Ribosomal MLST nucleotide identity (rMLST-NI), a rapid bacterial species identification method: application to Klebsiella and Raoultella genomic species validation

Abstract

Bacterial genomics is making an increasing contribution to the fields of medicine and public health microbiology. Consequently, accurate species identification of bacterial genomes is an important task, particularly as the number of genomes stored in online databases increases rapidly and new species are frequently discovered. Existing database entries require regular re-evaluation to ensure that species annotations are consistent with the latest species definitions. We have developed an automated method for bacterial species identification that is an extension of ribosomal multilocus sequence typing (rMLST). The method calculates an 'rMLST nucleotide identity' (rMLST-NI) based on the nucleotides present in the protein-encoding ribosomal genes derived from bacterial genomes. rMLST-NI was used to validate the species annotations of 11839 publicly available Klebsiella and Raoultella genomes based on a comparison with a library of type strain genomes. rMLST-NI was compared with two whole-genome average nucleotide identity methods (OrthoANIu and FastANI) and the k-mer based Kleborate software. The results of the four methods agreed across a dataset of 11839 bacterial genomes and identified a small number of entries (n=89) with species annotations that required updating. The rMLST-NI method was 3.5 times faster than Kleborate, 4.5 times faster than FastANI and 1600 times faster than OrthoANIu. rMLST-NI represents a fast and generic method for species identification using type strains as a reference.

Keywords: Klebsiella; Raoultella; bacterial genomes; bacterial species identification; nucleotide identity; rMLST.

Fig. 1.

Flowchart describing the stages required to scan a query genome against a library of rMLST alleles and calculate the rMLST nucleotide identity (rMLST-NI) and profile coverage values. These stages are implemented in the MLSS.pl program (Multilocus Sequence Search).

Fig. 2.

The rMLST nucleotide identity (rMLST-NI) for a query genome against each profile is calculated by extracting blast match information from each genome/allelic pairwise comparison (number of identical bases matched and the number of aligned bases). The rMLST-NI is defined as the total number of identical bases across all alleles in the profile divided by the sum of all aligned profile allele bases (expressed as a percentage). The profile coverage is similarly defined using the total number of aligned bases across all allelic matches. [··] indicates profile alleles and match information that have been removed. rMLST profiles typically contain allele indices for 53 loci. Allele indices and lengths shown are for illustrative purposes and do not represent the actual values.

Fig. 3.

Summary of 89 inconsistent species annotations where the NCBI Assembly database has a different species than the consensus result obtained from the four automated approaches used in this study (OrthoANIu, FastANI, rMLST-NI and Kleborate). The left-hand side represents the original species annotations from NCBI Assembly and the right-hand side represents the predicted species.

Fig. 4.

Phylogenetic tree of 30 Klebsiella/Raoultella isolates including 17 type strains and 13 non-type strain isolates where the NCBI Assembly species annotation was not consistent with the automated species identification scan results. Each isolate is annotated with (1) database identifier, (2) isolate name, (3) species (T indicates type strain), (4) species in source database (shown in square brackets if different). This tree was constructed in mega X (www.megasoftware.net) using the maximum-likelihood method and Tamura–Nei model (1000 replicates) from a multiple sequence alignment of the concatenated alleles of 51 non-paralogous rMLST loci generated by BIGSdb.

Fig. 5.

rMLST nucleotide identity webserver results for scanning UCI 27, (NCBI Assembly entry GCA_000534255.1, Klebsiella aerogenes ). The top five species are displayed in the results table and are colour coded according to the traffic light system of thresholds observed from the analysis of scan results from the 11750 dataset.