Scale-Dependent Drivers Shape Airborne Microbiomes across the Shanghai Subway Network

Abstract Urban subway systems represent important yet understudied environments for human exposure to airborne microorganisms. Therefore, we conducted an investigation of air microbiome across 43 underground stations spanning 13 lines of the Shanghai subway. Using stratified cluster sampling, we simultaneously collected total suspended particles (TSP, aerodynamic diameter ≤ 100 μm) and fine particulate matter (PM2.5, aerodynamic diameter ≤ 2.5 μm) at station entrances, halls, and platforms, integrating microbial taxonomy, transcriptional potential, PM2.5 chemical composition, microclimate, and network topology. We found bacteria dominated both particle-size fractions and accounted for the majority of transcriptional signals, whereas eukaryotic taxa were strongly enriched in TSP, reflecting pronounced particle-size filtering. Although transcriptional potential generally increased with abundance, dominant taxa such as Staphylococcus and Bacillus exhibited weak abundance–transcription coupling, indicating functional decoupling between numerical dominance and transcriptional contribution in subway air microbiomes. Microbial diversity and community composition remained broadly homogeneous across the subway network, consistent with extensive passenger mobility and high network connectivity. In contrast, finer-scale heterogeneity emerged within stations, particularly for eukaryotic communities, which declined from entrances to enclosed platform environments. Environmental associated factors exhibited kingdom-specific patterns: bacterial and viral diversity were primarily associated with local microclimatic conditions, especially temperature, whereas eukaryotic diversity was more strongly linked to urbanization and externally derived aerosol inputs. Overall, these findings demonstrate that subway airborne microbiomes are shaped by interacting effects of particle-size filtering, environmental selection, and network-driven microbial mixing.