Can genomics and meteorology predict outbreaks of legionellosis in urban settings?

Abstract

Legionella pneumophila is ubiquitous and sporadically infects humans causing Legionnaire's disease (LD). Globally, reported cases of LD have risen fourfold from 2000 to 2014. In 2016, Sydney, Australia was the epicenter of an outbreak caused by L. pneumophila serogroup 1 (Lpsg1). Whole-genome sequencing was instrumental in identifying the causal clone which was found in multiple locations across the city. This study examined the epidemiology of Lpsg1 in an urban environment, assessed typing schemes to classify resident clones, and investigated the association between local climate variables and LD outbreaks. Of 223 local Lpsg1 isolates, we identified dominant clones with one clone isolated from patients in high frequency during outbreak investigations. The core genome multi-locus sequence typing scheme was the most reliable in identifying this Lpsg1 clone. While an increase in humidity and rainfall was found to coincide with a rise in LD cases, the incidence of the major L. pneumophila outbreak clone did not link to weather phenomena. These findings demonstrated the role of high-resolution typing and weather context assessment in determining source attribution for LD outbreaks in urban settings, particularly when clinical isolates remain scarce.IMPORTANCEWe investigated the genomic and meteorological influences of infections caused by Legionella pneumophila in Sydney, Australia. Our study contributes to a knowledge gap of factors that drive outbreaks of legionellosis compared to sporadic infections in urban settings. In such cases, clinical isolates can be rare, and thus, other data are needed to inform decision-making around control measures. The study revealed that core genome multi-locus sequence typing is a reliable and adaptable technique when investigating Lpsg1 outbreaks. In Sydney, the genomic profile of Lpsg1 was dominated by a single clone, which was linked to numerous community cases over a period of 40 years. Interestingly, the peak in legionellosis cases during Autumn was not associated with this prevalent outbreak clone. Incorporating meteorological data with Lpsg1 genomics can support risk assessment strategies for legionellosis in urban environments, and this approach may be relevant for other densely populated regions globally.

Keywords: Legionella pneumophila; Legionnaires' disease; bacterial genomics; epidemiology; surveillance; whole genome sequencing.

Fig 1

Core genome phylogeny of 223 Lpsg1 genomes based on 1,231 core genes. Colored squares indicate BAPS cluster designation. Red dots indicate whether the isolate was collected as part of an outbreak investigation. The color bar on the perimeter indicates the year collected in 5-year increments according to the legend. A higher density of isolates recorded as part of an outbreak occurs in BAPS cluster 4. Visualization produced in iTOL (https://itol.embl.de/).

Fig 2

Minimum spanning tree of the 223 Lpsg1 genomes based on 1,521 loci with genomes with five allelic differences (CGRNA-4) collapsed and overlaid with (A) CGRNA-1. (B) BAPs cluster. (C) SBT sequence types. Each node represents either a single or a collection of genomes, color coded with pertinent information based on the different overlaid information. Allelic differences are scaled according to the scale bar. Visualization was produced with Grapetree (19).

Fig 3

Mean monthly weather observations vs reported LD case counts across NSW over the calendar year. Data range from years 1990 to 2019. LD case count means are represented statically across all four subfigures in green. Top left: relative humidity at 9 am, blue. Top right: relative humidity at 3 pm, pink. Bottom left: air temperature at 3 pm, orange. Bottom right: monthly precipitation, green. Results visualized with ggplot2 (20).

Fig 4

Fourteen-day lagged weather observations vs 14-day lagged moving average weather at the time of isolate collection by cgMLST. Individual cgMLSTs are indicated by colored circles. (A) Relative humidity at 9:00 am vs total rainfall in the previous 24 hours. (B) Relative humidity vs air temperature at 9:00 am. (C) Air temperature at 9:00 am vs air temperature at 3:00 pm. All circles are randomly jittered for clarity of isolated group sizes. Gray circles represent dates in which no isolates in this study were collected. Results visualized with ggplot2 (20).