From July 2013 to November 2015, a total of 131 BSIs patients hospitalized in the First Affiliated Hospital of Wenzhou Medical University were shown to carry K. pneumoniae. Among these, 35 patients were defined as XDR K. pneumoniae BSIs patients if their blood cultures were positive for XDR K. pneumoniae and with clinical signs of systemic inflammatory response syndrome. BSIs patients were further divided into two groups based on clinical outcome of antimicrobial treatment: death group (n = 14) and survivor group (n = 21). Medical records and clinical data (patient age, gender, ICU length of stay, reasons for ICU admission, intensive care procedures, comorbidities, medical care measures before BSIs, antibiotic administration, microbiological data, and outcomes) were collected and analyzed and are presented in Table 1.

All of the investigation protocols in this study were approved by The Ethics Committee of The First Affiliated Hospital of Wenzhou Medical University. Informed consent was waived because this retrospective study with retrospective observational nature mainly focused on bacteria and did no interventions to patients.

This retrospective study was conducted at the First Affiliated Hospital of Wenzhou Medical University, China. Based on the standardized international definition for XDR described by Li et al. (2012), 35 non-repetitive K. pneumoniae strains, isolated from nosocomial BSIs patients that were found to be susceptible to two or fewer antimicrobial categories, were collected. Initially, bacterial identification and antimicrobial susceptibility tests were conducted by the Vitek2 system (BioMèrieux, France). Then, the isolates were stored in frozen condition at −80°C with 30% glycerol. Subsequently, minimum inhibitory concentrations (MICs) of tigecycline and polymyxin B were determined by broth dilution method and interpreted by the recommendation of the European Committee on Antimicrobial Susceptibility Testing (EUCAST, 2015). Escherichia coli ATCC 25922 served as the control strain for susceptibility testing.

The presence of resistant mechanisms, including ESBLs genes (bla_CTX−M−1, bla_CTX−M−9, bla_TEM, bla_SHV, bla_VEB, and bla_PER), AmpC genes (bla_CMY, bla_FOX, bla_MOX, bla_DHA), carbapenemase genes (bla_KPC, bla_SPM, bla_IMP, bla_VIM, bla_GES, bla_NDM, bla_OXA−23, bla_OXA−48), PMQR genes [qnrA, qnrB, qnrC, qnrD, qnrS, qepA, aac(6′)-Ib-cr, oqxA, oqxB, gyrA, parC], AMEs [AAC(6′)-Ib, APH(3′)-Ia, AAC(3)-IV, ANT(2″)-Ia], and 16S-RMTase genes (armA, rmtA, rmtB, rmtC, rmtD, rmtE) were investigated by PCR and sequencing. For each isolate, DNA was extracted from fresh bacterial colonies using an AxyPrep Bacterial Genomic DNA Miniprep kit (Axygen Scientific, Union city, CA, USA). PCR assays were performed on a Veriti 96-well Thermal Cycler (Bio-Rad, USA) using specific primers, corresponding to related studies (Jacoby, 2009; Yu et al., 2009; Li et al., 2012; Ramirez and Tolmasky, 2013). Primer sequences are available on request. BLAST was utilized to align drug-resistance gene nucleotide sequences (http://blast.ncbi.nlm.nih.gov/Blast.cgi).

To determine the transferability of resistance determinants, Luria-Bertani (LB) mating experiments were implemented. Sodium azide-resistant E. coli J53 served as the recipient strain (Yi et al., 2012). Transconjugants were selected on Mueller-Hinton agar plates containing sodium azide (100 mg/L) and ertapenem (0.5 mg/L). The resistance genes successfully transferred from the donor strains were verified by PCR. MICs of antibiotics for the transconjugants were compared to donors and E. coli J53 to further confirm the transferable resistance genes.

Major plasmid incompatibility groups: F, FIA, FIB, FIC, HI1, HI2, I1, L/M, N, P, W, T, X, Y, K, A/C, B/O, FII, FrepB, were detected by a PCR-based replicon typing (PBRT) scheme (Carattoli et al., 2005). S1 nuclease converted supercoiled plasmids into full-length linear molecules, and S1-PFGE can be used to screen for megaplasmids simultaneously (Barton et al., 1995). Total DNA from the K. pneumoniae donor strains and E. coli transconjugants were isolated using an Axyprep Bacterial Genomic DNA Miniprep kit, digested with S1 nuclease (Takara Bio, Inc.) and analyzed in a CHEF-Mapper XA PFGE system (Bio-Rad). The gel was then subject to Southern blot analysis, after transfer to a positively charged nylon membrane (Roche Diagnostics, Branford, USA) by the capillary method. The membrane was subject to hybridization with labeled bla_KPC−2, rmtB, bla_CTX−M−9 probes according to the instructions of Detection Starter Kit II (Roche, Sant Cugat del Vallès, Spain). The plasmids of Salmonella H9812 served as size markers (Zhou et al., 2015).

A crude outer membrane protein (OMP) fraction was isolated by sonication, and OMPs were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (12% SDS-PAGE), and OmpK35 and OmpK36 were detected by immunoblotting (serum dilution 1:20,000). K. pneumoniae ATCC 13883 served as a control strain for OMP profiling (Webb et al., 2012; Zhou et al., 2015). For outer membrane purification, crude membranes were isolated from K. pneumoniae ATCC 13883 and subjected to sucrose density fractionation. Membrane fractions were then subject to analysis by SDS-PAGE and Coomassie blue staining. Alternatively, fractions were analyzed by immunoblotting to determine the localization of inner (F₁β) and outer (BamB) membrane proteins markers (Clements et al., 2009; Zhou et al., 2015).

Clonal relatedness for the 35 K. pneumoniae strains was analyzed by pulsed-field gel electrophoresis (PFGE). Genomic DNA was extracted using an AxyPrep Bacterial Genomic DNA Miniprep kit, subjected to complete digestion with the restriction endonuclease XbaI (Takara Bio, Dalian, China), and the diagnostic DNA fragments then separated in a PFGE CHEF-Mapper XA system (Bio-Rad) using 0.5 × Tris-borate-EDTA buffer at 120 V for 19 h, with pulse times ranging from 5 to 35 s. DNA fragments were stained with Gel Red (Biotium, USA) and analyzed using Quality one software (Bio-Rad). The DNA fingerprint patterns were analyzed according to the criteria proposed by Tenover et al. (Hu et al., 2013) and strains with more than 80% similarity were regarded as the same clone. Molecular typing of XDR K. pneumoniae strains was performed according to the protocol described on the Institut Pasteur K. pneumoniae MLST website (http://bigsdb.web.pasteur.fr/klebsiella/klebsiella.html). The sequences of seven housekeeping genes (i.e., gapA, infB, mdh, pgi, phoE, rpoB, and tonB) were amplified and sequence types (STs) were assigned by the MLST database according to Diancourt et al. (2005). The novel STs were submitted to the MLST database.

Categorical variables were compared with the chi-square test using SPSS software (version 17.0). Calculated p-values of < 0.05 were considered to be statistically significant.

There is a growing trend of mortality in hospital patients caused by infections with K. pneumoniae (Paczosa and Mecsas, 2016). A total of 35 nosocomial BSIs patients (age 14–90 years; male: female 4:1) were included in the present study (Table 1). These patients had been admitted in the intensive care unit (60.0%), hematology department (28.6%), and neurosurgery department (11.4%), data not shown. The BSIs patients were primarily elderly, suffered severe comorbidities and had been submitted to invasive procedures, such as central venous catheterization and urinary catheterization. As documented in Table 1, the BSIs resulted in lengthy ICU stays (1–58 days; median 16 days), within long overall hospital length of stay (1–86 days; median 30 days). The mortality rate attributed to BSIs caused by XDR K. pneumoniae was 40.0%, i.e., 14 of the 35 patients died from the BSI. Two groups of BSI patients had an increased risk of death, those that received urinary catheterization (34.3% died vs. 22.9% survived, P = 0.007), and patients with intracranial disease (17.1% died vs. 2.8% survived, P = 0.01). In terms of their treatment, combination therapy including meropenem proved to be ineffective against mortality of BSIs patients caused by XDR K. pneumoniae (P = 0.015), but tigecycline in combination with other antimicrobial agents was significantly effective and reduced the risk of death (P = 0.016). Furthermore, nine patients of the survivor group were treated with fosfomycin combined with more than three other antibiotics.

Molecular typing enables detection of nosocomial transmission of bacterial pathogens, and can assist in identifying the routes of transmission in hospital settings. Multilocus sequence typing (MLST) identified seven different STs with ST11 being the predominant clone (Table 2, Figure 1). This finding is consistent with the epidemic dissemination of K. pneumoniae carbapenemase (KPC)-producing K. pneumoniae in China described in another study (Zhou et al., 2015). Another six STs were also identified: ST1525 (FK688/13), ST290 (FK782/13), ST2281 (FK1468/14), ST268 (FK2152/15), ST14 (FK2203/15), ST656 (FK2180/15). The novel ST2281 (allelic profile: 4-1-1-22-7-4-35) which was a multiple locus variant, has not previously been documented and has been submitted to the MLST database.

The PFGE patterns of the 35 XDR K. pneumoniae isolates show a major cluster of 23 closely related isolates that exhibited >80% similarity: these are all KPC-2-producing resistant strains and are all ST11 clones (Figure 1). The rest 12 isolates which exhibited <80% similarity were assigned to seven STs (ST11, ST1525, ST290, ST2281, ST268, ST14, and ST656). Eighteen XDR K. pneumoniae isolated from ICU patients (directly admitted and transferred) belong to the predominant cluster, and the other five isolates belonging to the predominant cluster were separated from other departments. A worrying finding is that the close relatedness among the 23 XDR K. pneumoniae isolates suggests a phenomenon of clone dissemination within the clinic and the transfer of patients in different medical departments may accelerate clonal spread of pathogens.

MIC assays of the patient isolates revealed that each of the XDR K. pneumoniae strains were highly resistant to at least one of the β-lactams, carbapenems, aminoglycosides, quinolones, β-lactam/inhibitor combinations, and other clinical antimicrobial agents (Table 3). Consistent with the treatment success is the observation that tigecycline exhibited superior bacteriostasis activity in vitro. In addition, 33 isolates were susceptible to polymyxin B (Table 3).

Diverse resistance determinants were detected among the 35 XDR K. pneumoniae strains (Table 2). Of these, 33 strains (94.3%) were detected as co-harboring three or more resistance determinants. There is a global spread of KPC-2 carbapenemase (Naas et al., 2008; Nordmann et al., 2009) that usually explains such findings. However, two strains (FK688 and FK1934) did not express KPC-2 carbapenemase (highlighted in Figure 1). Other β-lactamases were also prevalent, including CTX-M-type ESBLs (31 isolates), SHV (28 isolates), and TEM (11 isolates). In addition, several strains harbored the β-lactamases CMY and DHA-1.

Additional genes conferring drug resistance were sequenced, including the DNA gyrase encoding gyrA and the DNA topoisomerase IV encoding parC. Amino acid substitutions detected in 31 (88.6%) of the strains were gyrA (S83I, D87G) and parC (S80I). In addition, mutations were detected in quinolone resistance-determining regions (QRDRs). Intriguingly, 29 of the 30 major ST strains (ST11, ST656) among the fluoroquinolone-resistant strains carried the favorable “double serine” mutations in the gyrA and parC genes plus an additional-energetically less favorable gyrA mutation. A single ST11 strain (2048/15) carried just one favorable serine mutation without any less favorable mutations. In contrast, fluoroquinolone-resistant minor clone strains either failed to harbor any gyrA or parC mutations (ST1525, ST290, ST268, ST14) or harbored one favorable serine mutation together with a less favorable mutation (ST2281) (Table 2). The 16S-rRNA methylase encoding gene rmtB was present in 28 (80%) of the strains, a gene that confers resistance to most clinically relevant aminoglycosides. The plasmid-mediated quinolone resistance genes qnrB, qnrS, and aac(6′)-Ib-cr were also determined as present in 11 (31.4%) of the strains. 35 (100%) of the strains possessed AMEs, including AAC(6′)-Ib (100%), APH(3′)-Ia (48.6%), AAC(3)-IV (11.4%), ANT(2″)-Ia (0%). Taken together, these results demonstrate carriage of multiple genetic determinants for drug resistance in the BSI strains.

KPC-2 and other resistance determinants are often carried on plasmids (Kuai et al., 2014), and S1-PFGE revealed the presence of one or more plasmids in each of the isolated strains (Figure 2). The detected plasmids by S1-PFGE ranged in size from ~50 to ~390 kb. For an initial assessment of the plasmids, transconjugation experiments were established using E. coli J53 as the recipient. Carbapenemase genes from 12 strains were successfully transferred to the recipient E. coli J53, but no transconjugants were recovered for the other 23 donor strains, despite repeated attempts, may due to the lack of sex factor F or conjugative plasmids which are key vehicles and contain simultaneously important elements, including an origin of transfer, DNA-processing factors (a relaxase and accessory proteins), and mating pair formation proteins (Achtman et al., 1971; Goessweiner-Mohr et al., 2014) (Table 2). The MIC values for the transconjugants determined for β-lactam antibiotics were consistent with the transfer of carbapenemase expression. Southern blot analysis revealed the presence of the bla_KPC−2 in the transconjugants and in 11 cases corresponded to a plasmid of ~50 kb (Figure 3), while strain J1743 carried the bla_KPC−2 gene on a plasmid of ~160 kb (Figure 3). These results are in accordance with the S1-PFGE analysis of the isolates (Figure 2). Some of these strains were found to be FrepB-positive by PBRT, showing that these plasmids belong to an IncF-type.

The major porins of E. coli, OmpF and OmpC, provide for most of the flux of molecules up to ~650 Da across the outer membrane (Nikaido et al., 1983; Sugawara et al., 2016) and thereby play a role in the susceptibility of E. coli to antibiotics (Nikaido, 2003; Zgurskaya et al., 2015). The homologous porins of K. pneumoniae, OmpK35 and OmpK36, have been highlighted as the primary channels through which beta-lactams cross the outer membrane (Kaczmarek et al., 2006). OmpK35 and OmpK36 are highly abundant proteins of the outer membrane in K. pneumoniae, and can be readily detected by Coomassie blue staining after SDS-PAGE analysis of the outer membranes (Rath et al., 2009). A procedure was optimized to purify outer membranes from the control strain K. pneumoniae ATCC 13883, using sucrose gradient fractionation to separate outer membranes from inner membranes (Figure 4A). Immunoblotting for marker proteins F₁β and BamB served as controls for the fractions (Figure 4A). Comparative analysis of outer membrane fractions isolated from the carbapenem-resistant, yet carbapenemase-negative strains FK688 and FK1934 using antibodies specific for OmpK35 and OmpK36 showed no detectable expression of OmpK35 and OmpK36 (Figure 4B). PCR primers designed to amplify regions of the corresponding loci failed to detect ompK35 and ompK36 in FK688 and FK1934 (data not shown).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.