Real-time investigation of a Legionella pneumophila outbreak using whole genome sequencing

Summary

Legionella pneumophila is the main pathogen responsible for outbreaks of Legionnaires' disease, which can be related to contaminated water supplies such as cooling towers or water pipes. We combined conventional molecular methods and whole genome sequence (WGS) analysis to investigate an outbreak of L. pneumophila in a large Australian hospital. Typing of these isolates using sequence-based typing and virulence gene profiling, was unable to discriminate between outbreak and non-outbreak isolates. WGS analysis was performed on isolates during the outbreak, as well as on unlinked isolates from the Public Health Microbiology reference collection. The more powerful resolution provided by analysis of whole genome sequences allowed outbreak isolates to be distinguished from isolates that were temporally and spatially unassociated with the outbreak, demonstrating that this technology can be used in real-time to investigate L. pneumophila outbreaks.

Legionella pneumophila is a Gram-negative facultative intracellular pathogen, and is a causative agent of legionellosis, a form of pneumonia that is potentially fatal [1]. L. pneumophila exists naturally in aquatic environments but can be found as a contaminant in air-conditioning cooling towers and hot-water systems where it can be transmitted by aerosol to cause infection [1]. As a species, L. pneumophila shows a high degree of genomic plasticity [2]; however, the presence of a globally distributed clone, responsible for outbreaks and sporadic cases in several countries has been described [3]. A study of L. pneumophila serogroup 1 in Australia demonstrated the presence of a predominant AFLP genotype in clinical isolates [4] although it is not clear if this genotype corresponds to that of the globally distributed clone mentioned above. The predominance of a particular L. pneumophila genotype can make epidemiological typing in outbreak situations difficult, since not all typing methods have the necessary resolving power to discriminate outbreak isolates from non-outbreak isolates. Whole genome sequencing provides a high level of resolution and has been used successfully to characterize disease outbreaks and to elucidate transmission sources and links between patients [5]. A retrospective pilot study into the use of whole genome sequence (WGS) analysis for the investigation of a L. pneumophila outbreak suggested that this approach could be used in conjunction with other typing techniques to identify links between isolates and environmental sources [6]. However, to the best of our knowledge, until now there have been no reports of WGS technology being used to investigate an outbreak in real-time in order to provide timely advice to public health units and aid in epidemiological investigation.

In May 2013, an outbreak of legionellosis caused by L. pneumophila serogroup 1 occurred in a large Australian hospital. Two patients were positive for L. pneumophila by culture, and one was positive by urinary antigen test; however, L. pneumophila was not isolated from this patient. Initial epidemiological investigations strongly suggested that the hot-water supply of the hospital was linked with these cases. Sequence-based typing (SBT) and virulence gene profiling were initially utilized as typing methods as part of the outbreak response but these were not able to discriminate outbreak isolates from non-outbreak isolates, probably due to the clonal nature of L. pneumophila present in the region. Methods such as spoligotyping and variable genetic element typing have been shown to have the potential to discriminate between isolates belonging to the same SBT sequence type [7, 8]; however, these methods can be labour intensive and may still have limited resolution compared to that possible by comparison of the whole genome. Whole genome sequencing of bacteria is becoming increasingly rapid and inexpensive due to advances in next-generation sequencing technology. It is now possible to sequence and analyse the entire genomes of several isolates within days and the information generated has the potential to provide a very high level of discrimination. For this reason, WGS analysis was used during this outbreak to compare isolates at the genomic level and provide further insight into possible links between the clinical and environmental isolates.

A total of nine L. pneumophila serogroup 1 isolates from the hospital where the outbreak occurred were referred to or isolated by the Public Health Microbiology laboratory as part of the outbreak response. Two of these were clinical isolates from the two legionellosis patients and seven isolates were from the hot-water system of the hospital. Culture, identification and serogrouping were confirmed by the Legionella reference laboratory using standard methods [9]. Additional isolates included as part of the WGS analysis were clinical and water isolates from the Public Health Microbiology reference collection, which are described in more detail below.

SBT performed using the seven loci scheme described by the European Legionnaires' Disease Surveillance Network [10] identified all nine isolates as belonging to sequence type (ST) 1, and virulence gene profiling using the method described by Huang et al. [11] demonstrated that all isolates were positive for the lvh and rtxA regions. However, the epidemiological conclusions that could be drawn based on this data alone were limited because the majority of L. pneumophila serogroup 1 isolates typed by the Public Health Microbiology laboratory have also been found to be ST1 and to possess the same virulence gene profile.

In order to expand upon the results obtained by SBT and virulence gene profiling, and to provide an improved level of discrimination between isolates from different sources, the isolates from the two legionellosis patients (P1 and P2) and an isolate from the hot-water supply of the hospital where the outbreak occurred (W1) were subjected to WGS analysis, as were three isolates from our reference collection that were considered to be temporally and/or spatially unassociated with this outbreak. These consisted of two clinical isolates, C1, isolated from a 2011 patient at the same hospital, and C2, isolated in 2008 from a patient at another hospital located ∼5 km from the hospital where the current outbreak occurred, and one environmental isolate, W2, isolated in 2012 from the hot-water system of a building located ∼30 km from the hospital where the current outbreak occurred.

DNA was extracted from the isolates using the MasterPure DNA extraction kit (Epicentre, USA) according to the manufacturer's instructions. Fragment libraries of the genomic DNA were generated using the Ion Plus Fragment library kit and were sequenced on an Ion Torrent PGM (Life Technologies, USA) according to the manufacturer's instructions. Analysis of the genomic data generated was performed using CLC Genomics Workbench v. 4·9 (CLC Bio, Denmark). Sequencing reads were mapped to a reference genome, L. pneumophila strain Paris (Genbank accession no. CR628336) [12].

Comparison of the genome sequences from the isolates indicated that they were highly similar to the L. pneumophila Paris strain, which is also ST1. However, in all of the isolates sequenced, the sequences for the resistance-related genomic island R1 described in the Paris strain by D'Auria et al. [13] were mostly absent, with only the first four genes present. P1, P2, W1 and C1 also lacked sequences for 72 of the 142 coding regions found on the Paris strain plasmid (pLPP, Genbank accession no. NC_006365) including the genes for the F-type IV secretion system (T4SSA). These isolates did, however, possess sequences for the T4SSA genes found on the Lorraine strain plasmid (pLELO, Genbank accession no. NC_018141) [2] as well as the sequences for several other coding regions from this plasmid. C2 and W2 differed from these isolates in that they possessed the sequences for the entire pLPP and showed no significant similarity to sequences from pLELO. Whole genome shotgun sequences for P1, W1, W2, P2, C1 and C2 have been submitted to Genbank with accession nos. AWQT00000000, AVAP00000000, AVNJ00000000, AWES00000000, AVOW00000000 and AVOV00000000, respectively.

Single nucleotide polymorphisms (SNPs) in the genomes of the isolates tested were identified using CLC Genomics Workbench v. 4·9, filtering for coding region SNPs that were present in 100% of mapped reads at regions with a minimum coverage of 15 reads. SNPs were analysed by generating a maximum-likelihood tree using BioNumerics v. 6·5 (Applied Maths, Belgium). This analysis revealed that the isolates formed two distinct groups separated by 1512 SNPs. The patient and environmental isolates P1, P2, and W1 clustered together into one group which was highly related genetically, differing by a maximum of 17 SNPs (Fig. 1). The other group consisted of C2, W2 and the L. pneumophila Paris reference strain. The C1 isolate clustered with the outbreak isolates and was found to differ from P1, P2 and W1 by only 20, 17 and 18 SNPs, respectively. This isolate was isolated in 2011 from a patient at the same hospital, meaning that although it was not temporally associated, it was spatially associated with the isolates from the recent outbreak and may indicate a persistent presence of this strain at this location.

Fig. 1. Maximum-likelihood tree of L. pneumophila isolate single nucleotide polymorphisms (SNPs). Branch numbers indicate the number of SNPs.

Phylogenetic analysis of the isolates was performed by multilocus sequence analysis (MLSA) using 28 genes that have previously been shown to be powerful for predicting the relatedness of Legionella and other bacterial genomes [2]. The sequences for three of the genes used in the original MLSA study (lepA, metG, thdF) were not complete in one or more of the genomes of the isolates tested and were therefore excluded from the analysis. The sequences of the remaining 28 genes were concatenated into a single sequence 41 250 bp long and the concatenated sequences were aligned using CLC Genomics Workbench v. 4·9. A pairwise comparison was performed using CLC Genomics Workbench v. 4·9 and a maximum-likelihood tree was built using Geneious v. 6·1 (Biomatters, New Zealand). The phylogenetic tree constructed using the concatenated sequences separated the isolates into two distinct groups (Fig. 2), with one group, group 1, consisting of P1, P2, C1 and W1. The isolates in this group had identical nucleotide sequences, with the exception of P2, which differed by one nucleotide. Group 2 consisted of C2, W2 and the reference strain. This group differed from group 1 by 71 nucleotides and within this group the sequences differed by 2 bp. These groupings correlate with the clustering produced by the SNP analysis.

Fig. 2. Phylogenetic tree built using 28 housekeeping gene sequences concatenated into one sequence 41 205 bp long. The numbers next to the branches represent the percentages of the support for the groups.

Overall, comparison of these L. pneumophila isolates at the genomic level was consistent with the reported clonal nature of L. pneumophila serogroup 1 ST1. Comparison of genomic features, SNP analysis and MLSA demonstrated that the level of genomic diversity that exists in L. pneumophila isolated from this geographical region is low, which is reflected by the inability of some typing methods to discriminate between spatially and temporally unassociated isolates. The WGS analysis determined that there was far less genetic diversity between the isolates from the outbreak-related patients and those from the hospital hot-water system than there was to other spatially and/or temporally related isolates. This information was provided to the state public health units during the outbreak investigation and used as part of the response to the outbreak [14]. Comparison of genomic features, SNP analysis and MLSA also demonstrated that isolate C1 was highly similar to, and likely to be the same as, the outbreak strain, producing molecular evidence that the case in the same hospital 2 years previously was related to the current outbreak. The use of whole genome sequencing during this outbreak demonstrates how this technology can be used to identify links between environmental and patient isolates and to inform the public health response in real-time outbreak situations of L. pneumophila, even in regions showing endemicity of a clonal strain.