RESEARCH

Genomic Definition of Hypervirulent

and Multidrug-Resistant Klebsiella
pneumoniae Clonal Groups
Suzanne Bialek-Davenet,1 Alexis Criscuolo,1 Florent Ailloud, Virginie Passet, Louis Jones,
Anne-Sophie Delannoy-Vieillard, Benoit Garin, Simon Le Hello, Guillaume Arlet,
Marie-Hélène Nicolas-Chanoine, Dominique Decré, and Sylvain Brisse

Multidrug-resistant and highly virulent Klebsiella pneu- Pseudomonas aeruginosa, and Enterobacter species) of bac-
moniae isolates are emerging, but the clonal groups (CGs) terial pathogens (3), raises serious therapeutic challenges.
corresponding to these high-risk strains have remained im- Most multidrug-resistant (MDR) K. pneumoniae isolates,
precisely defined. We aimed to identify K. pneumoniae CGs which produce extended-spectrum β-lactamases (ESBLs)
on the basis of genome-wide sequence variation and to and/or carbapenemases in combination with quinolone and
provide a simple bioinformatics tool to extract virulence and
aminoglycoside resistance, belong to particular clones (4–6).
resistance gene data from genomic data. We sequenced
48 K. pneumoniae isolates, mostly of serotypes K1 and K2,
Invasive community-acquired isolates are predominantly
and compared the genomes with 119 publicly available ge- of capsular serotypes K1 and K2 and appear to differ in
nomes. A total of 694 highly conserved genes were includ- clonal background from MDR isolates (7–11). Controlling
ed in a core-genome multilocus sequence typing scheme, the emergence of these 2 types of high-risk clones and miti-
and cluster analysis of the data enabled precise definition gating the alarming prospect of strains that would combine
of globally distributed hypervirulent and multidrug-resistant high virulence with multidrug resistance requires a precise
CGs. In addition, we created a freely accessible database, definition of clonal groups (CGs) and rapid identification of
BIGSdb-Kp, to enable rapid extraction of medically and epi- their medically relevant features. K. pneumoniae clones have
demiologically relevant information from genomic sequenc- been recognized so far by using multilocus sequence typing
es of K. pneumoniae. Although drug-resistant and virulent (MLST) based on 7 housekeeping genes (4,8,12). However,
K. pneumoniae populations were largely nonoverlapping,
MLST fails to draw clear discontinuities between CGs (4–
isolates with combined virulence and resistance features
were detected.
6). Rapid, high-throughput sequencing promises to revolu-
tionize medical microbiology and molecular epidemiology

K
(13,14) by improving discriminatory power and providing
lebsiella pneumoniae is a frequent cause of nosocomial
access to the resistome and virulome of clinical isolates.
infections and has also emerged as an agent of severe
However, it remains challenging to extract medically rel-
community-acquired infections, including pyogenic liver
evant information from genome sequences in a timely man-
abscess, pneumonia, and meningitis (1,2). The rise of anti-
ner. The objectives of this work were to delineate precisely,
microbial drug resistance in K. pneumoniae, a member of
based on genome-wide genotyping, CGs corresponding to
the ESKAPE group (Enterococcus faecium, Staphylococcus
highly virulent and MDR K. pneumoniae isolates; extract
aureus, Klebsiella pneumoniae, Acinetobacter baumannii,
the antimicrobial drug resistance and virulence-associated
genomic features of those CGs by using a rapid and simple
Author affiliations: Institut Pasteur, Paris, France (S. Bialek-
bioinformatics tool; and detect potential dual-risk isolates
Davenet, A. Criscuolo, F. Ailloud, V. Passet, L. Jones, A.-S. Delannoy-
carrying virulence and resistance genes.
Vieillard, S. Le Hello, S. Brisse); Centre National de la Recherche
Scientifique (CNRS), Paris (S. Bialek-Davenet, A. Criscuolo, V.
Materials and Methods
Passet, S. Brisse); Hôpital Beaujon, Clichy, France (S. Bialek-
Davenet, M.-H. Nicolas-Chanoine); Institut Pasteur, Antananarivo,
Isolate Selection for Genome Sequencing
Madagascar (B. Garin); Sorbonne Université, Paris (G. Arlet, D.
We sequenced 48 nonredundant K. pneumoniae isolates
Decré); Institut National de la Santé et de la Recherche Médicale
(Table S1, http://bigsdb.web.pasteur.fr/klebsiella/archives/
(INSERM), Paris (G. Arlet, M.-H. Nicolas-Chanoine, D. Decré);
Bialek_TechnicalAppendix.pdf). Forty-two of the isolates
Hôpitaux de l’Est Parisien, Paris (G. Arlet, D. Decré); and Faculté de
were of capsular serotype K1 or K2, the 2 serotypes pre-
Médecine, Université Paris Diderot, Paris (M.-H. Nicolas-Chanoine)
dominantly associated with community-acquired invasive
DOI: http://dx.doi.org/10.3201/eid2011.140206 1
These first authors contributed equally to this article.

1812 Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 20, No. 11, November 2014
 Definition of K. pneumoniae Clonal Groups

infections (2,15). To conduct genome sequencing, we used virulence-associated loci, and all currently described vari-
the HiSeq 2000 Sequencing System (Illumina, San Diego, ants of major antimicrobial drug resistance determinants
CA, USA) with a 2 × 100 nt paired-end strategy (see Sup- (namely, β-lactamases, aminoglycoside resistance–confer-
plemental Methods section at http://bigsdb.web.pasteur.fr/ ring enzymes and fluoroquinolone-resistance loci) were in-
klebsiella/archives/Bialek_TechnicalAppendix.pdf). cluded in the BIGSdb-Kp database. For selected strains, the
antimicrobial drug resistance phenotype was analyzed by
Genomes from Sequence Databases using conventional methods (see Supplemental Methods
Genomic sequences available as of July 15, 2013, for section at http://bigsdb.web.pasteur.fr/klebsiella/archives/
the entries Klebsiella, K. pneumoniae, and K. variicola, Bialek_TechnicalAppendix.pdf).
were downloaded from the NCBI (National Center for
Biotechnology Information) genome sequence repository Data Analysis
(http://www.ncbi.nlm.nih.gov/genome/). The downloaded Phylogenetic networks were constructed by using
sequences comprised 7 complete genomes and 115 SplitsTree v4.13.1 (18) based on comparison of allelic
whole-genome shotgun sequences available as scaffolds profiles using distance matrices corresponding to the per-
or contigs. Of these 115 draft genomes, 2 were discarded centage of distinct loci (excluding missing alleles), which
because of the poor quality of assembly and 1 was dis- were obtained using BioNumerics v6.6 (Applied-Maths,
carded because it corresponded in fact to the more distant Sint-Martens-Latem, Belgium). Clonal complexes (CCs)
species K. oxytoca. Thus, a total of 119 publicly available were defined as groups for which MLST profiles showed
genome sequences were included (Table S1, http://big- only 1 allelic mismatch with at least 1 other member of
sdb.web.pasteur.fr/klebsiella/archives/Bialek_Technical the group. Phylogenetic reconstruction based on cgMLST
Appendix.pdf). genes was performed by using minimum evolution analysis
after discarding putative homologous recombination biases
Serotype Determination and MLST Data Generation (see Supplemental Methods section at http://bigsdb.web.
The capsular serotype was determined by PCR, us- pasteur.fr/klebsiella/archives/Bialek_TechnicalAppendix.
ing K1- and K2-specific PCR primers (15). The serotype pdf). Selected genomes were annotated by using the Micro-
of some comparative isolates was initially determined by Scope annotation and comparative genomics platform (19).
using classical serology methods and/or by determining
their C-pattern (8). MLST data were generated by using the Data Availability
international K. pneumoniae MLST typing scheme (8,12). The annotated genomic sequences of strains BJ1-GA,
T69, SA1, and cur15505 (SB2390) were deposited in the
Definition of MLST and Core Genome MLST European Nucleotide Archive (ENA) and are available un-
(cgMLST) Schemes der accession nos. CBTU010000000, CBTV010000000,
A set of 634 genes, defined as the strict cgMLST set, CBTW010000000, and CCBO010000000, respectively.
was obtained by using stringent conservation and synteny The sequences of CIP 52.145 chromosome and plasmids are
criteria to maximize typeability and to minimize paralo- available under accession nos. FO834904 and FO834905
gous or xenologous loci (see Supplemental Methods sec- and FO834906, respectively. Sequence reads correspond-
tion at http://bigsdb.web.pasteur.fr/klebsiella/archives/ ing to the remaining 43 isolates have been deposited in the
Bialek_TechnicalAppendix.pdf). Two typing schemes, ENA Sequence Read Archive under study accession nos.
defined as sets of predefined loci, were implemented by us- ERS500935–ERS500961, ERS503301–ERS503315, and
ing the BIGSdb genome database software (16) in a newly ERS508075. The 48 assembled genomic sequences gener-
created database named BIGSdb-Kp (http://bigsdb.web. ated in this study are accessible from the BIGSdb-Kp data-
pasteur.fr). First, the MLST scheme, with its reference se- base through Institut Pasteur’s whole-genome MLST home
quences and sequence types (STs), was imported from the page (http://bigsdb.web.pasteur.fr).
international MLST database (http://www.pasteur.fr/mlst)
into BIGSdb-Kp, which now acts as the reference. Second, Results
a 694-gene cgMLST scheme was defined as the combina-
tion of the 7 MLST genes, the 53 ribosomal MLST genes Diversity of Virulent and MDR K. pneumoniae
(17), and the 634 strict cgMLST genes. Isolates as Determined by MLST
The average number of contigs and the N50 (i.e., the
Genes Associated with Virulence and Heavy Metal length for which half of the bases of a draft genome are
Resistance or with Drug Resistance situated in contigs of that length or longer) of the 48 ge-
Sequences of genes previously associated with viru- nomes generated in the present study were similar to those
lence or heavy metal resistance were used to define of the 112 publicly available draft genomes (Table S2,

Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 20, No. 11, November 2014 1813
RESEARCH

http://bigsdb.web.pasteur.fr/klebsiella/archives/Bialek_ allelic profiles were compared (Figure 1). Sharp disconti-

TechnicalAppendix.pdf). These 2 characteristics quan- nuities were apparent: groups of isolates were clustered in
tify the level of fragmentation of the sequence assemblies. homogeneous branches that were well separated from other
MLST allele sequences were retrieved from genomic se- groups. These results demonstrate the existence of clearly
quences imported into the BIGSdb-Kp database. For 46 of delineated K. pneumoniae clones. To define CGs nonarbi-
the 48 sequenced strains, MLST alleles were also obtained trarily, we analyzed the distribution of the number of al-
by the classical Sanger sequencing method, which, in all lelic mismatches (loci at which sequences differ) among all
cases, confirmed the genomic data. This result corresponds pairs of genomes (Figure S1, http://bigsdb.web.pasteur.fr/
to <1 error in 138,552 nt (i.e., 0.00072%), demonstrating klebsiella/archives/Bialek_TechnicalAppendix.pdf). The
the high quality of consensus base calls in the assemblies. results showed a high number of genome pairs with <100
Forty-three different STs were found in the entire da- mismatches or 500–600 mismatches; almost no genome
taset. Among the isolates we sequenced, the 20 K1 isolates pairs had 100–300 mismatches. Therefore, we propose to
represented 2 STs, among which ST23 (n = 18) predomi- define K. pneumoniae cgMLST CGs a groups of cgMLST
nated, whereas the 22 K2 isolates represented 10 distinct profiles having <100 allelic mismatches (i.e., 14.4% of the
STs, among which ST380 (n = 5) and ST86 (n = 6) predom- 694 alleles) with at least 1 other member of the group.
inated (Table S1, http://bigsdb.web.pasteur.fr/klebsiella/ar- By using this definition, we defined 14 CGs (Figure
chives/Bialek_TechnicalAppendix.pdf). Isolates belonging 1). The MDR-associated ST258 and its MLST single-lo-
to these 3 predominant STs, as well as ST57, ST65, and cus variants belonged to a single well-demarcated group,
ST375, were previously reported from invasive infections designated CG258. On the basis of MLST results, ST258
(7–11,20). Reference strains NTUH-K2044 and CG43 was previously associated with many other STs in a single
were found to belong to ST23 and ST86, respectively. Pub- heterogeneous CC (4,6). We showed that most of these
licly available genomes predominantly included (85/119, STs do not belong to CG258, demonstrating the power of
71%) STs previously associated with multidrug resistance: cgMLST to discard spurious MLST associations. With-
ST11, ST14, ST15, ST258, and ST512 (4,5,21,22). There- in CG258, isolates of ST258 and ST512 formed a main
fore, the dataset comprised multiple representative isolates cluster with, on average, only 2.1% allelic mismatches,
of both types of high-risk K. pneumoniae. whereas strains of ST11 and ST340 differed from this
cluster by 11.0%, indicating genetic substructure within
Definition of CGs based on cgMLST this CG. The other CC associated with multidrug resis-
Sequences corresponding to the 694 cgMLST loci were tance, CC14 (5), was split into 2 distinct CGs: CG14 and
extracted from the 167 genomes by using BIGSdb, and CG15 (Figure 1).
Figure 1. Phylogenetic network of the
167 Klebsiella pneumoniae genomes
as determined on the basis of the
allelic profiles of the 694 core genome
multilocus sequence typing (cgMLST)
genes. The network was constructed
by using the neighbor-net method
implemented in SplitsTree v4.13.1
(18). Nodes are colored according
to the clonal group (CG). Only the 10
most relevant CGs are highlighted;
note that ST35 was subdivided
into 2 CGs (CG35-A and CG35-B).
Gray shading indicates CG258.
Gray dots indicate phylogroups
KpII-B and KpIII (K. variicola). Bold
indicates reference strains. Scale bar
represents 100 allelic mismatches.
ATCC, American Type Culture Col-
lection; ST, sequence type.

1814 Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 20, No. 11, November 2014
 Definition of K. pneumoniae Clonal Groups

Most isolates of serotype K1, which is associated the differential clinical characteristics of infections they
with pyogenic liver abscess, belong to ST23 and ST57 cause, remain to be determined by using prospective col-
(7,8,11). All isolates of these 2 STs were placed together lections of isolates.
in CG23, a compact, clearly delineated CG. CG23 was To determine the population structure of K. pneumoni-
highly homogeneous (3.0% distinct alleles on average), ae, we performed phylogenetic analyses of cgMLST se-
even though epidemiologically unrelated isolates from quences (Figure 2). The results showed that most branches
distinct continents were included (Table S1, http://bigsdb. leading to CGs were long and branched deep in the tree. This
web.pasteur.fr/klebsiella/archives/Bialek_TechnicalAp- pattern suggests an evolutionary radiation of most groups
pendix.pdf). This result suggests that these strains recently at approximately the same time, presumably when K. pneu-
emerged from a common ancestor, which is in agreement moniae expanded into a novel niche. Whereas CG23 and
with epidemiologic evidence for the recent emergence of CG380 branched off early, CG375 and CG86 were placed
liver abscess caused by K. pneumoniae (2). STs associ- on a single branch together with ST25 and CG65, suggest-
ated with serotype K2 were distributed into 3 main CGs: ing a common ancestry for these 4 latter clones.
CG86, CG375, and CG380. In addition, K2 isolates of
MLST-defined CC65 (ST25, ST65, and ST375) repre- Reproducibility and Epidemiologic Relevance
sented 3 distinct CGs, which differed by at least 32%. of cgMLST
This result shows that MLST classification into CCs can One pair of public genomes in fact corresponded to the
conflate members of distinct CGs (Figure 2; Table S1, same strain: BAA-2146. This isolate, an MDR NDM-1–
http://bigsdb.web.pasteur.fr/klebsiella/archives/Bialek_ producing isolate belonging to ST11, was independently se-
TechnicalAppendix.pdf), and it illustrates the inability of quenced twice and was assembled by using 2 distinct meth-
MLST to reliably recognize K. pneumoniae CGs. In our ods (23) (GenBank accession no. APNN01). The 2 genomic
study, the use of cgMLST demonstrated that K2 isolates assemblies did not show a single allelic difference. This
associated with severe community-acquired infections observation indicates that cgMLST is highly reproducible.
are genetically more diverse than K1 isolates. The rela- To explore the ability of cgMLST to cluster epide-
tive prevalence of these serotype K2–associated CGs, and miologically associated isolates and distinguish them from

Figure 2. Phylogenetic tree of the 167 Klebsiella pneumoniae genomes as determined on the basis of core genome multilocus sequence
typing (cgMLST) genes and distribution of virulence and resistance features. The tree was inferred from minimum evolution analysis based
on aligned cgMLST sequences, with K. variicola and KpII-B sequences as outgroups. Terminal branches corresponding to different taxa
from the same clonal group (CG) or sequence type (ST) are shown as triangles of depth proportional to internal diversity. Bootstrap values
>50% based on 1,000 gene-by-gene replicates are given at branches. Scale bar represents 0.05% estimated sequence divergence. The
virulence and resistance gene content (indicated along the top of the figure) of identified clones is represented by squares, which are
colored in black proportionally to the percentage of presence of a gene or cluster among members of a given CG or ST.

Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 20, No. 11, November 2014 1815
RESEARCH

other genetically closely related isolates, we included in addition to the yersiniabactin siderophore cluster and
in our analysis 20 published genomes of isolates from a (except for CG23) the 2-component system kvgAS. These
2011 outbreak at the National Institutes of Health Clini- results reinforce the view that CG23, CG86, CG375, and
cal Center (Bethesda, Maryland, USA) (24). The outbreak CG380 represent hypervirulent CGs of K. pneumoniae (7–
isolates formed a unique cluster nested within the diversity 11,20,26). Several other STs and CGs (ST25, CG65, ST66
of CG258 (Figure 3), suggesting that cgMLST may be use- [corresponding to the virulent K2 reference strain CIP
ful for short-term epidemiologic questions and outbreak in- 52.145], ST90, and ST382) harbored virulence genes, sug-
vestigation. Additional studies on well-defined sets of out- gesting additional hypervirulent clones, which is also sup-
breaks will be needed to define levels of variation among ported by the clinical origin of some isolates of these STs
epidemiologically related and unrelated isolates. (Table S1, http://bigsdb.web.pasteur.fr/klebsiella/archives/
Bialek_TechnicalAppendix.pdf). In contrast, CG258 was
Detection of Virulence-Associated Genes almost entirely devoid of virulence genes. Microcin E492
in K. pneumoniae Genomes and colibactin synthesis clusters were present in almost all
Genes previously associated with virulence and gene isolates of CG23 and CG380. These 2 secreted molecules,
clusters coding for resistance to heavy metals, which can which cause damage to eukaryotic cells in vitro, may con-
be encoded on the virulence plasmid (25), were searched tribute to the virulence of these 2 CGs.
by using BIGSdb (Figure 2; Table S3, http://bigsdb.web.
pasteur.fr/klebsiella/archives/Bialek_TechnicalAppen- Detection of Drug Resistance–Associated Genes
dix.pdf). The type-3 fimbriae cluster mrkABCDF was al- in K. pneumoniae Genomes
most universally present, and plasmid-associated clusters We investigated the presence of genes associated with
pcoABCDERS (copper resistance) and silCERS (silver resistance to β-lactams, aminoglycosides, and quinolones.
resistance) were widely distributed. In contrast, the other Because single amino acid changes can have medically
genes were mainly associated with particular CGs. The al- relevant phenotypic effects, we searched the known pro-
lantoinase cluster was present only in members of CG23 tein variants of the major β-lactamase families (Figure 2;
and ST25. Some or all isolates of the major CGs associ- Table S4, http://bigsdb.web.pasteur.fr/klebsiella/archives/
ated with community-acquired invasive infections (CG23, Bialek_TechnicalAppendix.pdf). At least 1 variant of the
CG86, CG375, and CG380) harbored the 2 well-recog- SHV enzyme was found in 160 isolates, among which 159
nized virulence factors iucABCDiutA (aerobactin synthe- belonging to phylogroup KpI (27). The K. variicola and
sis) and rmpA (the positive regulator of mucoid phenotype) KpII-B isolates harbored blaLEN and blaOKP-B, respectively,

Figure 3. Phylogenetic network

of the 82 Klebsiella pneumoniae
strains belonging to clonal group
(CG) 258 as determined on the basis
of the allelic profiles of the 694 core
genome multilocus sequence typing
(cgMLST) genes. The 20 genomes
corresponding to isolates from the
2011 K. pneumoniae outbreak at the
National Institutes of Health Clinical
Center (Bethesda, Maryland, USA)
(24) are highlighted by gray shading.
Scale bar represents 10 allelic
mismatches. ST, sequence type.

1816 Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 20, No. 11, November 2014
 Definition of K. pneumoniae Clonal Groups

and 1 KpII-B isolate also contained the ESBL-encoding was resistant to ceftriaxone and ceftazidime, an antimicro-
blaSHV-18 gene. These results are consistent with previous bial drug resistance phenotype compatible with the produc-
associations of K. pneumoniae phylogroups KpI, KpII, tion of an SHV-type ESBL. However, the blaSHV allele pres-
and KpIII with chromosomally encoded β-lactamase fami- ent in this strain’s genome could not be precisely identified
lies SHV, OKP, and LEN, respectively (27,28). No gene by BIGSdb-Kp because it was situated at a contig extremity.
corresponding to a chromosomal β-lactamase was detect- Although isolates of hypervirulent CGs harbored al-
ed in the genomes of CIP 52.145 (ampicillin susceptible) most no resistance genes (Figure 2), we did identify 2 iso-
and 4_1_44. lates of CG23 (BG130 and BG141) that displayed genomic
In addition, blaTEM and blaOXA were detected in 45 and features responsible for high-level resistance to β-lactams,
80 genomes, respectively. In 48 of the latter, an internal aminoglycosides, and quinolones, as confirmed by analy-
stop codon was present within blaOXA-9, leading to a trun- sis of the resistance phenotype (Table S1, http://bigsdb.
cated protein. This truncation was previously described web.pasteur.fr/klebsiella/archives/Bialek_TechnicalAp-
on the multiple antimicrobial drug resistance plasmid pendix.pdf). The first strain, BG130, which was isolated in
pRMH712 isolated from strain 4003 (ENA accession no. Vietnam in 2008, contained an array of genes identical to
GU553923). Nine genomes contained a gene encoding the a part of the Tn6061 transposon from Pseudomonas aeru-
ESBL CTX-M-15. As expected, all 20 genomes from the ginosa and to the VEB-1-encoding region found in a K.
2011 outbreak at the National Institutes of Health Clinical pneumoniae isolate from Greece (33); these genes were
Center were found to harbor a gene encoding the carbapen- blaVEB-1, ant(2″)-Ia (aadB), arr-2, cmlA5, blaOXA-10, ant(3″)-
emase KPC-3 (24). A blaKPC gene was also detected in 62 Ia (aadA1), qacEΔ1, sulIΔ, cmlA9, tetR, and tetA. More-
other genomes. Several other β-lactamase–encoding genes, over, downstream of this gene array, strain BG130 carried
including blaNDM-1, were detected (Table S1, http://bigsdb. an additional region comprising gene rmtB, encoding an
web.pasteur.fr/klebsiella/archives/Bialek_TechnicalAp- rRNA methylase, and genes blaTEM-1 and blaCTX-M-15. Strain
pendix.pdf), consistent with earlier findings (29). BG130 also harbored gene qnrS1; the gene had the same
Regarding loci associated with resistance to quino- surroundings as those in plasmid pHS8, which was de-
lones, genes qnrB and qnrS were each present in 5 ge- scribed in a K. pneumoniae isolate from China (34). The
nomes, whereas a qnrA1 gene was found in the genome of second strain, BG141, which was isolated in Madagascar
strain 1162281, consistent with previous work (30) (Fig- in 2007, also carried numerous resistance genes. Those
ure 2; Table S5, http://bigsdb.web.pasteur.fr/klebsiella/ genes included blaCTX-M-15, blaTEM-1, blaOXA-1, aac(6′)-Ib-cr,
archives/Bialek_TechnicalAppendix.pdf). Genes coding and qnrB1, and the gene order was identical to that found
for the enzymatic targets of quinolones were detected in in plasmids pKDO1 and pKPX-2, which were identified in
all genomes, except for gene gyrB, which was absent in 2 MDR K. pneumoniae strains from the Czech Republic and
genomes. The gyrA and parC genes were mutated in their Taiwan, respectively (35,36). These results are evocative
quinolone resistance–determining region in all isolates of horizontal gene transfer that may have occurred between
of the MDR clone CG258, whereas no mutation of these virulent and MDR K. pneumoniae strains.
genes was found among the hypervirulent clones. In ad-
dition, the oqxAB locus encoding an efflux pump (31) was Discussion
detected in 163 of the 167 genomes, demonstrating that this K. pneumoniae clinical isolates are evolving toward
locus is highly conserved in K. pneumoniae (32). increasing levels of antimicrobial drug resistance, placing
We identified 16 aminoglycoside resistance–associ- this species among the infectious bacterial pathogens that
ated loci in the genomes studied (Figure 2; Table S6, http:// are most challenging to control (3). In parallel, even though
bigsdb.web.pasteur.fr/klebsiella/archives/Bialek_Techni- most infections still occur opportunistically in debilitated
calAppendix.pdf). The most frequently represented gene patients, K. pneumoniae is emerging as a cause of severe
(100 genomes) was ant(3″)-Ia. The aac(6′)-Ib-cr variant, community-acquired invasive infections (2). Recognizing
coding for an aminoglycoside- and quinolone-modifying the resistant or hypervirulent CGs is a prerequisite to better
enzyme, was found in 6 genomes (Table S5, http://bigsdb. understand and control their global emergence. We imple-
web.pasteur.fr/klebsiella/archives/Bialek_TechnicalAp- mented a genome database for K. pneumoniae, BIGSdb-
pendix.pdf). Genes conferring resistance to aminogly- Kp, and defined 694 genomic loci suitable for genotyping
cosides were mainly present in the MDR clones CG258 as well as loci associated with virulence and antimicrobial
and CG14. drug resistance. A genome-wide gene-by-gene approach
For all strains except cur15505, antimicrobial drug re- based on the BIGSdb system (16) was previously applied
sistance phenotypes were highly concordant with detected to Neisseria meningitidis (37) and Campylobacter species
genes (Table S1, at http://bigsdb.web.pasteur.fr/klebsiella/ (38). The BIGSdb system provides a simple tool for rapidly
archives/Bialek_TechnicalAppendix.pdf). Strain cur15505 extracting medically relevant information.

Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 20, No. 11, November 2014 1817
RESEARCH

The detection of genes with the BIGSdb system faces in Bethesda (24) from epidemiologically unrelated CG258
2 limitations. First, for multicopy genes, only 1 occur- isolates. However, local epidemiologic investigations will
rence might be detected. In K. pneumoniae, this problem require more resolution to decipher recent transmission
is posed by the possible co-occurrence of chromosomal events. Because the typical K. pneumoniae genome harbors
and plasmid-borne blaSHV genes. However, this issue ≈5,300 genes, there is ample room for defining additional
has been recently addressed since version 1.8.0 (http:// loci to improve discriminatory power. The BIGSdb-Kp da-
sourceforge.net/p/bigsdb/news/; K. Jolley, pers. comm.). tabase can host additional loci and schemes curated at a
Second, the number of loci for which allelic number at- distance by distinct laboratories (16).
tribution fails is strongly dependent on the assembly frag- Our findings demonstrate the existence of clearly
mentation (Figure S2, http://bigsdb.web.pasteur.fr/klebsi- distinguishable K. pneumoniae CGs and show that MDR
ella/archives/Bialek_TechnicalAppendix.pdf). Therefore, and hypervirulent populations of this species are largely
it seems advisable to exclude highly fragmented genomes nonoverlapping. However, hypervirulent K. pneumoniae
from the analysis. are beginning to evolve toward increasing levels of an-
Several K. pneumoniae international clones associated timicrobial drug resistance and may represent a new and
with multidrug resistance or hypervirulence were described serious threat to public health (40). The freely accessible
by using MLST (4–8,21). However, the low level of nu- BIGSdb-Kp database represents a novel tool for monitor-
cleotide sequence divergence among K. pneumoniae iso- ing the emergence of high-risk clones and for global col-
lates (8) makes it difficult to define borders between clones laboration on the population biology, epidemiology, and
(4–6). The 694-gene cgMLST genotyping scheme repre- pathogenesis of K. pneumoniae.
sents ≈100 times more sequence information than that pro-
vided by the 7-gene MLST. cgMLST demonstrated clear Acknowledgments
discontinuities in the K. pneumoniae genotypic space and We are grateful to K. Jolley for advice on BIGSdb instal-
showed that high-risk CGs were remarkably distinct from lation. We thank L. Lery and R. Tournebize for providing the
their closest genotypic neighbors. This result is in strik- complete genome sequence of strain CIP 52.145, J.-M. Thiberge
ing contrast to the genotypic continuum obtained by using for help in the initial setup of the BIGSdb-Kp database, and N.
MLST. Retrospectively, the failure of MLST to disclose Nihotte for contributing to the Institut Pasteur BIGSdb home
the sharp discontinuities among CGs can be attributed to a pages. The following colleagues are acknowledged for provid-
severe lack of resolution. ing K. pneumoniae isolates or reference strains: B. De Barbeyrac
The discovery of recognizable K. pneumoniae CGs (BD-DU), F. Randrianirina (BG94 and BG141), C. De Champs
opens the way to studying their genomic and biologic (CH137), E. Carbonelle (Zaire1), H. Courtade (100519185), E.
specificities. Different combinations of virulence and re- van Duikeren (V9902406), C. Forestier (LM21 and CH1031),
sistance genes were found among CGs, providing identi- F. Jauréguy (BP1011625), P. A. D. Grimont (CDC 4241-71), A.
fication markers and hinting that these clones had distinct Mérens (610356538), A. Merlet (20479), H.-L. Peng (CG43),
host–pathogen relationships and antimicrobial drug expo- D. Tainturier (MET1_63/88063), and J.-T. Wang (A3021 and
sure. The genomic signatures of CGs show that MDR and A5011). We thank E. Rocha for helpful comments.
hypervirulent populations of K. pneumoniae are, so far,
A.C. and genomic sequencing were supported financially
mostly nonoverlapping. However, genes encoding resis-
by a grant from Region Île-de-France. S.B.-D. was supported
tance to β-lactams, quinolones, and aminoglycosides were
by a postdoc grant from Assistance Publique–Hôpitaux de Par-
detected in 2 isolates of hypervirulent clone CG23: BG130
is and Institut Pasteur. This work was supported by the French
(Vietnam, 2008) and BG141 (Madagascar, 2007). These 2
Government’s Investissement d’Avenir program, Laboratoire
isolates are among the first documented members of CG23
d’Excellence Integrative Biology of Emerging Infectious Dis-
shown to harbor blaCTX-M-15 and genes conferring high-
eases (grant no. ANR-10-LABX-62-IBEID).
level resistance to aminoglycosides and fluoroquinolones.
A CTX-M-15–producing ST23 strain was also reported Dr Bialek-Davenet is a microbiologist working as a postdoc-
from South Korea (39). These results show that the gloomy toral fellow in the Microbial Evolutionary Genomics Unit, Institut
prospect of dual-risk K. pneumoniae strains, combining Pasteur. Her research interests include genomic epidemiology and
virulence and multidrug resistance features, is becoming resistance and virulence determinants of K. pneumoniae.
a reality.
Genome-wide genotyping is a powerful approach to
References
address fine-scale epidemiologic questions (37). Our find-
ings show that the discriminatory power of cgMLST, based 1. Podschun R, Ullmann U. Klebsiella spp. as nosocomial pathogens:
on 694 genes, makes it possible to distinguish a subgroup epidemiology, taxonomy, typing methods, and pathogenicity factors.
of isolates involved in the 2011 K. pneumoniae outbreak Clin Microbiol Rev. 1998;11:589–603.

1818 Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 20, No. 11, November 2014
 Definition of K. pneumoniae Clonal Groups

2. Shon AS, Bajwa RP, Russo TA. Hypervirulent (hypermucoviscous) 18. Huson DH, Bryant D. Application of phylogenetic networks in evo-
Klebsiella pneumoniae: a new and dangerous breed. Virulence. lutionary studies. Mol Biol Evol. 2006;23:254–67. http://dx.doi.
2013;4:107–18. http://dx.doi.org/10.4161/viru.22718 org/10.1093/molbev/msj030
3. Rice LB. Federal funding for the study of antimicrobial resistance in 19. Vallenet D, Belda E, Calteau A, Cruveiller S, Engelen S, Lajus A,
nosocomial pathogens: no ESKAPE. J Infect Dis. 2008;197:1079– et al. MicroScope–an integrated microbial resource for the cura-
81. http://dx.doi.org/10.1086/533452 tion and comparative analysis of genomic and metabolic data.
4. Woodford N, Turton JF, Livermore DM. Multiresistant Gram- Nucleic Acids Res. 2013;41:D636–47. http://dx.doi.org/10.1093/
negative bacteria: the role of high-risk clones in the dissemination nar/gks1194
of antibiotic resistance. FEMS Microbiol Rev. 2011;35:736–55. 20. Bialek-Davenet S, Nicolas-Chanoine MH, Decre D, Brisse S. Micro-
http://dx.doi.org/10.1111/j.1574-6976.2011.00268.x biological and clinical characteristics of bacteraemia caused by the
5. Breurec S, Guessennd N, Timinouni M, Le TA, Cao V, Ngandjio A, hypermucoviscosity phenotype of Klebsiella pneumoniae in Korea
et al. Klebsiella pneumoniae resistant to third-generation cepha- [letter]. Epidemiol Infect. 2013;141:188. http://dx.doi.org/10.1017/
losporins in five African and two Vietnamese major towns: mul- S0950268812002051
ticlonal population structure with two major international clonal 21. Samuelsen Ø, Naseer U, Tofteland S, Skutlaberg DH, Onken A,
groups, CG15 and CG258. Clin Microbiol Infect. 2013;19:349–55. Hjetland R, et al. Emergence of clonally related Klebsiella pneu-
http://dx.doi.org/10.1111/j.1469-0691.2012.03805.x moniae isolates of sequence type 258 producing plasmid-medi-
6. Baraniak A, Izdebski R, Fiett J, Sadowy E, Adler A, Kazma M, et al. ated KPC carbapenemase in Norway and Sweden. J Antimicrob
Comparative population analysis of Klebsiella pneumoniae strains Chemother. 2009;63:654–8. http://dx.doi.org/10.1093/jac/dkp018
with extended-spectrum β-lactamases colonizing patients in reha- 22. Warburg G, Hidalgo-Grass C, Partridge SR, Tolmasky ME,
bilitation centers in four countries. Antimicrob Agents Chemother. Temper V, Moses AE, et al. A carbapenem-resistant Klebsiella
2013;57:1992–7. http://dx.doi.org/10.1128/AAC.02571-12 pneumoniae epidemic clone in Jerusalem: sequence type 512 car-
7. Turton JF, Englender H, Gabriel SN, Turton SE, Kaufmann ME, rying a plasmid encoding aac(6’)-Ib. J Antimicrob Chemother.
Pitt TL. Genetically similar isolates of Klebsiella pneumoniae sero- 2012;67:898–901. http://dx.doi.org/10.1093/jac/dkr552
type K1 causing liver abscesses in three continents. J Med Micro- 23. Broberg CA, Palacios M, Miller VL. Whole-genome draft sequences
biol. 2007;56:593–7. http://dx.doi.org/10.1099/jmm.0.46964-0 of three multidrug-resistant Klebsiella pneumoniae strains available
8. Brisse S, Fevre C, Passet V, Issenhuth-Jeanjean S, Tournebize R, from the American Type Culture Collection. Genome Announc.
Diancourt L, et al. Virulent clones of Klebsiella pneumoniae: iden- 2013;1:e00312–3. http://dx.doi.org/10.1128/genomeA.00312-13
tification and evolutionary scenario based on genomic and pheno- 24. Snitkin ES, Zelazny AM, Thomas PJ, Stock F, Henderson DK,
typic characterization. PLoS ONE. 2009;4:e4982. http://dx.doi. Palmore TN, et al. Tracking a hospital outbreak of carbapenem-re-
org/10.1371/journal.pone.0004982 sistant Klebsiella pneumoniae with whole-genome sequencing. Sci
9. Decré D, Verdet C, Emirian A, Le Gourrierec T, Petit JC, Transl Med. 2012;4:148ra16.
Offenstadt G, et al. Emerging severe and fatal infections due to 25. Chen YT, Chang HY, Lai YC, Pan CC, Tsai SF, Peng HL. Sequencing
Klebsiella pneumoniae in two university hospitals in France. J Clin and analysis of the large virulence plasmid pLVPK of Klebsi-
Microbiol. 2011;49:3012–4. http://dx.doi.org/10.1128/JCM.00676-11 ella pneumoniae CG43. Gene. 2004;337:189–98. http://dx.doi.
10. Jung SW, Chae HJ, Park YJ, Yu JK, Kim SY, Lee HK, et al. org/10.1016/j.gene.2004.05.008
Microbiological and clinical characteristics of bacteraemia caused 26. Siu LK, Fung CP, Chang FY, Lee N, Yeh KM, Koh TH, et al.
by the hypermucoviscosity phenotype of Klebsiella pneumoni- Molecular typing and virulence analysis of serotype K1 Klebsi-
ae in Korea. Epidemiol Infect. 2013;141:334–40. http://dx.doi. ella pneumoniae strains isolated from liver abscess patients and
org/10.1017/S0950268812000933 stool samples from noninfectious subjects in Hong Kong, Singa-
11. Merlet A, Cazanave C, Dutronc H, de Barbeyrac B, Brisse S, pore, and Taiwan. J Clin Microbiol. 2011;49:3761–5. http://dx.doi.
Dupon M. Primary liver abscess due to CC23-K1 virulent clone of org/10.1128/JCM.00977-11
Klebsiella pneumoniae in France. Clin Microbiol Infect. 2012;18: 27. Haeggman S, Lofdahl S, Paauw A, Verhoef J, Brisse S. Diversity and
E338–9. http://dx.doi.org/10.1111/j.1469-0691.2012.03953.x evolution of the class A chromosomal beta-lactamase gene in Kleb-
12. Diancourt L, Passet V, Verhoef J, Grimont PA, Brisse S. Multilo- siella pneumoniae. Antimicrob Agents Chemother. 2004;48:2400–8.
cus sequence typing of Klebsiella pneumoniae nosocomial isolates. http://dx.doi.org/10.1128/AAC.48.7.2400-2408.2004
J Clin Microbiol. 2005;43:4178–82. http://dx.doi.org/10.1128/ 28. Fevre C, Passet V, Weill FX, Grimont PA, Brisse S. Variants of the Kleb-
JCM.43.8.4178-4182.2005 siella pneumoniae OKP chromosomal beta-lactamase are divided into
13. Köser CU, Ellington MJ, Cartwright EJ, Gillespie SH, Brown NM, two main groups, OKP-A and OKP-B. Antimicrob Agents Chemother.
Farrington M, et al. Routine use of microbial whole genome sequenc- 2005;49:5149–52. http://dx.doi.org/10.1128/AAC.49.12.5149-5152.2005
ing in diagnostic and public health microbiology. PLoS Pathog. 29. Leski T, Vora GJ, Taitt CR. Multidrug resistance determinants from
2012;8:e1002824. http://dx.doi.org/10.1371/journal.ppat.1002824 NDM-1-producing Klebsiella pneumoniae in the USA. Int J Anti-
14. Struelens MJ, Brisse S. From molecular to genomic epidemiology: microb Agents. 2012;40:282–4. http://dx.doi.org/10.1016/j.ijantimi-
transforming surveillance and control of infectious diseases. Euro cag.2012.05.019
Surveill. 2013;18:20386. 30. Kumar V, Sun P, Vamathevan J, Li Y, Ingraham K, Palmer L, et al.
15. Fang CT, Lai SY, Yi WC, Hsueh PR, Liu KL, Chang SC. Klebsi- Comparative genomics of Klebsiella pneumoniae strains with dif-
ella pneumoniae genotype K1: an emerging pathogen that causes ferent antibiotic resistance profiles. Antimicrob Agents Chemother.
septic ocular or central nervous system complications from pyo- 2011;55:4267–76. http://dx.doi.org/10.1128/AAC.00052-11
genic liver abscess. Clin Infect Dis. 2007;45:284–93. http://dx.doi. 31. Hansen LH, Johannesen E, Burmolle M, Sørensen AH, Sørensen
org/10.1086/519262 SJ. Plasmid-encoded multidrug efflux pump conferring resistance
16. Jolley KA, Maiden MC. BIGSdb: scalable analysis of bacterial to olaquindox in Escherichia coli. Antimicrob Agents Chemother.
genome variation at the population level. BMC Bioinformatics. 2004;48:3332–7. http://dx.doi.org/10.1128/AAC.48.9.3332-3337.2004
2010;11:595. http://dx.doi.org/10.1186/1471-2105-11-595 32. Perez F, Rudin SD, Marshall SH, Coakley P, Chen L, Kreiswirth
17. Jolley KA, Bliss CM, Bennett JS, Bratcher HB, Brehony C, BN, et al. OqxAB, a quinolone and olaquindox efflux pump, is
Colles FM, et al. Ribosomal multilocus sequence typing: universal widely distributed among multidrug-resistant Klebsiella pneu-
characterization of bacteria from domain to strain. Microbiology. moniae isolates of human origin. Antimicrob Agents Chemother.
2012;158:1005–15. http://dx.doi.org/10.1099/mic.0.055459-0 2013;57:4602–3. http://dx.doi.org/10.1128/AAC.00725-13

Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 20, No. 11, November 2014 1819
RESEARCH

33. Papagiannitsis CC, Miriagou V, Kotsakis SD, Tzelepi E, from whole-genome sequence data with rapid web-based analysis
Vatopoulos AC, Petinaki E, et al. Characterization of a transmissible methods. J Clin Microbiol. 2012;50:3046–53. http://dx.doi.org/10.
plasmid encoding VEB-1 and VIM-1 in Proteus mirabilis. Antimi- 1128/JCM.01312-12
crob Agents Chemother. 2012;56:4024–5. http://dx.doi.org/10.1128/ 38. Sheppard SK, Jolley KA, Maiden MC. A gene-by-gene approach
AAC.00470-12 to bacterial population genomics: whole genome MLST of Campy-
34. Hu FP, Xu XG, Zhu DM, Wang MG. Coexistence of qnrB4 and qnrS1 lobacter. Genes (Basel). 2012;3:261–77. http://dx.doi.org/10.3390/
in a clinical strain of Klebsiella pneumoniae. Acta Pharmacol Sin. genes3020261
2008;29:320–4. http://dx.doi.org/10.1111/j.1745-7254.2008.00757.x 39. Shin J, Soo Ko K. Single origin of three plasmids bearing
35. Dolejska M, Villa L, Dobiasova H, Fortini D, Feudi C, Carattoli A. blaCTX-M-15 from different Klebsiella pneumoniae clones. J Antimicrob
Plasmid content of a clinically relevant Klebsiella pneumoniae clone Chemother. 2014;69:969–72. http://dx.doi.org/10.1093/jac/dkt464
from the Czech Republic producing CTX-M-15 and QnrB1. Antimi- 40. Li W, Sun G, Yu Y, Li N, Chen M, Jin R, et al. Increasing occur-
crob Agents Chemother. 2013;57:1073–6. http://dx.doi.org/10.1128/ rence of antimicrobial-resistant hypervirulent (hypermucoviscous)
AAC.01886-12 Klebsiella pneumoniae isolates in China. Clin Infect Dis.
36. Huang TW, Chen TL, Chen YT, Lauderdale TL, Liao TL, Lee YT, 2014;58:225–32. http://dx.doi.org/10.1093/cid/cit675
et al. Copy number change of the NDM-1 sequence in a multi-
drug-resistant Klebsiella pneumoniae clinical isolate. PLoS ONE.
Address for correspondence: Sylvain Brisse, Microbial Evolutionary
2013;8:e62774. http://dx.doi.org/10.1371/journal.pone.0062774
37. Jolley KA, Hill DM, Bratcher HB, Harrison OB, Feavers IM, Genomics, Institut Pasteur, 28 rue du Dr Roux, 75724 Paris, France;
Parkhill J, et al. Resolution of a meningococcal disease outbreak email: sylvain.brisse@pasteur.fr

Sign up for Twitter and find the latest information about

emerging infectious diseases from the EID journal.
@CDC_EIDjournal

1820 Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 20, No. 11, November 2014