Hydrology masks warming effects on microbial communities in salt marsh soils

Abstract

Soil microbial communities play a pivotal role in salt marsh ecosystem functioning, driving processes such as organic matter decomposition and greenhouse gas cycling. Despite their importance, it remains unclear how climate warming will affect the diversity and activity of salt marsh soil microbial communities, limiting our ability to predict the fate of the vast stores of soil organic carbon in these so-called blue carbon ecosystems. Here, we leveraged the Marsh Ecosystem Response to Increased Temperature (MERIT) experiment to investigate the effects of sustained warming on the structure and function of the putatively active microbial community, as assessed by rRNA transcripts, alongside measurements of exo-enzymatic activities involved in carbon and nitrogen acquisition. Our results reveal that, after 5 years of experimental warming by +1.5°C and +3.0°C, the overall structure of the active microbial community remains remarkably stable, suggesting a high degree of resilience to elevated temperatures in this dynamic environment. However, warming selectively promoted drought-tolerant phyla, particularly Actinobacteriota and Firmicutes, which are known for their ability to degrade complex organic compounds and withstand desiccation. These findings suggest that while the active microbial community is broadly resistant to warming, subtle compositional shifts may enhance decomposition of recalcitrant soil carbon.

Long-term experimental warming in a salt marsh revealed that while the overall structure of the active soil microbial community remains stable, warming selectively enriches drought-tolerant, carbon-degrading phyla—suggesting increased microbial resilience and a potential shift toward enhanced decomposition of complex organic matter under climate change.

Introduction

Salt marshes—semiterrestrial wetlands that primarily occupy the intertidal zone at temperate latitudes—are important sinks for greenhouse gases. They are highly efficient at capturing and converting carbon dioxide to plant biomass and ultimately soil organic carbon (McLeod et al. 2011). Despite covering only a small fraction of the global ocean’s surface, salt marshes, along with other blue carbon ecosystems such as mangroves and seagrass beds, account for ∼50% of total organic carbon burial in marine soils and sediments (Duarte et al. 2013, Spivak et al. 2019). High rates of carbon accumulation in salt marsh ecosystems are the result of an imbalance in plant primary production and microbial decomposition of organic matter (Kirwan and Megonigal 2013, Temmink et al. 2022). Microbial decomposition is inhibited by the lack of more energy-efficient electron acceptors (i.e. oxygen), reducing overall soil carbon turnover (Almahasheer et al. 2017).

Salt marshes are characterized by hydrology-driven gradients (e.g. salinity, soil-redox), which are strongly reflected in gradients of biotic communities (e.g. plants, animals, and microbes) (Suchrow and Jensen 2010, Klink et al. 2016, Dinter et al. 2019). Both elevation and soil-depth gradients are characterized by strong differences in soil-redox potential (Rich et al. 2023, Tang et al. 2023, Mittmann-Goetsch et al. 2024). While soil redox does not always decrease with soil depth or decreasing surface elevation, the measurement of both parameters serves as an effective predictor for plant community composition (Davy et al. 2011). Many studies have shown that changes in abiotic conditions such as hydrology are reflected in changes in the microbial communities (Hernández et al. 2020, 2021, Unger et al. 2021, Kim et al. 2022). Microbial abundances and functioning (exo-enzymatic activities) were shown to have clear responses to hydrology (Mueller et al. 2020, Tang et al. 2021). Similarly, salinity can alter both organic matter decomposition rates (Luo et al. 2019) and microbial community composition (Neubauer et al. 2018, Dang et al. 2019). While hydrology-driven factors, such as salinity and soil redox, are strong predictors for biotic communities and their functioning, the combined influence of hydrology and warming has rarely been studied in situ.

Along with other global change factors, temperature rise is expected to alter carbon cycling (i.e. carbon sequestration) in salt marshes (Buschbaum et al. 2024). While salt marshes are highly effective carbon sinks, higher temperatures can accelerate soil respiration and potentially offset gains in carbon storage in these ecosystems (Bond-Lamberty and Thomson 2010). Yet, several studies highlight that salt marsh plants are resilient to moderate warming. This has been found for both aboveground (Noyce et al. 2019) and belowground plant net primary production (Charles and Dukes 2009, Smith et al. 2022). The response of organic matter decomposition to warming in salt marsh soils is still poorly understood; however, the substantial carbon reservoir stored in salt marsh soils presents a risk of positive feedback to global warming (Kirwan and Blum 2011). This is due to the lack of salt marsh warming experiments involving well-controlled and realistic warming treatments across whole-soil profiles (Rich et al. 2023).

First results from two existing field experiments employing whole-soil warming in coastal marshes reveal a potential for greater carbon loss (as respiration) with warming; this effect, however, is strongly mediated by hydrology and substrate composition (Smith et al. 2022, Tang et al. 2023). While in a microtidal, organic-rich salt marsh, belowground carbon accumulation was stimulated by moderate warming and decreased only with higher warming (Smith et al. 2022), organic matter decomposition rates were consistently higher with moderate and strong warming in a minerogenic, organic-poor salt marsh (Tang et al. 2023). This positive feedback can offset the carbon sequestration potential of salt marshes (Kirwan and Blum 2011). Hydrology and soil redox have strong leverage over warming effects. Anoxic conditions might “lock” carbon stocks away from warming effects, due to the accumulation of enzyme-inhibiting phenolic compounds (Freeman et al. 2001; but see Urbanová and Hájek 2021). Differences in site conditions across the elevational gradient of a salt marsh have been found to outweigh temperature effects (i.e. seasonal), to a large extent (Rinke et al. 2022, Tebbe et al. 2022). Studies from a peatland warming experiment corroborate these findings, showing a suppressed response of subsoil organic matter decomposition to warming compared to topsoil decomposition (Wilson et al. 2016, Hopple et al. 2020).

It has been posited that certain soil microbial groups are likely to thrive under warmed conditions. In particular, warming may enhance the ability of microbes to degrade more recalcitrant organic matter. Both Actinobacteriota and Firmicutes are well known for their ability to degrade complex organic compounds (Chen et al. 2016, Ni et al. 2024) and are expected to respond positively to increased temperatures given their resistance to desiccation (Oliverio et al. 2017, Wu et al. 2022). However, studies on the effects of warming on the soil microbial community structure in salt marshes are scarce, and even fewer examine the putatively active microbial pool. Limitations in commonly applied methods such as DNA-based amplicon sequencing may obscure our ability to detect changes. For example, extracellular DNA can persist in soils outside of living cells, masking shifts in community composition (Schnecker et al. 2024). Additionally, 16S rRNA gene sequencing of DNA includes dormant microbes, which are particularly abundant in salt marsh systems (Kearns et al. 2016). While approaches that separate extracellular DNA from intracellular DNA offer a promising way to better distinguish living microbial communities from relic DNA, these methods are technically challenging and require optimization tailored to specific soil types to yield reliable results (Alawi et al. 2014, Medina Caro et al. 2023). To overcome these challenges, total RNA analysis has emerged as a more precise approach for assessing changes in active microbial communities in salt marshes (Emery et al. 2019).

This study aims to assess the combined effects of climate warming and hydrology on microbial community structure and functioning in salt marsh soils. The study was conducted in a whole-ecosystem warming experiment, situated in a highly dynamic minerogenic salt marsh on the mainland coast of the German Wadden Sea, characterized by steep abiotic gradients in both salinity and soil redox (Rich et al. 2023). Both gradients are predominantly hydrology driven and vary along both elevation (marine-terrestrial) and soil-depth gradients. To our knowledge, this is the first study examining the effect of active belowground soil warming on the structure and function of salt marsh soil microbes in situ.

We hypothesized that (1) warming would increase microbial exo-enzymatic activity and (2) alter the community composition of putatively active microbes, characterized by higher relative abundances of microbes with the ability to degrade complex carbon compounds. We further hypothesized that (3) warming effects would be more pronounced in areas with less extreme hydrological conditions.

Material and methods

Site-description and experimental design

The MERIT (Marsh Ecosystem Response to Increased Temperature) experiment was established in 2018 in a Wadden Sea salt marsh on the Hamburger Hallig (54°36′06.7″N 8°48′57.4″E) in Germany, where tidal amplitude averages 3.4 m and accretion rates range from 3.0 to 12.9 mm yr⁻¹ (Nolte et al. 2013). MERIT is a whole-ecosystem warming experiment with passive aboveground and active belowground warming to a depth of 100 cm. Aboveground warming is applied by open-top chambers that trap heat and incoming radiation. Active belowground warming is accomplished by a combination of horizontal (GX-088L3106 GX, 9.8 Ω/m, 240 V) and vertical heating cables (GX 088L3100, 9.8 Ω/m, Danfoss, Denmark). MERIT consists of three 8-m² replicate plots in three elevational marsh zones located along the marine-terrestrial ecotone [pioneer zone, low marsh, high marsh] with three temperature treatments (ambient, +1.5°C, +3.0°C) leading to a total of 27 plots. The zones are characterized by differences in hydrology, soil redox conditions, and vegetation. The pioneer zone lies at the lowest elevation and is inundated almost daily during high tide. Here, soils remain waterlogged for extended periods, favoring strongly reduced conditions. Vegetation is sparse and dominated by Spartina anglica and Salicornia europaea with a low canopy height (∼24 cm) (Nolte et al. 2013, Rich et al. 2023). The low marsh is inundated less frequently, mainly during spring tides, resulting in a shorter hydroperiod and somewhat less reducing soil conditions compared to the pioneer zone. Vegetation is more diverse, forming a patchy community dominated by Puccinellia maritima, Atriplex portulacoides, and Limonium vulgare with a mean canopy height of ∼22 cm (Nolte et al. 2013, Rich et al. 2023). At the highest elevation, the high marsh is flooded only occasionally, typically during storm tides in autumn and winter. Soils here drain more rapidly and are less frequently reduced. Vegetation is dominated by tall Elymus athericus stands (mean canopy height ∼30 cm), forming dense monocultures that contrast with the more heterogeneous vegetation of the lower zones (Nolte et al. 2013, Rich et al. 2023). A detailed description of the experimental facility and its performance is given in Rich et al. (2023).

Soil sampling and processing

Soil sampling was conducted in September 2022. Sampling positions were chosen to have a similar distance to deep heating cables (d = 5 cm), sampling in close proximity without coring directly next to the heating cables, to avoid artificial effects of heating cables. Four replicate soil cores (100 cm deep) were taken from each of the N = 27 plots, using a 2.5-cm diameter gouge auger. The auger was rinsed with ethanol and DI water after each sampling to avoid microbial cross-contamination. Cores were cut into depth increments (0–5 cm, 5–10 cm, 20–30 cm, 40–50 cm, 80–100 cm) resulting in five samples per core. These samples were then stored at –20°C until further analysis. Subsamples for RNA analysis were taken during sampling, from one of the four replicate cores, using a sterilized spatula, carefully scraping material from the entire depth segment and storing it in a 5-ml Eppendorf tube. All tools used to cut and subsample increments were rinsed after every sample. RNA subsamples were immediately stored on dry ice in the field and during transport to the lab and were stored at –80°C until RNA extraction. The organic matter content was determined for every sample following standard loss on ignition protocols (Heiri et al. 2001).

Exo-enzymatic activity

Fluorometric exo-enzyme assays were conducted to measure activities of five hydrolytic enzymes involved in microbial carbon (ß-Glucosidase, GLU) and nitrogen acquisition (Chitinase, CHI; Leucine Aminopeptidase, LEU). Frozen field samples were homogenized, and a 2-g subsample was transferred into a 50-ml Falcon Tube. Soil slurries were prepared by adding 20-ml deionized water and mixing thoroughly. Exo-enzyme assays followed a standard protocol by Mueller et al. (2017). Assays were conducted in 96-well plates on a Multi-Detection Microplate Reader (Bio-tek Synergy ^TM HT, Winooski, USA). Substrate concentrations were at 1.6 mmol/l. Plates were incubated in the dark for 16–24 h at 20°C. Emission and excitation wavelengths were set to 460 and 365 nm, respectively. Due to the high carbonate buffer potential of the samples, assays were not buffered beforehand.

Soil RNA extraction and 16S rRNA amplicon sequencing

Total RNA was extracted from soil samples using the GeneMATRIX Environmental DNA & RNA Purification Kit (Roboklon, Germany) according to the manufacturer's protocol. Extracted RNA was purified using the Turbo DNA-free kit (Thermo Fisher, Germany). Reverse Transcriptase polymerase chain reaction (PCR) was performed to convert single-stranded RNA into complementary DNA (cDNA). In a PCR tube placed on ice, the following components were combined to a total volume of 13 µl: 10 µl of sterile distilled water, 1 µl of 10 mM dNTP mix (Invitrogen, USA), 1 µl of pd(N)₆ random hexamer primers (GE Healthcare, USA), and 1 µl of RNA sample. The mixture was heated at 65°C for 5 min in a PCR machine (Bio-Rad, USA) to denature secondary structures and then immediately cooled on ice. After a brief centrifugation, the following reagents were added: 1 µl of sterile distilled water, 1 µl of 0.1 M DTT, 1 µl of SuperScript III Reverse Transcriptase (Thermo Fisher, Germany), and 4 µl of 5x First-Strand Buffer. The reaction mixture was gently mixed by pipetting. To facilitate primer annealing, the reaction was incubated at 25°C for 5 min, followed by a 60-min incubation at 50°C for reverse transcription. The reaction was terminated by heating at 70°C for 15 min to inactivate the enzyme. cDNA concentrations were quantified using a Qubit 2.0 Flurometer (Invitrogen, USA) with the dsDNA HS and BR assay kits (Thermo Fisher, Germany). Amplicon libraries’ preparation was done by using in-house barcoded primer pairs targeting the V3–V4 hypervariable regions of the 16S rRNA (Uni515-F: GTGTGYCAGCMGCCGCGGTAA; Uni806-R: CCGGACTACNVGGGTWTCTAAT). PCR reactions (25 µl) consisted of 10x Pol Buffer C (Roboklon, Germany), 0.5 µM of each primer, 0.2-mM dNTP mix (Thermo Fisher, Germany), 2-mM MgCl₂, 1.25 U Optitaq Polymerase (Roboklon, Germany), PCR water, and 1 µl of template cDNA. PCR water and RNA extract were used as negative controls. The PCR program was set as follows: initial denaturation (5 min, 95°C), 32 cycles of denaturation (30 s, 95°C), annealing (30 s, 56°C), elongation (1 min, 72°C), and a final elongation (7 min, 72°C). Purification of the PCR product was done using the HighPrep™ PCR clean-up-reagents (Magbio Genomics inc, USA) according to the manufacturer's protocol. Concentrations of recovered PCR products were equilibrated and pooled together with positive and negative controls. The pool was sent for sequencing to Eurofins Genomics (Germany). The library preparation was done by PCR-free adapter ligation. Sequencing was done on the Illumina MiSeq platform using 2×300 bp paired-end mode using the V3 chemistry.

Processing of 16S rRNA amplicon data

Sequencing data were demultiplexed using cutadapt version 3.4 (Martin 2011) with the parameters -e 0.2 -q 15,15 -m 150 –discard-untrimmed. Amplicon sequence variants (ASVs, a proxy for phylogenetic species) were generated with trimmed reads and the DADA2 package version 1.20 (Callahan et al. 2016) in R version 4.1. For this, the pooled approach with the filtering parameters maxN = 10, truncQ = 20, rm.phix = TRUE, and minLen = 150 was used. Read pair merging was done using function “mergePairs” provided by the DADA2 package. Taxonomic assignment of the ASVs was done using DADA2 and the SILVA database version 138.1 (Quast et al. 2013). Further data processing was performed in R (version 4.4.2,R Core Team 2024) with the help of RStudio (version 2024.09.1+394, Posit team, 2024) and the packages ggplot2 (Wickham 2016), phyloseq (McMurdie and Holmes 2013), plyr (Wickham 2011), and vegan (Oksanen et al. 2025). ASVs that were assigned to chloroplasts or mitochondria and singletons were removed before rarefaction to 5000 reads per sample. To assess the metabolic potential of the assigned ASVs, we used the Functional Annotation of Prokaryotic Taxa (FAPROTAX) (Louca et al. 2016).

Quantification of 16S rRNA gene copy numbers

Quantitative polymerase chain reaction (qPCR) was employed to assess bacterial 16S rRNA gene abundance in soil samples collected in May 2023. Soil samples were collected to a depth of 30 cm using a gouge auger (d = 1.2 cm). The cores were sectioned into two depth increments (0–10 cm and 20–30 cm). DNA was extracted using the NucleoSpin® Soil Kit (NucleoSpin, Germany) according to the manufacturer’s protocol, with the exception that elution was carried out using autoclaved, ultrapure water. For the plasmid standard curve, Escherichia coli DH5α cells transformed with the 16S rRNA gene were cultured, and their plasmid DNA was extracted using the Presto™ Mini Plasmid Kit (Geneaid, Taiwan). DNA concentrations of both the soil samples and E. coli DH5α cells eluate were quantified at 280 nm using a NanoDrop 2000 spectrophotometer (Thermo Scientific, USA). The E. coli plasmid eluate was diluted to 30 ng/µl and used to create a five-step dilution series up to a dilution of 1:100 000, while the eluate from the soil samples was diluted once to 1:10. qPCR was carried out using the Eub341F and Eub534R primers (Muyzer et al. 1993) and SsoAdvanced™ Universal SYBR® Green Supermix on a Bio-Rad CFX96 system (Bio-Rad, USA). Cycling conditions included an initial denaturation at 98°C for 180 s, followed by 40 cycles of 95°C for 15 s, 60°C for 30 s, and 72°C for 15 s, followed by melt curve analysis. A melt curve analysis was performed from 60°C to 94°C with a ramp rate of 0.56°C per second. A standard curve was generated from the dilution series, relating measured Ct values to log₁₀-transformed copy numbers calculated from known plasmid (pDrive: 3851 bp) and vector length (194 bp) as well as measured DNA concentrations. Gene abundance was calculated as log₁₀ copies per gram organic matter.

Soil reduction, soil temperature, and elevation data

IRIS-sticks (Indicator of Reduction in Soils) were used to determine a soil reduction index (RI). The method followed Rabenhorst and Burch (2006) and was adapted by using flat PVC sticks, to simplify the scanning process (Mueller et al. 2020). Fe oxides were applied to white PVC-sticks (80 cm length) and deployed into the soils for 6 weeks. The capacity of solid-phase Fe (III) oxides to be reduced, by microbial Fe reducers, to soluble Fe (II) under anoxic conditions is used to assess soil reduction in soils. After 6 weeks of deployment, sticks were carefully taken out of the soil and rinsed with tap water. After drying, sticks were scanned using an overhead scanner (Viisan, China). Using a supervised classification, white areas on the sticks were determined (Mittmann-Goetsch et al. 2024). The RI was calculated as the share of white pixels from total pixels. Two consecutive campaigns, with duplicate 80 cm long IRIS sticks per plot, were conducted in 2023, spanning from 6 weeks prior to the soil sampling days in the previous year, until 6 weeks after. Depth increments were pooled together, to align with the soil layers from the microbial analysis (see 2.8 Statistical analysis). Per plot duplicate measurements of marsh surface elevation were conducted in 2024 using a DGPS device (ppm GmbH, Germany). Marsh surface elevations were calibrated to calculate elevation above mean high tide (MHT). Belowground temperatures were measured continuously in all plots using thermistors at 25 cm and 75 cm depth (Rich et al. 2023). Mean monthly temperatures were calculated per plot using temperature data ranging back one month before the soil-sampling date on the11th of September in 2022.

Statistical analyses

Rooting depth patterns at our sites guided the definition of two soil depth intervals. For microbial analyses, soil samples were pooled into a topsoil layer (0–30 cm) with strong live root influence and a subsoil layer (40–100 cm) with minimal live root influence. For redox characterization, reduction indices derived from IRIS sticks were pooled over the same topsoil layer (0–30 cm) but only down to 80 cm depth in the subsoil (40–80 cm), reflecting the installation depth of the probes. For all analyses, the enzymes Leucine-Aminopeptidase and Chitinase were combined into nitrogen-acquiring enzymes, to allow for a comparison with the carbon-acquiring enzyme ß-Glucosidase. We used nested mixed-effect ANOVA models with fixed effects (zone, layer, temperature) and random effects (core ID, Plot ID), to test for differences in microbial exo-enzymatic activities between the three elevational zones (pioneer zone, low marsh, high marsh, the soil-layers (topsoil 0–30 cm, subsoil 40–100 cm) and the temperature treatments (ambient, +1.5°C, +3.0°C). A similar model was used to test for differences in alpha diversity (Shannon index) of microbial communities without the random effect for core ID. The qPCR data were analyzed with a nested mixed-effect ANOVA model for differences in abundance of bacteria (EUB primer pairs) between zones and warming treatments. Prior to calculating mixed-effect ANOVA models, we visually inspected residuals for normal distribution and assessed the homogeneity of variances across groups. Given the fully balanced design of the study, potential moderate deviations from variance homogeneity were considered negligible for the validity of the ANOVA results (McGuinness 2002, Schielzeth et al. 2020). A nonmetric multidimensional scaling (NMDS) was performed based on Bray–Curtis distances from the microbial communities. Permutational multivariate analysis of variance (PERMANOVA) was performed, with Bray–Curtis distances to estimate the effect of experimental factors (zone, treatment, and layer) on the microbial communities assessed by rRNA transcripts. A correlation matrix between all metric environmental variables (soil elevation, soil reduction, soil temperature) and microbial response variables (exo-enzyme activities, alpha diversity) was calculated using Pearson correlations. Correlations were adjusted using the Holm correction. A distance-decay analysis was performed using the Bray–Curtis distances and Δ values of environmental variables (Δ soil elevation, Δ soil reduction, and Δ soil temperature). All statistical analyses were conducted with the software R Studio version 4.4.1 (R Core Team 2024). Data were prepared using R packages dplyr (Wickham et al. 2023), broom (Robinson et al. 2025), and lubridate (Spinu et al. 2024). All plots were created using R package ggplot2 (Wickham 2016) with extensions ggpubr (Kassambara 2020) and ggpmisc (Aphalo 2024). Nested mixed-effect ANOVA models were calculated using R package, lme4 package (Bates et al. 2015), and lmerTest (Kuznetsova et al. 2023). Tukey's pairwise comparisons were performed using the R package emmeans (Lenth et al. 2024). PERMANOVA was calculated using the R package vegan (Oksanen et al. 2025), and pairwise comparison was performed using pairwiseAdonis (Arbizu 2020).

Results

Exo-enzymatic activity

Exo-enzymatic activity of the carbon-acquiring enzyme ß-glucosidase differed significantly between elevational zones (F = 14.79, P < 0.001) and between soil layers (F = 90.75, P < 0.001, Fig. 1A). Both factors also showed a significant interaction effect (F = 4.16, P < 0.05). high marsh soils showed significantly higher carbon acquisition enzyme activity than samples from both pioneer zone and low marsh. Carbon acquisition enzyme activities were significantly higher in the topsoil layer compared to the subsoil layer (P < 0.001). While carbon acquisition enzyme activities did not significantly differ between the three warming treatments (ambient, +1.5°C, +3.0°C), there was a significant interaction effect between warming treatment and layer (F = 6.03, P < 0.01). Though not statistically significant, simple linear regression of carbon acquisition enzyme activities with mean plot temperature data (20 cm depth) showed a negative trend in activity with higher temperatures in the high marsh (Fig S1A). This trend was only apparent in the topsoil layer (0–30 cm soil depth) and showed an opposite trend in the subsoil layer (40–100 cm soil depth).

Figure 1.

Exo-enzymatic activity deviations (mean ± SE) in nmol * g OM^-1 * h^-1 from ambient baseline of (A) carbon acquiring enzyme β-glucosidase, (B) nitrogen-acquiring enzymes Leucine aminopeptidase and Chitinase across the elevational gradient (pioneer zone, low marsh, high marsh), between soil layers (topsoil 0–30 cm, subsoil 40–100 cm) and between warming treatments (+1.5°C, +3.0°C). Dashed line at 0 represents ambient baseline. Deviations show warming treatment effects relative to ambient control for each zone × layer combination.

The nested mixed-effects ANOVA revealed significant differences in combined nitrogen-acquiring enzymes Chitinase and Leucine-Aminopeptidase among elevational zones (F = 9.03, P < 0.01, Fig. 1B). Tukey HSD post-hoc showed that high marsh samples were significantly higher in nitrogen acquisition activity than pioneer zone samples. There were significantly lower nitrogen acquisition activities in subsoil samples (40–100-cm soil depth) than in topsoil samples (0–30-cm soil depth). Factors zone and layer had a significant interaction effect (F = 3.06, P < 0.05). There was no effect of warming treatment on the nitrogen acquisition enzyme activities. Across all zones and layers, the nitrogen acquisition enzymes did not relate to increased mean temperatures at the respective depths (Fig. S1B). However, there was a negative trend with higher temperatures in the activity of nitrogen-acquiring enzymes in high marsh soil across all layers, though this trend was not statistically significant (Fig. S1B).

Microbial diversity, community composition, and abundance

Alpha diversity (Shannon index of rRNA transcripts) differed significantly between soil layers, with higher diversity in the topsoil layer than in the subsoil layer (F = 26.56, P < 0.001). Zone had no significant effect on alpha diversity (F = 3.03, P = 0.07). Warming treatment showed no significant effects on the Shannon index for samples from top- and subsoil (Fig. 2A and Table 1). Though not significant, the alpha diversity within the high marsh showed a negative trend from ambient to +3.0°C samples (Fig. 2A). Similarly, there was a nonsignificant positive trend in the low marsh, where alpha diversity increased toward the +3.0°C treatment in the subsoil layer (Fig. 2A).

Figure 2.

(A) Alpha diversity, assessed using the Shannon diversity index, is compared across the elevational gradient (pioneer zone, low marsh, and high marsh), between the three warming treatments (ambient, +1.5°C, +3.0°C), and between soil layers (topsoil 0–30 cm, subsoil 40–100 cm) using boxplots. (B) Microbial community composition is visualized with NMDS plots based on Bray–Curtis distances between bacterial communities (16S rRNA sequenced as cDNA), illustrating patterns of beta diversity across the same environmental gradients. The subpanels show the results of one singular NMDS and share the same axis.

Table 1.

Summary table of statistical analyses reporting differences in microbial functioning variables (carbon and nitrogen acquiring enzymes) and microbial community variables from 16S rRNA sequencing (alpha- and beta-diversity) on soil samples taken in September 2022. Nested mixed-effect models (PERMANOVA for Bray–Curtis distances) were used to test for differences between factors: zones (pioneer zone, low marsh, and high marsh), layers (topsoil 0–30 cm, subsoil 40–100 cm), and warming treatments (ambient, +1.5°C, +3.0°C). For the abundance data (log₁₀ copies * g OM ⁻¹), a nested mixed-effect model was used to test for differences between factors zone and treatment on soil samples taken in May 2023. Shown are P-values, significant differences are highlighted in bold.

Factor/ Parameter	Carbon acquisition	Nitrogen acquisition	Alpha diversity (Shannon)	Beta diversity (Bray–Curtis distances)	Abundance (log10 copies)
Factorial approach
Zone	P < 0.001	P < 0.01	P = 0.072	P < 0.001	P < 0.01
Layer	P < 0.001	P < 0.001	P < 0.001	P < 0.001	/
Treatment	P = 0.686	P = 0.562	P = 0.759	P = 0.118	P = 0.459
Zone x treatment	P = 0.538	P = 0.774	P = 0.175	P = 0.151	P = 0.059
Zone x layer	P < 0.05	P = 0.094	P = 0.991	P < 0.001	/
Treatment x layer	P < 0.01	P = 0.764	P = 0.673	P = 0.636	/
Treatment x layer x zone	P = 0.154	P = 0.97	P = 0.395	P = 0.726	/

PERMANOVA showed that community composition, as assessed via rRNA transcripts, was significantly affected by both zone (F = 11.97, P < 0.001) and layer (F = 13.83, P < 0.001) (Table 1 and Fig. 2B). Post hoc pairwise comparisons revealed that community compositions differed between all elevational zones and the two depth layers. There is a gradual change in community composition from the pioneer zone to high marsh as well as from topsoil to subsoil (Fig. 2B). Both factors (zone, layer) also showed a significant interaction effect on microbial community composition (F = 4, P < 0.001). Warming treatment did not significantly alter the community composition (F = 1.25, P = 0.121).

The most abundant phyla, as assessed via rRNA transcripts, across all samples were Proteobacteria (27%), Chloroflexi (15%), and Firmicutes (11%), which represented more than half of the detected ASVs (53%). The next most abundant groups were Desulfobacterota (9%), Acidobacteriota (7%), Planctomycetota (7%), Actinobacteriota (5%), Myxococcota (3%), and Gemmatimonadota (2%). All other phyla accounted for less than 1% of all ASVs and were pooled together in the group “others” (13%). Analysis of taxonomic groups on phyla level revealed differences between zones and warming treatments (Fig. 3). While Firmicutes and Actinobacteriota increased from the pioneer zone, to the low marsh, and to high marsh, there was an opposite trend for Desulfobacterota and Chloroflexi which both decreased toward the higher-elevated zones. Differences between warming treatments were especially pronounced in high marsh. For Firmicutes, there was an increase in abundance between ambient and +3.0°C treatments of 9% in the high marsh zone. This effect was especially pronounced in the subsoil layer, where Firmicutes increased by 15%, while the topsoil layer showed an increase by 5%. The effect was reversed in low marsh where abundance decreased about 7% from +3.0°C to ambient (Fig. 3). In the pioneer zone, the changes in abundance of Firmicutes were comparably small (about 2%). Warming-induced changes in Actinobacteriota were found in the high marsh where the abundance from ambient to +3.0°C treatments increased about 4%. This increase in Actinobacteriota with warming was especially strong in the subsoil layer (8%) compared to the topsoil (1%).

Figure 3.

Stacked bar plots showing the relative abundance of the most abundant phyla, based on total transcribed RNA (16S rRNA sequenced as cDNA), in all samples across elevational zones (pioneer zone, low marsh, and high marsh) between different warming treatments (ambient, +1.5°C, +3.0°C), based on total transcribed RNA. Shown are the top nine most abundant phyla and all remaining phyla in category others.

FAPROTAX-based functional guild analysis revealed distinct differences among marsh zones and soil depths (Fig. 4A and B). In the topsoil (0–30 cm), warming was associated with increased relative abundances of taxa linked to aromatic compound degradation and hydrocarbon degradation, particularly in the high marsh plots. In contrast, subsoil communities (40–100 cm) showed stronger responses in fermentation-related functional potential, which increased under both warming levels across all zones (Fig. 4A). While functional group patterns in the pioneer zone and low marsh remained relatively stable, high marsh soils showed the strongest warming-related shifts, with enhanced functional potential for the breakdown of complex organic compounds. Deviation plots revealed differences in the most relevant functional guilds in temperature treatments (+1.5°C, +3.0°C) in %-change to the ambient controls (Fig. 4B). In the high marsh topsoil (0–30 cm) layer, hydrocarbon degradation showed a strong increase in the +3.0°C temperature treatments. In the subsoil (40–100 cm) of the high marsh both temperature treatments (+1.5°C, +3.0°C) led to increases in hydrocarbon degradation.

Figure 4.

Microbial degradation pathway abundances and warming responses across the elevational gradient. (A) Heatmap showing mean abundances of degradation pathways across warming treatments (ambient, +1.5°C, and +3.0°C) between soil layers (topsoil 0–30 cm, subsoil 40–100 cm) and across the elevational gradient (pioneer zone, low marsh, and high marsh). Functional predictions based on FAPROTAX database. (B) Percentage changes (mean ± SE) of key degradation pathways (hydrocarbon degradation, fermentation, aromatic compound degradation) relative to ambient conditions (dashed line at 0%) for +1.5°C and +3.0°C treatments. Only pathways with sufficient data coverage (>50% nonzero values) were included in the analysis. For more detailed information see Fig. S2.

qPCR analysis on samples from 2023 (topsoil 0–10 cm, 20–30 cm) were tested using nested mixed-effects ANOVA. The analysis revealed that microbial abundance (16S rRNA gene copy number) was significantly affected by zone (F = 3.61, P < 0.05); low marsh samples showed significantly higher microbial abundance than the pioneer zone (P < 0.05). Both temperature treatment (F = 0.79, P = 0.46) and the interaction of temperature treatment and zone (F = 0.33, P = 0.86) had no significant effect on the total microbial abundance (Fig. S3).

Linear regressions between dissimilarities (Bray–Curtis distances) and Δ elevation revealed positive significant effects for both topsoil and subsoil samples, where a higher difference in elevation resulted in higher dissimilarities (Fig. 5A and B, both P < 0.001). A similar trend was found for ΔRI in the topsoil (Fig. 5C, P < 0.001), but not for the subsoil samples (Fig. 5D). For Δ mean monthly temperature both topsoil and subsoil samples showed non-significant negative trends in dissimilarities. Higher differences in temperature between samples led to smaller dissimilarities (Fig. 5E and F, P = 0.061). This negative trend, however, might be largely driven by the strongly diverging community in one plot of the low marsh (Fig. 2B).

Figure 5.

Bray–Curtis distances between microbial communities based on total bacterial 16S rRNA (sequenced as cDNA) as a function of differences in elevation (A, B), in RI (C, D), and mean soil temperature (E, F). All differences are calculated for pairwise comparisons of the 27 plots of the MERIT experiment. Top panels (A, B, C) show samples from topsoil layer (0–30 cm) and bottom panels (D, E, F) show samples from subsoil layer (40–100 cm). R² and P-values are shown based on linear regression models. Regression lines are shown for significant linear regressions (P < 0.05).

Soil and environmental parameters

Across the elevational gradient, marsh surface elevation increased from the pioneer zone (−30.4 ± 3.6 cm) to the low marsh (−7.1 ± 2.4 cm) and was highest in the high marsh (26.2 ± 2.4 cm). Soil temperature showed little variation among zones, with values ranging from 18.9 ± 0.2°C in the low marsh to 19.5 ± 0.1°C in the pioneer zone, while the high marsh was slightly cooler at 19.4 ± 0.1°C. In contrast, the RI decreased with elevation, being highest in the pioneer zone (0.76 ± 0.07), intermediate in the low marsh (0.70 ± 0.06), and lowest in the high marsh(0.60 ± 0.07).

The RI showed a strong negative correlation with marsh surface elevation in topsoil samples (P < 0.05). The RI was lower in plots with lower marsh surface elevation. This relationship was not significant in the bottom soil samples (Fig. 6). Correlation analysis showed that the activity of the carbon and nitrogen-acquiring enzymes was positively correlated with soil elevation (P < 0.05). For carbon-acquiring enzymes, there was a negative correlation with soil reduction (P < 0.05). Mean monthly temperature did not correlate with any of the microbial parameters (Fig. 6). The nested mixed-effects ANOVA revealed no significant differences in RI between the temperature treatments (P > 0.05).

Figure 6.

Pearson correlation matrix of environmental parameters: RI, marsh surface elevation (m), soil temperature, and microbial variables [carbon acquisition, nitrogen acquisition, alpha diversity (Shannon index), abundance (log₁₀ copies * g OM⁻¹)]. Shown are two separately calculated matrixes for topsoil (0–30 cm) and subsoil (40–100 cm) samples. Significant correlations (P < 0.05) are highlighted and nonsignificant correlations (P > 0.05) are crossed out. Correlations were adjusted using the Holm correction.

Discussion

Warming effects on microbial exo-enzymatic activity are limited to the high marsh

Contrary to our first hypothesis, warming did not increase microbial exo-enzymatic activity in our study (Fig. 1A and B). In contrast, we see negative deviations in exo-enzymatic activities for both temperature treatments across the zones (Fig. 1A and B). Differences compared to the ambient plots were most pronounced in the high marsh for carbon-acquiring enzymes, while nitrogen-acquiring enzymes were most affected by the +1.5°C treatments in the pioneer zone (Fig. 1A and B). Interestingly, in the high marsh both carbon and nitrogen acquiring enzymes showed a negative trend in response to higher monthly temperature in the topsoil layer (Fig. S1A and C), while subsoil samples in the high marsh responded with higher activity of carbon-acquiring enzymes with warming (Fig. S1B). This finding contrasts with general catalytic assumptions and previous work in the MERIT experiment (Tang et al. 2023). In an early stage of the MERIT experiment (years 1 and 2 of experimental warming), Tang et al. (2023) showed that warming increases the initial organic matter decomposition rate across the three elevational zones. The results of this study (year 5 of experimental warming) can potentially be explained by (1) acclimation of microbial communities to warming and/or (2) substrate depletion over a longer time frame of warming. Both factors have been shown to cause a return to lower microbial activities after an initial warming-induced increase (Bradford et al. 2008, Romero-Olivares et al. 2017, Walker et al. 2018).

Acclimation of microbial communities after initial warming has been suggested as an important mechanism in terrestrial ecosystems (Luo et al. 2001, Crowther and Bradford 2013). Acclimation describes a widespread down-regulation of microbial respiration across species adapted to warmer temperatures in response to warming (Crowther and Bradford 2013). This mechanism would explain why there were no longer significant differences in microbial exo-enzymatic activity among temperature treatments in our study after 5 years of warming. We argue that substrate depletion also plays an important role in salt marshes in this study, as has been shown in other ecosystems (Walker et al. 2018). This is supported by the documented trend of decreasing C acquisition activity with higher temperatures (Fig. 1A and Fig. S1A). Additionally, organic matter content has continuously decreased in the high marsh from 2019 to 2022 (Fig. S4). While there are no significant differences in organic matter content between the warming treatments, it is likely that through higher temperatures and dryer soil conditions in the high marsh, resources are becoming limited (Schimel 2018, Metze et al. 2023). According to climate data, the sampling year 2022 has been among the driest and hottest in the past 20 years (Fig. S5), this can in fact cause drought conditions in less frequently flooded zones of salt marshes and accompanied changes in root distribution (Menzel et al. 2025).

Indication of drought stress within the putatively active microbial community structure of the high marsh

In this study, we present novel insights from 16S amplicon sequencing conducted on rRNA transcripts (sequenced as cDNA) from an in-situ salt marsh warming experiment. Previous findings from other studies showed that microbial community composition in salt marshes (based on 16S amplicon sequencing of DNA) changes little under experimental warming (Duchesneau et al. 2025). However, it remained unclear if changes in microbial communities might be obscured by extracellular DNA (see Schnecker et al. 2024), which can have comparably long retention times, if the DNA is protected (Bartholomäus et al. 2024). Contrary to our second hypothesis, we demonstrate that overall the putatively active microbial community composition changed little after 5 years of experimental warming. Nevertheless, there were some notable trends with regard to changes in relative abundances of indicator groups, as well as the functional potential of the communities according to FAPROTAX analysis (Fig. 3A and B and Fig. 4A and B).

Recent studies from the same experiment suggested that the high marsh zone is particularly vulnerable to drought stress (Ostertag et al. 2023, Menzel et al. 2025). Here, we observed higher abundances of Actinobacteriota and Firmicutes, phyla well-documented for their drought and heat tolerance due to their spore-forming capabilities (Oliverio et al. 2017, Wu et al. 2022). Actinobacteriota are well known for their ability to degrade complex, relatively recalcitrant soil organic matter compounds (Chen et al. 2016, Ni et al. 2024). Our FAPROTAX analysis reveals an increase in both aromatic carbon and hydrocarbon degradation functional potential in the warmed high marsh plots (Fig. 4A and B). This effect is especially pronounced in the topsoil layer, while the increasing temperatures in the subsoil lead to an increase in fermentation processes. It has previously been suggested that organic matter is especially vulnerable to warming effects in the high marsh (Tang et al. 2023). The enrichment of these dry-adapted groups in the high marsh suggests that drought conditions are already exerting selective pressure on the microbial community, or that such conditions are likely to occur. We see this further supported by the declining trend in alpha diversity with warming in the high marsh (Fig. 2A). Studies from upland ecosystems have shown sensitive negative responses of microbial alpha diversity to warming (Li et al. 2025). We argue that the declining alpha diversity is the result of a selection process towards more heat-tolerant taxa, a mechanism well known from upland ecosystems (Nottingham et al. 2022, Wu et al. 2022). It is also possible that the observed trend is due to a selection for microbial groups that can use carbon more efficiently when resources are limited. In August 2022—the month before sampling—there was an extreme drought, with the lowest rainfall and third highest temperature recorded since 2002 (Fig. S5). These dry conditions may have altered soil chemistry, making more electron acceptors available (Knorr and Blodau 2009). This could have supported microbes that break down more complex forms of carbon. Therefore, we cannot say for sure if the decline in microbial diversity was caused directly by higher temperatures or indirectly by changes in soil conditions due to drought.

Due to the lack of warming experiments in high-energy coastal systems like salt marshes (Rich et al. 2023), we so far had no evidence regarding the effects of warming on microbial diversity in these systems. To our knowledge, we report for the first time results demonstrating that high marsh zones respond to warming in a manner similar to upland ecosystems with respect to microbial diversity, whereas microbial diversity in low marsh and pioneer zones with more reducing conditions remains stable under warming (Fig 2A and B, and Fig 3A). Overall, the high marsh harbors conditions which could promote a positive feedback of soil organic matter decomposition (i.e. exo-enzymatic activities and hydrocarbon degradation) to global warming.

Hydrology (zonation and soil layer) as the main driver of salt marsh microbial functioning and community structure

In line with our third hypothesis, we report that, salt marsh zone and soil layer are main predictors for composition and functioning of microbial communities in Wadden Sea salt marshes (Figs. 1 and 2B). Given the strong relation between marsh surface elevation and RI (Fig. 6), we argue that hydrology is the main driver in determining composition and functioning of microbial communities. Both zonation (position along the marine-terrestrial ecotone) as well as marsh surface elevation of individual plots showed significant effects on exo-enzymatic activities, and on alpha and beta diversity (Fig. 5 and Table 1). Our findings are in line with other studies from wetland ecosystems, showing strong redox constraints on decomposition processes, with no indication of warming effects in permanently waterlogged (reducing) soil layers (Wilson et al. 2016, Hopple et al. 2020, Tang et al. 2023). Maxwell et al. (2024) showed that soil depth and elevation are the strongest predictors for SOC stocks in tidal marshes. Studies from the Wadden Sea area, where bacterial communities showed a clear zonation, further support our findings (Tebbe et al. 2022). This zonation is especially characterized by an increase in Firmicutes and Actinobacteriota from the pioneer zone to the high marsh, while Desulfobacterota and Planctomyceota decreased with increasing elevation (Fig. 3A).

We show that microbial activities were highest for both carbon and nitrogen acquisition in the high marsh (Fig. 1A and B). In addition, carbon enzyme activities were negatively correlated with soil reduction in the topsoil and nitrogen enzyme activities showed a negative trend with soil reduction in the topsoil (Fig. 6). In the least reduced zone (high marsh), organic matter breakdown is likely promoted. The high marsh zone is characterized by a low flooding frequency resulting in drier soils (Rich et al. 2023). We see that in areas with low reduction the activity of exo-enzymes was higher (Fig. 6). These findings are in line with the enzymatic latch hypothesis, stating that hydrolase enzymes are negatively affected by the accumulation of phenols in anoxic soils (Freeman et al. 2001). Our data suggest that the high marsh zone is distinct in its hydrology from lower elevated pioneer zone and low marsh and thus especially vulnerable to the influence of global warming.

Conclusion

In this study, we investigated how soil microbes are affected by both hydrology (i.e. zonation and soil layers) and climate (experimental warming). The study was conducted within the whole-ecosystem warming experiment MERIT located in a Wadden Sea salt marsh. Previously, litter decomposition was found to increase with warming in the early stages of the MERIT experiment (Tang et al. 2023). We find that 5 years after warming was initiated, hydrological conditions (i.e. zonation and soil layer) are the main drivers for both microbial functioning (i.e. carbon and nitrogen acquisition) and microbial community composition (i.e. alpha- and beta-diversity). Warming effects on the parameters investigated in this study are strictly limited to the higher-elevated, more terrestrial areas of the ecosystem. An additional reason for the unexpected finding that warming did not have significant effects on microbial community structure and functioning is because in our experiment we only employ passive aboveground warming, meaning plant-driven effects may be less prominent than in experiments with active aboveground warming. We show with our study that, in wetland ecosystems the effect of hydrology (here soil reduction and marsh surface elevation) is of overriding power, masking changes in microbial functioning and community structure through warming.

Author contributions

Julian Mittmann-Goetsch (Conceptualization, Data curation, Formal analysis, Investigation, Visualization, Writing—original draft), Peter Mueller (Conceptualization, Funding acquisition, Project administration, Supervision, Writing—review & editing), Kai Jensen (Conceptualization, Funding acquisition, Project administration, Supervision, Writing—review & editing), Susanne Liebner (Conceptualization, Writing—review & editing), Simon Thomsen (Methodology, Project administration, Writing—review & editing), Roy Rich (Data curation, Methodology, Project administration, Writing—review & editing), Alexander Bartholomäus (Formal analysis, Writing—review & editing), Johann Jaitner (Formal analysis, Investigation, Writing—review & editing), and Viktoria Unger (Conceptualization, Investigation, Project administration, Supervision, Writing—review & editing)

Conflict of interest

None declared.

Funding

Julian Mittmann-Goetsch was funded by Fischer Stiftung (Stifterverband für die Deutsche Wissenschaft) within the SEAL-C junior research group and the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) within the Research Training Group (RTG) 2530 Biota-mediated effects on carbon cycling in estuaries (grant no. 407270017). Peter Mueller was funded by DFG via the Emmy Noether Program (502681570).

Data availability

Data and code are available at https://github.com/JulianMiGoe/merit_microbes_repo. The raw sequencing data is publicly available on the European Nucleotide Archive (ENA) under project accession number PRJEB91652 (https://www.ebi.ac.uk/ena/browser/view/PRJEB91652).