Genomically Informed Surveillance for Carbapenem-Resistant Enterobacteriaceae in a Health Care System.

UNLABELLED: Carbapenem-resistant Enterobacteriaceae (CRE) are an urgent public health concern. Rapid identification of the resistance genes, their mobilization capacity, and strains carrying them is essential to direct hospital resources to prevent spread and improve patient outcomes. Whole-genome sequencing allows refined tracking of both chromosomal traits and associated mobile genetic elements that harbor resistance genes. To enhance surveillance of CREs, clinical isolates with phenotypic resistance to carbapenem antibiotics underwent whole-genome sequencing. Analysis of 41 isolates of Klebsiella pneumoniae and Enterobacter cloacae, collected over a 3-year period, identified K. pneumoniae carbapenemase (KPC) genes encoding KPC-2, -3, and -4 and OXA-48 carbapenemases. All occurred within transposons, including multiple Tn4401 transposon isoforms, embedded within more than 10 distinct plasmids representing incompatibility (Inc) groups IncR, -N, -A/C, -H, and -X. Using short-read sequencing, draft maps were generated of new KPC-carrying vectors, several of which were derivatives of the IncN plasmid pBK31551. Two strains also had Tn4401 chromosomal insertions. Integrated analyses of plasmid profiles and chromosomal single-nucleotide polymorphism (SNP) profiles refined the strain patterns and provided a baseline hospital mobilome to facilitate analysis of new isolates. When incorporated with patient epidemiological data, the findings identified limited outbreaks against a broader 3-year period of sporadic external entry of many different strains and resistance vectors into the hospital. These findings highlight the utility of genomic analyses in internal and external surveillance efforts to stem the transmission of drug-resistant strains within and across health care institutions. IMPORTANCE: We demonstrate how detection of resistance genes within mobile elements and resistance-carrying strains furthers active surveillance efforts for drug resistance. Whole-genome sequencing is increasingly available in hospital laboratories and provides a powerful and nuanced means to define the local landscape of drug resistance. In this study, isolates of Klebsiella pneumoniae and Enterobacter cloacae with resistance to carbapenem antibiotics were sequenced. Multiple carbapenemase genes were identified that resided in distinct transposons and plasmids. This mobilome, or population of mobile elements capable of mobilizing drug resistance, further highlighted the degree of strain heterogeneity while providing a detailed timeline of carbapenemase entry into the hospital over a 3-year period. These surveillance efforts support effective targeting of infection control resources and the development of institution-specific repositories of resistance genes and the mobile elements that carry them.

INTRODUCTION

Carbapenem-resistant Enterobacteriaceae (CRE) are an urgent problem since they cause infections with high morbidity and mortality and lengthen hospital stays (1, 2). Klebsiella pneumoniae carbapenemases (KPCs) most commonly confer carbapenem resistance among members of the Enterobacteriaceae. These class A serine beta-lactamases were first detected in 1996 in North Carolina and have subsequently spread worldwide (3). Among the 22 characterized variants (http://www.lahey.org/studies), KPC-2 and −3 occur most commonly (4), along with less common carbapenemases that include OXA-48, NDM, and VIM (5). Phenotypic resistance to carbapenem antibiotics can also occur with noncarbapenemase beta-lactamases, such as AmpC, and extended-spectrum beta-lactamases (ESBLs) when they occur with chromosomal porin defects and the AmpD and AmpR regulatory proteins (6, 7).

KPC genes commonly occur on Tn4401, a 10-kb Tn3 family element flanked by 38-bp inverted repeats and containing two interrupting insertion sequences (IS), ISKpn6 and ISKpn7 (8). Tn4401 has 5 isoforms, denoted as a to e, that are differentiated by deletions upstream from the KPC gene (8). Transposition generates identical 5-bp terminal direct repeats (TDR) at the insertion site (8). Tn4401 has been identified in many plasmids and in chromosomal insertions (8–10). Other carbapenemases are associated with different transposons, such as OXA-48 gene carriage by the Tn1999 transposon (11).

When present on mobile elements, carbapenem resistance may spread clonally and through inter- and intraspecies lateral gene transfer. Each mechanism presents different epidemiological risks within health care systems. Integrated analyses of chromosomal, plasmid, and transposable elements within strains can better support hospital surveillance by identifying resistance-carrying strains and the risks for intra- and interspecies transfer of mobile elements harboring resistance determinants over time (10, 12–14).

We used whole-genome sequencing of carbapenem-resistant clinical isolates of Klebsiella pneumoniae and Enterobacter cloacae to establish a hospital-specific database of resistance determinants, their genomic context, and time points of entry within the health care system. The resulting data set provided a locally informed genomic landscape of resistance genes and carrying vectors, used to support ongoing analyses and infection control efforts.

RESULTS

Carbapenem resistance in hospital isolates of K. pneumoniae and E. cloacae.

Phenotypic susceptibility testing identified unique populations of putative CREs and raised concerns regarding clonal populations within the hospital (Fig. 1). Among all K. pneumoniae isolates, 1.5% demonstrated resistance to ertapenem and imipenem and/or meropenem, while an additional 0.5% of isolates were resistant to ertapenem but susceptible to both imipenem and meropenem and 98% of strains were carbapenem susceptible (Fig. 1A). In contrast, 4.0% of Enterobacter cloacae isolates demonstrated pancarbapenem resistance, with 15.8% showing resistance to ertapenem but susceptibility to imipenem and meropenem (Fig. 1B). A cohort of 41 clinical isolates of K. pneumoniae and E. cloacae (see Table S1 in the supplemental material) underwent whole-genome sequencing to identify resistance determinants, their genomic contexts, and strain patterns over time.

FIG 1 

Carbapenem genotype-phenotype correlations of K. pneumoniae and E. cloacae. Percentages of phenotypic resistance to ertapenem and imipenem/meropenem in K. pneumoniae (A) and E. cloacae (B) isolates for all isolates cultured from 2011 to 2015. (C and D) Carbapenemases and ESBLs identified in sequenced isolates of K. pneumoniae (C) and E. cloacae (D). Asterisks indicate strains where the beta-lactamases also included a chromosomally encoded AmpC.

Genomic context of carbapenemases and other beta-lactamases.

Among the isolates sequenced, 17 strains had a carbapenemase gene that was detectable by genome sequencing (see Table S2 in the supplemental material). Targeted PCRs for KPC and OXA-48 genes confirmed these results (data not shown). Carbapenemase-producing strains also carried between 2 and 5 additional beta-lactamases, including CTX-M-15, as well as SHV, TEM, and OXA family enzymes. KPC-2 and KPC-3 occurred most commonly in K. pneumoniae strains, though KPC-4 and OXA-48 were also detected (Fig. 1C). In contrast, E. cloacae isolates with carbapenemases uniformly carried KPC-4 with 2 to 3 additional beta-lactamases, including a chromosomally encoded AmpC family enzyme (ACT and MIR family), as well as mobile OXA and TEM family enzymes (Fig. 1D; see also Table S2 in the supplemental material).

Among strains with phenotypic carbapenem resistance that did not harbor a detectable carbapenemase gene, 0 to 4 other beta-lactamase genes were identified per strain (Fig. 1C and D; see also Table S3 in the supplemental material). Carbapenemase gene-negative isolates of K. pneumoniae commonly carried chromosomal copies of SHV and Len family narrow-spectrum beta-lactamase genes. Four strains (BWH-NC5, -NC6, -NC7, and -NC36) (“BWH” in the strain designations denotes Brigham and Women’s Hospital, and “NC” denotes non-carbapenemase-carrying strains) also carried the epidemic ESBL CTX-M-15, along with OXA-1 (BWH-NC6) or OXA-1 with TEM-1 (BWH-NC7 and BWH-NC36). Of the E. cloacae non-carbapenemase-producing strains, all harbored chromosomally encoded AmpC family beta-lactamases, while BWH-NC28 and BWH-NC17 also carried (respectively) mobile-element-encoded CTX-M-15 and TEM-116 extended-spectrum beta-lactamases in addition to the chromosomal AmpC (Fig. 1D; see also Table S3). Interestingly, BWH-NC18, a highly resistant strain isolated 1 month previously from the same patient as BWH-NC17, did not have a detectable TEM-116 gene.

Phenotype-genotype concordance among carbapenemase-producing K. pneumoniae and E. cloacae isolates.

All isolates harboring a KPC carbapenemase gene demonstrated phenotypic resistance to ertapenem, imipenem, and meropenem (Fig. 1; see also Table S2 in the supplemental material). However, panresistance also occurred through other mechanisms, highlighted by the finding that 32.4% of meropenem-resistant strains, 39.5% of imipenem-resistant strains, and 52.3% of ertapenem-resistant strains did not carry a carbapenemase gene that was detectable by sequencing or targeted PCR (data not shown). Genome sequencing in these strains identified non-KPC etiologies, including other beta-lactamase genes in conjunction with disruptions in the genes encoding porins OmpC (OmpK36) and OmpF (OmpK35) or mutations in the gene encoding the AmpC-regulator AmpD (Fig. 1C and D; see also Tables S3 and S4) (6, 7, 15–19).

Transposon carriage of carbapenemase genes.

All KPC genes occurred in the context of Tn4401 transposons (Table 1). Among these, Tn4401a and Tn4401e isoforms were found to be carrying KPC-2 genes and the Tn4401b isoform was found to be carrying KPC-3 and KPC-4 genes. While the majority of KPC strains harbored only a single copy of Tn4401, transposon insertion site analyses identified two Tn4401 copies in BWH-C6 (carbapenemase carriage is denoted by “C” in the strain designation), one chromosomal and one plasmid borne, highlighting the capacity for the transposon cassette to mobilize within carrying strains.

TABLE 1 

Genomic context and transposon or plasmid identities in carbapenemase gene-carrying isolates

Strain	Speciesa	BLA(s)b	Transposon(s)	Plasmid Inc group(s) or genomic location	Plasmid or MEc	Closest reference plasmid	% identity
BWH-C1	KP	KPC-2	Tn4401a	IncR	pKPC-484	pKPC-484	>99
BWH-C2	KP	KPC-2	Tn4401a	IncR	pKPC-484	pKPC-484	>99
BWH-C3	KP	KPC-4	Tn4401b (interrupted by IS110)	IncN	pBWH-C3-KPC	pBK31551	75
BWH-C4	KP	OXA-48	Tn1999	IncL/M	E71T	E71T	>99
BWH-C5	KP	KPC-3	Tn4401b	Chromosomal
BWH-C6	KP	KPC-3	Tn4401b	Chromosomal
	KP	KPC-3	Tn4401b	IncI2	pBK15692	pBK15692	98
BWH-C7	KP	KPC-2	Tn4401e	IncA/C2	pBWH-C7-KPC	PR55	97
BWH-C8	KP	KPC-3	Tn4401b	Untypeable	p34399-43.500kb	p34399-43.500kb	98
BWH-C9	KP	KPC-3	Tn4401b	IncX3	pBWH-C9-KPC	p34618-43.380kb	>99
BWH-C10	KP	KPC-3	Tn4401b	Untypeable	p34399-43.500kb	p34399-43.500kb	98
BWH-C11	EC	TEM-1, OXA-1, KPC-4	Tn4401b (interrupted by Tn6901)	Unresolvable plasmid	ME-BWH-C11-KPC	pBK31551	90
BWH-C12	EC	TEM-1, OXA-1, KPC-4	Tn4401b (interrupted by Tn6901)	Unresolvable plasmid	ME-BWH-C11-KPC	pBK31551	90
BWH-C13	EC	KPC-4	Tn4401b	IncH12A/IncH12	pBWH-C13-KPC	pK29	91
BWH-C14	EC	KPC-4	Tn4401b	IncH12A/IncH12	pBWH-C13-KPC	pK29	91
BWH-C15	EC	KPC-4	Tn4401b	IncH12A/IncH12	pBWH-C13-KPC	pK29	91
BWH-C16	EC	KPC-4	Tn4401b	IncN	pBWH-C16-KPC	pBK31551	85
BWH-C17	EC	KPC-4	Tn4401b	IncN	pBWH-C16-KPC	pBK31551	85

KP, Klebsiella pneumoniae; EC, Enterobacter cloacae.

BLA, beta-lactamase; KPC, Klebsiella pneumoniae carbapenemase.

ME, mobile element.

In addition to KPC carbapenemase genes in the context of Tn4401, one instance of the OXA-48 carbapenemase gene was also found, on Tn1999 inserted in an incompatibility (Inc) group IncL/M plasmid in K. pneumoniae strain BWH-C4 (20, 21).

Klebsiella plasmids.

Carbapenemase gene-carrying transposons were borne by members of Inc group IncR, -N, -L/M, -I, and -X plasmids in Klebsiella (Table 1). While repetitive sequences can make plasmid assembly with short reads challenging, improved finished plasmid references in GenBank supported the creation of draft plasmid maps for most strains. These analyses identified several known KPC-carrying vectors, including pKPC-484 (strains BWH-C1 and -C2) (10), E71T (strain BWH-C4) (21), pBK15692 (strain BWH-C6) (22), and p34399-43.500kb (strains BWH-C8 and -C10) (GenBank accession number CP010387.1).

The analyses also identified vectors without close references. Strain BWH-C3’s plasmid, pBWH-C3-KPC, carried a KPC-4 gene (see Fig. S1a in the supplemental material). The plasmid backbone demonstrated only 75% identity to pBK31551, a plasmid that had Tn4401 interrupted by IS110 and Tn6901 elements (Fig. 2a) (23). However, unlike pBK31551, pBWH-C3-KPC lacks these interrupting elements (Fig. 2b). pBWH-C3-KPC also lacked the class I integron, an ISCRI element located upstream from Tn4401, and a region of plasmid transfer and stability machinery (Fip, Tra, and Stb genes) (Fig. 2b).

FIG 2 

pBK31551-derived mobile elements. (A) The pBK31551 sequence is shown, with open reading frames (ORFs) in blue, KPC gene-carrying Tn4401 transposons in green, other insertion sequences and transposons in orange, and class I integron sequences in grey. (B to D) Red boxes indicate homologous regions of pBK31551. (B) pBWH-C3-KPC lacks regions that correspond to class I integron resistance, Tra/Stb factors, and the IS110 and Tn6901 mobile elements within Tn4401. (C) pBWH-C16-KPC lacks the same regions except for Tra/Stb (grey box). (D) ME-BWH-C11-KPC covers an ~30-kb region of pBK31551 that includes Tn4401 (interrupted by Tn6901 [orange box]). This construct does not include the IS110 insertion found in the parent pBK31551 construct (yellow box in panel A).

Other plasmid vectors, previously described as non-KPC plasmids, were found with Tn4401 insertions. Strain BWH-C7 (KPC-2) carried an IncA/C plasmid with 97% sequence identity to PR55 (see Fig. S1b in the supplemental material), first identified in a clinical isolate of K. pneumoniae from France in 1969 (24). In this plasmid, designated pBWH-C7-KPC, Tn4401e inserted between the ter and kfrA genes (see Fig. S1b). In plasmid pBWH-C9-KPC, carried in strain BWHC9, Tn4401b inserted into an oxidoreductase gene in an IncX plasmid with >99% identity to p34618-43.380kb (unpublished data; GenBank accession number CP010395.1) (see Fig. S1c). This construct bears a close relationship to pKPSn90, a KPC-3-bearing IncX plasmid with Tn4401 inserted at a different location than in pBWH-C9-KPC (25).

While the majority of strains carried Tn4401 in a plasmid backbone, K. pneumoniae strains BWH-C5 and BWH-C6 carried chromosomal insertions. BWH-C6 also harbored a plasmid copy of Tn4401 in pBK15692. Interestingly, strain BWH-C5 also carried plasmid pBK15692, though Tn4401 appears to have been excised, highly suggestive of a transposon jump from pBK15692 to the chromosome in these strains (see Fig. S2a and b in the supplemental material).

Enterobacter plasmids.

KPC carriage has been less well characterized in Enterobacter species than in Klebsiella species. In this cohort, carbapenemase-bearing strains of E. cloacae contained a single, plasmid-borne copy of KPC-4 within Tn4401b. Mobile element analyses of these strains further identified three subgroups.

Group 1 strains carried mobile element ME-C11-KPC, which harbors Tn4401::blaKPC-4 and TEM-1 and OXA-1 genes in an approximately 28-kb segment with 100% identity to the IncN plasmid pBK31551, originally detected in K. pneumoniae (Fig. 2d). Of the two mobile elements that disrupt the Tn4401 insertion in pBK31551, IS1618 (IS110 family) and Tn6901, only Tn6901 is present. However, the group 1 Enterobacter cloacae strains lack the IS110 insertion but have the Tn6901 insertion. This strain carries two other Enterobacter plasmids with >99% identity to p35374-141.404kb and p34399-106.698kb, neither of which has been described as carrying a KPC or other beta-lactamase gene. With short-read analyses, it was not possible to link ME-BWH-C11-KPC into a larger plasmid backbone; however, the raw-read coverage was approximately 2× that of the plasmids, suggesting that the segment occurred as a duplication that collapsed into a chimeric contig, as described by Conlan et al. (10).

Group II strains, whose mobile elements are represented by pBWH-C13-KPC (see Fig. S1d in the supplemental material) carried a KPC-4 gene on Tn4401b embedded in a plasmid with 91% identity to the IncHI2 plasmid, pK29 (26). This construct was originally isolated from K. pneumoniae in Taiwan and shown to be negative for a KPC gene but carried the AmpC beta-lactamase CMY-8 gene and the ESBL CTX-M-3 gene, neither of which occurred in the backbone identified in plasmid pBWH-C13-KPC, though an OXA-129 gene was identified.

Group III strains, whose mobile elements are represented by pBWH-C16-KPC (Fig. 2C; see also Fig. S1e in the supplemental material), carried a KPC-4 gene on Tn4401b, along with TEM-1 on an IncN plasmid with 85% identity to pBK31551. This construct is quite similar to pBWH-C3-KPC (Fig. 2) in that the transposon is not disrupted by either of the mobile elements found in pBK31551. It additionally lacks the same regions of the upstream class I integron. However, unlike pBWH-C3-KPC, this construct contains the Tra/Stb-encoding region found downstream from the KPC-encoding region (Fig. 2C).

Epidemiology of carbapenemase elements over time.

To evaluate similarities among carbapenemase-carrying strains, distance trees based on core chromosomal single-nucleotide polymorphisms (SNPs) were compared with the strains’ multilocus sequence types (MLSTs), carbapenemase genes, and transposon and plasmid carriage profiles (Fig. 3 and 4).

FIG 3 

Maximum-likelihood phylogenetic tree of K. pneumoniae clinical isolates. SNPs in the core chromosome (excluding mobile elements) were calculated using strain KP-MGH-78578 as a reference (indicated by an asterisk). Isolates taken from the same patient (during separate inpatient stays) are denoted by double asterisks. Corresponding MLSTs, KPC variants, and carbapenemase gene-bearing plasmids (if present) are indicated. Strains with similar KPC gene-bearing constructs, i.e., chromosomal/pBK15692 (green), pKPC-484 (blue), and p34399-43.500kb (red), are indicated. Scale bar indicates a distance of 5,000 SNPs. Local support values are indicated at the nodes.

FIG 4 

Maximum-likelihood phylogenetic tree of E. cloacae clinical isolates. SNPs in the core chromosome (excluding mobile elements) were calculated using strain KP-ENHKUO1 as a reference (indicated by an asterisk). Isolates taken from the same patient (during separate inpatient stays) are denoted by double asterisks. Corresponding MLSTs, KPC variants, and carbapenemase gene-bearing plasmids (if present) are indicated. Strains with similar KPC-bearing constructs, i.e., pBWH-C13-KPC (green), pBWH-C16-KPC (blue), and ME-BWH-C11-KPC (red) are indicated. Scale bar indicates a distance of 2,000 SNPs. Local support values are indicated at the nodes.

For K. pneumoniae strains, the major branches corresponded to a variety of MLSTs (Fig. 3), including ST258 (8 strains), the major KPC-bearing clone in the United States (8). Additional MLST types included ST15 (n = 2), ST38, ST113, ST16, ST29, ST101, ST1562, and ST1393 (n = 1 each) and several strains with an unknown ST (n = 4). A single pancarbapenem-resistant ST258 strain, BWH-NC12, did not have a carbapenemase gene that was detectable by sequencing or targeted PCR but demonstrated disrupted OmpK35 and OmpK36 porin genes, likely contributing to its highly resistant phenotype (see Table S4 in the supplemental material).

The numbers of SNPs separating K. pneumoniae isolates in this study ranged widely, from 8 to 35,520 (Table 2). Among the ST258 isolates, strains differed by 35 to 612 SNPs. Two sets of KPC strains that carried identical plasmids and were cultured from different patients, BWH-C1 and -C2 and BWH-C8 and -C10, differed by 106 and 35 SNPs, respectively. Strains BWH-C5 and BWH-C6, which harbored chromosomal Tn4401 insertions in the same location, differed by 171 SNPs, suggesting that they are related but not immediately so. Furthermore, BWH-C6 also harbored a second copy of Tn4401 in pBK15692, suggesting an additional transposition event in an ancestor in common with BWH-C5.

TABLE 2 

SNP matrix for Klebsiella pneumoniae isolates

Strain	No. of SNPs that differ between indicated strains
	BWH-C1	BWH-C2	BWH-C3	BWH-C4	BWH-C5	BWH-C6	BWH-C7	BWH-C8	BWH-C9	BWH-C10	BWH-NC1	BWH-NC2	BWH-NC5	BWH-NC6	BWH-NC7	BWH-NC9	BWH-NC10	BWH-NC12	BWH-NC13	BWH-NC36
BWH-C1	0	106	16,131	17,219	539	400	349	612	16,533	601	32,692	17,009	34,958	15,714	17,001	16,961	16,359	373	16,671	16,999
BWH-C2	106	0	16,126	17,221	481	344	293	556	16,533	549	32,713	17,008	34,977	15,698	16,998	16,961	16,358	321	16,673	16,996
BWH-C3	16,131	16,126	0	17,033	16,127	16,094	16,093	16,064	16,311	16,047	32,209	16,632	34,566	16,913	16,915	16,813	16,388	16,135	16,245	15,814
BWH-C4	17,219	17,221	17,033	0	17,248	17,202	17,207	17,176	16,659	17,162	33,115	16,965	35,520	16,860	16,290	16,770	16,937	17,241	17,094	16,288
BWH-C5	539	481	16,127	17,248	0	171	248	473	16,565	454	32,668	17,010	34,962	15,710	17,024	16,956	16,370	300	16,651	17,022
BWH-C6	400	344	16,094	17,202	171	0	113	338	16,509	333	32,662	16,956	34,964	15,666	16,984	16,924	16,324	161	16,619	16,982
BWH-C7	349	293	16,093	17,207	248	113	0	289	16,496	284	32,687	16,951	34,987	15,665	16,998	16,937	16,326	108	16,618	16,996
BWH-C8	612	556	16,064	17,176	473	338	289	0	16,484	35	32,553	16,946	34,845	15,638	16,957	16,887	16,257	345	16,603	16,955
BWH-C9	16,533	16,533	16,311	16,659	16,565	16,509	16,496	16,484	0	16,485	32,741	16,719	35,206	16,195	16,655	16,699	15,817	16,530	14,805	16,653
BWH-C10	601	549	16,047	17,162	454	333	284	35	16,485	0	32,543	16,931	34,834	15,623	16,956	16,872	16,240	336	16,590	16,954
BWH-NC1	32,692	32,713	32,209	33,115	32,668	32,662	32,687	32,553	32,741	32,543	0	33,169	12,703	32,160	32,978	32,915	32,507	32,719	32,945	32,976
BWH-NC2	17,009	17,008	16,632	16,965	17,010	16,956	16,951	16,946	16,719	16,931	33,169	0	35,586	16,830	16,460	16,793	16,740	16,983	17,063	16,458
BWH-NC5	34,958	34,977	34,566	35,520	34,962	34,964	34,987	34,845	35,206	34,834	12,703	35,586	0	34,488	35,348	35,331	34,808	35,011	35,228	35,346
BWH-NC6	15,714	15,698	16,913	16,860	15,710	15,666	15,665	15,638	16,195	15,623	32,160	16,830	34,488	0	16,845	16,729	16,112	15,697	16,457	16,843
BWH-NC7	17,001	16,998	16,915	16,290	17,024	16,984	16,998	16,957	16,655	16,956	32,978	16,460	35,348	16,845	0	16,434	16,764	17,022	16,692	8
BWH-NC9	16,961	16,961	16,813	16,770	16,956	16,924	16,937	16,887	16,699	16,872	32,915	16,793	35,331	16,729	16,434	0	16,716	16,973	16,866	16,432
BWH-NC10	16,359	16,358	16,388	16,937	16,370	16,324	16,326	16,257	15,817	16,240	32,507	16,740	34,808	16,112	16,764	16,716	0	16,359	16,297	16,762
BWH-NC12	373	321	16,135	17,241	300	161	108	345	16,530	336	32,719	16,983	35,011	15,697	17,022	16,973	16,359	0	16,656	17,020
BWH-NC13	16,671	16,673	16,245	17,094	16,651	16,619	16,618	16,603	14,805	16,590	32,945	17,063	35,228	16,457	16,692	16,866	16,297	16,656	0	16,690
BWH-NC36	16,999	16,996	15,814	16,288	17,022	16,982	16,996	16,955	16,653	16,954	32,976	16,458	35,346	16,843	8	16,432	16,762	17,020	16,690	0

Clinical isolates of E. cloacae also belonged to a variety of MLSTs (Fig. 4), with the most common being ST78 (n = 6), in addition to ST171 (n = 3), ST190 (n = 3), ST252 (n = 1), ST162 (n = 1), ST133 (n = 1), and unknown MLST types (n = 6). Carbapenemase-harboring strains belonged to ST78 and ST171. Non-carbapenemase-carrying strains within ST78 included BWH-NC16 and BWH-NC28. The former was resistant to ertapenem but susceptible to imipenem and meropenem. This strain had a chromosomal AmpC gene, as well as a deletion in the C-terminal-domain region of the AmpD gene, which is associated with a carbapenem-resistant phenotype (16). BWH-NC28 was panresistant but was carbapenemase gene negative. It carried the CTX-M-15 ESBL gene in addition to the chromosomal AmpC gene, which in combination have been reported to cause phenotypic resistance to ertapenem (27). A disrupted porin (OmpF) likely also contributed to this isolate’s highly resistant phenotype (see Table S4 in the supplemental material).

The Enterobacter cloacae isolates demonstrated less overall SNP diversity than the Klebsiella strains, with a range of 0 to 14,550 SNPs among strains (Table 3). Isolates within the ST78 group of strains differed by 6 to 19 SNPs. Two sets of two strains carried a KPC-4 gene on the same mobile element; these were strains BWH-C16 and -C17 (pBWH-C16-KPC) and strains BWH-C11 and -C12 (ME-BWH-C11-KPC). Chromosomal SNP analyses showed them to be separated by 9 and 6 SNPs, respectively, making them the most similar pairs within the ST78 group. Among the ST171 group, strains BWH-C13, -C14, and -C15 were virtually identical (0 to 1 SNPs) and carried a KPC-4 gene on mobile element pBWH-C14-KPC (Fig. 4).

TABLE 3 

SNP matrix for Enterobacter cloacae isolates

Strain	No. of SNPs that differ between indicated strains
	BWH-C11	BWH-C12	BWH-C13	BWH-C14	BWH-C15	BWH-C16	BWH-C17	BWH-NC14	BWH-NC15	BWH-NC16	BWH-NC17	BWH-NC18	BWH-NC19	BWH-NC20	BWH-NC21	BWH-NC22	BWH-NC24	BWH-NC25	BWH-NC26	BWH-NC28	BWH-NC37
BWH-C11	0	9	5885	5886	5886	11	11	13,987	5862	14	5799	5798	14,288	5818	5775	13,916	5823	5803	5798	14	14,289
BWH-C12	9	0	5885	5886	5886	16	16	13,991	5862	19	5802	5801	14,290	5822	5779	13,920	5827	5807	5801	19	14,291
BWH-C13	5885	5885	0	1	1	5884	5884	13,873	1625	5883	4621	4620	14,234	4562	4564	13,721	4615	4601	4622	5879	14,231
BWH-C14	5886	5886	1	0	0	5885	5885	13,872	1626	5884	4622	4621	14,235	4563	4565	13,720	4616	4602	4623	5880	14,232
BWH-C15	5886	5886	1	0	0	5885	5885	13,872	1626	5884	4622	4621	14,235	4563	4565	13,720	4616	4602	4623	5880	14,232
BWH-C16	11	16	5884	5885	5885	0	6	13,990	5865	13	5800	5799	14,292	5819	5774	13,919	5822	5804	5799	13	14,295
BWH-C17	11	16	5884	5885	5885	6	0	13,988	5865	13	5798	5797	14,290	5819	5774	13,917	5822	5802	5797	13	14,293
BWH-NC14	13,987	13,991	13,873	13,872	13,872	13,990	13,988	0	13,828	13,987	13,989	13,988	14,104	14,023	14,014	10,540	14,069	13,947	13,987	13,991	14,213
BWH-NC15	5862	5862	1625	1626	1626	5865	5865	13,828	0	5860	4548	4547	14,250	4506	4511	13,686	4568	4525	4548	5858	14,223
BWH-NC16	14	19	5883	5884	5884	13	13	13,987	5860	0	5799	5798	14,288	5822	5775	13,916	5823	5801	5798	14	14,289
BWH-NC17	5799	5802	4621	4622	4622	5800	5798	13,989	4548	5799	0	1	14,312	1735	1770	13,972	1813	1935	31	5797	14,291
BWH-NC18	5798	5801	4620	4621	4621	5799	5797	13,988	4547	5798	1	0	14,313	1736	1771	13,971	1814	1936	32	5796	14,292
BWH-NC19	14,288	14,290	14,234	14,235	14,235	14,292	14,290	14,104	14,250	14,288	14,312	14,313	0	14,298	14,329	14,234	14,380	14,257	14,308	14,288	5062
BWH-NC20	5818	5822	4562	4563	4563	5819	5819	14,023	4506	5822	1735	1736	14,298	0	1851	13,971	1895	1922	1738	5818	14,258
BWH-NC21	5775	5779	4564	4565	4565	5774	5774	14,014	4511	5775	1770	1771	14,329	1851	0	13,987	79	1984	1771	5771	14,289
BWH-NC22	13,916	13,920	13,721	13,720	13,720	13,919	13,917	10,540	13,686	13,916	13,972	13,971	14,234	13,971	13,987	0	14,037	13,898	13,970	13,916	14,361
BWH-NC24	5823	5827	4615	4616	4616	5822	5822	14,069	4568	5823	1813	1814	14,380	1895	79	14,037	0	2028	1814	5819	14,337
BWH-NC25	5803	5807	4601	4602	4602	5804	5802	13,947	4525	5801	1935	1936	14,257	1922	1984	13,898	2028	0	1934	5801	14,232
BWH-NC26	5798	5801	4622	4623	4623	5799	5797	13,987	4548	5798	31	32	14,308	1738	1771	13,970	1814	1934	0	5796	14,288
BWH-NC28	14	19	5879	5880	5880	13	13	13,991	5858	14	5797	5796	14,288	5818	5771	13,916	5819	5801	5796	0	14,291
BWH-NC37	14,289	14,291	14,231	14,232	14,232	14,295	14,293	14,213	14,223	14,289	14,291	14,292	5062	14,258	14,289	14,361	14,337	14,232	14,288	14,291	0

Epidemiological detection of strain entry and outbreaks.

Patient diagnoses, lengths of inpatient stays, locations within the hospital, and coisolates cultured over the 3-year period of analyses were collected for all patients (Fig. 5). The analyses identified limited outbreaks within the hospital but, more prominently, noted sporadic detection of strains with identical plasmids months to more than a year apart. In the latter cases, a common hospital-based reservoir could not be identified. The most closely linked isolates in this cohort, BWH-C13, -C14, and -C15, differed by 0 or 1 SNP and were collected within 18 days of each other from patients being treated by the same clinical service.

FIG 5 

Integrating genomics with patient metadata. All members of each MLST with more than one representative (including at least one carbapenemase producer) were associated with patient metadata (dates of inpatient stays, diagnoses, and providers). Admissions and discharges are denoted by grey circles linked by thick lines (not visible for short stays). Dates of first positive CRE cultures are indicated by red asterisks. Associated carbapenemase gene-bearing plasmids are indicated to the right. Isolates of K. pneumoniae ST258 (A) and E. cloacae ST171 and ST78 (B) are shown.

DISCUSSION

Several studies have shown the utility of clinical microbial genome sequencing to aid in outbreak detection and the tracking of virulence and resistance factors (28, 29). We used genome sequencing to identify the genetic determinants of carbapenem resistance and their context within mobile elements or chromosomal sequences. In this manner, a hospital-specific repository of resistance genes, transposons, and mobile elements enabled more refined and rapid analyses of new isolates as they occurred (see Fig. S3 in the supplemental material).

Among the K. pneumoniae isolates, phenotypic resistance to carbapenems was mediated by KPC-2, -3, and -4 and OXA-48, as well as other beta-lactamases, in conjunction with accessory proteins, such as porins. OXA-48 has only recently been reported in the United States (30). In contrast, carbapenemase-producing E. cloacae strains harbored only KPC-4, which has been reported uncommonly in other CRE surveys (10, 12, 14, 31, 32). Non-carbapenemase-carrying isolates of E. cloacae with phenotypic carbapenem resistance carried ESBL genes and/or mutated AmpD and porin genes.

Klebsiella KPC genes were carried on different isoforms of Tn4401, a, b, and e, which were further inserted into seven different plasmids and, in two instances, chromosomally. We identified several known KPC-encoding constructs, such as pKPC-484, pBK15692, p34399-43.500kbp, and E71T.

pKPC-484 (Tn4401a::blaKPC-2) (10) and p34399-43.500kb (Tn4401b::blaKPC-3) were each identified in two patients with no obvious epidemiological connections. The IncI plasmid pBK15692, first identified in a K. pneumoniae strain isolated in 2005 in a New Jersey hospital, carries a KPC-3 gene within Tn4401b that has inserted into Tn1331 (22). Strain BWH-C6, which has a chromosomal Tn4401b, carries the pBK15692 plasmid with an additional copy of Tn4401b and a small deletion in Tn1331, removing the aad and blaOXA-9 genes. Interestingly, strain BWH-C5, with an identical chromosomal Tn4401b, carries the pBK15692 backbone with Tn4401b apparently excised. The only OXA-48 carbapenemase gene identified in this study was carried by plasmid E71T, an IncL/M plasmid first described in Ireland (21).

Novel plasmids identified in Klebsiella isolates included pBWH-C3-KPC, which carried the only KPC-4 gene identified in Klebsiella and showed the closest identity (75%) to pBK31551, a KPC-4 gene-bearing IncN plasmid that was first detected in New Jersey (23). In addition, plasmid pBWH-C7-KPC carried a Tn4401e insertion in the backbone of the IncA/C plasmid PR55, along with a large bacteriophage (24).

Plasmid analyses further refined the groupings among the KPC-carrying Enterobacter strains, identifying three distinct groups with pBWH-C13-KPC, pBWH-C16-KPC, or ME-BWH-C11-KPC. The backbone of the KPC-4 gene-carrying plasmid pBWH-C13-KPC showed 91% identity to the IncH12A/IncH12 plasmid pK29 (26). This construct occurred in the most closely related strains (0 to 1 SNPs), isolated within 18 days of each other from patients receiving care from the same inpatient service but housed on different floors of the hospital.

Notably, the other two Enterobacter carbapenemase gene-bearing elements, (pBWH-C16-KPC and, ME-BWH-C11-KPC), along with Klebsiella plasmid pBWH-C3-KPC, share homology with the IncN plasmid pBK31551, illustrating its capacity to transmit KPC-4 across species. We suggest one possible scenario linking the pBK31551-related constructs across species (see Fig. S4 in the supplemental material), namely, hospital entry within a K. pneumoniae ST834 or ST113 strain and spread to E. cloacae ST78 strains at a time prior to the start of genomic surveillance in 2011. The pBWH-C3-KPC construct in Klebsiella and the ME-BWH-C11-KPC and pBWH-C16-KPC derivatives in Enterobacter were then detected the following year.

Plasmid analyses among the K. pneumoniae MLST types further refined strain relationships. In particular, the relatively low number of SNPs among isolates carrying identical plasmids suggests the spread of clonal strains with their plasmids over the 3-year period, rather than significant transfer of plasmids to other strains. In contrast to Klebsiella, relatively little is known about the prevalence of KPC carriage among Enterobacter MLSTs, though an MLST scheme has recently been described for E. cloacae (33–35). These studies illustrated the multiclonal nature of drug-resistant Enterobacter strains by finding KPC-4 carriage in ST78 and ST171.

These results demonstrate the importance of analyzing resistance-carrying transposons and plasmids to enrich epidemiological tracking of resistance determinants within and across institutions. Among CRE strains from the same ST type that shared chromosomal SNP profiles, plasmid analyses improved the subclassification of strains and their nature as sporadic or associated with a potential internal outbreak or reservoir. The transposon, plasmid, and chromosomal SNP profiles further enabled the development of a hospital-specific repository of chromosomes and mobile elements that could additionally contribute to national surveillance efforts, which will be needed to more effectively identify and track resistance determinants across institutions.

Our data also highlight the strengths and weaknesses of short-read sequencing platforms to analyze mobile genetic elements with highly repetitive sequences. Currently, long-read sequencing technology remains beyond the capacity of most clinical microbiology laboratories. Closed reference-quality plasmid sequences are rarely obtained with short-read sequencing alone. However, assembled contigs from short-read platforms can be used to readily identify specific resistance genes and their immediate genomic context. The growing plasmid reference content in public databases further enhances the capacity to generate clinically valuable draft plasmid maps using short-read platforms. As more clinical laboratories move to perform such analyses, it will be possible to undertake more robust analyses of strain and mobile element transmission of drug resistance by institution, region, and timeframe, as well as in environments external to health care systems.

MATERIALS AND METHODS

Bacterial strains and data.

All strains and data were collected under IRB protocol 2011-P-002883 approved by the Partners Healthcare Internal Review Board. Strains were collected from the Clinical Microbiology Laboratory at Brigham and Women’s Hospital (BWH), a 793-bed hospital in Boston, MA, that supports a number of inpatient and outpatient services. Standing queries in the Crimson LIMS (36) flagged resistance to ertapenem, meropenem, and/or imipenem among members of the Enterobacteriaceae identified during routine clinical microbiologic testing. Thirty-seven isolates of carbapenem-nonsusceptible Klebsiella pneumoniae and Enterobacter cloacae, along with four pansusceptible and ESBL strains, were collected from October 2011 to October 2014 and included in genomic analyses. Phenotypic resistance was determined by MIC (Vitek 2 platform), Kirby-Bauer disk diffusion, and/or E-test strips (bioMérieux, Crapone, France) (37). Total DNA was isolated from each strain on the Qiagen EZ1 platform using the tissue DNA isolation kit (Qiagen, Venlo, Netherlands). Phenotypic resistance at BWH was tracked and charted using WHOnet (38). Patient data were analyzed in the hospital electronic medical records (EMR).

KPC PCR.

KPC genes were amplified according to the method of Mathers et al. (39).

Library preparation and sequencing.

Libraries were prepared using the Nextera XT system (Illumina, San Diego, CA). Strains were sequenced on the MiSeq platform (Illumina), using the V1 (150-bp paired-end reads) or V3 (300-bp paired-end reads) kit. The average sequencing depth resulted in 103× coverage.

Contig assembly and analysis.

De novo assembly was performed using SPAdes (version 3.1) (40), and the resulting contigs were assessed with QUAST (41), which showed an average N50 of 273,646 bp across isolates. Resistance genes were identified by BLAST against a database of resistance genes compiled from the Comprehensive Antibiotic Resistance Database (CARD), the Lahey Clinic (http://www.lahey.org/studies), and the Lactamase Engineering database (http://www.laced.uni-stuttgart.de) (42). The criteria used to positively call specific classes of beta-lactamases were (i) coverage of >97% query length of the putative gene, (ii) >97% identity with the matching reference sequence, and (iii) <5 mismatches, as well as no gaps in the alignment. To determine chromosomal versus mobile genomic context of the beta-lactamases discovered, the surrounding sequence neighborhoods on the parent contigs that contained beta-lactamase gene(s) were compared to the GenBank nt database using BLAST to assess matches to reference plasmids or chromosomes.

Transposon carriage was assessed by using BLAST to compare the de novo contigs against a set of transposon sequences derived from GenBank. To assess transposon terminal direct repeats (TDRs), copy numbers, and insertion sites, Bowtie2 was used to align raw reads to transposon junctions, which were grouped by the 5-bp TDR generated by the transposon (43). Five-base-pair TDRs that had a matching TDR on both the 5′ and 3′ end of the transposon were considered to be the borders of a complete transposon. The nontransposon sequence adjoining the TDR was compared to the GenBank nt database using BLAST in order to confirm the insertion site.

Plasmid incompatibility (Inc) groups were assessed using BLAST and the PlasmidFinder database from the Center for Genomic Epidemiology (CGE) (44). To identify all plasmid-associated contigs in an isolate, a BLAST search was conducted against the NCBI bacterial plasmid database (http://www.ncbi.nlm.nih.gov). Contigs that demonstrated a likely plasmid origin were selected and then compared to the GenBank nt database using BLAST for a more accurate identification and to arrive at a list of candidate plasmids, which were classified according to the plasmid replicons. To assess the strength of plasmid identifications and analyze antibiotic resistance regions, raw reads were aligned to plasmid reference sequences using Bowtie (45). Plasmids with the highest coverage by raw reads, where at least 65% of the backbone was accounted for, were selected as the best matches. To ascertain whether isolates contained regions not present in the reference plasmid, de novo contigs generated in SPAdes were ordered to the sequence of the reference plasmid using the MAUVE aligner and novel regions identified (46). Plasmid maps were then generated from each set of ordered contigs, annotated using the RAST engine, and given a new plasmid name if they were <98% identical to the reference construct or included novel insertions not present in the reference sequence (i.e., Tn4401) (47, 48). Carbapenemase-containing elements which could not be placed in a plasmid backbone were denoted as “ME,” for mobile element. Maps were visualized with MacVector software (MacVector, Inc., Cary, NC).

Chromosomal analyses of resistance genes and modifying mutations.

Genes involved in the regulation of AmpC (ampD and ampR), as well as the porin genes ompC and ompF (ompK36 and ompK35 in Klebsiella strains), were assessed for premature stop codons, disruptive insertion sequences (IS), and nonsynonymous mutations known to affect enzyme function (7, 15–19).

MLST.

Multilocus sequence typing (MLST) was done by using the MLST finder tool at the Center for Genetic Epidemiology (49).

SNP typing.

Chromosomal single-nucleotide polymorphisms (SNPs) across sequenced isolates were called in de novo-assembled contigs using the Nucmer and showSNPs tools in the Mummer package (50). Calls were made in comparison to chromosomal sequences from reference strains K. pneumoniae KP_MGH_78578 (NC_009648) and E. cloacae ENHKU01 (NC_018405). Default settings were used for Nucmer, and the “-CIlrT” options were used with showSNPs. Additional filtering steps applied to SNP calls were as follows: (i) removal of SNPs mapping within genes associated with bacteriophage or other mobile elements, (ii) removal of all SNPs within 20 bp of another SNP, (iii) removal of all SNPs within 20 bp of the end of a contig, (iv) removal of SNPs from noncoding regions, and (v) removal of SNPs from any region with greater than 2× the depth of coverage of the strain average. Concatenated SNPs were used to construct phylogenetic trees using the approximate-maximum-likelihood-based approach in FastTree with parameters “-nt -gtr” (51). Trees were visualized in FigTree using the midpoint-branching tree-building option (http://tree.bio.ed.ac.uk/software/figtree/). Local support values for each of the nodes were calculated in FastTree.

Sequence data accession numbers.

Accession numbers for draft sequence files (raw reads) are in Table S5 in the supplemental material. Reference plasmid sequences were downloaded from ftp://ftp.ncbi.nlm.nih.gov/genomes/Plasmids/. Reference transposon sequences used in this study are available at http://metagenomics.partners.org/PathogenGenomes/.

Novel carbapenemase-carrying plasmids identified in this study. Maps of ordered de novo contigs of carbapenemase plasmids with <98% identity to reference plasmids and/or novel insertions not present in reference plasmids. Boundaries of de novo contigs are shown in red, Tn4401 in green, and genes annotated by RAST in blue. Constructs include pBWH-C3-KPC (A), pBWH-C7-KPC (B), pBWH-C9-KPC (C), pBWH-C13-KPC (D), and pBWH-C16-KPC (E). Download

Figure S1, EPS file, 1.4 MB

Chromosomal Tn4401 insertions. BWH-C5 and BWH-C6 both have chromosomal copies of Tn4401 (green), while BWH-C6 has an additional copy on plasmid pBK15692 (green). (A) One set of TDRs was found in BWH-C5 (TTTAA), corresponding to the chromosomal Tn4401. (B) Two sets of TDRs were found in BWH-C6, one for the chromosomal Tn4401 (TTTAA) and one for the plasmid copy in pBK15692 (GTTCT). (A) In BWH-C5, alignment of raw reads to the pBK15692 reference demonstrates deep coverage over the pBK15692 backbone but low/no coverage over Tn4401 except for a low number of reads mapping to the central portion of the transposon likely derived from the chromosomal copy (only Tn4401 flanking regions, including the right- and left-end inverted repeats [IRR and IRL], are shown). (B) In contrast, the entire pBK15692 sequence is covered by raw reads at an even depth in BWH-C6 (including Tn4401). Download

Figure S2, EPS file, 2.1 MB

Workflow for whole-genome sequencing of carbapenem-resistant isolates. Intrainstitutional data analysis, storage, and reporting starts with automated detection of strains to be sequenced, based on their phenotypic resistance profiles. After isolation of genomic DNA, library preparation, and sequencing, the bioinformatics pipeline performs quality control assessments, de novo contig generation, antibiotic gene and transposon identification, preliminary plasmid assignment, and attachment of patient metadata. Outside resources are used to develop content for mobile element analyses. Manual review finalizes the plasmid assignment and epidemiological tracking of strains and plasmids to generate a strain “report card” for hospital infection control. This report defines resistance-carrying strains by the distance tree of chromosomal SNPs of current and prior isolates of interest that further incorporates the resistance genes identified and their carriage on transposon and plasmid mobile elements. Download

Figure S3, EPS file, 2.3 MB

Hypothetical scenario for spread of pBK31551-derived elements within a hospital. K. pneumoniae isolates of ST834 (red), ST113 (orange), and unknown STs (grey) and E. cloacae isolates of ST78 (blue) are shown. The plasmids that were observed are represented by solid lines, and hypothetical plasmids by dashed lines. pBK31551 was first detected in a K. pneumoniae ST834 strain isolated in a New Jersey hospital in 2005. pBWH-C3-KPC was observed at BWH in March 2012 in a K. pneumoniae ST113 strain. Subsequent derivatives of pBK31551 were found in E. cloacae ST78 strains in July 2012 and September 2012. Download

Figure S4, EPS file, 1.5 MB

Strain descriptions with patient characteristics.

Table S1, XLSX file, 0.1 MB

Beta-lactamases and phenotypic resistance in carbapenemase-producing clinical isolates.

Table S2, XLSX file, 0.04 MB

Beta-lactamases and phenotypic resistance in non-carbapenemase-producing clinical isolates.

Table S3, XLSX file, 0.05 MB

Mutations in AmpD, AmpR, OmpC (OmpK36), and OmpF (OmpK35).

Table S4, XLSX file, 0.1 MB

Strain fastq files loaded into the NIH Short Read Archive (SRA).

Table S5, XLSX file, 0.03 MB