Metabarcoding reveals a more complex cyanobacterial community than morphological identification

Abstract

Abstract Next generation sequencing has become a popular method to identify cyanobacterial diversity in ecosystems. Morphological identification poses challenges in deciphering cyanobacterial diversity because phenotypic variations can be misleading, picocyanobacteria are ignored easily, specialists familiar with classification systems are lacking, and updates of taxonomic treatises are not shared in a timely manner or adopted by researchers engaged in monitoring of phytoplankton. In this study, metabarcoding of 54 samples from 7 freshwater systems yielded saturated cyanobacterial sequences, and the compositions of the resulting communities were compared with a dataset generated by morphological identification. A total of 69 cyanobacterial operational taxonomic units (OTUs) were finally codetermined through the ribosomal database project (RDP) via the Silva database and blast analysis using a database maintained by the National Center for Biotechnology Information (NCBI). Blast resolved more OTUs, with 1 and 27 OTUs identified at the species level and 35 and 9 OTUs identified at the genus levels through Silva and NCBI, respectively. Only 13 OTUs were matched at genus levels by these two methods. The accuracy of blast was further verified through molecular phylogeny. Blast results were compared with morphological identification at the genus level. Each method identified 26 genera, and 12 genera were shared by both methods. Half of the genera only identified by metabarcoding have been described recently, and data on morphologically diagnostic characters were not available. Some genera shared morphological similarities, which led them to be misidentified. Ten of the 12 genera found by both methods yielded less than 2 morphological mismatches. These genera produced a higher proportion of occurrences in metabarcoding, which was indicative of higher sensitivity to cyanobacteria. A larger number of species were found in metabarcoding (8–38 OTUs) than in morphological surveys (1–16 morphospecies), nevertheless, the results were positively correlated (R2 = 0.56, p