Microbial ecological networks reveal community reorganization along a sediment biotoxicity gradient

Sediment quality assessment in shallow marsh wetlands often relies on diversity or chemical metrics, which may underestimate ecological impairment under pollution stress. In this study, We analyzed 26 sediment samples collected from the Baiyangdian marsh along a biotoxicity gradient using 16S rRNA gene sequencing, PICRUSt2-based functional inference, and microbial ecological network analysis. Luminescence inhibition ranged from 0.4% to 92.0%, with a mean of 23.0%, indicating an overall moderate toxicity level. Bacterial alpha-diversity remained relatively stable across toxicity groups, with Shannon indices ranging from 2.4 to 2.7. In contrast, microbial networks showed marked reorganization. Under moderate toxicity (50-80%), network complexity peaked, with 1656 nodes, 22030 edges, and a mean degree of 26.6. Under high toxicity (80-100%), average path length and characteristic path length declined to 1.720 and 1.696, while modularity dropped to 0.874, indicating reduced niche differentiation and greater structural centralization. Keystone taxa such as Phenylobacterium, Steroidobacter, and Fluviicola were enriched under high toxicity. Total nitrogen and total carbon were the major drivers of community and network variation. These findings suggest that microbial ecological networks can complement conventional toxicity classification for sediment quality assessment and ecological risk diagnosis.