Ecology of aerobic anoxygenic phototrophs on a fine-scale taxonomic resolution in Adriatic Sea unravelled by unsupervised neural network

Abstract

Abstract Background Aerobic anoxygenic phototrophs are metabolically highly active, diverse and widespread polyphyletic members of bacterioplankton whose photoheterotrophic capabilities shifted the paradigm about simplicity of the microbial food chain. Despite their considerable contribution to the transformation of organic matter in marine environments, relatively little is still known about their community structure and ecology at fine-scale taxonomic resolution. Up to date, there is no comprehensive (i.e. qualitative and quantitative) analysis of their community composition in the Adriatic Sea. Results Analysis was based on pufM gene metabarcoding and quantitative FISH-IR approach with the use of artificial neural network. Significant seasonality was observed with regards to absolute abundances (maximum average abundances in spring 2.136 ± 0.081 × 104 cells mL−1, minimum in summer 0.86 × 104 cells mL−1), FISH-IR groups (Roseobacter clade prevalent in autumn, other Alpha- and Gammaproteobacteria in summer) and pufM sequencing data agglomerated at genus-level. FISH-IR results revealed heterogeneity with the highest average relative contribution of AAPs assigned to Roseobacter clade (37.66%), followed by Gammaproteobacteria (35.25%) and general Alphaproteobacteria (31.15%). Community composition obtained via pufM sequencing was dominated by Gammaproteobacteria clade NOR5/OM60, specifically genus Luminiphilus, with numerous rare genera present in relative abundances below 1%. The use of artificial neural network connected this community to biotic (heterotrophic bacteria, HNA and LNA bacteria, Synechococcus, Prochlorococcus, picoeukaryotes, heterotrophic nanoflagellates, bacterial production) and abiotic environmental factors (temperature, salinity, chlorophyll a and nitrate, nitrite, ammonia, total nitrogen, silicate, and orthophosphate concentration). A type of neural network, neural gas analysis at order-, genus- and ASV-level, resulted in five distinct best matching units (representing particular environments) and revealed that high diversity was generally independent of temperature, salinity, and trophic status of the environment, indicating a potentially dissimilar behaviour of aerobic anoxygenic phototrophs compared to the general bacterioplankton. Conclusion This research represents the first comprehensive analysis of aerobic anoxygenic phototrophs in the Adriatic Sea on a trophic gradient during a year-round period. This study is also one of the first reports of their genus-level ecology linked to biotic and abiotic environmental factors revealed by unsupervised neural network algorithm, paving the way for further research of substantial contribution of this important bacterial functional group to marine ecosystems.