Microbial Ecology and Site Characteristics Underlie Differences in Salinity‐Methane Relationships in Coastal Wetlands

Methane (CH4) is a potent greenhouse gas emitted by archaea in anaerobic environments such as wetland soils. Tidal freshwater wetlands are predicted to become increasingly saline as sea levels rise due to climate change. Previous work has shown that increases in salinity generally decrease CH4 emissions, but with considerable variation, including instances where salinization increased CH4 flux. We measured microbial community composition, biogeochemistry, and CH4 flux from field samples and lab experiments from four different sites across a wide geographic range. We sought to assess how site differences and microbial ecology affect how CH4 emissions are influenced by salinization. CH4 flux was generally, but not always, positively correlated with CO2 flux, soil carbon, ammonium, phosphate, and pH. Methanogen guilds were positively correlated with CH4 flux across all sites, while methanotroph guilds were both positively and negatively correlated with CH4 depending on site. There was mixed support for negative relationships between CH4 fluxes and concentrations of alternative electron acceptors and abundances of taxa that reduce them. CH4/salinity relationships ranged from negative, to neutral, to positive and appeared to be influenced by site characteristics such as pH and plant composition, which also likely contributed to site differences in microbial communities. The activity of site‐specific microbes that may respond differently to low‐level salinity increases is likely an important driver of CH4/salinity relationships. Our results suggest several factors that make it difficult to generalize CH4/salinity relationships and highlight the need for paired microbial and flux measurements across a broader range of sites.


Introduction
The most recent report by the Intergovernmental Panel on Climate Change stated that increases in well-mixed greenhouse gases such as carbon dioxide (CO 2 ), methane (CH 4 ), and nitrous oxide (N 2 O) have contributed 1.0-2.0°C of warming to Earth's climate in the past century (IPCC, 2021).After CO 2 , CH 4 is the largest contributor, with a global warming potential 84-86 times that of CO 2 over a 20-year timeframe.CH 4 concentrations have increased from 675 ppb in the 1,700 s to 1,866 ppb in 2019, after being stable for most of the previous 800 years (Etheridge et al., 1998;IPCC, 2021).Wetlands are the largest natural source of CH 4 ; the methanogenic activity of anaerobic archaea in wetland soils contributes approximately 30% of all methane emissions globally (Saunois et al., 2020).It is thus important to understand the drivers of wetland CH 4 emissions, especially given that there has been much recent discussion of the potential for coastal and estuarine wetlands to sequester vast quantities of carbon in what has been referred to as "blue carbon" (Grimsditch et al., 2013;Kelleway et al., 2020;Mcleod et al., 2011;Nellemann, 2009).However, any increases in carbon (C) storage could potentially be offset by greenhouse gas emissions (Hemes et al., 2018;Rosentreter et al., 2021;Valach et al., 2021), which warrants further study and quantification.
Another topic in need of further research is how minor increases in salinity will affect greenhouse gas emissions, especially given that two principal factors driving coastal and estuarine wetland salinization-drought and sea level rise-are predicted to increase in the future with ongoing climate change (Herbert et al., 2015).Such salinization causes former tidal freshwater wetlands to become oligohaline (>0.5-5 ppt).Globally, sea levels have risen 20 cm since 1901 and are currently (2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018) rising at a rate of 3.7 mm per year, as a result of both melting ice and thermal expansion (IPCC, 2021).Droughts have increased in frequency and severity in many parts of the world, which decreases freshwater inputs into estuarine and coastal ecosystems, leading to low-level increases in salinity (Chamberlain et al., 2020;IPCC, 2021).Several other anthropogenic factors such as water management can also contribute to low-level salinization, which, in turn, precipitates ecosystem changes including loss of biodiversity and ecosystem services (Tully et al., 2019).
Salinity is a key environmental variable that affects plant, bacterial, archeal, fungal, and zooplankton community composition and productivity, as well as biogeochemical processes (Crain et al., 2004;Lozupone & Knight, 2007;Mohamed & Martiny, 2011;Weston et al., 2006;Zervoudaki et al., 2009).Only certain organisms in the tree of life have evolved to function at elevated salt concentrations, either by using compatible solutes to keep salt out or by having acidic proteomes and special ion pumps to function with high intracellular salt concentrations (Gunde-Cimerman et al., 2018;Oren, 2013).Effects of salinity on plant communities have been well documented, with drastic observed compositional changes along salinity gradients, including low-level, or oligohaline, salinities (Odum, 1988).High salinities are toxic to plants adapted to freshwater, while plants adapted to saltwater are outcompeted by freshwater-adapted plants in freshwater environments, and are therefore only found in saltwater environments (Crain et al., 2004).Salinity has been found to be a principal variable structuring microbial communities at global, regional, and landscape scales (Dragone et al., 2021;Hartman et al., 2024;Laas et al., 2022;Lozupone & Knight, 2007).At the chemical level, salinity also affects the solubility of dissolved organic carbon and other compounds, as well as the sorption of inorganic ions (Ardón et al., 2013;Neubauer et al., 2013;Weston et al., 2010).While oligohaline salinities are high enough to elicit responses in microbial communities (Hu et al., 2020;Lew et al., 2022), work done across broader salinity gradients including mesohaline (5-18 ppt) and polyhaline (18-30 ppt) areas suggests that the effects of salinization will likely be dependent on the magnitude of salinization (Hartman et al., 2024;Morina & Franklin, 2022;Weston et al., 2014).This may in turn affect the relationship between salinity and methane emissions, which has been shown to be non-linear (Hartman et al., 2024).
Theory predicts that CH 4 emissions will decrease with increased seawater influence as the concurrent increase in sulfate from seawater will promote sulfate-reducing organisms that can outcompete methanogens for shared resources, such as acetate and hydrogen (Achtnich et al., 1995;Kristjansson & Schönheit, 1983;Lovley & Klug, 1983;Lovley et al., 1982;Schönheit et al., 1982).Such a decrease, in fact a log-linear decrease in CH 4 emissions with salinity (including oligohaline salinities up to seawater salinities), is indeed what has been found in several studies and meta-analyses (Al-Haj & Fulweiler, 2020;Bartlett et al., 1987;Luo et al., 2019;Poffenbarger et al., 2011).However, a summary of laboratory microcosm experiments that tested salinity-methane relationships reported 8 negative relationships, 2 positive relationships, and 1 neutral relationship (Ardón et al., 2018).Similar variation has been observed for CO 2 emissions (Luo et al., 2019).Such variation in relationships has been attributed in part to hydrological setting (Ardón et al., 2018;Helton et al., 2019) and soil characteristics (Ury et al., 2022), but the roles of site context/history, including legacy effects of agriculture and fertilizer use (Ardón et al., 2017), as well as microbial ecology-relations of microbes to one another and to their physical surroundings-are two other main factors that could contribute to such discrepancies and are in need of further research.Wetlands with a history of agriculture use, and consequently fertilizer application, are characterized by what are known as biogeochemical legacies, including altered carbon and nutrient pools, such as legacy nitrogen and phosphorus (P) that can remain in soils for decades after the cessation of inputs (Ardón et al., 2017).Other important site legacies include the history of inundation and prior exposure to salinity (Hopple et al., 2022).CH 4 cycling is driven by the activity of, and interactions among, multiple distinct functional groups of microorganisms, including complex carbon degraders, fermenters, sulfate reducers, iron reducers, ammonia oxidizers, denitrifiers, methanogens, and methanotrophs (Bodelier & Steenbergh, 2014;Conrad, 2020;Lovley, 1991).These groups encompass functionally and taxonomically diverse organisms.Complex carbon degraders and fermenters are crucial players in the decomposition process, breaking down larger, more recalcitrant carbon compounds such as lignin and cellulose into fatty acids and eventually acetate, H 2 and CO 2 that fuel methanogenesis (Conrad, 2020).While sulfate reducers can compete with methanogens, they can also be syntrophic with methanogens, as certain taxa produce acetate and hydrogen which would then fuel methanogenesis from those substrates.Methanogens can perform one of four different methanogenesis pathways, including acetoclastic (CH 4 produced from acetate), hydrogenotrophic (CH 4 produced from hydrogen and CO 2 or other electron acceptors such as carbon monoxide or alcohols), methyl-dismutation (i.e., methylotrophic, CH 4 produced from methylated compounds), and methyl-reduction (CH 4 produced from methylated compounds and hydrogen) (Bueno de Mesquita et al., 2023;Kurth et al., 2020).Most methyl-dismutation pathways are likely not affected by competition with sulfate reducers (Maltby et al., 2016;Oremland & Polcin, 1982;Oremland et al., 1982;Xu et al., 2021).Methanotrophs include anaerobic archaea and both anaerobic and aerobic bacteria (Conrad, 2007;Guerrero-Cruz et al., 2021), the latter of which have been divided into separate types based on taxonomy and physiology (here we use types I, II, IIa) (Bodelier, 2011;Hanson & Hanson, 1996;Knief, 2015).Furthermore, prior work has suggested key interactions between the methane and nitrogen cycles, and among the taxa involved, particularly methanotrophs, ammonia oxidizers (AO) and nitrite oxidizing bacteria (NOB) (Bodelier & Steenbergh, 2014).Nitrogen can stimulate CH 4 oxidation in N limited conditions, while excess ammonia and nitrite can inhibit CH 4 oxidation (Alam & Jia, 2012;Cai et al., 2007;Dunfield & Knowles, 1995).Such diversity of microorganisms may lead to contrasting responses or a lack of response to low salinity, depending on the direct effects of salinity on these taxa, the indirect effects of other environmental variables on these taxa, the effects of interactions among these taxa, and the overall functional redundancy of the microbial community (Bodelier, 2011;Morrissey, Berrier, et al., 2014;Nyerges & Stein, 2009).
In this study we combine microbial and biogeochemical data from field and laboratory experiments in four different sites to assess how salinization affects microbial communities and biogeochemistry, including greenhouse gas fluxes.These data encompass tidal freshwater marshes, oligohaline wetlands, a freshwater forested wetland, control microcosms and microcosms amended with artificial seawater (ASW), field plots amended with artificial seawater, and tidal freshwater soils transplanted to an oligohaline wetland.With respect to microbial ecology, we hypothesized that (H1) low salinity would reduce alpha-diversity within each site due to the direct effects of NaCl on freshwater-adapted taxa, and (H2) broad site geographic, hydrologic, plant, and biogeochemical differences would lead to significant differences in microbial beta-diversity across the four geographic locations due to the effects of those variables on microorganisms.With respect to relationships with CH 4 flux, we hypothesized that (H3) a greater relative abundance of methanogens would fuel more methanogenesis and a lower relative abundance of methanotrophs would result in more CH 4 emitted to the atmosphere, (H4) increased availability of alternative electron acceptors would suppress methanogenesis due to competition, and (H5) greater overall decomposition rates would fuel more methanogenesis by providing methanogenic substrates.To test H5 we used correlates of decomposition, including CO 2 flux, organic carbon, ammonium, phosphate, pH, and relative abundance of hypothesized complex C degraders (Actinobacteriota, Firmicutes).While we did not directly measure decomposition rates, the literature shows that decomposition rates are positively correlated with carbon availability (Luo et al., 2015), as well as nutrient availability, including nitrogen and phosphorus (Barantal et al., 2012;Yan et al., 2020).Nutrient availability could also increase plant growth, but this in turn could increase carbon inputs and fuel increased decomposition.Long-term nutrient addition experiments have also been shown to increase CH 4 fluxes (Juutinen et al., 2018).Regardless of potential pH effects on methanogens, research has shown that overall decomposition rates are higher at more neutral pH compared to acidic pH (Walse et al., 1998).Firmicutes and Actinobacteriota were shown to harbor complex carbon degradation genes in wetlands (Hartman et al., 2024), so we also include them in this prediction.

Methods
We synthesized data from five research projects from four separate sites, chosen because they included measurements of methane fluxes in freshwater and oligohaline conditions, either due to natural variation or to field or lab manipulation, and had retained soil samples adequate for DNA extraction and sequencing for microbial community characterization, which had not been previously done at three of the four sites.The projects include field sampling across a natural salinity gradient, field manipulations of salinity with either seawater addition or transplanting, and laboratory microcosm incubation experiments in which soils were brought back to the laboratory to receive artificial seawater (ASW) additions for at least 12 weeks.Broadly, the four sites are the San Francisco Bay Estuary in California, USA ("SF") (Hartman et al., 2024), the Delaware River Estuary in New Jersey, USA ("DE") (Weston et al., 2014), the Alligator River Estuary in North Carolina, USA ("Alli") (Ardón et al., 2010(Ardón et al., , 2013)), and the Waccamaw River Estuary in South Carolina, USA ("Wacc") (Neubauer, 2013) (Table 1, Figure S1 in Supporting Information S1).These estuaries contain vast expanses of wetlands; for example, there are ∼59,000 ha of wetlands in the San Francisco Estuary and ∼445,000 ha and of wetlands in the Delaware Estuary.From all of the potential data available, we selected only samples with freshwater (<1 ppt) or oligohaline (1-5 ppt) salinity classes, as this is the most important salinity class to understand regarding seawater intrusion into freshwater wetlands and tidal marshes (Table 1).All soil samples were continuously covered with water (execpt Delaware lab) and come from similar depth classes within the 0-5 and 5-15 cm range.All soil samples were exposed to increased salinity for at least 12 weeks, but the length of exposure varied among studies, including decades (San Francisco), 3 years (Waccamaw), 1 year (Delaware field), and 12 weeks (Delaware lab, Alligator lab).CH 4 flux was measured at all sites; however, the surface area, volume, and depth of the sample upon which measurements were taken were different at each site, and some sites used a gas chromatograph while a portable analyzer was used at San Francisco (Table 1).Thus, methane flux data are comparable among treatments in the same site but are not directly comparable among the sites (Table 1).Additionally, CO 2 flux was measured at San Francisco, Waccamaw, and Alligator, and N 2 O flux was also measured at Delaware and Alligator.Suites of other soil and porewater variables were measured, and this varied among the sites and experiments as described below.Variables measured at only one site were included in the analysis if similar variables were measured at other sites (e.g., soil organic matter, dissolved organic carbon).

Sampling and Measurements
In San Francisco, soils were collected from 2 freshwater and 2 oligohaline wetland complexes in the delta formed by the Sacramento and San Joaquin rivers as they empty into the San Francisco Bay.Each salinity class included one unaltered reference wetland and one wetland restored from agricultural use as either cropland or pasture.At each wetland, three cores were taken with a split core auger with an airtight plastic sleeve with a 5 cm diameter and 15 cm depth.Each core in the plastic sleeve was immediately capped on both ends to maintain an anaerobic environment.The core was then placed into a 2 L Mason jar and the top cap was removed to quantify CH 4 and CO 2 flux over the course of 5 min on an ultraportable greenhouse gas analyzer (Los Gatos Research, San Jose, CA, USA).An additional core adjacent to each initial core was taken, split into the 0-5 cm depth segment and 5-15 cm depth segment, and aliquoted for DNA sequencing (stored at 20°C) and biogeochemical analyses of soil and porewater at the UC Davis Analytical Laboratory.Measured variables included: porewater dissolved organic carbon (DOC) and pH, soil C, nitrogen (N), C:N, ammonium (NH 4 + ), and Olsen phosphate (PO 4 3 ), and both porewater and soil nitrate (NO 3 ), sulfate (SO 4 2 ), iron (Fe), and manganese (Mn).Samples from San Francisco are all unmanipulated field samples (n = 72).
The site is flooded 10-30 cm during most high tides.Five replicate plots, at least 3 m apart from each other, received 40 L of either freshwater or seawater additions every 3-4 days over the course of 3 years.Seawater additions were from salt marsh tidal creek water from the flow-through seawater system at the Baruch Marine Field Laboratory (BMFL) that was diluted with freshwater to a salinity of 10.2 before being added to the marsh.
Freshwater additions were from a 180-m-deep groundwater well at the BMFL (salinity = 0.5 ± 0.04).Plot-scale exchanges of CO 2 and CH 4 were measured with large, transparent, temperature-controlled chambers 0.37 m 2 × 1.22 m height.Air within each chamber was stirred with four fans while pumps circulated air between the chambers and a LI-COR LI7000 CO 2 /H 2 O analyzer (LI-COR Biosciences, Lincoln, NE, USA).To determine CH 4 fluxes, air samples were collected from each chamber roughly every 5-7 min, stored in gas tight Hungate tubes, and analyzed for CH 4 concentration in the laboratory using a flame ionization detector on a Shimadzu GC14A gas chromatograph (Shimadzu Scientific Instruments, Columbia, MD, USA).Soils were collected at the end of the experiment by collecting one core of 6.4 cm diameter and 10 cm depth from each plot.Other measured  variables included soil organic matter (SOM), C, N, and C:N, and soil oxygen demand (SOD).Samples from Waccamaw are all manipulated field samples (n = 30).

Note
At Delaware, data come from two separate projects-a field transplant experiment and a laboratory incubation experiment.For the Delaware transplant experiment, large (29 cm diameter and approximately 30 cm depth) intact cores of tidal freshwater marsh (Rancocas Creek) soil and vegetation were transplanted into an oligohaline wetland (Salem) and a separate tidal freshwater marsh (Raccoon Creek) to serve as a control.Tidal freshwater marsh plant communities were dominated by Peltandra virginica and species of Bidens, Amaranthus, and Polygonum, while the oligohaline site was dominated by Zizania aquatica (wild rice).Transplants were done at water level as well as 40 cm below water level, but only the 40 cm flooded samples are included here, to more closely match the Delaware laboratory data (see below) and the Waccamaw field experiment (see above).Soil for DNA sequencing was taken from 0 to 4 cm and 10-14 cm depth.Greenhouse gas measurements were taken by placing covered (dark) acrylic chambers of 43 L volume over the soil cores.Carbon dioxide production inside the chamber was measured with a PPSystems infrared gas analyzer for several minutes.Methane exchange was measured by taking gas samples at 0, 10, and 20 min with a 60 mL syringe, injecting into 20 mL evacuated vials, and measuring CH 4 on an Agilent 6890N gas chromatograph with a flame ionization detector.Other measured variables here included porewater NH 4 + , PO 4 3 , SO 4 2 , and Fe 2+ .
For the Delaware laboratory microcosm incubation experiment, soils from the Woodbury Creek tidal freshwater marsh were collected to a depth of 25 cm with 10 cm diameter polyvinyl chloride tubes and sealed with gas-and water-tight caps (Weston et al., 2011).Holes were drilled in the core barrel just above the soil surface, and then cores were placed into separate 100 L tidal tanks in the dark at 20°C.The tidal tanks simulate the tidal cycle with 6 hr of air exposure followed by 6 hr of inundation.Tanks were filled with artificial freshwater that mimicked the freshwater chemistry of the Delaware River.After a 2-week pre-incubation period, water in one tank was replaced with artificial seawater with a salinity of 4.95 ppt.Both tanks were changed at least once weekly after the initial 2week period.Cores were incubated for 12 weeks after which the 0-4 and 10-14 cm depth portions were sampled and stored at 20°C for DNA sequencing, with an aliquot sent for porewater and soil biogeochemical measurements which included Cl , SO 4 2 , PO 4 3 , DOC, NO x , and acetate.Gas flux was measured by fitting a 1.2 L gas-tight cap onto the core.An initial headspace sample (3 mL) for CH 4 analysis was obtained with a gas-tight syringe; then final CH 4 samples were obtained after approximately 1 hr.CH 4 samples were analyzed immediately by flame ionization detection gas chromatography (Agilent 6890 N with Porapak Q column).The wetlands that were sampled in both experiments have no known prior history of agricultural use but are affected by runoff from surrounding agricultural and urban areas.Sample sizes from Delaware are n = 8 for the transplant experiment and n = 8 for the laboratory experiment (Table 1).
At Alligator, soils were collected from the Timberlake Observatory for Wetland Restoration, a forested freshwater wetland restored in 2004 from prior use as a corn field that is now used as a research site.The site had not experienced saltwater incursion for at least 20 years prior to sampling.The site where the soils were collected is characterized by Eutric Histosol soils and Atlantic white cedar vegetation (Ardón et al., 2013).A laboratory incubation was started with intact soil cores 2.5 cm in diameter and 30 cm deep.Alligator microcosms received either deionized freshwater (control), artificial seawater, artificial seawater without sulfate, or sulfate.In this analysis we only included the controls and artificial seawater additions, as in the Delaware experiment.To measure CO 2 , CH 4 , and N 2 O, cores were fitted with a gas tight lid with a Swagelok brass sampling port with a rubber septum (0.6 cm).Headspace gas samples were collected immediately and after 1 hr into evacuated 8 ml gas vials.Gases were quantified on a Shimadzu 17A gas chromatograph with electron capture detector (ECD), flame ionization detector (FID), and methanizer (Shimadzu Scientific Instruments, Columbia, MD, USA).As in Delaware, the experiment proceeded for 3 months, after which soils from 0-5 to 10-15 cm depths were collected, a portion stored at 20°C for DNA sequencing, and another aliquot analyzed for porewater biogeochemistry.Measured variables at Alligator were porewater NO 3 , SO 4 2 , total organic carbon (TOC), dissolved organic nitrogen (DON), dissolved inorganic nitrogen (DIN), total nitrogen (TN), NH 4 + , PO 4 3 , and soil pH, %C, %N and C:N.Samples from Alligator are all manipulated laboratory samples (n = 15).

Microbial Sequencing and Analysis
Soils were frozen at 20°C until DNA extraction.DNA was extracted from 0.3 g of soil with a Qiagen DNeasy PowerSoil kit following the manufacturer's instructions.PCR was then used to amplify the V4 region of the 16S rRNA gene, following the standard methods of the U.S. Department of Energy Joint Genome Institute (Tremblay et al., 2015).DNA was sequenced on a MiSeq 2000 (Illumina Inc., CA, USA) with paired-end 150 base pair chemistry.Raw data were processed with the iTagger pipeline to quality-filter reads, dereplicate sequences, and cluster sequences into operational taxonomic units (OTUs) at 97% sequence similarity (Tremblay et al., 2015).Taxonomy was assigned using the "assignTaxonomy" function in the dada2 R package (Callahan et al., 2016), with the SILVA 138.1 taxonomic database (Quast et al., 2013).The mctoolsr R package (Leff, 2022) was used to remove chloroplast and mitochondrial DNA and any taxa that were not assigned at least to Bacteria or Archaea at the domain level.We used known taxonomy-function relationships to assign functional guilds of interest for anaerobic biogeochemistry (Hartman et al., 2024) 1).Sequences were deposited to NCBI GenBank with BioProject ID PRJNA1004999.

Statistical Analysis
The number of OTUs observed per sample and Shannon diversity were used as microbial alpha-diversity metrics.
The effects of site, salinity class (freshwater or oligohaline), and depth (surface 0-5 cm range vs. deeper 5-15 cm range) on alpha-diversity were tested with ANOVA followed by Tukey's post hoc.Microbial community composition was assessed by calculating a Bray-Curtis dissimilarity matrix on square-root transformed abundances, and performing a PERMANOVA test implemented with the "adonis2" function in the vegan R package (Oksanen et al., 2022).Within-group multivariate homogeneity of dispersion was tested with PERMDISP implemented with the "betadisper" function in vegan.To compare a presence/absence-based metric with the Bray-Curtis metric, we also calculated the Jaccard dissimilarity metric with vegan.Unique and overlapping taxa in freshwater and oligohaline salinity classes were calculated with mctoolsr.Differences in relative abundances of taxa among the salinity classes were tested with Wilcoxon tests.Indicator species analysis to identify taxa associated with freshwater or oligohaline salinities was performed with the "multipatt" function in the indicspecies R package (De Cáceres & Legendre, 2009), with the "r.g" species-site group association function.A phylogenetic tree was built by aligning sequences with MUSCLE (Edgar, 2010) and then building a tree with fasttree (Price et al., 2009), both of which were implemented in QIIME (Caporaso et al., 2010).Nearest taxonomic index (NTI) was calculated for each sample using the "NTI.p"function in the iCAMP R package (Ning et al., 2020).NTI is a standardized measure of the phylogenetic distance to the nearest taxon for each taxon in a sample.NTI values >2 or < 2 suggest deterministic processes govern community assembly while NTI values between 2 and 2 suggest dominance of stochastic processes (Stegen et al., 2012).Models of environmental predictors of community composition were tested with distance-based redundancy analysis (dbRDA), implemented with the "capscale" function in vegan; the best combination of predictors was selected using backward model selection with the "ordistep" function in vegan.Communities were visualized with principal coordinates analysis (PCoA), with environmental vectors fit with the "envfit" function in vegan.Spearman correlations between CH 4 flux and chemical variables and certain microbial guild or taxa abundances were calculated and pvalues corrected with false discovery rate (FDR).All figures were made with either the ggplot2 (Wickham, 2016) or pheatmap (Kolde, 2019) R packages.All analyses were performed with R version 4.2.3 (R Core Team, 2023).

Results
All wetland soils measured in this study emitted a net flux of CH 4 to the atmosphere, the amount of which varied by at least three orders of magnitude (Figure 1).Among the five data sets, there were discrepancies in the relationship between salinity and CH 4 fluxes, with 2 positive relationships, 1 negative relationship, and 2 neutral relationships (Figure 1).
Microbial alpha diversity metrics, including OTU richness and Shannon diversity, differed significantly among sites, salinity classes, and depths (Figure 2, Table 2).In two sites, San Francisco and Waccamaw, richness in oligohaline samples was significantly lower than richness in freshwater samples, as was Shannon diversity in San Francisco and Alligator (Tukey HSD p < 0.05, Figure 2).Alpha diversity was not significantly affected by salinity class in the Delaware field or lab experiments.Microbial community composition was primarily driven by site and secondarily by salinity, pH, depth, and suites of other environmental factors that were unique to each site (Figure 3, Table 2).When analyzing all of the data together, samples from each site, including all depths, salinities, and environmental gradients within each site, clustered together, demonstrating a primary influence of site.Communities associated with higher salinities were also associated with lower CH 4 emissions (Figure 3a, "envfit" p < 0.05).
To further assess the role of environmental factors within sites, tests were also performed for each site/experiment separately.In San Francisco, a field study with the highest sample size, many environmental variables, including CO 2 flux, CH 4 flux, salinity, bulk density, soil C, N, C:N, NO 3 , NH 4 + , pH, Cu, Zn, Mn, Fe, PO 4 3 , and porewater DOC, and Zn, were correlated with community composition (Figure 3b).Stepwise redundancy analysis model selection suggested that bulk density, soil Zn, pH, Cl , C:N, and Mn, and porewater sulfate were the most important variables driving microbial community composition.In the Waccamaw field experiment, results were similar; samples in plots that received seawater additions and thus became oligohaline, and with higher C:N ratios and pH, were significantly different from control freshwater samples, which had higher CO 2 flux, soil N concentrations, net N 2 emissions, and soil oxygen demand (SOD).There was also a   1).Note that salinity and CH 4 were the only two continuous variables measured in all five studies (shown as vectors in panel a).
significant effect of depth (Figure 3c).The best model predicting community composition contained SOD and CO 2 flux as variables.In the Delaware field transplant experiment, oligohaline samples, associated with increased salinity and sulfate reduction, were significantly different from freshwater samples.The best model predicting community composition only contained salinity as a predictor variable.Depth was also a particularly important factor at this site; samples from the same depth but different salinities were more similar than vice versa (Figure 3d).Laboratory incubations with Delaware wetland soils did not cleanly cluster by depth and salinity class as the other experiments did, due to a high degree of variability in the control (freshwater salinity) samples and a lower sample size (Figure 3e).Community dispersion was not homogenous among the two salinity classes (PERMDISP, p < 0.05), indicating this higher degree of variability within the control group.On the other hand, the laboratory incubations with Alligator soils did cluster by salinity class and depth.Samples that received ASW addition making them oligohaline were associated with higher NH 4 + , Br , and CH 4 while controls were associated with higher CO 2 and N 2 O fluxes (Figure 3f).SO 4 2 , PO 4 3 , and CH 4 flux were the three variables selected to best predict microbial community composition in stepwise redundancy analysis.
The most abundant microbial functional guilds across all samples were sulfate reducing bacteria (mean = 6%), syntrophic sulfate reducing bacteria (2%), ammonia oxidizing bacteria (2%), ammonia oxidizing archaea (2%), nitrite oxidizing bacteria (1%), and type I methane oxidizing bacteria (1%) (Figure 4).All other guilds made up less than 1% of the community on average across all sites, although in some individual samples they had greater relative abundances.Across all samples, methanogens averaged 0.8% relative abundance and methanotrophs averaged 2.7% relative abundance.Most guilds varied significantly among sites and within sites some were differentially abundant between freshwater and oligohaline samples.However, unlike some of the dominant phyla, there were no consistent responses to salinity among guilds across multiple sites (Figure 4).
There were 220 methanogen OTUs: 146 hydrogenotrophic OTUs, 29 acetoclastic OTUs, 35 methyl-reducing OTUs, and 10 mixotrophic OTUs.44 methanogen OTUs were found in at least two sites.Methanogen community composition varied among the 4 sites (Figure S3 in Supporting Information S1).Hydrogenotrophs were the most abundant methanogen guild across the whole data set, followed by methyl reducers, acetoclasts, and mixotrophs, mirroring the trend in methanogen guild OTU richness (Figure 4, Figure S3 in Supporting Information S1).Hydrogenotrophs, acetoclasts, and methyl-reducers were generally positively associated with CH 4 flux in all sites (Figure 5).These three guilds also tended to be positively correlated with salinity, while mixotrophic methanogens had mixed relationships among the sites (Figure S3 in Supporting Information S1).Of the 13 methanogenic families identified across the whole data set, Methanobacteriaceae (containing hydrogenotrophs and methyl-reducers), was the most abundant, followed by Methanoregulaceae (hydrogenotrophs), Methanosaetaceae (acetoclasts), Methanomassiliicoccaceae (methyl-reducers), and Methanosarcinaceae (mixotrophs containing hydrogenotrophs, acetoclasts, methylotrophs, and methyl-reducers).Methanofastidiosaceae (hydrogenotrophs and methylotrophs (Nobu et al., 2016)) increased in abundance with salinity in both field studies (Figure S2 in Supporting Information S1).The ratio of methanogens to methanotrophs (MG:MT) was positively correlated with CH 4 flux in three sites, but only significantly in the higher sample size site (San Francisco) (Figure 5).
The aerobic type I methane oxidizing bacteria (MOB_I) were the most abundant methanotrophic guild across the whole data set, followed by MOB_II, MOB_IIa, and ANME.The most abundant methanotrophic genera included Methylocystis, Crenothrix, Methyloceanibacter, and Methylocaldum.Some taxa consistently responded to salinity in the two field studies; in both San Francisco and Waccamaw, Methyloceanibacter increased with salinity while Methylomagnum and Candidatus Methylospira decreased with salinity (Figure S4 in Supporting Information S1).The four methanotrophic guilds had variable relationships with CH 4 flux among the four sites (Figure 4).AO:NOB ratios were not significantly correlated with methanotroph abundances or and methanogen: methanotroph ratios except in San Francisco, where they were positively correlated, and Waccamaw, where the ratios were negatively correlated (Figure S5 in Supporting Information S1).
At the OTU level, there was a diverse group of 48 OTUs that could be considered strong indicators of either freshwater (n = 7) or oligohaline (n = 41) salinity conditions, with FDR corrected p-values < 0.05 and indicator correlation coefficients >0.45 (Figure S6 in Supporting Information S1).Freshwater indicator taxa included OTUs from 7 different phyla but none were classified to genus.Oligohaline indicator OTUs included taxa from 13 different phyla and encompass some known genera and functional guild classifications such as the sulfate reducers Desulfatiglans and SEEP-SRB1 and the nitrifiers Nitrospira, as well as taxa only identified to broader taxonomic levels.
Several biogeochemical variables and microbial guilds were significantly correlated with methane flux (Figure 5).
In terms of salinity and alternative electron acceptors expected to be negatively correlated with CH 4 flux, NO 3 A key contrasting result was an increase in CH 4 flux following ASW addition in the Alligator River estuary forested wetland laboratory incubations, and lack of increased CH 4 flux in week 12 of the Delaware River estuary laboratory incubations (Figure 1), both of which were performed using similar methodology.Soils from Alligator have low pHs (<5.5, Figure S7 in Supporting Information S1) and were dominated to a much greater extent by Acidobacteriota (Figure 4).In terms of the dominant methanogenic taxa, different OTUs within the Methanobacteriaceae family were dominant at each site, and these different OTUs responded differently to the +ASW treatment.Two OTUs at Alligator increased with ASW addition and were associated with greater CH 4 fluxes (Figure 6).
The percent of shared OTUs among salinity classes within each site was 44%-45% in laboratory experiments and 34%-40% in field samples (Figure S8 in Supporting Information S1).Nearest taxon index (NTI) was >2 in 131 of 133 samples (Figure S9 in Supporting Information S1).NTI was significantly lower in oligohaline samples than freshwater samples in San Francisco, Waccamaw, and Alligator but not Delaware (Figure S9 in Supporting Information S1).Dissimilarity between freshwater and oligohaline samples versus samples in the same salinity class was greater for both Bray-Curtis and Jaccard metrics; Cohen's d effect size was slightly greater for Jaccard dissimilarity, but this was not affected by field or lab conditions (Figure S10 in Supporting Information S1).

Microbial Alpha-Diversity (H1)
Salinity is expected to decrease the alpha-2024 diversity of freshwater microbial communities as increased salinity creates osmotic stress that requires organisms to either produce compatible solutes or, at extreme salinities, have specialized systems to function at high internal salt concentrations (Gunde-Cimerman et al., 2018;Oren, 2013).While there are many organisms that can tolerate a range of salinities, oligohaline salinities are high enough to cause stress or mortality in those organisms that have not evolved the ability to cope with a range of salinities (Georges et al., 2019).Our data partially support this hypothesis, as we found a negative effect of oligohaline salinity on richness in San Francisco and Waccamaw; however, richness in Delaware and Alligator was not affected.Decreases in bacterial richness have been found elsewhere, such as in freshwater streams affected by mining-induced salinity (Vander Vorste et al., 2019) and along salinity gradients in coastal wetlands (Zhao et al., 2020).In contrast, slight increases in salinity can also cause an increase in richness, which has been found in Louisiana wetlands up to 3.5 ppt salinity (Jackson & Vallaire, 2009), in Tibetan lakes up to 1 ppt salinity (Wang et al., 2011), and in Chinese soils exposed to up to 3.33 ppt irrigation water (Chen et al., 2017).Other work has suggested that bacterial communities are resistant to changes up to 3 ppt in salinity (Berga et al., 2017).

Microbial Beta-Diversity and Taxa (H2)
Microbial community composition is strongly linked to enzyme activity and is thus important to understand with respect to biogeochemistry (Morrissey, Gillespie, et al., 2014).Globally, salinity is a primary determinant of bacterial and archeal community composition (Auguet et al., 2010;Lozupone & Knight, 2007), so it was not surprising to find differences between freshwater and oligohaline communities in all five studies here, although the effect of salinity was secondary to the effect of site (Figure 3).The broad differences in communities among the four sites are likely a product of different site histories, plant communities, and biogeochemistry (Table 1, Figure S7 in Supporting Information S1).Despite these differences, there were many (50 phyla, 121 classes, 237 orders, 264 families, 291 genera) shared taxa among the four sites (Figure S11 in Supporting Information S1).The indicator species analysis also suggested that there are some taxa shared across the sites that consistently associate with either freshwater or oligohaline salinities.A high proportion of Proteobacteria was found in all sites and is consistent with other studies in coastal wetlands (Dang et al., 2019;Jackson & Vallaire, 2009;Zhao et al., 2020).The decline in Verrucomicrobiota in oligohaline sites has also been observed across the Baltic Sea salinity gradient (Herlemann et al., 2011) and demonstrates a possible preference of this phylum for non-saline environments, although some members have been cultured at seawater salinities (Schlesner et al., 2006).Myxococcota have been identified as key components of tidal freshwater wetlands elsewhere (Morina & Franklin, 2022), and have been shown to decline with increased salinity (Zhao et al., 2023), in concordance with our results.While we found positive effects of salinity on the relative abundance of Chloroflexi, another common freshwater and estuarine wetland phylum (Huang et al., 2022;Ikenaga et al., 2010), results in other studies have shown mixed responses to low-level salinity (Zhang et al., 2021;Zhao et al., 2020).Firmicutes, the other major phylum with consistent responses to salinity in more than one site, are also dominant members of other freshwater wetlands and have been shown to respond positively to salinity across estuarine (Zhang et al., 2021), groundwater (Sang et al., 2018), andupland soil (Van Horn et al., 2014;Zheng et al., 2017) salinity gradients.

Methanogens and Methanotrophs (H3)
While tidal wetlands have high functional redundancy for CO 2 production, there are stronger links between CH 4 production and microbes due to the conserved nature of methanogenic archaea (Morrissey, Berrier, et al., 2014).All four methanogenic guilds distinguishable with taxonomic data were present in the data set, and the most abundant methanogens were hydrogenotrophs and acetoclasts, suggesting that these pathways of methanogenesis dominate in these wetlands.All guilds were generally positively associated with CH 4 flux, except for mixotrophs in Waccamaw and Delaware.This discrepancy is likely explained by other controls on CH 4 flux at play in the samples with high mixotroph abundance.In the Delaware River Estuary, isotope-based measurements demonstrated that acetoclastic methanogenesis dominated over hydrogenotrophic methanogenesis, although both contributed to the observed CH 4 flux; methyl-based methanogenesis was not estimated (Weston et al., 2011(Weston et al., , 2014)).Acetoclastic methanogens identified in the Delaware 16S data include the pure acetoclastic genus Methanosaeta and the mixotrophic genus Methanosarcina.On the other hand, pure acetoclasts were absent from Alligator; acetoclastic methanogenesis may still contribute to CH 4 flux there via mixotrophic Methanosarcinaceae, which were present and positively correlated with CH 4 flux.Acetoclasts have been suggested to increase with low salinity (Berrier et al., 2022); in our data sets this was only the case in San Francisco.Though not as abundant, taxa in the methyl-reducing Massilliicococcales order were present in all sites and positively correlated with CH 4 flux (though only significantly in San Francisco).The relative contribution of methyl-based methanogenesis is expected to increase with increasing salinity and in marine to hypersaline conditions due to the non-competitive substrates (e.g., betaine, trimethylamine) produced as osmolytes (Bueno de Mesquita et al., 2023;Oren, 1999;Zhou et al., 2021), but is perhaps less relevant at low salinities.In publicly available metagenomic data sets from freshwater and coastal wetlands, the abundance of archeal genes involved in different methanogenesis pathways followed similar trends of a gene involved in acetoclastic methanogenesis (cdhD) being the most abundant, followed by hydrogenotrophic (frhA) and methylotrophic genes, with higher abundances in freshwater wetlands (Bueno de Mesquita et al., 2023).The ratio of methanogens to methanotrophs (MG:MT) was positively correlated with CH 4 flux in most sites, as seen elsewhere (Rey-Sanchez et al., 2019;Zhang et al., 2019), suggesting that this is an important microbial metric to calculate.Furthermore, MG:MT was positively correlated with salinity in Alligator soils, in line with the increase in CH 4 flux there.
A substantial proportion of the total CH 4 produced in methanogenic environments is consumed by methanotrophs (i.e., methane oxidizers) before it is emitted to the atmosphere, making them an important factor in regulating net CH 4 fluxes.In marine soils the amount has been estimated to range from 43% to 85% (Reeburgh, 2007).Four methanotrophic guilds were present at each site, yet their abundances were not consistently negatively correlated with CH 4 flux, highlighting the complexity of factors that contribute to observed CH 4 fluxes.MOB_I and MOB_IIa were negatively correlated with CH 4 flux in San Francisco, however, and their higher abundances may have suppressed emissions in a wetland complex with higher methanogen abundances (Hartman et al., 2024).Prior work has shown that both aerobic and anaerobic methanotrophy declined with salinity, with particular sensitivity of aerobic methanotrophy (Dalal et al., 2008).In our data set, aerobic bacterial methanotrophs (MOB_I, MOB_II, MOB_IIa) were much more abundant than anaerobic archeal methanotrophs (ANME).Two of the three aerobic guilds did decline with salinity, but only in San Francisco (Figure 3), so our results only weakly agree with prior work (Dalal et al., 2008).Methanotroph communities differed among the sites (Figure S4 in Supporting Information S1), with key abundant taxa including Crenothrix and Methylocystis.Crenothrix spp.are also dominant methanotrophs in stratified lakes (Oswald et al., 2017), while Methylocystis spp.are also dominant in peatlands (Chen et al., 2008).While some Methylocystis species are inhibited by salinities >5 ppt (Dedysh et al., 2007), Methylocystis was still abundant in Waccamaw oligohaline samples.Methylospira and Methylomagnum declined with salinity in both field studies, while Methyloceanibacter, containing known marine methanotrophs (Takeuchi et al., 2014), not surprisingly increased with salinity, highlighting the taxa-specific environmental preferences of methanotrophs to different salinities.This may explain the lack of relationship between aggregated guild abundances and salinity, as at the guild level, all of the potentially positively and negatively responding taxa are summed.
Previous work in rice paddies has suggested links between nitrogen cyclers and methanotrophs (Bodelier, 2011;Bodelier & Laanbroek, 2004;Bodelier & Steenbergh, 2014).AOA and AOB perform the first step of nitrification, oxidizing NH 3 to NO 2 .Then NOB perform the second step, oxidizing NO 2 to NO 3 .Thus, the combined activities of AO and NOB create NO 3 , which could suppress CH 4 flux due to the stimulation of nitrate reducers which can outcompete methanogens or by the inhibition of methane oxidizers.Alternatively, AO and NOB could increase CH 4 flux due to increased inorganic N available for organic carbon-producing plant growth and microbial growth.Furthermore, a high ratio of AO to NOB could indicate nitrite buildup, which, along with other forms of inorganic N, could be toxic to methanotrophs in excess (Bodelier & Laanbroek, 2004;Dunfield & Knowles, 1995).Our abundance data for AO:NOB and methanotrophs did not support this hypothesis of inhibition, with variation among sites and a lack of relationship in most cases.While influenced by agricultural runoff, our sites are unfertilized and therefore have lower inorganic nitrogen levels than the agricultural settings studied in previous work on inorganic N inhibition of methanotrophs (Cai et al., 2007).
While the results on methanogens and methanotrophs presented here are limited to 16S rRNA marker gene sequencing, they agree with shotgun metagenomic data from two of the sites-San Francisco and Alligator (Bueno de Mesquita et al., 2024;Hartman et al., 2024).In particular, relative abundances of methanogens and methanotrophs from 16S rRNA gene PCR amplicons matched relative abundances of 16S rRNA genes extracted from metagenomes with mTAGs (Bueno de Mesquita et al., 2024;Salazar et al., 2021), suggesting limited effects of any potential primer biases.Furthermore, relative abundances of mcrA and amoA, key marker genes of methanogenesis and methanotrophy, respectively, annotated from metagenomic data by sequence homology (Morgan-Lang et al., 2020), were strongly and significantly correlated with the relative abundances of methanogens and methanotrophs extracted taxonomically from 16S PCR amplicon data.While we did not perform qPCR of these key genes, the homology-based approach with shotgun metagenomic data was actually more strongly correlated with 16S taxonomic relative abundances than in silico PCR, which was assessed for mcrA and amoA (Hartman et al., 2024).

CH 4 , Salinity, Alternative e−Acceptors (H4)
While methane fluxes are generally expected to decrease with low salinity, as shown by several meta-analyses (Al-Haj & Fulweiler, 2020; Bartlett et al., 1987;Poffenbarger et al., 2011) and studies (Berrier et al., 2022;Chambers et al., 2011Chambers et al., , 2013;;Marton et al., 2012;Neubauer, 2013), data from a diverse range of individual sites and experiments show that salinity/methane relationships are highly variable and include positive, negative, and neutral relationships (Ardón et al., 2018;Helton et al., 2019).The degree of salinity and length of exposure to salinity are important, with different relationships seen above and below 7.5 ppt (Wang et al., 2017) and over different time periods; for example, previous work at Waccamaw demonstrated that long-term saltwater intrusion reduced both CO 2 and CH 4 flux, while short-term exposure increased CO 2 but decreased CH 4 (Neubauer et al., 2013).This is also an important source of variation among our 5 studies, where samples had been exposed to salinity for decades (San Francisco), 3 years (Waccamaw), 1 year (Delaware field), and 3 months (Delaware lab, Alligator lab) (Figure 5).Elsewhere, after 1 week of saltwater exposure CO 2 increased but CH 4 decreased (Dang et al., 2019).These results may be attributable to microbial community composition changes and adaptations.Intrusion events followed by recovery from salinity involve dynamic microbial interactions that affect carbon cycling (Berrier et al., 2022).Temporary changes in salinity are expected to select for different microbial taxa (e.g., adapted to a range of salinities) compared to long-term changes in salinity, which are expected to select for taxa optimized for growth at the given salinity (Chambers et al., 2013).For example, when freshwater and marine microbial communities were mixed, the community shifted toward the marine community composition, and when freshwater communities were exposed to sterile marine water, more freshwater taxa were lost from the community than saltwater taxa were lost when saltwater communities were exposed to freshwater (Rocca et al., 2020).Another important variable is hydrologic setting-in permanently flooded soils, low salinity suppressed methanogenesis, while in intermittently flooded soils, low salinity increased methanogenesis (Helton et al., 2019).Flooded samples have greater abundances of both methanogens and methanotrophs than non-flooded samples (Rey-Sanchez et al., 2019).In our data set all samples were continuously covered with water except for Delaware laboratory incubations which underwent a simulated tidal cycle of 6 hr of flooding and 6 hr of air exposure.We sought to use additional data on microbial community composition and biogeochemistry to explain discrepancies in salinity/methane relationships.
Seawater exposure is expected to increase the availability of terminal electron acceptors for microbial growth, which affects the pathways by which organic matter is degraded (Sutton-Grier et al., 2011;Weston et al., 2006).In particular, nitrate, sulfate, iron, and manganese are key electron acceptors for microbial metabolism in anaerobic environments, and in tidal freshwater marshes specifically (Lovley, 1991;Megonigal et al., 2019;Weston et al., 2006).We hypothesized that salinization by seawater would suppress CH 4 flux due to competition between organisms reducing those electron acceptors and methanogens.Furthermore, these acceptors are used by methanotrophs during methane oxidation, whose activity would further decrease CH 4 flux (Guerrero-Cruz et al., 2021).Nitrate was negatively correlated with CH 4 emissions at Alligator, consistent with what has been shown in rice paddies (Cai et al., 2007).This is also consistent with findings that increases in NO 3 increase denitrification rates (Morrissey & Franklin, 2015), and denitrifier abundances increase with low salinity (Franklin et al., 2017).Fluxes of N 2 O, a product of denitrification of nitrate, were significantly negatively correlated with CH 4 flux at Alligator, consistent with potential competitive interactions between denitrifiers and methanogens.Of the alternative electron acceptors, sulfate has been the most widely implicated as a control on methanogenesis in freshwater (Lovley & Klug, 1983) and estuarine (Oremland & Polcin, 1982) environments.However, here we find no support for suppression of CH 4 production by sulfate at oligohaline salinities in the sites studied.Sulfatereducing bacteria (SRB) were the most abundant microbial functional guild and several sulfate reducing taxa were indicators of oligohaline conditions.However, SRB were actually positively correlated with CH 4 flux in Alligator soils.Relationships between iron and manganese and salinity are less clear; both increased with salinity in our data sets, but this could also be caused by changes in oxidation states.However, there is no evidence in our data that increased Fe-reducing bacteria led to suppression of methanogenesis.Elsewhere, salinity has initially temporarily increased iron reduction rates, but this effect diminished over time (Weston et al., 2006).Another alternative electron acceptor in wetland soils is humic substances, which are used by certain anaerobic methanotrophs (ANME) (Bai et al., 2019), certain sulfate reducers (Cervantes et al., 2002), and other taxa (Coates et al., 1998).The abundance of humic substances could suppress CH 4 emissions by either increasing activity of ANME (which consume CH 4 ), increasing the activity of SRB and other humic substance reducers that can outcompete methanogens, or by causing methanogens to switch their metabolism from methanogenesis to humic substance reduction, which does not produce methane (Valenzuela & Cervantes, 2021).We did not quantify humic and phenolic compounds in our studies, but we hypothesize they could be important in the forested and more acidic Alligator site, as SUVA 254 , a common metric of aromaticity and humics in water, was negatively correlated with CH 4 flux earlier in the same experiment on which we report here (Ardón et al., 2018); this remains an important avenue for future research.
While we only present results here from week 12 of the lab experiments, for which there was microbial data available, results from Alligator (increase in CH 4 following artificial seawater addition) were consistent with the entire time course of the experiment (Ardón et al., 2018).CH 4 emissions were significantly elevated starting at day 21 of the experiment and continuing until the experiment's conclusion on day 112.Although the limited amount of flux data available from cores with microbial data (week 12) from Delaware did not show elevated CH 4 flux (Figure 1), the complete time series of that experiment showed increased CH 4 emissions within 1 week after artificial seawater addition and continuing for 5 months (Weston et al., 2011).
We predicted that regardless of how they were affected by salinity, CH 4 fluxes would be coupled with CO 2 flux and metrics of organic C, which should fuel metabolic pathways producing CO 2 and CH 4 (Oikawa et al., 2017;Seo et al., 2014;Sutton-Grier et al., 2011).We also predicted positive correlations between CH 4 flux and ammonium and phosphate concentrations, pH, and the relative abundance of Firmicutes and Actinobacteriota.Ammonium and phosphate are important nutrients for microbial growth, low pH may inhibit certain methanogens (Van Kessel & Russell, 1996), and Firmicutes and Actinobacteriota are known to degrade complex carbon sources, which may supply substrates for methanogenesis (Gavande et al., 2021;Hartman et al., 2024).There was mixed support for these predictions.CO 2 and CH 4 were indeed coupled in two field study sites (San Francisco, Waccamaw), but not in Alligator.Notably, although CO 2 is generally produced in much higher quantities than CH 4 , it is also produced during acetoclastic and methylotrophic methanogenesis, but not hydrogenotrophic methanogenesis and the most abundant methanogens at Alligator were hydrogenotrophs.Different C, N, and P variables were measured among the sites/experiments, but there were some instances of positive correlations with CH 4 flux, such as DOC in San Francisco, and NH 4 + in Alligator soils.Ammonium can either increase or decrease CH 4 emissions based on the balance between its effects at the plant/ecosystem level, the microbial community level, and the biochemical level (Schimel, 2000).In Alligator soils, it was associated with increased CH 4 flux, possibly due to increased N available for plant and methanogen growth (Ardón et al., 2018).Phosphate was measured in two sites but was not significantly correlated with CH 4 flux, suggesting that it does not exert a strong influence on CH 4 in these coastal wetlands.Notably, these sites are likely not phosphate limited, which may partially explain this lack of effect; in P-limited systems phosphate abundance may be more important (Herbert et al., 2020;Ket et al., 2011).Methanogenesis relies on upstream depolymerization of polymeric organic matter, as well as degradation of fatty acids such as butyrate (Berrier et al., 2022).In this way methanogenesis is coupled to both photosynthesis and complex carbon degrading microorganisms.Metagenomic data has demonstrated positive correlations between hemicellulose and cellulose degrading genes and CH 4 flux, and that Firmicutes and Actinobacteriota were the dominant taxonomic groups containing those genes (Hartman et al., 2024).Firmicutes and Actinobacteriota were associated with increased CH 4 flux at both San Francisco and Alligator.They have also been implicated in the depolymerization of rice straw, particularly in cellulose and hemicellulose degradation, fueling methanogenesis in rice paddies (Gavande et al., 2021;Kausar et al., 2011); Actinobacteriota were also associated with methane production in Sphagnum peat bogs (Pankratov et al., 2006).A lack of relationship in Delaware and Waccamaw could be due to the lower overall abundance of Firmicutes in these sites; other taxa, including those in the Proteobacteria and Bacteroidetes phyla, may be responsible for complex C degradation in those locations.

Site-Specificity
What could be driving such discrepancies in CH 4 /salinity relationships among the sites/experiments?One clear difference is the low pH and high abundance of Acidobacteriota and Firmicutes in Alligator laboratory-incubated soils compared to Delaware laboratory-incubated soils (Figure 3, Figure S7 in Supporting Information S1).This is likely driven by broad differences in plant community composition between the sites, which affects organic carbon quality and quantity which in turn affect microbial communities and decomposition rates (Bueno de Mesquita et al., 2019;Megonigal et al., 2019;Shahbaz et al., 2017;Sutton-Grier et al., 2011).However, another laboratory experiment with soils from a tidal forested wetland similar to Alligator showed a decrease in CH 4 production with seawater addition, highlighting the need for more studies in forested wetlands to be able to make generalizations (Marton et al., 2012).Alligator also notably had the smallest and least rich methanogen population in all samples, particularly in oligohaline samples (Figure S4 in Supporting Information S1).While humic substances were not quantified, they are expected to be higher in the acidic Alligator wetland compared to Delaware Estuary wetlands, and this could suggest another mechanism for increases in CH 4 flux with salinity at that site, if salinity decreases the concentration of humics.
There is also support for the idea that site-specific microbial consortia that are functionally relevant have different responses to low salinity.While many broader taxonomic groups were shared among the four sites, relatively few were shared at the OTU level (Figure S11 in Supporting Information S1).A case in point are the methanogens in the most abundant methanogen family, Methanobacteriaceae (Figure 6).Most Methanobacteriaceae reads were from the hydrogenotrophic Methanobacterium genus, but even within the genus there were different OTUs and some evidence for different responses to oligohaline salinity.The idea that different species in the same genus, or even different strains of the same species, have different environmental preferences is not new, especially when considering pathogens (Thelaus et al., 2018), and has been greatly expanded upon in recent years with the help of full genome sequencing and metabolic modeling.For example, metabolic modeling of 55 Escherichia coli strains showed differences in niches among the strains (Monk et al., 2013).Such site-specific OTU niche differences and salinity responses, particularly of taxa involved in CH 4 cycling, can then contribute to discrepancies in CH 4 / salinity relationships.

Field Versus Lab Settings
Another potential source of discrepancy in CH 4 /salinity relationships among the five data sets examined here could arise from differences in field and laboratory experimental conditions.Field samples are, not surprisingly, characterized by greater richness and diversity than the laboratory samples taken after 3 months in the laboratory (Figure 2).Laboratory experiments also did not include plants and therefore do not take into account changes in C inputs as plants grow, release exudates, and senesce.To test for differences in the ecological dynamics between laboratory and field samples, we analyzed taxa overlap and nearest taxon index (NTI), and compared a presence/ absence metric (Jaccard) with an abundance-based metric (Bray-Curtis).Field samples are expected to have more unique taxa in freshwater and oligohaline conditions than samples incubated in the lab; field samples are mixed with the surrounding environment and microbes can disperse into the sampled soil from the surrounding environment, whereas samples incubated in the laboratory are cut off from the regional species pool after they are initially collected from the field.Dispersal and immigration are key processes in microbial community assembly (Sloan et al., 2006).Residence times of estuarine microbes in the field are on the order of days to weeks, indicating the potential for dynamic temporal turnover (Crump et al., 2004).
As expected, laboratory experiments had a higher degree of overlap in OTUs among salinity classes (Figure S8 in Supporting Information S1).However, differences in community composition among salinity classes were similar when using presence/absence and abundance-based metrics, and there were no differences among field and lab samples (Figure S10 in Supporting Information S1).Furthermore, virtually all samples, whether from field or lab, were dominated by deterministic assembly processes (Figure S9 in Supporting Information S1), even in the field where stochasticity could have potentially played a larger role due to dispersal and immigration.The NTI values >2 suggest phylogenetic clustering in most samples.Overall this suggests that similar ecological dynamics are at play in the lab in the field, namely the deterministic processes of selection, competition, and fitness differences among species (Vellend, 2010).However, the difference in alpha diversity is notable and may affect the degree of microbial community functional redundancy, and consequently the effects of salinity on C mineralization.

Conclusions
Salinization is a major issue caused by sea level rise, storms and tides, drought, water management, and waterbody connectivity, which leads to increased ionic strength, alkalinization, and sulfidation, which can result in coastal forest loss, species invasion, yield declines, eutrophication, marsh migration, and changes in microbial communities and biogeochemical cycling (Tully et al., 2019).An outstanding question regarding salinization of coastal wetlands is whether CH 4 emissions will increase or decrease, with important implications for C storage and climate change (Ardón et al., 2018).Our synthesis of microbial and biogeochemical data from multiple studies and methods from multiple geographic locations highlights both some consistencies in the wetland soil microbial communities that regulate CH 4 cycling, as well as some important differences that can qualitatively affect how CH 4 emissions respond to increases in low salinity, including microbial community responses and length of exposure.Key similarities at all sites included a high proportion of Proteobacteria, dominance of hydrogenotrophic methanogens, and general positive relationships between methanogens and the methanogen: methanotroph ratio and CH 4 flux.Some of the main differences were that salinity-CH 4 relationships differed in direction and magnitude, alpha-diversity did not always decrease with salinity, methanogenic genera and responses to salinity differed among sites, and methanotrophic guilds were not always negatively correlated with CH 4 fluxes.Importantly, our results do not support the assumption that seawater intrusion will generally decrease CH 4 emissions from coastal wetlands, at least in the short-term.Further systematic paired sampling of microbial communities and CH 4 flux measurements across a greater number of sites exposed to the same amount of salinity for the same amount of time is needed to directly analyze CH 4 /salinity relationship as a response variable, and to be able to make more generalizations about CH 4 /salinity relationships in tidal freshwater marshes and other coastal and estuarine wetlands.

Figure 1 .
Figure 1.Methane (CH 4 ) flux from the observational study (Obs), both field experiments, and both lab experiments.Note the difference in y-axis scales among panels and that each study used a different method for measuring CH 4 flux, so absolute values of fluxes are not comparable across panels, only within panels.Sample sizes per salinity class per study are 36 for San Francisco, 15 for Waccamaw, 4 for Delaware field, 4 for Delaware lab, and 7 for Alligator (Table 1).Different letters represent significant differences (t-test, p < 0.05).Data are only shown for cores with microbial data.Data from the full experiments at Waccamaw and Alligator are consistent with results shown here.Data from the full Delaware lab experiment show increased CH 4 flux in the oligohaline treatment.

Figure 2 .
Figure 2. OTU richness (a) and Shannon diversity (b) of bacterial and archeal soil communities across four different sites, two salinity classes, and two depth ranges.DE = Delaware, SF = San Francisco.Sample sizes per salinity class per study are 36 for SF, 15 for Waccamaw, 4 for Delaware field, 4 for Delaware lab, and 8 or 7 for Alligator (Table1).Different letters represent significant pairwise comparisons within each panel (Tukey posthoc, p < 0.05).The x-axis is sorted by increasing ASV richness by study (combined freshwater and oligohaline data for each study), with freshwater on the left and oligohaline on the right, for each study.Study type (lab, field, or observational (Obs)) is stated in the x-axis labels.

Figure 3 .
Figure 3. Principal coordinates analysis of Bray-Curtis dissimilarities at the OTU level for bacterial and archeal soil communities for all data from all four estuaries (a), San Francisco Bay field salinity gradient (b), Delaware River estuary field experiment (c), Waccamaw River estuary field experiment (d), Delaware River laboratory incubation experiment (e), and Alligator River estuary laboratory incubation experiment (f).Sample sizes per salinity class per study 36 for SF, 15 for Waccamaw, 4 for DE field, 4 for DE lab, and 8 or 7 for Alligator (Table1).Note that salinity and CH 4 were the only two continuous variables measured in all five studies (shown as vectors in panel a).

Figure 5 .
Figure 5. Correlations between chemical and microbial variables and methane flux.Rows are annotated by type (flux, porewater or soil), variable (chemical or microbial), and the predicted relationship with methane (positive or negative).Blocks are labeled with hypotheses H3, H4, or H5, which correspond to the stated hypotheses about methanogens/ methanotrophs, alternate electron acceptors, and overall decomposition.The heatmap shows Spearman's rho values and asterisks indicate if the test was statistically significant ( p < 0.05).Guild abbreviations are given in Figure 4 caption.Gray cells indicate data not present.Data shown here from DE are from the field experiment only.MG:MT = methanogen: methanotroph ratio, SR = sulfate reduction rate; for chemical abbreviations see the methods text.

Figure 6 .
Figure 6.Methanobacteriaceae percent relative abundances for each individual sample in the laboratory incubation experiments conducted with wetland soils from the Alligator River and Delaware River estuaries.

Table 1
Summary of the Different Sources of Data Combined in This Paper , using the "Get_16S_guilds_alt" function in a publicly available R script (https://github.com/cliffbueno/SF_microbe_methane/blob/main/modules/AssignGuilds.R).Sequencing depth was 131,845 ± 3373 SE sequences per sample, and data were rarefied to 31,264 sequences per sample.Only one sample from San Francisco with very few reads (1,296) was dropped.The final sample size analyzed was 133 (Table

Table 2
Statistical Results for Microbial Alpha-and Beta-Diversity Across the Whole Data Set Note.B-C = Bray-Curtis.