DNA–RNA congruence and inter-domain co-occurrence networks reveal contrasting microbial association regimes in eastern versus western mangrove sediments of Shenzhen

Mangrove ecosystems are highly vulnerable to coastal environmental change and anthropogenic disturbance. The mangrove forests distributed along the eastern and western coasts of Shenzhen experience contrasting environmental regimes. The eastern coast is influenced by oligotrophic waters of the northern South China Sea and is dominated by native mangrove stands, whereas the western coast, located within the Pearl River Estuary, is subject to elevated nutrient inputs, intensive urbanization and extensive mangrove restoration. In this study, sediment prokaryotic and eukaryotic communities were characterized through parallel 16S and 18S rRNA gene sequencing at both DNA and RNA levels. Significant differences in community composition were detected between eastern and western sites in both microbial domains. Eastern sediments were enriched in sulfate-reducing bacteria and ammonia-oxidizing archaea and were dominated by pennate diatoms, whereas western sediments contained higher proportions of copiotrophic bacterial taxa and were dominated by centric diatoms. Inter-domain co-occurrence networks constructed at both DNA and RNA levels exhibited substantially greater complexity in eastern mangroves than in western mangroves. Hub subnetworks further revealed a pronounced difference in DNA–RNA congruence, with 307 taxa shared between DNA and RNA levels in the east compared with only 12 taxa in the west, indicating a strong decoupling between DNA-associated and RNA-associated communities under western conditions. Consistent with these network patterns, Mantel tests showed that eastern microbial communities were only weakly associated with abiotic variables, whereas western communities responded more strongly to environmental gradients, suggesting that external forcing plays a greater role in structuring western assemblages. Collectively, these results indicate that the relatively stable eastern environment supports complex, highly congruent microbial networks, while the more dynamic western environment is characterized by sparse and decoupled networks. Environmental regimes therefore reorganize not only taxonomic composition but also the architecture and DNA–RNA congruence of microbial co-occurrence networks, highlighting the potential of these network-level attributes as complementary indicators for mangrove ecosystem assessment.

Mangrove forests are among the most productive and ecologically important coastal ecosystems in the tropics and subtropics, providing critical services including carbon sequestration, coastal protection, nutrient filtration, and biodiversity support (Alongi et al., 2005; Carugati et al., 2018). Beneath these functions, sediment microbial communities drive organic matter mineralization, sulfur cycling, nitrogen transformation, and methane metabolism, making them central to the biogeochemical cycling that sustains mangrove ecosystem health (Kristensen et al., 2008; Helfer and Hassenrück, 2021). Periodic tidal inundation generates dynamic fluctuations in salinity, redox potential, and nutrient availability, and microbial communities in these environments respond to such changes while sustaining essential ecosystem functions (De Santana et al., 2021). The adaptation and succession of mangroves are closely tied to the diversity and activity of sediment microbes, including bacteria, archaea, fungi, and protists, which maintain primary productivity, nutrient transformations, and vegetation conservation (Mai et al., 2021; Lai et al., 2022). Understanding the composition, activity, and interaction structure of these communities is therefore central to assessing the functional status of mangrove ecosystems (Allard et al., 2020; Zhuang et al., 2020).

Over the past decade, high-throughput sequencing has substantially advanced our knowledge of mangrove sediment microbial diversity and its environmental drivers (Zhang et al., 2019; Gómez-Acata et al., 2023; Jamon-Haon et al., 2025). These studies have revealed that mangrove sediment microbiomes are highly diverse and likely perform a wide range of ecological functions (Allard et al., 2020). However, most of this work has examined prokaryotic or eukaryotic communities separately, leaving their inter-domain relationships poorly understood (Birnbaum et al., 2025; Ng et al., 2025). Recent evidence from plant microbiome research demonstrates that inter-domain networks reveal hub species and ecological patterns invisible to single-domain analyses, and that host evolution or environmental disturbance can fundamentally reshape these inter-domain interaction architectures (Lee et al., 2022). In addition, metabarcoding using environmental DNA (eDNA) as a template, which captures the total community, is widely applied in field surveys to understand the diversity and function of sediment microbiomes (Ruppert et al., 2019; Pawlowski et al., 2022). However, the DNA-associated communities retrieved from DNA do not distinguish the RNA-associated fraction due to the presence of dormant and deceased organisms, especially in mangrove sediments (Von Hoyningen-Huene et al., 2022; Yu et al., 2025). There has been increasing recognition of the importance of environmental RNA (eRNA) in field applications, because RNA is predominantly generated by living cells and is therefore used as a proxy for the transcriptionally active fraction (Zaiko et al., 2018; Yates et al., 2021). Moreover, eRNA decays much faster than eDNA and the ubiquitous presence of RNases in the environment further contributes to its rapid degradation (Marshall et al., 2021; Giroux et al., 2022). Parallel DNA–RNA surveys have uncovered marked discrepancies between DNA-associated communities and RNA-associated, transcriptionally active proxies (Wu et al., 2017; Lu et al., 2020; Giroux et al., 2022; Lu et al., 2023). Therefore, eDNA profiling recovers both present and past organisms, whereas eRNA methods provide a transcriptionally active proxy that more closely reflects the contemporary community.

In recent years, co-occurrence network analysis has been increasingly applied to provide new insight into potential microbial interactions and their responses to environmental variations (Deng et al., 2012; Banerjee et al., 2016). Network topological features can identify hub taxa and their ecologically meaningful associations within the community (Berry and Widder, 2014). Inter-domain networks constructed at both DNA and RNA levels can further reveal how interaction structure differs between the total and active communities (Du et al., 2023). Despite growing recognition of the importance of sediment microbes, most previous studies have been largely descriptive, focusing primarily on community composition (Allard et al., 2020). Microbial composition and potential microbial interactions are still rarely incorporated into the assessment of mangrove ecosystem status. This is particularly relevant because mangrove restoration has expanded considerably across China and Asia, yet restoration outcomes are typically evaluated through vegetation metrics alone, while the microbial communities that underpin sediment biogeochemical functions receive limited attention (Lai et al., 2022). How microbial communities and their association structures differ between mangrove systems under contrasting coastal regimes, and whether these differences can inform ecosystem assessment, remains largely unexplored.

The mangrove forests distributed along the eastern and western coasts of Shenzhen represent an ideal natural laboratory for comparing sediment microbiomes under contrasting coastal regimes (He et al., 2025). Bisected by the Hong Kong Special Administrative Region, the eastern coast, influenced by oligotrophic waters from the northern South China Sea with stable high salinity and native-dominated mangrove communities; and the western coast, situated within the Pearl River Estuary, subject to high riverine nutrient loading, pronounced salinity fluctuations, rapid urbanization, and extensive mangrove restoration (Harrison et al., 2008; Yi et al., 2021; He et al., 2025). This environmental contrast offers an opportunity to investigate how different regimes shape the structure, activity, and interaction networks of sediment microbial communities.

In this study, we conducted a comprehensive survey across the Shenzhen coastline, simultaneously characterizing prokaryotic (16S rRNA gene) and eukaryotic (18S rRNA gene) communities at both DNA and RNA levels. Inter-domain co-occurrence networks were constructed to identify hub taxa and functional modules. Our aims are to (1) reveal differences in the DNA-associated and RNA-associated sediment microbiomes between eastern and western mangrove ecosystems, and (2) identify distinct microbial co-occurrence regimes between contrasting environmental settings through inter-domain co-occurrence networks. By comparing DNA-associated and RNA-associated communities and their co-occurrence structures, this study seeks to provide new insight into the microbial dimensions of mangrove ecosystem functioning and assessment.

Mangrove forest field survey and environmental data collection

A ground-based vegetation survey was conducted across 105 plots (each 10 m × 10 m) distributed along the eastern (n = 33) and western (n = 72) coasts of Shenzhen between June and August 2023. Within each plot, mangrove community composition and structural attributes were recorded, including species identity, stem density, and life-stage classification (adult, sapling, and seedling). Surface sediment samples (0–5 cm) were collected concurrently within the same plots. For DNA analysis, sediments were placed in sterile zip-lock bags; for RNA analysis, sediments were transferred into 50 mL Falcon tubes and immediately immersed in RNAlater (Ambion, Austin, TX, USA). All samples were kept in a foam-insulated box with dry ice during transport and stored at −80 °C until nucleic acid extraction. Additional surface sediments (0–5 cm) were collected in triplicate within each plot at low tide for physicochemical analyses. Total carbon (TC), total nitrogen (TN), and total sulfur (TS) contents were determined using an elemental analyzer (vario EL cube; Elementar Analysensysteme GmbH, Germany). Grain size distribution was measured with a laser particle size analyzer after dispersing the samples in deionized water. Sediment salinity was determined as electrical conductivity (EC) in deionized water using a mass ratio of 1:5 (sediment to water) (Bao, 2000; Chai et al., 2017). Total phosphorus (TP) was determined spectrophotometrically by the vanadium-molybdenum yellow method after digestion with concentrated sulfuric acid (Kitson and Mellon, 1944; Bao, 2000). Tidal parameters (tidal range, mean tide level, and frequency of high–low tidal cycles) and meteorological data (mean annual temperature, precipitation, and humidity) were obtained from the local marine and meteorological monitoring departments.

Nucleic acid extraction, amplicon amplification and sequencing

Total DNA and RNA were extracted from approximately 0.2 g of sediment using the DNeasy PowerSoil Pro Kit (Qiagen, Hilden, Germany) and the RNeasy PowerSoil Total RNA Kit (Qiagen, Hilden, Germany), respectively, following the manufacturer’s instructions. RNAlater was removed following the procedure described in Xu et al. (2013). Reverse transcription of RNA was performed immediately after extraction using the PrimeScript RT Reagent Kit with gDNA Eraser (Perfect Real Time) (Takara, Japan), according to the manufacturer’s protocol.The DNA and cDNA were used as templates for downstream PCR and qPCR amplifications.

Prokaryotic (bacteria and archaea) and eukaryotic amplicons were amplified using the primer sets from the Earth Microbiome Project (EMP) (Thompson et al., 2017). The bacterial and archaeal 16S rRNA gene fragments were amplified using barcoded primers 515F (5′-barcode+ GTGYCAGCMGCCGCGGTAA-3’) and 806R (5′-barcode+ GGACTACNVGGGTWTCTAAT-3’) (Caporaso et al., 2012; Apprill et al., 2015; Parada et al., 2016). Triplicate PCR reactions were performed in a 25 µL reaction volume containing 10 µL Premix Ex Taq (Takara, Japan), 0.5 µM of each primer, and 2 µL of template DNA. The PCR thermal cycling conditions consisted of an initial denaturation at 94 °C for 3 minutes, followed by 25 cycles of 94 °C for 30 seconds, 50 °C for 45 seconds, and 72 °C for 45 seconds, with a final extension at 72 °C for 10 minutes. Triplicate PCR products were examined by gel electrophoresis. The triplicate PCR products were pooled together and sequenced on the Illumina NovaSeq 6000 platform (Illumina, San Diego, CA, USA).

The eukaryotic 18S rRNA gene fragments were amplified using barcoded primers 1391F (5′-barcode+GTACACACCGCCCGTC-3’) and EukBr (5′-barcode+TGATCCTTCTGCAGGTTCACCTAC-3’) (Amaral-Zettler et al., 2009; Stoeck et al., 2010). Triplicate PCR reactions were performed using the same composition as for the 16S rRNA amplification. The thermal cycling conditions were modified to optimize yield: initial denaturation at 94 °C for 3 minutes, followed by 28 cycles of 94 °C for 30 seconds, 57 °C for 45 seconds, and 72 °C for 45 seconds, with a final extension at 72 °C for 10 minutes. Triplicate PCR products were examined by gel electrophoresis. The triplicate PCR products were pooled together and sequenced on the Illumina NovaSeq 6000 platform (Illumina, San Diego, CA, USA).

Bioinformatic analysis

To assess prokaryotic community structure, 16S rRNA gene amplicons were sequenced from 209 libraries generated from the same batch of sediment samples, comprising 104 DNA-level and 105 RNA-level libraries. For eukaryotic community structure, 18S rRNA gene amplicons were sequenced from 185 libraries generated from the same DNA and cDNA templates, comprising 96 DNA-level and 89 RNA-level libraries. DNA and RNA were extracted from the same sediment samples. A small number of libraries were unavailable because PCR amplification failed or because the sequencing data did not meet quality-control criteria, which accounts for the unequal numbers of DNA- and RNA-level libraries. Sample metadata, including collection sites and the DNA/RNA identity of each library, are provided in Supplementary Table S2. Amplicon sequencing data from both the 16S and 18S rRNA gene fragments were analyzed using the mothur software package for microbial ecology (Schloss et al., 2009). Quality control was performed by discarding low-quality reads (e.g., those with low average quality scores, incorrect lengths, ambiguous bases, homopolymers longer than 8 bp, or chimeric sequences). High-quality reads were clustered into operational taxonomic units (OTUs) at 97% similarity in mothur (Schloss et al., 2009), a resolution suitable for comparing community structure, DNA–RNA congruence, and co-occurrence networks. OTUs with fewer than 10 reads across all samples were excluded from downstream analysis. Representative sequences were randomly selected using the get.oturep command in mothur. Taxonomic classification of both prokaryotic and eukaryotic sequences was performed in mothur (v.1.48.0) using the classify.seqs command with the Wang method and an 80% bootstrap cutoff. Taxonomic classification of bacterial and archaeal sequences was performed against the SILVA SSU reference database (release 138; Quast et al., 2013), using the mothur-formatted files silva.nr_v138.align and silva.nr_v138.tax. Eukaryotic sequences were classified against the Protist Ribosomal Reference database (PR2 version 5.0.0; Guillou et al., 2013). Rarefaction curves and Good’s coverage were calculated in mothur (v.1.48.0). The sequencing dataset is deposited at the National Center for Biotechnology Information (NCBI) under BioProject PRJNA1506730. Sample metadata corresponding to the 209 prokaryotic and 185 eukaryotic libraries are provided in Supplementary Table S2.

Diversity indices, including observed richness, Shannon diversity, Pielou’s evenness, and Margalef richness, were calculated from rarefied OTU tables in R (version 4.4.2) using the vegan package (version 2.7-2; Oksanen et al., 2026) and visualized with ggplot2 (Wickham, 2016). Relationships between environmental variables and community composition were first examined with Mantel tests. Collinearity among the 12 environmental variables was then assessed with pairwise Pearson correlations and variance inflation factors (VIF) across the 105 vegetation plots. All VIF values were below 10. Annual climate and tidal variables (temperature, humidity, precipitation, HLCount, Tide_Diff and Tide_level) showed little or no within-coast variation and were therefore excluded from within-coast constrained models. Distance-based redundancy analysis was performed separately for each coast × template combination using capscale in vegan, with Bray–Curtis dissimilarities of the taxon-composition tables (prokaryotic phyla and Proteobacteria classes; eukaryotic higher taxa) constrained by the plot-scale sediment variables total sulfur (TS), total carbon (TC), salinity, total phosphorus (TP), total nitrogen (TN) and clay content. TN was omitted from the eastern models because it had insufficient within-coast variation. Model significance was tested by permutation (999 permutations). Unique contributions of individual variables were assessed with marginal tests (anova and adonis2, by = “margin”).

Co-occurrence network analysis

Co-occurrence networks were constructed using 184 pairs of prokaryotic 16S and eukaryotic 18S rRNA gene sequencing data obtained from the same batch of DNA and cDNA (RNA template) samples. The dataset was categorized into four groups according to geographic location (Eastern/Western of Shenzhen coast) and source (DNA/RNA): East-DNA (n = 32), East-RNA (n = 30), West-DNA (n = 63), and West-RNA (n = 59). Prior selection was performed to ensure the representativeness of OTUs; only those present in ≥15 prokaryotic samples or ≥10 eukaryotic samples were retained for subsequent network robustness analysis. Pairwise Spearman correlations between OTUs were calculated using the “picante” package in R (version 1.8.2; Kembel et al., 2010). Only robust (r > 0.8 or r < -0.8) and statistically significant (p < 0.01) correlations were retained for network construction (Hu et al., 2017). Spearman p-values were obtained from “picante::cor.table” and were not Bonferroni- or FDR-adjusted. Edge retention followed the dual criterion of |r| > 0.8 and raw p < 0.01. At the sample sizes used here (n ≥ 30), |r| > 0.8 already corresponds to very small raw p-values, and a Benjamini–Hochberg correction would not remove additional edges that meet the |r| threshold. Network attributes, including degree, modularity, betweenness centrality, and clustering coefficient, were analyzed using the “igraph” package in R (version 2.2.1; Csardi and Nepusz, 2006). Random networks were by generating 1000 Erdös-Rényi networks with an equal number of nodes and edges as the corresponding real-world networks (Xue et al., 2018; Erdős and Rényi, 1960).

Topological features of the network were further analyzed to characterize the roles of nodes within the network. Within-module connectivity (Zi), reflecting how well a node is connected to other nodes within its module, and among-module connectivity (Pi), reflecting how well a node is connected to nodes in other modules, were calculated (Deng et al., 2012). Nodes were classified into four categories based on their roles: (i) Peripheral nodes (Zi ≤ 2.5, Pi ≤ 0.62), which have few connections mainly within their modules; (ii) Connectors (Zi ≤ 2.5, Pi > 0.62), which link to nodes in other modules; (iii) Module hubs (Zi > 2.5, Pi ≤ 0.62), which are highly connected to nodes within their modules; and (iv) Network hubs (Zi > 2.5, Pi > 0.62), which serve as both module hubs and connectors (Deng et al., 2012). To evaluate the repeatability of hub identification, 10% of samples were removed without replacement and the four networks were rebuilt with the same Spearman cutoff (|r| > 0.8, p < 0.01). The leave-out procedure was repeated 10 times for East-DNA (n = 32) and East-RNA (n = 30), and 20 times for West-DNA (n = 63) and West-RNA (n = 59). After each reconstruction, within-module connectivity (Zi) and among-module connectivity (Pi) were recalculated, and hub taxa were identified with the original thresholds (Zi > 2.5 or Pi > 0.62). Recovery was calculated as the proportion of hub taxa identified in the full network that were again classified as hubs in the leave-out network. Because closely related OTUs may occupy the same topological position across reconstructions, recovery was scored at the OTU, genus and family levels.

Environmental characteristic of the eastern and western mangrove ecosystem

Field surveys encompassed 105 sample plots distributed across the eastern (n = 33) and western (n = 72) coasts, spanning 28 mangrove and associated plant species (Figure 1). In the total mangrove survey dataset, true mangroves dominated the vegetation (90.3% of all stems), with semi-mangroves (8.7%) and associate species (1.0%) as minor components. However, community composition and structure diverged markedly between coasts. The eastern community contained 26.2% semi-mangrove stems, with true mangroves accounting for 73.8% of stems dominated by Kandelia obovata (38.4%) and Aegiceras corniculatum (20.2%). Native stems comprised 99.5% of the community. Structural attributes included a stem density of 70.1 ± 89.3 stems per plot, an adult tree proportion of 70.3%, saplings at 27.3% and a seedling proportion of 2.4%. The western community was overwhelmingly dominated by true mangroves (98.8% of stems), with semi-mangroves reduced to 0.1% (three individuals). While K. obovata (39.0%) and A. corniculatum (25.7%) remained co-dominant, introduced stems comprised 13.5%, predominantly Sonneratia apetala (13.5%). Western stands exhibited higher stem density (90.9 ± 60.5 stems per plot), with saplings at 33.8% and seedlings at 5.1%.

Quantitative assessment confirmed significant environmental heterogeneity between coastal sectors. PERMANOVA indicated overall differentiation (p < 0.001), with Welch’s t-tests revealing significant distinctions in 10 of 12 abiotic parameters (p < 0.05) (Supplementary Table S1). Western sites recorded higher mean annual temperature, greater tidal range, and higher mean tide level, whereas the eastern coast experienced higher annual precipitation and increased frequency of high-low tidal cycles. Sediment analyses showed western sites had significantly elevated total sulfur, total carbon, total phosphorus, and total nitrogen contents, along with higher clay content (≤4 μm fraction). Intertidal sediment salinity showed no significant difference between regions.

Prokaryotic and eukaryotic community composition and environmental drivers in the eastern and western mangrove

Sequencing of the 16S rRNA gene produced 6,125,777 high-quality sequences that clustered into 32,632 operational taxonomic units (OTUs) at 97% similarity across 209 samples (104 DNA-level and 105 RNA-level). Good’s coverage averaged 0.92, and rarefaction curves reached a plateau, confirming sufficient sequencing depth (Supplementary Figure S1). Overall, the communities were dominated by Proteobacteria (62.2% of sequences), comprising Gammaproteobacteria (28.4%), Deltaproteobacteria (26.2%), and Alphaproteobacteria (7.5%). Other prominent phyla included Bacteroidetes (11.0%), Chloroflexi (6.5%), Epsilonbacteraeota (2.5%), Planctomycetes (2.5%), Verrucomicrobia (2.5%), Acidobacteria (2.1%), and Actinobacteria (1.6%), with Archaea accounting for 2.3% of sequences (mainly Thaumarchaeota and Euryarchaeota) (Figures 2A, B).

Community structure showed strong spatial partitioning between the eastern and western coasts. NMDS ordination revealed clear regional clustering (ANOSIM, r = 0.24, p = 0.001) (Figure 2C). SIMPER analysis quantified an overall Bray-Curtis dissimilarity of 83.3% between eastern and western communities, with nearly all of this dissimilarity (98.9%) attributable to OTUs enriched in eastern sites. All of the top 20 individual OTUs with the highest average contribution were strongly East-enriched (fold-change 1.2–53.7×), including multiple sulfate-reducing Deltaproteobacteria (e.g., Desulfatiglans and the Sva0081 sediment group) and ammonia-oxidizing Thaumarchaeota (Candidatus Nitrosopumilus in Nitrosopumilales). At the phylum level, Proteobacteria contributed the largest share of the observed divergence (16.86%), followed by Bacteroidetes (2.66%); at the class level, Gammaproteobacteria (6.73%), Deltaproteobacteria (6.36%), and Alphaproteobacteria (3.56%) were the primary drivers, almost entirely through East-enriched OTUs.

In addition, diversity metrics differed markedly between the DNA-associated and RNA-associated communities. Across species richness, Shannon diversity, Pielou’s evenness, and Margalef richness, RNA-level values were significantly higher than DNA-level values in both regions (p < 0.05) (Figure 2D). A significant but relatively moderate dissimilarity was also observed between DNA- and RNA-based assemblages across the full dataset (ANOSIM, R = 0.196, p = 0.001). Taxonomic composition further highlighted these DNA–RNA discrepancies. In both eastern and western sediments, the RNA-associated community was strongly enriched in Proteobacteria, which accounted for 64.63% and 70.59% of the DNA–RNA dissimilarity contribution, respectively. This enrichment was mainly driven by Gammaproteobacteria (Oceanospirillales and Betaproteobacteriales) and Deltaproteobacteria (Desulfobacterales). Conversely, the DNA-associated community showed consistent enrichment of Bacteroidetes, Chloroflexi (mainly Anaerolineales), and Verrucomicrobia.

Sequencing of the 18S rRNA gene produced 5,123,675 high-quality sequences that clustered into 21,241 OTUs at 97% similarity across 185 samples (96 DNA-level and 89 RNA-level). Good’s coverage averaged 0.94, and rarefaction curves reached a plateau, confirming sufficient sequencing depth (Supplementary Figure S1). The communities were dominated by Gyrista (49.5% of sequences), primarily Bacillariophyceae and Mediophyceae. Other abundant groups included Ciliophora (13.6%), Euglenozoa (10.4%), Fungi (6.4%), Opisthokonta (3.8%), Chlorophyta (3.2%), Cercozoa (2.0%), Foraminifera (1.8%), Alveolata (1.7%), and Rotosphaerida (1.3%) (Figures 3A, B).

Community structure likewise exhibited clear spatial partitioning between coasts. NMDS ordination showed distinct regional clustering (ANOSIM, r = 0.27, p = 0.001) (Figure 3C). SIMPER analysis revealed that the overall dissimilarity was driven almost entirely by OTUs enriched in eastern sediments. All of the top 20 individual OTUs with the highest average contribution were strongly East-enriched (fold-change ranging from 2.2- to >666,000-fold), including multiple chlorophytes (e.g., Cladophora sp.) and rhodophytes (e.g., Caloglossa leprieurii). At the phylum level, Gyrista contributed the largest share (8.77%), followed by Fungi (4.46%), Ciliophora (4.03%), Chlorophyta (2.55%), Opisthokonta (1.94%), Cercozoa (1.59%), Euglenozoa (1.29%), Foraminifera (0.78%), and Rhodophyta (0.71%). At the class level, Bacillariophyceae, Opisthokonta, Ascomycota, and Gyrista were the main contributors, all through East-enriched OTUs.

Diversity metrics again differed consistently between the DNA-associated and RNA-associated communities, with RNA-level values significantly higher than DNA-level values across all indices (p < 0.05) in both regions (Figure 3D). A significant but moderate dissimilarity was also observed between DNA- and RNA-based eukaryotic assemblages across the full dataset (ANOSIM, R = 0.196, p = 0.001). Taxonomic shifts between DNA-associated and RNA-associated communities were region-specific. In the eastern sediments, RNA-associated fraction were enriched in chlorophytes and fungi, while DNA-associated fraction showed stronger representation of pennate diatoms (Naviculaceae) and foraminifera. In the western sediments, Gyrista (dominated by centric diatoms of the Mediophyceae and Thalassiosirales) was enriched in the DNA-associated fraction, whereas Euglenozoa (primarily Kinetoplastea) dominated the RNA-associated fraction.

Mantel tests showed an apparent asymmetry in environment–community correlations (Figure 4; Supplementary Figure S2). Significant associations were detected mainly in western assemblages for both prokaryotes and eukaryotes at DNA and RNA levels (r = 0.08–0.35, p < 0.05). Eastern prokaryotic communities showed no significant single-variable Mantel correlations. Eastern eukaryotic communities were associated only with tidal variables (HLCount, Tide_Diff and Tide_level) and only at the RNA level. Western prokaryotic communities correlated with eight variables at the DNA level (TS, TC, TP, TN, temperature, humidity, precipitation and HLCount; r = 0.17–0.29, p < 0.01) and seven variables at the RNA level (TC, TP, TN, temperature, humidity, precipitation and HLCount; r = 0.15–0.35, p < 0.05). Western eukaryotic communities correlated with six variables at the DNA level (temperature, humidity, precipitation, HLCount, Tide_Diff and Tide_level; r = 0.15–0.31, p < 0.01) and three variables at the RNA level (salinity, humidity and HLCount; r = 0.08–0.20, p < 0.05).

Several of these Mantel correlates (climate and tide) do not vary appreciably among plots within a coast. After removing these regional variables and retaining plot-scale sediment predictors (TS, TC, salinity, TP, TN and clay; TN omitted in the east), db-RDA models were significant in all eight coast × domain × template combinations (permutation p = 0.001–0.038), but explained only a modest fraction of compositional variation (adjusted R² = 0.09–0.20; Supplementary Table S4). Marginal tests identified a small number of unique predictors: TC in the western prokaryotic DNA model (p = 0.016) and clay in the western prokaryotic RNA model (p = 0.022); TS and TP in the eastern eukaryotic DNA model (p = 0.022 and 0.036) and TP in the eastern eukaryotic RNA model (p = 0.008); and clay in the western eukaryotic DNA model (p = 0.002). Thus both coasts show detectable sediment–community structure at the multivariate level, whereas the stronger western signal in the Mantel analysis is contributed in part by region-scale climate and tidal variables that distinguish the two coasts rather than by independent plot-scale gradients.

Topological features of the inter-domain co-occurrence network between the eastern and western mangrove

To examine potential co-occurrence patterns and interaction structures among prokaryotic and eukaryotic taxa, inter-domain networks were constructed at both DNA and RNA levels after rigorous pre-selection of metacommunities (OTUs retained only if present in ≥ 15 prokaryotic or 10 eukaryotic samples) (Figure 5). Robustness was confirmed by comparison with Erdős–Rényi random networks (Supplementary Figure S3); real networks exhibited distinct small-world properties, with significantly higher modularity, clustering coefficients, and small-world coefficients than their random counterparts (Table 1).

Marked spatial heterogeneity in network complexity was observed between the eastern and western coasts. Eastern networks (both DNA- and RNA-based) were substantially larger and more interconnected, containing 6,117–8,101 nodes and 20,015–60,438 edges, compared with only 1,424–2,707 nodes and 2,195–10,557 edges in western networks. In parallel, clear differences emerged between DNA- and RNA-based networks across both regions. DNA-based networks were consistently denser and more highly connected (10,557–60,438 edges; clustering coefficient 0.45–0.56) than RNA-based networks (2,195–20,015 edges; clustering coefficient 0.31–0.39). RNA networks, however, showed higher modularity (0.74–0.89 versus 0.83–0.89 for DNA networks) and slightly longer average path lengths (8.66–8.95 versus 8.51–8.91). The eastern DNA network displayed the highest mean degree (~14.9), while RNA-based networks remained sparser overall.

Prokaryotes dominated all networks (77–87% of nodes), with eukaryotes accounting for the remaining 13–23%. Core prokaryotic phyla were largely conserved across datasets: Deltaproteobacteria (11.17–23.34%), Gammaproteobacteria (13.90–16.95%), and Bacteroidetes (9.16–11.84%) were most abundant, followed by Alphaproteobacteria (6.34–10.93%) and Chloroflexi (4.51–7.53%). Eukaryotic phyla, particularly Gyrista (3.96–7.09%) and Ciliophora (2.95–4.49%), were enriched in RNA-based networks, with Gyrista reaching 7.09% in the western RNA network. Spatial divergence in taxonomic composition and DNA–RNA partitioning was also evident between coasts. In the east, Deltaproteobacteria dominated both DNA (20.31%) and RNA (23.34%) networks. In the west, Deltaproteobacteria remained dominant in DNA networks (21.65%) but decreased markedly in RNA networks (11.17%), where Gammaproteobacteria (16.95%) and Bacteroidetes (11.84%) increased; western RNA networks also contained a higher proportion of eukaryotic nodes (22.75%) than eastern RNA networks (14.83%).

Hub taxa and their associated hub subnetworks across DNA and RNA levels

Hub taxa in each network were identified by calculating within-module connectivity (Zi) and among-module connectivity (Pi) and visualized in ZP-plots, where module hubs and connectors were defined as hub taxa (Supplementary Figure S4). These hub taxa showed distinct compositional patterns across the four networks. Eastern networks contained substantially more hub taxa (East-DNA: 239, including 219 module hubs and 20 connectors; East-RNA: 316, with 158 of each type) than western networks (West-DNA: 59, with 56 module hubs and 3 connectors; West-RNA: 27, with 21 module hubs and 6 connectors). To test whether the same taxa would still be identified as hubs if the sample combination changed, each network was reconstructed after removal of 10% of samples (Supplementary Table S3). In the DNA networks, a median of 60% (East-DNA) and 59% (West-DNA) of the original hub OTUs were recovered as hubs. Recovery was lower in the RNA networks (24% in East-RNA and 46% in West-RNA), consistent with the greater sensitivity of the RNA-associated community to which libraries are included. When hub identity was allowed to be the same genus or the same family rather than the same OTU, median recovery rose to 0.57–0.77 at genus level and 0.62–0.77 at family level in all four groups. Thus the precise OTU that occupies a hub position can change among sample draws, particularly at the RNA level, but the position itself continues to be filled by members of the same genera and families. Hub taxa in this study are therefore treated as repeatable topological roles at the genus/family level, not as a fixed list of OTU barcodes. Prokaryotes dominated all groups, but their proportions varied: bacteria accounted for 87.4% (209 taxa) in East-DNA and 83.2% (263 taxa) in East-RNA, increasing to 91.5% (54 taxa) in West-DNA and dropping to 63.0% (17 taxa) in West-RNA. Eukaryotes were more prominent in RNA-based networks (13.6% in East-RNA and 37.0% in West-RNA) than in DNA-based networks (8.4% in East-DNA and 8.5% in West-DNA), while archaea were minor (3.2–4.2%) and restricted to eastern networks. Deltaproteobacteria led in East-DNA (47 taxa), East-RNA (77 taxa), and West-DNA (20 taxa) but were minimal in West-RNA (2 taxa); Gammaproteobacteria and Bacteroidetes gained prominence in West-RNA. Among eukaryotes, Gyrista and Ciliophora were the most frequent, with fungi notable only in East-RNA (5 taxa). No hub taxa were shared across all four networks, indicating strong regional and molecular differentiation. Eastern networks shared 15 common hub taxa between DNA and RNA levels (dominated by Deltaproteobacteria, including sulfate reducers such as Bac000003 and Bac000010 in Desulfobacterales), while western networks shared only one (Euk003012, a Colpodea ciliate).

Hub subnetworks were extracted from each of the four co-occurrence networks by retaining hub taxa (module hubs and connectors) together with their directly associated nodes. Eastern subnetworks were substantially larger and more modular than their western counterparts (East-DNA: 1,492 nodes across 14 modules; East-RNA: 1,702 nodes across 32 modules versus West-DNA: 406 nodes across 10 modules; West-RNA: 202 nodes across 15 modules), reflecting greater community complexity in the eastern coastal environment. Overlap between DNA- and RNA-based subnetworks differed sharply by coast. Eastern subnetworks shared 307 common taxa between DNA and RNA levels (representing ~18–21% of nodes in each subnetwork), dominated by Deltaproteobacteria (85 taxa, primarily Desulfobacterales and Syntrophobacterales), Bacteroidetes (42 taxa, largely Bacteroidales), Gammaproteobacteria (38 taxa), and Gyrista (12 taxa, mainly Bacillariophyceae). In contrast, western subnetworks shared only 12 common taxa (~3–6% of nodes). Cross-coast overlap was modest: 66 taxa were shared between the two DNA subnetworks (dominated by Deltaproteobacteria at 18 taxa, Alphaproteobacteria at 9 taxa, and Gammaproteobacteria at 8 taxa) and 33 taxa between the two RNA subnetworks (led by Gammaproteobacteria at 7 taxa, Verrucomicrobia at 6 taxa, and Gyrista at 4 taxa). Their 12 common taxa (the intersection of DNA and RNA cross-coast overlaps) were primarily bacteria such as Bacteroidetes (3 taxa) and Deltaproteobacteria (2 taxa), with Thaumarchaeota (1 taxon), underscoring a cross-regional core of taxa potentially involved in decomposition and nutrient cycling.

Modules within these subnetworks revealed distinct taxonomic compositions and potential functional asscociations. In the eastern subnetworks, a recurring potential association among taxa putatively involved in sulfate reduction–polymer degradation–benthic production was evident, integrating Deltaproteobacteria (primarily Desulfobacterales and Syntrophobacterales) with Bacteroidetes (Bacteroidales) and Gyrista (Bacillariophyceae), often together with Ciliophora. Additional eastern modules linked Deltaproteobacteria with Gammaproteobacteria (Betaproteobacteriales) and Gyrista, consistent with potential sulfur- and carbon- cycling associations. In the western subnetworks, modules were smaller and functionally distinct: West-DNA Module 1 grouped Gammaproteobacteria (Thiotrichales sulfur oxidizers) with Gyrista, a potential sulfur oxidation–primary production association, while West-RNA Module 3 was dominated by Gammaproteobacteria (Betaproteobacteriales) with Opisthokonta, consistent with a potential heterotrophic-degradation association under high organic loading. Across all subnetworks, intra-module connections were predominantly positive, indicating predominantly positive co-occurrence within modules, consistent with potential associations among taxa occupying shared niches and suggesting that module hubs and connectors participate in potential functional associations.

Distinct microbial communities between eastern and western mangrove sediments

Sediment microbiomes showed significant difference between the oceanographic eastern coast and the western Pearl River Estuary, which is a pattern consistent across both prokaryotes and eukaryotes. Regional clustering was significant for both domains (ANOSIM, prokaryotes r = 0.24, eukaryotes r = 0.27, both p = 0.001), with SIMPER attributing 83.3% of prokaryotic dissimilarity to coast-level differences (Figures 2C, 3C). The east-west divergence was overwhelmingly driven by eastern-enriched taxa across both domains, indicating profoundly asymmetric community structure. This inter-domain concordance points to fundamental ecological divergence rather than taxon-specific stochastic variation, reflecting the cumulative imprint of contrasting oceanographic regimes, geochemical conditions, and human disturbance intensities on the sediment microbiome.

This asymmetry reflects contrasting oceanographic and coastal regimes. The western coastline lies within the Pearl River Estuary, where DIN concentrations reach ~100 μM with N:P ratios exceeding 100:1 during wet seasons (Harrison et al., 2008), and riverine discharge delivers substantial terrestrial organic matter to coastal sediments (Yu et al., 2021). In contrast, the eastern coast receives oligotrophic waters from the northern South China Sea with stable high salinity (~30–33 PSU) and low nutrient concentrations (Lu and Gan, 2023). Our geochemistry confirmed this gradient; western sites exhibited significantly elevated total sulfur, total carbon, total phosphorus, and total nitrogen relative to eastern sites (p < 0.01; Supplementary Table S1), consistent with pronounced terrestrial organic loading in nearby subtropical mangrove sediments (Zhao et al., 2023). Beyond geochemistry, the coasts diverged sharply in anthropogenic pressure and ecosystem integrity. High-intensity human activities dominate the western coastal zone, driven by urban expansion and land reclamation that increased artificial shorelines by 120% since 1986 (He et al., 2025), while the eastern sector experiences scattered, low-intensity disturbance (Yi et al., 2021). This pressure gradient is also mirrored in mangrove community composition. Western stands contained 13.5% introduced Sonneratia apetala from active restoration programs, whereas eastern communities comprised 99.5% native species and included centuries-old stands under ecological protection. S. apetala was introduced to China in 1985 as a fast-growing restoration species and has been planted in Shenzhen Bay since the early 1990s; in southern China it can expand into native forest and change stand structure (Ren et al., 2009; Li R. et al., 2017; Li Y. et al., 2017). Plantations of this species in Shenzhen Bay enrich sediment carbon, nitrogen and sulfur relative to adjacent mudflats (Li R. et al., 2017), and its introduction has been shown to shift sediment diazotrophic assemblages toward sulfur-oxidizing lineages at the expense of sulfate-reducing groups (Huang et al., 2022). These changes in organic-matter input and microbial assembly (Lai et al., 2022) therefore sit within the same western setting already marked by estuarine nutrient loading and modified shorelines. Our survey was conducted from June to August 2023, during the wet season in which the greater part of Shenzhen’s annual rainfall is delivered and Pearl River discharge, temperature and sediment wetting are at their annual highs. Under these conditions, freshwater and nutrient pulses into the estuary intensify, and comparable subtropical mangrove sediments in the Pearl River Estuary and neighboring coasts show seasonal reorganization of benthic microbial abundance, extracellular enzyme activity and carbon metabolic profiles (Luo and Gu, 2015; Yang et al., 2022; Wang et al., 2024). The east–west contrast reported here is therefore the pattern expressed in that wet-season window; whether the same community and network differences persist through the dry season remains to be established.

The taxa driving this east-west divergence were predominantly East-enriched. Among prokaryotes, sulfate-reducing Deltaproteobacteria (Desulfatiglans and the Sva0081 sediment group) and ammonia-oxidizing Thaumarchaeota (Candidatus Nitrosopumilus) ranked highest. Sulfate-reducing bacteria are among the most critical functional guilds in mangrove sediments, where the sulfur cycle accounts for approximately 50% of total metabolic activity (Helfer and Hassenrück, 2021). Globally, Desulfatiglans and the Sva0081 sediment group are recognized as core mangrove microbiota with versatile metabolic capabilities in carbon and sulfur transformations (Gómez-Acata et al., 2023; Jamon-Haon et al., 2025). Their strong enrichment in eastern sediments indicates a well-established sulfur cycling network tightly coupled with organic matter mineralization, an indicator of redox stratification in mature mangrove ecosystems. Concurrent enrichment of ammonia-oxidizing archaea further signals active nitrogen cycling capacity in the east, pointing to coupled sulfur and nitrogen biogeochemical processing in these stable, oligotrophic sediments (Li et al., 2011).

Eukaryotic communities showed parallel spatial partitioning. Gyrista, primarily pennate diatoms, contributed the largest share of eukaryotic dissimilarity, followed by Ciliophora, Chlorophyta, and Foraminifera, all East-enriched. Eastern sediments contained pennate diatoms (Bacillariophyceae, particularly Naviculaceae in DNA fractions) and chlorophytes (Cladophora spp. in RNA fractions). Raphid-pennate diatoms possess a raphe enabling vertical migration to the sediment surface during low tides to optimize light exposure, an adaptation favoring stable substrates with moderate nutrient levels (Zhu et al., 2017). In contrast, western DNA fractions were dominated by centric diatoms (Thalassiosirales), predominantly planktonic forms deposited through sedimentation of suspended particles under high-turbidity estuarine conditions. Western RNA fractions showed strong transcriptional activity of kinetoplastid flagellates, bacterivorous grazers that play a central role in the microbial loop. The high activity of these grazers in nutrient-loaded western sediments indicates intensified microbial loop dynamics fueled by elevated bacterial biomass and labile organic matter turnover. Foraminifera enrichment in eastern sediments is consistent with their established sensitivity to environmental conditions and preference for well-oxygenated, oligotrophic marine environments, where they serve as reliable bioindicators of sedimentary environmental quality (Gómez-León et al., 2018; Daché et al., 2024). Together, these eukaryotic patterns reinforce the prokaryotic evidence; the eastern coast selects for benthic autotrophs in stable substrates, while the western coast favors heterotrophic grazers and planktonic deposition in eutrophic, disturbed systems.

Simultaneous DNA and RNA profiling distinguishes an RNA-associated, transcriptionally active proxy from the DNA-associated community. DNA persists from living, dormant and deceased cells, whereas RNA degrades rapidly in sediments and is therefore treated as a transcriptionally active proxy rather than as a census of living cells. Across both coasts, RNA-level communities were consistently more diverse than DNA-level communities (p < 0.05), indicating that a broader range of taxa maintain transcriptional activity than the DNA pool alone would indicate. Bacteroidetes, Chloroflexi (Anaerolineales), and Verrucomicrobia were consistently enriched in DNA-associated fractions, representing a reservoir of less transcriptionally active or dormant taxa. DNA–RNA partitioning also differed markedly between coasts. In eastern sediments, the RNA-associated fraction was dominated by Desulfobacterales and Oceanospirillales, consistent with coupled sulfate reduction and heterotrophic activity. In western sediments, the RNA-associated fraction was enriched in kinetoplastids and other eukaryotic grazers, suggesting intensified microbial-loop interactions under high nutrient loading. These contrasting RNA-associated community profiles add a functional dimension to the spatial divergence observed at the DNA level. Paired DNA and RNA templates therefore separate a DNA-associated standing community from an RNA-associated, transcriptionally active proxy, and they show which taxa are transcriptionally represented at the time of sampling. This proxy has several limits. Copy number of the rRNA operon varies by more than an order of magnitude among bacterial genomes, so read abundance is not a direct measure of cell abundance (Větrovský and Baldrian, 2013). Ribosomal RNA content likewise does not scale uniformly with growth rate across taxa, and dormant cells may retain ribosomes, so an RNA-associated signal cannot be equated with a currently growing population (Blazewicz et al., 2013). Preservation, extraction and reverse transcription can further influence RNA recovery. These constraints preclude taxon-level claims of in situ growth. They do not remove the broader pattern recovered here: eastern and western sediments differ in both DNA-associated and RNA-associated composition, and they differ sharply in DNA–RNA congruence, a contrast that remains informative for comparing the two coastal regimes. Collectively, these results suggest that the east–west difference in sediment microbiomes is the combined outcome of contrasting water-mass properties, nutrient and sediment regimes, shoreline development and mangrove restoration history, rather than the imprint of any single factor, and that this composite setting is expressed in both the DNA-associated and the RNA-associated communities.

Divergent inter-domain co-occurrence networks reveal contrasting microbial co-occurrence regimes

The pronounced divergence between eastern and western microbial communities documented above raises the question of whether such taxonomic and metabolic differences also extend to the underlying patterns of microbial association. Prokaryotes and microbial eukaryotes are well-recognized drivers of the biogeochemical cycling in mangrove sediment (Kristensen et al., 2008; Helfer and Hassenrück, 2021) to benthic primary production, bacterivorous grazing, and nutrient regeneration by protists (Zhu et al., 2017; Hu et al., 2021). Yet their co-occurrence relationships and inter-domain interactions remain poorly characterized in mangrove systems. To address this, inter-domain co-occurrence networks were constructed for the four sample groups (East-DNA, East-RNA, West-DNA, West-RNA), each incorporating both prokaryotic and eukaryotic OTUs. Pre-selection of representative taxa and validation against Erdős–Rényi random networks confirmed the robustness of all four networks, which displayed significant small-world properties. Eastern networks were substantially larger and more interconnected than western counterparts, with correspondingly more hub taxa (Table 1). When 10% of samples were withheld and the networks rebuilt, the original hub OTUs were only partly recovered, especially in the RNA networks, but the same genera and families repeatedly occupied hub positions (Supplementary Table S10). The East–West difference in hub taxa therefore reflects a repeatable taxonomic pattern rather than an incidental outcome of network construction. DNA networks were consistently denser, while RNA networks exhibited higher modularity. DNA-associated networks integrate the standing stock of taxa, including dormant organisms that may constitute a microbial seed bank (Lennon and Jones, 2011), and therefore capture a broader spectrum of co-occurrence relationships. The more modular RNA-associated networks, by contrast, reflect the co-occurrence among taxa detected in the transcriptionally active proxy, where modular guilds correspond to specific metabolic pathways such as sulfur cycling, nitrogen transformation, and methanogenesis (Du et al., 2023). Greater network complexity has been associated with enhanced functional redundancy and stronger biodiversity–multifunctionality linkages (Banerjee et al., 2018; Wagg et al., 2019), whereas environmental perturbation tends to reduce network complexity and cooperative interactions (De Vries et al., 2018; Yuan et al., 2021). The denser eastern networks therefore suggest a greater capacity for stable potential functional associations within the sediment microbiome, while the chronically variable estuarine conditions of the western coast appear to constrain the establishment and persistence of such relationships.

To resolve the core microbial assemblages underpinning these networks, hub subnetworks were extracted by retaining hub taxa, identified from within-module and among-module connectivity, together with their directly associated nodes (Figure 6). This approach isolates the structurally and central co-occurrence links and provides a focused view of how prokaryotic and eukaryotic taxa co-assemble into potential functional associations. The degree of overlap between DNA and RNA subnetworks diverged sharply between coasts. Eastern subnetworks shared 307 common taxa across DNA and RNA levels (~18–21% of each subnetwork), dominated by Deltaproteobacteria (85 taxa, primarily Desulfobacterales and Syntrophobacterales), Bacteroidetes (42 taxa, largely Bacteroidales), Gammaproteobacteria (38 taxa), and Bacillariophyceae diatoms (12 taxa). Western subnetworks, in contrast, shared merely 12 common taxa (~3–6% of each subnetwork), led by Bacteroidetes (3 taxa) and Deltaproteobacteria (2 taxa). The high DNA–RNA overlap in the east signifies that the taxa forming the structural backbone of the community are largely the same taxa recovered in the RNA-associated, transcriptionally active proxy, a hallmark of functional congruence between the DNA-associated and RNA-associated community fractions. Such congruence reflects both the reliability of the observed network in capturing in situ ecological relationships and the long-term environmental stability that allows inter-domain interactions to be established and maintained without disruption. The minimal overlap in the west indicates near-complete decoupling between the DNA-associated and RNA-associated interaction cores, consistent with a seed-bank-driven regime in which the DNA-associated pool retains a dormant reservoir accumulated over past conditions while the RNA-associated pool responds to transient environmental fluctuations (Lennon and Jones, 2011). These contrasting overlap patterns do not reflect a difference in ecological quality between coasts but rather two distinct ecological scenarios: in the eastern system, a stable environment sustains consistent microbial associations in which DNA-associated and RNA-associated communities coincide, whereas in the western system, pronounced environmental variability decouples the DNA-associated and RNA-associated layers, with microbial composition and interactions tracking the prevailing conditions at each point in time.

Modular structures within these subnetworks provided further resolution on the functional logic of microbial co-occurrence in each system. Modules in co-occurrence networks are typically interpreted as operationally coherent groups of taxa with shared environmental preferences or metabolic interdependencies, and frequently correspond to recognizable biogeochemical guilds (Deng et al., 2012; Du et al., 2023). In the eastern subnetworks, a recurring potential association among taxa putatively involved in sulfate reduction–polymer degradation–benthic production was identified across both DNA and RNA levels. East-DNA Module 3 integrated Desulfobacterales (Deltaproteobacteria), Bacteroidales (Bacteroidetes), Bacillariophyceae diatoms, and Spirotrichea ciliates within a single module. Within this assemblage, Desulfobacterales mediate the terminal step of anaerobic organic matter mineralization, with sulfate reduction accounting for up to 60–90% of total carbon oxidation in subtropical mangrove sediments (Alongi et al., 2005; Kristensen et al., 2008). Bacteroidetes (including Bacteroidales) are widely recognized as important contributors of carbohydrate-active enzymes involved in the degradation of complex polysaccharides in marine and mangrove sediments (Paixão et al., 2021; Chen et al., 2024). Bacillariophyceae diatoms contribute autochthonous labile carbon through benthic photosynthesis (Zhu et al., 2017), supplementing allochthonous detrital inputs, while Spirotrichea ciliates regulate bacterial biomass and accelerate dissolved nutrient regeneration through bacterivorous grazing (Hu et al., 2021). The co-occurrence of these guilds within shared modules outlines an integrated sediment potential functional association linking polymer hydrolysis, fermentation, terminal anaerobic respiration, benthic primary production, and grazing-mediated nutrient regeneration. East-RNA Modules 4 and 5 recapitulated this assemblage in the active fraction, with the addition of Anaerolineales (Chloroflexi). Anaerolineaceae are recognized primary fermenters producing acetate, propionate, and H 2 (McIlroy et al., 2017; Xia et al., 2016), substrates among the preferred electron donors for sulfate-reducing bacteria (Plugge et al., 2011), and their co-occurrence with Desulfobacterales in the active modules is consistent with a potential fermentation–terminal reduction associations. East-RNA Module 2 further linked Syntrophobacterales with Euryarchaeota methanogens, raising the possibility of syntrophic fermentation–methanogenesis associations (Stams and Plugge, 2009), although direct metabolic exchange awaits experimental confirmation.

Western subnetworks reflected a markedly different mode of microbial organization. West-DNA Module 1 grouped Thiotrichales sulfur oxidizers with Gyrista diatoms, a potential sulfur oxidation–primary production association. The dominance of sulfur oxidizers rather than reducers in the western sulfur-cycling guild is consistent with frequent redox oscillations driven by tidal flushing and salinity fluctuations, which favor sulfide re-oxidation over sustained sulfate reduction (Kristensen et al., 2008). West-RNA Module 3 grouped Betaproteobacteriales with Opisthokonta fungi into a potential heterotrophic degradation association geared toward the rapid turnover of allochthonous organic matter delivered by Pearl River discharge. Rather than the potential multi-trophic, inter-domain associations sustained under stable eastern conditions, the western sediments support functionally targeted modules that may be individually less complex but are responsive to the prevailing environmental conditions. The contrast between the two coasts is therefore not one of network complexity alone but of underlying mechanism: the eastern system sustains complex, potential inter-domain associations linking autotrophic production, polymer hydrolysis, fermentation, terminal anaerobic respiration, and eukaryotic grazing, consistently recovered at both DNA and RNA levels. The western system supports environment-responsive modules with minimal inter-domain integration and pronounced decoupling between structural and active layers. These contrasting interaction architectures suggest that the ecological consequences of differing environmental regimes extend beyond shifts in species composition to reshape the fundamental organization of microbial metabolic networks in mangrove sediments (Banerjee et al., 2018).

Microbial network architecture and DNA–RNA congruence as potential indicators for mangrove ecosystem assessment

Integration of DNA- and RNA-based community profiling with inter-domain co-occurrence network analysis reveals that the eastern and western coast host distinct microbial assemblages with divergent ecological strategies. Eastern networks were substantially larger and more densely connected, exhibiting high DNA–RNA congruence (307 shared taxa within hub subnetworks). Such features reflect specialized, tightly coupled assemblages dominated by anaerobic fermenters and sulfate reducers. In contrast, western networks were sparse and fragmented, with merely 12 taxa shared between DNA and RNA datasets. This pattern points to loosely connected generalist communities enriched in opportunistic heterotrophs, alongside strong decoupling between DNA-associated and RNA-associated fractions. These divergent interaction regimes demonstrate that environmental conditions not only shift taxonomic composition but fundamentally reshape the architecture of microbial association networks.

Traditional mangrove ecosystem assessments mainly rely on vegetation structure, coverage area and selected abiotic parameters. Although indispensable, these metrics frequently lag behind shifts in underlying ecological processes. Notably, multi-level degradation of community structure and function may persist even where mangrove coverage has recovered or areas receive formal protection (Wang et al., 2020). Mangrove plants exhibit slow growth and strong physiological tolerance; hence, detectable shifts in vegetation traits (stem density, biomass, species composition) only emerge following long-term stress, usually over multi-year to decadal timescales (Alongi, 2008; Xiong et al., 2019). Furthermore, systematic field vegetation surveys require considerable time and manpower, restricting monitoring frequency and spatial coverage. By contrast, high-throughput sequencing of sediment DNA and RNA can be finished within weeks post-sampling and easily expanded to broad sampling ranges using standardized protocols, offering near-real-time insights into the current organizational state of the sediment microbiome.

A key methodological strength of this study is the parallel use of DNA and RNA templates. DNA-associated profiles represent the standing community, encompassing living, dormant and relic organisms, whereas RNA-associated profiles provide a transcriptionally active proxy. Parallel DNA–RNA profiling mitigates confounding effects induced by relic DNA and helps identify taxa that are transcriptionally represented at the time of sampling, and therefore more likely to contribute to contemporaneous biogeochemical processes. Analyzing such datasets via inter-domain co-occurrence networks generates quantifiable metrics, including network size, connectivity, modularity, and critically, the extent of DNA–RNA overlap, to characterize the organizational state of sediment microbiomes. As observed at eastern mangrove sites, high network complexity accompanied by strong DNA–RNA congruence indicates ecosystems where potential functional associations (polymer degradation, fermentation, sulfate reduction, benthic production) appear structurally persistent. In contrast, low network complexity and limited DNA–RNA overlap at western sites suggest an opportunistic community regime shaped by stronger external environmental forcing. Such network-based traits reveal ecosystem organizational features undetectable via vegetation surveys or bulk environmental monitoring, and potentially respond faster to environmental changes and restoration progression.

Systematic reviews have shown that microbial communities vary distinctly across intact, restored and disturbed mangrove forests, and serve as core drivers of nutrient cycling and ecosystem adaptability (Lai et al., 2022). The Mangrove Microbiome Initiative also advocates shifting research focus from descriptive taxonomic profiling to hypothesis-testing investigations on microbial interactions and their associations with ecosystem functions and restoration success (Allard et al., 2020). By connecting environmental conditions, inter-domain interaction structure and DNA–RNA congruence, this study establishes a framework to incorporate microbial network metrics into mangrove ecosystem evaluation and restoration monitoring. We recognize inherent limitations of the amplicon framework used here. Primer choice and PCR amplification introduce taxonomic bias: even widely used EMP primers do not recover all lineages equally, and a single mismatch can systematically under- or over-represent particular marine groups (Parada et al., 2016). The 16S pair 515F/806R and the 18S V9 pair 1391F/EukBr were adopted for broad coverage, but the latter is not fully domain-specific and can co-amplify prokaryotic templates in bacteria-rich matrices (Amaral-Zettler et al., 2009; Stoeck et al., 2010). Differential amplification can therefore shift relative abundances independently of ecological change. We used the same primer sets and PCR conditions for paired DNA and cDNA templates from the same sediments and pooled triplicate reactions, so these biases are shared across templates and coasts. The east–west contrast is drawn from community and network patterns that persist under this common amplification framework. The same tables constrain network inference. High-throughput amplicon data are compositional: sequencing imposes an arbitrary total, so taxon proportions are not independent (Gloor et al., 2017). Under this constraint, standard correlations, including Spearman rank correlations, can generate spurious associations and can even reverse the sign of true relationships, particularly in lower-diversity assemblages (Friedman and Alm, 2012). Sequencing data alone also cannot show that taxa interact metabolically. Network edges are therefore interpreted as potential functional associations. To limit weak links, only strong and significant correlations (|r| > 0.8, p < 0.01) among OTUs that passed prevalence filters were retained, and the four empirical networks were compared with Erdős–Rényi random networks of equal size. The east–west contrast in network size, modularity and DNA–RNA overlap was recovered in both DNA-associated and RNA-associated networks, indicating that the main organizational difference is not the product of a single correlation threshold. Node and edge counts should be read comparatively. Future studies combining shotgun metagenomes and metatranscriptomes, together with compositionally robust network inference, would better test whether this organizational contrast holds beyond the present amplicon design. Nevertheless, the consistent network patterns observed across DNA-associated and RNA-associated datasets, together with their linkage to core biogeochemical pathways within modules, support the utility of microbial network architecture as a complementary indicator for evaluating mangrove ecosystem status. Future work should test whether these network metrics can track ecosystem recovery along restoration chronosequences. Follow-up manipulative experiments and long-term longitudinal monitoring are required to validate our findings. The contrasting microbial interaction regimes between the two coasts—complex, highly congruent networks dominated by anaerobic fermenters and sulfate reducers in the east versus sparse, decoupled networks enriched in opportunistic heterotrophs in the west—are conceptually summarized in Figure 7. With mangrove ecosystems facing escalating pressures from coastal development and climate change, elucidating the microbial underpinnings of ecosystem functioning is essential for designing conservation strategies that protect both the forests and the microbiomes that sustain their ecological services.