Introduction

The emergence and spread of carbapenem-resistant Enterobacterales (CRE) is considered to be one of the most serious antibiotic resistance threats [1]. The majority of CRE is Klebsiella pneumoniae, which are carbapenem resistant due to the production of carbapenemases (CRKP) [2]. This worldwide trend is also observed in countries of the Arabian Peninsula [3,4,5]. Spread of CRKP is mainly driven by clonal dissemination of so called high-risk multi-drug resistant (HiR MDR) clones [2]. K. pneumoniae ST14 is one of such clones producing various carbapenemases that have been reported from four continents [2, 5, 6] including the Arabian Peninsula [3, 7,8,9]. Earlier, this clone was shown to have high prevalence among CRE isolated in the United Arab Emirates [8, 10]. To understand the reasons of its successful emergence and spread, we used whole genome sequencing (WGS) to assess the resistome, the capsular locus, the carbapenemase carrying plasmid types, as well as the level of core genome similarity of selected K. pneumoniae ST14 isolated in countries of the Arabian Peninsula, which were also compared to whole genomes of K. pneumoniae ST14 isolated at other geographical locations.

Materials and methods

Identification of CRKP-ST14 isolates

Altogether 761 independent CRKP strains isolated between June 2011 and June 2016 in 33 hospitals of five countries of the Arabian Peninsula from a collection described earlier [11] were screened by PFGE [12], to select isolates exhibiting ≥ 80% similar patterns to 23 known CRKP-ST14 identified earlier in the UAE [7,8,9]. These were further divided into sub-clusters based on 90% PFGE pattern similarity threshold (Supplementary Fig. 1). At least one isolate from these PFGE sub-clusters carrying unique carbapenemase genes or their combinations were subjected to multi-locus sequence typing [13] to confirm their sequence type.

Testing of antibiotic susceptibility, carbapenemase production, and detection of carbapenemase genes

Phenotypic confirmation of the isolates’ carbapenemase production and methods testing their antibiotic susceptibility was described earlier in [11]. Presence of the common carbapenemase genes (blaNDM, blaOXA-48-like, blaKPC, blaVIM, blaIMP) were screened by PCR and identified by direct sequencing of the amplicons [9].

Characterization of carbapenemase gene-carrying plasmids

Electrophoresis gels of plasmids extracted by the alkaline lysis methods from the wild-type CRKP strains were capillary transferred to Hybond N + membranes (Roche Diagnostics, Mannheim, Germany) and the membranes were probed with blaNDM-1 or blaOXA-48 PCR amplicons labelled and detected with the DIG DNA-labelling and detection kit (Roche Diagnostics, Mannheim, Germany) [14]. Selected carbapenemase bearing plasmids were conjugally transferred into a sodium-azide resistant derivative of rifampicin resistant Escherichia coli J53 (J53RAZ) [7]. If conjugation failed, 1 μg of plasmid DNA purified with Plasmid Maxi Prep Kit (Qiagen, Hilden, Germany) of a wild type isolate was heat-shock transformed into competent cells of Escherichia coli DH5α [7]. Transfer experiments were considered successful if a single-plasmid containing derivative was obtained, confirmed by plasmid electrophoresis and PCR targeting the respective carbapenemase gene [7]. Plasmid incompatibility typing was performed by PCR as described earlier in [3, 7, 15].

Analysis of the whole genomes

Genomic DNA of selected CRKP-ST14 isolates was extracted using the QIAamp DNA Mini Kit (QIAGEN, Crawley, UK), according to the manufacturer’s instructions. Bacterial DNA libraries were prepared using the Illumina Nextera®XT DNA Library Preparation Kit (Illumina, USA) and sequenced on Illumina MiSeq platform using 2 × 250 bp or 2 × 300 bp paired-end sequencing technology. Reads were quality checked and assembled using the Velvet 1.1.04. algorithm built in the Ridom SeqSphere+ software. Core genome MLST (cgMLST) comparison was performed using Ridom SeqSphere+ (version 7.2.3). The assembled contigs were analysed by Kleborate for the virulence gene and capsule locus related information [6]. The acquired resistance gene content was assessed using ResFinder 4.0 (https://cge.cbs.dtu.dk//services/ResFinder/) [16]. In case the Velvet assembly lacked an expected carbapenemase gene, the respective reads were assembled using SPAdes 3.11 and the contigs produced were re-analyzed in the ResFinder 4.0. The integrity of lipopolysaccharide synthesis regulating genes mgrB, phoP, phoQ, pmrA, and pmrB [17], and that of the main porin genes (ompC and ompF) was compared to the respective genes of K. pneumoniae MGH78578 (GenBank Acc. No. CP000647). The sequence data generated were uploaded to the European Nucleotide Archives under project number PRJEB44852.

Comparison of local CRKP-ST14 isolates to ones isolated globally

In order to compare K. pneumoniae ST14 isolates of the Arabian Peninsula genomes of three carbapenemase gene-carrying K. pneumoniae ST14 isolates: PittNDM01, KP617 and NUHL24835 (GenBank accession numbers: CP006798–CP006802, CP012753–CP012755, CP014004–CP014006) and further 170 carbapenemase carrying K. pneumoniae ST14 genomes from the PathogenWatch (https://pathogen.watch/genomes/all?genusId=570&mlst=14&speciesId=573 accessed on 02–06-2021) were downloaded. All were compared by cgMLST using Ridom SeqSphere+ software. A total of 173 genomes available from the databases were also screened for their respective capsular locus types. One-hundred and nine global isolates from the databases were isolated in the same period as the recovery of our isolates from the Arabian Peninsula. Further 50 isolates used in the comparison were collected between 2017 and 2019, and 14 did not have the collection date specified.

Results

Identification of Klebsiella pneumoniae ST14 isolates

Based on PFGE, a total of 172 K. pneumoniae strains exhibited patterns ≥ 80% similar to known CRKP-ST14 isolates. The majority of them (n = 162) were from hospitals in the UAE, but nine isolates from Riyadh, Saudi Arabia, and one from Riffa, Bahrain, also exhibited similar PFGE patterns. These 172 isolates could further be divided into 14 PFGE sub-clusters applying the 90% similarity threshold (Supplementary Fig. 1). All but one carried a carbapenemase gene(s), with a remarkable high rate of double carbapenemase producers (42%). The types and distribution of carbapenemase alleles is shown in Supplementary Table 1.

Based on the PFGE subclusters and on the type of carbapenemase genes carried, 70 isolates were subjected to MLST that confirmed sequence type 14 in each case (Supplementary Fig. 1).

Analysis of the carbapenemase genes and their localizations

The 70 CRKP-ST14 isolates were further investigated by testing the plasmid localization of their respective carbapenemase genes. Based on the alleles of blaNDM and blaOXA carbapenemase genes and on their association with different plasmids, strains were divided into 13 subgroups (SGs) (Table 1). In one isolate (ABC225) no carbapenemase genes or the production of such enzymes could be detected. Two strains (ABC220 and ABC224, already described in [9]) harboured blaKPC-2 on IncX3 plasmids, while all the others carried blaNDM-1 alone or in combination with various alleles of blaOXA-48-like. The blaOXA-232 gene detected by PCR could not be localized by hybridization, neither was the carbapenemase gene transferable by conjugation nor by transformation from one isolate (ABC334). On the other hand, although the plasmid electrophoresis gel of isolate ABC643 did not hybridize with the NDM-probe, from this strain the blaNDM-1 was successfully transferred by conjugation into E. coli J53RAZ and the carbapenemase gene was identified on an approx. 240 kb IncHI1B plasmid from the transconjugant. Consequently, ABC643 was considered being member of subgroup D (Table 1). From each of the 13 subgroups identified, up to five isolates, overall 39 isolates were selected for whole genome sequencing (Table 2).

Table 1 Subgroups of CRKP-ST14 with various alleles of carbapenemases located on different plasmids
Full size table
Table 2 Characteristics, cgMLST clustering, and resistome of strains subjected to whole genome sequencing
Full size table

Clonal relatedness and resistome of the carbapenem-resistant K. pneumoniae ST14

The assembly statistics of the 39 isolates subjected to WGS is summarized in Supplementary Table 2. The cgST types are listed in Table 2. Core genome MLST revealed seven singletons and three clusters that differed by up to 14 out of 2358 alleles (Table 2 and Fig. 1).

Fig. 1
figure 1

Minimum spanning tree based on cgMLST of the 39 K. pneumoniae ST14. Distance is based on number of differences of the 2358 alleles in K. pneumoniae sensu lato cgMLST scheme with cluster alert distance set at 15

Full size image

As it is shown in Table 2, Cluster 1 and Cluster 2 contained isolates from various hospitals in different cities carrying different carbapenemases, whereas Cluster 3 consisted of isolates exclusively from one hospital in Saudi Arabia.

Beyond the carbapenemase and other beta-lactamase genes, all isolates carried acquired resistance genes coding for aminoglycoside, trimethoprim and sulfamethoxazole resistance, and they variably possessed macrolide, tetracycline, rifampicin and plasmid mediated quinolone resistance genes. Importantly, the majority of isolates in the three clusters carried the armA 16S methylase gene. The array of acquired antibiotic resistance genes matched the antimicrobial susceptibility profiles of the isolates (Supplementary Table 3), e.g., high-level tetracycline resistance was observed with isolates possessing the tet(D) gene. Similarly, the few isolates susceptible to chloramphenicol were the ones lacking any phenicol resistance genes. However, the lack of acquired phenicol resistance genes did not translate always to susceptibility to chloramphenicol, suggesting alternative resistance mechanisms, possibly efflux. The colistin resistance observed was not due to the presence of mobile colistin resistance genes (mcr1-9) and did not correlate with certain clusters. Colistin resistance in 7 of the 13 colistin resistant isolates was due to insertional, frameshift or non-synonymous point mutations of the mgrB gene. In further two isolates point mutations in phoQ, and in another three point mutations in pmrB were responsible for this phenotype. In one isolate the regulatory genes examined were all wild-type (Table 2), but further investigation of the colistin resistance mechanism in this single isolate was beyond the scope of our current study.

We assume that the only carbapenemase non-producing strain (ABC225) was resistant to carbapenems due to the production of multiple beta-lactamases, including ones with extended-spectrum (blaTEM-1A, blaOXA-1, blaOXA-9, blaCTX-M-15) combined with deletions in the ompF gene: a 19 base-long one starting at nucleotide 676 and a single nucleotide one at position 872, while the ompC gene was wild-type (data not shown).

All isolates, but ABC681 possessed the yersiniabactin siderophore genes encoded on ICEKp5 mobile element, but none of them had genes characteristic to hypervirulent K. pneumoniae, i.e., rmpA, rmpA2, and iutA. Interestingly, isolates in Cluster 2 and further two singletons (RH60 and ABC334) possessed capsular locus KL64, not the KL2, i.e., one typically associated with K. pneumoniae ST14 and also present in the remaining isolates (Fig. 1). Although isolates carrying the KL64 locus clustered by their core genome, the carbapenemase genes carried by them were variable: some of them possessed blaNDM-1, others blaOXA-232, or both of these carbapenemases (Table 2 and Fig. 2).

Fig. 2
figure 2

Neighbour-joining tree based on the cgMLST of 212 K. pneumoniae ST14. Color code is based on the capsular locus type. The strains marked with red dotted line are isolated in 2017–2019, i.e., after the strains investigated in the present study

Full size image

Comparison of K. pneumoniae ST14 isolated in the Arabian Peninsula to international K. pneumoniae ST14 genomes

The neighbour-joining tree generated by comparing the cgMLST of local as well as global, altogether 212 K. pneumoniae ST14 isolates is presented in Fig. 2. K. pneumoniae ST14 genomes with KL2 capsular locus formed several clusters, whereas isolates of the ST14 clone having other capsular locus types (KL16, KL64, KL107, and KL133) clearly separated from the previous groups. The core genomes of K. pneumoniae ST14 harbouring KL64 from the UAE and Bahrain formed a distinct cluster with NDM-1 or NDM-5 and/or OXA-232 producing K. pneumoniae ST14 isolates from the Netherlands, Australia, India and the USA and with one isolate from India with capsular locus KL107, which also contained the wzi64 allele.

Discussion

Numerous high-risk K. pneumoniae clones, among them ST14, have now been identified [2]. ST14 strains have been reported to produce a wide range of carbapenemases: apart from being the sequence type from which the blaNDM-1 gene was first identified [18], various blaKPC [9, 19, 20], blaNDM [21, 22], and blaOXA-48-like [6, 23,24,25,26] alleles or double carbapenemase producers have been reported from different parts of the world, including the Middle East [7, 8, 27,28,29,30,31]. Furthermore, ST14 strains appear to be one of the most wide-spread, highly resistant CRE clones in Dubai [8].

Even in the light of this variety of CRKP-ST14 strains, the diversity of carbapenem-resistant mechanisms encountered in the relatively small pool of ST14 strains isolated in countries of the Arabian Peninsula was remarkable (Table 1). Beyond strains carrying blaNDM-1, blaOXA-232, blaKPC-2, and combinations of blaNDM-1 either with blaOXA-232, or with blaOXA-48, we identified isolates with blaOXA-162. To the best of our knowledge, this is the first observation of this allele alone or in combination with blaNDM-1 in this high-risk clone of K. pneumoniae. This variety of carbapenemase gene content further confirms the earlier observations describing this high-risk clone as one which is particularly fit to acquire resistance plasmids and spread the resistance traits [2, 6]. Furthermore, we also encountered a strain, which did not produce a dedicated carbapenemase, yet was resistant to carbapenems due to ESBL production and ompF mutation, a phenomenon otherwise not uncommon among CRKP [10, 11].

The diversity of ST14 strains was further accentuated by the various types of, mostly conjugative, plasmids carrying the carbapenamase genes. Even the same allele of a carbapenemase gene (blaNDM-1) could be localised on two different plasmids, IncHI1b and IncFII, in our isolates. On the other hand, all blaOXA-48 and blaOXA-162 were located on IncL type plasmids, while blaOXA-232 was associated with IncColKP3 type plasmids and in a single isolate (ABC240) with an IncColKP3-IncR composite plasmid (Table 1).

The cgMLST results did not completely overlap with these carbapenemase gene-, and plasmid-based groups (Table 1 and 2). Considering only the carbapenemase genes, members of cgMLST cluster 3 were the most homogenous, as all such strains produced NDM and OXA-48 with the exception of a single isolate, presumably having lost the plasmid coding for the latter carbapenemase. This lack of heterogeneity was not surprising as all these strains were isolated in the same hospital in Saudi Arabia. Interestingly, they clustered with Saudi isolates of 2018, and German isolates of 2016–2019 deposited in the PathogenWatch database (Fig. 2). On the other hand, members of the other two cgMLST clusters were more heterogeneous: despite being highly similar at the genome level: they carried a variety of plasmids coding for a variety of carbapenemase alleles coded on different plasmids. In some cases, such differences were even seen among isolates from the same hospital (e.g., see Cluster C1 isolates from Hospital G in Abu Dhabi, Table 2). This considerable heterogeneity within the same ST, even within the same cgMLST cluster, suggests that horizontal transfer of carbapenemase gene-coding plasmids considerably contributes to the epidemiology of ST14 even in a relatively small region. Nevertheless, the fact that same cgMLST group strains possessing the same carbapenemase bearing plasmid could be isolated from multiple hospitals (e.g., Cluster C1 strains of plasmid group A from hospitals A in Sharjah, C in Al Ain, and B in Abu Dhabi) indicates the importance of clonal, inter-hospital spread, as well.

The variety and wide array of genes coding for resistance to numerous antibiotic classes within clusters were also notable findings supporting the previous notion. It also explains how members of the clone develop to being extremely drug resistant, or even pan-drug resistant, which was already observed in the region [8, 30].

Recently, and concerningly the convergent evolution of antimicrobial resistance and virulence was observed in K. pneumoniae, mostly, but not exclusively, in South-East Asia [6, 32, 33]. Some of these hypervirulent isolates expressed non-K1 capsular types, i.e., K2, K54, or K64 [33, 34]. Although virulence genes typically associated with hypervirulent K. pneumoniae (rmpA/A2, iutA) were not identified in our strains (data not shown), the two capsular types (KL2 and KL64) detected in carbapenem-resistant K. pneumoniae ST14 have commonly been linked to hypervirulence and invasive diseases [6, 33, 35]. Further studies are needed to investigate whether this capsular switch in CRKP-ST14 from the well-known K2 type to K64 increases or changes virulence, or simply helps the pathogen to avoid immune recognition, and helps the dissemination of this clone, as was hypothesized in case of KPC-producing K. pneumoniae CC258 with capsular variations [35, 36]. Our data does not allow us to speculate on the molecular mechanisms involved in the emergence of K64 strains. Nevertheless, as the K2 and K64 loci are considerably different [37], the role of recombination rather than local, mutation-based evolution seems to be more likely. However, irrespective of the mechanism behind, the clustering of cgMLST of ST14-K64 isolates collected in 2013–2016 from the UAE and Bahrain and those of other three continents, i.e., Australia from 2016, USA and India from 2017, and the Netherlands (Fig. 2) suggests a potentially emerging subgroup of the K. pneumoniae ST14 clone. The dissemination and virulence potential of this subgroup needs further monitoring.