Comparative Analysis of the Two Acinetobacter baumannii Multilocus Sequence Typing (MLST) Schemes

Abstract

Acinetobacter species assigned to the Acinetobacter calcoaceticus-baumannii (Acb) complex, are Gram-negative bacteria responsible for a large number of human infections. The population structure of Acb has been studied using two 7-gene MLST schemes, introduced by Bartual and coworkers (Oxford scheme) and by Diancourt and coworkers (Pasteur scheme). The schemes have three genes in common but underlie two coexisting nomenclatures of sequence types and clonal complexes, which complicates communication on A. baumannii genotypes. The aim of this study was to compare the characteristics of the two schemes to make a recommendation about their usage. Using genome sequences of 730 strains of the Acb complex, we evaluated the phylogenetic congruence of MLST schemes, the correspondence between sequence types, their discriminative power and genotyping reliability from genomic sequences. In silico ST re-assignments highlighted the presence of a second copy of the Oxford gdhB locus, present in 553/730 genomes that has led to the creation of artefactual profiles and STs. The reliability of the two MLST schemes was tested statistically comparing MLST-based phylogenies to two reference phylogenies (core-genome genes and genome-wide SNPs) using topology-based and likelihood-based tests. Additionally, each MLST gene fragment was evaluated by correlating the pairwise nucleotide distances between each pair of genomes calculated on the core-genome and on each single gene fragment. The Pasteur scheme appears to be less discriminant among closely related isolates, but less affected by homologous recombination and more appropriate for precise strain classification in clonal groups, which within this scheme are more often correctly monophyletic. Statistical tests evaluate the tree deriving from the Oxford scheme as more similar to the reference genome trees. Our results, together with previous work, indicate that the Oxford scheme has important issues: gdhB paralogy, recombination, primers sequences, position of the genes on the genome. While there is no complete agreement in all analyses, when considered as a whole the above results indicate that the Pasteur scheme is more appropriate for population biology and epidemiological studies of A. baumannii and related species and we propose that it should be the scheme of choice during the transition toward, and in parallel with, core genome MLST.

Keywords: Acinetobacter baumannii; clonal complexes; comparative genomics; multilocus sequence typing; phylogeny; sequence types.

FIGURE 1

(A) Distribution of the MLST loci on the genome. A panel is used to highlight the relative position of two loci whenever they come from the same gene. (B) Scatterplot comparing loci length and variability in terms of number of alleles registered on the Pubmlst database (https://pubmlst.org/abaumannii).

FIGURE 2

Maximum Likelihood phylogeny of the 104 gdhB variants detected in the dataset. Alleles of the traditional gdhB locus are highlighted in red; alleles of the alternative locus (including 38 non-registered alleles) are highlighted in blue.

FIGURE 3

Minimum spanning trees of A. baumannii, A. nosocomialis, A. pittii, A. seifertii, and A. dijkshoorniae (23 isolates) using (A) Pasteur and (B) Oxford MLST scheme. The colors corresponding to Acinetobacter species are shown in the legend. Minimum spanning trees representing the structure of the A. baumannii international clone II (422 isolates) as reconstructed using Pasteur (C) and Oxford (D) MLST schemes. Numbers inside each circle indicate the ST types. Circle size is proportional to the number of isolates belonging to the same ST type. Colors in (C,D) represent sub-branches identified by eBURST using the Oxford MLST scheme.

FIGURE 4

Sankey diagram of the MLST classification of the 730 genomes in use, as performed with the Pasteur and Oxford schemes. Two-way corresponding STs were removed to improve image clarity. Captions show the corresponding STs belonging to (A) International Clone 1, (B) International Clone 2, and (C) all the other genomes.

FIGURE 5

Maximum Likelihood phylogenies of 730 genomes of the Acb complex, inferred from (A) a concatenate of 1409 orthologous core genes, (B) a concatenate of 68,340 SNP positions, (C) a concatenate of the seven alleles used in the Pasteur MLST scheme, and (D) a concatenate of the seven alleles used in the Oxford MLST scheme. Major clonal complexes are highlighted: International clone I in blue, International clone II in red, International clone III in green.

FIGURE 6

Heatmap showing the levels of agreement of each MLST locus to a reference alignment based on the core genome. Agreement levels are shown in terms of regression line slopes obtained from distance values between genome pairs. For each locus four regression were obtained, based, respectively, on the entirety of genome-wide distances and three subranges. White color indicates that a MLST locus represents the genome-wide distances well, showing a similar evolutionary pace. Red shades (indicating a negative correlation) and blue shades (indicating that the MLST locus shows an evolutionary pace much higher than the genome wide distances) indicate that the locus variation does not well represent genome variation.