed as:

$$-\frac{\partial }{\partial {{{\rm{x}}}}}\left\{-\frac{D\left(x\right)}{{\theta }^{2}\left(x\right)}\frac{\partial \left[{{\varnothing }}\left(x\right)C\left(x\right)\right]}{\partial x}+\left[{{\varnothing }}\left(x\right)b\left(x\right)+{{\varnothing }}\left(x\right)v\left(x\right)\right]C\left(x\right)\right\}+R(x)=0$$
(4)

Where \(C\left(x\right)\) is sulfate concentration in the porewater, x is depth below the seafloor, \(\varnothing \left(x\right)\) is porosity, \(D\left(x\right)\) is the molecular diffusion coefficient of sulfate, \({\theta }^{2}\left(x\right)\) is tortuosity, \(b\left(x\right)\) is porewater burial velocity, \(v\left(x\right)\) is externally driven flow velocity, \(R(x)\) is the rate of diagenetic reaction per unit volume of sediment65.

The model applied a 5-point Gaussian filter to the sulfate concentration profile, with the weighting on the 5 points of: [0.06, 0.24, 0.4, 0.24, 0.06]. All parameters used for modeling are listed in Supplementary Table 2. Porosity profile was measured using QDN-G1 samples. The averaged porosity was applied as the constant porosity value for both cores assuming similar sediment properties in the study area. Diffusion coefficients of sulfate in the two cores were obtained using R package marelac66 with the input of respective salinity, temperature and pressure values (Supplementary Table 2). The tortuosity \({\theta }^{2}\) was calculated by the relation, \({\theta }^{2}=\varnothing f\), where \(f\) is the formation factor, which can be calculated based on the empirical relationship \(f=1.3\times {\varnothing }^{-1.45}\)67,68. We used the sedimentation rate near the seafloor 1.2 × 10−4 m yr−1 as the porewater burial velocity for both cores, assuming a similar sedimentation rate in the study area, which is calculated by the radiocarbon dating ages of the first two sampling layers below the seafloor in QDN-G1. A constant external flow advection velocity near the sediment-water interface of 10−5 m yr−1 was used65. A minimum of three measured concentration data points were used to determine each reaction zone65. The model quantifies uncertainties in the rate estimates by using a Monte Carlo technique65. The depth-integrated net reaction rate yielded the total flux at steady state for the sediment column.

Estimation of TOC and FeR-OC remineralization fluxes in the SMTZ

The remineralization fluxes of TOC and FeR-OC in the SMTZ can be estimated by the difference of their average accumulation rates between the sediment column above and within the SMTZ, assuming that all the TOC and FeR-OC lost in the SMTZ is remineralized by heterotrophic microorganisms as opposed to mobilized and moved into another organic carbon pool.

The accumulation rates of TOC and FeR-OC are calculated by multiplying their contents in sediments by the sediment mass accumulation rate69:

$${{{\rm{TOC}}}}\; {{{\rm{accumulation}}}}\; {{{\rm{rate}}}}={{{\rm{LSR}}}}\times {{{\rm{DBD}}}}\times {{{\rm{TOC}}}}\%$$
(5)
$${{{\rm{F}}}}{{{{\rm{e}}}}}_{{{{\rm{R}}}}}{\mbox{-}}{{{\rm{OC}}}}\; {{{\rm{accumulation}}}}\; {{{\rm{rate}}}}={{{\rm{LSR}}}}\times {{{\rm{DBD}}}}\times {{{\rm{F}}}}{{{{\rm{e}}}}}_{{{{\rm{R}}}}}-{{{\rm{OC}}}}\%$$
(6)

where DBD is the dry sediment bulk density in g cm−3, LSR is the linear sedimentation rate in cm kyr−1, TOC% and FeR-OC% are the weight percentage of TOC and FeR-OC in sediments, respectively. The TOC and FeR-OC accumulation rates are expressed in mol m−2 kyr−1. The average sedimentation rate (11.1 cm kyr−1) and the average dry bulk density (0.9 g cm−3) of QDN-G1 was used for the calculation for core QDN-14B, assuming similar sedimentation rate and bulk sediment properties in the study area.

When assigning the background levels of TOC and FeR-OC accumulation rates, which are independent of remineralization within the SMTZ, the average accumulation rate in the sediment column above the SMTZ was used. As TOC and FeR-OC contents in core QDN-G1 begin to steadily decline above the SMTZ (Fig. 2b), this approach may lead to an overestimation of remineralization fluxes within the SMTZ. Therefore, for core QDN-G1, we estimated a lower boundary of background TOC and FeR-OC accumulation rates, represented by the four samples above the SMTZ (560–720 cmbsf) as average.

DNA extraction, qPCR and amplicon sequencing

The 16 S rRNA gene data was obtained for core QDN-G1 according to the following procedures.

The DNA for qPCR and V4 region of 16 S rRNA gene sequencing was extracted from ~0.25 g of sediments using DNeasy® PowerSoil® Pro Kit (Cat. No. 47014, Qiagen, Germany), according to the manufacturer’s instructions.

The extracted DNA was used as template for qPCR to determine the abundance of bacterial 16 S rRNA gene with the primer set 331 F (5ʹ-TCCTACGGGAGGCAGCAGT-3ʹ) /797 R (5ʹ-GGACTACCAGGGTATCTAATCCTGTT-3ʹ)70. Standard curves were constructed using a 10-fold series dilution of the plasmids for six gradients carrying the bacteria 16 S rRNA gene. qPCR was carried out in a volume of 20 μL, including 10 μL 2 × PowerUp™ SYBR™ Green Master Mix (Cat. No. A25742, Thermo Scientific, USA), 1.6 μL each primer (10 μM), 2 μL template DNA and 4.8 μL steriled deionized water. The qPCR program consisted of an initial cycle of 95 °C for 5 min; 40 cycles of 95 °C for 30 s, 60 °C for 30 s, 72 °C for 30 s, 80 °C for 10 s, and the data was collected at the final step of each cycle. The melting curve was generated using default program. All samples were subject to qPCR measurement with three technical replicates. The bacterial cell number was evaluated based on the abundance of 16 S rRNA gene, applying the average copy number of the 16 S rRNA gene on genomes of Bacteria (5.3 copies/genome, rrnDB version 5.8)71.

The V4 region of 16 S rRNA gene was amplified using the primer set 515 F (5ʹ-GTGYCAGCMGCCGCGGTAA-3ʹ) /806 R (5ʹ-GGACTACNVGGGTWTCTAAT-3ʹ)72. DNA was amplified using the following cycling conditions: 95 °C, 5 min; 30 cycles (95 °C, 30 s; 50 °C, 30 s; 72 °C, 30 s); 72 °C, 7 min. The PCR products of samples were sent to Shanghai Personal Biotechnology Co., Ltd. (Shanghai, China) for high-throughput sequencing of the 16 S rRNA gene using the Illumina Novaseq PE250 platform.

Sequence analysis

The raw reads of 16 S rRNA gene of both cores QDN-G1 and QDN-14B were processed and analyzed using the QIIME 2 platform (version 2020.11)73. The primers and adapters were first trimmed out using Cutadapt (version 3.1)74. Raw sequences were then processed using DADA275, including quality filtering, denoising, paired-end sequence merging, chimera filtering and producing amplicon sequence variants (ASVs) and ASV Table. Taxonomy was assigned using q2-feature-classifier (a scikit-learn naive Bayes machine-learning classifier)76 with Silva database release 13877. Multiple sequence alignment and phylogenetic tree construction were performed using the QIIME 2 plugin q2-phylogeny (align-to-tree-mafft-iqtree). Unassigned sequences, singletions and sequences affiliated with eukaryotes were discarded. Eventually, to eliminate uneven sequencing depths, the ASV table was rarefied to 14935 and 71773 sequences per sample for QDN-14B and QDN-G1, respectively, determined by the sample with the fewest sequences (Supplementary Data 1 and 2).

Statistical analyzes

Statistical analyzes were carried out in R (version 4.1.3). The Wilcoxon rank sum test was applied to compare TOC and FeR-OC records between SMTZ and non-SMTZ sediments in two sediment cores studied, including TOC content, FeR-OC content, fFeR-OC, δ13CTOC, and δ13CFeR-OC.

One-way analysis of variance (ANOVA) was used to determine if there are significant differences in fFeR-OC in surface sediments of different marine environments. Tukey Honest Significant Difference (HSD) test was used for performing multiple pairwise-comparison between the environments to determine which among them were significantly different.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.