Skip to main content
Browse Subject Areas

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Virulent Clones of Klebsiella pneumoniae: Identification and Evolutionary Scenario Based on Genomic and Phenotypic Characterization

  • Sylvain Brisse ,

    Affiliations Institut Pasteur, Genotyping of Pathogens and Public Health, Paris, France, Institut Pasteur, Biodiversité des Bactéries Pathogènes Emergentes, Paris, France

  • Cindy Fevre,

    Affiliations Institut Pasteur, Genotyping of Pathogens and Public Health, Paris, France, Institut Pasteur, Biodiversité des Bactéries Pathogènes Emergentes, Paris, France

  • Virginie Passet,

    Affiliations Institut Pasteur, Genotyping of Pathogens and Public Health, Paris, France, Institut Pasteur, Biodiversité des Bactéries Pathogènes Emergentes, Paris, France

  • Sylvie Issenhuth-Jeanjean,

    Affiliation Institut Pasteur, Biodiversité des Bactéries Pathogènes Emergentes, Paris, France

  • Régis Tournebize,

    Affiliations Institut Pasteur, Unité de Pathogénie Microbienne Moléculaire, Paris, France, Unité INSERM U786, Institut Pasteur, Paris, France

  • Laure Diancourt,

    Affiliations Institut Pasteur, Genotyping of Pathogens and Public Health, Paris, France, Institut Pasteur, Biodiversité des Bactéries Pathogènes Emergentes, Paris, France

  • Patrick Grimont

    Affiliation Institut Pasteur, Biodiversité des Bactéries Pathogènes Emergentes, Paris, France


Klebsiella pneumoniae is found in the environment and as a harmless commensal, but is also a frequent nosocomial pathogen (causing urinary, respiratory and blood infections) and the agent of specific human infections including Friedländer's pneumonia, rhinoscleroma and the emerging disease pyogenic liver abscess (PLA). The identification and precise definition of virulent clones, i.e. groups of strains with a single ancestor that are associated with particular infections, is critical to understand the evolution of pathogenicity from commensalism and for a better control of infections. We analyzed 235 K. pneumoniae isolates of diverse environmental and clinical origins by multilocus sequence typing, virulence gene content, biochemical and capsular profiling and virulence to mice. Phylogenetic analysis of housekeeping genes clearly defined clones that differ sharply by their clinical source and biological features. First, two clones comprising isolates of capsular type K1, clone CC23K1 and clone CC82K1, were strongly associated with PLA and respiratory infection, respectively. Second, only one of the two major disclosed K2 clones was highly virulent to mice. Third, strains associated with the human infections ozena and rhinoscleroma each corresponded to one monomorphic clone. Therefore, K. pneumoniae subsp. ozaenae and K. pneumoniae subsp. rhinoscleromatis should be regarded as virulent clones derived from K. pneumoniae. The lack of strict association of virulent capsular types with clones was explained by horizontal transfer of the cps operon, responsible for the synthesis of the capsular polysaccharide. Finally, the reduction of metabolic versatility observed in clones Rhinoscleromatis, Ozaenae and CC82K1 indicates an evolutionary process of specialization to a pathogenic lifestyle. In contrast, clone CC23K1 remains metabolically versatile, suggesting recent acquisition of invasive potential. In conclusion, our results reveal the existence of important virulent clones associated with specific infections and provide an evolutionary framework for research into the links between clones, virulence and other genomic features in K. pneumoniae.


Klebsiella pneumoniae is responsible for a variety of diseases in humans and animals [1][3]. Most notoriously, K. pneumoniae is a prominent nosocomial pathogen mainly responsible for urinary tract, respiratory tract or blood infections [4]. Isolates from hospitals often display antibiotic resistance phenotypes [5], [6], while resistance isolates and genetic elements may also spread into the community [7], [8]. Nosocomial infections are caused by highly diverse K. pneumoniae strains that may be considered as opportunistic, rather than true pathogens, since they mostly affect debilitated patients [4]. In contrast, serious community infections due to K. pneumoniae can affect previously healthy persons. Historically, K. pneumoniae was described as the agent of Friedländer's pneumonia, a severe form of lobar pneumonia with a high mortality [9]. K. pneumoniae is still one of the leading causes of community acquired pneumoniae in some countries [10], [11]. Recently, K. pneumoniae pyogenic liver abscess (PLA), sometimes complicated by endophthalmitis or meningitis, emerged in Taiwan and other Asian countries, as well as in other continents [12][17]. Rhinoscleroma and atrophic rhinitis (also called ozaena) are two chronic and potentially severely disturbing diseases of the upper respiratory tract, associated respectively with K. pneumoniae subsp. rhinoscleromatis and K. pneumoniae subsp. ozaenae [3], [18][21]. Other K. pneumoniae infections that are severe but more rarely reported include meningitis, necrotizing fasciitis and prostatic abscess [22][24]. Finally, granuloma inguinale (donovanosis) [25] is caused by uncultivated bacteria, which may belong to K. pneumoniae [26], [27].

Factors that are implicated in the virulence of K. pneumoniae strains include the capsular serotype, lipopolysaccharide, iron-scavenging systems, and fimbrial and non-fimbrial adhesins [3], [28][31]. The abundant polysaccharidic capsule that typically surrounds K. pneumoniae protects against the bactericidal action of serum and impairs phagocytosis, and may be regarded as the most important virulence determinant of K. pneumoniae. Among the 77 described capsular (K) types of the serotyping scheme, types K1, K2, K4 and K5 are highly virulent in experimental infection in mice and are often associated with severe infections in humans and animals [1], [32][35]. K1 isolates were frequent among Friedländer's peumonia cases [1], [36] and are prominent among PLA cases, especially those with complications. Serotypes K2, K4 and K5 are frequent causes of metritis in mares and were also associated with community-acquired pneumonia [1], [36]. Isolates causing rhinoscleroma are always of type K3 [18], [27]. Finally, although their role as a direct cause of ozaena is not fully established, K. pneumoniae subsp. ozaenae isolates from cases of atrophic rhinitis are of serotype K4 or more rarely K5 [18].

In contrast with the extensive knowledge that has been gathered on the genotype-virulence relationships in the closely related species Escherichia coli and Salmonella enterica, virulent clones of K. pneumoniae remain virtually undefined [31], [37], [38]. Critically, it is unknown whether particular diseases are caused by specific clones or rather, by the expression of particular virulence determinants. This distinction is important, as virulence factors may be horizontally transferred among strains and could be weakly associated with the genomic background that harbor them, with clear implications for emergence of new pathogens and for diagnostic purposes. It is currently unknown whether capsular types characterize specific clones, in which case the K type may be useful to identify such clones and to predict the presence of other associated virulence determinants. Alternately, as is the case in e.g. Streptococcus pneumoniae [39], K types may be distributed across many unrelated clones due to frequent horizontal transfer of the cps operon, which is responsible for the synthesis of the capsular polysaccharide. In this case, a more complex picture is to be expected for the association of capsular types, other virulence determinants, and strain genomic background. More generally, the genetic structure of K. pneumoniae remains virtually unexplored [40], [41], and the phylogenetic relationships among virulent strains causing identical or distinct diseases are therefore unknown. In addition, the relationships between environmental, carriage or virulent K. pneumoniae isolates are undocumented. As a consequence, limited information on how these strains evolved to become pathogenic is currently available.

Evolution towards increased virulence can be accompanied by ecological changes that reflect specialization of pathogenic bacterial clones to their new lifestyle. For example, evolution of the particular pathogenic pattern of Shigella or Salmonella enterica serotype Typhi has been paralleled by host restriction and reduction of metabolic capabilities [42][44]. With the exception of the well-known reduced metabolic capabilities of K. p. subsp. rhinoscleromatis and K. p. subsp. ozaenae [27], it is not known whether the virulent strains of K. pneumoniae belong to ecologically specialized pathogenic clones.

The purposes of this study were (i) To determine the population genetic structure of K. pneumoniae, with a particular emphasis on the definition of virulent clones and their distinctness from other strains; (ii) To determine the extent of horizontal transfer of capsular synthesis (cps) operons among clones; and (iii) To characterize the virulent clones with respect to capsular type, other known virulence factors, experimental virulence to mice, and metabolic properties.


1. Restricted levels of genetic diversity and recombining population structure

Alignment of the seven genes sequences from 235 isolates showed no insertion/deletion (indel) in six genes. In gene tonB, one insertion of two codons (isolate SB3336) and three deletion events (one of four codons, and two of two codons) were observed. Excluding these four indels, 129 (4.3%) of the 3,012 nucleotides positions were polymorphic, four of them corresponding to tri-allelic single nucleotide polymorphisms (SNPs), thus implying a total of 133 mutations. The maximal level of nucleotide divergence among alleles ranged from 0.37% (gapA) to 1.74% (phoE), while the diversity index π (the average number of nucleotide differences per site between any two sequences chosen randomly from the study sample) ranged from 0.14% (for gapA) to 1.0% (for tonB) (Table 1). Synonymous substitutions were 12 times more frequent than non-synonymous substitutions. Despite this high degree of sequence conservation, a total of 117 haplotypes or sequence types (STs) were distinguished.

Table 1. Nucleotide polymorphism among 235 Klebsiella pneumoniae isolates.

Visual inspection of the repartition of polymorphic sites across the phylogeny of the concatenated sequence suggested that many polymorphisms have been shuffled by genetic exchange. The strong network structure obtained after split decomposition analysis (Figure 1B) confirmed the high level of incompatibility among sites, indicative of a pervasive history of intra- and/or intergenic recombination. Recombination was detected by LDhat with statistical significance in the two most polymorphic genes, tonB (r/m ratio, 22.3; p = 0.02) and phoE (r/m ratio, 18.1; p = 0.0084), indicative of intragenic recombination in these genes. Frequent polymorphisms were too scarce (no more than 1 or 2 polymorphisms above the 0.1 frequency level) in the remaining five genes to test for recombination. Evidence of homologous recombination was also provided by the observation of multiple nucleotide substitutions between STs differing by a single allele (single locus variants, or SLVs). First, ST16 and ST60 differed only by six nucleotides at gene tonB (alleles tonB-4 and tonB-8). Second, ST65 and ST243 were identical except for five SNPs between their alleles tonB-13 and tonB-25. For these two cases, import of a recombining segment, rather than independent mutations in the mismatched gene, seems compelling. In total, eight allelic mismatches (24%) between SLV pairs involved more than one SNP and are likely to result from genetic exchange. Of note, single nucleotide changes may also have been introduced by homologous recombination, given the very high sequence relatedness among most alleles. In conclusion, recombination appears frequent among housekeeping genes in K. pneumoniae.

Figure 1. Clonal diversity and relationships among 235 Klebsiella pneumoniae isolates.

A. Minimum spanning tree (MStree) analysis of multilocus sequence typing (MLST) data for 235 K. pneumoniae isolates, representing 117 sequence types (STs). Isolates of capsular serotypes K1 to K5 are colored according to serotype. Each circle corresponds to a sequence type (ST); ST number is given inside each circle. Grey zones surround STs that belong to the same clonal complex (CC), which is named according to the central ST (the likely founder of the CC). CC65-K2 is delimited by the red triangle (see text). The lines between STs indicate inferred phylogenetic relationships and are represented as bold, plain, discontinuous and light discontinuous depending on the number of allelic mismatches between profiles (1, 2, 3 and 4 or more, respectively); note that discontinuous links are only indicative, as several alternative links with equal weight may exist. The STs of reference genome strain MGH78578 (ST38) and of the type strain ATCC 13883T (ST3) are indicated. B. Split decomposition analysis of concatenated sequences of the seven genes. Numbers at the tip of branches are ST numbers. Note the bushy network structure indicative of pervasive homologous recombination. Branches were colored for the clones that are highlighted on panel A. Note the distribution into unrelated branches of strains with a given capsular (K) type.

2. Identification of virulent clones of K. pneumoniae

As expected given frequent recombination and low levels of sequence divergence, sequence-based phylogenetic analysis using PhyML [45] revealed a bushy tree (data not shown) with no conspicuous internal structure and only few strongly supported nodes. ClonalFrame [46] failed to estimate the population parameters, probably due to insufficient polymorphism. The splits decomposition network (Figure 1B) revealed no internal phylogenetic structure, with few obvious haplotype associations. In order to reveal relationships among closely related haplotypes with an approach that is less sensitive to recombination, we used the minimum spanning tree (MStree) method based upon allelic profiles. When allowing only one allelic mismatch to assign isolates to a given clonal complex (CC), 12 CCs were disclosed (the same groups were identified using eBURST [47]), while the remaining isolates were distributed into 72 singletons (Figure 1A).

Remarkably, six CCs corresponded to serotypes K1 to K4, and these CCs were characterized by their distinctive K type or pathological origins. Notably, K1 isolates were distributed into two CCs, CC23K1 and CC82K1 (Figure 1A). CC23K1 was composed of ST23 and ST57, which comprised only K1 isolates, and ST26 and ST163, which included K35 and K61 isolates. Differently, CC82K1 included only K1 isolates. Remarkably, the four K1 isolates from cases of PLA belonged to CC23K1: three isolates from Taiwan belonged to ST23 and one isolate from Zaire had ST57. Other isolates of ST23 were isolated from horses and from human blood infections (four isolates each; see Table S1). CC82K1 comprised the 15 remaining K1 isolates, none of which was involved in liver abscess. These isolates included the reference strain of serotype K1 (A5054, ST82) as well as 11 isolates from blood or respiratory infection in France between 1976 and 1984.

Likewise, all K2 isolates except CIP 52.204 (ST86) could be grouped into two distinct and apparently unrelated groups (Figure 1A). First, CC14K2 comprised STs 14, 78 and 80, all being composed of isolates of serotype K2, and also included some K24 isolates (ST15). ST14 included ESBL-producing isolates from Curaçao [48] and isolates collected in France and Italy between 1981 and 2002 from urinary, respiratory, blood and cerebro-spinal fluid (CSF) infections. A second group of K2 isolates was composed of STs 65, 25 and 243, which together formed one CC, and of ST66, which differed from ST65 by only two genes (infB and rpoB, one SNP each). ST65 includes an isolate from a cat infection, one isolate that caused an epidemics in monkeys at a French zoo [49], and one human clinical isolate from an anal abscess, while ST25 corresponds to a nosocomial blood isolate (Table S1). ST66 corresponds to the reference strain of capsular serotype K2 (B5055) and the virulent strain CIP 52.145 [34], as well as to isolate 675, which was used as a vaccine in animals (Table S1). For simplicity, and given the genetic and phenotypic similarity of ST66 strains with ST65 strains, we will consider ST66 as a member of CC65K2, even if ST66 does not belong to CC65K2 sensu stricto.

Sets of isolates which belong to a single clonal complex and share many other common features including K type, virulence factor content and metabolic profile (see below), likely descend from a common ancestral strain from which they inherited their common properties. Therefore, CC23K1, CC82K1, CC14K2 and CC65K2 may be regarded as four distinct clones. However, their precise demarcation is rendered difficult, based on MLST data alone, by the high degree of allele sharing with other K. pneumoniae STs, possibly due to their recent evolutionary emergence or to ongoing allelic exchange with other STs (see discussion).

All K. pneumoniae subsp. rhinoscleromatis isolates were identical at the seven genes (ST67) except one isolate (839, France, 1982), which had a single SNP in tonB, resulting in ST68 (Figure 1A). ST67 included CIP 52.210T, the type strain of K. pneumoniae subsp. rhinoscleromatis, and C5046, the reference strain of capsular serotype K3. In addition, ST67 included nine unrelated clinical isolates from rhinoscleroma cases, isolated from six countries between 1954 and 2003. Clearly, subspecies K. pneumoniae subsp. rhinoscleromatis is highly homogeneous and appears to correspond to a single clone, which we refer to as clone Rhinoscleromatis. Interestingly, all Rhinoscleromatis isolates differed from all other strains, including 10 non-rhinoscleroma isolates with serotype K3, by four or more allelic mismatches. MLST thus clearly demarcates Rhinoscleromatis isolates from all other K. pneumoniae members. However, it is important to stress that Rhinoscleromatis clearly belongs to a single genetic pool together with K. pneumoniae (Figure 1B): the average genetic distance between Rhinoscleromatis and the 115 other STs is 0.54%, while distances among the 115 STs ranged from 0.033% to 0.70%.

All K. pneumoniae subsp. ozaenae isolates formed a single clonal complex, CC91oz. ST91 could be inferred as the genetic founder of this clone (Figure 1A), as all other STs of CC91oz differed from ST91 by a single mismatch, whereas they differed among them by two mismatches, with the single exception of the pair ST95 and ST96. ST91 included CIP 52.211T, the type strain of K. pneumoniae subsp. ozaenae, and the K4 reference strain D5050, as well as the reference strain of type K5 (CIP 52.212 = E5051) and two clinical isolates. The other STs of CC91oz included K4 clinical isolates from ozena cases and blood infections, as well as two isolates from patients with granulomas (ST90 and ST96). Isolates from ozaena cases were distributed in the three genotypes ST90, ST91 and ST95. These results indicate that all K. pneumoniae subsp. ozaenae isolates can be considered as descending from a single ancestor, forming clone Ozaenae. This clone was well demarcated from the remaining isolates, as there was only one K. pneumoniae strain (SB169-2, ST97, C-pattern C16a) that had only two allelic mismatches with clone Ozaenae, while all other K. pneumoniae isolates, including three non-Ozaenae K5 isolates, had at least four mismatches with any member of clone Ozaenae. Of note, clone Ozaenae (6 STs) is more heterogeneous than clone Rhinoscleromatis (2 STs) based on the present strain collection, possibly reflecting a more ancient evolutionary emergence and/or a more rapid diversification.

3. Capsular types are not strongly associated with genomic background

No close phylogenetic relatedness was apparent between the two K1 groups, between the two K2 groups, and between clones Rhinoscleromatis and Ozaenae with other K. pneumoniae strains of serotypes K3, K4 or K5 (Figure 1), indicating an independent origin in distinct genomic backgrounds, rather than a common ancestral origin. Two evolutionary mechanisms could result in identical K-types being distributed in unrelated genomic backgrounds: horizontal transfer of the cps operon, or evolutionary convergence. In the latter scenario, similar capsular polysaccharide antigenic structures would be synthesized by phylogenetically unrelated cps operons that are functionally identical. In order to estimate the phylogenetic relatedness of cps operon structures, isolates were analyzed by PCR-RFLP of the cps operon [50], which can disclose unrelated C-patterns among isolates of a given K type. The C-pattern of 211 isolates could be established (Table S1; C-patterns available upon request). Clearly, indistinguishable or highly similar C-patterns were observed in unrelated MLST genotypes (Figure 2): K1 isolates from both CC23K1 and CC82K1 had C-pattern C1a, while K2 isolates of CC14K2 and CC65K2 exhibited the highly similar patterns C2b to C2e (CC14K2) and C2a (CC65K2). All K3 isolates from clone Rhinoscleromatis and from the 10 K3 K. pneumoniae isolates in five other STs, had C-pattern C3a or the highly similar patterns C3b to C3d (Figure 2). In particular, C3a was observed in all Rhinoscleromatis isolates as well as in the unrelated ST3 and ST13. The three variant K3 C-patterns were observed in ST8 (C3c), ST71 (C3b) and ST153 (C3d). Likewise, the C-pattern C5a was observed in clone Ozaenae K5 isolates and in K. pneumoniae K5 isolates (ST60, ST61 and ST149), which do not appear phylogenetically related (Figure 1). Altogether, these data are suggestive of several independent historical events of horizontal transfer of the cps operon between isolates belonging to distinct clones.

Figure 2. Distribution of related capsular operon regions in unrelated clones.

The splits tree represents the relationships among gnd alleles as obtained after split decomposition analysis. The distribution of the gnd alleles found in isolates and reference strains of capsular serotypes K1 to K4 is indicated by black coloration of sequence types (STs) in the MStree of the corresponding insets. Below the MStree displays are represented the C-pattern of the corresponding isolates. Note that similar or identical gnd and C-patterns are distributed in unrelated STs.

In order to fully demonstrate suspected cases of cps region horizontal transfer, we sequenced in 177 relevant isolates, a 360-nt internal portion of gene gnd, which genomic location is just adjacent of the cps operon [50], [51]. A high level of nucleotide polymorphism was encountered (Table 1), with 136 (38%) polymorphic sites, with no indel. Thirteen isolates had a gnd sequence that differed from the 164 other sequences by 6% to 18%. The remaining 164 sequences were still much more variable than the seven MLST genes. LDhat analysis demonstrated a strong intra-genic recombination pattern (r/m ratio = 12.5, p = 0.008), consistent with the highly reticulated structure obtained using SplitsTree (Figure 2). However, despite frequent intra-genic recombination, gnd alleles were remarkably similar for isolates of the same K type (Figure 2), indicating that the association between the gnd gene and the cps operon was not broken down.

Notably, the gnd sequences of K1 isolates from CC23K1 and CC82K1 were undistinguishable (gnd-12 in both CCs, and gnd-11 in two strains of ST82, differing at a single SNP from gnd-12). Similarly, the gnd sequence of Rhinoscleromatis isolates (alleles gnd-42, gnd-45 and gnd-46) were either identical or highly similar (Figure 2) to those of K. pneumoniae K3 isolates (gnd-42, gnd-43 and gnd-44), demonstrating their recent common ancestry and their horizontal transfer into distinct genomic backgrounds. Likewise, K5 isolates of clone Ozaenae and K5 K. pneumoniae isolates (ST60 and ST61) had identical gnd sequences (gnd-5). Together with identity of C-patterns, these data demonstrate a common evolutionary origin of the cps-gnd region in Rhinoscleromatis and K3 K. pneumoniae isolates, in Ozaenae and K5 K. pneumoniae isolates, and in both K1 groups. Horizontal gene transfer of entire gnd-cps region is the most likely explanation for the current distribution of cps-gnd regions with a unique origin in distinct genomic backgrounds. The transfer of the gnd-cps region could be inferred for other K types as well (data not shown).

Different from the above, gnd sequences in K2 isolates of CC14K2 (gnd-38) and CC65K2 (gnd-16 or gnd-17) were unrelated (Figure 2). Because the C-pattern from these two CCs are highly related (Figure 2), it is likely that the cps operon was transferred horizontally without the gnd gene, or that the gnd gene was replaced in one of the CCs subsequently to a cps-gnd co-transfer event. However, as allele gnd-38 was also observed in K2 strain CIP 52.204 (ST86), horizontal transfer of the cps-gnd region has occurred between this ST and CC14K2.

It is remarkable that the gnd allele in K4 isolates of clone Ozaenae was the same as observed in most K1 isolates (gnd-12). This is fully consistent with the observation that K1 and K4 C-patterns (Figure 2) are highly similar [50] and indicates a close evolutionary relationships of the cps operons that determine K1 and K4 serotypes.

4. Virulence is associated with clone, rather than with K-type

In order to determine whether the above-identified clones differ by their virulence potential, the presence of 10 genetic factors implicated in Klebsiella virulence was assessed by PCR. A total of 102 representative isolates were characterized (Table 2). While the three genes uge, wabG and ureA gave a positive PCR reaction in all isolates, other factors showed unequal repartition across CCs, resulting in distinctive virulence factor fingerprints of major CCs (Table 2). Notably, we found sharp differences in virulence gene content between CC23K1 and CC82K1, as well as between CC14K2 and CC65K2. Consistent with the location of magA within the cps operon of K1 isolates [52], both K1 groups were magA positive, while magA was not detected in any other isolate. However, CC23K1 differed from CC82K1 by the presence (100% vs. 0%, respectively) of genes mrkD coding for the type 3 fimbriae adhesin, which facilitates adhesion to the basement membranes of several human tissues [53], [54], and allS, coding for the activator of the allantoin regulon [30]. Interestingly, allS was specific for K1 isolates of CC23K1 members, as it was undetected in CC82K1 and in non-K1 members of CC23K1 (ST26-K61 and ST163-K35). CC23K1 was also characterized by a higher prevalence (80%) of non-fimbrial adhesin CF29K [55], whereas CC82K1 and most other isolates were negative.

Table 2. Virulence gene content of Klebsiella pneumoniae clones (a).

The two K2 groups CC14K2 and CC65K2 also differed by their virulence gene content. Particularly, isolates of CC14K2, including its K24 members, were all positive for the iron uptake marker kfu [31], whereas all CC65K2 isolates were negative. In contrast, rmpA, the regulator of mucoid phenotype [56], was undetected in CC14 whereas rmpA PCR was positive in 71% of CC65K2 isolates (Table S1).

Clone Rhinoscleromatis was characterized by the complete absence of kfu and the presence of rmpA. These characteristics also distinguished Rhinoscleromatis from other K3 K. pneumoniae isolates (Table S1). Ozaenae isolates shared the unique property, together with CC82K1, of being negative for mrkD (except for one isolate).

To determine whether the two K1 clones and the two K2 clones differ in their virulence, four to nine strains per clone were tested in mice (Table S1). There was a clear difference in the virulence of CC14K2 and CC65K2, as no strain (0 out of nine) of the former was lethal, whereas four out of six CC65 strains killed mice after five days. The two avirulent CC65 strains were either rmpA negative (as were all CC14 K2 strains) or negative for fim and mrkD (Table S1). Likewise, out of seven CC23K1 strains, four K1 strains (ST57 and three of ST23; all rmpA positive) were lethal to mice. The three avirulent strains were one ST23 K1 strain and the two non-K1 strains of ST26 and ST163; these three strains lacked rmpA. In contrast, of the four CC82K1 strains assayed, only one was slightly virulent to mice, even though rmpA PCR was positive (Table S1). Hence, virulence to mice of K1 and K2 strains appeared to differ, depending on the clone they belonged to.

5. Metabolic versatility and evolution of virulent K. pneumoniae clones

In order to determine whether virulent clones of K. pneumoniae are truly in the process of adapting to a pathogenic lifestyle, rather than simply representing classical K. pneumoniae strains with particular combinations of virulence factors, the ability to utilize 99 carbon sources was compared between representative isolates of the virulent clones and other K. pneumoniae isolates (Figure 3; Table 3). A total of 32 substrates were either utilized by all isolates (n = 16) or by none (n = 16, Figure 3 legend); some of these substrates are useful for identification of the K. pneumoniae species [27]. However, the remaining substrates showed differences among K. pneumoniae strains. Interestingly, the pattern of carbon source utilization correlated closely with MLST-defined clones (Figure 3; Table 3). Clone Rhinoscleromatis showed a restricted substrate utilization pattern, with the distinctive loss of the ability to use seven substrates, including D-glucuronate and D-galacturonate (0% vs. 100%) and protocatechuate, an intermediate in the degradation of lignin. Isolates of CC82K1 also had a clearly distinctive pattern, in particular with the loss of L-fucose, D-(+)malate and succinate utilization. Clone Ozaenae isolates exhibited three groups of metabolic profiles (A, B and C on Figure 3), each of these consisting of the loss of a number of substrates, with trans-aconitate in common. Finally, the remaining isolates formed a large group that comprised K1 isolates of CC23K1, showing that these can be differentiated from CC82K1 by the utilization of several carbon sources (Table 3). In addition, CC23K1 were almost exclusive among K. pneumoniae isolates in using dulcitol and D-tagatose as sole carbon source, while they differed from the remainder of the large cluster by the loss of benzoate utilization. Differently, CC14K2 and CC65K2 both belonged to the large biotype cluster and were weakly distinguished, although L-sorbose utilization was found only in CC65.

Figure 3. Biotype profiling of K. pneumoniae clones.

Cluster analysis (simple matching coefficient) of K. pneumoniae isolates and reference strains based on metabolic profiles as assessed by biotype-100 strips. Codes above the column correspond to substrate code (Table S3). A blue square means the strain grew on the corresponding substrate as sole carbon source. Dark blue, growth was observed after two days; light blue, growth observed after four days. Note the strong homogeneity of biotype-100 profiles within clones. Three clones (Ozaenae, Rhinoscleromatis and CC82 K1) have lost the ability to utilize a number of substrates, including some common substrates between the three clones (see text). Note that three tests measure coloration, not growth: hydroxyquinoline-beta-glururonide (black color), tryptophane (brown color: hydrolysis into indole-pyruvic acid) and histidine (red color). The following substrates were utilized by all assayed strains: D-Glucose, D-fructose, D-trehalose, D-Melibiose, D-Raffinose, Maltotriose, Maltose, D-Cellobiose, 1-O-Methyl-B-D-glucoside, D-Arabitol, Glycerol, Adonitol, N-Acetyl-D-glucosamine, D-Gluconate, L-Alanine and L-Serine. The following substrates were always negative: hydroxyquinoline-beta-glucuronide, D-Lyxose, i-Erythritol, 3-O-Methyl-D-glucose, Tricarballylate, Tryptophan, Gentisate, 3-Hydroxybenzoate, 3-Phenylpropionate, Trigonelline, Betaine, Caprylate, Tryptamine, Itaconate, Propionate, 2-Ketoglutarate.

Table 3. Utilization of carbon sources by Klebsiella pneumoniae clones (a).

On average, isolates of the largest cluster were able to utilize more carbon sources (65±3), whereas isolates of clone Rhinoscleromatis were those with the lowest metabolic abilities (47±2.4) (Table 3). CC82K1 and Ozaenae isolates (groups A, B and C together) used 48±2.2 and 50±5.9 substrates, respectively. It was striking that several substrates were lost in common by the three metabolically-restricted clones. For example, D-Malate and succinate were lost by Ozaenae (group A) and CC82K1, trans-aconitate was lost by Ozaenae groups A and B and by CC82K1, 1-O-Methyl-a-D-glucoside and lactulose were lost by CC82K1 and Rhinoscleromatis, whereas several substrates (e.g. 5-aminovalerate) were lost by the three groups. The loss of the same metabolic abilities indicates convergent evolution in these clones, possibly indicative of parallel specialization to a similar niche.


The population of K. pneumoniae appears to be characterized by a low level of nucleotide divergence among orthologous genes, contrasting with related species such as S. enterica and E. coli. This restricted polymorphism cannot be attributed to a biased sampling, as our dataset included isolates from the environment and animals, in addition to human isolates from different clinical sources and large geographic and temporal scales. The genetic distance that separates K. pneumoniae from its closest phylogenetic relatives (KpII and KpIV [40], [57]) calculated based on the same seven genes is nearly 4%. Therefore, the species K. pneumoniae may have undergone a bottleneck relatively recently, long after its separation from its closest relatives. Still, K. pneumoniae is much more diverse than notorious monomorphic pathogens such as Y. pestis or S. enterica serotype Typhi [58], [59].

A high number of distinct genotypes were disclosed by MLST despite restricted nucleotide polymorphism. Our analyses suggest that homologous recombination has more impact on sequence evolution than mutation, although it is difficult to obtain a reliable estimate of the recombination/mutation ratio with such a low level of polymorphism. A high recombination rate would shuffle polymorphisms among clones and lineages, generating many genotypes that can be distinguished by MLST. As a consequence, the clonal frame of K. pneumoniae clones will diversify more rapidly than it would by a purely mutational process, and the disclosed STs may not be highly stable over long periods of time.

Gene gnd was atypical by its high level of polymorphism. Because this gene is located between the rfb and cps operons responsible for the synthesis of the two major surface polysaccharides, the lipopolysaccharide and capsule, its evolution is probably highly influenced by the likely positive selection operating at these two neighboring loci, as demonstrated for E. coli or Salmonella [60], [61]. In addition, exchange of the cps operon between E. coli and K. pneumoniae was reported [60], and the divergent gnd alleles encountered in the present study clearly indicate incorporation into K. pneumoniae isolates of nucleotide sequences from other Enterobacteriaceae species.

Determining the phylogenetic relationships within a recombining species is difficult and may even be meaningless if recombination has erased the pattern of descent among strains. In particular, analysis based on allelic profiles can be misleading and may result in the clustering of unrelated STs into long straggly chains of genotypes [62]. One can therefore be suspicious about the true clonal link between isolates of CC17, which consists of chains of STs with distinct K-types, with the exception of some possibly meaningful terminal groupings such as three K5 STs (Figure 1).

Identification of clones within species with high rates of recombination is possible if these clones spread in the population [63], [64]. The fact that several clonal complexes disclosed herein are relatively homogeneous with respect to several features including K type, virulence factor content and metabolic profile, demonstrates that they correspond to clones, i.e. descend from a common ancestral strain from which they inherited their common properties. So far, K serotyping has been the dominant common language for recognition of related Klebsiella strains in epidemiological and virulence studies, but it was unknown whether isolates with the same K type belonged to single clones. Our data clearly reject this simple view. Indeed, most K-types (with the exception of K4) that were represented by several isolates were dispatched in unrelated STs. We could show that the shared K type resulted from horizontal transfer of the cps operon among these unrelated genotypes, generally with the co-transfer of the adjacent gnd gene. Therefore, knowledge of the K type provides unreliable prediction of clone identity. Given their close physical linkage, recombination between gnd and cps is probably unfrequent, and gnd sequencing could therefore be used as a proxy for K typing, which is technically demanding [50], [65], [66]. However, the finding of unrelated gnd sequences in the two major CCs of K2 isolates shows that this method would not be totally reliable.

Isolates with serotypes K1 to K4 were preferentially included in this study; therefore, our isolate collection does not reflect K type frequency in natural populations. Our selected collection allowed the discovery of six clones comprising isolates that are considered as particularly pathogenic based on clinical features in animal and humans and on experimental evidence [1], [32][35]. Our data provide the first evidence that the agent of rhinoscleroma on the one hand, and isolates recovered from cases of ozaena on the other hand, each correspond to a single clone. It is remarkable that these highly homogeneous clones include isolates that were isolated over a time span of several decades from several countries in Asia, Africa and Europe. Hence, these two pathogens, both involved in chronic infections, can be viewed as monomorphic pathogens, similar in this respect to e.g. Mycobacterium leprae [67]. Nevertheless, isolates of clone Ozaenae appear to be slightly more heterogeneous than Rhinoscleromatis based on MLST data, K type and biotype. Ozaenae isolates have also been implicated in distinct types of infections such as bacteremia, urinary tract infections [68] or splenic abscess [69], and were variable for the presence of several virulence factors. These observations may reflect a more diverse lifestyle for Ozaenae than for the intracellular human-restricted pathogen Rhinoscleromatis.

K. ozaenae” and “K. rhinoscleromatis” could not be separated from K. pneumoniae by DNA relatedness [70]. For this reason, K. ozaenae and K. rhinoscleromatis were treated as subspecies of K. pneumoniae in the early editions of the Bergey's Manual [18], [71]. However, these two clones appear to have evolved from the genetic pool taxonomically regarded as K. pneumoniae subsp. pneumoniae, which does not form a phylogenetic lineage distinct from the other two subspecies (this study and [40]). Therefore, it is appropriate to consider isolates associated to rhinoscleroma and ozaena as clones of K. pneumoniae that acquired particular pathogenic properties, rather than separate phylogenetic entities that deserve subspecies status. Our data do not indicate a close affiliation of clone Rhinoscleromatis with clone Ozaenae. The uncultivable agent of donovanosis, or granuloma inguinale, has been included in the genus Klebsiella as K. granulomatis [26], [72]. Its phoE sequence [26], allele phoE-1, was encountered in several K. p. subsp. pneumoniae STs (including CC14K2), and is distinct from phoE-15 found in Rhinoscleromatis. Despite the similarities in the pathologies they cause [21], [25], [26], it was thus not possible to equate K. granulomatis with clone Rhinoscleromatis, but a close evolutionary link cannot be excluded. In any case, phoE data indicate that K. granulomatis does not represent a distinct genomic species, but instead belongs to K. pneumoniae as well.

This study demonstrates for the first time that K1 isolates that cause PLA are genetically distinct from K1 isolates from cases of respiratory infections and septicemia. Even though our identification of CC23K1 as the only clone associated with PLA is based on only four PLA-causing isolates, this result is fully consistent with a previous report based on a worldwide collection [38]. Recent progress stimulated by the emergence of K. pneumoniae PLA has provided important clues as to the bacterial factors involved in this infection [30], [31], [52], [73]. Our PCR tests show that among the genetic factors that have been associated with K1 PLA strains, only allS appears to be totally specific for this pathogen. In addition, we show for the first time that allS is not universally present in K1 strains [30], [74]. Our data provide the novel observation that CF29K is particularly prevalent in this clone. CF29K corresponds to adhesin CS31A found in E. coli strains and involved in human diarrhea and in septicemia in calves [55]. Our data suggest that this factor could either be directly implicated in the pathogenesis of PLA, or linked to another pathogenicity factor on the 185 kb plasmid that harbors gene cf29A [75]. In agreement with others [76], we found that magA is present in K1 isolates not involved in PLA and should thus not be considered as a marker of PLA-causing isolates [52]. The respiratory or blood origin of most CC82K1 isolates, together with previous reports of the frequent implication of K1 strains in Friedländer's pneumonia, is consistent with this clone being a prominent agent of this severe form of pneumonia. The existence of two K1 groups that differ by their pathological potential is of high relevance for understanding the bacterial determinants of PLA and acute pneumonia caused by K. pneumoniae. K2 isolates also clearly appear to be distributed into several unrelated genotypes. For both K1 and K2 serotypes, we could show that the pathogenic potential of strains depends on their genotype, rather than on their K type. CC14K2 comprises isolates for serotypes K2 and K24, but we did not observe any difference in virulence gene content or in virulence to mice between CC14K2 members of both serotypes. In contrast, some virulence factors distinguished CC14K2 (including its K24 isolates) and CC65K2, e.g. gene kfu (100% vs. 0%, respectively). These results show that at least for K1 and K2 isolates, the clonal complex is a better predictor of virulence gene content and of virulence to mice than the K type, and that previous associations of virulence factors with K-types [74], [77] should be revisited by analysis of isolates from distinct CCs. Thus, even if the capsular polysaccharide is a prominent pathogenicity determinant, the long-held belief that K type is predictive of virulence should be discontinued.

The nature of K. pneumoniae pathogenic clones and their history of interaction with their animal and human hosts, including possible specialization, remain largely unknown. Our biotype data demonstrated that the three clones Rhinoscleromatis, Ozaenae and CC82K1 have each lost several metabolic abilities, some of which in common, probably by parallel evolution. It has long been recognized that the two former clones are biotypes of K. pneumoniae with less nutritional versatility [18], [27] and together with some K1 strains, require specific factors for growth [27]. To our knowledge, Rhinoscleromatis, Ozaenae and K1 isolates have neither been reported from the environment, nor in intestinal carriage, and it is perhaps significant that several substrates that are not utilized by these clones belong to plant product degradation pathways. We hypothesize that these three clones are engaged in evolutionary specialization to a restricted ecological niche, possibly represented by the upper respiratory tract of humans. A restriction in ecological niche may in turn reduce the opportunity for encounter with other K. pneumoniae strains. The intracellular lifestyle of Rhinoscleromatis provides the most achieved example, and this pathogen may now be evolving independently from its ancestral species K. pneumoniae. Consistent with this hypothesis, Rhinoscleromatis and Ozaenae were genetically the most distinct of the 117 STs (Figure 1), which had in general no more than three allelic mismatches among themselves. This observation suggests that these two clones are less frequently involved in allelic exchange with other strains. Finally, it is interesting to notice that the gene coding for the adhesin MrkD was undetected specifically in CC82K1 and Ozaenae, also suggestive of niche reduction. In contrast, the typical biotype profile of the PLA-associated CC23K1 does not suggest ecological specialization. Hence, the acquisition by this clone of its particular set of virulence determinants is possibly recent in time, consistent with epidemiological data [12], and the pathogenicity of clone CC23K1 may be uncoupled from any particular adaptation to humans. Infection of the liver is believed to take place from the intestine. Because liver infection and metastasis to the eye and brain are unlikely to provide any specific selective advantage to this clone, pathogenesis can be viewed as accidental. Given that the natural habitat of this clone is probably indistinct from its non-virulent ancestor, keeping an intact metabolic versatility may be a key requirement for successful competition of this clone with other generalist K. pneumoniae clones. It is therefore unlikely that reductive evolution by specialization will be observed in this important emerging clone.

Materials and Methods

Bacterial isolates

A total of 235 K. pneumoniae reference strains or isolates were included in this study (Table S1). Capsular (K) serotypes K1 to K4 were included preferentially in order to estimate their genetic diversity. The collection included 25 isolates with serotype K1 from cases of pyogenic liver abscess (n = 4), other clinical sources (n = 17) and reference strains (n = 4). Nineteen K2 isolates, 16 K. pneumoniae subsp. rhinoscleromatis isolates (all being K3) and 14 K. pneumoniae subsp. ozaenae (12 K4 and 2 K5) were included. For comparison purposes, we included K3 (n = 10) and K5 isolates (n = 4) of K. pneumoniae subsp. pneumoniae. Type strains of the three subspecies and reference strains of serotypes K1 to K5 as well as laboratory strain KP52.145, were included. In some cases (Table S1, column ‘probable duplicate’), two or more subcultures of the same original strain were included, because they were obtained from different sources (e.g., the Orskov collection of K-type reference strains, the Collection de l'Institut Pasteur [CIP] and the ATCC). This is due to the fact that the K-type reference strain and the taxonomic type strain or other laboratory strains are sometimes derived from the same initial strain.

The remaining isolates were included to represent different sources, without consideration of their K type. Most isolates were of human clinical origin. For comparison purposes, we collected 13 isolates from the environment and 18 from fecal samples using a selective medium based on citrate and inositol [78], and gathered 30 horse isolates and 8 other animal isolates from previous studies [49], [79][81]. The 67 isolates from nosocomial infections previously analyzed [82] were included. Isolates originated from 20 countries from Europe, North America, Asia and Africa. The most represented countries were France and the Netherlands (Table S1).

Species and subspecies identification and biotyping

Isolates were initially identified as Klebsiella pneumoniae sensu stricto, K. pneumoniae subsp. rhinoscleromatis or K. pneumoniae subsp. ozaenae using standard, recommended biochemical tests [27]. Identification of the later two subspecies was controlled by capsular serotyping using the capsular swelling method and using the following biochemical tests: Voges-Proskauer, urease, ONPG, lysine decarboxylase, citrate, malonate, and gas production. Isolates could be identified as belonging to K. pneumoniae sensu stricto, i.e., phylogenetic group KpI [40], based on phylogenetic clustering of the seven MLST genes and gyrA [40], using KpII-A, KpII-B [57], KpIII [40] and K. oxytoca, K. planticola and K. terrigena isolates for comparison.

Re-identification and biotyping of isolates at the species level (or subspecies level within K. pneumoniae) was performed using Biotype-100 strips (BioMérieux, Marcy l'Etoile, France), which contain 99 substrates in cupules [27]. Minimal medium 1 was used and isolates were identified using software Recognizer (P.A.D. Grimont, Institut Pasteur) against the Enterobacteriaceae database constructed in the laboratory (version 2000). Substrates that were particularly useful for species discrimination were m-coumarate, gentisate, histamin, 3-hydroxybenzoate, D-melezitose, 3-O-methyl-D-glucose, and tricarballylate [27]. Minimal medium 2 was used for isolates of K. pneumoniae subsp. rhinoscleromatis or K. pneumoniae subsp. ozaenae [27]. Reproducibility of biotype-100 profiles was controlled by inclusion of strain ATCC 13883T in each batch and by the independent analysis of synonymous strains (Table S1; Figure 3).

Capsular serotyping

Serotyping was determined by the capsular swelling method [49], [79], [80], and the K-type of some isolates were controlled by the agglutination method [65]. The K-serotype of the type strains and reference strains was known prior to this study.

cps PCR-RFLP (molecular serotyping)

The determination of the C-pattern was determined as previously described [50]. A reference C-pattern database was constituted by the C-patterns obtained for the 77 reference strains of the International serotyping scheme and for the study isolates for which the K-type was determined by classical serotyping. C-patterns that were encountered in isolates of defined capsular type were labeled with ‘C’ followed by the number capsular (e.g., C2) followed by a letter denoting the successive banding patterns found for isolates of this serotype (e.g., C2a, C2b …). Some isolates were analyzed by cps PCR-RFLP but not by classical serotyping. For the C-patterns that had a match in the reference database, the same K-type as that of the reference strain(s) was inferred. For the remaining isolates, the obtained C-pattern had no match in the reference database; these C-patterns were numbered consecutively, starting at C100 (Table S1).

Multilocus Sequence Typing (MLST)

MLST was performed as previously described [82] with the following modification: universal sequences were added upstream of each forward (GTT TTC CCA GTC ACG ACG TTG TA) and reverse (TTG TGA GCG GAT AAC AAT TTC) primers. All amplifications were performed at 50°C, and sequences were obtained using the two universal sequencing primers given above. Further details are available on the K. pneumoniae MLST web site (

gnd gene sequencing

The sequence of a 360-bp portion of the gnd gene was established on both strands by using primers gnd-1F (TGA AGC AGC AAA CAA AGG TAC) and gnd-8R (TCA TCG GCG ATC TGC TTA AAG T), which amplify an internal portion of 457 bp of the gene. The annealing temperature was 46°C (30 cycles of 30 sec, 94°C; 30 sec., 46°C; 30 sec., 72°C, followed by 1 min at 72°C). When amplification failed, primer gnd-2 (ACA TCA CGC AGC GCC TGC TGA T) was used instead of gnd-8R, with 50°C as annealing temperature. Sequencing primers were gnd-9, TGA TGA (A/G)GC nGC (A/c)AA CAA AGG TAC, and gnd-10, TCA TCa GC(a/G) ATC TG(C/t) TTG AAG Ta(c/t).

Virulence PCR

PCR assays were performed to check for the presence of 10 genes that have previously been associated with virulence in K. pneumoniae. Target genes, primers used and specific annealing temperature of PCR are given in Table S2. After 5 min at 94°C, there were 35 cycles of 94°C, 30 sec.; annealing temperature, 30 sec.; and 72°C, 1 min. followed by a final elongation of 1 min at 72°C. Strains NTUH-K2044, KP52145 and MGH 78578 [31], [34], [52], [83] were used as PCR controls. PCR products from several STs were systematically sequenced to control that the amplified PCR products corresponded to the expected gene.

Infection of mice

Female Balb/cJ mice were purchased from R. Janvier Breeding Center (Le Genest St. Isle, France) and housed under standard conditions of feeding, light and temperature with free access to food and water. Experiments were performed according to the Institut Pasteur guidelines for laboratory animals husbandry. Seven to eight weeks-old mice were first anesthetized, with 80 microliters intramuscular injection of ketamine (Imalgene, 31.25 mg/kg, Merial) and Acepromazine (Calmivet, 1.5 mg/kg, Vetoquinol) and then infected by inoculation of 20 microliters of bacteria suspension (106 bacteria) into their right nostril. Eight mice per strain were infected. The number of surviving mice was monitored every day during twelve days.

Data analysis

For each MLST locus, an allele number was given to each distinct sequence variant (confirmed by at least two chromatogram traces), and a distinct sequence type (ST) number was attributed to each distinct combination of alleles at the seven genes. Allele and profile numbers were incremented successively in the order in which they were discovered. In order to define the relationships among isolates at the microevolutionary level, we performed allelic profile – based comparisons using a minimum spanning tree (MStree) analysis with the BioNumerics v5.10 software (Applied-Maths, Sint Maartens-Latem, Belgium). MStree analysis links profiles so that the sum of the distances (number of distinct alleles between two STs) is minimized [84]. Isolates were grouped into clonal complexes (clonal families), defined as groups of profiles differing by no more than one gene from at least one other profile of the group [85]. Accordingly, singletons were defined as STs having at least two allelic mismatches with all other STs.

Split decomposition analysis was performed using SplitsTree version 4.10 [86], [87]. Neighbor-joining tree analysis was performed using MEGA v4 [88]. Nucleotide diversity indices were calculated using DNAsp v4 [89]. ClonalFrame analysis [46] was performed with 50,000 burn-in iterations and 100,000 subsequent iterations.

The relative contribution of recombination and mutation on the short term was calculated using eBURST and the clonal diversification method [90], [91]. For each pair of allelic profiles that differ by a single allelic mismatch (single locus variants, or SLVs), the number of nucleotide changes between the alleles that differ is counted. A single nucleotide difference is considered to be likely caused by mutation, whereas more than one mutation in the same gene portion is considered to derive from recombination, as it is considered unlikely that two mutations would occur on the same gene while the other genes remain identical. No correction was made for single nucleotide differences possibly introduced by recombination.

The population recombination rate was estimated by a composite-likelihood method with LDhat [92]. LDhat employs a parametric approach, based on the neutral coalescent, to estimate the scaled parameter 2Ner where Ne is the effective population size, and r is the rate at which recombination events separate adjacent nucleotides. The crossing-over model L was used for the analysis of biallelic sites, with frequency of the less frequent allele >0.1.

Nucleotide sequences

Sequences generated in this study are available at for the seven MLST genes. In addition, gnd alleles have been deposited in GenBank/EMBL/DDBJ databases under the accession numbers FJ769917-FJ769969.

Supporting Information

Table S1.

Strains. Characteristics of the 235 strains included in the study.

(0.13 MB XLS)

Table S2.

Primers used for virulence genes PCR.

(0.04 MB XLS)

Table S3.

Carbon sources assayed with the Biotype-100 strips.

(0.01 MB PDF)


We are grateful to Engeline van Duijkeren, J. Verhoef, A. Fluit, Jin-Town Wang, Etienne Carbonelle and Christophe De Champs for providing isolates. The Collection de l'Institut Pasteur is acknowledged for providing reference and type strains. Our study benefited greatly from the historical Klebsiella collection of the Institut Pasteur Enterobacteriaceae Unit, gathered by the late Claude Richard.

Author Contributions

Conceived and designed the experiments: SB CF RT PG. Performed the experiments: SB CF VP SIJ RT LD. Analyzed the data: SB CF RT. Wrote the paper: SB.


  1. 1. Ørskov I (1981) The genus Klebsiella (Medical aspects). In: Starr MP, Stolp H, Truper HG, Balows A, Schlegel HG, editors. The Prokaryotes: a handbook on habitats, isolation, and identification of bacteria. Berlin: Springer-Verlag. pp. 1160–1165.
  2. 2. Seidler RJ (1981) The genus Klebsiella (Nonmedical aspects). In: Starr MP, Stolp H, Truper HG, Balows A, Schlegel HG, editors. The Prokaryotes: a handbook on habitats, isolation, and identification of bacteria. Berlin: Springer-Verlag. pp. 1166–1172.
  3. 3. Brisse S, Grimont F, Grimont PAD (2006) The genus Klebsiella. In: Dworkin M, Falkow S, Rosenberg E, Schleifer K-H, Stackebrandt E, editors. The Prokaryotes · A Handbook on the Biology of Bacteria. 3rd edition ed. New York: Springer.
  4. 4. Podschun R, Ullmann U (1998) Klebsiella spp. as nosocomial pathogens: epidemiology, taxonomy, typing methods, and pathogenicity factors. Clin Microbiol Rev 11: 589–603.
  5. 5. Paterson DL, Ko WC, Von Gottberg A, Mohapatra S, Casellas JM, et al. (2004) International prospective study of Klebsiella pneumoniae bacteremia: implications of extended-spectrum beta-lactamase production in nosocomial Infections. Ann Intern Med 140: 26–32.
  6. 6. Woodford N, Reddy S, Fagan EJ, Hill RL, Hopkins KL, et al. (2007) Wide geographic spread of diverse acquired AmpC beta-lactamases among Escherichia coli and Klebsiella spp. in the UK and Ireland. J Antimicrob Chemother 59: 102–105.
  7. 7. Colodner R, Rock W, Chazan B, Keller N, Guy N, et al. (2004) Risk factors for the development of extended-spectrum beta-lactamase-producing bacteria in nonhospitalized patients. Eur J Clin Microbiol Infect Dis 23: 163–167.
  8. 8. Keynan Y, Rubinstein E (2007) The changing face of Klebsiella pneumoniae infections in the community. Int J Antimicrob Agents 30: 385–389.
  9. 9. Carpenter JL (1990) Klebsiella pulmonary infections: occurrence at one medical center and review. Rev Infect Dis 12: 672–682.
  10. 10. Ko WC, Paterson DL, Sagnimeni AJ, Hansen DS, Von Gottberg A, et al. (2002) Community-acquired Klebsiella pneumoniae bacteremia: global differences in clinical patterns. Emerg Infect Dis 8: 160–166.
  11. 11. Yu VL, Hansen DS, Ko WC, Sagnimeni A, Klugman KP, et al. (2007) Virulence characteristics of Klebsiella and clinical manifestations of K. pneumoniae bloodstream infections. Emerg Infect Dis 13: 986–993.
  12. 12. Liu YC, Cheng DL, Lin CL (1986) Klebsiella pneumoniae liver abscess associated with septic endophthalmitis. Arch Intern Med 146: 1913–1916.
  13. 13. Cheng DL, Liu YC, Yen MY, Liu CY, Wang RS (1991) Septic metastatic lesions of pyogenic liver abscess. Their association with Klebsiella pneumoniae bacteremia in diabetic patients. Arch Intern Med 151: 1557–1559.
  14. 14. Wang JH, Liu YC, Lee SS, Yen MY, Chen YS, et al. (1998) Primary liver abscess due to Klebsiella pneumoniae in Taiwan. Clin Infect Dis 26: 1434–1438.
  15. 15. Fang FC, Sandler N, Libby SJ (2005) Liver abscess caused by magA+ Klebsiella pneumoniae in North America. J Clin Microbiol 43: 991–992.
  16. 16. Fang CT, Lai SY, Yi WC, Hsueh PR, Liu KL, et al. (2007) Klebsiella pneumoniae genotype K1: an emerging pathogen that causes septic ocular or central nervous system complications from pyogenic liver abscess. Clin Infect Dis 45: 284–293.
  17. 17. Rahimian J, Wilson T, Oram V, Holzman RS (2004) Pyogenic liver abscess: recent trends in etiology and mortality. Clin Infect Dis 39: 1654–1659.
  18. 18. Ørskov I (1984) Klebsiella Trevisan 1885. In: Krieg NR, Holt JG, editors. Bergey's manual of systematic bacteriology. Baltimore: Williams and Wilkins Co. pp. 461–465.
  19. 19. Chand MS, MacArthur CJ (1997) Primary atrophic rhinitis: a summary of four cases and review of the litterature. Otolaryngol Head Neck Surg 116: 554–558.
  20. 20. Shehata MA (1996) Atrophic rhinitis. Am J Otolaryngol 17: 81–86.
  21. 21. Hart CA, Rao SK (2000) Rhinoscleroma. J Med Microbiol 49: 395–396.
  22. 22. Lu CH, Chang WN, Chang HW (2002) Klebsiella meningitis in adults: clinical features, prognostic factors and therapeutic outcomes. J Clin Neurosci 9: 533–538.
  23. 23. Wong CH, Kurup A, Wang YS, Heng KS, Tan KC (2004) Four cases of necrotizing fasciitis caused by Klebsiella species. Eur J Clin Microbiol Infect Dis 23: 403–407.
  24. 24. Kohler JE, Hutchens MP, Sadow PM, Modi BP, Tavakkolizadeh A, et al. (2007) Klebsiella pneumoniae necrotizing fasciitis and septic arthritis: an appearance in the Western hemisphere. Surg Infect (Larchmt) 8: 227–232.
  25. 25. Richens J (1985) Donovanosis–a review. P N G Med J 28: 67–74.
  26. 26. Carter JS, Bowden FJ, Bastian I, Myers GM, Sriprakash KS, et al. (1999) Phylogenetic evidence for reclassification of Calymmatobacterium granulomatis as Klebsiella granulomatis comb. nov. Int J Syst Bacteriol 49: 1695–1700.
  27. 27. Grimont PAD, Grimont F (2005) Genus Klebsiella. In: Brenner DJ, Krieg NR, Staley JT, editors. Bergey's manual of Systematic Bacteriology Volume 2: The Proteobacteria, Part B: The Gammaproteobacteria. New York: Springer-Verlag. pp. 685–693.
  28. 28. Williams P, Tomas JM (1990) The pathogenicity of Klebsiella pneumoniae. Reviews of Medical Microbiology 1: 196–204.
  29. 29. Regue M, Climent N, Abitiu N, Coderch N, Merino S, et al. (2001) Genetic characterization of the Klebsiella pneumoniae waa gene cluster, involved in core lipopolysaccharide biosynthesis. J Bacteriol 183: 3564–3573.
  30. 30. Chou HC, Lee CZ, Ma LC, Fang CT, Chang SC, et al. (2004) Isolation of a chromosomal region of Klebsiella pneumoniae associated with allantoin metabolism and liver infection. Infect Immun 72: 3783–3792.
  31. 31. Ma LC, Fang CT, Lee CZ, Shun CT, Wang JT (2005) Genomic heterogeneity in Klebsiella pneumoniae strains is associated with primary pyogenic liver abscess and metastatic infection. J Infect Dis 192: 117–128.
  32. 32. Simoons-Smit AM, Verweiji-van Vught AMJJ, Kanis IYR, MacLaren DM (1984) Virulence of Klebsiella strains in experimentally induced skin lesions in the mouse. Journal of Medical Microbiology 17: 67–77.
  33. 33. Mizuta LG, Ohta M, Mori M, Hasegawa T, Nakashima I, Kato N (1983) Virulence for mice of Klebsiella strains belonging to the O1 group: relationship to their capsular (K) types. Infect Immun 40: 56–61.
  34. 34. Nassif X, Sansonetti PJ (1986) Correlation of the virulence of Klebsiella pneumoniae K1 and K2 with the presence of a plasmid encoding aerobactin. Infection and Immunity 54: 603–608.
  35. 35. Ofek I, Kabha K, Athamna A, Frankel G, Wozniak DJ, et al. (1993) Genetic exchange of determinants for capsular polysaccharide biosynthesis between Klebsiella pneumoniae strains expressing serotypes K2 and K21a. Infect Immun 61: 4208–4216.
  36. 36. Eickoff TC, Steinhauer BW, Finland M (1966) The Klebsiella-Enterobacter-Serratia division. Biochemical and serological characteristitcs and susceptibility to antibiotics. Annales of Internal Medicine 65: 1163–1179.
  37. 37. Cheng HP, Chang FY, Fung CP, Siu LK (2002) Klebsiella pneumoniae liver abscess in Taiwan is not caused by a clonal spread strain. J Microbiol Immunol Infect 35: 85–88.
  38. 38. Turton JF, Englender H, Gabriel SN, Turton SE, Kaufmann ME, et al. (2007) Genetically similar isolates of Klebsiella pneumoniae serotype K1 causing liver abscesses in three continents. J Med Microbiol 56: 593–597.
  39. 39. Coffey TJ, Dowson CG, Daniels M, Zhou J, Martin C, et al. (1991) Horizontal transfer of multiple penicillin-binding protein genes, and capsular biosynthetic genes, in natural populations of Streptococcus pneumoniae. Mol Microbiol 5: 2255–2260.
  40. 40. Brisse S, Verhoef J (2001) Phylogenetic diversity of Klebsiella pneumoniae and Klebsiella oxytoca clinical isolates revealed by randomly amplified polymorphic DNA, gyrA and parC genes sequencing and automated ribotyping. Int J Syst Evol Microbiol 51: 915–924.
  41. 41. Nouvellon M, Pons JL, Sirot D, Combe ML, Lemeland JF (1994) Clonal outbreaks of extended-spectrum beta-lactamase-producing strains of Klebsiella pneumoniae demonstrated by antibiotic susceptibility testing, beta-lactamase typing, and multilocus enzyme electrophoresis. J Clin Microbiol 32: 2625–2627.
  42. 42. Maurelli AT, Fernandez RE, Bloch CA, Rode CK, Fasano A (1998) “Black holes” and bacterial pathogenicity: a large genomic deletion that enhances the virulence of Shigella spp. and enteroinvasive Escherichia coli. Proc Natl Acad Sci U S A 95: 3943–3948.
  43. 43. Dagan T, Blekhman R, Graur D (2006) The “domino theory” of gene death: gradual and mass gene extinction events in three lineages of obligate symbiotic bacterial pathogens. Mol Biol Evol 23: 310–316.
  44. 44. Parkhill J, Dougan G, James KD, Thomson NR, Pickard D, et al. (2001) Complete genome sequence of a multiple drug resistant Salmonella enterica serovar Typhi CT18. Nature 413: 848–852.
  45. 45. Guindon S, Gascuel O (2003) A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood. Syst Biol 52: 696–704.
  46. 46. Didelot X, Falush D (2007) Inference of bacterial microevolution using multilocus sequence data. Genetics 175: 1251–1266.
  47. 47. Feil EJ, Li BC, Aanensen DM, Hanage WP, Spratt BG (2004) eBURST: inferring patterns of evolutionary descent among clusters of related bacterial genotypes from multilocus sequence typing data. J Bacteriol 186: 1518–1530.
  48. 48. Van Westreenen M, Paauw A, Fluit AC, Brisse S, Van Dijk W, et al. (2003) Occurrence and spread of SHV extended-spectrum beta-lactamase-producing Klebsiella pneumoniae isolates in Curaçao. J Antimicrob Chemother 52: 530–532.
  49. 49. Richard C (1989) Epidémiologie des infections à Klebsiella pneumoniae dans deux élevages de singes-écureuils et de lémuriens. Bull Soc Pathol Exot 82: 458–464.
  50. 50. Brisse S, Issenhuth-Jeanjean S, Grimont PA (2004) Molecular serotyping of Klebsiella species isolates by restriction of the amplified capsular antigen gene cluster. J Clin Microbiol 42: 3388–3398.
  51. 51. Arakawa Y, Wacharotayankun R, Nagatsuka T, Ito H, Kato N, et al. (1995) Genomic organization of the Klebsiella pneumoniae cps region responsible for serotype K2 capsular polysaccharide synthesis in the virulent strain Chedid. J Bacteriol 177: 1788–1796.
  52. 52. Fang CT, Chuang YP, Shun CT, Chang SC, Wang JT (2004) A novel virulence gene in Klebsiella pneumoniae strains causing primary liver abscess and septic metastatic complications. J Exp Med 199: 697–705.
  53. 53. Hornick DB, Allen BL, Horn MA, Clegg S (1992) Adherence to respiratory epithelia by recombinant Escherichia coli expressing Klebsiella pneumoniae type 3 fimbrial gene products. Infect Immun 60: 1577–1588.
  54. 54. Tarkkanen AM, Virkola R, Clegg S, Korhonen TK (1997) Binding of the type 3 fimbriae of Klebsiella pneumoniae to human endothelial and urinary bladder cells. Infect Immun 65: 1546–1549.
  55. 55. Di Martino P, Bertin Y, Girardeau JP, Livrelli V, Joly B, et al. (1995) Molecular characterization and adhesive properties of CF29K, an adhesin of Klebsiella pneumoniae strains involved in nosocomial infections. Infect Immun 63: 4336–4344.
  56. 56. Nassif X, Fournier J-M, Arondel J, Sansonetti PJ (1989) Mucoid phenotype of Klebsiella pneumoniae is a plasmid-encoded virulence factor. Infection and Immunity 57: 546–552.
  57. 57. Fevre C, Passet V, Weill FX, Grimont PA, Brisse S (2005) Variants of the Klebsiella pneumoniae OKP chromosomal beta-lactamase are divided into two main groups, OKP-A and OKP-B. Antimicrob Agents Chemother 49: 5149–5152.
  58. 58. Roumagnac P, Weill FX, Dolecek C, Baker S, Brisse S, et al. (2006) Evolutionary history of Salmonella typhi. Science 314: 1301–1304.
  59. 59. Achtman M, Zurth K, Morelli G, Torrea G, Guiyoule A, et al. (1999) Yersinia pestis, the cause of plague, is a recently emerged clone of Yersinia pseudotuberculosis. Proc Natl Acad Sci U S A 96: 14043–14048.
  60. 60. Nelson K, Selander RK (1994) Intergeneric transfer and recombination of the 6-phosphogluconate dehydrogenase gene (gnd) in enteric bacteria. Proc Natl Acad Sci U S A 91: 10227–10231.
  61. 61. Bisercic M, Feutrier JY, Reeves PR (1991) Nucleotide sequences of the gnd genes from nine natural isolates of Escherichia coli: evidence of intragenic recombination as a contributing factor in the evolution of the polymorphic gnd locus. J Bacteriol 173: 3894–3900.
  62. 62. Turner KM, Hanage WP, Fraser C, Connor TR, Spratt BG (2007) Assessing the reliability of eBURST using simulated populations with known ancestry. BMC Microbiol 7: 30.
  63. 63. Smith JM, Smith NH, O'Rourke M, Spratt BG (1993) How clonal are bacteria? Proc Natl Acad Sci U S A 90: 4384–4388.
  64. 64. Spratt BG, Maiden MC (1999) Bacterial population genetics, evolution and epidemiology. Philos Trans R Soc Lond B Biol Sci 354: 701–710.
  65. 65. Ørskov I, Ørskov F (1984) Serotyping of Klebsiella. In: Bergan T, editor. Methods in Microbiology. London: Academic Press. pp. 143–164.
  66. 66. Hansen DS, Skov R, Benedi JV, Sperling V, Kolmos HJ (2002) Klebsiella typing: pulsed-field gel electrophoresis (PFGE) in comparison with O:K-serotyping. Clin Microbiol Infect 8: 397–404.
  67. 67. Monot M, Honore N, Garnier T, Araoz R, Coppee JY, et al. (2005) On the origin of leprosy. Science 308: 1040–1042.
  68. 68. Goldstein EJC, Lewis RP, Martin WJ, Edelstein PH (1978) Infections caused by Klebsiella ozaenae: a changing disease spectrum. Journal of Clinical Microbiology 8: 413–418.
  69. 69. Sarma PS (2007) Klebsiella ozaenae splenic abscess following odontogenic infection in a girl with sickle cell disease. Int J Infect Dis 11: 86–87.
  70. 70. Brenner DJ, Steigerwalt AG, Fanning GR (1972) Differentiation of Enterobacter aerogenes from Klebsiellae by deoxyribonucleic acid reassociation. International Journal of Systematic Bacteriology 22: 193–200.
  71. 71. Ørskov I (1974) Klebsiella. In: Buchanan RE, Gibbons NE, editors. Bergey's manual of determinative bacteriology. 8th ed. Baltimore: Williams & Wilkins. pp. 321–324.
  72. 72. Kharsany AB, Hoosen AA, Kiepiela P, Kirby R, Sturm AW (1999) Phylogenetic analysis of Calymmatobacterium granulomatis based on 16S rRNA gene sequences. J Med Microbiol 48: 841–847.
  73. 73. Lai YC, Lin GT, Yang SL, Chang HY, Peng HL (2003) Identification and characterization of KvgAS, a two-component system in Klebsiella pneumoniae CG43. FEMS Microbiol Lett 218: 121–126.
  74. 74. Yu WL, Ko WC, Cheng KC, Lee CC, Lai CC, et al. (2008) Comparison of prevalence of virulence factors for Klebsiella pneumoniae liver abscesses between isolates with capsular K1/K2 and non-K1/K2 serotypes. Diagn Microbiol Infect Dis.
  75. 75. Darfeuille-Michaud A, Jallat C, Aubel D, Sirot D, Rich C, et al. (1992) R-plasmid-encoded adhesive factor in Klebsiella pneumoniae strains responsible for human nosocomial infections. Infect Immun 60: 44–55.
  76. 76. Struve C, Bojer M, Nielsen EM, Hansen DS, Krogfelt KA (2005) Investigation of the putative virulence gene magA in a worldwide collection of 495 Klebsiella isolates: magA is restricted to the gene cluster of Klebsiella pneumoniae capsule serotype K1. J Med Microbiol 54: 1111–1113.
  77. 77. Yeh KM, Kurup A, Siu LK, Koh YL, Fung CP, et al. (2007) Capsular serotype K1 or K2, rather than magA and rmpA, is a major virulence determinant for Klebsiella pneumoniae liver abscess in Singapore and Taiwan. J Clin Microbiol 45: 466–471.
  78. 78. van Kregten E, Westerdaal NAC, Willers JMN (1984) New, simple medium for selective recovery of Klebsiella pneumoniae and Klebsiella oxytoca from human feces. Journal of Clinical Microbiology 20: 936–941.
  79. 79. Richard C (1973) Etude antigénique et biochimique de 500 souches de Klebsiella. Annales de Biologie Clinique 31: 295–303.
  80. 80. Tainturier D, Richard C (1986) Métrites en série pendant la saison de monte chez des juments de trait breton provoquées par Klebsiella pneumoniae type capsulaire K2, biotype d. Revue de Médecine Vétérinaire 137: 49–58.
  81. 81. Brisse S, Duijkeren E (2005) Identification and antimicrobial susceptibility of 100 Klebsiella animal clinical isolates. Vet Microbiol 105: 307–312.
  82. 82. Diancourt L, Passet V, Verhoef J, Grimont PAD, Brisse S (2005) Multilocus sequence typing of Klebsiella pneumoniae nosocomial isolates. J Clin Microbiol 43: 4178–4182.
  83. 83. McClelland M, Florea L, Sanderson K, Clifton SW, Parkhill J, et al. (2000) Comparison of the Escherichia coli K-12 genome with sampled genomes of a Klebsiella pneumoniae and three Salmonella enterica serovars, Typhimurium, Typhi and Paratyphi. Nucleic Acids Res 28: 4974–4986.
  84. 84. Schouls LM, van der Heide HG, Vauterin L, Vauterin P, Mooi FR (2004) Multiple-locus variable-number tandem repeat analysis of Dutch Bordetella pertussis strains reveals rapid genetic changes with clonal expansion during the late 1990s. J Bacteriol 186: 5496–5505.
  85. 85. Feil EJ (2004) Small change: keeping pace with microevolution. Nat Rev Microbiol 2: 483–495.
  86. 86. Huson DH (1998) SplitsTree: analyzing and visualizing evolutionary data. Bioinformatics 14: 68–73.
  87. 87. Huson DH, Bryant D (2006) Application of phylogenetic networks in evolutionary studies. Mol Biol Evol 23: 254–267.
  88. 88. Kumar S, Tamura K, Jakobsen IB, Nei M (2001) MEGA2: molecular evolutionary genetics analysis software. Bioinformatics 17: 1244–1245.
  89. 89. Rozas J, Rozas R (1999) DnaSP version 3: an integrated program for molecular population genetics and molecular evolution analysis. Bioinformatics 15: 174–175.
  90. 90. Feil EJ, Maiden MC, Achtman M, Spratt BG (1999) The relative contributions of recombination and mutation to the divergence of clones of Neisseria meningitidis. Mol Biol Evol 16: 1496–1502.
  91. 91. Guttman DS, Dykhuizen DE (1994) Clonal divergence in Escherichia coli as a result of recombination, not mutation. Science 266: 1380–1383.
  92. 92. McVean G, Awadalla P, Fearnhead P (2002) A coalescent-based method for detecting and estimating recombination from gene sequences. Genetics 160: 1231–1241.