Genomic Epidemiology Links Burkholderia pseudomallei from Individual Human Cases to B. pseudomallei from Targeted Environmental Sampling in Northern Australia

ABSTRACT

Each case of melioidosis results from a single event when a human is infected by the environmental bacterium Burkholderia pseudomallei. Darwin, in tropical northern Australia, has the highest incidences of melioidosis globally, and the Darwin Prospective Melioidosis Study (DPMS) commenced in 1989, documenting all culture-confirmed melioidosis cases. From 2000 to 2019, we sampled DPMS patients’ environments for B. pseudomallei when a specific location was considered to be where infection occurred, with the aim of using genomic epidemiology to understand B. pseudomallei transmission and infecting scenarios. Environmental sampling was performed at 98 DPMS patient sites, where we collected 975 environmental samples (742 soil and 233 water). Genotyping matched the clinical and epidemiologically linked environmental B. pseudomallei for 19 patients (19%), with the environmental isolates cultured from soil (n = 11) and water (n = 8) sources. B. pseudomallei isolates from patients and their local environments that matched on genotyping were subjected to whole-genome sequencing (WGS). Of the 19 patients with a clinical-environmental genotype match, 17 pairs clustered on a Darwin core genome single-nucleotide polymorphism (SNP) phylogeny, later confirmed by single sequence typing (ST) phylogenies and pairwise comparative genomics. When related back to patient clinical scenarios, the matched clinical and environmental B. pseudomallei pairs informed likely modes of infection: percutaneous inoculation, inhalation, and ingestion. Targeted environmental sampling for B. pseudomallei can inform infecting scenarios for melioidosis and dangerous occupational and recreational activities and identify hot spots of B. pseudomallei presence. However, WGS and careful genomics are required to avoid overcalling the relatedness between clinical and environmental isolates of B. pseudomallei.

INTRODUCTION

The infectious disease melioidosis is caused by the environmental Gram-negative pathogen Burkholderia pseudomallei. Melioidosis presents with pneumonia and a multitude of other manifestations, with overall mortality of 10 to 40%, rising to over 90% in severe sepsis in some locations (1, 2). First recognized in Rangoon, Burma (Yangon, Myanmar), in 1911, B. pseudomallei is found in the rhizosphere, soil, and water of tropical and subtropical regions (3, 4). The greatest numbers of cases are reported annually in Southeast Asia, South Asia, and northern Australia (5, 6). Since 2012, the U.S. Centers for Disease Control and Prevention have listed B. pseudomallei as a tier 1 select agent. The global distribution of B. pseudomallei and the burden of disease have been poorly characterized, but recent modeling suggests that many previously unrecognized environments across the globe are receptive to B. pseudomallei (7), and reported case numbers have increased exponentially, including in Africa and the Americas (8).

Each case of melioidosis results from a single event when a human or animal is infected by exposure to B. pseudomallei in the environment. Zoonotic melioidosis and person-to-person transmission have been reported but are rare. Infection occurs through broken skin (percutaneous), inhalation (B. pseudomallei in aerosols), or ingestion (B. pseudomallei in water) (9). Transmission is facilitated by certain environmental conditions, such as severe weather events (tropical storms, cyclones, hurricanes, and typhoons) and environmental perturbation, such as urban construction, with case clusters sometimes resulting. In addition, specific occupations and outdoor activities are known to increase the risk of exposure to B. pseudomallei (10).

The extensive genetic diversity across B. pseudomallei strains is associated with the highly plastic nature of the B. pseudomallei genome (11), and B. pseudomallei can survive in diverse environments. For example, B. pseudomallei can survive in the harshest of environments, such as distilled water, nutrient-depleted soil, and desert and temperate environments (7, 12, 13). B. pseudomallei has been detected in water, and ingestion from unchlorinated water supplies has been linked to high case numbers in rural Thailand (9), while two historical outbreaks of melioidosis in remote Aboriginal communities in northern Australia were linked by bacterial genotyping to unchlorinated or inadequately chlorinated water supplies (14, 15). In 2015, B. pseudomallei was detected for the first time in aerosols in Taiwan, where cases were subsequently reported downwind during severe weather (16). We subsequently used whole-genome sequencing (WGS) to link the bacteria found from air sampling outside a patient’s room to the clinical isolate from that patient, who had presented with mediastinal melioidosis consistent with inhalational melioidosis (17).

Darwin is the capital of the Northern Territory, Australia, and is a coastal city in the tropical “Top End” of the Northern Territory. Darwin has the highest reported incidences of melioidosis globally, with diverse genotypes of B. pseudomallei being ecologically well established in the urban environment and highly prevalent in soil and drain water (soil, 75%; drains, 56%) (18). The Darwin Prospective Melioidosis Study (DPMS) commenced in 1989, documenting all culture-confirmed melioidosis cases in the Top End, each with corresponding epidemiological and clinical data and B. pseudomallei genomes (19). During the DPMS, we reported on three melioidosis case clusters which were resolved using WGS. Two clusters of cases were linked to unchlorinated water, with one cluster involving nine cases of melioidosis and four deaths in a remote Aboriginal community (14, 20). A third cluster involved two cases that were attributed to hand wash detergent contaminated with B. pseudomallei (21). Despite these genomic investigations to determine source attribution of case clusters, a substantial gap remains in our knowledge of the genomic epidemiology of melioidosis in individual Darwin patients.

Over a 19-year period we sampled the local environments of melioidosis patients for B. pseudomallei when a local source of infection seemed likely, based on patient history that was collected on admission to Royal Darwin Hospital. The primary aim of this study was to use our comprehensive genomic and epidemiological data set to determine if epidemiologically linked patient clinical and environmental isolates were genetically similar and thus likely linked by transmission. To compare clinical and epidemiologically linked environmental B. pseudomallei isolates, we used a combination of genotyping, core genome single nucleotide polymorphism (SNP) phylogenies, and pairwise comparative genomics. We took advantage of our genomic epidemiology analysis to address a second aim, which was to investigate the genomic approach that provided the highest genetic resolution for inferring a transmission event. To address this, we assessed the impact of reference genome relatedness (close versus distant) and genetic diversity (high versus low diversity in alignment) on the number of core genome SNPs separating the clinical isolates from the epidemiologically linked environmental isolates.

MATERIALS AND METHODS

Ethics approval.

This study was approved by the Human Research Ethics Committee of the Northern Territory Department of Health and the Menzies School of Health Research (MSHR) (HREC 02/38). Patient consent was obtained for sampling to be undertaken at the patients’ residences.

Environmental sampling at sites epidemiologically linked to infection.

From 2000 to 2019, we performed environmental sampling at 98 DPMS melioidosis patient residential addresses or linked exposure sites to pinpoint infection locations. The 98 patients were located across the Top End in northern Australia, with most in the Darwin city region and its rural surroundings. We collected ∼3 water samples, mainly from unchlorinated bores, and/or ∼10 soil samples from each of the epidemiologically linked sites (see Fig. S1 in the supplemental material). Where possible, soil samples were spaced 10 m apart and taken from a 30-cm depth. Global Positioning System (GPS) coordinates were recorded at each site using a Garmin GPS device (Garmin eTrex30).

B. pseudomallei isolates, growth, confirmation, and DNA extraction.

Immediately after the collection of clinical and environmental samples, we attempted to culture B. pseudomallei using methods previously established at the Menzies laboratory (17, 22, 23). Confirmation of B. pseudomallei was performed using a real-time PCR assay targeting the B. pseudomallei-specific 115-bp segment within the type III secretion system 1 (TTS1) gene (24). Following B. pseudomallei confirmation in both the environmental and clinical specimens, genomic DNA was extracted from purified B. pseudomallei colonies (25).

BOX-PCR typing and multilocus sequence typing assignment.

BOX-PCR typing using the BOX Air 1 primer was performed on the linked clinical and environmental B. pseudomallei strains (25). Strains with a matching BOX-PCR pattern were subsequently genotyped using multilocus sequence typing (MLST) to assign a sequence type (ST) to each strain. MLST was done either by seven-allele sequencing as previously described (26) or by determining the allele sequence in silico from bacterial WGS read data using the BIGSdb tool, which is accessible on the B. pseudomallei MLST website (https://pubmlst.org/bpseudomallei/) (27).

Genome selection, quality control, and genome assembly.

For this study, we sequenced 123 isolates (Table S1), with genomes sequenced on Illumina platforms and each sequencing platform generating 150-bp paired-end reads. Genomic analysis included an additional 519 publicly available B. pseudomallei genomes, including genomes that captured the genetic diversity of B. pseudomallei from the Northern Territory, a selection of genomes that represented the five most common STs in Darwin, and environmental genomes that belonged to patients investigated in this study (Table S1). Reads were trimmed with Trimmomatic v0.39 (28), and quality was assessed with FastQC (https://www.bioinformatics.babraham.ac.uk/projects/fastqc). Following adapter trimming and quality assessment, draft assemblies were generated using the MGAP pipeline (https://github.com/dsarov/MGAP---Microbial-Genome-Assembler-Pipeline). Assemblies were deemed to be of good quality and so were included in downstream genomic analysis (contigs < 400 and N₅₀ > 80,000 bp).

Variant calling and phylogenetic analysis.

All sequence read mapping and identification of core genome SNPs was performed using default settings in SPANDx v3.2 (29). The SNP variants identified by SPANDx v3.2 were used for phylogenetic reconstruction using maximum parsimony (MP) in PAUP 4.0.b5 (30). Bootstrapping using 1,000 replicates was carried out for phylogenies to establish node support. Phylogenetic trees were visualized and manipulated using ggtree in R or Interactive Tree of Life (ITOL; https://itol.embl.de).

Rationale for genome alignments and phylogenetic analysis.

We constructed six phylogenies (i.e., from six core genome SNP alignments) to (i) establish the relatedness of epidemiologically linked clinical-environmental isolates and (ii) investigate the impact of reference genome relatedness and genetic diversity on the number of core genome SNPs that differentiated each clinical isolate from its linked environmental isolate. The rationale was that a more closely related reference genome would reduce the chance of mismapping and increase the regions in the reference genome against which reads would be mapped. The first phylogeny included the ST-matched study isolates and a set of genetically diverse Northern Territory B. pseudomallei genomes. The closed Australian B. pseudomallei genome MSHR1153 (distantly related to our study isolates) was used as the reference for read mapping (N₅₀, 4,032,226 bp; contigs, 2; size, 7,312,903 bp; CP009271 and CP009272) (31). We next constructed a core genome SNP alignment and subsequent phylogenetic tree for each of the five most prevalent disease-causing STs isolated from Darwin patients (ST36, ST109, ST132, ST553, and ST562), and each alignment included the ST-matched patient and environmental genomes from this study that belonged to those STs. For each sequence type alignment, a reference genome of the same sequence type (closely related and genetically similar to study isolates) was used for read mapping (reference genome for ST36, MSHR3763 [CP017052 and CP017053]; reference genome for ST109, MSHR4462 [JPQM00000000]; reference genome for ST132, MSHR465J [JPZW00000000]; reference genome for ST553, MSHR5855 [CP008784 and CP008783]; and reference genome for ST562, MSHR5858 [CP008892 and CP008891]).

Pairwise analysis for detection of genome nucleotide differences and large deletions.

For further comparative genomics refinement, we performed pairwise comparative genomics of ST-matched clinical and environmental isolates linked to each patient to identify variants and large deletions between isolates. For each patient, the corresponding environmental isolate draft assembly was used as the reference genome for the clinical isolate. In brief, the clinical paired-end reads in FASTQ format were mapped to the annotated environmental genome to identify SNPs and large deletions using SPANDx v3.2. Additionally, the paired-end reads from each environmental reference were simultaneously mapped to itself as a draft assembly to increase accuracy in SNP and large-deletion identification. For confirmation of SNP differences, the BAM file from each patient (the B. pseudomallei clinical isolate genome) was visualized against the environmental reference genome using Tablet (James Hutton Institute, Scotland, UK). For identification of large deletions in the accessory and core genome, the SPANDx BED tools locus presence/absence output was visualized for each ST-matched pair belonging to each patient.

Data availability.

Sequencing reads generated during this study were submitted to the Sequence Read Archive database on NCBI under BioProject no. PRJNA746348.

RESULTS

Environmental sampling at sites epidemiologically linked to human infections.

The Darwin melioidosis study recorded 1,212 episodes of melioidosis in 1,148 individuals over 30 years (1989 to 2019) (32). Of the 1,148 cases, 13 were previously associated with three melioidosis case clusters (summarized in Table S3) and the remaining 1,135 cases were not known to be part of clusters (Fig. 1). Eight hundred thirty-nine of these cases occurred from 2000 onwards, and in 98 (12%) of these, we sampled the patient’s local environment based on clinical history suggesting a potential exposure site (Fig. 1). The 98 sites that were epidemiologically linked to infection were predominantly located in the Darwin urban areas (n = 53 [54%] sites) and Darwin rural regions (n = 41 [42%] sites), with an additional 2 sites in remote communities and 1 in the small township of Katherine, ∼318 km south of Darwin. At 84 sites, we performed environmental sampling within 6 months of a reported case, with a median time of 23 days and an interquartile range of 38 days after clinical presentation. For 14 sites, environmental sampling was performed 6 months to 5 years after a case was reported, with a median time of 617 days and an interquartile range of 459 days after clinical presentation.

FIG 1.

A total of 1,148 melioidosis patients were reported to have melioidosis as a part of the 30-year Darwin study, and 13 patients were involved in three case clusters, as previously demonstrated by genomics. For this study, we classified the remaining 1,135 patient cases as endemic cases, each considered likely to be linked to a single point source. Of these, 839 patients were from 2000 onward, and we investigated the source of infection for 98 (12%) of these patients, which are further stratified by time between case and environmental sampling, environmental culture of B. pseudomallei, and ST match.

From the 98 sites epidemiologically linked to melioidosis cases, we collected 975 environmental samples, which included 742 soil (median number collected at each site = 8) and 233 water (median number collected at each site = 3) (Fig. 2a). B. pseudomallei was prevalent in environments epidemiologically linked to human disease and was cultured from 12% of soils (92 of 742 samples collected) and 22% of water samples (52 of 233 samples collected) (see Fig. 2c for a breakdown of positives and negatives for each geographical region), which is consistent with our previous B. pseudomallei environmental isolation rate (18).

FIG 2.

Environmental sampling framework for B. pseudomallei. (a) Numbers of clinical and environmental samples from melioidosis patients and household water or soil. (b) Temporal distribution of patient investigations and collection of environmental samples that were stratified by culture of B. pseudomallei. (c) Geographical distribution of environmental samples collected.

Overall, B. pseudomallei was recovered from the environment of 50 (51%; see Table S4 for the frequency of recovery of B. pseudomallei from various water/soil samples) patient sites, with 48 (49%) patient sites being culture negative (Fig. 1). B. pseudomallei was identified in the environment of urban Darwin and the surrounding Darwin rural region over a large spatial scale with a linear distance of ∼30 km (Fig. 3).

FIG 3.

Sites epidemiologically linked to melioidosis infections. (A) Map of Australia. The region sampled during this study is located within the white box. (B) Map of sites sampled, which are stratified by culture of B. pseudomallei from the environment, ST match, or ST mismatch. Three sites not included on the map were from outside the greater Darwin area.

Genotyping linked clinical isolates to environmental isolates.

From the 126 B. pseudomallei isolates (51 clinical, 20 water, 54 soil, and 1 air) that were genotyped as a part of our surveillance study, we identified 43 distinct MLST genotypes (Table S2). ST109 was the most frequently observed genotype (n = 35), followed by ST132 (n = 11), ST36 (n = 7), and ST562 (n = 7). Fourteen (33%) of the B. pseudomallei STs found from environmental samples were also present in clinical patient isolates.

At sites positive for B. pseudomallei (50 sites, corresponding to 50 melioidosis patients), genotyping matched the ST of the clinical B. pseudomallei isolate to the epidemiologically linked environmental isolate for 19 patients, with the ST mismatching for 31 patients (Fig. 1). For the 19 patients with an ST match, soil was linked to 11 cases and bore water on rural properties was linked to 8 cases, with the ST matches geographically dispersed across the urban Darwin and surrounding rural regions (Fig. 3). Of the 19 ST-matched clinical and environmental isolates, 8 were ST109, 3 were ST132, and 1 each was ST36, ST131, ST279, ST325, ST326, ST337, ST553, and ST562.

Epidemiology of 19 patients with a genotype match.

Detailed histories of the 19 patients where genotyping matched the ST of the clinical isolate to the ST of the epidemiologically linked environmental isolate are provided in Text S1. In brief, infecting scenarios included recreational and occupational exposure activities related to outdoor physical activity, play, gardening, and outdoor maintenance, such as cutaneous exposure through cuts and trauma and presumptive inhalation of aerosols from severe weather events or use of a whipper snipper (weed whacker). Infecting scenarios also included exposure to B. pseudomallei from unchlorinated bore water located on patient properties in the Darwin rural region, with percutaneous infection, ingestion, and aerosolization from showering all potential modes of transmission.

Evidence of genomic linkage.

Of the 19 patients for whom clinical isolate and environmental isolate were the same ST, 17 patients had isolates that were phylogenetically closely related (range, 0 to 9 SNPs; median SNP difference = 1) (Table 1). Clinical and environmental isolates belonging to the two other patients (P290 and P449) were separated by 989 and 113 SNPs, respectively, precluding a likely direct infection link in these 2 scenarios despite the matching STs (Table 1 and Fig. 4). Of the six patients with B. pseudomallei in bore water from their rural Darwin property linked to their clinical B. pseudomallei by 0 to 7 SNPs (Table 1), three (P379, P710, and P821) had cutaneous melioidosis, with percutaneous inoculation of B. pseudomallei in bore water being the likely mode of infection. Of the 11 patients with soil B. pseudomallei linked to their clinical B. pseudomallei isolate by 0 to 9 SNPs (Table 1), one (P493) had an exposure location which was not his residential address but an urban Darwin public roadside location where he used a whipper snipper to trim grass. Cultures from soil sampled from the roadside verge yielded a B. pseudomallei isolate which was separated by 2 SNPs from his blood culture B. pseudomallei isolate; his presentation with pneumonia was considered consistent with inhalational melioidosis from B. pseudomallei-containing aerosols generated by the whipper snipper.

TABLE 1.

SNP differences between B. pseudomallei clinical case isolates and B. pseudomallei environmental isolates from the 19 patients with ST matches^a

DPMS patient and isolate	ST	Date of collection (interval [days])	Sample origin	SNP difference			Likely transmission eventb	Exposure	Presumptive mode of infection	Primary presentation
				NT phylogeny	Single-ST phylogeny	Pairwise comparison
P289							Yes	Recreational activities on property	Inhalation	Bacteremic and pulmonary melioidosis
MSHR1415	109	11 July 2001	Environmental (soil from P289’s property)	—	—	—
MSHR0910	109	1 Jan 2000 (557)	Clinical (rectal swab)	1	1	4
P290							No	NA	NA	Chronic pulmonary melioidosis
MSHR1868	337	15 May 2004	Environmental (P290’s bore water)	—	—	—
MSHR0912	337	18 Jan 2000 (1,579)	Clinical (sputum)	989	NA	1,294
P314							Yes	Unchlorinated bore water on property	Ingestion	Pulmonary melioidosis
MSHR1435	131	3 Oct 2002	Environmental (P314’s bore water)	—	—	—
MSHR1043	131	11 July 2000 (814)	Clinical (throat swab)	7	NA	8
P379							Yes	Unchlorinated bore water on property	Percutaneous inoculation	Cutaneous melioidosis
MSHR1588	326	1 Mar 2003	Environmental (P379’s bore water)	—	—	—
MSHR1504	326	28 Jan 2003 (32)	Clinical (foot swab)	5	NA	4
P449							No	NA	NA	Cutaneous melioidosis
MSHR2154	36	2 Aug 2005	Environmental (soil from sports oval)	—	—	—
MSHR2078	36	17 May 2005 (77)	Clinical (groin lesion)	113	136	152
P486							Yes	Recreational activities on property	Inhalation	Bacteremic melioidosis
MSHR2494	109	11 Jan 2007	Environmental (soil from P486’s property)	—	—	—
MSHR2436	109	29 Dec 2006 (13)	Clinical (blood culture)	9	9	9
P493							Yes	Using a whipper snipper	Inhalation	Bacteremic and pulmonary melioidosis
MSHR2571	132	6 Mar 2007	Environmental (soil)	—	—	—
MSHR2472	132	20 Feb 2007 (14)	Clinical (blood culture)	2	2	2
P500							Yes	Playing in yard	Percutaneous inoculation	Cutaneous melioidosis
MSHR2850	109	9 Aug 2007	Environmental (soil from P500’s property)	—	—	—
MSHR2535	109	6 Mar 2007 (156)	Clinical	0	0	0
P629							Yes	Trauma to finger while removing irrigation from yard	Percutaneous inoculation	Cutaneous melioidosis
MSHR4340	109	5 Aug 2010	Environmental (soil from P629’s property)	—	—	—
MSHR4196	109	12 July 2010 (24)	Clinical (wound swab)	1	1	1
P648							Yes	Recreational exposure in yard	Percutaneous inoculation	Soft-tissue abscess melioidosis
MSHR4459	279	26 Oct 2010	Environmental (soil from P648’s property)	—	—	—
MSHR4181	279	25 June 2010 (123)	Clinical (abscess pus)	4	NA	6
P679							Yes	Gardening in yard	Inhalation	Genitourinary melioidosis
MSHR4569	109	25 Jan 2011	Environmental (soil from P679’s property)	—	—	—
MSHR4441	109	16 Dec 2010 (40)	Clinical (urine)	1	1	1
P692							Yes	Air at accommodation	Inhalation	Septic shock, pulmonary melioidosis
MSHR4681	562	18 Feb 2011	Environmental (soil from P692’s accommodation)	—	—	—
MSHR4687	562	18 Feb 2011	Environmental (air from P692’s accommodation)	—	—	—
MSHR4515	562	13 Jan 2011 (36)	Clinical (blood culture)	0	3	1
P710							Yes	Unchlorinated bore water on property	Percutaneous inoculation	Cutaneous melioidosis
MSHR4997	132	8 Apr 2011	Environmental (P710’s bore water)	—	—	—
MSHR4758	132	16 Mar 2011 (23)	Clinical (breast abscess)	0	0	2
P821							Yes	Unchlorinated bore water on property	Percutaneous inoculation	Cutaneous melioidosis
MSHR7797	109	22 Jan 2013	Environmental (P821’s family friends’ bore water)	—	—	—
MSHR7723	109	12 Dec 2012 (32)	Clinical (skin lesion)	1	1	1
P1018							Yes	Unchlorinated bore water on property	Ingestion	Genitourinary (tubo-ovarian) melioidosis
MSHR9292	325	20 Apr 2016	Environmental (P1018’s bore water)	—	—	—
MSHR9221	325	2 Apr 2016 (18)	Clinical (pus)	0	NA	0
P1034							Yes	Recreational exposure on property	Percutaneous inoculation	Septic arthritis melioidosis
MSHR9684	109	25 Oct 2016	Environmental (soil from P1034’s property)	—	—	—
MSHR9436	109	22 July 2016 (95)	Clinical (blood culture)	5	5	6
P1127							Yes	Unchlorinated bore water on property	Ingestion	Hepatic abscess
MSHR11808	132	13 June 2018	Environmental (P1127’s bore water)	—	—	—
MSHR11634	132	4 Apr 2018 (70)	Clinical (blood culture)	1	1	4
P1131							Yes	Digging in soil with no gloves, exposing psoriasis wounds	Percutaneous inoculation	Bacteremic melioidosis
MSHR11805	109	5 June 2018	Environmental (P1131’s soil)	—	—	—
MSHR11821	109	5 June 2018	Environmental (P1131’s soil)	—	—	—
MSHR11824	109	5 June 2018	Environmental (P1131’s soil)	—	—	—
MSHR11640	109	17 Apr 2018 (49)	Clinical (blood culture)	6 (MSHR1821 and MSHR11805)	9 (MSHR11805); 11 (MSHR11821)	15 (MSHR11805)
P1159							Yes	Unchlorinated bore water on property	Ingestion	Pulmonary and genitourinary melioidosis
MSHR12486	553	16 Apr 2019	Environmental (water from creek at P1159’s property)	—	—	—
MSHR12252	553	4 Feb 2019 (71)	Clinical (blood culture)	6	6	7

^aNA, not applicable.

^bBased on corresponding patient epidemiology data and strain genomics.

FIG 4.

Maximum-parsimony phylogeny of B. pseudomallei isolated from DPMS patients and their epidemiologically linked environmental B. pseudomallei isolates (n = 19 patients), with a set of Darwin genomes for local genetic context (n = 373) and rooted with MSHR0668. Phylogeny is based on 177,247 core genome SNPs. Homoplasy index = 0.4539; consistency index = 0.5461. Colored node circles indicate Darwin patients that had an ST match, the strip delineates isolate source (human, soil, air, animal, or water), and five prevalent B. pseudomallei STs found in Darwin melioidosis patients are shown (ST36, ST109, ST132, ST562, and ST553). Nodes were well supported for the 17 patient clusters.

To determine if single ST phylogenies provide higher genetic resolution compared to the Darwin phylogeny, we next created single-ST phylogenies for the five most prevalent STs found in Darwin melioidosis patients (Fig. S2 to S6; Table 1). ST-only trees supported the clustering observed in the larger Northern Territory phylogeny, but for two patients (P449 and P692), additional SNPs were identified between respective clinical and environmental isolates. However, the median SNP difference was concordant between the single-ST-phylogeny approach with the Darwin large phylogeny, and so overall, the single-ST phylogeny approach did not provide higher genetic resolution.

Pairwise genomics confirms same strain and supports transmission events.

To identify the maximum number of SNP differences and large deletions separating clinical and environmental isolates belonging to the 19 patients, we next performed pairwise comparative genomics (Table 1). When the clinical sequencing reads were mapped to the matched environmental isolate assembly, isolates belonging to 17 patients were found to be separated by 0 to 15 SNPs, with a median difference of 4 SNPs, and we did not identify large deletions in core or accessory genome regions (BED tools locus presence/absence output). This further confirmed the isolate genetic match for 17 of the 19 patients with concordant clinical and environmentally linked B. pseudomallei STs.

For patients P290 and P449, who on initial phylogeny had large SNP separations between their ST-concordant clinical and environmentally linked B. pseudomallei isolates, mapping of the respective clinical sequencing reads to the matched environmental isolate assembly identified 1,294 and 152 SNP differences, respectively, confirming that for these 2 cases, infection was not from the environmentally linked isolate.

Pairwise genetic differences between the epidemiologically linked isolates (n = 19) did not correlate with time (Fig. S7). For example, zero SNP differences were noted between linked clinical and environmental B. pseudomallei isolates belonging to both P1018 and P500, but isolates belonging to P1018 were separated by 18 days, while isolates belonging to P500 were separated by 156 days (Fig. S7).

DISCUSSION

To identify the source of B. pseudomallei infection, we sampled 98 melioidosis patients’ local environments for B. pseudomallei when clinical history suggested that infection had occurred at a specific location. We performed the greatest number of epidemiological investigations in 2010 and 2011, when Darwin was impacted by above-average rainfall due to a strong La Niña event and case numbers were particularly high (32). We identified B. pseudomallei at 50 patient sites, and comparative genomics found that environmental B. pseudomallei isolates from each patient’s linked site matched the clinical B. pseudomallei isolates for 17 patients. Minimal core genome SNP differences, lack of deletions between genome pairs, and corresponding patient epidemiology supported the idea that patients acquired their infections from the environment at their epidemiologically linked site.

Our results provide support for potential modes of infection that were described by patients to occur at the locations where the genomically linked environmental B. pseudomallei isolates were recovered. Studies from Darwin and Thailand have suggested that broken skin, inhalation, and ingestion are important routes for the development of melioidosis (9, 17, 20, 33). Each of these speculative modes of infection was represented among the 17 patients with genomically linked environmental and clinical B. pseudomallei isolates, and fuller descriptions are provided in our 30-year prospective melioidosis study review (19). The data support the current Northern Territory public health guidelines that recommend wearing appropriate footwear and gloves, covering sores, and wearing masks while undertaking outdoor recreational or occupational activities such as gardening or high-pressure hosing.

Using genomics and linked patient epidemiological data, we identified an infection location for 17 of the 50 patients that had a linked B. pseudomallei environmental isolate. Tracking transmission of the remaining patients may have been hampered by the genetic diversity of organisms found within a host or environmental source. However, in human B. pseudomallei infection, polyclonality is thought to be rare, as demonstrated by genotyping of multiple colonies from single melioidosis patients (34). Thus, for melioidosis, a single colony from a patient is sufficient to reconstruct the chain of transmission, except for chronic cases. In contrast, single environmental samples can yield cultures of polyclonal B. pseudomallei isolates with up to 4 genotypes detected in single soil and bore water samples (35, 36). A single colony isolated from an environmental sample is likely insufficient in areas of endemicity to reconstruct chains of transmission between otherwise epidemiologically linked clinical and environmental sources. To capture the genetic diversity within an environmental sample, numerous colonies could be sequenced and might provide greater sensitivity for identifying B. pseudomallei transmission events.

A SNP cutoff or distance threshold for inferring transmission of B. pseudomallei from environment to human has not been established. Our study adds value, as SNP differences between isolates with a strong epidemiological link are required when establishing a genetic similarity benchmark (37). Previous epidemiological investigations on Australian melioidosis animal and human clusters have used SNP differences ranging from 0 to 5 SNPs for inferring an environmental transmission event or 0 SNPs for identifying human case clusters where infection likely came from a common source but where no environmental isolate is available to confirm transmission (17, 20, 33, 38–41). For this study we found a maximum of 15 SNPs in the 17 case-environment isolate matches for which we inferred a causal transmission. This is concordant with SNP cutoffs established for other clinically relevant bacteria, with 2 to 37 SNP differences being reported (42–51). For example, ≤37 SNPs has been used as the cutoff for inferring transmission of Pseudomonas aeruginosa (51), a good comparison to B. pseudomallei, due to its similar genetic makeup and biology. As in these epidemiological studies, we used core genome SNPs and closely related isolates as reference genomes, with our analysis demonstrating that pairwise genomics provided the highest genetic resolution, which can be useful for determining the origin of melioidosis outbreaks involving clonal strains.

B. pseudomallei isolates belonging to a single ST can be genetically clonal but separated by many years. For example, ST284 isolates from southwest Western Australia (WA) differed by only 532 SNPs despite being collected over a 51-year period (40). Similarly, work on ST562, an Asian B. pseudomallei genotype estimated to have been introduced into the Darwin environment around 1988, showed a limited diversity of 141 SNPs among 71 isolates over the 30+ years since the presumptive single introduction (41). The limited genetic diversity among B. pseudomallei STs can significantly confound transmission dynamics. Consistent across these WA and Darwin STs is that epidemiologically unrelated isolates differed by ≤1 SNP; a combination of epidemiology and phylogenetic analysis including closely related local isolates for context is therefore required to refute or rule in a link (41). Notably, there is an assumption of a positive correlation between number of SNPs and time since the transmission event. We detected clinical and environmental isolates separated by months or a few days, but in both instances, linked isolates were separated by 0 SNPs. The lack of correlation between genetic divergence and time has been demonstrated previously for B. pseudomallei and other important pathogens, including Francisella tularensis, Yersinia pestis, and Legionella pneumophila, and could be attributed to the bacteria potentially existing in a dormant nonreplicative state (40, 52–54). Consequently, nucleotide mutation rate, horizontal gene transfer, and selective pressures would need to be built into any future model using SNP number to predict B. pseudomallei transmission.

An inherent limitation of core genome SNP analysis is that it is restricted to genome regions present in all analyzed isolates, which means that useful information in the accessory genome is discarded. To overcome this limitation, we performed pairwise comparative analysis of genomes using the respective environmental isolate as an index isolate, which enabled us to interrogate both accessory and core genome regions. In the 17 matched pairs, pairwise analysis failed to detect large-scale genetic differences across the accessory and core genomes, supporting the phylogenetic analyses. It is anticipated that in the future, a combination of genomic approaches, such as combined analysis of variation in core, accessory, and regulatory genome regions, will be used for epidemiology investigations. The combined genomic approach has been undertaken on the important Escherichia coli lineage ST131, providing a superresolution view into the global epidemiology of ST131 (55). The multigenome approach is valuable for genomic investigations of clonal melioidosis outbreaks and for in-depth analysis of the epidemiology of B. pseudomallei STs that are found in multiple geographies.

Here, we present a genomic study on Darwin human melioidosis cases demonstrating that environments at Darwin residential properties and public sites are linked to human melioidosis infections. Our study highlights that percutaneous inoculation, inhalation, and ingestion are likely modes of infection in the Darwin region, as supported by the genomics and linked patient epidemiology. A vaccine that protects against B. pseudomallei is currently unavailable, and so alternative solutions for preventing melioidosis in our community have been put in place by the Northern Territory Government public health team. Our findings support the current melioidosis public health guidelines and can shape future melioidosis genomic surveillance programs locally and globally.