Genomic Analysis of Healthcare-associated Isolates

We sequenced 122 K. pneumoniae colonies (termed isolates) picked from primary culture plates from 23 stools/17 participants (median, 5 isolates per stool; range, 1–15). These were assigned to 21 STs (Figure 1B and Supplementary Table 1). Most patients (15/17) carried a single ST, with the 2 remaining cases each carrying 3 STs (Supplementary Table 1). No 2 patients carried the same ST, indicating an absence of patient-to-patient transmission. Pairwise analysis of SNPs in the core genome of isolates from the same patient/same ST demonstrated a median (range) difference of 1 (0–30) SNPs for the 17 patients.

We sequenced 24 isolates from 18 environmental swabs. These belonged to a more restricted population of 4 STs (Figure 1B and Supplementary Table 1). Integration of genetic and epidemiological data suggested 2 broad environmental contamination events. A cluster of 11 K. pneumoniae ST23 isolates were cultured from 2 adjacent single rooms and 1 more distant single room in ward A, all positive on a single date in April 2014. Pairwise core genome SNP analysis demonstrated a median (range) difference of 2 (range, 0–5) SNPs after removing 2 outliers (≥45 SNPs different). A second cluster of 8 K. pneumoniae ST307 isolates were cultured from 2 adjacent single rooms in ward A and 1 bedside in a 2-bedded bay in ward B, all positive on a single date in September 2015. The 8 isolates were identical at the core genome level. In addition, K. pneumoniae ST268 and ST3021 were each isolated once from different single rooms. The ST3021 isolate was cultured from the same room and day as an ST23 isolate, suggesting a wider contamination event and/or inadequate cleaning.

Comparison of isolates from stool and the environment showed that 2 STs (ST268 and ST307) were identified from both sources (Figure 2A). ST268 was isolated from a single stool from a patient (D034) and their environment in August 2015 and October 2015, respectively. Patient D304 had 17 admissions to the hematology day unit and 2 transfers to ward A during the period between the positive stool and environment sample. The most closely related stool–environmental ST268 isolate pair had highly similar core genomes (1 SNP different), which is consistent with recent patient–environment transmission. ST307 was isolated from stool in August 2015 from a patient (C029) in a single room on ward B and 18 days later from the environment of 2 rooms in ward A and 1 room in ward B (a different room from the index case). The most closely related stool and environmental ST307 isolate pair was 64 core genome SNPs different (see Figure 3A for phylogeny), which is not consistent with recent patient–environment transmission.

Open in a separate window

Figure 2.

Relatedness of Klebsiella pneumoniae isolated from different sources. A, Distribution of multilocus sequence types (STs) by source of isolation. B, Maximum-likelihood core genome phylogeny of K. pneumoniae from humans, the hospital environment, hospital sewage, livestock, and municipal wastewater from the East of England. The 3 clades are highlighted in orange (KpI), blue (KpII), and green (KpIII). The 2 right-hand columns (from left to right) show isolate source and those cases where STs were the same for both human and nonhuman isolates. Abbreviation: SNP, single-nucleotide polymorphism.

Open in a separate window

Figure 3.

Phylogeny of sequence type (ST) ST307 Klebsiella pneumoniae isolates and characterization of plasmids present in isolates from patients and their environment. A, Maximum likelihood tree of ST307 K. pneumoniae isolates from patient stool, blood, and the hospital environment based on single-nucleotide polymorphisms in the core genome after removal of recombination. B, Antimicrobial resistance genes and associated mobile elements in the pKPN3-307 Type B (GenBank KY271405.1) reference plasmid and in the human (GenBank MH745929) and environmental plasmids (GenBank MH745930). The shared 31 533-bp antimicrobial resistance region containing the bla_CTX-M-15 element is highlighted in gray. Arrows indicate the orientation of features, with the forward direction defined as the direction of transcription for genes, toward the main part of the attC site for cassettes, in integrons toward attI for 5’ flanking regions, away from the cassette array for 3’ flanking regions, relative to the direction of transcription of the transposase gene for insertion sequences and transposons (Tn) (ie, inverted repeat left to inverted repeat right) and to the direction of the reverse transcriptase for Group II introns. The missing end of a feature is shown by a zig-zag line. Abbreviation: SNP, single-nucleotide polymorphism.

We sequenced 16 K. pneumoniae isolates from 7 blood cultures/5 patients (1 isolate from each of the archived collection for 6 cultures/4 patients before or after the 6-month study and 10 primary plate colonies from the case that occurred during the study). These belonged to 5 STs, with 1 ST per patient. Four of the 5 STs were not identified in any other patient during the 6-month study, including the ST isolated from the patient who was bacteremic in this period. The exception was ST307, isolated from a patient (B024) with a bloodstream infection at 2 time points in September 2014 and December 2014 (at least 5 months before the prospective study began). Blood culture isolates from September and December differed by 4 SNPs, indicating relapse or reinfection with the same strain. Comparison between these and the prospective study isolates showed that the earlier blood culture isolates were related to the ST307 patient C029 stool isolates (13 SNPs; range, 12–15) and less closely to the ST307 environmental isolates (54 SNPs; range, 52–57; Figure 3A). This is indicative of either a persistent environmental reservoir of ST307 K. pneumoniae or ongoing transmission between unsampled patients. Sequencing and pairwise core genome analysis of 10 colonies from a single blood culture from 1 patient (B022, ST6) demonstrated a median (range) difference of 1 (0–2) SNP. This patient did not provide a stool sample.

Isolation of K. pneumoniae From Livestock, Retail Meat, and Sewage

The number of sampling points at each farm was maximized by taking pooled samples containing up to 50 aliquots of freshly passed fecal material from each major area in a given farm (eg, a pen). Culture of 136 pooled fecal samples from 29 livestock farms (Figure 4A) demonstrated that 6 (4%) samples from 4 (14%) farms (1 dairy cattle and 3 turkey) were positive for K. pneumoniae (Supplementary Table 1). Culture of 97 prepackaged fresh meat products from 11 countries purchased at 11 major Cambridge supermarkets were all negative for K. pneumoniae (Supplementary Table 2). Longitudinal sampling of sewage (4 samples over 16 months) at the study hospital (Figure 4A) resulted in K. pneumoniae being identified in 3/4 samples. A survey of 20 municipal wastewater treatment plants (Figure 4A) led to the recovery of K. pneumoniae from 17/40 water samples (11 untreated and 6 treated wastewater), taken from 12/20 treatment plants, with 5/12 plants releasing K. pneumoniae into the environment. Multiple colonies were selected from each positive sample for sequencing.

Open in a separate window

Figure 4.

Sampling locations and phenotypic antimicrobial susceptibility across the Klebsiella pneumoniae phylogeny. A, Map of the East Anglia region of the United Kingdom, showing the locations of the farms (images indicate the animal species), wastewater treatment plants (water drops), and hospitals in the region (indicated by a white “H” in a blue square). The hospital where the clinical and environmental isolates were collected is shown with H surrounded by a square. Adapted from Gouliouris et al [29]. Copyright 2018 Gouliouris et al. Adapted with permission. B, Maximum likelihood core genome phylogeny of 249 K. pneumoniae from human stool, blood, livestock, municipal wastewater, and hospital sewage (left) and their phenotypic antimicrobial susceptibility (right). The first vertical column shows isolate source and the remainder, the susceptibility to 19 antimicrobial drugs. Abbreviation: SNP, single-nucleotide polymorphism.

Genome-based Comparison of K. pneumoniae From the Hematology Ward and Elsewhere

We sequenced 87 K. pneumoniae isolates from livestock (32 isolates), municipal wastewater treatment plants (28 isolates), and hospital sewage (27 isolates). From this we identified 4 STs from livestock isolates, 24 STs from municipal wastewater, and 6 STs from hospital sewage (Figure 2A). For the livestock isolates, 2 STs (ST1609 and ST661) were identified from a single cattle farm, and 1 ST (ST3010) was isolated from all 3 turkey farms together with 1 colony of its single locus variant (SLV) from 1 of these (ST2585). All isolates from 3 turkey farms (including the SLV) were a median (range) of 12 core genome SNPs (range, 0–17) different, suggesting linkage between farms. Of these, 2 farms were located approximately 24 miles apart and the third was approximately 82 miles and 66 miles from the 2 other farms, respectively. All 3 farms were owned by the same company, and it is possible that transmission may have occurred as a result, although the route of transmission is unknown. STs were compared between isolates from the patient cohort, their ward environment, livestock, and municipal and hospital waste (Figure 2A). There was no overlap in STs from the different sources with 2 exceptions (ST661 from patient stool and livestock, and ST20 from patient stool and municipal wastewater). The 2 ST611 isolates differed by 2641 core genome SNPs, and the 6 ST20 isolates differed by a median (range) of 849 (848–851) SNPs. This indicates that isolates from different sources that belonged to the same ST were not closely related.

A maximum-likelihood phylogenetic tree of the 249 study isolates (hematology ward [n = 162], livestock/sewage/wastewater [n = 87]) was constructed based on 499 378 core gene SNPs (Figure 2B). This demonstrated high genetic diversity and 3 distinct populations (208 KpI [K. pneumoniae], 22 KpII [Klebsiella quasipneumoniae], and 19 KpIII [Klebsiella variicola]) isolates) (Figure 2B), consistent with previous descriptions [1]. Each clade contained isolates from human stool and wastewater. Isolates from the hospital environment and blood were confined to KpI and KpIII, while isolates from cattle and turkeys resided in different clades (turkey in KpI, cattle in KpII).

Antibiotic Resistance in K. pneumoniae From the Hematology Ward and Elsewhere

Phenotypic antibiotic susceptibility of the 249 isolates to 19 antibiotics is summarized in Figure 4B. Resistance to meropenem and ertapenem was detected for 2 isolates from hospital sewage, both of which contained bla_KPC-2. No colistin resistance genes (MCR-1, MCR-1.2, MCR-2, MCR-3, MCR-4, MCR-5) were detected in any of the isolates. Colistin resistance in K. pneumoniae can also occur through modification of lipopolysaccharides that arise from mutations in the pmrA/pmrB and phoP/phoQ two-component regulatory systems. We identified the L96P mutation in phoQ associated previously with colistin resistance [27] in 3 isolates recovered from a single wastewater treatment plant (accession numbers ERR985164, ERR985165, ERR985166). This indicates the importance of wastewater treatment plants as a source of resistance variants. Almost half of all isolates had an ESBL phenotype (118/249, 47%). Nearly all ESBL K. pneumoniae (n = 113) carried bla_CTX-M-15, the remainder carrying bla_CTX-M-1 (n = 3) or bla_KPC-2 and bla_SHV-12 (n = 2). Isolates positive for bla_CTX-M-15 were distributed across 12 STs and in all reservoirs tested, while bla_CTX-M-1 and bla_KPC-2 were reservoir specific and restricted to ST277 and ST258, respectively (Supplementary Table 1).

Plasmid analysis for the 113 bla_CTX-M-15 positive isolates was initially performed based on comparison between contigs containing the gene and the National Center for Biotechnology Information (NCBI) nucleotide database. This demonstrated close similarity to several reference plasmids, against which short-read data for the 113 isolates were then mapped. All isolates from turkey farms contained a bla_CTX-M-15 plasmid with high sequence similarity (ID >99%, coverage >99%) to pKpN01-CTX [28]. Wastewater isolates that belonged to ST902 (municipal plant) and ST1436 (hospital sewage) showed high sequence match (ID >99%, coverage >98%) to pKPN3-307 type A plasmid [30]. All ST268 isolates from stool and the ward environment mapped (ID >99%, coverage >99%) to the plasmid pCTXM15 (GenBank accession CP016925.1), consistent with these being associated with patient–environmental transmission. By contrast, while ST307 isolates from blood and stool mapped to IncFIB(K) pKPN3-307 type B (GenBank KY271405.1; ID >97%, coverage >99%), ST307 isolates from the hospital environment were not positive for this plasmid and no plasmid reference was identified in the NCBI nucleotide database. This is consistent with these not being associated with patient–environment transmission. Long-read sequencing of 1 bloodstream and 1 environmental bla_CTX-M-15 ST307 isolate showed that both contained an IncFIB(K) plasmid (GenBank MH745929 and MH745930, respectively). The environmental plasmid had only 78% coverage (ID >99%) against the human plasmid; however, this shared sequence included a 31 533-bp antimicrobial resistance region containing the bla_CTX-M-15 element (Figure 3B and Supplementary Table 3). Mapping of all ST307 bla_CTX-M-15 isolates to these 2 plasmids showed that clinical isolates contained MH745929 and the environmental isolates contained MH745930 (Figure 3A).

All bla_CTX-M-15 plasmids encoded a 13.4-kb region containing multiple antimicrobial resistance genes including bla_CTX-M-15, bla_TEM, strA, strB, and sul2 (see Figure 3B and AMR_Cluster1 in Supplementary Table 4). In total, 88/113 bla_CTX-M-15 positive isolates from humans, the hospital environment, livestock, and wastewater had high coverage mapping (>99%) and sequence ID (>99%) to this cluster, which shared synteny with an antimicrobial resistance region found previously in other K. pneumoniae plasmids [30]. In addition, 17 isolates from the hospital sewer and humans shared a 7.2-kb resistance cluster (see AMR_Cluster2 in Supplementary Table 3), containing bla_CTX-M-15 and aph(3) (encoding aminoglycoside resistance).

Analysis of bla_CTX-M-1 and bla_KPC-2 positive isolates demonstrated that the 3 bla_CTX-M-1 isolates (ST277, from wastewater) mapped with high similarity to the p369 plasmid (GenBank accession KT779550; ID >99%, coverage >99%) [31], and the 2 carbapenemase-producing K. pneumoniae (bla_KPC-2) isolates (ST258) from hospital sewage showed >99% mapping coverage and >99% sequence ID to the whole sequence of KPC reference plasmid pKpQIL-UK [32].