Abstract

Industrial activities impose complex pollutant mixtures that reshape soil microbiomes through interacting stressors. Although previous studies have examined microbial responses to composite contamination, the regime-specific coupling between co-occurrence network stability and microbial assembly processes in coal-industrial soils remains insufficiently resolved. Here, we integrated 16S rRNA gene sequencing, microbial co-occurrence networks, and iCAMP assembly modeling to investigate microbiome responses across coal-industrial zones (mining/CM, washing/CW, combustion/CC, background/CK) in northwestern China. CW emerged as a composite pollution hotspot, with high levels of PAHs (6246.5 μg/kg), heavy metals (e.g., Pb), and human-fecal taxa ( Escherichia-Shigella , 9.2% relative abundance). CW soils exhibited the highest β-diversity, lowest network robustness, and strongest drift-driven assembly, indicating ecological fragmentation. Conversely, CC soils under heavy metal stress (Cd: 10.8 mg/kg) showed more deterministic assembly (homogeneous selection: 20.8%) and cohesive modules dominated by metal-tolerant taxa like Nocardioides. CM soils co-contaminated with PAHs and metals exhibited steep taxonomic turnover and drift-dominated assembly, likely driven by mining disturbance. Despite these pronounced changes, α-diversity did not differ significantly among the four sites. Across sites, composite stressors in CW weakened generalist-mediated cohesion, while persistent metals in CC promoted deterministic filtering and specialist stabilization. We propose tailoring restoration to site-specific assembly mechanisms and network vulnerabilities. These findings reveal how distinct ecological processes shape microbial network resilience under composite contamination and provide a mechanistic foundation for future microbiome-based soil recovery frameworks.