Multilocus sequence typing and capsule genotyping misclassify K. variicola and K. quasipneumoniae isolates as K. pneumoniae.

Over the years, K. variicola and K. quasipneumoniae strains have been misclassified as K. pneumoniae, resulting in the deposition of misidentified reference strains in public repositories, including the public K. pneumoniae multilocus sequence typing (MLST) schema (21,–24). We used the whole-genome sequence data to generate a K. pneumoniae MLST assignment for each strain. In our collection of 13 ESBL-producing K. variicola isolates, 6 (46%) had a defined K. pneumoniae MLST type and 3 (23%) had a novel combination of known K. pneumoniae MLST alleles (Table 2). Among the 15 ESBL-producing K. quasipneumoniae isolates, 12 (80%) had a defined K. pneumoniae MLST type, and 1 (7%) had a novel combination of known alleles (Table 2). Similarly, among the 12 non-ESBL-producing K. quasipneumoniae strains, 3 (25%) had a defined K. pneumoniae MLST type, and 6 (50%) had a novel combination of known alleles (Table 2). Consistent with the extensive genetic diversity observed within each Klebsiella clade (Fig. 2 and ​and3),3), multiple different MLST designations were found for K. variicola and K. quasipneumoniae strains in our collection (Table 2).

Open in a separate window

FIG 2

Phylogenetic tree of K. quasipneumoniae from humans. Polymorphisms in the K. quasipneumoniae strains were called against the 700603 reference genome (black cross). The clade KpIIA is at the top left, while the KpIIB clade is at the bottom right. ESBL-producing strains sequenced for this study are represented by circles. Non-ESBL-producing strains of both clades identified in our collection are represented by diamonds. Strains previously sequenced by Holt et al. (8) are represented by squares. Strains associated with human invasive infection sites are indicated in red, and human colonization is indicated in blue.

Open in a separate window

FIG 3

Phylogenetic tree of K. variicola from humans and animals. Polymorphisms in the K. variicola strains were called against the At-22 reference genome (black cross). All K. variicola isolates identified were ESBL producers. There is one outlier Klebsiella, KPN1705, which projects beyond the border of the figure to the top left. Strains sequenced for this study are represented by circles, while those previously sequenced by Holt et al. (8) are represented by squares. Strains associated with human invasive infection sites are indicated in red, human colonization is indicated in blue, and animal-associated isolates are indicated in green.

Capsule serotype has been used for close to a century to classify Klebsiella species in general (25). Capsule genotype, as assessed by sequencing the genes in the capsule locus, has been used more recently as a molecular tool to subclassify K. pneumoniae strains (26,–28). Among our collection of K. variicola and K. quasipneumoniae isolates, every strain had a K. pneumoniae capsule locus detected (Table 2). Again, consistent with the extensive genetic diversity observed within each clade, multiple different capsule genotypes were found, including some strains with novel combinations of capsule genes (Table 2). Taken together, these data demonstrate that the majority of K. variicola and K. quasipneumoniae strains recovered from human patients would be misidentified by typing methods such as K. pneumoniae MLST and capsule genotyping.

Whole-genome sequencing reveals extensive genetic diversity within and between Klebsiella clades.

To begin assessing phylogenetic relationships between K. pneumoniae, K. variicola, and K. quasipneumoniae strains, single nucleotide polymorphisms (SNPs) were determined relative to the ST258 K. pneumoniae reference genome NJST258_2 (GenBank accession number CP006918.1) (6). Consistent with a previous report using strains collected from a variety of human, animal, and environmental sources (8), each Klebsiella species formed a distinct clade, with K. quasipneumoniae divided into two subclades (Fig. 1). Phylogenetic analysis demonstrated that the ESBL-producing and non-ESBL-producing K. quasipneumoniae strains are intermingled on the cladogram, suggesting that they are derived from a common genetic pool (Fig. 2). However, this does not imply that our ESBL-producing and non-ESBL-producing K. quasipneumoniae strains are identical, because both groups include strains with a diverse array of different MLST, capsule, and plasmid genotypes (Tables 2 and ​and33).

Phylogenetic comparison of our K. quasipneumoniae isolates recovered from infected patients with publicly available sequence data from strains collected from humans (8) showed much genetic intermixing (Fig. 2). Similarly, analysis of K. variicola strains recovered from our patients with invasive infection (Table 1) or colonized humans and animals (8) also revealed no clear phylogenetic segregation (Fig. 3). Taken together, these data suggest that K. quasipneumoniae and K. variicola strains recovered from infected patients and colonized humans or animals are derived from a common genetic pool.

One outlier strain, KPN1705, was present in the K. variicola population (Fig. 3). The KPN1705 strain was allied phylogenetically with K. variicola, yet it was 251,390 SNPs distant from the K. variicola At-22 reference genome. This was striking compared to the average distance to the At-22 reference of 38,056 SNPs (range, 31,777 to 45,299) in our other K. variicola strains. Since hybrid strains of Klebsiella with large chromosomal recombination events have been described previously (8), we compared the reads from strain KPN1705 against At-22 (K. variicola), NJST258_2 (K. pneumoniae), and 700603 (K. quasipneumoniae) reference genomes. The SNP distribution was uniform across the three reference genomes, which made the possibility of KPN1705 being the result of a single large recombination event less likely (data not shown). To further evaluate its phylogeny, we determined that KPN1705 was a single locus variant of the K. pneumoniae ST1155. The only other ST1155 strain reported in the literature is designated 10982 and is proposed to represent a novel species (22). Strain 10982 was isolated from a perianal swab from an intensive care unit (ICU) patient in Maryland in 2005. Thus, our strain KPN1705 appears most closely related to a novel Klebsiella species that colonizes the gastrointestinal (GI) tract of humans and may be capable of causing human disease.

SHV-LEN-OKP core chromosomal beta-lactamases are not restricted by Klebsiella species.

The SHV-LEN-OKP beta-lactamases are core chromosomal genes of Klebsiella that have been suggested to be differentiators of Klebsiella species: K. pneumoniae (SHV restricted), K. quasipneumoniae (OKP restricted), and K. variicola (LEN restricted) (10, 29, 30). However, one potential confounder is that the SHV beta-lactamase genes can also be carried on plasmids (31). We next used the whole-genome sequence data to assess SHV-LEN-OKP gene content in our 40 K. variicola and K. quasipneumoniae strains. Among our collection, 7/13 (54%) K. variicola strains carried SHV on plasmids in addition to the chromosomal LEN gene (Table 2). Similarly, 7/15 (47%) K. quasipneumoniae strains carried SHV on a plasmid in addition to the chromosomal OKP gene. Only 1/12 (8%) non-ESBL-producing K. quasipneumoniae isolate carried SHV on a plasmid in addition to the chromosomal OKP genes. The presence of SHV on plasmids, sometimes in multiple copies, may complicate the identification and analysis of Klebsiella strains.

Unexpectedly, we discovered one K. variicola strain (designated KPN807; Table 2) carrying a chromosomal copy of OKP rather than the expected LEN gene that is typical of K. variicola strains. This OKP gene was carried within a 7-kb segment of chromosomal DNA that had a higher sequence identity to K. quasipneumoniae than K. variicola, suggesting recombination of this region (Fig. 4).

Open in a separate window

FIG 4

Comparison of the 7-kb region of recombination in strain KPN807. KPN807 assembled contigs are compared to the At-22 K. variicola reference genome (top) and the 700603 K. quasipneumoniae reference genome (bottom). The consensus is indicated by a red bar, with SNPs marked as colored vertical lines. KPN807 has a marked increase in SNP density in the center of the region relative to At-22 and the flanking regions relative to 700603, suggesting that this 7-kb region of the chromosome has recombined into the K. variicola KPN807 from a K. quasipneumoniae isolate.

K. variicola and K. quasipneumoniae have similar antimicrobial resistance and plasmid replicon content to K. pneumoniae, including KPC and NDM-1.

The emergence of multidrug resistance among K. pneumoniae strains is a cause of public concern. A table of the MIC values for the 40 K. variicola and K. quasipneumoniae strains is provided in Table S1 in the supplemental material. Next, we used the whole-genome sequence data to compare plasmid replicons carried by the different Klebsiella species. There were many plasmid replicons and antimicrobial resistance genes among the ESBL-producing K. variicola and K. quasipneumoniae isolates that were also identified in K. pneumoniae (Table 3) (18). In general, plasmid replicons tended to be associated with genomically allied strains (Fig. 5). In particular, the FIBk, FIIk, and FII replicons were detected in the majority of our K. variicola and K. quasipneumoniae isolates. Several strains also possessed the KPC-2 carbapenemase and CTX-M beta-lactamase genes associated with multiple drug resistance. Of note, KPC-positive K. variicola and K. quasipneumoniae strains have not been previously reported in the United States.

Open in a separate window

FIG 5

Beta-lactamase gene content of K. pneumoniae, K. variicola, and K. quasipneumoniae. The phylogenetic relationship between strains based upon the presence of beta-lactamase genes is shown on a rectangular cladogram. Polymorphisms were called against the K. pneumoniae reference genome NJST258_2. K. quasipneumoniae are represented by circles, while K. variicola are represented by squares. ESBL-producing Klebsiella are shown in red, while non-ESBL producing Klebsiella are shown in blue. To the right of the cladogram, the common beta-lactamase genes are shown. The first column lists the core genome SHV-OKP-LEN beta-lactamase: LEN in red, OKP-A in green, and OKP-B in blue. No SHV was found in the core chromosome. The most common allele(s) for OKP and LEN is represented with a darker shade of the primary color. The second column represents the plasmid SHV-OKP-LEN beta-lactamase content, with the most common allele SHV-12 in purple and all others in gray. The third column indicates the CTX-M-15 alleles in teal, with CTX-M-3 in gray. The fourth column shows the TEM alleles, with the most common allele TEM-198 in orange and all others in gray. The fifth and sixth columns represent the presence of KPC-2 and NDM-1 alleles in black, respectively.

Unexpectedly, one K. variicola strain (designated KPN1481; Table 3) carried a plasmid, pKPN1481-1, with the New Delhi metallo-beta-lactamase 1 (NDM-1) gene (GenBank accession numbers CP020847 to CP020852). This represents the first NDM-1-producing K. variicola strain to be reported. To better characterize the plasmid and strain containing the NDM-1 gene, it was sequenced to closure using single molecule real-time (SMRT) sequencing. Results confirmed the presence of the NDM-1 gene on pKPN1481-1, a 342-kb plasmid (18). Alignment of pKPN1481-1 to NDM-1 gene-containing plasmids from K. pneumoniae strains recovered from our health care system and elsewhere revealed extensive similarity (18, 32, 33). Importantly, the 10,273-bp region containing the NDM-1 gene was identical in sequence to four NDM-1 plasmids carried by K. pneumoniae isolates in our collection.

Gene content comparison between K. pneumoniae, K. variicola, and K. quasipneumoniae.

We compared the gene content between our ESBL-producing K. variicola strains, K. quasipneumoniae strains, and 12 representative K. pneumoniae strains from our previously published collection (18). We identified a total of 30,075 unique genes present in the pangenome. A Klebsiella core genome was identified consisting of 2,800 unique genes that were present in >95% of the isolates. Most of the accessory genes were present in less than 15% of the isolates (23,334 genes; 77.5% of the total). A binary tree of accessory gene content demonstrates that the accessory gene content is largely differentiated by the three major clades, suggesting a clade-specific core genome for each ESBL Klebsiella species as well as the presence of unique mobile genetic content (Fig. 6). This finding is further reinforced when considering the core genome of the K. variicola and ESBL-producing K. quasipneumoniae in our study. The ESBL-producing K. quasipneumoniae strains have a core genome of 3,338 unique genes, while the K. variicola strains have a core genome of 3,960 genes. This difference may be in part a reflection of the division of K. quasipneumoniae into two distinct clades. The presence of nif operon genes, which facilitate nitrogen fixation, has been associated with agricultural isolates of Klebsiella. We found nif genes present in 12/13 (92.3%) ESBL-producing K. variicola human isolates, 11/15 (73.3%) ESBL-producing K. quasipneumoniae isolates, and 10/12 (83.3%) non-ESBL-producing K. quasipneumoniae isolates, suggesting that nif genes may persist in isolates found in human infections. The presence or absence of select genes is included in Tables S2 and S3.

Open in a separate window

FIG 6

Binary accessory genome tree of K. pneumoniae, K. variicola, and K. quasipneumoniae. The phylogenetic relationship between strains based upon the presence or absence of genes in the pangenome is shown. The three clades are visible, with the K. pneumoniae strains (black circles), K. variicola strains (blue circles), and K. quasipneumoniae strains (magenta circles) indicated. The K. pneumoniae strains separate into two clades based upon their MLST type, while the K. quasipneumoniae strains separate into the two distinct clades representing K. quasipneumoniae subsp. quasipneumoniae and K. quasipneumoniae subsp. similipneumoniae. One strain of K. quasipneumoniae (KPN1470) is allied with K. pneumoniae, due to similarities in accessory genome content.

Whole-genome sequencing of Klebsiella.

To prepare whole-genome sequencing libraries, the cryopreserved stocks were grown on tryptic soy agar containing 5% sheep blood. Genomic DNA was extracted using standard methods (Qiagen, Valencia, CA), and NexteraXT libraries were prepared using the manufacturer’s protocols (Illumina, San Diego, CA) and sequenced on an Illumina MiSeq or NextSeq instrument.

Bioinformatic analysis of strains.

The single nucleotide polymorphism calling pipeline and additional bioinformatic pipelines were described previously (18). Strains from Holt et al. were downloaded from European Nucleic Acid Archive (accession number ERP000165). BLAST was performed using the NCBI BLAST toolkit as well as CLC Genomics Workbench v.10.1. Visualization of SNP distribution was performed using CLC Genomics Workbench v.10.1. FASTQ files were assembled into contigs using Spades v3.9.0, and contigs were annotated using Prokka v1.12 (43, 44). Gene content analysis was performed using Roary v3.6.1 (45). The 12 strains of ESBL-producing K. pneumoniae included for pangenome comparison are KPN1, KPN2, KPN9, KPN11, KPN12, KPN17, KPN18, KPN133, KPN1998, KPN2000, KPN2108, and KPN2129 (18). Assembly of SNPs into phylogenetic trees was accomplished with the scripts prephix v3.3.0, phrecon v4.6.0, and FastTreeMP v2.1 (46). Prephix and phrecon are available from GitHub (https://github.com/codinghedgehog).

MALDI-TOF MS identification.

Identification of isolates was performed in the Houston Methodist clinical microbiology laboratory using a Bruker Biotyper MALDI-TOF MS as part of the standard clinical microbiology practice as described previously (19). Briefly, colonial material from agar plates or pelleted cells from liquid blood cultures were transferred to the target plate, dried at room temperature for approximately 2 min, and covered with alpha-cyano-4-hydroxycinnamic acid (HCCA) matrix. Spectra were collected using the default instrument settings and interpreted using the research use only microorganism reference library on the Biotyper (Bruker, Billerica, MA). At the time of isolation in 2011 to 2015, Bruker microorganism reference library versions starting with 4.0.0.0 were used. The strains were reanalyzed in 2017 using version 6.0.0.0, which was installed in January 2017.

This work was supported by funds from the Fondren Foundation and the National Institute of Allergy and Infectious Diseases, National Institutes of Health, Department of Health and Human Service (contract HHSN272201400027C).

We acknowledge Kathryn Holt for graciously providing sequence data and Kathryn Stockbauer for help in preparing the manuscript.