Contrasting bacterial and eukaryotic microbial communities across a stratified water column in the northern South China Sea.

Bacterial and eukaryotic microbial communities may reflect water-column structure differently across spatial and taxonomic scales. We analyzed 74 matched samples from eight stations spanning 5-800 m in a stratified anticyclonic-eddy system of the northern South China Sea using 16S and 18S rRNA gene amplicon sequencing, ordination and leave-one-station-out (LOSO) random forests. Bacterial OTU composition was dominated by depth-layer structure (marginal PERMANOVA R2 = 0.625), whereas station identity explained little additional variation (R2 = 0.031). For eukaryotic microbes, station-associated heterogeneity was much stronger (R2 = 0.474) than the depth-layer effect (R2 = 0.034). The eukaryotic station-associated pattern was robust and persisted below the upper water column. LOSO classification achieved 1.000 accuracy for bacterial OTU and genus profiles, compared with 0.595 for eukaryotic OTUs and 0.878 after genus-level aggregation. Across stations, bacterial composition tracked the shared vertical environmental gradient more consistently than eukaryotic composition, whose performance improved after genus-level aggregation. Thus, the two microbial domains captured different aspects of variation within the same stratified water column.