Wetlands store approximately one-third of global soil organic carbon (SOC) and play important roles in regulating and stabilizing global climate1,2,3. CH4 and CO2, the two greatest contributors to the greenhouse effect, are the dominant gaseous end-products from the mineralization of organic carbon in wetland ecosystems4,5. Carbon fluxes in a wetland ecosystem are closely linked to its hydrological features6. Climate changes may introduce specific hydrology-related stressors. For instance, freshwater wetlands are increasingly vulnerable to drought events, which lower water tables and introduce O2 into wetland soils. This leads to more frequent exposure of soil organic matter to O2, thereby affecting the organic carbon balance7. In addition, freshwater wetlands face perturbations as a result of seawater intrusions. As the sea level rises, the inundation of wetlands by seawater brings a high concentration of sulfate ions (SO42-), substantially altering the sediment chemistry2,8,9. These stressors can change the production patterns of CH4 and CO2, resulting in feedback that is poorly characterized qualitatively in terms of the direction of the changes and quantitatively in terms of the magnitudes of the changes. Understanding how environmental disturbances affect the dynamic of metabolic processes and the succession of ecological communities is critical for accurately modeling changes in greenhouse gas emissions from wetland ecosystems in future climate scenarios.

Microbial interactions are vital in modulating organic matter decomposition and greenhouse gas emissions within wetland ecosystems10,11. These interactions are driven by fermentation and respiration, coupling the oxidation of organic compounds by electron transfer to electron acceptors12,13,14. In the absence of disturbance by droughts and seawater intrusions, obligatory anaerobic fermenters become dominant15. Syntrophic interactions among these diverse anaerobic microorganisms facilitate the stepwise breakdown of complex organic substrates into simpler, chemically stable compounds, ultimately resulting in the production of CH4 and CO216. However, O2 exposure introduced during drought can promote aerobic microbial respiration of carbon stocks, with CO2 as the primary product17. The introduction of SO42- from seawater inundation as an alternative electronic acceptor leads to sulfate-reducing bacteria (SRB) to producing more H2S, utilizing H2 and/ or organic acids as electron donors. The competition from SRBs for H2 and organic acids may reduce methane production by methanogens18,19. Characterizing these metabolic interactions provides insight into biochemical transformations within the community under changing redox conditions, and allows for investigations into how changes in these conditions, induced by climate change, will affect the microbial metabolic networks that control carbon cycling in wetland ecosystems.

Carbon cycling and gas emissions in wetlands have been studied extensively in both field and laboratory conditions20,21,22,23. Elevated CO2 levels, reduced carbon accumulation, and decreased CH4 emissions have been observed when introducing O2 into previously anoxic wetland soil22,24. Previous studies have also demonstrated lower CH4 emissions to the atmosphere from lacustrine wetlands following seawater intrusion [21-23]. However, few studies have characterized how the critical syntrophic interactions controlling CH4 and CO2 productions respond to environmental redox perturbations, such as increased O2 from periods of drought or SO42- from seawater intrusion.

Establishing a clear connection between environmental disturbance and microbial adaptations remains a challenge, due to the complex and dynamic nature of microbial communities25,26. In this study, we set up laboratory microcosms to investigate the effects of SO42- exposure and O2 exposure alone or in combination on microbial activities and interactions, as well as the resulting fate of carbon within wetland soil (Fig. 1). We used proteogenomics to characterize the biochemical and physiological responses of microbial communities to individual perturbations and their combined effects. Stoichiometric models were employed to deconvolute carbon exchanges among the main functional guilds. By integrating geochemical, metaproteomic, and stoichiometric analyses, we aimed to develop a molecular understanding of how climate change-induced stressors modulate wetland microbial communities and greenhouse gas fluxes.

Fig. 1: Schematic diagram of study design.
figure 1

Emissions of methane and carbon dioxide from wetland ecosystems can be severely perturbed by climate-change-induced stressors, such as seawater intrusions and droughts. To investigate this, we constructed laboratory microcosms using freshwater wetland soils under four incubation conditions, including anoxic non-sulfate addition (S-O-), anoxic sulfate-addition (S+O-), oxic non-sulfate addition (S-O+) and oxic sulfate-addition (S+O+) conditions. The soil communities were characterized by integrating geochemical analysis, metagenomics, metaproteomics and stoichiometric modeling. The results uncovered the molecular mechanism on how wetland microbial communities modulate greenhouse gas fluxes under future climate scenarios.


SO4 2- and O2 additions changed community structure and activities

Laboratory microcosms were constructed using lacustrine wetland soils and incubated under four conditions: anoxic non-sulfate-addition (S-O-), anoxic sulfate-addition (S+O-), oxic non-sulfate-addition (S-O+ ), and oxic sulfate-addition (S+O+ ) (Fig. 1). Sterile air was introduced under S-O+ and S+O+ conditions. The final concentration of SO42- following addition under the S+O- and S+O+ conditions was approximately 7.2 mM, falling within the range of natural seawater concentration27.

Metaproteomic analysis identified a total of 10598 proteins based on assignments of unique peptide identifications to individual proteins. These proteins originated from 807 microbial species and 50 phyla, with Pseudomonadota (~51% of total proteome abundance) and Euryarchaeota (~33% of total proteome abundance) being the dominant phyla (Fig. 2a and Supplementary Table 2). All experimental treatments resulted in an increase of Pseudomonadota in proteome abundance compared to the S-O- condition. However, the increase under dual exposure (S+O+ , 82%, p-value < 0.000, t-test) was more than the additive effect of the increase observed under the S+O- condition (45%, p-value = 0.001, t-test) and S-O+ condition (7%, p-value = 0.5, t-test). This suggests a synergistic interaction between the effects of SO42- and O2 on the proteome abundance of Pseudomonadota, where their combined impact on microbial communities is greater than the sum of their individual effects. Conversely, Euryarchaeota consistently decreased in proteome abundance upon stress exposure, although this reduction was less under dual exposure (S+ O+ , 52%, p-value = 0.0006, t-test) than the sum of the decreases under S+O- condition (29%, p-value = 0.008, t-test) and S-O+ condition (41%, p-value = 0.001, t-test) (Fig. 2a). This suggests an antagonistic interaction between the effects of SO42- and O2 on the proteome abundance of Euryarchaeota, where their combined impact on microbial communities is less than the sum of their individual effects.

Fig. 2: Effects of SO42- and O2 exposures on microbial community composition and activities.
figure 2

a Relative proteomic abundance of microbial phyla under four treatment scenarios. The relative proteomic abundance is expressed as the proportion of total proteins attributable to each microbial species. b, c Community species diversity indices, with b showing the richness and c the evenness across the four experimental conditions. Significant differences with p-values determined using Student’s t-test and adjusted for the false discovery rate, are marked by * for p-value < 0.05, ** for p-value < 0.01 and *** for p-value < 0.001. d Correlation between microbial species diversity and the accumulations of CH4 and CO2, where each dot represents an experimental measurement (X-axis for geochemical measurements, Y-axis for metaproteomic measurements) from the four conditions with three replicates, and the red line displays best linear fit. r is the Pearson correlation coefficient, and p values were determined by t-statistic. e, f Cumulative amounts of fermentation products are measured by carbon molarity e and electron molarity f within the microcosms. The error bars are defined as standard deviation.

Compared to the S-O- condition, the metaproteome abundances of active communities showed increased richness under both the S+O- (28%, p-value = 0.02, t-test) and S-O+ conditions (36%, p-value = 0.0009, t-test). There was also an increase in evenness under S+O- (5%, p-value = 0.04, t-test) condition, S-O+ condition (8%, p-value < 0.002, t-test), and S+O+ condition (9%, p-value = 0.001, t-test) (Fig. 2b, c). This suggests that SO42- and O2 exposures contributed to the proliferation of low-abundance microbial species. Moreover, we found a negative correlation between the increased richness and evenness of the active community and the accumulation of CH4 and a positive correlation with CO2 accumulation (Fig. 2d).

The primary metabolic products in the microcosm system were quantified by carbon (Fig. 2e) and electron molarity (Fig. 2f). Under the S-O- condition, the soil carbon was fermented into 0.32 ± 0.03 mmole/g dry soil of organic acids, 0.44 ± 0.02 mmole/g dry soil of CO2, and 0.29 ± 0.01 mmole/g dry soil of CH4 over seven days of incubation. Compared to S-O-, the SO42- addition increased the carbon transfer to organic acids by 94% (p-value = 0.0003, t-test) and decreased the CH4 production by 73% (p-value < 0.0001, t-test), and the O2 exposure decreased the carbon transfer to organic acids by 76% (p-value = 0.0002, t-test), increased that to CO2 by 70% (p-value < 0.0001, t-test), and eliminated the CH4 production. Under the S-O- condition, 1.36 ± 0.13 mmole/g dry soil of electrons was transferred to organic acids, 0.19 ± 0.01 mmole/g dry soil to H2, and 2.32 ± 0.63 mmole/g dry soil to CH4. In comparison, the SO42- addition diverted 1.44 ± 0.07 mmole/g dry soil of electrons to H2S and decreased the electron transfer to H2 by 40% (p-value = 0.0004, t-test), and the O2 exposure decreased the electron transfer to H2 by 55% (p-value = 0.001, t-test). The effects of the combined exposure to both SO42- and O2 on the carbon and electron transfers were similar to the effects of the O2 exposure only.

SO4 2- and O2 additions altered depolymerization of plant polysaccharides

The formation of hydrolysis products derived from plant polysaccharides, such as glucose, xylose, mannose, and galacturonic acid, was affected by exposure to SO42- and O2 over time (Supplementary Fig. 1a1d). Under both anoxic and oxic conditions, the addition of SO42- slowed down the accumulation of glucose. This resulted in a 12% lower accumulation under anoxic conditions (p-value = 0.04, t-test) and a 14% reduced glucose accumulation (p-value = 0.02, t-test) under oxic conditions (Supplementary Fig. 1). Xylose reached its peak accumulation later under anoxic conditions compared to oxic conditions. The highest accumulation of xylose was observed under the S-O- condition, while its accumulation was relatively similar under the other three conditions. Mannose and galacturonic acid continuously accumulated under anoxic conditions. Their accumulation peaked between 16 to 23 days and then decreased under oxic conditions.

The metaproteomic analysis identified five key enzymes from species involved in the degradation of plant polysaccharides. These species had the highest proteome abundance under anoxic conditions. Pairwise comparisons of proteome richness and evenness of these species, between the S-O- condition and the S-O+ condition, as well as between the S+O+ condition and the S+O+ condition, revealed no significant changes under different oxic conditions, which suggests that the actual species composition remained relatively stable across four conditions. A strong negative correlation (r = -0.70, p-value = 0.01, t-test) was observed between the total proteome abundance of plant polysaccharide-degrading microbes and the accumulation of CO2 (Fig. 3a). Conversely, a weak positive correlation was observed between the total proteome abundance of these microbes and the accumulation of CH4 (Fig. 3a).

Fig. 3: Effects of SO42- and O2 exposures on plant polysaccharide degradation.
figure 3

a Correlation analysis between the relative abundances of plant polymer degradation species and CAZymes with the accumulation of CH4 and CO2, where each dot represents experimental measurements (X-axis for geochemical measurements, Y-axis for metaproteomic measurements) from four conditions with three replicates, and the red line is the result of linear fitting. r is the Pearson correlation coefficient, and p values were determined by t-test. b Protein abundances of CAZyme under the four conditions. The error bars are defined as standard deviation.

The identified carbohydrate-active enzymes (CAZymes), including pectate lyase, beta-glucosidase, xylosidase, beta-galactosidase, and starch phosphorylase, showed the highest total abundance under the S-O- condition, and the lowest abundance under the S+O+ condition (Fig. 3b). This indicates that elevated SO42- and O2 levels suppressed the activities of plant polysaccharide degradation.

Among the identified CAZymes, beta-glucosidase was detected across all conditions, with the highest abundance observed under the S+O- condition (fold change > 4.0, q-value < 0.007). Xylosidase, involved in xylan degradation, showed higher abundance under the S+O- condition (fold change > 3.2, q-value < 0.0001) (Supplementary Table 3), indicating enhanced cellulose and xylan degradation capacity under this condition. Beta-galactosidase was only detected under the S+O+ condition (Supplementary Table 3). This enzyme can release galactose from hemicellulose, which enters the glycolysis pathway after conversion into glucose 1-phosphate by glucuronate isomerase (UXA). The highest abundance of UXA proteins was detected under the S+O+ condition (fold change > 1.7, q-value < 0.0001) (Supplementary Table 3), suggesting enhanced hemicellulose utilization under this condition. These observations collectively demonstrate that changes in SO42- and O2 levels altered the preference of the microbial community in degrading plant polysaccharides. Specifically, the enzymes enriched under oxic conditions use more complex structural substrates (e.g., hemicellulose) compared to those enriched under anoxic conditions (e.g., cellulose).

SO4 2- and O2 decreased methanogenesis and promoted methane oxidation

Under the S-O- condition, the microbial community continuously produced CH4 (0.29 ± 0.007 mmole/g dry soil) and CO2 (0.44 ± 0.02 mmole/g dry soil) throughout the experiment (Fig. 4a, b). The two-way ANOVA indicated a significant interaction between SO42- and O2 on CH4 accumulation (F (1,8) = 1590, p-value < 0.0001). CH4 release was reduced by 73% (p-value < 0.0001, t-test) under the S+O- condition, and by 97% (p-value < 0.000, t-test) under the S-O+ condition (Fig. 4a and Table 1), highlighting the higher sensitivity of CH4 accumulation to O2 than to SO42-. However, the same analysis showed no significant interaction between SO42- and O2 (F (1, 8) = 0.1, p-value = 0.75). The main effects were observed, CO2 accumulation increased by 11% higher under the S+O- condition (p-value = 0.03, t-test), by 70% higher under the S-O+ condition (p-value < 0.0001, t-test), and by 83% higher under the S+O+ condition (p-value = 0.0002, t-test), compared to the S-O- condition (Fig. 4b), suggesting an additive effect of SO42- and O2 on CO2 accumulations.

Fig. 4: Modulation of methane metabolism by SO42- and O2 exposures.
figure 4

a, b Changes in molar mass of CH4 a and CO2 b under the four conditions, over a 30-day incubation. ce Comparison of methanogens and methanotrophs in terms of relative abundance c, richness d, and evenness e under the four conditions. Relative abundance is quantified as the aggregate of protein abundances attributable to a specific species. f Schematic representation of methane metabolic pathways. The boxes in the pathway map, arranged from left to right, correspond to the S-O-, S+O-, S-O+ and S+O+ conditions. The color gradient from red to blue represents the relative abundance ranking of proteins among the four conditions, with red indicating the highest ranked relative abundance and blue the lowest. The gray color indicates proteins not detected by LC-MS/MS. Significant differences with p-values determined using Student’s t-test, and adjusted for the false discovery rate, are marked by * for p-value < 0.05, ** for p-value < 0.01 and *** for p-value < 0.001. The error bars are defined as standard deviation.

Table 1 Accumulation of metabolic products in the microcosms under the four incubation conditions

Metaproteomic analysis identified 634 proteins from methanogens (Supplementary Table 1). These methanogens showed the highest proteome abundance under the S-O- condition, constituting 22% of the total metaproteome (Supplementary Table 1). The addition of SO42- and O2, either individually or in combination, led to a reduction in the proteome abundance (p-value < 0.0001, one-way ANOVA), richness (p-value = 0.005, one-way ANOVA), and evenness (p-value = 0.008, one-way ANOVA) of methanogenic populations (Fig. 4c–e). Furthermore, strong positive correlations were observed between the CH4 accumulation and the proteome abundance (r = 0.9, p-value < 0.0001, t-test), richness (r = 0.9, p-value = 0.0001, t-test), as well as evenness (r = 0.8, p-value = 0.002, t-test) of methanogens (Supplementary Fig. 2). In summary, the observed relationships among SO42- and O2 availability, CH4 accumulation, and the composition of methanogen populations could serve as diagnostics for more predictive climate modeling.

Metaproteomics revealed significant suppression of both hydrogenotrophic and acetoclastic methanogenesis by SO42- and O2. Many marker proteins involved in methanogenesis, including methyl-coenzyme M reductase (Mcr), 5,10-methylenetetrahydromethanopterin reductase (Mer), tetrahydromethanopterin S-methyltransferase (Mtr), acetyl-CoA decarbonylase/synthetase (CdhA) and methylenetetrahydromethanopterin dehydrogenase (Mtd) were most abundant under S-O- condition (Fig. 4f and Table 2). The addition of SO42- and O2, either individually or in combination, led to a significant decrease in the abundance of these proteins (Fig. 4f). This decrease coincided with the decrease in CH4 emissions observed when exposed to SO42- and O2 stressors (Fig. 4a).

Table 2 Key proteins in methane metabolism and sulfate reduction pathways

Furthermore, the SO42- and O2 exposures suppressed hydrogenotrophic methanogenesis much more than acetoclastic methanogenesis. Specifically, after SO42- addition to the anoxic sediment, the abundance of acetyl-CoA decarbonylase/synthetase essential for acetoclastic methanogenesis, decreased by 1.2-fold (q-value = 0.0001). And there was a more pronounced decrease in proteins for hydrogenotrophic methanogenesis (Fig. 4f, Table 2), including a 3.4-fold decrease of the 5,10-methylenetetrahydromethanopterin reductase (q-value = 0.003) and a 5.8-fold decrease of the tetrahydromethanopterin S-methyltransferase (q-value < 0.0001). This indicated a stronger suppression of hydrogenotrophic methanogenesis by SO42-. Under S-O+ condition, we observed a decrease in proteins for both the acetoclastic and hydrogenotrophic pathways of methanogenesis when compared to S-O- condition. In the acetoclastic pathway, there was a 2.6-fold decrease in acetyl-CoA synthetase (Acs) (q-value = 0.01) and a 2.1-fold decrease in methyl-coenzyme M reductase (q-value < 0.0001) (Fig. 4f, Table 2). In the hydrogenotrophic pathway, there was a 7.0-fold decrease in the methylenetetrahydromethanopterin dehydrogenase (q-value < 0.0001) and a 3.1-fold decrease in methyl-coenzyme M reductase (q-value = 0.02). Similarly, under S+O+ condition, compared to the S+O- condition, there was a more substantial decrease in key enzymes associated with hydrogenotrophic methanogenesis, including a 9.2-fold decrease in 5,10-methylenetetrahydromethanopterin reductase (q-value < 0.0001) and a 5.0-fold decrease in tetrahydromethanopterin S-methyltransferase (q-value < 0.01), compared to a 1.5-fold decrease in acetyl-CoA decarbonylase/synthetase (q-value = 0.006). These results highlight that O2 had a greater suppression of hydrogenotrophic methanogenesis than acetoclastic methanogenesis. Moreover, compared to the S-O- condition, under S+O+ condition, the abundance of methyl-coenzyme M reductase decreased by 1.7-fold (q-value = 0.01), and the abundance of tetrahydromethanopterin S-methyltransferase decreased by 8.7-fold (q-value < 0.0001). This suggests that the combined addition of SO42- and O2 had a more significant impact on hydrogenotrophic methanogenesis compared to acetoclastic methanogenesis.

In addition, metaproteomics identified 208 proteins from methanotrophs (Supplementary Table 1). Under the S+O+ condition, methanotrophs showed the highest richness (p-value = 0.0001, one-way ANOVA), and evenness (p-value = 0.0004, one-way ANOVA) (Fig. 4c–e). Moreover, significant negative correlations were found among the CH4 accumulation and proteome abundance (r = -0.7, p-value = 0.02, t-test), richness (r = -0.9, p-value < 0.0001, t-test), and evenness (r = -0.8, p-value = 0.002, t-test) of active methanotrophs (Supplementary Fig. 2).

Key enzymes involved in methane oxidation, including methane/ammonia monooxygenase (Pmo) and lanthanide-dependent methanol dehydrogenase (XoxF), showed distinct responses to elevated SO42- and O2 (Fig. 4f and Table 2). Under the anoxic condition, SO42- addition increased the abundance of methane/ammonia monooxygenase and lanthanide-dependent methanol dehydrogenase by 2.3-fold (q-value < 0.0001) and 1.4-fold (q-value = 0.0002), respectively. However, under the oxic conditions, SO42- addition decreased the abundance of lanthanide-dependent methanol dehydrogenase by 1.7-fold (q-value = 0.005). Oxygen addition led to an increase in the abundance of methane/ammonia monooxygenase and lanthanide-dependent methanol dehydrogenase by 3.9-fold (q-value < 0.0001) and 2.5-fold (q-value = 0.006), respectively, under the non-sulfate-addition condition. Under the sulfate-addition condition, their abundance increased by 1.7-fold (q-value = 0.005) and 3.0-fold (q-value < 0.0001), respectively. Furthermore, when compared to S-O- condition, the abundance of methane/ammonia monooxygenase and lanthanide-dependent methanol dehydrogenase increased by 4.2-fold (q-value < 0.0001) and by 3.9-fold (q-value = 0.001) under S+O+ condition (Fig. 4f, Table 2).

SO4 2- and O2 additions changed metabolic behaviors of sulfate-reducing bacteria

Under the sulfate-limited conditions, some sulfate-reducing bacteria (SRB) function as syntrophic oxidizers, metabolizing lactate to produce acetate, H2, and CO2. With added SO42-, SRB may oxidize lactate to acetate and CO2 and concurrently reduce SO42- to H2S28. Two-way ANOVA indicated a significant interaction between SO42- and O2 on lactate accumulation (F (1,8) = 66.51, p-value < 0.0001), with significant main effects of SO42- (F (1, 8) = 55.4, p-value < 0.0001) and O2 (F (1, 8) = 74.76, p-value < 0.0001). Under the S-O- condition, lactate substantially accumulated, reaching its peak of 0.033 ± 0.004 mmole/g dry soil on day 9, and then decreased to 0.018 ± 0.002 mmole/g dry soil over the next 7 days. Under the S+O- condition, lactate accumulation continually increased, reaching 0.037 ± 0.004 mmole/g dry soil at the end of incubation. Under oxic conditions, lactate accumulation remained consistently low at less than 0.012 ± 0.004 mmole/g dry soil over 30 days (Fig. 5a). Similarly, the two-way ANOVA for H2S accumulation revealed a significant interaction between SO42- and O2 (F (1, 8) = 1222, p-value < 0.0001), with significant main effects for both SO42- and O2. Specifically, a substantial accumulation of H2S (0.18 ± 0.008 mmole/g dry soil) was observed under the S+O- condition, while only trace amounts of H2S (0.003 ± 0.0002 mmole/g dry soil) were detected under the S+O+ condition (Fig. 5b).

Fig. 5: Effects of SO42- and O2 exposures on sulfate reduction and sulfite/sulfide oxidation processes.
figure 5

a, b Changes in molar mass of lactate a and H2S b over the 30 days of incubation. ce Relative abundances c, richness d, and evenness e of sulfate-reducing bacteria (SRB) and sulfide/sulfur-oxidizing bacteria (SOB). Significant differences with p-values determined using Student’s t-test, and adjusted for the false discovery rate, are marked by * for p-value < 0.05, ** for p-value < 0.01 and *** for p-value < 0.001. f Sulfur metabolism pathways. The boxes in the pathway map, arranged from left to right, correspond to the S-O-, S+O-, S-O+ and S+O+ conditions. The color gradient from red to blue represents the relative abundance ranking of proteins among the four conditions, with red indicating the highest ranked relative abundance and blue the lowest. The gray color indicates proteins not detected by LC-MS/MS. The error bars are defined as standard deviation.

Metaproteomics identified a total of 145 proteins from SRB species, with the highest SRB proteome abundance observed under S+O- condition (fold change > 1.3, p-value < 0.006, t-test) (Fig. 5c, Supplementary Table 1). SO42- addition enhanced the activities of both lactate oxidation and dissimilatory sulfate reduction within this community (Fig. 5d–f and Table 2). In the lactate oxidation pathway, SO42- addition resulted in an 8.0-fold increase in the abundance of pyruvate ferredoxin oxidoreductase (Por) (q-value < 0.0001) and a 5.3-fold increase in acetyl-CoA synthetase (q-value < 0.0001). In the dissimilatory sulfate reduction pathway, the abundance of adenylylsulfate reductase (Apr) increased by 1.8-fold (q-value < 0.0001), sulfate adenylyltransferase (Sat) increased by 1.4-fold (q-value = 0.002), and dissimilatory sulfate reductase (Dsr) increased by 1.5-fold (q-value = 0.005) following SO42- addition (Fig. 5f, Table 2).

Oxygen addition significantly suppressed the activities of lactate oxidation and dissimilatory sulfate reduction. The response of key proteins in both pathways to O2 varied with the availability of SO42-. In the lactate oxidation pathway, the reduction in the abundance of acetyl-CoA synthetase was more significant when comparing S-O- to S-O+ condition (2.6- fold decrease, q-value = 0.01), than when comparing S+O- to the S+O+ condition (1.5-fold decrease, q-value = 0.006). A 1.4-fold reduction (q-value < 0.0001) in pyruvate ferredoxin oxidoreductase was observed only when comparing S-O- to S-O+ condition, suggesting that O2 showed a stronger suppressive effect under the sulfate-sufficient condition. Conversely, in the dissimilatory sulfate reduction pathway, the suppressive effect of O2 was intensified under sulfate-limited conditions, as indicated by a larger decrease in the abundance of sulfate adenylyltransferase when comparing S-O- to S-O+ condition (2.3-fold decrease, q-value = 0.007), than when comparing S+O- to S+O+ condition (1.2-fold decrease, q-value = 0.01). Additionally, a 1.4-fold (q-value = 0.0001) decrease in the abundance of dissimilatory sulfate reductase was only observed when comparing S-O- to S-O+ condition (Fig. 5f, Table 2) and no significant change when comparing S+O- to S+O+ condition.

Furthermore, metaproteomics analysis identified a total of 779 proteins from sulfide/sulfur-oxidizing bacteria (SOB), with the highest SOB proteome abundance observed under S+O+ condition (Fig. 5c and Supplementary Table 1). The proteome abundance of SOBs was negatively correlated with the CH4 accumulation (r = -0.7, p-value = 0.01, t-test), while the evenness of SOBs showed a positive correlation with the CH4 accumulation (r = 0.8, p-value = 0.0006, t-test) (Supplementary Fig. 2).

Modeling of carbon and energy flow through the microbial communities

A stoichiometric model was constructed to model the fluxes through the primary carbon metabolism and energy conservation pathways among the predominant microbial guilds in the wetland communities. The overall model comprised nine reactions, including aerobic heterotrophy, lactate fermentation, hydrogenic acetogenesis, sulfidogenic lactate oxidation, sulfidogenic hydrogen oxidation, hydrogenotrophic methanogenesis, acetoclastic methanogenesis, methane oxidation, and butyrate production (Supplementary Fig. 3). The fluxes of these reactions were fitted to the cumulative amounts of acetate, lactate, butyrate, H2, CO2, CH4, and H2S produced or consumed over the 30 days of incubation. The correlation coefficient R2, exceeded 0.9 in all four conditions (Supplementary Fig. 4). The consistency between the experimental data and the modeling results indicated a plausible estimation of the carbon and energy flows.

Under the S-O- condition, the primary reactions involved were hydrogenic acetogenesis, hydrogenotrophic methanogenesis, and acetoclastic methanogenesis. SO42- addition significantly increased the flux from 0 to 0.33 ± 0.005 mmole/g dry soil through sulfidogenic hydrogen oxidation (Fig. 6a, b). This observation was consistent with the increased abundance of functional proteins, including sulfate adenylyltransferase (1.4-fold, q-value = 0.002), adenylylsulfate reductase (1.8-fold, q-value < 0.001) and dissimilatory sulfate reductase (1.5-fold, q-value = 0.001) in the dissimilatory sulfate reduction pathway (Fig. 5c). Furthermore, SO42- addition halted hydrogenotrophic methanogenesis and significantly reduced the flux by 1.7- folds through acetoclastic methanogenesis (p-value = 0.0014, Wald type II χ² tests) (Fig. 6b). These reductions were supported by the fact that the abundance of functional proteins for hydrogenotrophic methanogenesis decreased much more than for acetoclastic methanogenesis. Under the two oxic conditions, the dominant reaction shifted to aerobic heterotrophy. The fluxes through dominant reactions observed under anoxic conditions were consistently lower than oxic conditions, such as hydrogenic acetogenesis (fold change > 3.9, p-value = 0.000, Wald type II χ² tests), hydrogenotrophic methanogenesis (fold change = 6.0, p-value = 0.004, Wald type II χ² tests), and acetoclastic methanogenesis (fold change > 35.4, p-value = 0.0003, Wald type II χ² tests). Moreover, O2 also eliminated sulfidogenic hydrogen oxidation and butyrate production (Fig. 6c, d). Under S+O+ condition, sulfidogenic lactate oxidation replaced sulfidogenic hydrogen oxidation for sulfate reduction compared to S+O- condition (Fig. 6b, d). These findings indicate that the addition of SO42- and O2 restructured the carbon flows across the metabolic network of the wetland community.

Fig. 6: Stoichiometric modeling of the wetland microbial community.
figure 6

ad fluxes of the key reactions under four experimental conditions. eh Production (+) and consumption (-) of CO2 and CH4 by these reactions. The error bars are defined as standard deviation.

The models estimated the contributions of each reaction to the CO2 and CH4 accumulations. Under S-O- condition – where CO2 accumulation was lowest – it was estimated 51% of CO2 was derived from hydrogenic acetogenesis (0.28 ± 0.01 mmole/g dry soil), 32% from acetoclastic methanogenesis (0.18 ± 0.002 mmole/g dry soil), and 17% from butyrate production (0.091 ± 0.01 mmole/g dry soil). Additionally, 0.14 ± 0.008 mmole/g dry soil CO2 was used for hydrogenotrophic methanogenesis to produce CH4. SO42- addition reduced the contributions of butyrate production (0.064 ± 0.014 mmole/g dry soil) and acetoclastic methanogenesis (0.099 ± 0.002 mmole/g dry soil) to CO2 accumulation by 31% and 45%, respectively. However, the contribution of CO2 accumulation from hydrogenotrophic acetogenesis increased by 13%, and no flux was detected from hydrogenotrophic methanogenesis using CO2 to produce CH4. These overall changes resulted in a 16% increase in total CO2 accumulation. The highest CH4 accumulation was observed under S-O- condition, with acetoclastic methanogenesis contributing 56.6% (0.18 ± 0.002 mmole/g dry soil) and hydrogenotrophic methanogenesis contributing 43.4% (0.14 ± 0.008 mmole/g dry soil) (Fig. 6e). Under the S+O- condition, only acetoclastic methanogenesis contributed to CH4 accumulation, and the total CH4 accumulation decreased by 69% compared to the S-O- condition (Fig. 6f). Under the two oxic conditions, aerobic heterotrophs contributed 94% and 96% of the CO2 produced under non-sulfate addition and sulfate addition conditions, respectively (Fig. 6g, h).


Climate change will result in increased hydrological challenges to wetlands, including seawater intrusion and changing regional regimes of drought and flooding. These stressors lead to recurrent perturbations in wetland microbial communities, primarily due to elevated O2 during periods of drought and elevated SO42- from seawater intrusion22,29. Our findings demonstrated that the addition of SO42- and O2 diminished CH4 emissions and increased CO2 emissions, which is consistent with observations in natural wetland ecosystems exposed to increased seawater and O21,30,31. However, it remains challenging in natural wetland soil to determine how SO42- and O2 shape microbial processes that ultimately impact CH4 and CO2 emissions32,33. In this study, we developed a conceptual food web connecting the metabolic activities of key microbial populations to CH4 and CO2 emissions, under the perturbation of elevated SO42- and O2 (Fig. 7). Our findings indicated that elevated availabilities of SO42- and O2 changed the compositions of functional guilds and their metabolic activities, including plant polymer breakdown, methane production and oxidation, as well as sulfide/sulfur oxidation. Elevated SO42- reduced CH4 emission by 3.7-fold and increased CO2 emission by 1.1-fold. Exposure to O2 resulted in a 33.3-fold decrease in CH4 emission and only a 1.7-fold increase in CO2 emission. Because the warming effect of CH4 is approximately 28-36 times greater than that of CO234, the elevated SO42- and O2 exposures may reduce the overall warming effect of gas emissions from wetlands.

Fig. 7: Conceptual food web responding to SO42- and O2 perturbations.
figure 7

The S-O- condition serves as the baseline. Circle size corresponds to the relative abundance of marker proteins within the reaction, with larger circles indicating higher abundance, and smaller circles denoting lower abundance. Numbers alongside circles indicate the fold change in metabolic fluxes as inferred from stoichiometric models relative to the control. Edge thickness represents the magnitude of fluxes that consume or produce metabolites. Dashed lines indicate fluxes for nearly eliminated hydrogenotrophic methanogenesis post-treatment. Statistical significance was assessed by Wald type II χ² tests, with p-values adjusted for the false discovery rate indicated by: *** p < 0.001, ** p < 0.01, * p < 0.05.

The reduction in the CH4 production by SO42- exposure was attributed primarily to the sharp decrease in H2 available for hydrogenotrophic methanogenesis and acetate needed for acetoclastic methanogenesis. Notably, hydrogenotrophic methanogenesis appeared to be more sensitive to elevated SO42- compared to acetoclastic methanogenesis. This sensitivity is rooted in the thermodynamic basis of competition for H2 and carbon by methanogenic consortia and SRB, as hydrogenotrophic sulfate reduction is more energetically favorable than either hydrogenotrophic methanogenesis or acetoclastic methanogenesis22,35,36. Even under oxic conditions, SO42- addition facilitated the co-existence of SOB and SRB. Oxygen consumed by sulfide-oxidizing SOB fosters the growth of anaerobic sulfide-generating SRB. In contrast, acetoclastic methanogens do not directly compete with SRB, as most SRBs have an incomplete TCA cycle, which prevents them from completing the oxidation of acetate to CO237,38. In addition, our results indicated that a key factor contributing to the increase in CO2 emission due to the addition of SO42- is the inhibition of hydrogenotrophic methanogenesis, which consumes CO2. SO42- addition reorganized carbon fluxes through the community due to SRB outcompeting hydrogenotrophic methanogens, in turn driving accumulation of short-chain fatty acids and increasing CO2 emissions.

O2 suppressed both hydrogenotrophic methanogenesis and acetoclastic methanogenesis. In the absence of SO42-, elevated O2 suppressed reactions that produce H2, such as hydrogenic acetogenesis. The decreased H2 production constrained the hydrogenotrophic methanogenesis39. Under the sulfate addition conditions, the introduction of O2 was found to enhance the consumption of H2S by SOB. Consequently, this promoted H2 uptake by sulfate reduction and intensified the competition between hydrogenotrophic methanogens and SRB38. The proteomic abundances of carbohydrate degradation enzymes also indicated that O2 altered the preference for the degradation of plant polysaccharides. Exposure to O2 favored microbial communities that hydrolyze polysaccharides with more complex structures. Upon recovery from oxic conditions, this may lead to prolonged release of CO224.

Approximately 80% of the studies investigating the impact of anthropogenic pressures on ecosystem functions have primarily focused on analyzing the effect of individual stressors in isolation40,41. However, climate change may introduce many co-occurring stressors, which need to be investigated in combination. In this study, we found an antagonistic interaction between the effects of sulfate and O2 on CH4, and CO2 emissions, where the reduction in CH4 accumulation and the increase in CO2 resulting from the combined treatment of SO42- and O2 was less pronounced compared to the sum of individual effects. This non-additive effect can be attributed to the high functional redundancy stemming from extraordinarily diverse functional guilds in soils41,42,43. Furthermore, both SO42- and O2 affect certain metabolic processes in a similar manner. For instance, elevated levels of either SO42- or O2 nearly completely halted hydrogenotrophic methanogenesis, resulting in a reduction in CH4 accumulation. Therefore, when SO42- and O2 exposures are combined, the observed methanogenesis decrease is not additive.

In this study, we generated a functional understanding of the decreased CH4 and increased CO2 emissions in wetland microcosms subjected to SO42- and O2 pressure by integrating geochemical analysis, proteogenomics, and stoichiometric modeling. The geochemical analysis quantified the end-products’ accumulation levels and the intermediates’ standing levels in the metabolic network, which were used by the stoichiometric modeling to estimate the flux rates across key metabolic processes under different conditions. Many of the changes in the estimated flux rates were supported by the concordant changes in the protein abundances of enzymes involved in those metabolic processes. The obtained results may enable climate models to better estimate greenhouse gas emissions under changing environmental conditions.


Sampling and soil microcosm set up

Triplicate soil samples were collected from a seasonally flooded urban freshwater wetland connected to Lake Washington in Seattle, WA, USA (Coordinates: 47.642196oN, -122.296236oW) in May 2022 using a soil core sampler with 2” x 6” plastic liners (AMS, Inc., American Falls, ID). The core sampler was forced approximately 60 cm into the sediment bed with a slide hammer to cut roots and other plant material. The three replicate soil cores were collected approximately one meter apart from each other, avoiding large tree branches or roots. The broader location was selected based on its variably submerged nature, with the water table generally being below the surface during the fall and winter months and above the surface during the spring and summer months. At the time of sampling, the water table was above the soil surface and the soil was completely saturated with water. The cores were sealed air-tight upon removal from the natural habitat, to protect the redox state of samples. Samples were then transported to the lab on ice, for immediate processing.

Microcosms were set up by filling 160 mL bottles with 40 g fresh sediment soil slurry in an anaerobic chamber. The chamber gas contained N2 : H2 (97:3). All bottles were preincubated for five days at room temperature and then incubated at 30 °C for 30 days. Soil microcosms were divided into four groups, with each group receiving specific treatments every two days: anoxic non-sulfate-addition, anoxic sulfate-addition (resulting in a final SO42- concentration of approximately 7.2 mM), oxic non-sulfate-addition (where half of the gas phase was replaced by sterile air), and oxic sulfate-addition, which was treated with both SO42- and sterile air. Sampling was conducted on days 2, 9, 16, 23, and 30, with each group sacrificing three bottles for sampling at the indicated time points. During sampling, 10 mL of gas phase products were collected in vacuumed Labco exetainer gas vails for later analysis. Then 10 ml of PBS buffer was injected into the bottle, and shaken at 200 rpm for 10 min, after which 10 ml of well-shaken culture liquid was sampled, and stored at -80 °C for subsequent analysis.

Chemical analysis

Headspace pressure was measured using a pressure gauge (Cole Parmer SK-68900-24) before sampling. pH measurements were taken immediately after sampling. Analysis of H2, CO2 and CH4 was carried out using a gas chromatograph (GC-8A, Shimadzu, Japan). H2S in the gas phase was measured by an H2S monitor (Forensics, FD-103-H2S, US), the H2S dissolved in slurry was calculated based on the pH and the amount of H2S in the gas phase. Absolute gas composition was calculated with the ideal gas law. The soluble sugars (glucose, xylose, mannose and galacturonic acid) and fermentation products (acetate, lactate, and butyrate) in the slurry were analyzed by high-performance liquid chromatography (HPLC) as reported previously44. The metabolic measurements were standardized using the gram of dry soil at the endpoint, which constitutes 22.9 ± 0.1% of soil slurry.

DNA extraction and metagenomic sequencing

Only the initial soils from the field were sequenced using metagenomic analysis. The total DNA in the triplicate field soil samples was extracted using the PowerMaxSoil DNA isolation kit as described previously45. Metagenomic library preparation and DNA sequencing were performed at the Joint Genome Institute. The metagenomic libraries were prepared for sequencing using 2 × 151-bp lanes on the Illumina NovaSeq S4 platform. A total of 404,050,411 ± 23,970,278 sequence reads were obtained.

Metagenomic data processing

Metagenomic reads were preprocessed using BBTools for removing adaptors, trimming reads, and sequencing error correction46. The pre-processed reads from three replicates were co-assembled into a combined metagenome with SPAdes v3.15.547, A total of 857,101 scaffolds with a length > 1 kbp were retained. The total length and L50 of the assembly were 1,939,450,947 bp and 2420 bp, respectively. The percentage of sequencing reads mapped onto the scaffolds is 87%. Genes were called from these retained scaffolds using the Prodigal algorithm48. A total of 26,768,938 genes were identified. Gene functions were annotated using Kofamscan49. Taxonomic classification was carried out at the scaffold level using Kaiju, based on reference species in the NCBI RefSeq50. Metagenome-assembled genomes (MAGs) were obtained using MetaBAT v2.12.1 with default parameters51. The quality of MAGs was estimated using CheckM v1.1.2. The MAGs obtained were classified into high-quality MAGs with completeness > = 50% and contamination < 5%. The taxonomy classifications of high-quality MAGs were inferred using GTDB-Tk v2.3.252.

Metaproteomics measurement

Soil cultures at the end of the experiment were analyzed by metaproteomics. Proteins were extracted as described previously. Briefly, soil samples were suspended in lysis buffer (containing 10 mM tris-HCl, 4% SDS and 10 mm dithiothreitol), boiled for 5 min and further disrupted by sonication for a 2 min 10% pulse 5 times. The supernatant was collected by centrifugation at 14,000 g for 10 min. Proteins were then precipitated by TCA (trichloroacetic acid) overnight and pelleted by centrifugation. Protein pellets were washed with ice-cold acetone three times and resuspended in guanidine buffer. Protein concentrations were quantified by Bicinchoninic Acid Assay. Twenty mg samples of proteins were processed using FASP (filter-aided sample preparation) and digested by Trypsin/LysC mixture. Each sample was analyzed using 2D-LC-MS/MS (two-dimensional liquid chromatography-tandem mass spectrometry) on an Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific, USA) at the IDeA National Resource for Quantitative Proteomics. Tryptic peptides were separated into 46 fractions on a 100 × 1.0 mm Acquity BEH C18 column (Waters) using an UltiMate 3000 UHPLC system (Thermo) with a 50 min gradient from 99:1 to 60:40 buffer A:B ratio under basic pH conditions, then consolidated into 12 super-fractions. Each super-fraction was then separated by reverse phase XSelect CSH C18 2.5 um resin (Waters) on an in-line 120 × 0.075 mm column using an UltiMate 3000 RSLCnano system (Thermo). Peptides were eluted using a 60 min gradient from 98:2 to 65:35 buffer A:B ratio. Eluted peptides were ionized by electrospray (2.4 kV) followed by mass spectrometric analysis on an Orbitrap Fusion Tribrid mass spectrometer (Thermo). MS data were acquired using the FTMS analyzer in profile mode at a resolution of 240,000 over a range of 375 to 1500 m/z. Following HCD activation, MS/MS data were acquired using the ion trap analyzer in centroid mode and normal mass range with normalized collision energy of 28–31% depending on charge state and precursor selection range. Protein identification and quantification followed established procedures. Briefly, mass spectrometry spectra were searched using Sipros Ensemble against a protein database constructed from the metagenome. Raw search results were filtered to achieve 1% FDR (false discovery rate) at the peptide level, which was estimated using the target-decoy approach. Peptide identifications are assigned to protein or protein groups in accordance with the parsimonious rule. To avoid ambiguity in data analysis, protein groups were excluded from biological analysis. Protein quantification was achieved through Intensity-based label-free analysis using ProRata14. Protein abundances were quantified by the total peak height of all quantified peptides from a protein, normalized against the average total of all data sets53,54,55.

Following previously described methods56,57,58, the taxonomic annotations were assigned to the identified peptides based on the matched corresponding scaffold’s taxonomic annotation information, which had a length greater than 1000 bp. The relative abundance of a species is expressed as the sum of the abundances of all proteins detected that belong to this species. If a protein was identified in more than one species, its abundance was equally divided among these identified species. The richness and evenness of the communities were calculated using the vegan R package, based on the total protein abundance of species59.

Construction of the stoichiometric model

A set of overall reactions depicting the documented metabolism of major functional populations was fit to the measured amounts of metabolites in the microcosms (Supplementary Fig. 3, Supplementary Fig. 4). The code used in fitting can be found on the Github The resulting fits provide a basis for quantifying metabolic contributions. Deviations between the inferred metabolic transformations and measured transformations indicate uncertainty in our understanding of the metabolic transformations or limitations in analytic capacity.

Statistical analysis

Pairwise comparisons for metabolic measurement data were carried out using Student’s t-test, while metabolic fluxes from modeling were analyzed using Wald type II χ² tests. One-way or two-way ANOVA was used for comparisons among groups of more than two. Differences in protein abundance in the metaproteome were analyzed by Deseq2 R package, which is based on a model using the negative binomial distribution to account for variance and mean linkage through local regression60,61,62. The p-values were adjusted to q-values using the Benjamini and Hochberg method for multiple comparison correction63. Differences with q-value < 0.05 were regarded as statistically significant.