INTRODUCTION

Wetland ecosystems have a disproportionately large role in the global carbon cycle and are responsible for the storage of up to 70% of terrestrial organic carbon (Mitra et al., 2005). Wetlands are critical for organic carbon storage because they create anoxic benthic environments in highly productive ecosystems, leading to slow decomposition, and net accumulation of carbon (Jackson et al., 2014). Understanding carbon processing requires multiple ecological frameworks, including stoichiometric, metabolic, and redox constraints to capture limiting factors for the assimilatory and dissimilatory processes of microbial decomposers (Brown et al., 2004;Helton et al., 2015). The limitations imposed on decomposers is a key topic of study in wetland restoration, as the combination of organic matter (carbon limitation) and environmental chemistry (macronutrient limitation, redox limitations) will determine the degree of organic matter storage as compared to mineralization, controlling the development (or loss) of wetland soils (Chambers et al., 2019).

When considering energetic limitations to microbial communities, the bioavailability of carbon can play an important role, occasionally leaving communities in carbon-rich systems to become limited not by total carbon, but by biologically available carbon (Berggren et al., 2015;Soares et al., 2017). Additionally, recalcitrant carbon can be bound to other limiting nutrients, which can prime organic matter mineralization for the purpose of releasing bound nutrients (Guenet et al., 2010;Howard-Parker et al., 2020). Macrophyte and microbial communities change along freshwater to marine gradients, which can lead to significant changes in the quality and quantity of organic matter deposited on and contributing to wetlands soils (Servais et al., 2020;Smith et al., 2019;Zhao et al., 2023). Coastal wetlands are dynamic systems, where both tidal and seasonal changes in marine influence are key drivers of the system, but long-term increases in salinity caused by saltwater intrusion have the potential to drastically change the system, with changes to both macrophyte communities and the total chemistry of the system affecting the breakdown of organic matter (Neubauer, 2013;Tully et al., 2019). The press of saltwater intrusion has been found to reduce microbial diversity but to enhance the abundance of microbial taxa taking advantage of marine subsidies, raising questions about the consequences of these shifting community structures on the mineralization of organic matter (Mobilian et al., 2023).

Saltwater intrusion exposes microbial decomposers to a series of novel stressors (osmotic stress) and subsidies (macronutrients, sulfates) that have been linked to increases in the breakdown of organic matter in coastal wetlands (Morrissey, Gillespie, et al., 2014;Neubauer et al., 2019). As marine water brings iron and sulfates to the environment, changing redox conditions which are especially important in anaerobic wetland soils, can constrain microbial activity when highly reducing environments cause the availability of terminal electron acceptors to become limiting (Falkowski et al., 2008;Tully et al., 2019). Redox fluctuations can be ecologically important to the function of wetlands, such as when systems are shifted toward, or away from the production of methane, which plays an important role in the global carbon cycle, or the use of sulfate as alternate electron acceptors (Dean et al., 2018;Helton et al., 2015). Redox conditions are especially important in wetlands which are characterized by both temporal and spatial limitations to oxygen availability, and even in wetlands with higher oxygen there is often both spatial and temporal heterogeneity in oxygen availability (Lacroix et al., 2023). Microbes can circumvent oxygen limitation of respiration through the use of alternative electron acceptors of varying efficiencies, which leads to especially tight linkages wetland carbon (such as dissolved organic carbon) and nitrates which are a highly efficient electron acceptor (Helton et al., 2015;Taylor & Townsend, 2010). After the pool of nitrates are reduced microbes begin to utilize iron and sulfates as terminal electron acceptors for redox reactions (Froelich et al., 1979). Understanding how the relative availability of carbon, nutrients, and redox conditions affect microbial communities will be important for predicting how changing conditions will change the processing of carbon, and where limitations will change for microbial communities.

We understand that microbial communities vary within coastal wetlands, and that carbon lability plays a major role in its breakdown (Berggren et al., 2015;Smith et al., 2019). Similarly, we understand that redox gradients along freshwater to marine gradients play an important role in driving microbial activity (Morrissey, Gillespie, et al., 2014;Neubauer et al., 2019). What we don't understand is how microbial communities respond to the interactions between shifting carbon lability, nutrient availability, and shifting redox gradients across different litter species caused by sea-level rise.

In this study, we investigated the relative importance of carbon quality, litter stoichiometry, and environmental stressors and subsidies to the function and composition of benthic microbial communities, and their role in the mineralization of organic matter. The Florida Coastal Everglades (FCE) is the ideal environment for this study, as it has macrophytes that produce litter of variable quality, is experiencing sea-level rise that is changing redox conditions where seawater is intruding and has two major drainages that vary in elemental stoichiometry and relative phosphorus limitation. We used the FCE as a model ecosystem to study how variations in organic matter quality drives the composition and function of microbial communities across freshwater to marine gradients. To better understand the drivers of the mineralization of organic matter and the structure and function of microbial functional groups we asked three questions: (1) How do benthic microbial functional groups and anaerobic processing of carbon change across different litter species with distinct carbon lability and nutrient concentrations? (2) How do benthic microbial functional groups and anaerobic processing of carbon change along freshwater to marine gradients? (3) How do benthic microbial functional groups and anaerobic processing of carbon vary seasonally as levels of freshwater influence vary? We predicted that anaerobic activity would increase with the availability of sulfate and phosphorus, and where low flow and minimal tidal exchange facilitated anoxic conditions.

METHODS

Site description and experimental design

We deployed litterbags along freshwater to marine gradients in wetlands in Everglades National Park (Florida, USA), an International Biosphere Reserve, a World Heritage Site, and a Ramsar Wetland of International Importance. Everglades wetlands are highly oligotrophic and spatially heterogenous with wide variation in hydrology, productivity, and relative nutrient limitation (Castañeda-Moya et al., 2013;Childers et al., 2003;Noe et al., 2001). Sea-level rise currently introduces both subsidies and stressors into the ecosystem, which are shaping and altering macrophyte communities along coastal gradients (Tully et al., 2019). One major subsidy is phosphorus, which is higher in the marine waters of Florida Bay and the Gulf of Mexico that contribute to the major flow paths of the Everglades than that of the extremely oligotrophic freshwater wetlands further inland (Boyer, 2006;Fourqurean & Zieman, 2002). Phosphorus concentrations have increased in coastal estuarine sites over time, but disentangling the direct effects of sea-level rise is difficult because of concurrent restoration efforts that also introduce subsidies and stressors to the system (Dessu et al., 2018). Marine water additionally brings subsidies of sulfate, which can shape soil microbial communities (Pester et al., 2012;Zhao et al., 2023).

For the past 20 years, the FCE Long Term Ecological Research (LTER) program has studied the ecology of coastal wetlands along salinity and phosphorus gradients. We performed a reciprocal transplant of four different litter species across seven FCE-LTER sampling sites within the two major drainages of the Everglades: Shark River Slough (SRS) and Taylor Slough/Panhandle (TS/Ph; Figure 1). SRS is a high-productivity, long-hydroperiod wetland that transitions from sawgrass-dominated ridge and slough peat marshes, to tidal riverine mangroves (Castañeda-Moya et al., 2013;Childers et al., 2006;Ewe et al., 2006). TS/Ph is a lower productivity, short-hydroperiod wetland that transitions from sawgrass and periphyton-dominated marl prairies to microtidal mangrove scrub forests. SRS has higher concentrations of phosphorus, and in the ecotone and mangrove forests, has significant diurnal exchanges of marine water (Cawley et al., 2014). TS/Ph has a more seasonal pattern of marine inputs into the mangrove ecosystems (Anderson et al., unpublished data). We deployed litterbags in December 2020 at three sites in each drainage corresponding to marsh (SRS-2, TS/Ph-2), ecotone (SRS-4, TS/Ph-3), and mangrove (SRS-6, TS/Ph-7), and a single site in the seagrass meadows in Florida Bay (TS/Ph-10). We collected samples after 1 month (January 2021) and 4 months (April 2021).

Surface water physicochemistry

We averaged data collected every 3 days from ISCO autosamplers for the month that litter was collected at all sites except for TS/Ph-10 for concentrations of total nitrogen (TN), total phosphorus (TP), and salinity (Gaiser & Childers, 2022;Troxler, 2022aTroxler, , 2022b)). At TS/Ph-10, we averaged data collected from monthly grab samples during the month that litter was collected (Briceno, 2020). TP was measured following the method of Sol orzano and Sharp (1980). TN and total carbon (TC) were measured using an Antek TN analyzer (Antek Instruments, Houston, Texas, USA). All water chemistry analyses were conducted by the CREST CAChE Nutrient Core Facility who are NELAC Certified for non-potable water-General Chemistry under State Lab ID E76930. Surface water temperature was measured at the locations of gas flux towers for SRS-2, SRS-6, TS/Ph-1, TS/Ph-7, and TS/Ph-10. Missing data were estimated using water temperatures at the closest tower: water temperature data at TS/Ph-1, SRS-6, and TS/Ph-7 were used as an estimate for TS/Ph-2, SRS-4, and TS/Ph-3, respectively.

Litter breakdown and chemistry

Eleocharis cellulosa (spikerush) and Cladium jamaicense (sawgrass) are the dominant species in Everglades freshwater marshes, Rhizophora mangle (red mangroves) dominates at mangroves and coincides with higher salinity, and Thalassia testudinum (seagrass) is the dominant species in seagrass meadows of Florida Bay. We collected litter to be deployed as live stems from SRS-2 (sawgrass and spikerush), SRS-6 (red mangrove), and TS/Ph-10 (seagrass). All litter was air dried for at least one week and weighed prior to being sealed into litterbags, except for seagrass. Seagrass was not dried as drying makes seagrass very brittle and can accelerate mass loss through breakage. The wet mass of seagrass was measured for each litterbag and we oven-dried a subset of seagrass to calculate the ratio of dry to wet mass of litter deployed in each litterbag. All deployed seagrass was collected within 48 h of being deployed and was stored at 4 C until being deployed.

We deployed litterbags with litter of a single species inside coarse mesh (5 mm) bags with spikerush, sawgrass, or mangrove litter at each of the seven sites, or seagrass at each mangrove site (TS/Ph-7, SRS-6) and in Florida Bay (TS/Ph-10). We deployed four litterbags at two sub-sites for each site for a total of n = 12 litterbags of each species deployed to each site. We collected 2 litterbags from each subsite at 1 and 4 months after deployment. For each recovered litterbag, we rinsed the litter with deionized water and removed a subset of litter for transcriptomics, or productivity measurements. We measured the wet mass of both the subset and remaining coarse mesh litter to correct for the total dry mass. After rinsing and taking subsets, we oven-dried all litter samples at 45 C for at least 3 days. After measuring the dry mass remaining, we ground the dry litter using a ball mill (SPEX Certiprep 8000D, Metuchen, New Jersey, USA). We calculated the ash-free dry mass (AFDM) for each sample by combusting a subsample at 550 C and measuring the remaining mass. We calculated the breakdown rate (k) for each sample as: k = ln (AFDM t / AFDM 0 )/days incubated. We calculated degree-days as: degree-days = summed daily mean water temperature (in degrees Celsius)/number of days deployed. We then scaled breakdown rate (k) by temperature as: k/degree-day.

We measured TN and TC for each litter sample using a CE Flash 1112 Elemental Analyzer. We measured TP using a UV-2101 Shimadzu Spectrophotometer using a modified colorimetric method (Sol orzano & Sharp, 1980). We measured the recalcitrance of each litter sample with ramped pyrolysis, combusting at a series of thermal intervals associated with the loss of different qualities of carbon. We combusted at four intervals: T1: 180 C (hemicellulose), T2: 300 C (cellulose) T3: 400 C (lignin), 550 C (inorganic carbon), using thermal intervals validated by Trevathan-Tackett et al. (2017). After each combustion we measured the mass lost and the percentage of the remaining material that was carbon using a CE Flash 1112 Elemental Analyzer (Thermo Scientific, Waltham, Massachusetts, USA). We calculated the percent carbon lost at each thermal interval as the mass of carbon post-combustion divided by the mass of carbon pre-combustion × 100%.

Bacterial productivity

We measured the maximal bacterial productivity of each litter type as the uptake of tritiated thymidine, using a modification of the method by Wetzel and Likens (2000). All litter samples were returned from the field on ice and were processed within 24 h of collection. We homogenized a known mass (~0.2 g) of each litter sample in 200 mL deionized water using a small bullet blender. We added 10 mL of the homogenized litter into four clean falcon tubes, immediately fixing one with 2% bacteria-free formalin, for n = 3 live replicates and a killed control sample. We added tritiated thymidine to each sample for a final concentration of 10 nM and incubated for 1 h. After incubation, we halted microbial activity with the addition of 50% trichloroacetic acid, and incubated samples at 0 C for 15 min. After the final incubation we filtered each sample and after 24 h, ran on a liquid scintillation counter (Beckman Model 3801, Beckman Instruments, Fullerton, CA). The values for the blanks were subtracted from each sample to calculate the mols of thymidine taken up per gram of material per hour.

Microbial molecular analyses

For each site and litter pair we collected a subset of 2-3 g wet mass of litter, a grab sample of soil, and a grab sample of periphyton for each site. All subsamples were preserved at -20 C until extraction, which took place up to a year after initial collection. Samples were sent to Novogene (Novogene Co. Ltd., Beijing, China) for the total RNA extraction followed by metatranscriptome sequencing. Briefly, the total RNA was extracted using TRIzol reagent (Rio et al., 2010) and the quality and quantity of the RNA were assessed using the Agilent 2100 bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) and Nanodrop ND-1000 (ThermoScientific, Waltham, MA, USA), respectively. After the total RNA samples passed the quality check, cDNA libraries were prepared from total RNA using poly(A) enrichment of the mRNA to remove rRNA resulting in the construction of 250-300 bp insert cDNA libraries and sequenced by paired-end (PE) sequencing (PE 2 × 150 bp) using an Illumina NovaSeq 6000 platform (NovaSeq Reagent Kits, Illumina, Inc., San Diego, CA, USA).

Raw reads were processed using the Simple Annotation of Metatranscriptomes by Sequence Analysis 2.0 (SAMSA2) pipeline (Westreich et al., 2018) with slight modification. Briefly, low-quality bases were trimmed using Trimmomatic v0.39 (Bolger et al., 2014) and overlapping paired-end reads were merged into single sequences using PEAR v0.9.11 (Zhang et al., 2014). Ribosomal RNA reads were removed with SortMeRNA v2.1 (Kopylova et al., 2012) and the cleaned transcripts were annotated by DIAMOND v0.9.36 (Buchfink et al., 2021) against the National Center for Biotechnology Information Reference Sequence (RefSeq) database (O'leary, 2016) for taxonomic and functional characterization. The resulting annotation files were aggregated and merged with custom Python and R scripts included in the SAMSA2 pipeline (Westreich et al., 2018).

We selected n = 12 genes/gene families encoding for focal enzymes to investigate which are important to the breakdown of organic matter: Dioxygenases (associated with aerobic respiration), sulfatases (associated with the release of sulfates from complex molecules), sulfite reductases (associated with sulfite reduction), methyl coenzyme M reductase and formylmethanofuran (associated with methanogenesis), nitrite reductases (associated with nitrite reduction), cellobiosidase, glucosidase, and xylosidase (associated with cellulose breakdown), phenol oxidase (associated with lignin breakdown), acid phosphatase (associated with phosphate acquisition in acidic environments), and alkaline phosphatase (associated with phosphate acquisition in basic environments; Table 1). For each gene/family of interest, we searched all annotated transcripts for all entries corresponding to that gene/family and combined all values for a total expression.

We selected n = 6 monophyletic microbial functional groups, representing sulfate reducers, sulfate oxidizers, methane oxidizers, methanogens, nitrite oxidizers, and ammonia oxidizers associated with sulfate and methane cycling. We filtered all annotated transcripts for all species with the following in the name: "desulfo" for sulfate reducers, "sulfito" for sulfite oxidizers, "methylo" for methyl/methane oxidizers, "methano" for methanogens, "nitro" for nitrite oxidizers, and "nitroso" for aerobic ammonia oxidizers.

Data analyses

We used principal components analysis to synthesize variation in litter chemistry between January and April sampling. Litter chemistry data was scaled and centered prior to principal components analysis. We used two-way ANOVA to test the effects of site and species on bacterial productivity after one and four months of litter breakdown. We fit linear mixed-effects models to explain litter breakdown rates with month as a random effect, and site as a random effect nested within drainage with the lme4 package in R (Bates et al., 2015). We calculated corrected Akaike information criterion (AIC c ) for each model and selected the best-fitting model as the one with the highest adjusted R 2 score that had a ΔAIC less than or equal to 2. We calculated coefficient estimates for each fixed effect in the best-fitting model and tested for significant relationships between coefficients and our response variable with the LmerTest package in R (Kuznetsova et al., 2017). We fit the model with species, site TN, site TP, site DOC, and site salinity as fixed effects. We used linear regressions to test significant correlations between site chemistry, breakdown rates, microbial activity, and microbial functional group abundances. For nonlinear relationships we fit generalized additive models with a cubic regression spline using the package "mgcv" (Wood, 2017). We performed all analyses with R version 4.2.0 (R Core Team, 2022). All plots were constructed with the "ggplot2" package (Wickham, 2016).

RESULTS

Surface water physicochemistry

Phosphorus concentrations in surface waters were high and increased from freshwater to the coast in January 2021 in SRS with averages (±SD) ranging from 0.84 (±0.12) μmol/L in the freshwater marsh to 1.74 (±2.09) μmol/L in the coastal mangroves, but decreased from freshwater to the coast in TS/Ph with averages ranging from 0.67 (±0.05) μmol/L in the freshwater marsh to 0.37 (±0.11) μmol/L in the coastal mangroves. April phosphorus concentrations increased from freshwater to coastal in both drainages, but were marginally lower than January in SRS, ranging from 0.74 (±0.30) μmol/L in the freshwater marsh to 1.80 (±1.03) μmol/L in the coastal mangroves, and TS/Ph ranging from 0.27 (±0.03) μmol/L in the freshwater marsh to 0.56 (±0.22) μmol/L in the coastal mangroves. Florida Bay had the lowest total phosphorus concentration with an average of 0.32 μmol/L, but our site is in a notably nutrient poor pocket of Florida Bay (Appendix S1: Table S1). Salinity significantly increased moving from freshwater to more marine sites and was significantly higher in SRS ecotone and mangrove sites than the corresponding TS/Ph sites (ANOVA, F(5,54) = 38.86, p < 0.001; Appendix S1: T A B L E 1 We selected n = 12 gene families for focal enzymes to investigate, which are important to the breakdown of organic matter.

Gene of interest Microbial usage

Aerobic respiration Dioxygenases

Releases sulfates from complex molecules Sulfatases Sulfite reduction Sulfite reductases Methanogenesis Methyl coenzyme M reductase, formylmethanofuran Nitrite reduction Nitrite reductases Cellulose breakdown Cellobiosidase, glucosidase, xylosidase Lignin breakdown Phenol oxidase Phosphate acquisition Acid phosphatase, alakline phosphatase Note: We selected enzymes involved in 5 metabolic pathways: Oxygen metabolism, sulfate metabolism, methane metabolism, carbon degradation, and phosphorus acquisition.

Table S1). Temperature was not significantly different among sites (ANOVA, F(5,54) = 1.41, p = 0.24).

Litter chemistry and breakdown

We synthesized metrics of litter chemistry using principal components analysis of seven measured parameters: percent total carbon (%TC), percent phosphorus (%P), percent nitrogen (%N), and estimates of carbon quality measured by ramped pyrolysis: % hemicellulose, % cellulose, % lignin, % inorganic carbon (Figure 2). In January, the first principal component explained 51% of the variation, driven equally by %N (22%), % cellulose (20%), % lignin (19%), % inorganic C (19%), and %P (19%). The second principal component explained 22% of the variation and was primarily driven by % hemicellulose (53%), and %TC (23%). In April, the first principal component explained 51% of the variation, driven equally by % cellulose (24%), % TC (19%), % lignin (18%), and % inorganic C (18%). The second principal component explained 28% of the variation and was primarily driven by %N (42%), and %P (28%). Nutrient concentrations varied between litter species, but not over time. Across both months average (±SD) percent carbon was highest in Cladium (45.3 ± 3.0) and Rhizophora (44.6 ± 4.8), followed by Eleocharis (40.9 ± 5.5), and then Thalassia (32.2 ± 4.9; ANOVA, F (3,183) = 53.37, p < 0.001). Percent nitrogen was higher in Thalassia (2.0 ± 0.6) and Rhizophora (2.0 ± 0.4) compared with Cladium (0.7 ± 0.2) and Eleocharis (0.8 ± 0.1; ANOVA, F (3,183) = 219.5, p < 0.001). Percent phosphorus was highest in Rhizophora (0.08 ± 0.02), followed by Thalassia (0.06 ± 0.02), Cladium (0.02 ± 0.00), and Eleocharis (0.03 ± 0.02; ANOVA, F (3,183) = 189.7; p < 0.001).

Litter breakdown rates were higher in January than April. Average breakdown rates (±SD) were highest in Thalassia (0.0008 ± 0.0002) followed by Rhizophora (0.0006 ± 0.0001) and Eleocharis (0.0006 ± 0.0002), and finally Cladium (0.0003 ± 0.0001; Figure 3). Variation within sites was highly variable, as each species responded differently across sites.

We found that breakdown rates were best explained by species and salinity. To understand how species and site characteristics interact to drive breakdown we fit linear mixed-effects models with site and month as random effects, and included species, TN, TP, and salinity as fixed effects. We fit the model with species as a factor as opposed to litter chemistry because litter species clustered in our principal components analysis that had distinctly different breakdown rates (e.g., Cladium and Eleocharis), while breakdown rates were similar for species with distinct chemistry (e.g., Eleocharis and Rhizophora). This suggests that species identity played a more important role than our measures of litter chemistry. The best-fitting model included species and salinity, but only species was a significant predictor of breakdown rate (Table 2).

Bacterial productivity

We found no significant difference in maximal bacterial productivity of litter between either species (ANOVA, F (2,37) = 1.90; p = 0.16), or site (ANOVA, F (6,33) = 1.41; p = 0.24) after the first or for species (ANOVA, F (2,39) = 0.45; p = 0.64), or site (ANOVA, F (6,35) = 2.62;

F I G U R E 2 Principal components (PC) analysis of litter chemistry of four litter species (Eleocharis cellulosa, Rhizophora mangle, Cladium jamaicense, and Thalassia testudinum) after one month (January) and four months (April) of incubation. We constructed PC from seven measured parameters: Percent total carbon (%TC), percent phosphorus (%P), percent nitrogen (%N), and estimates of carbon quality measured by ramped pyrolysis: % hemicellulose, % cellulose, % lignin, % inorganic carbon. p = 0.07) after the fourth month of litter breakdown. There was no significant correlation between BP and environmental variables including salinity (R 2 = 0.00, p = 0.94), total nitrogen (R 2 = 0.05, p = 0.18), total phosphorus (R 2 = 0.01, p = 0.45), or dissolved organic carbon (R 2 = 0.01, p = 0.68).

Relative gene abundance

Gene abundance varied in the degree of aerobic versus anaerobic genes expressed between soils, litter, and periphyton. We detected the presence of all genes of interest except for methyl coenzyme M reductase (Table 1). Most genes related to anaerobic pathways were more abundant in soils than in litter and periphyton, including sulfatases (ANOVA, F (2,30) = 3.58, p < 0.05) and formylmethanofuran (ANOVA, F (2,30) = 13.72, p < 0.001; Figure 4). Dioxygenases, which indicate aerobic pathways, were more abundant in litter than in soil (ANOVA, F (2,30) = 3.11, p < 0.05), as were nitrite reductases (ANOVA, F (2,30) = 18.09, p < 0.001).

To test how seasonal variation in freshwater affected microbial communities along freshwater to marine gradients we tested relationships between salinity and gene expression. After the first month of breakdown there was a significant positive correlation between salinity and sulfatase gene abundance, and a significant negative correlation between salinity and acid phosphatase gene abundance (Figure 5; Appendix S1: Table S2A). There effect Estimate SE df t value p value Intercept 2.59E-04 1.48E-04 1.11E+00 1.752 0.311 Species: Eleocharis 1.87E-04 2.91E-05 1.78E+02 6.441 1.07E-09 Species: Rhizophora 2.15E-04 2.89E-05 1.78E+02 7.424 4.53E-12 Species: Thalassia 4.05E-04 3.96E-05 1.81E+02 10.233 <2E-16 Salinity -3.31E-06 1.85E-06 5.88E+00 -1.79 0.125 Note: Site and month were both included as random effects, where site was nested within drainage. The full model contained species, salinity, total nitrogen, total phosphorus, and DOC as fixed effects, but the best-fitting model only contained salinity and species as fixed effects.

was no significant relationship between salinity and glucosidase, cellobiosidase, nitrite reductase, alkaline phosphatase, xylosidase, phenol oxidase, sulfite reductases, dioxygenase or formylmethanofuran (Appendix S1: Table S2A).

After four months of breakdown, there was a significant negative correlation between salinity and glucosidase and xylosidase (Figure 5; Appendix S1: Table S2B). There was a significant nonlinear relationship between salinity and sulfite reductase with a peak at intermediate salinity (Figure 5; Appendix S1: Table S2B). There was a weak positive correlation between salinity and sulfatase, and a weak negative correlation between salinity and phenol oxidase (Appendix S1: Table S2B). There was no significant relationship between salinity and dioxygenase, formylmethanofuran, nitrite reductase, cellobiosidase, alkaline phosphatase, and acid phosphatase (Appendix S1: Table S2B).

After the first month of breakdown, there was a weak positive correlation between average water column phosphorus concentration and dioxygenase (Appendix S1: Table S3A). There was no significant relationship between water column phosphorus concentration and any other variables. After four months of breakdown, there was a weak negative correlation between average water column phosphorus concentration and glucosidase (Appendix S1: Table S3B).

There were no significant correlations between any of gene family and k/degree day after one month (Appendix S1: Table S4A), or four months of litter breakdown (Appendix S1: Table S4B).

Microbial functional groups

We investigated organismal abundance based on the relative gene abundance for six monophyletic microbial groups, each representing sulfate reducers, sulfite oxidizers, methane oxidizers, methanogens, nitrite oxidizers, and ammonia oxidizers. We found significantly higher abundance of sulfite reducers (ANOVA, F (2,63) = 6.28, p < 0.01), methane oxidizers (ANOVA, F (2,63) = 7.88, p < 0.001), methanogens (ANOVA, F (2,63) = 6.86, p < 0.01), nitrite oxidizers (ANOVA, F (2,63) = 15.99, p < 0.001), and ammonia oxidizers (ANOVA, F (2,63) = 44.36, p < 0.001) in soils than in litter. There was no significant difference in abundance between litter and soils for sulfite oxidizers (ANOVA, F (2,63) = 0.49, p = 0.61).

Of the microbial functional groups we investigated, the relative gene expression of sulfite reducers after both one (11.64-0.31) and four months (16.23-0.03) had the highest gene abundances but were highly variable across sites. Methane oxidizers for one month (2.56-1.02) and four months (2.93-0.43), and nitrite oxidizers for one month (0.31-4.28) and four months, (0.06-3.13) were also abundant but more even across sites. Ammonia oxidizers (0.007-1.10), sulfite oxidizers (0-0.76), and methanogens (0-0.03) had lower relative gene expression.

After the first month of breakdown, there was a significant positive correlation between salinity and relative gene expression of sulfite oxidizers and ammonia oxidizers on leaf litter (Appendix S1: Table S5A). There was no correlation between salinity and relative gene abundance of sulfite reducers, methane oxidizers, methanogens, or nitrite F I G U R E 4 Gene expression for sulfatases, dioxygenases, and formylmethanofuran across leaf litter, periphyton, and soil along Everglades freshwater to marine gradients. Boxplots represent the interquartile range for each variable, and the solid line is the median. Error bars represent the 95% CIs. Letters above boxplots indicate significant (α ≤ 0.05) differences among groups using one-way ANOVA.

oxidizers on leaf litter or soil (Appendix S1: Table S5A). After four months of breakdown, there was a significant positive correlation between salinity and relative gene expression of sulfite oxidizers on leaf litter (Appendix S1: Table S5B). After four months, there was no correlation between salinity and relative gene abundance of ammonia, sulfite reducers, methane oxidizers, methanogens, or nitrite oxidizers on leaf litter or soil (Appendix S1: Table S5B).

DISCUSSION

Our goal was to understand how site-specific and litter-specific environmental factors interact to drive microbial processing of litter, with a specific focus on salinity, both directly and as a proxy for marine subsidies, and phosphorus. Our results suggest that litter quality in terms of percent carbon, nitrogen, and phosphorus does not play a role in litter breakdown, but litter species S2.

identity does. Additionally, we found that salinity somewhat explained breakdown rates, and was a major driver of microbial activity related to sulfate uptake and reduction. Salinity, marine subsidies, and phosphorus all can play major roles in driving both the structure and function of coastal microbial communities (Guevara et al., 2014;Ikenaga et al., 2010;Lozupone & Knight, 2007;Sundareshwar et al., 2003). We found that variation in litter breakdown was largely driven by litter species, and while the best-fitting model included salinity, it was not a significant predictor of breakdown rates. Our data show that bacterial productivity did not vary significantly across our samples suggesting that microbial communities were at early stages of establishment across our samples. This combination of species-specific differences driving breakdown, a limited effect of salinity, and limited variation in microbial productivity suggests that early breakdown rates were largely influenced by leaching as opposed to microbial activity (Benner & Hodson, 1985;Berg & Staaf, 1981). In contrast, we found a distinct relationship between salinity and both microbial gene transcripts and functional group abundances, but we did not find a relationship between microbial gene transcripts or functional group abundances and litter species. Our results suggest that environmental chemistry is largely determining the initial makeup of the functional groups breaking down litter in coastal wetlands, which can play a significant role in determining later rates of breakdown (Bradford et al., 2016). However, we did not find a significant role of the metrics of litter chemistry that we measured.

We found that salinity and associated subsidies were the major environmental factors shaping microbial activities, and our results show negative relationships between labile carbon-degrading enzymes (glucosidase) and salinity, but not enzymes responsible for breaking down more recalcitrant carbon. Salinity is often found to inhibit enzyme activity, including carbon-degrading enzymes (Servais et al., 2020;Yun et al., 2010). The impact of salinity on carbon processing, however, is still unclear with contradictory findings of both increased (Saviozzi et al., 2011;Wang et al., 2018) and decreased carbon processing in response to increasing salinity which could be related to variation in carbon recalcitrance (Neubauer, 2013;Rejm ankov a & Houdkov a, 2006). Our work highlights the need for a more in-depth understanding of the microbial contribution to mineralization of carbon in changing coastal wetlands.

Salinity is often coupled with increasing sulfate concentrations, which can shift dominant anaerobic pathways to the more efficient sulfate production from methanogenesis, potentially leading to increases in organic matter mineralization (Neubauer, 2013;Neubauer et al., 2005;Weston et al., 2011). Our data showed that despite higher freshwater inflows (as indicated by higher water depths) in January, salinity concentrations were similar to April. Marine influence increases during pulsed tidal events at a shorter timescale than our measurements of salinity, indicating higher marine subsidies to our litter samples in April. This is further supported by the changing seasonal phosphorus dynamics, where April phosphorus concentrations are higher at marine sites from both sloughs, and phosphorus trends at Taylor Slough switch to increase in concentration closer to the coast in April where we predict higher marine subsidies. In our study, we measured sulfatases, which indicate sulfate acquisition, and sulfate reductase, which indicates dissimilatory reduction of sulfate (Morrissey, Gillespie, et al., 2014). We found that sulfate acquisition was strongly correlated with salinity during January when water levels are higher in coastal sites. In April when there are lower water levels we found no correlation with sulfate acquisition. We found no relationship in January between dissimilatory sulfate reduction and salinity; however, in April there was high activity at intermediate salinity sites (Morrissey, Gillespie, et al., 2014;Servais et al., 2021). The correlation with salinity across space allows us to understand how increasing marine influence along the gradient affects microbial activity, while the comparison between January and April allows us to test the effect of decreasing freshwater inputs, which is likely associated with increases in tidally pulsed marine subsidies, along the coastal wetland gradient. This may indicate that sulfate acquisition is most limiting to microbial communities (and thus most expressed) when freshwater inputs are high and marine influence is low because sulfate is less available. However, we find that sulfate is being actively used for dissimilatory reduction when water depths decrease, and marine influence decreases. This suggests that dissimilatory reduction of sulfate is seasonally limited, primarily happening when marine influence brings in sulfate at low freshwater flow. The difference in expression of sulfate acquisition and reduction enzymes could also be related to the relative redox conditions at our sites. Sulfatase activity was significantly higher in our soils than litter, while dioxygenases (which require the presence of oxygen) were higher, suggesting that while not associated with the expression of sulfite reductase, sulfatases are being produced by more anaerobic communities.

Although Everglades ecosystems are highly phosphorus-limited, the activities of microbial communities are also highly constrained by redox conditions. Our results showing no correlation between phosphatase activity and salinity suggests that marine phosphorus subsidies do not play a major role in regulating the microbial functional group abundances, and the lack of a relationship found with phosphorus concentrations may suggest that microbes are not acquiring phosphorus from the water column, but from the litter. We found that alkaline phosphatase was nearly 10 times more expressed than acid phosphatase, likely because the Everglades landscape is underlain by limestone and therefore buffered by bicarbonate. The lack of a relationship between alkaline phosphatase and water column phosphorus concentrations could also occur because Everglades is extremely phosphorus limited, and phosphatase expression may be saturated even at sites with relatively higher concentrations of phosphorus (Price et al., 2006). While phosphorus certainly plays a role in the breakdown of litter, our data may suggest that in highly phosphorus-limited systems like the Everglades, variation in redox conditions and carbon limitation may play similarly important roles for the microbes colonizing leaf litter (e.g., Helton et al., 2015).

When observing the composition of anaerobic microbial communities, we found high abundance of sulfite reducers, which were up to 6% of the microbial community. Sulfite reducer abundance was highest in marine sites and was especially high at TS/Ph-7, which is both stagnant and shaded; likely one of the sites experiencing anoxic conditions most frequently. The combination of the taxonomic makeup of the microbial communities, along with the variation in gene expression we see, suggests that Everglades surface soils and leaf litter experience periodic anaerobic conditions that sulfite and sulfate reducers exploit where sulfate is available, but also oxygen is frequently available on the soil surface. This is further supported by clear gradients of sulfate reducers at these same sites, but in deeper soils (Ikenaga et al., 2010;Zhao et al., 2023). As redox and nutrient conditions of coastal wetlands change, we are likely to see shifts in microbial community structure and function as well (Cheung et al., 2018;Maietta et al., 2020).

Our results show that microbial communities on litter at early stages of breakdown are functionally distinct from those on nearby surface soils, but do not contribute to variation in early breakdown rates. Additionally, microbial communities in surface soils and litter are a mix of aerobic and anaerobic microbes, and seasonal variation in sulfate cycling likely plays an important role in the establishment of microbial communities that contribute to the breakdown of litter. Seasonal shifts play a major role in coastal wetland carbon dynamics, and we show that seasonal shifts may also change microbial activity and communities as marine stressors and subsidies are altered by incoming freshwater (Malone et al., 2014;Wilson et al., 2018). It is possible that the changes we see are more related to changes related to litter breakdown rates, however, the strong relationship between microbial activity and salinity suggests that changes in seasonal chemistry are more important. We specifically found that during higher freshwater inputs in January, salinity was associated with higher uptake of sulfates. In April when marine inputs increased, salinity was instead associated with higher dissimilatory use of sulfates, but only at intermediate salinities. This indicates a strong sulfate limitation during periods with higher freshwater, when sulfate acquisition is the limiting factor for sulfate reduction. That pattern shifts to high dissimilatory sulfate reduction when sulfates become available as freshwater inputs decrease. An in-depth understanding of microbial community composition and functional changes will allow us to better predict how increasing sea-level rise will change the processing of carbon within coastal wetlands, as both substrate and environmental chemistry are changing (Kirwan et al., 2013;Smoak et al., 2013). These data provide a starting point for further studies on how sea-level rise and resultant changes to macrophyte community structure will change the deposition and breakdown of organic matter across wetlands. We show that early breakdown of litter is driven by species-specific differences, while microbial community activity is limited seasonally by sulfate availability. Our study captures broad spatial dynamics across the Everglades but is not designed to account for finer spatial heterogeneity that may play an important role, especially in areas with variation in tidal exchange of marine water. Our results highlight the need for further studies that are specifically focused on how coastal microbial communities and litter breakdown respond to changing hydrologic regimes influenced by both marine water from sea-level rise and freshwater from the restoration of coastal wetlands. Additionally, we highlight the importance of further study on how microbial activity develops at later stages of litter breakdown, and different metrics of litter quality that relate to the differences in breakdown between litter species.