Traditional microbiological and molecular analyses
Figure 1 shows the antimicrobial susceptibility findings and clonal relationships between the isolates obtained by means of traditional methods. The 68 isolates showed resistance to all β-lactams and ciprofloxacin, and all but one were carbapenemase producers. Almost 90% were resistant to at least one of the two aminoglycosides tested; 47% were resistant to tigecycline; seven samples (10.30%) were resistant to colistin; and two (3.1%) were intermediate/resistant to all of the tested antibiotics (Additional file 1: Table S1).
The use of REP-PCR revealed 13 DiversiLab patterns that clustered into five groups. The most prevalent was cluster 3 (n = 46 samples) followed by cluster 1 (n = 13), 2 (n = 4), 4 (n = 2) and 5 (n = 2). One isolate was a singleton (Fig. 1).
Whole-genome sequencing and in silico data analysis
The isolates’ final assembly obtained by means of WGS ranged from 96 to 237 contigs of >500 bps/sample, with N50 values of between 68,340 and 151,723, thus covering ~5.8 Mb of the K. pneumoniae genome. The molecular analysis was made in order to obtain information about the MLST, resistomes, virulomes, plasmid replicon types and core SNP genotypes of all of the strains (Fig. 2).
The MLST analysis, which was based on seven housekeeping genes (rpoB, gapA, mdh, pgi, phoE, infB and tonB), revealed six different sequence types, four of which (ST258, ST512, ST745 and ST1519) were members of clonal complex (CC) 258. ST258 was the “founding” CC genotype whose diversification was reflected in the appearance of the other STs: ST512 was a single-locus variant (SLV) of the ancestral ST, whereas ST745 and ST1519 were double-locus variants (DLVs). ST101 and ST307 were evolutionarily independent of the CC. The GoeBURST radial diagram (Additional file 2: Figure S1) shows the relatedness and patterns of evolutionary descent among the defined STs: the most representative were ST512 (46/68) and ST258 (16/68), whereas three strains belonged to ST307 and a further three were identified as ST101, ST745 and ST1519.
The in silico β-lactamase characterisation of the 68 sequenced isolates analysed 23 gene loci reported on the Pasteur website. The most frequent carbapenemase producing genes were blaKPC and blaSHV (both 98.53%), followed by blaTEM (80.88%), blaCTX-M (4.41%) and blaOXA (2.94%). The dissemination of blaKPC genes was sustained by the blaKPC-2 variants (ST258 and two ST307 isolates) and the blaKPC-3 allele (ST512 and their ST745 and ST1519 single-locus variants, and one ST307 isolate) (Fig. 2). Many blaKPC genes are associated with the Th4401 promiscuous transposon-related structure (which is approximately 10 kb in size and consists of a transposase, a resolvase, the blaKPC gene, and two insertion sequences, ISKpn6 and ISKpn7) contained in a number of different plasmids often belonging to incompatibility (Inc) group F [33]. BLAST alignments between the KPC-2 and KPC-3 contigs and the IncF Tn4401a pKpQIL-IT plasmid region (Plasmid Accession No. JN233705.2) showed that blaKPC-2 and blaKPC-3 genes were embedded in the Tn4401a transposon isoform.
The ST101 isolate (Kp27) was resistant to carbapenems in the absence of blaKPC genes and, in this isolate alone, the possible loss or alteration of outer membrane porins was investigated. GenBank accession numbers for K. pneumoniae strains used as reference for OmpK35 and OmpK36 gene sequences were AJ011501 and FJ577673, respectively. The ompK35 sequence of Kp27 showed a perfect match of 1195/1202 bp with the reference, and a nucleotide deletion that leads to a truncated protein product of 63 amino acids. The ompK36 sequence of Kp27 perfectly matched only 1098 of the 1151 ompK36_FJ577673 nucleotides and 20 bp gaps led to amino acid deletions and substitutions (see Kp27_ST101_omp35_omp36_sequences_analysis.doc in the Additional file 3).
In most of our isolates, the presence of β-lactamase genes was associated with sequences encoding for resistance to other classes of antibiotics: aminoglycosides (aadA2, aadA5, aac(3′)-Ia, aac(6′)-Ib, strA, strB and armA genes), fluoroquinolone (oqxA, oqxB, qnrB66 and aac(6′)-Ib-cr genes), MLS (msrE, mphA and mphE genes), phenicol (catA1 and catB3 genes), sulfonamide (sul1 and sul2 genes), tetracycline (tetA gene) and trimethoprim (dfrA-12/14/17 alleles) (Fig. 2).
In order to investigate AcrAB-TolC efflux pump-dependent and pump-independent tigecycline resistance mechanisms, we analysed the sequences of the acrR, ramR, marR, soxR, lon, rpsJ and rpoC genes [25, 26] by aligning our isolates with the reference genes of the tigecycline-susceptible isolate K. pneumoniae subsp. pneumoniae MGH78578 (GenBank Accession No. CP000647). The most frequently observed mutations were SNPs, but the gene sequences varied in terms of their presence, mutations and deletions. However, it was not possible to correlate these genomic features with tigecycline resistance because of the presence of a common mutation in susceptible, intermediate and resistant strains.
Many of the transcriptional regulation systems controlling lipopolysaccharide (LPS) modifications are involved in colistin resistance and, recently, the mobile mcr-1 and mcr-2 genes have been related to the increasing colistin resistance induced by the over-use of polymyxins in animals [27, 28]. We studied the mgrB, pmrA/pmrB, phoP/phoQ chromosomal genes taking the colistin-susceptible isolate K. pneumoniae subsp. pneumoniae MGH78578 (GenBank Accession No. CP000647) as reference, and also looked for the presence of plasmidic mcr-1 (reference: E. coli pHNSHP45 Accession No. KP347127.1) and mcr-2 genes (reference: E. coli KP37 pKP37-BE Accession No. NG_051171.1). The BLAST results revealed some changes in the chromosomal genes, and nucleotide variations were found in the pmrA/pmrB and phoP/phoQ genes of all of the colistin-resistant strains (Kp6, Kp19, Kp34, Kp37, Kp64, Kp65 and Kp68) and in colistin-sensitive isolates. Conversely, there wee some interesting variations in the mgrB gene in Kp19 and Kp37, with the locus being disrupted by an IS5-like element (best match: KKBO-4 isolate, Accession No. HG008893.1). Isolates Kp6, Kp64, Kp65 and Kp68 showed a wild-type mgrB gene, whereas the Kp34 mgrB contig started with the query sequence, and covered only 116/144 bps of the reference. The plasmid hits mcr-1 and mcr-2 were not found.
Figure 2 also shows the virulence repertoire detected in our samples. The mrk operon, which encodes type 3 fimbrial adhesins was detected in all of the isolates, and mrkD and mrkH genes were respectively missing from some ST512 and ST258 isolates. Aerobactin (irp2) genes, pesticin/yersiniabactin receptor (fya) genes, iron uptake system (kfu) genes, and alternative siderophore yersiniabactin (ybt) genes were detected in only one ST512 sample (Kp23), in all of the ST258 isolates except for Kp36, and in ST101 and ST307. Comparison of the genomes of our isolates with a reference K. pneumoniae strain (Kp13, GenBank Accession No. CP003999) showed that the virulence genes were located chromosomally. The iron acquisition systems were in a region of genomic plasticity consisting of a ~ 65 kbp integrative and conjugative element related to the Yersinia pestis high pathogenicity island (HPI), which is present in many Enterobacteriaceae genera (Additional file 4: Figure S2). The presence of genes coding for the yersiniabactin siderophores (irp1 and irp2) are characteristic of this HPI, and their occurrence in more virulent K. pneumoniae strains has been previously reported [34]. Finally, the K-antigens related to the wzi alleles detected in our isolates were K type-41/wzi-29 (all ST258 samples), K not defined/wzi-154 or wzi-173 (ST307 and ST512) and K type-17/wzi-137 (ST101).
Using the PlasmidFinder web server, a total of nine plasmid replicon types were detected and characterised: IncFIB(pQil), IncFIB(K), ColRNAI, IncX1, IncX3, IncFII(K), IncN, IncL/M(pMU407 and IncFIA(HI1) (Accession Nos. of reference plasmids currently included in the Genbank database: JN233705, JN233704, DQ298019, EU370913, JN247852, CP000648, AY046276, U27345 and AF250878). The similarity in alignment between the best matching plasmid in the database and the corresponding sequence in the input genome ranged from 97.97% to 100%. Between three and five replicons were identified in each K. pneumoniae isolate and in various combinations: four arrangements of three plasmid replicons, five different combination of four replicon types, and two different combinations of five plasmid replicons (Fig. 2).
A maximum likelihood core SNP tree of the 68 K. pneumoniae isolates was constructed in order to provide high-resolution strain tracking and discrimination. Of the 38,120 SNPs identified, the 27,203 shared by all of the isolates could be divided into four major and genetically distinct lineages: A, B, C and D (Fig. 2). Lineage A had two major subgroups: A1 and A2. Subgroup A1 included seven ST512 isolates and the ST1519 isolate, whose resistome, virulome and plasmid content was almost identical, with the main exception of isolate Kp23 (ST512), which carried the Yersinia pestis HPI and belonged to the K type-41 group. The A2 isolates (n = 40) clustered into two major branches A2a (six ST512 isolates) and A2b (33 ST512 isolates and one ST745 isolate). A2a included identical strains, with the exception of the absence of the mrkH-7 gene in isolate Kp9; the isolates in A2b were more variable and, interestingly, all of the strains lacking the fimbrial adhesin mrkD gene were in two separate sub-branches.
Lineage B had a series of six subgroups consisting of double or single branches (nine ST258 isolates), and included the strains with the most variable virulome, resistome and plasmid content. However, seven of these isolates, which were identical except for the presence of the plasmid replicons IncX1 in isolate Kp22 and IncL/M (pMU 407) in isolate Kp13, clustered together.
Lineage C included only the ST101 isolate, whereas lineage D (three ST307 isolates) could be divided into two subgroups on the basis of the resistome, virulome and plasmid replicon analyses, with the two closest isolates (containing the blaOXA genes and allele blaKPC-2) being grouped together.
The core SNP analysis was highly consistent with the MLST results and the resistome, virulome and plasmid profiles.