Combining morphological and metabarcoding approaches reveals the freshwater eukaryotic phytoplankton community

Background Phytoplankton diversity can be difficult to ascertain from morphological analyses, because of the existence of cryptic species and pico- and concealed phytoplankton. In-depth sequencing and metabarcoding can reveal microbial diversity, and identify novel diversity. However, there has been little comparison of metabarcoding and morphological datasets derived from the same samples, and metabarcoding studies covering total eukaryotic phytoplankton diversity are rare. In this study, the variable V7 region of the 18S rDNA gene was employed to explore eukaryotic phytoplankton diversity in 11 Chinese freshwater environments, and further compared with the dataset obtained through morphological identification. Results Annotation by the evolutionary placement algorithm (EPA) rather than alignment with the SILVA database improved the taxonomic resolution, with 346 of 524 phytoplankton operational taxonomic units (OTUs) being assigned to the genus or species level. The number of unassigned OTUs was greatly reduced from 259 to 178 OTUs by using the EPA in place of the SILVA database. Metabarcoding detected 3.5 times more OTUs than the number of morphospecies revealed by morphological identification; furthermore, the number of species and the Shannon-Wiener index inferred from the two methods were correlated. A total of 34 genera were identified via both methods, while 31 and 123 genera were detected solely in the morphological or metabarcoding dataset, respectively. Conclusion The dbRDA plot showed distinct separation of the phytoplankton communities between lakes and reservoirs according to the metabarcoding dataset. The same pattern was obtained on the basis of 10 environmental variables in the PCO ordination plot, while the separation of the populations based on morphological data was poor. However, 30 morphospecies contributed 70% of the community difference between lakes and reservoirs in the morphological dataset, while 11 morphospecies were not found by metabarcoding. Considering the limitations of each of the two methods, their combination could substantially improve phytoplankton community assessment.