Whole Genome Sequencing of Extended Spectrum β-lactamase (ESBL)-producing Klebsiella pneumoniae Isolated from Hospitalized Patients in KwaZulu-Natal, South Africa

Extended spectrum β-lactamase (ESBL)-producing Klebsiella pneumoniae remain a critical clinical concern worldwide. The aim of this study was to characterize ESBL-producing K. pneumoniae detected within and between two hospitals in uMgungundlovu district, South Africa, using whole genome sequencing (WGS). An observational period prevalence study on antibiotic-resistant ESKAPE (i.e. Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, Enterobacter spp.) bacteria was carried out in hospitalized patients during a two-month period in 2017. Rectal swabs and clinical specimens were collected from patients hospitalized and were screened for ESBL-producing, Gram-negative ESKAPE bacteria using cefotaxime-containing MacConkey agar and ESBL combination disk tests. Nine confirmed ESBL-K. pneumoniae isolated from six patients and two hospitals were whole genome sequenced using an Illumina MiSeq platform. Genome sequences were screened for presence of integrons, insertion sequences, plasmid replicons, CRISPR regions, resistance genes and virulence genes using different software tools. Of the 159 resistant Gram-negative isolates collected, 31 (19.50%) were ESBL-producers, of which, nine (29.03%) were ESBL-K. pneumoniae. The nine K. pneumoniae isolates harboured several β-lactamase genes, including blaCTX-M-15, blaTEM-1b, blaSHV-1, blaOXA-1 concomitantly with many other resistance genes e.g. acc(6′)-lb-cr, aadAI6, oqxA and oqxB that confer resistance to aminoglycosides and/or fluoroquinolones, respectively. Three replicon plasmid types were detected in both clinical and carriage isolates, namely ColRNAI, IncFIB(K), IncF(II). Sequence type ST152 was confirmed in two patients (one carriage isolate detected on admission and one isolate implicated in infection) in one hospital. In contrast, ST983 was confirmed in a clinical and a carriage isolate of two patients in two different hospitals. Our data indicate introduction of ESBL-producing K. pneumoniae isolates into hospitals from the community. We also found evidence of nosocomial transmission within a hospital and transmission between different hospitals. The Clustered Regularly Interspaced Palindromic Repeats (CRISPR)-associated cas3 genes were further detected in two of the nine ESBL-KP isolates. This study showed that both district and tertiary hospital in uMgungundlovu District were reservoirs for several resistance determinants and highlighted the necessity to efficiently and routinely screen patients, particularly those receiving extensive antibiotic treatment and long-term hospitalization stay. It also reinforced the importance of infection, prevention and control measures to reduce the dissemination of antibiotic resistance within the hospital referral system in this district.

www.nature.com/scientificreports www.nature.com/scientificreports/ Multi-locus sequence type analysis (MLst) and core genome multi-locus sequence type analysis (cgMLst). Analyses of MLST profiles has shown high variation among the seven housekeeping genes and identified five different sequence types (STs) including ST152 (n = 4), ST983 (n = 2), and three singleton ST432, ST607 and ST17 ( Table 2). The four K. pneumoniae ST152 strains isolated from two patients were detected in clinical (n = 1) and carriage (n = 3) samples in the tertiary hospital while the two K. pneumoniae ST983 were each identified in carriage and clinical sample of patients admitted in the district and tertiary hospital, respectively ( Table 2). The single-locus variants ST432, ST607 and ST17 were isolated from tertiary (n = 1) and district (n = 2) hospital, respectively.
The cgMLST K. pneumoniae scheme was defined with NCBI data using K. pneumoniae K069 as the reference genome. The close relatedness between a batch of carriage (A105R1B5) and clinical (ED01500733) ST 983 strains isolated from the district and tertiary hospital, respectively was evident, with 100% identity and an allelic distance of zero (Fig. 1). Similarly, high genetic similarity was observed between carriage (G702R3B2, G702R1B5, G702R2B5) and clinical K. pneumoniae (ED01503757) ST152 strains originating from the tertiary hospital with 99% identity and an allelic distance of zero (Fig. 1).

Discussion
The increasing prevalence of ESBL-KP remains a major clinical concern worldwide. To understand the molecular epidemiology of ESBL-KP, we studied using WGS, antibiotic resistance genes, MGEs and genetic lineages associated with circulating ESBL-KP isolated from carriage and clinical samples of hospitalized patients in uMgungundlovu district, South Africa. www.nature.com/scientificreports www.nature.com/scientificreports/ The increasing prevalence of ESBL-KP has been associated with high mortality in developing country 14 . However, a 5% prevalence of ESBL-KP was detected in faecal carriage and clinical samples in our study. This is consistent with the report of Jallad et al. 15 , which shown 9.7% of ESBL-KP from faecal carriage among healthy patients in nursing homes in Lebanon 15 . In contrast, this finding is lower than that described by Perovic et al. 16 , where a 68.9% prevalence of ESBL-KP was detected in bloodstream infections in the public healthcare sector in Free State, Gauteng, Limpopo, KwaZulu-Natal, and Western Capes provinces in South Africa 16 . Similarly, Rashid et al. 5 reported a 32.43% prevalence of ESBL-KP from faecal carriage in healthy patients hospitalized in tertiary hospital in India 5 . The discrepancies observed in the ESBL-KP prevalence could be attributed to variation of the geographic location, level of exposure to healthcare settings, hospital levels, antibiotic stewardship programs and antibiotic use.
The molecular characterization of diverse resistance determinants associated with the circulating ESBL-KP strains was undertaken following the health referral system. High level of resistance was detected in the tertiary hospital with one isolate, G702R2B5 (ST152) harboring 48 resistance genes in contrast to the district hospital where a maximum of 36 resistance genes were identified in the isolate A111R1B2 (ST17). The presence of genes encoding resistance to β-lactams, aminoglycosides, fluoroquinolones, fosfomycin, rifampicin, sulphonamide were reported in both clinical and carriage samples in the tertiary hospital. This is consonant to that reported in the literature where bla CTX-M-15 , bla SHV-28 , and bla TEM-1B and fosA3 were the common genes implicated in the resistance of cephalosporins, monobactams and fosfomycin identified in carriage and clinical K. pneumoniae isolate in   18 . This suggest that ESBL-KP either in clinical or carriage sample, could be a probable reservoir of resistance genes for other bacterial species and be responsible for genetic transfer to other species. The dissemination of ESBL-KP in these healthcare settings could probably be attributed to a lack of effective infection, prevention and control (IPC) measures for their containment.
An interesting finding of this study was the detection of the clonal lineage ST152 (n = 4; 44.5%) circulating in both carriage (at admission, n = 3) and clinical sample (n = 1) of the tertiary hospital. These isolates were characterized by their multidrug resistance which was confirmed by the concomitant presence of several β-lactam (bla CTX-M-15 , bla SHV-11 , bla TEM-1B and bla OXA-1 ) resistance genes. This is consistent with the literature which confirmed that the bla SHV gene is a normal chromosomal gene in K. pneumoniae and that CTX-M-15 is the most  , trimethoprim (dfrA27) and sulphonamide (sul1, sul2) were also identified in these isolates. It is acknowledged that aac(6′)-Ib-cr is a variant of the aac(6′)-Ib gene which acetylates fluoroquinolones and has a low-level resistance to aminoglycosides. Mutation in gyrA (Ser83F) and parC (Ser80L) have further been detected in these four ST 152 isolates. Our MICs corroborate these findings since all these isolates exhibited high level resistance to fluoroquinolones, except for K. pneumoniae ED01503757 where moderate fluoroquinolone resistance was observed. ESBL-KP harboring similar resistance genes have been reported in Italia 20 and Lebanon hospital 8 . Tokajian et al. 17 , showed that CTX-M-15 was associated with MDR-K. pneumoniae and revealed that qnrB6 was frequently observed in African countries 17 . Several studies showed that ESBL-producing K. pneumoniae ST152 is associated with resistance to carbapenems [21][22][23] .
Taken all together, the fact that ESBL-KP ST152 strains isolated in the tertiary hospital harbored the same resistance genes and mobile genetic elements including plasmids [ColRNAI, IncFIB(K), IncF(II)] and integrons (IntIPac) suggests that this clone could be associated with intra-and/or inter-hospital dissemination. This clonal spread was corroborated in our cgMLST analysis, where in clade II a high genetic relationship (99.90% identity and allelic distance of zero), was observed in our collection of carriage (G702R3B2, G702R1B5, G702R2B5) and clinical K. pneumoniae (ED01503757) ST152 strains, all originating from the tertiary hospital. In addition, a close relationship was observed with the K. pneumoniae ST152 strain K069 (NXKY0000000) detected in a clinical sample in Pretoria, South Africa. Of further interest is that one of the K. pneumoniae ST152 (ED01503757) isolates was carbapenem susceptible whereas the other ST152 isolates showed reduced susceptibility to carbapenems. Although the contribution of efflux mechanisms was neither determined by a MIC reduction assay nor gene expression assay, we hypothesized that the overexpression or repression of the numerous multidrug resistance efflux pumps detected in these isolates could be associated with the various MICs of imipenem and meropenem as described 24 .
Another interesting finding of this study, was the detection of two  A-:B-)]. They were isolated in two patients hospitalized in the district (A105R1B5) and tertiary (ED01500733) hospital, suggesting the probable clonal spread of the ESBL-KP ST983 inter-hospital in uMgungundlovu district as a result of the health referral system. The cgMLST analysis confirms that the two clinical (ED01500733) and carriage (A105R1B5) K. pneumoniae ST 983 in clade I were closely genetically related with 100% identity, an allelic distance of zero and an allele difference of one. Meanwhile, they shared a common ancestor with the K. pneumoniae ST 17 (A111R1B2) detected in the carriage sample of the district hospital. The two ST983 isolates ED01500733 and A105R1B5 were as such closely related to this ST17 isolate with an allelic distance of zero, a 99.41% percent identity and allele differences of 16640 and 16639 between A111R1B2 and A105R1B5, and between A111R1B2 and ED01500733, respectively. Our findings thus suggest that these ST 152 and ST 983 lineages could spread between patients in the same or different  . CRISP arrays detected in the K. pneumoniae G702R2B5 (3A) and A105R2B2 (3B). Two different Characterization of CRISPR arrays detected including (CRISPR 2) and (CRISPR 1 and 2) in K. pneumoniae G702R2B5 strain. The first CRISPR2 array composed of six direct repeated sequences and nine spacer sequences was located at nucleotides 11242 to 11731. CRISPR1 array composed twelve direct repeated sequences and eleven spacer sequences was located at nucleotides 194435 to 195045, CRISPR2 composed twenty-two direct repeated sequences and twenty-one spacers was located at nucleotides 203887 to 205176. www.nature.com/scientificreports www.nature.com/scientificreports/ Whilst Ambler classes A, B, C and D carbapenemase-producing K. pneumoniae strains gained worldwide attention due to the high resistance conferred to carbapenems, K. pneumoniae has evolved to become resistant to almost all β-lactams without harbouring carbapenemase genes 13 . This phenomenon has been possible with the concomitant use of multiple resistance mechanisms such as acquisition of an Ambler class A or C β-lactamases, with the loss of the OmpK35 and OmpK36 porins and/or overexpression of MDR efflux pumps 13 . In fact, some studies 25,26 have established the impact of MDR-efflux pumps and porin losses on the membrane permeability of K. pneumoniae. The carbapenem resistance detected phenotypically (meropenem 16 mg/L; MIC imipenem 8-64 mg/L) in some isolates was not corroborated genotypically with the detection of specific carbapenemase encoding genes by WGS. This could hence be explained by the fact that resistance was not mediated by specific carbapenemase genes but rather by porin loss and/or MDR efflux pumps present in all isolates. Efflux systems have been reported in several clinically important bacteria and the overexpression of MDR efflux pumps can lead to high-level multi-drug resistance 25,26 . All the isolates harbored several MDR efflux pumps including CmeA, CmeB, MATE, MFS, MacA, MarcB, AcrB, MarA, OML, RND and AcrAB. We postulate that MDR efflux pumps were implicated in multi-drug resistance observed in the majority of isolates. K. pneumoniae contains three well-known porins including the two major porins OmpK35 and OmpK36 that are homologous to the OmpF and OmpC of Escherichia coli respectively, as well as the small porin OmpK37. Given that OmpK35 and OmpK36 porins play a critical role in the cell penetration of antibiotics, their loss can lead to reduce susceptibility or resistance to cephalosporins and carbapenems, especially in strains harboring Ambler class A, B, C or D β-lactamase 12,13 . The detection of OmpK37 porin that allows penetration by carbapenems but not other β-lactams, may explain why all isolates expressed high level resistance to cefoxitin (except for clinical isolates), cefotaxime and ceftazidime, and null or moderate resistance to carbapenems. We thus hypothesized that the deficiency in OmpK35 and OmpK36 coupled with the presence of bla CTX-M-15 , and MDR efflux pumps, could play a significant role in conferring K. pneumoniae resistance to carbapenems and third generation cephalosporins in our study.
ABR is generally mediated by intrinsic or acquired resistance genes located on chromosome or MGEs, respectively. The CRISPR-Cas system was demonstrated to cleave plasmid DNA, thereby protecting bacteria from transduction (phage infection) and other horizontal gene transfers. They were supposed to be a defense mechanism against infection by diverse extra-chromosomal agents 26 . The correlation between the presence of CRISPR-Cas system and antibiotic resistance has already been studied and an inverse correlation between its presence and acquisition of antibiotic resistance was described in 48 Enterococcus faecalis strains 26 . In K. pneumoniae, CRISPR-Cas system has been detected in a very few strains worldwide. Apart from our isolates A105R2B2 and G702R2B5, only two complete K. pneumoniae genomes (NC_018522, NC_012731) and five draft genomes sequences (NC_012731, NZ_ANGH02000012, NZ_APGM01000001, NZ_JH930419, NZ_JH930428) harbor it. Even though CRISPR-Cas system serves to protect bacteria against phage infections and horizontal gene transfer, their presence among ESBL-KP ST607 (A105R2B2) and ST152 (G702R2B5) that were the most resistant isolates, suggest their probable implication in the acquisition of resistance genes. This is consistent with a report which demonstrated that in Klebsiella genomes, CRISPR-Cas systems are located among genes encoding for proteins that are likely involved in metabolism as well as resistance to antibiotics 22 . Additionally, the detection of several phages in these two highly resistant strains along with CRISPR-associated cas3 genes shed light on areas for further investigation on the emergence of ABR and transmission of antibiotic resistance genes.
In summary, our findings reveal the dissemination of ESBL-producing K. pneumoniae within and between wards and hospitals in uMgungundlovu district, South Africa. It shows that hospital is a reservoir for several resistance determinants and highlights the necessity to efficiently and routinely screen patients, particularly those receiving extensive antibiotic treatment and long-term hospitalization stay. It also reinforces the need for infection, prevention and control measures to reduce the dissemination of ABR in this district. Methods ethical approval. Ethical approval was obtained from the Biomedical Research Ethics committee (BREC) (No. BF512/16, sub-study of BCA444/16) of the University of KwaZulu-Natal, South Africa. Permission to conduct the research was also granted from the Department of Health, uMgungundlovu District and hospital managers. All methods were performed in accordance with the relevant guidelines and regulations. study design and bacterial isolates. This study took place in two hospitals at different level of care (district and tertiary), from May to July 2017 in uMgungundlovu district, South Africa. The district and tertiary hospitals were approximately 70 Km apart. Oral and written informed consent were obtained from all study participants after explanation of the procedure and purpose of the study. Rectal swabs were collected aseptically with Amies swabs from all admitted in-patients >18 years old to form the carriage sample. Isolates routinely processed in the microbiological laboratory during the sampling period formed the clinical sample. Every patient included in this study was screened for the presence of Gram-negative ESKAPE bacteria. All samples were cultured onto MacConkey agar with and without cefotaxime (2 mg/L). After incubation for 18-24 h at 37 °C, each morphotype growing on MacConkey with cefotaxime (MCA + CTX) was subjected to Gram staining, catalase and oxidase tests, followed by biochemical identification with API 20E (bioMérieux, Marcy l'Etoile, France) and the Vitek ® 2 System (bioMérieux, Marcy l'Etoile, France).
The strains sequenced in this study were isolated from carriage (A105R2B2, A105R1B2, A111R1B2, G702R1B5, G702R2B5, G702R3B2) and clinical samples including sputum (ED01503757, ED01502268) and urine (ED01500733) of six patients hospitalised in the district or tertiary hospital. The isolates A105R2B2, A105R1B2, A111R1B2 were detected in rectal swabs of two patients (A105 and A111) admitted in the medical ward of the district hospital while the isolates G702R1B5, G702R2B5, G702R3B2 were recovered from rectal swabs of a single patient (G702) but at different time-point (admission, after 48 h and at discharge) in the tertiary hospital. The clinical isolates ED01503757, ED01502268, ED01500733 were identified from three patients in the tertiary hospital. These www.nature.com/scientificreports www.nature.com/scientificreports/ isolates were closely related on enterobacterial-repetitive-polymerase chain reaction (ERIC-PCR) 27 analysis 28 . Given that we aimed to evidence clonal spread of ESBL-KP, within each ERIC-cluster, representative isolates of related strains originating from different level of care were considered for WGS. These isolates were more representative because they belonged to the same ERIC-PCR cluster and antibiotic resistant patterns. phenotypic screening of esBL-production and antimicrobial susceptibility testing. All isolates were phenotypically screened for ESBL, AmpC, KPC and MBL production using combination disk test sets (ROSCO DIAGNOSTICA, Taastrup, Denmark). Minimum inhibitory concentrations (MICs) were determined broth microdilution for all selected isolates. Ampicillin, cefoxitin, cefuroxime, cefotaxime, ceftriaxone, imipenem, meropenem, amikacin, gentamicin, trimethoprim, ciprofloxacin, moxifloxacin, nitrofurantoin, tetracycline, were tested and interpreted according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) breakpoints 29 . E. coli ATCC 25922, K. pneumoniae ATCC 700603 and K. pneumoniae ATCC 51503 were used as controls.
Identification of the resistome, virulome and mobile genetic elements. The GoSeqIt tool was used to annotate and determine known antimicrobial resistance genes, virulence factors and plasmids using ResFinder 32 , VirulenceFinder 33 and PlasmidFinder 34 , respectively. The RAST SEED viewer aided the identification of integrons and transposases flanking the β-lactamase genes 35 . The identification, annotation and visualization of prophage associated regions were performed using PHAge Search Tool (PHAST) server 36 . Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) and insertion sequence elements were investigated with the CRISPRFinder server (http://crispr.i2bc.paris-saclay.fr/Server/) and ISFinder (https://www-is.biotoul.fr/) 37 , respectively. Outer membrane porin genes were analyzed with the Sequence Search Antibiotic Resistance Tool (SSTAR, version 1.1.01, https://github.com/tomdeman-bio/ Sequence-Search-Tool-for-Antimicrobial-Resistance-SSTAR-) 38 software that used a standalone Basic Local Alignment Search Tool (BLAST) and a database combining ARG-ANNOT and ResFinder to identify known antibiotic resistance genes, detect putative new variants, modification and/or truncated genes. In addition, the Comprehensive Antibiotic Resistance Database (CARD; https://card.mcmaster.ca) was used to corroborate the results. Finally, the contigs of the K. pneumoniae G702R2B5 were mapped against the complete genome of K. pneumoniae U25 (CP012043) for visualization of the genomic organization 39 .
Multilocus sequence typing (MLst) and core genome multi-locus sequence type analysis (cgMLst). The scheme of Diancourt et al. 38 , which considers the allelic variation amongst seven housekeeping genes (gapa, infb, mdh, pgi, phoe, rpob and tonb) to assign STs was used for in silico multi-locus sequence type (MLST)-analyses and WGS data were used for the MLST assignment of K. pneumoniae isolates 40 .
A genome-wide gene-by-gene comparison approach was used to assess the clonal relatedness between isolates within and across wards and hospitals. The core genes were determined from the annotated genome assemblies, predicted coding regions were extracted and converted into protein sequences. A phylogeny was drawn for K. pneumoniae using Rapid large-scale prokaryote pangenome analysis (Roary; https://sanger-pathogens.github.io/ Roary/) to estimate the tree for the core genome. The genome of K. pneumoniae strain K069 (Accession number NXKY01000005.1) served as reference genome and the following 12 query international K. pneumoniae genomes (Accession numbers JUBG00000000, JTKD00000000, JUBL00000000, JUBM00000000, AZAP00000000, CP012743, CP012744, CP012043, NXLE01000020.1, NXKY01000005.1, NXKX01000020.1, CP022922.1, CP033901.1) obtained from NCBI database were used to assess the cgMLST target genes. Altogether, 2944 core genes were extracted with an alignment length of 2,852,207 bp shared by the nine K. pneumoniae genomes. The allelic distance from the cgMLST was visualized using Figtree v1.4.3 (http://tree.bio.ed.ac.uk/software/figtree/) in a maximum likelihood phylogenetic tree including isolate name, ST type and country.