Analysis of genetic relatedness of Haemophilus influenzae isolates by multilocus sequence typing

Abstract

The gram-negative bacterium Haemophilus influenzae is a human-restricted commensal of the nasopharynx that can also be associated with disease. The majority of H. influenzae respiratory isolates lack the genes for capsule production and are nontypeable (NTHI). Whereas encapsulated strains are known to belong to serotype-specific phylogenetic groups, the structure of the NTHI population has not been previously described. A total of 656 H. influenzae strains, including 322 NTHI strains, have been typed by multilocus sequence typing and found to have 359 sequence types (ST). We performed maximum-parsimony analysis of the 359 sequences and calculated the majority-rule consensus of 4,545 resulting equally most parsimonious trees. Eleven clades were identified, consisting of six or more ST on a branch that was present in 100% of trees. Two additional clades were defined by branches present in 91% and 82% of trees, respectively. Of these 13 clades, 8 consisted predominantly of NTHI strains, three were serotype specific, and 2 contained distinct NTHI-specific and serotype-specific clusters of strains. Sixty percent of NTHI strains have ST within one of the 13 clades, and eBURST analysis identified an additional phylogenetic group that contained 20% of NTHI strains. There was concordant clustering of certain metabolic reactions and putative virulence loci but not of disease source or geographic origin. We conclude that well-defined phylogenetic groups of NTHI strains exist and that these groups differ in genetic content. These observations will provide a framework for further study of the effect of genetic diversity on the interaction of NTHI with the host.

FIG. 1.

Maximum-parsimony majority-rule consensus tree for 359 ST, plotted with ST65 as the outgroup. Colored regions of the tree numbered 1 through 13 indicate branches (clades) that were found in 100% of trees, with the exception of clades 9 and 11, which were found in 92% and 81% of trees, respectively. The tree is aligned with a plot showing the serotype and the eBURST group for each ST. Each ST corresponded either to a single serotype or to one or more NTHI strains. H. influenzae biogroup aegyptius (H. ae.) strains are plotted separately from other NTHI strains. The ST for 15 sequenced strains are indicated on the tree. Table S1 in the supplemental material lists the position of each ST on the tree as well as its eBURST group and the serotype of associated strains.

FIG. 2.

Characterization of NTHI strains and correlation of biotype and hmw locus with clustering on the majority-rule consensus tree. The tree differs from that in Fig. 1 in that each ST found in more than one NTHI strain was expanded to allow plotting of data for each strain, while serotype-specific clusters were collapsed (indicated by triangles). The tree is aligned with a plot showing clade (as in Fig. 1), biotype, the presence of the hmw loci (determined by PCR), and the type of specimen from which the strain was cultured. Commensal, strains cultured from oropharyngeal or nasopharyngeal swabs from healthy children; otitis, strains from middle ear aspirates from children with otitis media; eye, strains from children with conjunctivitis; mucosal, strains from other mucosal samples, primarily sputum or endotracheal aspirates; invasive, strains from blood or cerebrospinal fluid. The biotype and PCR data, including the results of PCR of the LPS biosynthetic locus flanked by infA and ksgA, are summarized in Table 1.

FIG. 3.

Comparison between different analyses of the complete MLST database. The dendrogram was generated from MLST sequences by the neighbor-joining method, using software in the START2 suite of MLST analytic tools. In alignment with each ST on the tree is plotted the serotype of associated strains, and the assignment of the ST to clade and eBURST group as previously described is indicated. Table S1 in the supplemental material lists the position of each ST on this tree and on three other trees generated using other programs in START2.

FIG. 4.

Comparison between 16S typing and MLST. The dendrogram is a majority-rule tree derived from maximum-parsimony analysis of 84 16S sequences. Each branch is labeled with a red numeral representing the percentage of trees contributing to the majority-rule consensus tree for that branch. Branches representing encapsulated strains are shaded in color and labeled by serotype. For each branch, the MLST phylogenetic group (clade or eBURST group 2) for the same strains is indicated by a colored arc. Strains with 16S ST 3, 4, 5, 31, and 80 (indicated by red circles) each belonged to more than one MLST phylogenetic group.