Patients’ characteristics

Our study enrolled all culture-positive plates of Gram-negative bacteria recovered from extraintestinal infections during a 16-month period. A total of 123 Klebsiella spp. were identified through biochemical testing among these consecutive samples. Of those, 51 patients (53.7% female, 46.3% male, age range was from 10 to 93 years) with a clinical history consistent with carbapenem non-susceptible (including either imipenem, meropenem or ertapenem disk diffusion test results) K. pneumoniae infections were subjected for further analyses. In addition, three patients infected with carbapenem-susceptible hypervirulent K. pneumoniae were included for comparative process. The majority of the isolates were from urine cultures (29, 53.7%), respiratory (20, 37%), wound (2, 3.7%), and blood (2, 3.7%). Patients were admitted or stayed at four distinct wards; Intensive Care Unit (ICU) (66.7%), Internal (22.2%), Cardiac Care Units (CCU) (3.7%) and Emergency/observation service (7.5%).

Amikacin was the most potent antibiotic against carbapenem non-susceptible K. pneumoniae

Resistance rates against imipenem, meropenem, ertapenem, cefepime, and trimethoprim/sulfamethoxazole antibiotics were 75.9%, 70.4%, 83.3%, 72.2%, and 72.2%, respectively. Isolates were highly resistant against ampicillin/sulbactam and amoxicillin/clavulanic acid (88.9% resistant rates for both), and piperacillin/tazobactam and aztreonam (83.3% resistance rates for both). Amikacin was figured out as the most potent antibiotic with a 22.2% resistance rate, followed by tobramycin (63%) gentamicin (64.8%), and levofloxacin (66.7%). The MDR phenotype was found in 87% of the isolates tested. The MICs of carbapenems against blaOXA-48 and blaNDM-1 producers are shown in Table 1.

Table 1 Demographic characteristics of K. pneumoniae subjected to WGSCC147 was detected as the dominant clone

The study isolates were categorized into 15 STs and CC147 (including ST147 [9 strains] and its single locus variant (SLV) ST392 [16 strains]), was identified as the dominant clone, accounting for 25 (46.2%) episodes of infection. The other prevalent clones were ST45 (9.2%), and two high-risk clones, ST14 and ST48 (7.4%, for each).

Average nucleotide identity (ANI), with ANI values obtained for each species exceeding 97%, supported clustering by drawing the phylogenetic tree. Based on phylogenetic analysis of 88,634 genome-wide SNPs, strains were divided into K. pneumoniae KpI (51, 94.4%) and KpII subgroups of K. quasipneumoniae subsp. quasipneumoniae KpII-A (3, 5.6%) that belonged to ST1308 (Fig. 1). The distribution of capsular locus (KL) types across the phylogenetic tree revealed their highly clonal nature, with each KL type observed only in one ST, except for KL-2 (O locus; O1v1), which was shared by ST48 and two hvKp ST86 strains. We identified 14 different K loci and 5 O loci, and these K/O loci provided 14 different combinations in our collections. The most frequently identified KL types were KL-27 and KL-64 which were linked to ST392 and ST147, respectively (Table 1). The O1/O2v1 loci were the most prevalent O antigens, detected in 22 (40.7%) isolates of different STs including ST14, ST48, ST86, ST147, ST348, ST2159 and ST6510. The O4 serotype was the second most common O locus and was identified in 20 strains of ST392, ST815, and ST1308. We observed the same O-Locus combined with a distinct K-Locus, i.e., O1v1 was associated with KL-2/KL-16/KL-62, O2v1 in KL-64/KL22, O101 was associated with KL-155/KL-52 /KL-47, and O4 was linked to KL-144/KL-27/KL-107 (Table 1).

Fig. 1

Phylogenetic tree of study 54 K. pneumoniae isolates. Maximum-likelihood phylogenic tree clustered different clones (STs) and CC147 strains (ST147, ST392) were identified as the dominant clone with the highest prevalence of varied resistance elements. The tree was constructed from the whole genome SNPs arising by mutation and labeling and visualization were done by using the iTOL web-interface. The data presented are the date of isolation for CC147, sequence types, capsular types, Integrative and conjugative Elements (ICEs), acquired resistance markers and amikacin/gentamicin susceptibility phenotypes. The hvKp strains have been shown in red

The majority of patients from whom CC147 strains recovered were admitted to an intensive care unit (internal ICU, surgical ICU), where strains were isolated during their stay. Besides the ICU ward, one of the CC147 infection episodes was recovered from the urine culture of a patient admitted to the CCU (strain no. 447, ST392). Temporally, the first NDM positive strain (strain no. 2, ST392) was isolated shortly after starting of sample collection (March 2018), and all except one ST392 strain (strain no. 659) were recovered during 2018. In contrast, ST147 appeared in June 2018 and replaced ST392 during the entire sampling period (Fig. 1). In terms of carbapenem MICs, the ST147 clone did better than the ST392/NDM + strains against this family of antibiotics (Table 1).

To identify genomic clusters that are likely to be epidemiologically linked, 23 SNP pairs were used as a pairwise distance cutoff. The ST147 strains were clustered as the significant phylogroup with limited SNP (5–19) divergence, suggesting an outbreak of this clone in our hospital. In contrast, ST392 was relatively heterogeneous, with more detectable SNPs (2–48), indicating lower clonality in this variant.

Clustering of CC147 genomes based on a minimum spanning tree revealed a high level of proximity among ST147 strains (maximum of 11 allelic distances), while ST392 exhibited shallow branching and one strain (No. 45) (OXA-48+/NDM−) had a relatively high allelic distance (43 allele differences) from the nearest ST392 node (Fig. 2).

Fig. 2

Minimum spanning tree of CC147 based on the core genome MLST. The 25 CC147 strains were analyzed based on comparing 2365 alleles calculated in Ridom SeqSphere. Cluster was defined based on the maximum allelic distance of ≤ 15 alleles. Inside the bubbles the isolate ID numbers are shown, and the allelic distances between isolates are represented on the lines connecting them. Clusters 1 (ST392) and 2 (ST147) are shown by pink and pale blue zones, respectively

The highest prevalence of resistance markers detected in ST147

Except for fosA and oqxA/oqxB (efflux pumps), which were detected in all study populations, the most common resistance marker was blaSHV- (51, 94.4%), which was chromosomally located in all positive strains, followed by blaCTX-M-15 (39, 72.2%), dfrA (38, 70.4%), sul1 (37, 68.5%), aac3IIa (27, 50%), qnrS (25, 46.3.1%), blaOXA-1 (20, 37.03%) and aac6-Ib-cr /aph(3´)-Ib / aph(6´)-Id / aadA5 (19, 35.1% for each).

The blaOXA-48, blaNDM-1 and blaOXA-48/NDM-1 were detected in 18 (33.3%), 14 (24.9%), and 6 (11.1%) strains, respectively. While blaOXA-48 was distributed among different STs including ST14, ST45, ST48, CC147, and ST377, the dual carbapenemases of blaOXA-48/blaNDM-1 was detected in CC147 and mostly in ST392 (5 strains). Focusing on the CC147 resistance gene content revealed that this clonal group harbored a significantly higher number of different resistance markers compared to other STs. However, it wasn’t uniform when the two members were considered separately. It means ST147 had a more homogeneous combination of resistance markers than ST392, including blaNDM-1 (9 [100%] vs. 11 [68.8%], P: 0.1), sul1 (9 [100%] vs. 14 [87.5%], P: 0.02), blaOXA-1 (9 [100%] vs. 8 [50%], P: 0.02), blaOXA-10 (7 [77.8%] vs. 0, P < 0.001), armA (7 [77.8] % vs. 0, P < 0.001), aac(6´)-Ib-cr (9 [100%] vs. 7 [43.8%], P: 0.008) and blaTEM-1B (9, [100%] vs. 0, P < 0.001). The exception was blaOXA-48 which was markedly harbored by ST392 (9 [56.2%], P: 0.04) (Fig. 1). Furthermore, mutations in the OmpK36 porin (Gly134Asp135 duplication in loop 3, OmpK36GD) that constrict this porin channel and render it resistant to antibiotics were found in all ST147 and two ST377 strains.

The association between carrying resistance elements and resistant phenotypes was assessed. Isolates harboring blaOXA-1 were significantly resistant to aztreonam (P: 0.01) and cefepime (P < 0.001). Moreover, the blaCTX-M-15 carriage was in association with a resistant phenotype to aztreonam, piperacillin/tazobactam, ampicillin/sulbactam, ceftazidime, cefepime, amoxicillin/clavulanate (P < 0.001 for all), and cefotaxime (P: 0.01). Resistance against amikacin was associated with the carriage of aac(6´)-Ib-cr (9 [75%], P: 0.002) and armA (7 [58.3%], P < 0.001), while gentamicin resistance was detected in strains harboring aac(6´)-Ib-cr (18 [51.4%], P: 0.001), aac3Iia (27 [77.1%], P < 0.001), and armA (8 [22.9%], P: 0.04). The least frequent resistance marker was the 16S rRNA methylase gene rmtC which harbored by two ST147 strains, both were resistant against all aminoglycosides. The aminoglycoside nucleotidyltransferase aadA5 was detected in 18 CC147 strains and was in strong association with resistant phenotype against gentamicin (18 [94.7%], P: 0.001) and tobramycin (19 [100%], P < 0.001). Even when the nucleotidyl-phosphatases aph(6’)-Id and aph(3’)-Ib were combined, resistance to the three aminoglycosides tested was not associated (Table 2). The proportions of acquired resistance markers which provide resistant phenotypes against the study antibiotics have been shown in sunburst plot (Additional file 1).

Table 2 Resistance genotypes in association with phenotypic characteristics

The fluoroquinolone-resistant (FQ-R) phenotype (non-susceptibility to either ciprofloxacin or levofloxacin) was found in 42 (77.7%) of the study isolates and was associated with parC (S80I) + gyrA (S83I) mutations (27 [64.3%], P: 0.001) or carrying the qnrS1 gene (25 [59.5%], P: 0.001) and aac(6´)-Ib-cr (19 [45.2%], P: 0.004). Among the CC147, all were concomitantly positive for qnrS1 and gyrA S83I / ParC S80I, except two ST147 and two ST392 strains, which were negative for qnrS1. The aforementioned genotypes were mostly prevalent in CC147 strains. The other PMQR found was qnrB1/B4, which was detected in 13 ST45, ST48, ST348, and ST1308 strains but was not linked to fluoroquinolone resistance (Table 2).

Different plasmid repertoires were found in CC147 members

Concerning plasmid replicon types, a total of 23 replicon types were found. The IncF was the dominant plasmid type that was identified in all isolates, followed by the IncHI1B (31, 57.4%), IncL (24, 44.4%), and the col_pHAD28 (24, 44.4%). The most common IncF-type plasmids were IncFII_K (26, 48.1%), IncFIB_K (18, 33.3%), and IncFIB(K) (pCAV1099-114) (16, 29.6%). Furthermore, the two members of CC147 showed different patterns of replicon types. The IncFIA (HI1), IncFII_K, IncFII_Yp, IncFIB (pQil), IncFIB (K) (pCAV 1099-114), and IncFIB (pB171) (P: 0.001 for all) were detected significantly in ST147, and the IncHI1B (pNDM-MAR), IncL (pHAD28), Col440I, and IncFII were remarkably harbored by ST392. The three types of plasmids, IncFII(Yp), IncFIB (pB171), and pKPC-CAV 1321, were exclusively detected in ST147 strains (Fig. 3).

Fig. 3

Bar-plot displaying the prevalence of different replicon types and their association with STs. The Y-axis represents the number of positive isolates in each STs

On three occasions, it was possible to link replicon type and resistance elements to the same contig: (i) ST147 strains carried blaNDM-1 on an IncFIB (pB171). (ii) ST392 harboured blaNDM-1 on a Col440I, and (iii) blaOXA-48 was embedded in a Tn1999 composite transposon in association with the IncL/M replicon type, the notorious plasmid lineage responsible for the worldwide dissemination of blaOXA-48.

The bla NDM-1+ transconjugants was successfully developed

By using conjugation experiment, the NDM-1+ plasmids of 3 studied isolates (ST147 and 2 ST392 strains) were successfully transferred to recipient, conferring resistance to carbapenems and cephalosporins in transconjugants. Plasmid gel extraction and the following PCR experiment of the transconjugants revealed that the blaNDM-1 was located on conjugative plasmids. However, in the experiments the OXA-48+ transconjugants were not detected.

Clones differed in their content of virulence genes

We analyzed additional important virulence factors such as yersiniabactin, aerobactin, and salmochelin siderophore systems in the study population, whose bacterial survival has been shown to be enhanced by acquisition of iron from the host. Aerobactin iutA and capsule wzi were the most commonly identified virulence factors, and they were found in all isolates. Type 1 (fim) and type 3 (mrk) fimbriae gene clusters are major adhesins to biotic and abiotic surfaces. Type 1 gene cluster was identified in all isolates. The “mannose-resistant Klebsiella-like (type 3) fimbriae cluster” (mrkABCDFHIJ) was detected in 47 (87%) of our collection.

Accessory genomes are those virulence markers that have variable presence [40]. Ferric uptake system, kfuABC, as an accessory genome was present in 9 (16.9%) strains, including ST1308, ST6510, high-risk clone ST14, and hvKp ST2159. Yersiniabactins, including “yersiniabactin receptor gene” (fyuA), irp1/2, and “yersiniabactin siderophore cluster” (ybtAEPQSTUX) were detected in 46.3% of strains and in association with 4 chromosomally “integrated conjugative elements” (ICEs), mainly from ST14, ST40, ST45, ST86, and ST392. The majority of yersiniabactin-carrying (ybt+) isolates, 22 out of 25 (88%), harbored ICEKp3 and ICEKp4 related to ybt 9 and ybt 1 lineages, corresponding to isolates from ST392/ ST86 and ST14/ST45/ST48, respectively. Other detected MGEs and ybt lineages with low frequency (< 5%) were ybt 16 /ICEKp12 and ybt 14 /ICEKp5 (Table 1). Aerobactin iucABCD (iuc1) and salmochelin (iroBCDN) (iro1) were detected among three hvkp strains, including two ST86 (positive for ICEKp3) and one ST2159 strains (Fig. 4).

Fig. 4

The virulence genes content of K. pneumoniae. A BLAST Ring Image Generator (BRIG) image shows the presence of virulence factors among different STs. The FFN format file of hvKp (strain No. 290 identified as ST86) was downloaded from VFDB website, and open reading frames (ORFs) are annotated based on the VFDB results and used as reference strains to draw the image. Aligning between the regions of interest and each genome are shown as Colored segments (indicate > 70% similarity), and gray segments (indicate > 50% similarity). From inside to outside of the figure: “deep sky blue “rings”; ST14, “blue” rings; ST40, “purple” rings; ST45, “yellow” rings; ST48, “aquamarine” rings; ST147, “red” rings: ST392, “blue-green [teal]” rings; ST377, “pale green” rings; ST1308. The other undefined colored rings are STs that include one or two strains

Iranian ST147 strains are phylogenetically more related to European genomes

By mapping 183 genomes of K. pneumoniae, including our nine ST147 strains and 174 publicly available ST147/NDM + genomes to K. pneumoniae strain 4/1–2 (GenBank ID: CP023839.1, NDM + strain), 2943 SNPs were identified. Isolates were clustered into different groups and the largest one was consisted of United States (Chicago region) genomes. The majority of downloaded genomes (n = 128) originated in the United States (except for two genomes; SRR3228444 and SRR8984905, all were from the Chicago region), followed by Europe (n = 36), Southeast of Asia (n = 4), South America (Peru; n = 5), and Canada (n = 1). The Iranian strains were clustered with three European/one Canadian (KL64/ O2v1) strains with ≤ 30 SNP differences (Fig. 5). Focusing on resistance markers revealed that the United States isolates harbor different pattern of these elements compared to other genomes, including blaCTX-M-15 (USA; 1.6% vs. 92.3%, P < 0.001), blaOXA-48 (USA; 0.8% vs. 41.5%, P < 0.001), blaOXA-1 (USA; 3.1% vs. 67.7%, P < 0.001), blaOXA-10 (USA; 0.8% vs. 12.3%, P < 0.001), qnrS (USA; 3.1% vs. 46.2%, P < 0.001), armA (USA; 93.8% vs. 26.2%, P < 0.001), qnrB (USA; 2.3% vs. 33.8%, P < 0.001), qnrE (USA; 0 vs. 7.7%, P: 0.004), aac6Ib-cr (USA; 6.3% vs. 84.5%, P < 0.001), aac3Iia (USA; 2.3% vs. 55.4%, P < 0.001), rmtC (USA; 1.6% vs. 10.8%, P: 0.007), rmtF (USA; 0 vs. 23.1%, P < 0.001), aph3Via (USA; 94.5% vs. 16.9%, P < 0.001), blaOXA-9 (USA; 0.8% vs. 13.8%), and aac3IId (USA; 93.8% vs. 0, P < 0.001). The link between the ST147 and the KL64 was found in this collection, except for nine strains that were identified as KL10, KL107, and KL112 (Fig. 5).

Fig. 5

Maximum-likelihood phylogeny of ST147/NDM + strains based on whole genome SNPs analysis. A phylogenetic reconstruction of 183 strains (including the 174 downloaded genomes and our 9 study strains) with K. pneumoniae strain 4/1–2 (GenBank ID: CP023839.1, NDM + strain), as the reference strain is shown. Colored strips surrounding the phylogram represent the country of origin of each strain, capsular types, ICE types, carriage of NDM/CTX-M-15, OXA-types, and genes involved in gentamicin and amikacin resistance

The comparable plasmid harboring blaNDM-1 was found in four downloaded genomes clustered with our strains, and IncF pMLST revealed it to be of the K2/K5:A22: B36 replicon type. Considering the ICE and Yersiniabactin types, all United States strains were negative. While 19 ICEKp4 (ybt 1), nine ICEKp12 (ybt 16) and 11 ICEKp3 (ybt 9) positive isolates were identified. Genomic characteristics of ENA downloaded genomes have been shown in Additional file 2.

ST147 K. pneumoniae genomes reported from neighboring countries in our geographic region were included in the cgMLST analysis to identify genetic relatedness. Using a ≤ 15 allele difference threshold in cgMLST analysis, strains were clustered in 13 complex types (CTs) and 18 singletons (Fig. 6). The CTs were formed by genomes from the same countries, except for one cluster (CT1) which was the biggest and included isolates from all countries. Iranian genomes (our 9 isolates and 3 previously reported) were grouped into cluster 3 (11 strains) and one strain remained as a singleton (Ir-2). Considering the clustered strains, the highest diversity was found among Indian strains that were categorized into 5 clusters, followed by genomes from Russia /Turkey (4 clusters) and Pakistan (2 clusters) (Fig. 6). Characteristics of downloaded genomes deposited from neighboring countries are shown in Additional file 3.

Fig. 6

Minimum Spanning Tree showing distance based on cgMLST of 1826 genes using the parameter ‘pairwise ignoring missing values for 130 ST147 genomes. Each circle is named with the geographical origin, including DTU: our ST147 strains, Ir: Iran, Tu: Turkey, In: India, P: Pakistan, Is: Israel, Le: Lebanon, Ru: Russia, UAE: United Arab Emirates. The data of carbapenemase genes is shown by the coloured rings inside each bubble

Diversity of genetic context involved in spreading of bla NDM-1, armA and bla CTX-M-15

Analysis of the genetic context of blaNDM-1 in ST147 and ST392 as the positive clones corroborated the previous conserved sequence of this gene, showing that the NDM-1 gene is located downstream of truncated ISAba125 and is followed by bleMBL, Isomerase, protein disulfide reductase dsbD, cutA, and GroES-GroEL genes; however, there were some differences between these two clones based on the arrangement of mobile genetic elements (MGEs) surrounding this segment. Among ST392 strains, only one isoform of MGEs, including ΔISKpn26-NDM core structure-TnAs3-ΔIS3000-Tn5403 was found on plasmid col440I (Fig. 7a). In ST147, the NDM-1 core structure was discovered upstream of the rmtC 16S rRNA methylase gene, flanked by MITESen1 and IS903B in one strain and MITESen1/MITESen1 composite transposon in the other (Fig. 7b, c). In rmtC negative ones, the NDM core structure was bracketed by MITESen1/ISSen4 (Fig. 7d). The plasmid type of NDM in ST147 was IncFIB (PB171).

Fig. 7

Genetic context of resistance markers among study population. a contig harbouring blaNDM-1 in ST392, b–d three forms of MGEs bracketing NDM-1, e, f two isoforms of MGEs arrangements in blaCTX-M-15 harbouring contigs, g the integron 1 cassette structure harbouring the armA

The blaCTX-M-15 harbored by ST14, ST45, ST48, ST377 and all except one CC147 strain. The core segment was arranged by a truncated and/or intact form of ISEcp1 upstream of CTX-M-15, and the "WbuC family protein fold metalloprotein" gene was located downstream but in a reversed direction along with a copy of Tn2. This core structure was co-localized with (i) blaTEM-1B, tnpA, aph(6)-Id, aph(3)-Ib and sul2 genes, which were bracketed by IS15DIV and IS5075 (IS110 family), and (ii) IS2 and qnrS1. The length of CTX-M-15 carrying contigs varied across STs; however, data suggest a common origin via horizontal transfer across multiple sequence types. (Fig. 7e, f).

The other 16S rRNA methylase gene, armA, was detected in seven out of nine ST147 and one ST48 strain (no. 639). The core structure which was located on the class 1 integron structure was consisted of armA bracketed by ISEc28 and ISEc29, along with ARR-2, cmlA1, blaOXA-10, aadA1, qacE, and sul1 in downstream, and msr(E)/mph(E) in upstream. The genetic environment of this structure was conserved in all positive strains and was flanked by two copies of IS15DIV (Fig. 7g).