Abstract

Keystone taxa represent low-abundance microorganisms that exert disproportionate influence on ecosystem function and community stability, but their identification remains challenging due to limitations in taxonomic resolution and genome completeness using traditional short-read sequencing approaches. This study employed long-read metagenomics using Oxford Nanopore Technologies (ONT) and Pacific Biosciences (PacBio) platforms to achieve high-resolution taxonomic classification and functional annotation for keystone taxa identification across diverse soil ecosystems. We analyzed 187 soil samples from six biomes including temperate forests, grasslands, agricultural systems, wetlands, arid regions, and tundra environments, generating 2.8 Tb of long-read sequencing data with average read lengths of 8.2 kb (ONT) and 12.7 kb (PacBio). Advanced binning algorithms and hybrid assembly approaches enabled reconstruction of 1,247 high-quality metagenome-assembled genomes (MAGs) with >90% completeness and <5% contamination. Network analysis identified 89 putative keystone taxa characterized by high centrality measures (betweenness centrality >0.15) and low relative abundance (<1% of total community). Long-read sequencing improved species-level taxonomic resolution by 340% compared to short-read approaches, enabling precise identification of closely related taxa with distinct ecological roles. Functional annotation revealed that keystone taxa were significantly enriched in genes related to stress response (2.8-fold enrichment), secondary metabolite production (3.4-fold enrichment), and inter-species signaling (4.1-fold enrichment). Temporal analysis across 24 months demonstrated that keystone taxa maintained stable network positions despite seasonal fluctuations in overall community composition (average stability coefficient 0.73). Experimental validation through selective removal and addition experiments confirmed keystone effects, with targeted taxa removal causing 23-47% reduction in community stability metrics and significant alterations in nutrient cycling processes. The study identified previously unknown keystone species including Candidatus Solibacter variabilis, Bacillus keyensis sp. nov., and several uncultured members of the Verrucomicrobia and Planctomycetes phyla. Comparative analysis revealed that keystone taxa networks were conserved across similar biomes but differed significantly between ecosystem types, suggesting environment-specific selection for keystone functions. The findings demonstrate that long-read metagenomics provides unprecedented resolution for keystone taxa identification, enabling deeper understanding of microbial community dynamics and supporting targeted interventions for ecosystem management and restoration.