Introduction

Coastal zones serve as critical interfaces where marine and terrestrial ecosystems interact. Rivers and other surface runoffs were historically considered as the primary nutrient source1. Moore et al. (1999) introduced the concept of the “subterranean estuary” (STE) to describe coastal groundwater mixing zones2. Studies have revealed that STE-derived nutrient fluxes can equal or exceed terrestrial inputs3, attributable to the ubiquity of submarine groundwater discharge (SGD) across coastlines4. SGD, defined as water flow through permeable sediments or karst conduits, transports terrestrial nutrients, pollutants, and trace elements to coastal waters4,5,6, significantly influencing coastal ecosystems and biogeochemical processes7,8. However, excessive SGD-derived nutrients may cause ecological events including eutrophication, harmful algal blooms, and coral reef degradation9,10. Despite reducing anthropogenic nutrient discharges from wastewater and agricultural runoff globally, coastal eutrophication persists due to SGD11,12,13, severely impacting fisheries, aquaculture, and tourism industries5,14. Previous research on SGD-derived chemical loads focused on environmental variables like temperature, salinity, organic carbon supply, and solute concentrations15,16,17,18. Chemical composition of groundwater entering marine systems is ultimately regulated by STE microbial communities19,20,21, whose characteristics and metabolic functions remain poorly understood, with only limited studies addressing SGD microbial diversity and functionality22,23.

STEs function as dynamic mixing zones and continuously transport diverse microbial assemblages derived from both terrestrial and marine sources24. The steep physicochemical gradients within STEs (e.g., freshwater-seawater interfaces, oxic-anoxic transitions) shape spatially highly heterogeneous microbial communities24, while hydrodynamic processes such as tidal fluctuations can induce drastic niche shifts within hours, leading to high microbial diversity25,26. In such dynamic environments, a central scientific challenge lies in distinguishing which microorganisms represent the truly active and adapted24. This distinction is critical for accurately understanding STE microbial functionality. Multiple studies have observed that, despite fluctuating environmental conditions, a core microbial community persists in STEs, comprising taxa such as Proteobacteria, Bacteroidota, Firmicutes, and Desulfobacterota25. The relative sequence abundances of these core taxa exhibit both resilience and stability across different STE environments25, with compositions showing limited temporal or spatial variability27,28. These core taxa drive biogeochemical cycling of key elements (e.g., carbon, nitrogen, phosphorus, sulfur)1 by utilizing terrestrial plant-derived organic matter29, marine-sourced organic material24, or degrading organic pollutants1,20,30, thereby serving as vital engines for coastal ecological sustainability. Of particular importance is the evolution of highly adaptive strategies among these microbes to cope with the complex and fluctuating environmental pressures in STEs, including anthropogenic disturbances. Among these, dormancy mechanisms enable survival under unfavorable conditions and facilitate rapid reactivation when conditions improve31—this may be a key mechanism maintaining the stability and functional resilience of the core community25. Nevertheless, the influences of human activities on the active state, community stability, and ecological functions of STE microbes remains poorly understood24.

Our research originally planned to use “high anthropogenic disturbance” and “low anthropogenic disturbance” as experimental controls, but designing field sites simultaneously meeting both conditions in coastal zones-areas typically subject to intensive human activities—proved challenging. The COVID-19 pandemic unexpectedly provided ideal experimental conditions. Lockdown restricted nearly half the global population32,33, drastically reducing ecological interference of human activities. Urban closures not only mitigated pandemic spread but also suspended industries like aquaculture, trade, and tourism, leading to unprecedented declines in anthropogenic emissions within STE regions. For this study, we selected Sanggou Bay in Shandong Province, China as our field site. During summer 2018 (Pre-COVID), the area hosted extensive tourism and marine cultural activities, whereas the COVID-19 period (2020–2022) led to restrictions on tourism, mariculture, and farming, minimizing human impacts. According to stress gradient theory34, such dramatic shifts in human activity intensity represent varying environmental stress levels that may alter benthic microbial communities. The transition from Pre-COVID (summer 2018) to Post-COVID (summer 2022) transformed the local environment from high to low human stress, creating a coastal ecological rest phase. In restoration ecology, rest constitutes mitigation through reconstruction, rehabilitation, or governance; here, it manifested as progressive reduction of human disturbance. We hypothesize that this rest phase may enhance microbial community stability and ecosystem service functionality. The concept of ‘rest’ has been applied in practices such as agricultural fallowing and rest-grazing in pastoral systems. These practices enhance soil microbial functional diversity and metabolic potential35, and improve soil quality by reshaping microbial communities36. However, coastal ecosystem rest effects remain severely understudied.

Through integrated 16S rRNA gene sequencing, metagenomic analysis, and environmental parameter measurements, we examined microbial community changes mediated by coastal resting effects from multiple perspectives: community diversity, assembly processes, network stability, and functional traits. The study objectives were: (1) to compare human-mediated physicochemical characteristics in STE seepage faces between the Pre-COVID and Post-COVID environmental states; (2) to investigate diversity patterns and mechanistic drivers of microbial communities under contrasting levels of anthropogenic nutrient stress; and (3) to analyze resting-induced microbial stability and metabolic potential, elucidating microbial responses to anthropogenic pressure differentials. As a comprehensive investigation of pandemic-era STE microbial dynamics, this research establishes theoretical foundations for understanding coastal resting effects on sediment microbiomes.

Results

Changes in physicochemical properties

To characterize the environmental stress gradients induced by varying anthropogenic disturbances during Pre-COVID and Post-COVID periods, physicochemical parameters were determined (Fig. S2). Results demonstrated significant variations in all measured indicators except MPS between the two periods (Fig. 1), with marked decreases (P < 0.01) in sediment TOC (Pre-COVID: 62.50–95.80 μmol/g, Post-COVID: 23.7–48.4 μmol/g), TN (Pre-COVID: 4.07–6.24 μmol/g; Post-COVID: 2.73–5.45 μmol/g), CDOM, TS, RP, Chl-a, and SP, all exhibiting similar declining trends (Fig. S2). Sanggou Bay sediments showed MPS ranges of 187–249 μm and SP dominated by sandy particles (>86%), with significant physical changes post-pandemic (Fig. 1). Vertically, most parameters displayed no spatial variation except for MPS, TS, RP, and SP in Post-COVID samples (Fig. S3).

Fig. 1: The box plot illustrates the differences in physicochemical factors between Pre-COVID (n = 15) and Post-COVID (n = 15) periods.
figure 1

Significance levels: *P < 0.05, **P < 0.01, and ***P < 0.001. Other abbreviations indications are as in Fig. S2.

Full size image

Microbial community diversity, composition, and assembly processes

A total of 1,015,193 high-quality bacterial 16S rRNA sequences were obtained (19,204–84,750 sequences/sample), normalized to 10,402 ASVs for analysis. The index of α-diversity differed significantly between periods (p < 0.001; Fig. 2A). PCoA based on Bray-Curtis distances showed distinct clustering (p < 0.001; Fig. 2B), with Post-COVID exhibiting lower dissimilarity (Fig. 2C). Spatial analysis revealed Pre-COVID communities had decreasing α-diversity from the upper to the lower intertidal zones (Fig. S4A) and significant heterogeneity (Fig. S4C), while Post-COVID showed no spatial variation (Fig. S4D). Among 45 phyla, the top 10 accounted for 94.13% abundance, dominated by Proteobacteria (39.48%) and Bacteroidota (23.84%) (Fig. 2D). Pre-COVID had more unique ASVs (Fig. 2E), but shared core ASVs (2385) were more abundant in the Post-COVID samples (Fig. 2F). Further analysis of core taxa revealed that within the top 100 core ASVs, Proteobacteria and Bacteroidota collectively represented the dominant phyla (Fig. S5A), and exhibited significant abundance differences between the Pre- and Post-COVID periods (Fig. S5B). At the genus level, Woeseia, Sediminicola, and Gramella were identified as dominant taxa (Fig. S5C). Null model analysis indicated deterministic processes dominated Pre-COVID assembly versus stochastic processes Post-COVID (Fig. 2G, H).

Fig. 2: Microbial community diversity, composition, and assembly processes.
figure 2

The box plot (A) illustrates differences in α-diversity (ACE and Shannon) indices of benthic microorganisms between Pre-COVID and Post-COVID periods, while PCoA (B) visualizes bacterial community β-diversity. Bray-Curtis dissimilarity (BCD) analysis is shown in boxplots (C), and the relative richness of dominant bacterial phyla at the seepage surface is displayed (D). A Venn diagram (E) highlights shared and unique ASVs between the two periods, and a box plot (F) compares the relative abundance of core ASVs. Phylogenetic and taxonomic turnover distributions, based on βNTI and RCbray values, are presented (G), followed by the proportions of assembly processes across Pre and Post-COVID periods (H). Significance levels: *P  < 0.05, **P < 0.01, ***P <  0.001.

Full size image

The microbial community is affected by physicochemical factors

To elucidate how environmental factors influenced microbial community succession pre- and post-pandemic, redundancy analysis (RDA) was employed to reveal relationships among environmental variables, samples, and microbiota. RDA demonstrated that interannual differences in environmental variables were primary drivers of microbial shifts, explaining 70.78% of total bacterial community variation (Fig. 3A). Axis 1 represented the dominant influencing gradient, with TOC, CDOM, Chl-a, TS, TN, RP, and MPS all exerting significant effects (P < 0.05). TOC and CDOM showed highest explanatory power (r² = 0.882 and r² = 0.845), identifying them as core determinants of community structure. Mantel tests (Fig. 3B, Table S1) revealed significant associations (P < 0.05) between microbial community metrics and environmental drivers. Specifically, microbial composition correlated positively with TOC, CDOM, and SP; species richness and diversity were linked to TOC, CDOM, TS, RP, and Chl-a (with diversity additionally tied to MPS); and core species abundance showed strong connections to TOC, CDOM, and Chl-a. Notably, the strongest correlation emerged between core species abundance and TOC (r = 0.551, p = 0.001), underscoring its pivotal role in shaping microbial dynamics. Environmental variables were categorized as chemical factors (TOC, CDOM, Chl-a, TN, TS, RP) and physical factors (MPS, SP). Variation partitioning analysis (VPA, Fig. 3C) confirmed chemical factors’ greater relative importance (36.14% vs. physical factors’ 4.85%), with their joint contribution accounting for 12.02% of the variation. Collectively, measured environmental variables explained 53.01% of microbial community changes, underscoring their pivotal role in shaping seepage-face microbiota during pandemic-induced transitions. Random forest modeling validated CDOM, TOC, Chl-a, and TS as key predictors of bacterial composition, richness, diversity, and core species abundance dynamics.

Fig. 3: The microbial community is affected by physicochemical factors.
figure 3

Redundancy analysis (RDA) for physicochemical factors and benthic bacterial communities (A), Mantel test evaluates correlations between bacterial composition, richness, diversity, core speciesrelative abundance and various physicochemical parameters (B). Variance partitioning analysis (VPA) quantifies the relativecontributions of chemical and physical factors to benthic microbial community variations (C). Random forest analysis reveals thepredictive importance (percentage increase in mean square error) of environmental factors driving bacterial community succession incomposition (D), richness (E), diversity (F), and core species relative abundance (G). Significance levels: *P < 0.05, **P < 0.01, ***P < 0.001, with “ns” indicating non-significance (other abbreviations as in Fig. S2).

Full size image

Microbial co-occurrence networks, keystone taxa, and community stability

To investigate non-random assembly patterns of microbial communities in subterranean estuary seepage-face sediments under different environmental variables across two timepoints, we constructed co-occurrence networks using Spearman correlation (|r| > 0.6, P < 0.05; Fig. 4A, B) after ASV screening for Pre-COVID and Post-COVID periods. The microbial network topology shifted markedly: Pre-COVID networks contained 177 nodes and 757 edges (Fig. 4A), while post-COVID showed marginally fewer nodes (175) but increased edges (1205) (Fig. 4B). Positive interactions declined from 76.62% to 64.48%, whereas negative interactions rose from 23.38% to 35.52% (Fig. 4A, B). Topological parameter analysis revealed significant increases in Degree and Closeness centrality during Post-COVID (Fig. 4E), while Betweenness and Eigenvector centrality remained unchanged. Additionally, analysis of microbial network cohesion revealed significantly higher cohesion in Post-COVID compared to Pre-COVID (Fig. 4E), demonstrating stronger microbial interaction strengths in the Post-COVID period. Assessment of community stability via the AVD showed that Post-COVID AVD values were significantly lower than those of Pre-COVID (Fig. 4E), further supporting enhanced structural stability in Post-COVID microbial communities. Zi-Pi analysis identified keystone species (Fig. 4C, D): Pre-COVID had 2 module hubs (ASV101, ASV128, Table S2) and 22 connectors (Table S3), while Post-COVID featured 16 connectors but no module hubs (Table S4). Collectively, these findings indicate that despite reduced overall diversity during the post-COVID period, microbial communities developed stronger interconnections and complexity, yielding greater stability.

Fig. 4: Comparative analysis of bacterial community networks across pandemic periods.
figure 4

Co-occurrence networks illustrate bacterial interactions during Pre-COVID (A) and Post-COVID (B) periods, where nodes represent ASVs (size scaled to degree, colors denoting phylum affiliation), with red/blue edges indicating positive/negative correlations, respectively. Zi-Pi plot visualizations for Pre-COVID (C) and Post-COVID (D) periods reveal taxa interaction patterns. Network topology metrics (degree, betweenness, closeness, eigenvector centrality) alongside cohesion and average variation degree show statistically significant differences (Wilcoxon rank-sum test, *P < 0.05, **P < 0.01, ***P < 0.001) between periods (E).

Full size image

To investigate the regulatory mechanisms of environmental factors on microbial community structure and stability before and after the pandemic, we employed Partial Least Squares Path Modeling (PLS-PM) to elucidate multi-level causal relationships (Fig. 5A). Key findings demonstrated that chemical factors (Chemical) served as the central hub of environmental driver networks, exhibiting significant negative regulation on core microbial relative abundance (Core) (β = −0.874, p < 0.001). Core microbes displayed dual regulatory roles: a negative effect on richness (β = −0.576, p = 0.016) and a stronger negative correlation with diversity (β = −0.783, p = 4.42e−06). Standardized effects analysis (Fig. 5B) identified core microbes (Core) as the most robust positive predictor of stability (β = 0.545, p = 0.023), highlighting their pivotal role in maintaining community resilience despite pandemic-induced environmental changes. These results systematically unravel how pre-versus post-pandemic environmental shifts reconfigured microbial interaction networks, with chemical drivers predominantly modulating core populations that ultimately determined community stability thresholds.

Fig. 5: Path analysis of environmental and microbial drivers shaping community stability and metabolic potential.
figure 5

The partial least squares path model (PLS-PM) reveals direct/indirect effects of physical-chemical factors, bacterial richnessersity/composition, andcore species relative abundance on community stability (A), with standardized effect magnitudes shown in bar charts (B). Modelvalidity was assessed via goodness-of-fit (GOF), where solid/dashed arrows indicate positive/negative relationships, respectively(arrow width reflects path coefficient strength; latent variable values show determination coefficients). Significance thresholds: *P < 0.05, **P < 0.01, ***P < 0.001. Ordinary Least Squares (OLS) quantifies the relationships between: C/N ratio and core species relative abundance (C), core species abundance and community stability (D), and community stability and metabolic potential (Metabolism) (E).

Full size image

Changes in microbial functions and related genes

The resting effect in coastal zones significantly influenced microbial community functionality. Principal coordinates analysis (PCoA) of metabolic function distribution revealed patterns similar to bacterial composition (Fig. S6A), with significant clustering differences between Pre-COVID and Post-COVID periods (PERMANOVA, p < 0.001). Six level-1 metabolic pathways were identified from the Kyoto Encyclopedia of Genes and Genomes (KEGG) database annotations as representative microbial metabolic routes in seepage face sediments (Fig. S6B), with Metabolism predominating at 69.71% abundance among KEGG L1 pathways. From the Pre-COVID to the Post-COVID period, Metabolism showed a significant increase (p < 0.001; Fig. 6A), indicating enhanced metabolic activity.

Fig. 6: Differential abundance of KEGG functional pathways before and after the COVID-19 lockdown.
figure 6

Wilcoxon rank-sum tests identify significant differences in KEGG level 1 functional pathways between Pre-COVID and Post-COVID states (A). Wilcox rank sumtests identify significant differences in KEGG level 2 functional pathways between Pre-COVID and Post-COVID states (B).

Full size image

Metagenomic annotation of microbial genes related to biogeochemical cycles enabled detailed analysis of carbon (C), nitrogen (N), phosphorus (P), and sulfur (S) cycling gene variations between periods, assessing microbial functional potential in these processes (Table S5). Pre-COVID communities preferred lignin degradation (Fig. 7A), while Post-COVID shifted toward fatty acid synthesis (Fig. 7B), reflecting altered carbon source availability; nitrification dominated Pre-COVID (potentially increasing N2O emissions), whereas Post-COVID showed enhanced denitrification (likely reducing greenhouse gas release; Fig. 7D); Post-COVID exhibited improved inorganic phosphorus utilization capacity, potentially alleviating phosphorus limitation (Fig. 7E); and dynamic sulfur oxidation-reduction pathway shifts may reflect environmental sulfur speciation changes or redox state fluctuations (Fig. 7F). Further species contribution analysis of differentially abundant genes identified key microbial drivers (Fig. S7). Species contribution analysis revealed that although dominant core taxa like Woeseia and Sediminicola contributed substantially to functional shifts, unclassified taxa also demonstrated high contributions (Fig. S7).

Fig. 7: Comparative relative abundance of microbial genes involved in biogeochemical cycling before and after the COVID-19lockdown.
figure 7

C-hydrolysis genes (A), C fixation genes (B), methane metabolism genes (C), N cycling genes (D), P cycling genes (E), and S cycling genes (F). Values formatted as X/Y denote relative abundances in Pre-COVID/Post-COVID samples respectively. Significance levels: *P < 0.05, **P < 0.01, ***P < 0.001.

Full size image

Discussion

The global scale outbreak of COVID-19 from 2019 and subsequent global lockdown measures led to significant reductions in tourism, fisheries, and industrial activities, resulting in an extended period of coastal resting that provided unexpected opportunities for ecological recovery. Benthic microbial community stability and functional metabolic potential play crucial roles in subterranean estuary seepage-face environments23,24. Previous studies on spatiotemporal variations of coastal bacterial communities have demonstrated that environmental changes can induce significant structural shifts37,38,39,40. However, few reports have addressed how sustained reductions in environmental stress might affect the structure, stability, and metabolic potential of microbial communities in subterranean estuary seepage faces24,25. This study provides novel evidence supporting the positive impacts of coastal resting on seepage-face microbiomes through comprehensive analyses of environmental characteristics, microbial community structural stability, and functional metabolic potential. These findings corroborate our hypothesis that the coastal rest period led to a more stable microbial community structure, as indicated by reduced compositional variability and enhanced network complexity, alongside an increased functional metabolic capacity. This suggests a potential mechanism for improving ecosystem service functions in subterranean estuaries under reduced anthropogenic pressure.

The significant differences in physicochemical parameters in Sanggou Bay sediments before and after COVID-19 reflect varying environmental stress levels, particularly regarding nutrient levels and pollution. During Pre-COVID, this STE experienced intensive urbanization, tourism, and agricultural activities that directly contributed to substantial nutrient inputs. A portion of these organic materials was transported to coastal waters via submarine groundwater discharge (SGD), becoming key chemical components of the sedimentary organic matter41, while the remainder stimulated surface water primary productivity through excessive nutrient loading42,43. This created escalating threats of eutrophication and ecological imbalance44,45. Elevated eutrophication promotes algal blooms and subsequent organic deposition46, with high sediment TOC concentrations linked to algal proliferation47. Additionally, large-scale mariculture operations globally raise concerns about eutrophication risks, particularly through sediment phosphorus accumulation48,49. Intensive farming practices that increase stock density and feed inputs result in low nutrient utilization efficiency50—11.7–27.7% for nitrogen and 8.7–21.2% for phosphorus51, with only 4–27.4% of feed actually consumed50. Excess feed deposits as organic-rich sediment containing elevated C, N, and P, potentially stimulating sulfide overproduction52. However, the two-and-a-half-year coastal resting period Post-COVID mediated significant reductions in sediment TOC, CDOM, TN, TS, RP, and Chl-a levels Fig. 1), demonstrating its pronounced effect on alleviating nutrient loads in subterranean estuary seepage face zones.

The C/N ratio is widely used to preliminarily assess the primary sources of sedimentary organic matter53,54. Different biological groups exhibit distinct C and N concentrations in their organic compositions—vascular plants contain abundant carbon compounds like cellulose, whereas algae are rich in nitrogen-containing proteins55,56. Marine algae typically exhibit C/N ratios between 4–8, while terrestrial plants show values >1257. In our study, seepage-face sediment C/N ratios decreased from 12.20–16.77 during the Pre-COVID to 7.09–11.92 during the Post-COVID period, indicating a shift in organic matter sources within the subterranean estuary. Higher C/N ratios generally suggest terrestrial organic matter dominance, while lower values reflect increased marine organic inputs39. This transition likely resulted from reduced human activities during the pandemic, diminishing terrestrial contributions while marine sources became relatively more significant. Furthermore, Pre-COVID mariculture operations involving artificial cage systems enhanced sediment deposition rates58, where cage structures trapped fine particles, explaining the higher sand proportion (SP) observed during the pre-pandemic. Minimal spatial variation in physicochemical factors across seepage faces may stem from the STE zone’s unique hydrogeological characteristics18. As critical ecotones, seepage faces effectively regulate hydrological processes and nutrient cycling5,6.

Seepage faces demonstrated significant and multi-faceted microbial responses to altered nutrient stocks and organic substrate sources under coastal resting conditions. Microbial diversity variations reflect habitat changes, typically increasing with resource availability and declining with nutrient reduction59,60. Our study identified TOC, CDOM, TS, RP, and Chl-a as key determinants of microbial richness and diversity (Fig. 3B, E, F). The pre-pandemic eutrophic state in Sanggou Bay supported higher α-diversity (Fig. 2A), where diverse organic/inorganic nutrients facilitated specialized species’ growth61,62. Environmental heterogeneity also shaped community structure, with PCoA and PERMANOVA revealing distinct bacterial clustering between periods (Fig. 2B). Community variations correlated strongly with nutrients (TOC, CDOM, TN, TS, RP, Chl-a) and physical factors (SP) (Fig. 4B–D), with differences diminishing as nutrients declined (Fig. 2C). These findings aligned with null model results (Fig. 2G, H): Pre-COVID communities were primarily governed by heterogeneous selection (84.4%, deterministic processes), whereas Post-COVID assembly shifted toward homogeneous dispersal (72%) and selection (14.2%), emphasizing stochasticity. This shift can be biologically explained by reduced environmental heterogeneity following the dramatic decline in human activities. During the Pre-COVID period, strong and varied anthropogenic pressures (e.g., nutrient inputs from aquaculture and tourism) created heterogeneous selective conditions that favored deterministic assembly. In contrast, the coastal “rest” period significantly lowered nutrient loads and shifted organic matter sources toward more homogeneous marine-dominated substrates, thereby weakening selective filtering. Analogous patterns were observed in wastewater treatment systems, where microbial communities transitioned from deterministic to random assembly under disturbances63.

At the phylum level (Fig. S5), dominant bacterial groups aligned with those observed in global subterranean estuary studies24,64, where Proteobacteria and Bacteroidota predominated, accompanied by Chloroflexi, Actinobacteriota, Acidobacteriota, Gemmatimonadota, Myxococcota, Desulfobacterota, and Firmicutes as characteristic taxa. Notably, core species analysis revealed significantly higher relative abundance in Post-COVID versus Pre-COVID (Fig. 2F). Environmental filters mediated this shift23,65, reflecting microbial adaptive adjustments to altered conditions. Our prior work demonstrated strong substrate dependency in Sanggou Bay’s microbial succession21, where higher sediment C/N ratios indicate poorer carbon availability (lower substrate quality) and vice versa66. The substrate source transition stimulated core species proliferation (Fig. 5C), with phylum-level preferences becoming evident: Proteobacteria, Acidobacteriota, and Chloroflexi were favored by high-C/N substrates, while Firmicutes and Bacteroidota thrived under low-C/N conditions66, their abundance changes matching these metabolic strategies. Coastal resting severed terrestrial inputs Post-COVID, shifting seepage-face substrates from land-derived to marine-dominated sources. Under resource limitation, communities transitioned from “diversity exploration” to “efficient utilization,” enriching core species adapted to labile marine substrates.

Coastal resting elevated core species’ relative abundance and restructured microbial community architecture, mediating enhanced stability and complexity (Fig. 5). Microbial co-occurrence networks (where nodes represent individual bacterial taxa and edges represent significant correlations between them) exhibited marginally reduced node counts but significantly increased edge numbers from Pre- to Post-COVID (Fig. 4A, B), indicating strengthened microbial interactions despite decreased α-diversity. High α-diversity doesn’t necessarily correlate with robust microbial associations67, suggesting interaction intensification likely stemmed from core species enrichment. Significant rises in bacterial degree and closeness centrality (Fig. 4E) revealed augmented node connectivity and more frequent/core species interactions, mirroring findings reported by Yuan et al.68. These modifications potentially accelerated coordinated environmental responses, bolstering community adaptability and stability. Cohesion and AVD analyses confirmed Post-COVID’s stabilized communities (Fig. 4E), with PLS-PM modeling identifying core species abundance as the primary stability regulator (p < 0.001), despite its inverse relationship with chemical factors (p < 0.001) (Fig. 5A, B). The OLS results indicated that the regulatory mechanism of chemical factors on core species may involve substrate source shifts (p < 0.001) (Fig. 5C). While Xun et al. demonstrated an association between overall richness and stability using AVD69, we identified core species abundance as the pivotal stability determinant—communities with enriched core taxa demonstrated superior stability.

Microbial interaction patterns further elucidate stability-driving mechanisms, where decreased positive interaction ratios and increased negative interactions reflect weakened cooperation and intensified competition. Pre-COVID’s relatively eutrophic conditions created diversified micro-niches and resource partitioning, promoting community differentiation that potentially alleviated nutrient/niche competition and resulted in weaker microbial associations70,71. Conversely, Post-COVID’s oligotrophic transition and substrate shifts forced microbial adaptation through competition for limited resources72, driving community restructuring from “diversity-cooperation” to “core species-competition” networks. This competitive enhancement of stability echoes the ecological theory proposed by Coyte et al.73 while cooperative networks enable efficient information processing, they exhibit lower stability compared to competition-driven systems that resist perturbations through resource limitation, functional redundancy, and dynamic reorganization74. Collectively, post-pandemic coastal resting reduced seepage-face microbial diversity but elevated core species abundance, intensified microbial interconnectivity, and enhanced complexity—ultimately conferring greater community stability and resilience.

The microbiome differences mediated by coastal resting in the subterranean estuary seepage face significantly influenced microbial biogeochemical cycling potential. Distinct functional clustering between Pre-COVID and Post-COVID microbial communities emerged, where altered environmental factors—particularly nutrient input changes—closely correlated with metabolic activity and functional performance37. KEGG analysis revealed significant Post-COVID enhancement in Metabolism, indicating improved environmental adaptation and resource-use efficiency. This aligns with studies showing low-nutrient conditions may increase microbial stochasticity, ultimately strengthening network stability to elevate functional potential75, as corroborated by our linear regression demonstrating significant stability-Metabolism correlations (Fig. 5E). Core species likely activated more metabolic genes under resource scarcity76. KEGG level-2 analyses showed Pre-COVID dominance in Energy Metabolism, Cell Motility, Membrane Transport, Folding/Sorting/Degradation, and Cofactor/Vitamin Metabolism (Fig. 6B), reflecting higher physiological activity to process terrestrial-dominated, eutrophic substrates. Lignin-degrading genes (pox, glx) exhibited greater Pre-COVID activity (p < 0.05; Fig. 7A), confirming carbon utilization preference shifts. Post-COVID’s elevated Carbohydrate/Amino Acid Metabolism indicated efficient simple-substrate assimilation, transitioning from “complex-substrate breakdown” to “high-efficiency labile-substrate utilization” to promote N/C cycling. Enhanced Transport and Catabolism highlighted improved nutrient uptake efficiency. Pre-COVID functions revealed survival strategies in terrestrial-rich environments: high-energy metabolism and motility for complex substrate degradation (Fig. 6B). Post-COVID’s functional shift toward metabolic efficiency prioritization provides theoretical foundations for coastal restoration strategies.

Analysis of differentially abundant genes in carbon hydrolysis (pox and glx) revealed significant contributions from Paracoccus, Mycolicibacterium, and Nocardioides (Fig. S7A). These taxa degrade complex organic carbon compounds, including pollutants and natural organic matter, via robust enzymatic systems, thereby facilitating carbon cycling77,78,79. For the accA gene, high contributions were observed from dominant taxa Woeseia and Sediminicola (Fig. S7B). Notably, Woeseia was identified as a keystone taxon in the co-occurrence network (Tables S2–4), functioning as both module hubs and connectors across both Pre- and Post-COVID periods80,81. This genus, widely distributed in global coastal environments, performs diverse biogeochemical functions, utilizing various organic substrates through heterotrophic metabolism and playing a critical role in carbon and nitrogen cycling. Sediminicola also showed high contribution not only to accA but also to multiple nitrogen cycling genes (nosZ, nirK, ureC, napA) (Fig. S7C), with recent isolates demonstrating nitrate reduction and starch hydrolysis capabilities82. Low-abundance taxa such as Pseudomonas contributed substantially to differential genes, supported by their versatile metabolic capacities for degrading organic pollutants (Figs. S7C, D and S5A). Functional redundancy and metabolic complementarity among multiple species facilitated collective responses to environmental changes, reflecting microbial plasticity in C-N-P-S cycling under long-term selection. This study assessed relative gene abundances rather than absolute quantities, indicating shifts in metabolic potential rather than process rates. Despite lower biomass Post-COVID, increased diversity of carbon metabolic genes—particularly in fatty acid synthesis and carbon fixation—suggested enhanced carbon use efficiency during coastal rest. The significant contributions from unclassified taxa indicate potential novel functions or unrecognized keystone groups in the STE (Fig. S7), warranting further isolation or metagenomic assembly to resolve their genomic and functional traits.

This study, through a multi-faceted, mechanistic analysis of subterranean estuary microbial communities during the unique historical context of COVID-19, demonstrates that reduced coastal activities significantly decreased nutrient loads and terrestrial organic matter inputs, driving microbial communities to transition from a “diversity exploration” to an “efficiency prioritization” strategy. This shift enhanced core species abundance, community stability, and metabolic efficiency. Our work not only provides novel insights into the dynamic characteristics of subterranean estuary ecosystems but also offers valuable information for future ecological management and conservation strategies. Future research should focus on investigating the long-term impacts of human activities on microbial diversity and ecological functions, particularly how microbes continuously adapt and adjust their ecological roles under sustained disturbance reduction scenarios. Future research and pilot studies are critically needed to evaluate the logistical feasibility, economic costs, and long-term benefits of integrating such rest periods into adaptive management frameworks, with the goal of balancing ecological and societal needs.

Methods

Study site and field surveys

Our sampling site was located in the inner coast of Sanggou Bay (122°29′07″ E, 37°08′25″N), Rongcheng City, Shandong Province, China (Fig. S1A). Characterized by a temperate monsoon climate with annual precipitation of approximately 820 mm, Sanggou Bay supports local economies through aquaculture of kelp, oysters, scallops and other marine products83, serving as a national aquaculture center. Surface runoff and SGD contribute substantial quantities of nutrients to the bay84. Terrestrial materials are transported via surface flows mainly from the Gu, Ba, Shili and Sanggan Rivers85, while porewater from STE constitutes another crucial nutrient source from coastal sedimentary environments86. Sampling was conducted during Pre-COVID (July 2018) and Post-COVID (July 2022) periods at the seepage face within the Sanggou Bay STE. All sampling activities for both Pre-COVID and Post-COVID campaigns were conducted during the neap tide period at low tide to ensure hydrological consistency. This approach minimizes the transient influence of tidal pumping and ensures that the sampled porewaters are representative of the steady-state groundwater discharge, thereby allowing for a robust comparison of the microbial communities and geochemical conditions between the two time periods18. As shown in Fig. S1B, three sampling points were selected across the upper, middle and lower intertidal zones of the seepage face. Sediment samples (0–20 cm depth) were collected using corers, immediately sectioned vertically into five layers (1–4 cm, 5–8 cm, 9–12 cm, 13–16 cm, and 17–20 cm) after discarding the surface 1 cm layer to remove shell fragments and plastic debris. This sampling design resulted in a total of 30 sediment samples (2 periods × 3 zones × 5 layers = 30) for subsequent analysis. Samples were stored in sterile bags at −20 °C during transportation. For physicochemical analysis (Fig. S2), subsamples were freeze-dried, while remaining aliquots were preserved at −80 °C for subsequent DNA extraction.

Laboratory analyses

Sediment samples were freeze-dried until a constant weight was achieved, and algal and shell debris were removed. For the determination of total organic carbon (TOC), inorganic carbon was removed using 2 mol/L hydrochloric acid, and the remaining carbon was quantified using a CHNOS elemental analyzer (Vario EL III, Germany). Colored dissolved organic matter (CDOM) was analyzed using a spectrophotometer (TU-1901, PERSEE®, China), and its content was estimated based on the absorption coefficient at 355 nm (a355)87. Total nitrogen (TN) and total sulfur (TS) in the sediment were measured using a CHNS Cube88. Reactive phosphorus (RP) content was extracted using 0.5 mol/L potassium sulfate solution and quantified using a flow injection system (SAN plus, SKALAR Analytical b.v., Netherlands). Sediment chlorophyll a (Chl-a) concentration was analyzed using a fluorometer89. For particle size analysis, organic components were removed using hydrogen peroxide (10% H2O2), and sodium hexametaphosphate (5%) was added as a dispersant before measurement with a Coulter LS 100Q instrument (Coulter Company, USA)90. The carbon-to-nitrogen ratio (C/N) for each sample was calculated using the following formula: \({\rm{C}}/{\rm{N}}=\frac{[\mathrm{TOC}]}{[\mathrm{TN}]}\) where [TOC] and [TN] represent the content (μmol·g−1) of total organic carbon and total nitrogen, respectively.

Total genomic DNA was extracted from sediment samples using the FastDNA® SPIN Kit for Soil (MP Biochemicals, Solon, OH, USA) according to the manufacturer’s instructions. DNA integrity was confirmed by 1% agarose gel electrophoresis, and concentrations were measured with a NanoDrop 2000 spectrophotometer (Thermo Fisher Scientific, USA). The bacterial 16S rRNA genes of the V3-V4 regions were amplified for sequencing using the universal bacterial primers 341F (5′-CCTACGGGNGGCWGCAG-3′) and 806R (5′-GGACTACHVGGGTATCTAAT-3′), which were previously developed and validated by Caporaso et al. (2011)91. The PCR reaction system (25 μL) included 20 ng of template DNA, 2.5 μL of TransStart® buffer (10×), 2 μL of dNTPs (2.5 mmol/L), 1 μL each of forward and reverse primers (10 μmol/L), 0.5 μL of TransStart® Taq DNA polymerase (2.5 U/μL), and ddH2O to make up the volume. The PCR conditions were as follows: initial denaturation at 94 °C for 3 min; 29 cycles of denaturation at 94 °C for 5 s, annealing at 57 °C for 90 s, and extension at 72 °C for 10 s; and a final extension at 72 °C for 5 min. PCR products were verified by 1.5% agarose gel electrophoresis, purified using AMPure XP beads (Beckman Coulter, USA), and subjected to Illumina bridge PCR amplification. The purified products were quantified using NanoDrop and mixed in a 1:1 mass ratio. The target fragments were recovered by gel extraction using 1.8% agarose gel electrophoresis. The constructed libraries were quality-controlled and sequenced on the Illumina MiSeq platform (Illumina Inc., San Diego, CA, USA) using a 2 × 300 bp paired-end sequencing protocol.

For metagenome sequencing, the extracted DNA was fragmented to an average size of approximately 400 bp using the Covaris M220 (Gene Company Limited, China). Paired-end libraries were constructed using the NEXTflexTM Rapid DNA-Seq kit (Bioo Scientific, Austin, TX, USA). Adapters containing complete primer hybridization sites were ligated to the blunt ends of the fragments. Paired-end sequencing (2 × 150 bp) was performed on the Illumina NovaSeq platform (Illumina Inc., San Diego, CA, USA) using NovaSeq reagents according to the manufacturer’s instructions (refer to manufacturer’s protocol).

Data processing

Following demultiplexing, sequences underwent quality filtering using fastp92 and paired-end merging via FLASH93. High-quality sequences were then processed with the DADA2 plugin in the QIIME2 pipeline94,95 using recommended parameters to perform sample-specific error profile-based denoising, achieving single-nucleotide resolution. Denoised sequences were used to generate amplicon sequence variants (ASVs) through DADA2, with subsequent steps including paired-read merging, chimera filtering, and construction of ASV tables at 100% sequence similarity for 16S rRNA genes. Prior to downstream analysis, sequences annotated as mitochondrial or chloroplast origin were systematically removed. To mitigate sequencing depth biases in alpha/beta diversity analyses, sequences from each sample were rarefied to 19,204, retaining an average Good’s coverage of 94.16%. Taxonomic classification of ASVs was conducted using the Vsearch consensus classifier integrated in QIIME2, referencing the SILVA 16S rRNA database (v138.1).

For the acquired metagenomic data, the quality control pipeline included filtering and trimming raw FASTQ files using fastp (v0.12.4)92. Each metagenomic sample (averaging ~30 Gb per sample) was assembled independently using MEGAHIT (https://github.com/voutcn/megahit) with a de Bruijn graph-based approach, retaining contigs >300 bp. For the assembled metagenomic data: Open reading frames (ORFs) were predicted using MetaGeneMark96. A non-redundant gene catalog was generated by clustering all sequences at 95% sequence identity (90% coverage) with CD-HIT (v4.8.1). Non-redundant gene sequences were searched against the NCBI non-redundant protein database using DIAMOND (v0.9.14.115) with an e-value cutoff of 1e−5. Pathway annotation was performed using the KEGG database (Kyoto Encyclopedia of Genes and Genomes; http://www.genome.jp/kegg/) with DIAMOND, applying an e-value cutoff of 1e−5. Functional gene content analysis of metagenomic sequences was conducted using multiple cycle-specific databases: CCycDB (Carbon Cycle Database), MCycDB (Methane Cycle Database), NCycDB (Nitrogen Cycle Database), PCycDB (Phosphorus Cycle Database), and SCycDB (Sulfur Cycle Database). Sequences underwent initial quality filtering (Phred score ≥20) and were aligned to these databases using DIAMOND (v2.0.15) with an e-value cutoff of 1e−5. KEGG Orthology (KO) assignments were determined via a best-hit approach, requiring a minimum identity threshold of 70% and alignment coverage of 70%. Functional pathway abundance was quantified by counting reads mapped to each KO, normalized by the total mapped reads per sample to account for sequencing depth biases. Normalized abundances were then aggregated based on their roles in specific steps of CNPS (Carbon, Nitrogen, Phosphorus, and Sulfur) cycling pathways.

Statistical analysis

All statistical analyses were conducted in R v4.4.2 with RStudio (2023.09.1 Build 494). Alpha diversity indices (ACE and Shannon) of benthic bacterial communities were calculated using the vegan package. Differences in alpha diversity and environmental factors between Pre-COVID and Post-COVID periods were assessed using the Wilcoxon rank-sum test, while comparisons among Upper, Middle, and Low zones were performed using the Kruskal-Wallis test, followed by Dunn’s post-hoc test for pairwise comparisons. Community compositional and functional dissimilarities were visualized using principal coordinate analysis (PCoA). Permutational multivariate analysis of variance (PERMANOVA; vegan package) based on Bray-Curtis dissimilarity matrices was applied to evaluate beta diversity differences between Pre-COVID and Post-COVID periods. Wilcoxon rank-sum tests identified the top 10 differentially abundant bacterial taxa and KEGG pathways (L1 and L2 levels) between the two periods. Functional gene pathways were classified based on methodologies established by Sun et al. and Zheng et al., and Fisher’s exact test was used to evaluate pathway-specific functional potentials (p < 0.05)97,98. Ordinary least squares (OLS) regression analyzed relationships between sediment C/N ratios and core species relative abundance, community stability, and metabolic activity. Redundancy analysis (RDA; vegan package) identified environmental and spatial variables significantly correlated with bacterial composition. Variation partitioning analysis (VPA; vegan package) quantified the relative contributions of physical and chemical factors to community variation. Mantel tests (LinkET package) evaluated correlations between bacterial composition, richness, diversity, core species abundance, and environmental variables. Random Forest modeling (via the RandomForest package), a machine learning technique, determined the relative importance of individual factors driving microbial succession99. Partial least squares path modeling (PLS-PM; plspm package) assessed interactions among physical factors, chemical factors, microbial composition/diversity, and community stability. Model validity was assessed using the coefficient of determination (R²) and goodness-of-fit (GOF): GOF > 0.4 indicated acceptable model structure, while GOF > 0.7 denoted excellent fit100.

The concept of the “core microbiome” has been widely explored in microbial ecology across diverse systems, including human microbiomes (oral, gut), environmental niches (soil, seawater), and industrial processes (wastewater treatment, fermentation)101. A core microbiome is comprised of the members common to two or more microbial assemblages associated with a habitat102,103,104. Despite its increasing use, there remains no consensus on practical methodologies for defining core microbiomes101,102. Core species identification is critical for predicting global change impacts on biogeochemical cycles105. While definitions of the core microbiome remain debated, it is widely acknowledged that ‘shared taxa are critical’106. Accordingly, we employed the most prevalent identification approach, defining core taxa as microbial lineages shared across two or more samples from a specific host or environment101,102. Here, shared bacterial taxa across samples were identified using Venn diagrams (VennDiagram package) and defined as core species107.

Null modeling was used to uncover community assembly processes, following the approach established by Stegen et al.108. The beta nearest taxon index (betaNTI) and Raup-Crick metric (RC) were utilized to identify the underlying processes governing bacteria community assembly. Values of betaNTI <−2 indicated homogeneous selection, while values >2 indicated heterogeneous selection. For cases where |betaNTI| ≤ 2, the RC metric was applied to classify the remaining processes, with RC < − 0.95 denoting homogenous dispersal, RC > 0.95 indicating dispersal limitation, and |RC| ≤ 0.95 signifying ecological drift.

Co-occurrence networks for Pre-COVID and Post-COVID bacterial communities were constructed using Spearman correlations (psych package). Pairwise correlations with r > 0.6 and FDR-adjusted p < 0.05 were retained. ASVs with mean relative abundance <0.01% or detection frequency <60% were excluded to reduce false positives109. Networks were visualized and analyzed in Gephi (v0.10.1) to compute topological features (degree, betweenness, closeness, eigenvector centrality)110.

Keystone species were identified using within-module connectivity (Zi) and among-module connectivity (Pi) thresholds111: Module hubs: Zi > 2.5, Pi < 0.62 (high intra-module connectivity); Connectors: Zi < 2.5, Pi > 0.62 (high inter-module connectivity); Network hubs: Zi > 2.5, Pi > 0.62 (global centrality); Peripherals: Zi < 2.5, Pi < 0.62 (low connectivity); Non-peripheral nodes (Module hubs, Connectors, Network hubs) were classified as keystone taxa due to their metabolic significance69,112.

To validate the results of network analysis, we calculated network cohesion, a metric quantifying species interaction strength based on pairwise correlations between taxa, adjusted via null models and weighted by taxon abundances34. Specifically: Pairwise correlation calculation: A taxon relative abundance × sample matrix was used to compute all pairwise correlations (e.g., Spearman) between taxa. Null model adjustment: Correlations were statistically validated by subtracting random expectations generated through iterative null model permutations (e.g., shuffling taxon abundances across samples). Cohesion computation: For each sample, positive cohesion (sum of significantly positive correlations) and negative cohesion (sum of significantly negative correlations) were calculated using abundance-weighted sums of null model-adjusted correlations. The formulas are defined as follows (where n = total taxa in the network):

$${\text{Cohesion}}=\mathop{\sum }\limits_{i=1}^{n}{{\text{abundance}}}_{i}\times {{\text{connectedness}}}_{i}$$
(1)

Microbial community stability was quantified using the average variation degree (AVD)69. Lower AVD indicates higher stability (stability = 1 − AVD). The variation degree (\(\left|{a}_{i}\right|\)) of each species is calculated using the following formula, where (\({X}_{i}\): Abundance of the species in a single sample; \({\bar{X}}_{i}\): Mean abundance of the species across a sample group; δi: Standard deviation of the species’ abundance within the sample group):

$$|{a}_{i}|=\frac{|{X}_{i}-{\bar{X}}_{i}|}{{\delta }_{i}}$$
(2)

The AVD value is calculated using the following formula (where k = number of samples in a group, n = number of species):

$$AVD=\frac{{\sum }_{i=1}^{n}\frac{|{X}_{i}-{\bar{X}}_{i}|}{{\delta }_{i}}}{k\times n}$$
(3)