Gene content dissimilarity for subclassification of highly similar microbial strains

Abstract

Background: Identification and classification of highly similar microbial strains is a challenging issue in microbiology, ecology and evolutionary biology. Among various available approaches, gene content analysis is also at the core of microbial taxonomy. However, no threshold has been determined for grouping microorgnisms to different taxonomic levels, and it is still not clear that to what extent genomic fluidity should occur to form a microbial taxonomic group.

Results: By taking advantage of the eggNOG database for orthologous groups, we calculated gene content dissimilarity among different microbial strains based on the orthologous gene profiles and tested the possibility of applying gene content dissimilarity as a quantitative index in classifying microbial taxonomic groups, as well as its potential application in subclassification of highly similar microbial strains. Evaluation of gene content dissimilarity to completed microbial genomes at different taxonomic levels suggested that cutoffs of 0.2 and 0.4 can be respectively used for species and family delineation, and that 0.2 gene content dissimilarity cutoff approximately corresponded to 98 % 16S rRNA gene identity and 94 % ANI for microbial species delineation. Furthermore, application of gene content dissimilarity to highly similar microbial strains suggested it as an effective approach in classifying closely related microorganisms into subgroups.

Conclusions: This approach is especially useful in identifying pathogens from commensals in clinical microbiology. It also provides novel insights into how genomic fluidity is linked with microbial taxonomy.

Keywords: Gene content dissimilarity; Genomic fluidity; Highly similar strains; Microbial subclassification.

Fig. 1

The flowchart of applying gene content dissimilarity for microbial delineation and classification. Three main steps were included. First, orthologous gene profiles were obtained for all selected microbial genomes by searching against the eggNOG database. Second, pairwise gene content dissimilarity as measured by Bray-Curtis dissimilarity was calculated for all pairs of microbial strains. Third, microbial strains were clustered into different groups

Fig. 2

Distribution of gene content dissimilarity for the retrieved microbial genomes at different taxonomic levels, including species (a), genus (b), family (c), and order (d). Cutoffs of 0.2 and 0.4 were recommended for microbial species and family delineation, respectively

Fig. 3

Comparison of 16S rRNA gene identity, ANI, and gene content dissimilarity in microbial species delineation. A cutoff of 0.2 corresponded to 98 % 16S rRNA gene identity and 94 % ANI in species delineation. A total of 5008 intra-species and 8642 intra-genus comparisons were plotted. Red dots falling in the Q1 quadrant were mostly several clostridium strains, for which misclassification may have occurred. Red dots represented intra-species comparisons, and blue dots indicated intra-genus comparisons

Fig. 4

Application of gene content dissimilarity in classifying microbial strains belonging to Enterobacteriaceae. a PCoA clustering of all selected microbial strains belonging to Enterobacteriaceae. b PCoA clustering of highly similar microbial strains including E. coli and Shigella. A clear separation of Shigella and E. coli O157:H7 from other E. coli strains could be observed

Fig. 5

Application of gene content dissimilarity in classifying Streptococcus strains. Clear separation of different species into different groups could be observed. Highly similar strains belonging to S. mitis, S. oralis, and S. pneumoniae were also well separated