• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of jcmPermissionsJournals.ASM.orgJournalJCM ArticleJournal InfoAuthorsReviewers
J Clin Microbiol. Aug 2004; 42(8): 3388–3398.
PMCID: PMC497587

Molecular Serotyping of Klebsiella Species Isolates by Restriction of the Amplified Capsular Antigen Gene Cluster


The objective of the present work was to develop a molecular method that would enable determination of the capsular serotypes of Klebsiella isolates without the use of antiserum. PCR amplification of the capsular antigen gene cluster (cps) was followed by digestion with the restriction enzyme HincII (cps PCR-restriction fragment length polymorphism [RFLP] analysis). The profiles (C patterns) obtained for 224 strains representing the 77 known K serotypes showed 3 to 13 fragments ranging in size from 0.2 to 4.4 kb. A total of 97 distinct C patterns were obtained; 100% of 61 pairs of samples tested twice showed reproducible C patterns. The C patterns were K-type specific; i.e., the C pattern(s) of any K serotype was distinct from the C patterns of all other K serotypes, with the only exceptions being serotypes K22 and K37, which are known to cross-react. For 12 of 17 K types for which at least two strains were included, C-pattern variations were found among strains with the same K serotype. Therefore, cps PCR-RFLP analysis has a higher discriminatory power than classical K serotyping. C-pattern identity was observed among strains with a given K type that were collected many years apart and from distinct sources, indicating C-pattern stability. Only 4.5% of the strains were nontypeable, because of unsuccessful PCR amplification (whereas 8 to 23% are nontypeable by classical K serotyping). Three of four noncapsulated strains analyzed showed recognizable C patterns. The K serotypes of 18 (82%) of 22 recent Klebsiella pneumoniae clinical isolates could be deduced from their C patterns. In conclusion, cps PCR-RFLP analysis allows determination of the K serotype, while it is easier to perform and more discriminatory than classical serotyping.

Isolates of the genus Klebsiella, especially those of the species Klebsiella pneumoniae, are important causes of nosocomial and community-acquired infections (13, 20, 31). Intra- and interhospital outbreaks of infections caused by such strains are common (2, 3, 23). Klebsiella strains are also responsible for severe infections in animals, including metritis in mares (29, 36).

Most Klebsiella strains are encapsulated by a polysaccharidic capsule of considerable thickness responsible for the glistening, mucoid aspect of colonies on agar plates. Following the pioneering work of Julianelle (17), who described the first three capsular serotypes, 77 K antigens are now included in the international K-typing scheme (27, 28). Determination of the K type has been used since the 1920s to type Klebsiella isolates (17) and has long been the preferred method for the investigation of epidemiological relationships among strains (11, 13, 14, 18, 28). Because it is highly discriminatory and reproducible, K serotyping is particularly useful for the comparison of strains from different geographical locations and different times. K-serotype determination is also important for confirmation of K. pneumoniae subspecies: almost all strains of K. pneumoniae subsp. rhinoscleromatis, the agent of rhinoscleroma, are of serotype K3, whereas klebsiellae of serotype K4 are found almost exclusively to be K. pneumoniae subsp. ozaenae and represent the vast majority of strains of this subspecies.

The capsule is considered a major virulence factor of Klebsiella, and important differences in virulence among serotypes appear to exist. For example, laboratory studies with mouse models have led to the view that serotypes K1 and K2 are the most virulent, in contrast to other serotypes, such as K7 and K21, which are considered less pathogenic (18, 22, 26, 30). Particular serotypes of K. pneumoniae have been incriminated in epidemics in animals (36). Strains of the K1, K2, and K5 serotypes appeared to be associated with metritis epidemics in mares, whereas serotype K7 strains were carried by healthy stallions but did not cause disease in mares (29). Capsular material is the most promising substance that could be used as a vaccine candidate to protect human populations exposed to Klebsiella infections (9) and has proved useful in the control of Klebsiella outbreaks in monkeys and lemurs (36).

The considerations mentioned above emphasize the critical importance of K-serotype determination. A major drawback of serotyping is the large number of cross-reactions that occur among the 77 capsule types. Therefore, in several cases it is necessary for individual sera to be absorbed with the cross-reacting K antigens for diagnostic use. In addition, a substantial proportion, ranging from 8 to 23%, of Klebsiella strains are nontypeable (12, 24, 37), either because they harbor K antigens that have not been incorporated in the reference collection or because they are noncapsulated. Serotyping is not easy to perform, and interpretation of the results is subjective. Because only specialized laboratories can produce and maintain good-quality antisera, the practice of serotyping is restricted to a very few centers, which greatly limits its applicability for surveillance of the prevalent serotypes.

The genomic organization of the chromosomal region that is responsible for capsular polysaccharide synthesis in a K2 K. pneumoniae strain was determined by Arakawa et al. (1). Nineteen possible open reading frames (ORFs) were identified in the 24,329-bp nucleotide sequence of the cps region. The aim of this study was to devise a method for amplifying from all species of the Klebsiella genus the cps gene cluster responsible for the synthesis of the K antigens and to characterize K serotypes by restriction fragment length polymorphism (RFLP) analysis of the amplified cps cluster, which generates C patterns. As a result, a database of C patterns was initiated, which should render it possible to determine capsular types without the use of antisera.


Bacterial strains.

A total of 246 samples were included in this work (Table (Table1).1). Of these, 47 were duplicate cultures of some reference strains (see below), but for simplicity all 246 samples are hereafter referred to as “strains.”

Overview of the 246 strains evaluated in this study

A total of 120 strains were reference strains of the described Klebsiella K serotypes. Seventy of these reference strains were obtained from the World Health Organization Collaborative Center for Escherichia coli and Klebsiella (Statens Serum Institute, Copenhagen, Denmark). Reference strains of serotypes K5, K6, K8, K11, K38, K50, and K68 were lost from our collection. To complete the panel of K-serotype reference strains and to check the reproducibility, 50 serotype reference strains were additionally obtained from the Collection de l'Institut Pasteur (CIP), Institut Pasteur, Paris, France. These strains represented serotypes K1 to K54, with the exception of serotypes K20, K32, K41, and K52. Serotypes K55 to K82 were not available at CIP. Thus, in total, the only reference strain that was not included was that of serotype K68, and 44 reference strains were included twice, with one culture obtained directly from the World Health Organization reference center and one culture obtained from CIP.

Seven reference or type strains of K. pneumoniae subspecies were included. For the type strain of K. pneumoniae subsp. pneumoniae, we included both the culture obtained from CIP (CIP 82.91) and the culture obtained from the American Type Culture Collection (ATCC; ATCC 13883). The type strain of K. pneumoniae subsp. rhinoscleromatis was also included twice, as CIP 52.210 and ATCC 13884. The type strain of K. pneumoniae subsp. ozaenae obtained from the ATCC (ATCC 11296) was included; it corresponds to the K4 serotype reference strain. Additionally, reference strain K. pneumoniae subsp. rhinoscleromatis ATCC 6908 and K. pneumoniae subsp. pneumoniae genome reference strain MGH78578 (sequenced at the Genome Sequencing Center at Washington University Medical School http://genome.wustl.edu/), kindly provided by M. McClelland, were included.

A total of 97 Klebsiella isolates from our laboratory, collected from 1972 to 2001, were included. The K serotypes of these strains had been determined prior to this study. Two strains of serotype K68, which was not represented by a reference strain, and one strain of serotype K59, the reference strain of which was negative upon PCR amplification, were included. Strains representative of the frequent K serotypes in our collection were selected, as were four noncapsulated strains and five strains which reacted with two distinct antisera: three K1,58 strains, one K22,37 strain, and one K21,31 strain. Close examination of the strains labeled as noncapsulated under the microscope after China ink staining confirmed the absence of a visible capsule. These 97 strains mostly originated from diverse human infections, including blood, respiratory tract, and urinary tract infections (Table (Table1).1). Strains from horses (n = 13) and environmental sources (n = 5) were also included.

Finally, 22 recent clinical isolates of unknown K serotype were included in order to evaluate the potential of cps PCR-RFLP analysis for K-serotype determination. Twenty of these isolates were collected from 1997 to 2000 and were previously analyzed for their susceptibilities to antimicrobials (5, 6). These clinical isolates were epidemiologically nonrelated; came from nine European countries; and were obtained from blood (n = 14), urinary tract infections (n = 4), and wound or soft tissue infections (n = 2). These clinical isolates were selected because they showed resistance to ceftazidime (MICs > 64 mg/liter), probably resulting from extended-spectrum beta-lactamase production. The two last isolates were collected in 2003 from two French patients; one was a K. pneumoniae subsp. rhinoscleromatis isolate, and one was a K. pneumoniae subsp. pneumoniae isolate from an anal abscess.

Overall, the species and subspecies distributions were as follows: 175 strains were K. pneumoniae subsp. pneumoniae, 16 were K. pneumoniae subsp. ozaenae, 14 were K. pneumoniae subsp. rhinoscleromatis, 25 were K. planticola, 12 were K. oxytoca, 3 were K. terrigena, and 1 was K. ornithinolytica. When such strains were available in our collection, we purposely included strains of the same K type but of different taxa.


Serotypes were determined prior to this study by the capsular swelling method (35, 36, 38), and the K types of some strains were controlled during this study by the agglutination method (28). The K serotypes of the type strains and reference strains were known prior to this study but were controlled by agglutination when it was believed to be necessary. Strain and antiserum controls were always used in order to control agglutination specificity.


Identification was performed prior to this study by standard, recommended biochemical tests (13). Reidentification of strains to the species level (or to the subspecies level for K. pneumoniae isolates) was performed by using Biotype-100 strips (BioMérieux, Marcy l'Etoile, France), which contain 99 carbohydrate substrates in cupules. Minimal medium 1 was used, and strains were identified with Recognizer software (version 2000; P. A. D. Grimont, Institut Pasteur) against the database of strains of the family Enterobacteriaceae constructed in the laboratory. Substrates that were particularly useful for species discrimination were m-coumarate, gentisate, histamin, 3-hydroxybenzoate, d-melezitose, 3-O-methyl-d-glucose, and tricarballylate (13). For strains with dubious identities, the identification was confirmed by gyrA gene sequencing (6). To confirm the identification of K. pneumoniae subsp. ozaenae and K. pneumoniae subsp. rhinoscleromatis isolates, the following biochemical tests were used (14): Voges-Proskauer test and tests for urease, o-nitrophenyl-β-d-galactopyranoside, lysine decarboxylase, citrate, malonate, and gas production.

Bacterial DNA preparation.

Bacteria were plated on tryptocasein soy (TCS) agar, and one colony was grown overnight with shaking at 37°C in 10 ml of TCS broth. DNA was extracted by use of the Wizard Genomic DNA purification kit (Promega, Charbonnières-les-Bains, France). Briefly, 1 ml of culture suspension was centrifuged (13,000 × g, 5 min). Nuclei lysis solution (600 μl) was added to the pellet, the pellet was gently resuspended, and the samples were incubated at 80°C for 5 min. After equilibration to room temperature, 3 μl of RNase solution was added; the samples were gently mixed 25 times by inversion and were then incubated at 37°C for 30 min. Protein precipitation solution (200 μl) was added, before vigorous vortexing for 30 s. The tubes were then cooled for at least 5 min on ice. After 4 min of centrifugation (13,000 × g), the supernatant was transferred to tubes containing 600 μl of isopropanol, and the samples were gently mixed and then centrifuged (13,000 × g for 3 min). The supernatants were thoroughly eliminated, and the pellets were washed with 600 μl of 70% ethanol and centrifuged (13,000 × g for 2 min). The supernatants were thoroughly eliminated, and the pellets were dried at room temperature for 1 h. Fifty to 100 μl of rehydration solution was added, and the DNAs were dissolved by gently mixing them overnight at room temperature. The DNAs were stored at −20°C. Control of DNA quality was performed by 0.8% agarose gel electrophoresis in 0.5× TBE (Tris-borate-EDTA) buffer.

Amplification of cps genomic region.

Among the ORFs identified in the cps region (1), the gnd gene, which codes for 6-phosphogluconate dehydrogenase, was identified downstream of the longest transcriptional unit. We designed two primers flanking the cps cluster, based on (i) the sequence of the cps genomic region of a K2 serotype (1), (ii) the sequence of the K. pneumoniae genome project K52 strain MGH78578 of the Genome Sequencing Center at Washington University Medical School (http://genome.wustl.edu/), and (iii) the sequence diversity of the gnd gene among Klebsiella species (S. Brisse, unpublished data). Primer CPS-1 (5′-GCT GGT AGC TGT TAA GCC AGG GGC GGT AGC G) was complementary to the JUMPstart sequence (15) that is located just upstream of ORF3 on the sequence from strain K2 Chedid (1, 33). This JUMPstart sequence was found to be conserved in genome project strain MGH78578. Reverse primer rCPS (5′-TAT TCA TCA GAA GCA GCA CGC AGC TGG GAG AAG CC) was complementary to a conserved region of the gnd gene sequence (gene positions 1042 to 1007). A second reverse primer, which was named rCPS2 (5′-GCG CTC TGG CTG GTC CAT TTA CCG GTC CCT TTG) and whose sequence was specific for a region located 232 nucleotides upstream of primer rCPS (which was thus 232 bp closer to the forward primer), was designed in order to amplify the cps regions of strains for which the amplification with primer rCPS failed. The molecular sizes of the products amplified from strain MGH78578 are expected to be 18,766 bp in PCRs with primers CPS-1 and rCPS and 18,534 bp in PCRs with primers CPS-1 and rCPS2.

PCR amplifications were performed with the Expand Long Template PCR system (Roche, Mannheim, Germany). In a final volume of 50 μl, 2.5 μl of template DNA (diluted 1/10 extemporaneously just after 1 h of gentle mixing at room temperature) was added to the amplification solution containing 5 μl of 10× kit buffer 2, 3.75 U of Taq polymerase, each deoxynucleoside triphosphate at a final concentration of 0.5 mM, and each primer at a final concentration of 0.3 μM. Cycling conditions were as follows: one denaturation step of 2 min at 94°C and 10 initial cycles of 10 s at 94°C, 30 s at 63°C, and 15 min at 68°C, followed by 20 iterative cycles of 10 s at 94°C, 30 s at 63°C, and 15 min plus 20 s for each new cycle at 72°C. A final elongation step of 7 min at 72°C was added. Amplified products were controlled by 0.75% agarose gel electrophoresis in 0.5× TBE buffer.

In silico restriction analysis and comparative analysis of several restriction enzymes.

In order to select restriction enzymes that would generate easily interpretable and informative restriction profiles, the numbers and size distributions of the expected restriction fragments that would be obtained with the two published sequences of the Klebsiella cps cluster were investigated. Sequence files were prepared from the genome sequence with GenBank accession number D21242 and the MGH78578 genome sequence by pasting the primer sequences at the two extremities of the expected amplified region. The expected PCR products were 17,057 and 18,766 nucleotides, respectively. Restriction maps of both sequences were first generated by using all available restriction enzymes in the MapDraw software of the Lasergene package (DNASTAR, Madison, Wis.). Enzymes that exhibited between 10 and 17 restriction sites were selected and were used to generate lists of restriction fragment sizes ordered by decreasing size by using the web page http://genome-www.stanford.edu/Sacch3D/analysis/. On the basis of the expected profiles and prices, enzymes PvuII, EcoRI, EcoRV, AvaII, HaeII, and HincII were selected and tested with the products amplified from serotype K1 to K12 strains. The enzyme HincII was finally selected as the most promising on the basis of its discriminatory ability and interpretability.

RFLP analysis.

Between 10 and 20 μl of each of the amplified products, depending on their concentrations, as revealed by control gel electrophoresis, was digested for 150 min at 37°C in a final volume of 25 μl containing 2.5 μl of 10× buffer (Buffer 3; BioLabs, St. Quentin en Yvelines, France), 20 U of the HincII restriction enzyme (BioLabs), and sterile distilled water. Gels of 20 cm in length and with a final agarose concentration of 1.5% were prepared by mixing 1:1 Metaphor agarose (TEBU, Le Perray en Yvelines, France) and NuSieve (3:1) agarose (TEBU) in 0.5× TAE buffer (20 mM Tris-acetate, 0.5 mM EDTA [pH 8.2]). DNA size markers were prepared by mixing 12 μl of loading buffer containing bromphenol blue, 60 μl of Amplisize Molecular Ruler (10 ng/μl; Bio-Rad Laboratories, Marnes la Coquette, France), 12 μl of bacteriophage lambda HindIII (0.5 ng/μl; Promega), and 28 μl of the product amplified from K52 K. pneumoniae strain MGH78578 digested with EcoRV (the sizes of which are known from the genome sequence). Markers were loaded in lanes 1, 7, 13, and 20 of each 20-lane gel. After the samples were loaded, the restriction products were separated by electrophoresis for 16 h at 55 V. The gels were then incubated in a small volume of distilled water containing ethidium bromide (final concentration, 0.5 μl/ml) for 40 min, briefly rinsed in 0.5× TAE buffer, and visualized under UV light. Gel images were electronically captured by use of a charge-coupled device video camera interfaced to a microcomputer (Genomic, Collonges-sous-Salève, France). Tagged image file format (TIFF) images were analyzed with the programs in the Taxotron package (version 2000; Institut Pasteur) and with BioNumerics software (version 3.5; Applied Maths, Sint-Martens-Latem, Belgium). The molecular sizes of the fragments were estimated by use of the Spline algorithm in Taxotron software. Bands corresponding to fragments smaller than 200 bp were not considered because they were faint. Cluster analysis was performed by the unweighted pair group method with arithmetic averages procedure with BioNumerics software, based on the Dice coefficient with a 2% tolerance parameter.


Amplification of the cps gene cluster.

In order to initiate the construction of a database of reference patterns, a first set of 224 Klebsiella strains was tested to determine the ability of PCR to amplify the cps region (Table (Table1).1). The set included 215 strains of known K serotypes, 4 noncapsulated strains, and 5 strains that reacted with two distinct antisera. For 211 (94%) of these strains, unique and intense PCR products were obtained when primers CPS-1 and rCPS were used. Amplification succeeded for three of four noncapsulated strains, as well as for the five strains that reacted with two distinct antisera.

The PCR product was longer than 10 kb for 207 of the 211 strains. The four exceptions were the four strains corresponding to the two cultures of each of the K15 and K34 reference strains, which showed amplification products of approximately 7 kb. In order to control for the possibility that the amplified products with atypically small sizes did not result from nonspecific amplification, a new PCR primer, rCPS2, which targets a position located 232 nucleotides upstream of rCPS on the gnd gene, was defined. When primer rCPS2 was used in combination with primer CPS-1, the sizes of the PCR products obtained for the four atypical samples were undistinguishable from those obtained initially.

For 13 of the 224 strains studied, we could not obtain a unique and intense PCR product using primers CPS-1 and rCPS (Table (Table1).1). First, the amplification product was too weak for nine strains. These strains were of serotype K2, K6 (two strains), K7, K16, K24, K59 (the reference strain for this serotype), K68, and one strain that was noncapsulated (Table (Table1).1). Five of these strains were K. pneumoniae subsp. pneumoniae, two were K. terrigena, one was K. planticola, and one was K. oxytoca. Second, four strains showed, in addition to an intense PCR product in the expected size range, an additional PCR product of a smaller molecular size. These strains were of serotypes K3, K4, K6, and K9 and belonged to K. pneumoniae subsp. pneumoniae (three strains) and K. pneumoniae subsp. ozaenae. The initial species identifications of these 13 strains were confirmed by gyrA gene sequencing. Other PCR amplification conditions were tested for the 13 strains by varying the composition of the PCR buffer, but amplification was still unsuccessful.

Because it was suspected that the poor PCR results obtained with these 13 strains could be due to variations in the gnd sequence at the rCPS primer-annealing site, amplification with primer rCPS2 was tested. PCR amplification was successful for 5 of the 13 problematic strains. These strains were of serotypes K6 (two strains), K7, K9, and K68. The remaining eight strains were still unsuccessfully amplified.

Restriction analysis and correspondence between cps PCR-RFLP patterns and K serotypes.

The products amplified from the 211 strains that were successfully amplified with primers CPS-1 and rCPS were digested with HincII, and the restriction fragments were separated by agarose gel electrophoresis. Examples of the cps PCR-RFLP patterns, hereafter referred to as C patterns, are presented in Fig. Fig.1.1. Since the smallest fragment observed was longer than 200 bp for most strains and since fragments smaller than this were difficult to observe reproducibly due to their low intensities, we chose to limit the analysis to fragments ranging in size from 200 to 4,361 bp (the last size corresponds to the size of the largest size marker fragment).

FIG. 1.
cps PCR-RFLP analysis of 12 Klebsiella strains. Lanes M, molecular size markers (in base pairs). The C-pattern codes are indicated above the lanes, and the K serotypes and strain names are given in parentheses. Small differences can be observed between ...

In order to determine the accuracy of the molecular size estimation performed with Taxotron software and to confirm that the C patterns were composed of fragments belonging to the cps region, the C patterns obtained for two serotypes, serotypes K2 and K52, for which the nucleotide sequence of the cps region had been fully determined (see reference 1 and http://genome.wustl.edu/), were compared with the expected profiles obtained by in silico restriction (the figure is available upon request). The standard deviation of the difference between the expected and the estimated sizes was 1.2% (17 fragments were included in the comparison). The differences were higher for the smaller fragments, with the highest percent difference being 3.3% (8 nucleotides) for the 242-nucleotide fragment of pattern C2e.

C-pattern reproducibility and stability were controlled in the following ways. First, 17 strains were randomly selected and assayed twice, starting from the DNA extraction step. Second, for 44 K serotypes, two distinct cultures of the reference strain kept in different laboratories for many years were tested (see Materials and Methods and Table Table1).1). Indistinguishable C patterns were obtained for the two samples of each of these 61 pairs (data not shown).

Ninety-six distinct C patterns were obtained when the 211 strains were analyzed (Fig. (Fig.22 and Table Table1).1). The C patterns were easily interpretable and comprised 3 to 13 fragments.

FIG. 2.
Overview of the 102 distinct C patterns found during this study. The five C patterns obtained for clinical strains whose K serotypes were initially unknown and which did not exactly match a reference C pattern are given. The patterns obtained with primer ...

Remarkably, after analysis of the 77 K serotypes, represented either by their reference strain or by strains from our collection (two K68 strains and one K59 strain), all C patterns were K serotype specific; i.e., the C pattern of any K serotype was distinct from the C patterns of all other K serotypes. Moreover, the C patterns of distinct K types always differed by at least two fragments. The only exception was the serotype pair K22 and K37, which could not be distinguished by cps PCR-RFLP analysis. We called their common C pattern C22,37a. Interestingly, serological cross-reactions between these two serotypes are known to occur (28). Strain 1622 from our collection reacted with both antisera, and its C pattern was also C22,37a (Table (Table11).

The C patterns of the three noncapsulated strains that were successfully amplified by PCR were determined. These strains had C patterns C2e, C3a, and C14a, respectively (Table (Table1).1). These three strains did not agglutinate with the corresponding antisera. Three strains that reacted with both the K1 and the K58 antisera had pattern C1a. Strain 607, which reacted with both the K21 and the K31 antisera, showed a unique C pattern, called C21,31a (Fig. (Fig.22).

The PCR products from the five strains that were amplified with primer rCPS2 (but not when with primer rCPS) were cut with HincII and the patterns were compared to those in the database (Fig. (Fig.3).3). For four of these strains, the pattern obtained closely resembled the C pattern for a strain of the same K serotype, whereas the pattern obtained for K68 strain 1578 was unrelated to C pattern C68a (Fig. (Fig.22 and and33).

FIG. 3.
Comparison of the patterns obtained after amplification with the alternative primer, gnd-specific primer rCPS2, with closely matching C patterns obtained with primer rCPS (and BioNumerics software). In most cases, the patterns obtained with primer rCPS2 ...

C-pattern variation within a given K serotype.

In order to evaluate the variations in C patterns among strains within a given K serotype, 97 strains were selected from our laboratory collection (Table (Table11 and Materials and Methods), based on availability, with emphasis on the serotypes that were highly represented in our collection: K1 to K4, K7, K24, K25, K30, K35, and K62. In addition, seven reference or type strains distinct from the serotype reference strains were included (Table (Table11 and Materials and Methods). These additional 104 strains belonged to 17 distinct K serotypes.

For 5 of these 17 K serotypes, the C patterns of our collection strains or reference strains did not differ from those of the corresponding K-serotype reference strains. These were serotypes K1 (15 strains other than the K-serotype reference strains were analyzed), K3 (n = 19), K5 (n = 3), K7 (n = 4), K16 (n = 1), and K24 (n = 1).

For the remaining 12 K serotypes (K2, K4, K6, K8, K10, K25, K30, K35, K39, K47, K52, and K62), we observed C-pattern variations within a given K serotype (Table (Table11 and Fig. Fig.11 and and2).2). The K serotypes within which at least three distinct C patterns were found included K4 (six C patterns), K2 (five C patterns), and K35 and K62 (three C patterns each). All variant C patterns found were K serotype specific; i.e., they were distinct from the C patterns observed for strains of other K serotypes.

In general, the various C patterns observed among strains within a given K serotype were similar, as illustrated by cluster analysis (Fig. (Fig.2),2), but exceptions were found. For example, strains of serotype K2 showed five distinct C patterns that differed by only one or a few bands (Fig. (Fig.11 and and2).2). On the other hand, the 12 K. pneumoniae subsp. ozaenae strains of serotype K4 showed six distinct C patterns. Four of these C patterns were similar and grouped together upon cluster analysis, whereas patterns C4e and C4f were very distinct (Fig. (Fig.11 and and22).

The K serotypes of all strains showing atypical C patterns were verified. No disagreement was found between the K serotypes determined upon verification and the K serotypes determined initially.

C patterns of strains belonging to the same K serotype but to distinct species or subspecies.

It is well known that some K serotypes are shared by distinct Klebsiella species or K. pneumoniae subspecies (28). Strains belonging to distinct species were included for five K serotypes: K8, K35, K39, K52, and K62 (Table (Table1).1). For these five K serotypes, the C patterns of strains belonging to distinct species differed (Table (Table11 and Fig. Fig.2).2). Moreover, the C patterns of different species were generally very different, as opposed to the similarity of variant C patterns observed within a species. For example, C patterns C62a and C62b of K. pneumoniae K62 strains were similar to each other, but C pattern C62c of K. oxytoca K62 strain 199 was very distinct (Fig. (Fig.2).2). One exception was the resemblance of patterns C39a and C39b, observed in K. planticola and K. pneumoniae subsp. pneumoniae, respectively (Table and Fig. Fig.22).

Strains of distinct subspecies of K. pneumoniae but with the same serotype had the same C patterns. K5 strains of K. pneumoniae subsp. ozaenae and K. pneumoniae subsp. pneumoniae had C pattern C5a, whereas K3 strains of K. pneumoniae subsp. rhinoscleromatis and of K. pneumoniae subsp. pneumoniae had C pattern C3a (Table (Table1).1). The identities of all of these strains at the subspecies level were confirmed by biochemical tests.

Families of distinct but related C patterns.

Comparison of the fragments present in all possible pairs of C patterns revealed that the C patterns of strains belonging to different K serotypes sometimes shared a number of fragments in common, which suggested an evolutionary relatedness of the corresponding cps clusters. Apart from the C pattern already mentioned, C22,37a, which was shared by K22 and K37 strains, the clearest example was the pair of C patterns C1a and C4a. Indeed, although these two patterns could be distinguished without ambiguity, most of the fragments were common (Fig. (Fig.2).2). The following pairs of C patterns also showed several fragments in common: C1a and C4d, C2c and C58a, C2e and C50a, C3a and C4e, C5a and C24a, C7a and C80a, C17a and C61a, C18a and C23a, C18a and C27a, C18a and C52a, C21a and C33a, C28a and C53a, C30a and C41a, C36a and C81a, C39 and C62a-C62b, C52a and C62a, and C52a and C21,31a. These pairs of serotypes have not been described as cross-reacting (28). Surprisingly, with the exception of the C patterns of serotype K22 and K37 strains, we did not observe fragment relatedness among the C patterns of serotypes that are known to cross-react (28), with only three exceptions: K12, K29, and K42; K24 and K47; and K27 and K46.

Evaluation of fragment size errors.

The 44 pairs of cultures of the reference strains were used to evaluate the variations associated with fragment size interpolation. This set of 88 patterns comprised 346 fragments. The standard deviation of the fragment size differences was 1.2%, and a higher variation was associated with a smaller fragment size. A 99% confidence interval would correspond to 2.575 times the standard deviation, i.e., 3.17% of the fragment size.

Molecular serotyping of recent clinical isolates.

The C patterns of 22 epidemiologically unrelated clinical isolates of unknown serotype were determined (Table (Table1).1). Twenty of these strains were ceftazidime resistant and, thus, potential extended-spectrum beta-lactamase producers. When the 16 distinct C patterns obtained were compared to the reference database of 96 C patterns obtained so far, 11 C patterns, representing 16 isolates, were undistinguishable from a reference C pattern by use of a 3% tolerance of fragment size variation (Table (Table1).1). The K serotypes that could thus be deduced from the C patterns were confirmed by classical K serotyping, with no exception. The five remaining C patterns included two patterns, called C24like-a and C68like-a, that were closely similar to reference C patterns C24a and C68a, respectively (one and three fragment differences, respectively; Fig. Fig.2).2). Using classical serotyping, we confirmed that the K serotypes of these strains were K24 and K68, respectively. Thus, in total, the K serotypes of 18 (82%) of 22 clinical strains could be correctly deduced from their C patterns. However, we would recommend that the deduced serotype be confirmed by classical serotyping when the C pattern does not exactly match a reference pattern. Three new C patterns (called CNEW-1, CNEW-2, and CNEW-3) did not appear to be related to any reference pattern (Fig. (Fig.2).2). The two isolates with pattern CNEW-3 were nontypeable (no reaction with any antisera) and possibly represent a new serotype, whereas strain SB164 (CNEW-1) and strain SB204 (CNEW-2) had serotypes K35 and K22,37, respectively.


Over several decades considerable knowledge has accumulated on the epidemiology, prevalence, and pathogenic potential of capsular serotypes of Klebsiella species (13, 14, 28). However, due to technical limitations, the practice of serotyping is declining, and molecular methods, such as pulsed-field gel electrophoresis (14, 32), amplified fragment polymorphism analysis (16), randomly amplified polymorphic DNA analysis (42), and ribotyping (3, 6), are increasingly preferred for use in the discrimination of Klebsiella strains for typing and identification purposes. Unfortunately, none of these methods allows the determination of the capsular serotype, and it is therefore impossible to compare new data with the corpus of knowledge built over many decades on the basis of serotyping data.

Molecular serotyping methods have recently been proposed as ways of determining the antigenic compositions of bacterial strains through characterization of the genomic regions that are responsible for antigen synthesis (7, 8, 10, 19, 21). Here we show that it is possible to determine the capsular serotypes of Klebsiella strains without the use of antiserum. After analysis of strains of the 77 described Klebsiella K serotypes, the C patterns obtained for any given K serotype were distinct from the C patterns of all other K serotypes, with the only exception being serotypes K22 and K37, which are known to cross-react (28). However, for those serotypes for which only one or a few strains were tested, C-pattern specificity (i.e., the fact that the C patterns of strains of a given K serotype are distinct from those of strains of other K serotypes) will need to be confirmed by analyzing more strains. The fact that very distinct C patterns were obtained for most of the K serotypes indicates that the differences in antigenic specificity among serotypes are due to differences in gene content and the organization of the cps gene cluster rather than only to point mutations or genetic differences in other parts of the genome. This is consistent with the molecular basis of O-antigen variation in E. coli, Shigella, and Salmonella enterica (34, 40, 41).

The fact that C patterns were obtained from noncapsulated strains shows that the absence of a capsule, at least in these strains, is not due to the absence of the cps gene cluster. Rather, it may be due to a few genetic differences affecting either the regulation of capsular gene expression or the complete assemblage of the polysaccharide antigen.

Overall, the C patterns of only 6.5% of the strains (13 of 199 strains, if one sample for each of the 47 duplicate strains is left out) could not be determined, and this rate fell to 4.5% (9 of 199 strains) when alternative gnd-specific primer rCPS2 was used. This rate of nontypeable strains is lower than that obtained by classical K serotyping, which ranges from 8 to 23% (4, 12, 24, 37). Unsuccessful PCR amplifications can probably be attributed to sequence variation at the gnd primer-annealing site. The rate of variation in the sequence of the gnd gene is very high among Klebsiella strains, even within a single species (Brisse, unpublished). In E. coli and S. enterica, the amount of nucleotide diversity in gnd has been shown to be much higher than those in other genes (25, 39). Non-K. pneumoniae strains represented a high proportion (4 of 13 strains) of the strains that failed to produce an amplification product, and the only reference strain that failed to produce an amplification product, the serotype K59 reference strain, was K. planticola. The design of alternative gnd-specific primers for PCR should allow PCR amplification of the recalcitrant strains. Importantly, most C patterns derived by restriction of the PCR products obtained with the alternative primer, rCPS2, were comparable to the C patterns obtained with the original gnd-specific primer, rCPS. Given that the sequences for which the primers are specific are located only 232 bp apart, the impact of the difference at the extremity of the large PCR product on the C pattern was expected to be limited and possibly caused the small differences observed in Fig. Fig.33 compared to the corresponding C pattern.

When classical serotyping is performed, cross-reactions between some antisera are frequently observed (28). The C patterns of five strains, the K serotypes of which could not be determined because they consistently reacted with two antisera, were determined. In one case, a new pattern (pattern C21,31a) was found, and this pattern was distinct from the patterns of strains reacting either only with K21 antisera or only with K31 antisera. This suggests that the cross-reacting antigen may in fact be distinct from the K21 and K31 antigens, while it may share epitopes with them. In all other cases, the C patterns could be matched with those of strains of known K serotypes, eliminating the ambiguity due to serological cross-reactions. Despite the cross-reactions often observed between K1 and K58 antisera, pattern C58a of the K58 reference strain was distinct from pattern C1a (Fig. (Fig.22).

For 12 of 17 K serotypes for which at least two strains were included, C-pattern variations were found within a given K serotype. Thus, cps PCR-RFLP analysis has a higher discriminatory power than classical K serotyping. Two kinds of intra-K-serotype C-pattern variations could be distinguished. First, some C patterns for strains of a given K serotype differed by only one or a few fragments. In these cases, C-pattern variation is probably explained by evolutionary events that cause a limited modification of the sequence or organization of the cps gene cluster but that leave the antigenic specificity unaffected. Second, the C patterns of some strains appeared to be completely unrelated to the C patterns of other strains of the same K serotype. In these cases, C-pattern variation may be due to major reorganization events at the cps locus, but it may also reflect the existence of cps clusters that have distant evolutionary relationships but that direct the synthesis of cross-reacting or identical capsular polysaccharides. Determination of the nucleotide sequence of the cps cluster should distinguish between these possibilities.

It was remarkable to observe C-pattern conservation among strains of a given K serotype that were collected many years apart and from distinct sources (Table (Table1).1). This was the case, for example, for all K1 strains (collected from 1954 and 1984), for all K3 strains (collected from 1954 to 2001), and for K4 strains sharing C pattern C4a (collected from 1930 to 1988). This suggests a high degree of stability of the C pattern across time and across ecological environments and indicates that cps PCR-RFLP analysis is suitable for the long-term epidemiological monitoring of capsular types.

The value of cps PCR-RFLP analysis for K-serotype determination was tested with a set of 22 recent clinical strains of unknown K serotypes, using as a reference the database of C patterns obtained in this study. The high rate of successful determination of the K serotypes for K. pneumoniae isolates indicates that cps PCR-RFLP analysis should allow the determination of the K serotypes of the vast majority of clinical strains of this species. However, future implementation of the reference database by addition of new C patterns is needed to improve the rate of success of the method. For species other than K. pneumoniae, which were not represented in large numbers in this study, a more important effort probably remains necessary in order to achieve good coverage of the diversity of C patterns.


We thank Francine Grimont for helpful tips and comments.


1. Arakawa, Y., R. Wacharotayankun, T. Nagatsuka, H. Ito, N. Kato, and M. Ohta. 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. [PMC free article] [PubMed]
2. Arlet, G., M. Rouveau, I. Casin, P. J. Bouvet, P. H. Lagrange, and A. Philippon. 1994. Molecular epidemiology of Klebsiella pneumoniae strains that produce SHV-4 beta-lactamase and which were isolated in 14 French hospitals. J. Clin. Microbiol. 32:2553-2558. [PMC free article] [PubMed]
3. Bingen, E. H., P. Desjardins, G. Arlet, F. Bourgeois, P. Mariani-Kurkdjian, N. Y. Lambert-Zechovsky, E. Denamur, A. Philippon, and J. Elion. 1993. Molecular epidemiology of plasmid spread among extended broad-spectrum beta-lactamase-producing Klebsiella pneumoniae isolates in a pediatric hospital. J. Clin. Microbiol. 31:179-184. [PMC free article] [PubMed]
4. Blanchette, E. A., and S. J. Rubin. 1980. Seroepidemiology of clinical isolates of Klebsiella in Connecticut. J. Clin. Microbiol. 11:474-478. [PMC free article] [PubMed]
5. Brisse, S., D. Milatovic, A. C. Fluit, J. Verhoef, and F. J. Schmitz. 2000. Epidemiology of quinolone resistance of Klebsiella pneumoniae and Klebsiella oxytoca in Europe. Eur. J. Clin. Microbiol. Infect. Dis. 19:64-68. [PubMed]
6. Brisse, S., and J. Verhoef. 2001. Phylogenetic diversity of Klebsiella pneumoniae and Klebsiella oxytoca clinical isolates revealed by randomly amplified polymorphic DNA, gyrA and parC gene sequencing and automated ribotyping. Int. J. Syst. E vol. Microbiol. 51:915-924. [PubMed]
7. Coimbra, R. S., F. Grimont, and P. A. D. Grimont. 1999. Identification of Shigella serotypes by restriction of amplified O-antigen gene cluster. Res. Microbiol. 150:543-553. [PubMed]
8. Coimbra, R. S., F. Grimont, P. Lenormand, P. Burguiere, L. Beutin, and P. A. D. Grimont. 2000. Identification of Escherichia coli O-serogroups by restriction of the amplified O-antigen gene cluster (rfb-RFLP). Res. Microbiol. 151:639-654. [PubMed]
9. Cryz, S. J., E. Fürer, and R. Germanier. 1985. Purification and vaccine potential of Klebsiella capsular polysaccharides. Infect. Immun. 50:225-230. [PMC free article] [PubMed]
10. Dauga, C., A. Zabrovskaia, and P. A. D. Grimont. 1998. Restriction fragment length polymorphism analysis of some flagellin genes of Salmonella enterica. J. Clin. Microbiol. 36:2835-2843. [PMC free article] [PubMed]
11. Edwards, P. R., and M. A. Fife. 1952. Capsule types of Klebsiella. J. Infect. Dis. 91:92-104. [PubMed]
12. Fung, C. P., B. S. Hu, F. Y. Chang, S. C. Lee, B. I. Kuo, M. Ho, L. K. Siu, and C. Y. Liu. 2000. A 5-year study of the seroepidemiology of Klebsiella pneumoniae: high prevalence of capsular serotype K1 in Taiwan and implication for vaccine efficacy. J. Infect. Dis. 181:2075-2079. [PubMed]
13. Grimont, F., P. A. D. Grimont, and C. Richard. 1992. The genus Klebsiella, p. 2775-2796. In A. Balows, H. G. Trüper, M. Dworkin, W. Harder, and K.-H. Schleifer (ed.), The prokaryotes, vol. 3. Springer-Verlag, New York, N.Y.
14. Hansen, D. S., R. Skov, J. V. Benedi, V. Sperling, and H. J. Kolmos. 2002. Klebsiella typing: pulsed-field gel electrophoresis (PFGE) in comparison with O:K-serotyping. Clin. Microbiol. Infect. 8:397-404. [PubMed]
15. Hobbs, M., and P. R. Reeves. 1994. The JUMPstart sequence: a 39 bp element common to several polysaccharide gene clusters. Mol. Microbiol. 12:855-856. [PubMed]
16. Jonas, D., B. Spitzmüller, F. D. Daschner, J. Verhoef, and S. Brisse. 2004. Discrimination of Klebsiella pneumoniae and Klebsiella oxytoca phylogenetic groups and other Klebsiella species by use of AFLP. Res. Microbiol. 155:17-23. [PubMed]
17. Julianelle, L. A. 1926. Immunological relationships of encapsulated and capsule-free strains of Encapsulatus pneumoniae (Friedlander's bacillus). J. Exp. Med. 44:683-696. [PMC free article] [PubMed]
18. Kauffmann, F. 1949. On the serology of the Klebsiella group. Acta Pathol. Microbiol. Scand. 26:381-406.
19. Kilger, G., and P. A. D. Grimont. 1993. Differentiation of Salmonella phase 1 flagellar antigen types by restriction of the amplified fliC gene. J. Clin. Microbiol. 31:1108-1110. [PMC free article] [PubMed]
20. Ko, W. C., D. L. Paterson, A. J. Sagnimeni, D. S. Hansen, A. Von Gottberg, S. Mohapatra, J. M. Casellas, H. Goossens, L. Mulazimoglu, G. Trenholme, K. P. Klugman, J. G. McCormack, and V. L. Yu. 2002. Community-acquired Klebsiella pneumoniae bacteremia: global differences in clinical patterns. Emerg. Infect. Dis. 8:160-166. [PMC free article] [PubMed]
21. Machado, J., F. Grimont, and P. A. D. Grimont. 2000. Identification of Escherichia coli flagellar types by restriction of the amplified fliC gene. Res. Microbiol. 151:535-546. [PubMed]
22. Mizuta, L. G., M. Ohta, M. Mori, T. Hasegawa, I. Nakashima, and N. Kato. 1983. Virulence for mice of Klebsiella strains belonging to the O1 group: relationship to their capsular (K) types. Infect. Immun. 40:56-61. [PMC free article] [PubMed]
23. Monnet, D. L., J. W. Biddle, J. R. Edwards, D. H. Culver, J. S. Tolson, W. J. Martone, F. C. Tenover, R. P. Gaynes, et al. 1997. Evidence of interhospital transmission of extended-spectrum beta-lactam-resistant Klebsiella pneumoniae in the United States, 1986 to 1993. Infect. Control Hosp. Epidemiol. 18:492-498. [PubMed]
24. Mori, M., M. Ohta, N. Agata, N. Kido, Y. Arakawa, H. Ito, T. Komatsu, and N. Kato. 1989. Identification of species and capsular types of Klebsiella clinical isolates, with special reference to Klebsiella planticola. Microbiol. Immunol. 33:887-895. [PubMed]
25. Nelson, K., and R. K. Selander. 1994. Intergeneric transfer and recombination of the 6-phosphogluconate dehydrogenase gene (gnd) in enteric bacteria. Proc. Natl. Acad. Sci. USA 91:10227-10231. [PMC free article] [PubMed]
26. Ofek, I., K. Kabha, A. Athamna, G. Frankel, D. J. Wozniak, D. L. Hasty, and D. E. Ohman. 1993. Genetic exchange of determinants for capsular polysaccharide biosynthesis between Klebsiella pneumoniae strains expressing serotypes K2 and K21a. Infect. Immun. 61:4208-4216. [PMC free article] [PubMed]
27. Ørskov, I., and M. A. Fife-Asbury. 1977. New Klebsiella antigen K 82 and the deletion of five of those previously assigned. Int. J. Syst. Bacteriol. 27:386-387.
28. Ørskov, I., and F. Ørskov. 1984. Serotyping of Klebsiella. Methods Microbiol. 14:143-164.
29. Platt, H., J. G. Atherton, and I. Ørskov. 1976. Klebsiella and Enterobacter organisms isolated from horses. J. Hyg. Camb. 77:401-408. [PMC free article] [PubMed]
30. Podschun, R., and U. Ullmann. 1992. Klebsiella capsular type K7 in relation to toxicity, susceptibility to phagocytosis and resistance to serum. J. Med. Microbiol. 36:250-254. [PubMed]
31. Podschun, R., and U. Ullmann. 1998. Klebsiella spp. as nosocomial pathogens: epidemiology, taxonomy, typing methods, and pathogenicity factors. Clin. Microbiol. Rev. 11:589-603. [PMC free article] [PubMed]
32. Poh, C. L., S. C. Yap, and M. Yeo. 1993. Pulsed-field gel electrophoresis for differentiation of hospital isolates of Klebsiella pneumoniae. J. Hosp. Infect. 24:123-128. [PubMed]
33. Rahn, A., J. Drummelsmith, and C. Whitfield. 1999. Conserved organization in the cps gene clusters for expression of Escherichia coli group 1 K antigens: relationship to the colanic acid biosynthesis locus and the cps genes from Klebsiella pneumoniae. J. Bacteriol. 181:2307-2313. [PMC free article] [PubMed]
34. Reeves, P. R., M. Hobbs, M. A. Valvano, M. Skurnik, C. Whitfield, D. Coplin, N. Kido, J. Klena, D. Maskell, C. R. Raetz, and P. D. Rick. 1996. Bacterial polysaccharide synthesis and gene nomenclature. Trends Microbiol. 4:495-503. [PubMed]
35. Richard, C. 1973. Antigenic and biochemical study of 500 Klebsiella strains. Ann. Biol. Clin. (Paris) 31:295-303. [PubMed]
36. 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. [PubMed]
37. Simoons-Smit, A. M., A. M. Verweij-Van Vught, I. Y. Kanis, and D. M. MacLaren. 1985. Biochemical and serological investigations on clinical isolates of Klebsiella. J. Hyg. (London) 95:265-276. [PMC free article] [PubMed]
38. Tainturier, D., and C. Richard. 1986. Métrites en série pendant la saison de monte chez des juments de trait breton provoquées par Klebsiella pneumoniae type capsulaire K2, biotype d. Rev. Med. Vet. 137:49-58.
39. Thampapillai, G., R. Lan, and P. R. Reeves. 1994. Molecular evolution in the gnd locus of Salmonella enterica. Mol. Biol. Evol. 11:813-828. [PubMed]
40. Wang, L., K. Andrianopoulos, D. Liu, M. Y. Popoff, and P. R. Reeves. 2002. Extensive variation in the O-antigen gene cluster within one Salmonella enterica serogroup reveals an unexpected complex history. J. Bacteriol. 184:1669-1677. [PMC free article] [PubMed]
41. Wang, L., and P. R. Reeves. 2000. The Escherichia coli O111 and Salmonella enterica O35 gene clusters: gene clusters encoding the same colitose-containing O antigen are highly conserved. J. Bacteriol. 182:5256-5261. [PMC free article] [PubMed]
42. Wong, N. A., C. J. Linton, H. Jalal, and M. R. Millar. 1994. Randomly amplified polymorphic DNA typing: a useful tool for rapid epidemiological typing of Klebsiella pneumoniae. Epidemiol. Infect. 113:445-454. [PMC free article] [PubMed]

Articles from Journal of Clinical Microbiology are provided here courtesy of American Society for Microbiology (ASM)
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...