Assessing microbial internal drivers of nitrogen-cycle functional potential in lakeside wetlands: insights from random forest modelling and co-occurrence network analyses

Environmental factors are widely assumed to influence nitrogen (N) cycle processes in lakeside wetlands. However, recent studies suggested that shifts in environmental conditions do not consistently correspond to proportional changes in N-functional potential. To elucidate the contributions of different drivers, multiple influences, including environmental factors, microbial diversity, functional microbial genera, and gene abundance, were examined by integrating a random forest model (RFM), variation partitioning analysis (VPA), network co-occurrence analysis, and structural equation modeling (SEM). Almost 66 % of functional genes in RFM analysis exhibited a preferable R2 fit over 0.6. VPA analysis revealed that functional microbial genes (55.1 %) and genera (35.1 %) accounted for over 90 % of the explanatory contribution rate, surpassing environmental influences (4.3 %) and diversity indices (5.4 %). Network co-occurrence demonstrated that denitrification genes (norB/C/D/E; napA) formed central connective nodes in the network linked to nitrogen-cycle functional potential, and functional genera-gene linkages were closely aligned with variations in N-functional potential. SEM results further revealed that environmental factors did not exhibit a significant direct association with functional genes; instead, their effects were mediated indirectly through the restructuring of microbial community composition. Under long-term hydrological and nutrient fluctuations, microbial communities exhibited internal co-occurrence networks that were associated with N-cycle functional potential. Recognizing these internally structured microbial associations provides new insight into microbial contributions to N-cycle functional potential and may guide the design strategies of sustainable lakeside wetland restoration.