Resolution of a meningococcal disease outbreak from whole-genome sequence data with rapid Web-based analysis methods

Abstract

The increase in the capacity and reduction in cost of whole-genome sequencing methods present the imminent prospect of such data being used routinely in real time for investigations of bacterial disease outbreaks. For this to be realized, however, it is necessary that generic, portable, and robust analysis frameworks be available, which can be readily interpreted and used in real time by microbiologists, clinicians, and public health epidemiologists. We have achieved this with a set of analysis tools integrated into the PubMLST.org website, which can in principle be used for the analysis of any pathogen. The approach is demonstrated with genomic data from isolates obtained during a well-characterized meningococcal disease outbreak at the University of Southampton, United Kingdom, that occurred in 1997. Whole-genome sequence data were collected, de novo assembled, and deposited into the PubMLST Neisseria BIGSdb database, which automatically annotated the sequences. This enabled the immediate and backwards-compatible classification of the isolates with a number of schemes, including the following: conventional, extended, and ribosomal multilocus sequence typing (MLST, eMLST, and rMLST); antigen gene sequence typing (AGST); analysis based on genes conferring antibiotic susceptibility. The isolates were also compared to a reference isolate belonging to the same clonal complex (ST-11) at 1,975 loci. Visualization of the data with the NeighborNet algorithm, implemented in SplitsTree 4 within the PubMLST website, permitted complete resolution of the outbreak and related isolates, demonstrating that multiple closely related but distinct strains were simultaneously present in asymptomatic carriage and disease, with two causing disease and one responsible for the outbreak itself.

Fig 1

Timeline of the outbreak.

Fig 2

NeighborNet graphs generated using the Genome Comparator tool of BIGSDB with seven-locus MLST data (A), 20-locus extended MLST data (B), rMLST data (C), or the 944 variable loci (D) identified when the FAM18 annotation was compared against all isolates.

Fig 3

Location of nucleotide changes among isolates corresponding to each of strains 1, 2, and 5, mapped onto the FAM18 genome. The red circles represent the loci that varied within isolates designated strain 1 (8 loci varied, none of which were single base changes); the blue circles represent loci that varied within strain 2 (27 loci varied, 10 of which were single base changes); the green circles represent the loci that varied within strain 5 (17 loci varied, 7 of which were single base changes). Full details of the locus changes can be found in Tables S1 to S3 of the supplemental material. The figure was generated using CGView (44).

Fig 4

The BIGSdb process workflow. Following sequencing, rapid analysis for strain identification and genome comparison can be performed in minutes.