Anthropogenic nitrogen pollution poses a systemic threat to microbial interaction networks and biogeochemical cycling in groundwater ecosystems, yet the underlying mechanisms remain poorly understood. Employing an endpoint gradient comparison, we conducted metagenomic analyses of urban groundwater under severe nitrogen stress (Shanghai, China; with NH₄⁺ and NO₃⁻ concentrations ∼28× and ∼10× background levels, respectively) versus a near-pristine mountain aquifer (Calistoga, USA). This revealed a multi-level collapse and adaptive restructuring of microbial communities under nitrogen stress. Pollution triggered a fundamental restructuring of bacterial communities, with system type (urban vs. mountain) explaining 74 % of the compositional variation, accompanied by a significant reduction in bacterial alpha-diversity (Shannon index decreased by 34 %) and a taxonomic shift from Actinomycetota-dominated mutualistic networks in the mountain system to Pseudomonadota-dominated communities (> 0.86 relative abundance) in urban groundwater. Functionally, urban systems exhibited multi-pathway suppression of energy-intensive processes, including nitrification (e.g., hao, nxrB genes), methanogenesis, and inorganic sulfur oxidation, aligning with the theory of "pollution-induced metabolic decoupling." To survive, the microbial community pivoted to low-energy strategies, significantly enriching genes for organic sulfur metabolism (e.g., dddT, tsdB), which may exacerbate nitrogen retention by inhibiting denitrifiers via metabolites like H₂S. Co-occurrence network topology analysis indicated a catastrophic loss of complexity in urban groundwater, with a ∼90 % reduction in connectivity and a collapse in modularity (from 19.94 to 3.33), alongside an abnormally high proportion of positive correlations (94.4 %), signaling a major loss of ecosystem stability and functional redundancy. Random Forest and redundancy analyses jointly identified ammonium (NH₄⁺) as the core environmental driver of this cascading failure, explaining 86 % of the variance in functional gene profiles and likely disrupting the nitrification pathway through specific suppression of the rate-limiting hao gene (which explained 76 % of the variance in nitrification rates). Based on these insights, we propose a dual-track restoration framework that couples external NH₄⁺ source control with internal microbial network rewiring (e.g., restoring keystone taxa, regulating sulfur feedback loops) to break the nitrogen-sulfur inhibition cycle and restore ecological function. Our findings underscore the critical importance of integrating microbial network resilience into strategies for managing and rehabilitating contaminated groundwater ecosystems.