Genomic Epidemiology of an Outbreak of Klebsiella pneumoniae ST471 Producing Extended-Spectrum β-Lactamases in a Neonatal Intensive Care Unit

Introduction

Klebsiella pneumoniae, one of the “ESKAPE” pathogens, has emerged as an important nosocomial pathogen that mainly causes infections in the respiratory tract, urinary tract, and blood.^1,2 The prevalence of multi-drug resistant (MDR) K. pneumoniae has increased dramatically in recent years. MDR K. pneumoniae frequently exhibits resistance to extended-spectrum cephalosporins because of its production of extended-spectrum β-lactamases (ESBLs).³ ESBL-producing bacteria cause higher morbidity, mortality, and fiscal burden.^4,5 This limits efficient clinical treatment tremendously, resulting in undesirable treatment outcomes.

The outbreak of multidrug-resistant bacteria is a serious public health problem, and nosocomial outbreaks caused by ESBL-producing K. pneumoniae have been reported worldwide.^6–11 ESBL-producing K. pneumoniae is the most frequently implicated pathogen causing outbreaks in neonatal intensive care units (NICUs), with mortality rates up to 31% among infected infants.¹² Traditional molecular typing methods, such as pulsed-field gel electrophoresis (PFGE) and multilocus sequence typing (MLST), can define the outbreak strains well, but cannot reveal the transmission routes between strains.

Over the past decade, whole-genome sequencing (WGS) technology has developed rapidly and has been used widely in epidemiological investigations of bacterial infectious diseases, to identify outbreaks,^13,14 trace pathogen transmission,^15,16 discover new modes or ways of transmission,¹⁷ identify new clones,¹⁸ and track nosocomial infections and outbreaks.¹⁹ In the present study we report an outbreak of K. pneumoniae producing ESBLs that took place in a NICU and attempt to reveal the fine transmission route of the strains within this outbreak.

Materials and Methods

Review of NICU Records and Collection of Patient Information

The outbreak occurred in the NICU of the Affiliated Hospital of Chifeng University, a 2000-bed general tertiary care and University-affiliated teaching hospital in Inner Mongolia, China. This hospital accepts 2200 outpatient and emergency patients each day. The NICU has a total of 25 beds.

The medical records of patients from whom ESBL-producing K. pneumoniae was isolated were reviewed, including time of stay in the NICU, bed site, clinical symptoms, diagnosis, data of the ESBL-producing K. pneumoniae isolation, and outcomes.

Bacterial Identification and Antimicrobial Susceptibility

Bacterial identification was initially performed using the VITEK-2 automated system (BioMérieux, Craponne, France) and further confirmed using matrix-assisted laser desorption/ionization time of flight mass spectrometry (MALDI-TOF-MS) (Bruker, Leipzig, Germany).

Antibiotic susceptibility testing was performed using the broth dilution method. We measured the minimum inhibitory concentration (MIC) of ampicillin (AMP), ampicillin/sulbactam (AMS), amoxicillin/clavulanate (AMC), cefazolin (CFZ), cefoxitin (CFX), cefotaxime (CTX), ceftazidime (CAZ), cefepime (FEP), aztreonam (AZM), imipenem (IMI), meropenem (MEM), amikacin (AMI), kanamycin (KAN), azithromycin (AZI), tetracycline (TET), doxycycline (DOX), tigecycline (TGC), ciprofloxacin (CIP), levofloxacin (LEV), trimethoprim-sulfamethoxazole (SXT), chloramphenicol (CHL), and colistin (CO). All antibiotics, except tigecycline and colistin, were interpreted according to the approved standard of Clinical and Laboratory Standards Institute (CLSI) 2016 guidelines.²⁰ Susceptibility to tigecycline and colistin was defined based on the criteria proposed by the European Committee on Antimicrobial Susceptibility Testing-2018 (http://www.eucast.org/fileadmin/src/media/PDFs/EUCAST_files/Breakpoint_tables/v_8.0_Breakpoint_Tables.pdf). Escherichia coli ATCC 25922 was used as quality control strain for the antibiotic susceptibility testing.

We screened genes for resistance to beta-lactamase (bla_CTX-M, bla_TEM, bla_SHV, bla_OXA, bla_PER, bla_VEB) in all strains, as described previously.²¹ DNA sequencing was performed on both strands of the PCR amplification products. The results were compared and aligned with reference sequences using the online BLAST database.

Pulsed-Field Gel Electrophoresis (PFGE)

We used the 1-day, standardized PFGE protocol for K. pneumoniae.²² Cell suspensions were placed in polystyrene tubes (Falcon 12 × 75 mm, from Fisher Scientific, Hampton, NH, USA), and their optical density was adjusted to 3.8–4.0 using a Densimat photometer (BioMérieux). K. pneumoniae slices were digested using 50 U per slice of XbaI (Takara, Dalian, China) for 4 h at 37 °C, and electrophoresis was performed using a CHEF-DRIII system (Bio-Rad Laboratories, Hercules, CA, USA). Electrophoresis was performed using a switch time of 6 s to 36 s for 18.5 h, and images were captured using a Gel Doc 2000 system (Bio-Rad) and converted to TIFF files. The TIFF files were analyzed using BioNumerics version 5.1 software (Applied Maths, Kortrijk, Belgium).

Multilocus Sequence Typing (MLST)

MLST using seven genes (gapA, infB, mdh, pgi, phoE, rpoB, and tonB) was performed on the isolates according to the protocol described on the K. pneumoniae MLST website (http://bigsdb.pasteur.fr/). Alleles and sequence types (STs) were assigned using the MLST database (http://bigsdb.pasteur.fr/klebsiella/klebsiella.html).

Whole-Genome Sequencing (WGS), Detection of Single Nucleotide Polymorphisms (SNPs), and Minimum Core Genome (MCG) Sequence Typing

The ten outbreak strains were selected for WGS. Genomic DNA from the bacterial strains was sequenced using Illumina sequencing by constructing two paired-end (PE) libraries with average insertion lengths of 500 bp and 2000 bp, respectively. Sequences were generated using an Illumina GA IIx apparatus (Illumina Inc., San Diego, CA, USA). Raw data was processed in four steps, including removing reads with 5 bp of ambiguous bases, removing reads with 20 bp of low quality (≤ Q20) bases, removing adapter contamination, and removing duplicated reads. Finally, 100× libraries were obtained with clean PE read data. Assembly was performed using SOAPdenovo v1.05.²³

The whole-genome sequence of K. pneumoniae KPR0928 (NCBI accession number: NZ_CP008831.1) was used as the reference sequence, and clean reads of the sequenced isolates were mapped to the reference genome using bowtie 2 software under the default parameters.²⁴ SNPs were then identified using Samtools and combined together according to the reference.²⁵ SNPs with low quality reads (read depth < 5) and those located within 5 bps on the chromosome were removed to avoid the effect of recombination, as described in previous studies.^17,26 The isolates were clustered and a heatmap was generated using the heatmap package in R. The transmission route was then reconstructed based on the emergence of different SNPs in each isolate and the case information, including the time of the onset of infection and the hospitalization time.

The multilocus sequences were identified using the pMLST 2.0 database (https://cge.cbs.dtu.dk/services/MLST/).²⁷ Antimicrobial resistance genes were identified by searching the Comprehensive Antibiotic Research Database (CARD, https://card.mcmaster.ca/).²⁸

Nucleotide Sequence Accession Numbers

The data from the WGS project have been deposited at GenBank under the BioProject ID PRJNA578669.

Results

Overall Description of the Outbreak

In mid-January, 2015, we identified five cases (cases 1–5 in Table 1 and Figure 1) of neonatal pneumonia caused by ESBL-producing K. pneumoniae in the NICU of our hospital. This NICU consists of two units with a total of four wards and 25 beds, and has been open since 2013.

Table 1 Descriptions of the Five Cases Involved in the February 16th Outbreak in the NICU

Figure 1 The timeline of patient admission and ESBL-producing K. pneumoniae isolation. Shadows on the timeline represent the duration of the case. Different wards are indicated in different colors. The red arrows indicate the isolation of the strains. Light red arrows represent sampling where no K. pneumoniae was isolated.

The index case (case 1) of this outbreak was identified on February 13, 2015. On February 16, four more cases (cases 2–5) in the same unit in the NICU were observed (Table 1). The timeline of patient admission and ESBL-producing K. pneumoniae isolation is shown in Figure 1. An outbreak was declared and stringent infection control measures were implemented, including contact precautions, strengthening of hand hygiene, environmental cleaning, and enhancing antimicrobial stewardship.

Although stringent infection control measures were carried out, seven more ESBL-producing K. pneumoniae were isolated during the hospitalization of the five outbreak cases; among them, four, one, and one strains were isolated from cases 2, 3, and 5, respectively, and another one was isolated from a new case (case 7, Figure 1) in this NICU. Unfortunately, we failed to recover four of the strains (three from case 2 and one from case 5) because of poor preservation; therefore, they could not be used for subsequent experiments.

From March 27, 2015 (the discharge date of the last patients with ESBL-producing K. pneumoniae), over a period of six months, no further ESBL-producing K. pneumoniae were isolated in the NICU.

Searching for Outbreak-Related Strains Using PFGE

PFGE was used to screen outbreak-related K. pneumoniae strains among the strains isolated from the same hospital during or around the outbreak. A total of 22 K. pneumoniae strains, which were isolated between January 4 and March 17, were analyzed by PFGE. Interestingly, two strains (CF125 and CF934) were clustered together with the eight outbreak strains (Figure 2). CF125 was isolated from case 1 on January 20, 23 days before the outbreak. CF934 was isolated from the tip of a trachea cannula of a new case (case 6, Figure 1) who had been in NICU before, on January 17, 26 days before the outbreak.

Figure 2 Clustering of the 30 K. pneumoniae strains based on their PFGE patterns. The outbreak and outbreak-related strains are framed. The information for strain ID, isolation date, ward, type of specimen, and PFGE type is listed to the left of the patterns. Abbreviations: NICU, neonatal intensive care unit; PFGE, pulsed-field gel electrophoresis.

PFGE divided the ten outbreak-related strains into five different PFGE types (PT01–PT05). The dominant PFGE type (PT04) contained five isolates. The other four PFGE types showed one to three bands that were different to PT04. All ten isolates were the “same” or “closely-related” strains according to the interpretation criteria of PFGE patterns proposed by Tenover et al.²⁹ The other 20 K. pneumoniae strains isolated from the same hospital during or around the outbreak showed more than 20 bands that were different to outbreak-related strains.

Using MLST, all ten outbreak-related strains were assigned as ST471 (allele numbers of 3, 1, 1, 68, 4, 1, and 39 for gapA, infB, mdh, pgi, phoE, rpoB, and tonB, respectively). By searching the K. pneumoniae MLST database (http://bigsdb.pasteur.fr.), we found that there was only one ST471 strain in the database, which was isolated from a human from Spain in 2012.

Antibiotic Susceptibility and Characterization of Resistance Genes

All 10 ESBL-producing K. pneumoniae strains showed same results for the antibiotic susceptibility test. All strains showed insensitivity to ampicillin, ampicillin/sulbactam, amoxicillin/clavulanate, cefazolin, cefotaxime, cefepime, trimethoprim-sulfamethoxazole, and aztreonam, and were susceptible to other 14 antibiotics (cefoxitin, ceftazidime, imipenem, meropenem, amikacin, kanamycin, azithromycin, tetracycline, doxycycline, tigecycline, ciprofloxacin, levofloxacin, chloramphenicol, and colistin). All ten K. pneumoniae strains were confirmed as positive for ESBL production using double-disk diffusion tests.

By PCR and sequence alignment of beta-lactamase genes, all 10 isolates were shown to harbor bla_CTX-M-15, bla_TEM-1, bla_SHV-108, and bla_OXA-1 genes, and tested negative for bla_PER, and bla_VEB. Based on the WGS results, we further characterized the antibiotic resistance genes among the 10 isolates. In addition to the four beta-lactamase genes (bla_CTX-M-15, bla_TEM-1, bla_SHV-108, and bla_OXA-1), all isolates were confirmed to possess genes for resistance to aminoglycoside (aac(3)-II, aac(6ʹ)-Ib, aph(33)-Ib and aph(6)- Id), sulfonamide (sul2), fosfomycin (fosA), trimethoprim (dfrA14, dfrA22), chloramphenicol (catB3), polymyxin (arnA), and bacitracin (bacA). Furthermore, several resistance pumps were identified, including a plasmid-mediated tetracycline resistance pump (tet_efflux).

Comparison of Outbreak Isolates Based on WGS-Based SNPs

WGS and MCG typing were performed for the 10 ESBL-producing K. pneumoniae strains, with the aim of using the SNPs found in their genomes to determine a putative transmission map of this outbreak. Genomic comparisons revealed a total of 7 MCG SNPs among the 10 strains (Figure 3). The evolutionary relationships based on MCG typing of the 10 strains are outlined and presented in Figure 4. The strains isolated from case 6 (CF934) in January showed the same SNP profile as the reference strain K. pneumoniae KPR0928. The first strain isolated from case 1 (CF125) had one SNP (T608763C) that was different to the reference strain. Another case 1 strain (CF728) had one more SNP (C154003T). The first strain (CF987) isolated from case 2 had the same SNP (T608763C) to that of the first case 1 strain (CF125). Another case 2 strain (CF124) had one more SNP (G953229A). The strains isolated from cases 4 (CF119) and 5 (CF127), both of which were isolated on February 16, had one (G2665757T) and two (C996839T, C2820669A) SNPs that were different to the case 1 and 2 strains (CF125, CF987), respectively. The first strain (CF919) of case 3, which was also isolated on February 16, had one SNP (G2550784A) that was different to the case 4 strain (CF119). The second case 3 strain (CF126), which was isolated 14 days later, showed one more SNP (G3771456A).

Figure 3 Heat map of SNP distribution of ten ESBL-producing K. pneumoniae.

Figure 4 Evolutionary relationships based on MCG typing of ten ESBL-producing K. pneumoniae. Isolates from different cases are indicated in different colors. SNPs are marked on the connecting lines.

Identification of the Origin and the Construction of a Putative Transmission Route of the Outbreak

The genomic and epidemiological information were integrated to construct the transmission route. The most likely transmission route is shown in Figure 5.

Figure 5 Putative map of ESBL-producing K. pneumoniae transmission during the outbreak. The transmission map was constructed using genetic and patient trace data. Nodes represent patients, and arrows indicate a transmission event directly from one patient to another.

First, case 6 was suspected of being the source of the outbreak isolates, because the historical strain from case 6 showed the same SNP patterns to the reference strain K. pneumoniae KPR0928. This was also supported by the epidemiological information. The case 6 strain was the earliest isolated ESBL-producing K. pneumoniae strain, which was isolated 1 month before the February 16 outbreak. Thus, case 6 was the most likely source of the outbreak isolates.

Second, the first strains isolated from cases 1 and 2 showed one SNP difference to the isolates of case 6, suggesting that the strains from cases 1 and 2 were not transmitted from case 6. However, combining the epidemiological information, that the first strain of case 1 was isolated on January 20, closer to the isolation of case 6 strain; and the first strain of case 2 was isolated on February 16, suggesting that the most possible transmission route was that this ESBL-producing K. pneumoniae strain was passed from case 6 to case 1 and then to case 2. Interestingly, the second strains from both cases 1 and 2 had one SNP difference with the first strains, suggesting that there were mutations in vivo.

Third, the strains from cases 4 and 5 had one and two SNPs different to the first strains of cases 1 and 2. However, the second strain of case 1 was isolated three days before the isolation of the strains from cases 4 and 5 and showed another SNP, which suggested that the strains of cases 4 and 5 were not transmitted from case 1. Thus, case 2 was the most likely source of the strains isolated from cases 4 and 5.

Fourth, the first strain of case 3, which was also isolated on February 16, had one SNP difference with the case 4 strain. The second case 3 strain, which was isolated 15 days later, showed one more SNP. Thus, the isolates of case 3 were transmitted from case 4, which then persisted in case 3 and mutated in vivo.

Fifth, the strain from case 7 was isolated 20 days after the February 16 outbreak and showed same SNP profile as the second strain of case 2, suggesting that case 2 directly transmitted this strain to case 7.

On the whole, consistent with the genomic and epidemiological information, the integrated map identified that: (1) Case 6 was the source of the February 16 outbreak; (2) before the outbreak, the ESBL-producing K. pneumoniae was a long-term resident in case 1; (3) case 2 played a key role, as a “superspreader”, in the outbreak, probably directly causing cases 4, 5, and 7, and indirectly causing case 3.

Discussion

In this study, we described an outbreak caused by ESBL-producing K. pneumoniae in a NICU of a large university hospital in China. The ESBL-producing K. pneumoniae isolates belonged to ST471, and all harbored bla_CTX-M-15, bla_TEM-1, bla_SHV-108, and bla_OXA-1 beta-lactamase genes. ST471 is a highly uncommon ST, which has only been reported in one human case in Spain and clinical bovine mastitis in Tunisia.³⁰ The Tunisian study pointed out the necessity to improve farm management to avoid the dissemination in this sector of ESBL-producing K. pneumoniae. It is worth noting that the outbreak reported here occurred in Chifeng City, which belongs to Inner Mongolia, the largest pastoral area in China. Therefore, a future investigation should determine whether this ST471 strain infected humans via cattle, sheep, or other livestock. Furthermore, as far as we know, this is not only the first study to report an outbreak caused by K. pneumoniae belonging ST471, but also is the first to report an outbreak caused by a K. pneumoniae strain co-harboring the bla_CTX-M-15, bla_TEM-1, bla_SHV-108 and bla_OXA-1 genes, which suggests that more attention should be paid to the K. pneumoniae isolates producing epidemic ESBLs.

CTX-M-15 is the main CTX-M-type of clinical ESBL-producing K. pneumoniae in China and other parts of the world.^31–37 Furthermore, bla_CTX-M-15-positive K. pneumoniae has been also widely detected in various animals, including farm cattle, pet dogs, and wild birds,^38–40 indicating communication between the genetic pools of the bla_CTX-M-15 gene in animals and humans. In future studies, we should investigate how these ESBLs are transmitted from animals to humans, and then implement measures to block this route to avoid such infections and outbreaks. TEM-1 is the main TEM-type of K. pneumoniae isolated from Tanzania, India, and Iran.^41–43 OXA-1 and SHV-108 are rarely detected from K. pneumoniae. In a recently multicenter survey of ESBL-producing K. pneumoniae causing intra-abdominal infections from nine tertiary hospitals in China, no TEM-type and OXA-1 strain was detected, and the most common genotype for K. pneumoniae was SHV-2a.³² However, in another survey of K. pneumoniae from pediatric patients in Shanghai, China, TEM-1 and OXA-1 were detected in 68.3% and 14.6% of strains.⁴⁴ Although there is no global survey of the distribution of these resistance genes at present, it can be seen from the current research reports that there are regional differences in the distribution of these beta-lactamase genes between different countries, different cities, even between different hospitals of the same region.

PFGE used to be the gold standard for molecular subtyping of outbreaks. However, although PFGE could identify outbreak-related strains, it could not reveal the evolutionary relationship and transmission route between them. Our previous research showed that WGS and MCG typing could reveal the details of transmission within a K. pneumoniae nosocomial outbreak.¹⁶ MCG typing was first used to study the population structure of Streptococcus suis and then Legionella pneumophila.^17,26 In the present study, we further confirmed that WGS and MCG typing can reveal the fine transmission map and identify the “superspreader” of an outbreak. If WGS and MCG typing were applied in the early stage of an outbreak, the outbreak could be controlled. In addition, this method has the potential to be used to track other human-borne pathogens, such as Neisseria meningitidis, Bordetella pertussis, and Streptococcus pneumoniae.

Conclusion

In summary, we describe the integration of epidemiological information with high-resolution WGS and MCG typing to enhance the investigation of an outbreak of ESBL-producing K. pneumoniae infection. Our data showed that epidemiological inferences are greatly improved by interpretation in the context of WGS and that K. pneumoniae strains isolated from the same outbreak contain sufficient SNP differences to refine the epidemiological linkages on the basis of genetic lineage.