RESUMEN
Despite the importance of bacteria in aquatic ecosystems and their predictable diversity patterns across space and time, biomonitoring tools for status assessment relying on these organisms are widely lacking. This is partly due to insufficient data and models to identify reliable microbial predictors. Here, we show metabarcoding in combination with multivariate statistics and machine learning allows to identify bacterial bioindicators for existing biological status classification systems. Bacterial beta-diversity dynamics follow environmental gradients and the observed associations highlight potential bioindicators for ecological outcomes. Spatio-temporal links spanning the microbial communities along the river allow accurate prediction of downstream biological status from upstream information. Network analysis on amplicon sequence veariants identify as good indicators genera Fluviicola, Acinetobacter, Flavobacterium, and Rhodoluna, and reveal informational redundancy among taxa, which coincides with taxonomic relatedness. The redundancy among bacterial bioindicators reveals mutually exclusive taxa, which allow accurate biological status modeling using as few as 2-3 amplicon sequence variants. As such our models show that using a few bacterial amplicon sequence variants from globally distributed genera allows for biological status assessment along river systems.
Asunto(s)
Biomarcadores Ambientales , Microbiota , Flavobacterium , Aprendizaje Automático , Microbiota/genética , RíosRESUMEN
Microbial communities are major players in the biogeochemical processes and ecosystem functioning of river networks. Despite their importance in the ecosystem, biomonitoring tools relying on prokaryotes are still lacking. Only a few studies have employed both metabarcoding and quantitative techniques such as catalysed reported deposition fluorescence in situ hybridization (CARD-FISH) to analyse prokaryotic communities of epilithic biofilms in river ecosystems. We intended to investigate the efficacy of both techniques in detecting changes in microbial community structure associated with environmental drivers. We report a significant correlation between the prokaryotic community composition and pH in rivers from two different geographical areas in Norway. Both CARD-FISH and metabarcoding data were following the pattern of the environmental variables, but the main feature distinguishing the community composition was the regional difference itself. Beta-dispersion analyses on both CARD-FISH abundance and metabarcoding data revealed higher accuracy of metabarcoding to differentiate regions and river systems. The CARD-FISH results showed high variability, even for samples within the same river, probably due to some unmeasured microscale ecological variability which we could not resolve. We also present a statistical method, which uses variation coefficient and overall prevalence of taxonomic groups, to detect possible biological indicators among prokaryotes using metabarcoding data. The development of new prokaryotic bioindicators would benefit from both techniques used in this study, but metabarcoding seems to be faster and more reliable than CARD-FISH for large scale bio-assessment.