Translating genomics into practice for real-time surveillance and response to carbapenemase-producing Enterobacteriaceae: evidence from a complex multi-institutional KPC outbreak

Abstract

Background: Until recently, Klebsiella pneumoniae carbapenemase (KPC)-producing Enterobacteriaceae were rarely identified in Australia. Following an increase in the number of incident cases across the state of Victoria, we undertook a real-time combined genomic and epidemiological investigation. The scope of this study included identifying risk factors and routes of transmission, and investigating the utility of genomics to enhance traditional field epidemiology for informing management of established widespread outbreaks.

Methods: All KPC-producing Enterobacteriaceae isolates referred to the state reference laboratory from 2012 onwards were included. Whole-genome sequencing was performed in parallel with a detailed descriptive epidemiological investigation of each case, using Illumina sequencing on each isolate. This was complemented with PacBio long-read sequencing on selected isolates to establish high-quality reference sequences and interrogate characteristics of KPC-encoding plasmids.

Results: Initial investigations indicated that the outbreak was widespread, with 86 KPC-producing Enterobacteriaceae isolates (K. pneumoniae 92%) identified from 35 different locations across metropolitan and rural Victoria between 2012 and 2015. Initial combined analyses of the epidemiological and genomic data resolved the outbreak into distinct nosocomial transmission networks, and identified healthcare facilities at the epicentre of KPC transmission. New cases were assigned to transmission networks in real-time, allowing focussed infection control efforts. PacBio sequencing confirmed a secondary transmission network arising from inter-species plasmid transmission. Insights from Bayesian transmission inference and analyses of within-host diversity informed the development of state-wide public health and infection control guidelines, including interventions such as an intensive approach to screening contacts following new case detection to minimise unrecognised colonisation.

Conclusion: A real-time combined epidemiological and genomic investigation proved critical to identifying and defining multiple transmission networks of KPC Enterobacteriaceae, while data from either investigation alone were inconclusive. The investigation was fundamental to informing infection control measures in real-time and the development of state-wide public health guidelines on carbapenemase-producing Enterobacteriaceae surveillance and management.

Keywords: Antimicrobial resistance; Klebsiella pneumoniae carbapenemase; Microbial genomics; Multidrug-resistant organisms; Outbreak investigation; Transmission modelling; Whole-genome sequencing.

Figure 1. Workflow summary diagram of methods used in this study.

Microbiology methods (orange), bioinformatic methods (blue), and epidemiological methods (green) used to generate data for analysis and reporting to public health authorities (purple) are shown. Results of combined prospective epidemiological and genomic analyses performed iteratively during the outbreak were reported to the Department of Health and Human Services in real-time. CPE, carbapenemase-producing Enterobacteriaceae; KPC, Klebsiella pneumoniae carbapenemase; MALDI-TOF, matrix-assisted laser desorption/ionisation time-of-flight; PCR, polymerase chain reaction; WGS, whole-genome sequencing; SNP, single nucleotide polymorphism; AMR, antimicrobial resistance; MLST, multi-locus sequence typing.

Figure 2. Incidence of new KPC-producing Enterobacteriaceae cases referred to MDU PHL, 2012–2015.

Blocks are coloured by the species and KPC allele of the referred isolates. Repeated detections of KPC-producing isolates from the same patient have been excluded.

Figure 3. The initial maximum likelihood phylogenetic tree comprised three dominant clades.

The tree includes 29 ST258 K. pneumoniae isolates collected from 29 patients prior to June 2014, with external nodes coloured according to healthcare facility at time of sample collection. Recurrent isolates from each patient have been excluded. Multiple patients in Facility A (purple) and Facility F (orange) were colonised/infected with KPC-producing K. pneumoniae, corresponding to known previous outbreaks in those hospitals. Support values (%) from 1,000 bootstrap replicates are shown for major branches. Major phylogenetic clades have been labelled cluster A (green shading), B (orange shading), and C (purple shading) in the order that the clades emerged, with the larger clade B comprising two subclades, B1 and B2. Corresponding clusters identified through Bayesian analysis of the population structure (BAPS) are also shown. The tree was rooted using an outgroup isolate (K. pneumoniae NJST258_1, GenBank accession CP006923.1; not shown in the tree) from a different ST258 clade.

Figure 4. Epidemiological links based on overlapping patient admissions were unable to resolve where and when transmission was occurring for most isolates.

Links in the above network connect patients based on overlapping patient admissions in the same hospital ward at the same time (minimum of one day overlap). Links that occurred after detection of bla_KPC in both patients have been excluded. Nodes are numbered by patient, and coloured by healthcare facility (as in Fig. 3) at the time of sample collection. The epidemiological network connecting patients 22, 23, 24 and 33 correlates with the closely related genomes in cluster C (Fig. 3).

Figure 5. Combined analysis, with genomic relationships between isolates overlaid upon epidemiological data, delineated multiple transmission networks.

A maximum likelihood phylogenetic tree is shown on the left, labelled with major genomic clusters and supporting branch bootstrap values (%) from 1,000 replicates for major branches. Nodes of the tree are coloured by healthcare facility at the time of sample collection as in Figs. 3 and 4. Coloured horizontal bars on the right indicate healthcare facility admissions over time (x axis), with different colours representing different healthcare networks. Black diamonds (♦) indicate first detection of KPC for each patient.

Figure 6. Combined analysis of genomic and epidemiological data of the cluster C network identified secondary transmission.

Example case study: Patients 22 and 23 both reported overseas hospitalisation in the 12 months prior to first detection of KPC—patient 22 in Vietnam for spinal surgery following a motor vehicle accident, and patient 23 in Greece for stem cell therapy. Both patients had undergone rectal screening on admission to Ward 1 in Facility A, with patient 22 being placed in intensive contact precautions for the duration of his hospital admissions after isolating another multidrug-resistant organism, though was required to use shared bathroom facilities with patients in the adjacent room. Having required treatment with meropenem for both hospital-acquired pneumonia and a surgical wound infection, patient 22 was later diagnosed with a KPC-producing K. pneumoniae indwelling catheter-associated urinary infection in January 2014. Twelve days later, patient 23 was subsequently found to have a polymicrobial sacral wound infection, with cultures including KPC-producing K. pneumoniae from sacral tissue. In response to this, all patients on the ward who had been admitted to the same room and/or shared bathroom facilities with patients 22 and 23 were screened, with the subsequent identification of patient 24. Alerts were placed on the records of patients meeting the criteria who had been previously discharged. Environmental screening of the rooms and bathrooms was conducted, with no KPC-producing organisms identified, and extended bleach cleaning with changes of curtains, chairs and other furnishings was conducted for the entire ward. Patient 33 was also admitted to Ward 1 in Facility A in February 2014, subsequent to the identification of KPC-2 in patients 22, 23 and 24. This patient was not screened as he had not been admitted to the same room, nor had he shared a bathroom with the identified cases. He also reported no recent history of overseas travel. However, he was identified in July 2014 through routine screening at Facility A following transfer from Ward 3, Facility U, located 25 km away. A KPC-producing isolate from patient 39 was identified in September 2014, and although the isolate genomically clustered with isolates from patients 22, 23, 24 and 33 identified at Facility A, she had no previous presentations to that healthcare facility. However, immediately prior to identification of KPC, she had also been in Facility U on Ward 3, though she was admitted there 13 days after patient 33’s discharge. Given this was the only plausible epidemiological link to the other cluster C patients, secondary transmission was presumed to have occurred in Ward 3, Facility U.

Figure 7. Bayesian evolutionary analysis indicates each of the phylogenetic clades corresponding to the genomic clusters emerged prior to the detection of KPC in Victoria.

A maximum clade credibility timed phylogeny from Bayesian evolutionary analysis of local CC258 K. pneumoniae isolate genomes are shown, with median node heights displayed. The thin red bars indicate 95% highest posterior density (HPD) intervals for the most recent common ancestor (MRCA) for major clades and defined genomic clusters (indicated on the right). The shaded grey region indicates the recent period when KPC isolates were detected in Victoria.

Figure 8. An inferred transmission tree shows that undetected colonisation was significant in propagating the outbreak.

Solid nodes represent the posterior mean time of KPC acquisition by individuals and are coloured by the corresponding genomic cluster, with empty circles representing inferred unsampled individuals contributing to the transmission tree. Branches are shaded by number of missing links in the transmission tree, with lighter branches representing increasing numbers of missing links implicated.

Figure 9. Klebsiella pneumoniae carbapenemase plasmids from the C. farmeri isolates were almost identical to KPC plasmids from K. pneumoniae outbreak isolates, despite differing replication proteins.

BLAST comparison between an IncFIB (pQIL-like) plasmid genome from a ST258 Klebsiella pneumoniae isolate, AUSMDU00008079 (above), and an IncR plasmid genome from a Citrobacter farmeri isolate, AUSMDU00008141 (below), from the outbreak. The grey shading indicates corresponding DNA regions of high nucleotide identity transcribed in opposing directions, with the Tn4401 transposon harbouring bla_KPC-2 highlighted in red. The plasmid genomes have been orientated to their respective replicons and downstream plasmid partitioning genes, parA and parB.

Figure 10. Genomes of isolates from the same host group together in the phylogeny.

Maximum likelihood tree of the study isolates with additional isolates obtained up to 1 November 2016 included. Multiple isolates from the same patient have been coloured by patient, with the accompanying graph indicating the collection dates for the corresponding isolates. Thirty-two isolates from six clinical samples obtained from patient 70 (blue) over eight months have been highlighted to illustrate the within-host lineages emerging in this patient. Six environmental isolates from the room of patient 75 are also shown (light pink). Bootstrap values (%) from 1,000 replicates have been displayed for major branches in the tree.