A total of 106 non-duplicated K. pneumoniae isolates (identified using VITEK 2 compact system [BioMérieux, Marcy-l'Étoile, France]) exhibiting a carbapenem-resistant phenotype (imipenem minimum inhibitory concentration [MIC] ≥4 µg/mL and/or ertapenem MIC ≥2 µg/mL) were consecutively collected between June 2013 and September 2017 at the Second Hospital of Anhui Medical University in Anhui, eastern China. The hypermucoviscosity phenotype of the carbapenem-resistant K. pneumoniae (CRKP) isolates was detected using the string test, as described previously [7]. Briefly, all the tested isolates were cultured overnight on Columbia blood agar plates (BioMérieux, Shanghai, China) at 37℃, and bacterial colonies were then stretched with an inoculation loop. The formation of a viscous string over five mm in length was defined as a positive result; CRKP isolates with a positive string test were defined as CR-HMKP.

Antimicrobial susceptibility testing of CR-HMKP isolates was initially performed using the VITEK 2 compact system. Subsequently, ampicillin, amikacin, aztreonam, cefazolin, cefepime, ceftazidime, ceftriaxone, ciprofloxacin, levofloxacin, gentamicin, trimethoprim-sulfamethoxazole, tobramycin (National Institutes for Food and Drug Control, Beijing, China); ertapenem, imipenem (Merck Sharp & Dohme Corp, Hangzhou, China); meropenem (Sumitomo Pharmaceuticals, Suzhou, China); and piperacillin-tazobactam (Pfizer, New York, NY, USA). MICs were confirmed using the agar dilution method according to the recommendations and breakpoints proposed by the CLSI [13^,14]. Tigecycline (Pfizer) and colistin (Sigma Aldrich, Shanghai, China) MICs were determined by the broth microdilution method, using the European Committee on Antimicrobial Susceptibility Testing (EUCAST) guidelines and breakpoint [15]. The standard strains Escherichia coli ATCC25922 and Pseudomonas aeruginosa ATCC27853 were used for quality control.

Both the modified Hodge test (MHT) and imipenem-EDTA double-disk synergy test (DDST) were performed to determine the carbapenemase phenotypes as described previously [16]. Carbapenemase-encoding genes (including bla_KPC, bla_NDM, bla_IMP, bla_VIM, and bla_OXA-48) and common extended-spectrum β-lactamase (ESBL)-encoding genes (including bla_CTX-M, bla_SHM, and bla_TEM) were screened by PCR as described previously [17^,18]. All positive PCR products were sequenced and compared with the published sequences in GenBank (www.ncbi.nlm.nih.gov/blast/).

The capsular polysaccharide serotypes of the CR-HMKP were determined by PCR and sequencing of the K-serotype specific wzi allele, which is closely associated with K-type [19]. The PCR products were sequenced and assigned using an online database (http://bigsdb.pasteur.fr/klebsiella/klebsiella.html), and then compared with the reference sequences in GenBank [20]. The corresponding capsular types were determined according to the criteria of ≥94% DNA identity for the same types and ≤80% identity for different types [21]. Fifteen virulence-associated genes, including aerobactin, allS, entB, fimH, iroB, iucA, kfuBC, magA, mrkD, rmpA, rmpA2, ureA, wabG, ybtS, and ycfM, were detected by PCR and sequenced using previously described primers [7^,22].

The genotypes of the CR-HMKP isolates were determined by pulsed-field gel electrophoresis (PFGE) and multilocus sequence typing (MLST). Briefly, following digestion with the XbaI restriction endonuclease (Takara, Dalian, China), the genomic DNA of the tested isolates and the reference strain Salmonella H9812 were separated on agarose gels using a CHEF Mapper XA PFGE system (Bio-Rad, Hercules, CA, USA). Next, the gel was stained, and dendrograms were constructed from the PFGE data using the BioNumerics software (Applied Maths, Sint-Martens-Latem, Belgium) with the Dice similarity coefficient. The same pulsotype (PT) was defined when the isolates shared >90% similarity. MLST was performed as described previously [23]. The number of alleles was determined and the sequence type (ST) was then assigned using the MLST website (http://bigsdb.pasteur.fr/klebsiella/klebsiella.html).

The virulence of the CR-HMKP isolates was assessed by serum killing and G. mellonella lethality assays. For references, the clinical hypermucoviscous K. pneumoniae strain KPN54798 (ST23, serotype K1, carbapenem-susceptible) was employed as the hypervirulent control (H-control), while the clinical non-hypermucoviscous K. pneumoniae isolate KPN49 (ST11, serotype K47, harboring bla_NDM-1) was employed as the low-virulence control (L-control). The clinical and microbiological characteristics of the two reference strains are listed in Supplemental Data Table S1. For the serum killing assay, human blood was collected from five healthy volunteers, and the serum was then separated and stored at −80℃. The assays were performed and the results were expressed as previously described [21^,24]. For the G. mellonella infection model, different concentrations of bacterial suspensions were diluted in 10 mM phosphate-buffered saline (PBS) and inoculated into the hemocoel of 20 larvae of G. mellonella (weighing 300±25 mg) (Kaide Ruixin Co., Ltd., Tianjin, China) at final concentrations of 1×10³−10⁷ colony-forming units (CFUs)/larva. Following injection, the larvae were incubated at 37℃ and monitored every 12 hours for 72 hours. Experiments were performed three times on separate occasions. The survival rates were recorded, and the 50% lethal dose (LD₅₀) was determined as described previously [25]. Negative control groups included larvae that were inoculated with 10 µL of PBS.

The clinical information of patients infected with CR-HMKP was collected from electronic medical records. All patients were evaluated based on the US Centers for Disease Control and Prevention (CDC) criteria to determine whether the infections were due to CR-HMKP [26]. This retrospective study was approved by the ethical committee of the Second Hospital of Anhui Medical University, with waiver of informed consent (approval number PJ-YX2018-001).

The Shapiro-Wilk method was used to test normality. Normally distributed variables were summarized as mean±SD, and non-normally distributed variables were summarized as median and range. The serum killing assays data were summarized as mean±SE. Survival data were plotted using the Kaplan-Meier method and analyzed using log-rank tests with GraphPad Prism 7.0 (GraphPad Software, La Jolla, CA, USA). The LD50 values were calculated using the Bliss method, and the results are expressed as log10 (lg) transformed values. A one-sample t-test was used to compare the data and was performed using SPSS v. 21.0 (IBM Corp., Armonk, NY, USA). A two-sided P<0.05 was considered statistically significant.

Of the 106 non-duplicated CRKP isolates, 12.3% (13/106) were positive for the string test and identified as CR-HMKP. The patients with CR-HMKP isolates were distributed across several departments, and their median length of hospitalization was 36 days (range 6–174 days). Notably, 76.9% (10/13) of the patients had a history of a stay in the intensive care unit (ICU). Of these patients, seven (53.8%) were males, and their mean age was 55.3±18.6 years. All but one patient received invasive treatment prior to infection with CR-HMKP. Eleven patients exhibited a high fever symptom following infection with CR-HMKP; four of these patients subsequently developed septic shock and eventually died because of failure of the anti-infection treatment (Table 1). Notably, seven cases occurred between July and August 2016, and all these cases were assessed as hospital-acquired (HA) infections. Five of the patients had an ICU stay history in July 2016, and their inpatient bed positions were very close, while the other two patients were in adjacent beds during hospitalizations in the Department of Plastic Surgery (PS) in August 2016.

The CR-HMKP isolates were primarily isolated from secretion/pus (N=6, 46.2%), followed by sputum (N=4, 30.8%), blood (N=2, 15.4%), and puncture fluid (N=1, 7.7%). All were resistant to imipenem, ertapenem, and meropenem but were susceptible to tigecycline and colistin (Table 2). The carbapenemase phenotype tests revealed that seven isolates were DDST-positive, and the others were MHT-positive. Consistent with these phenotypes, seven isolates harbored bla_NDM-1, while the remaining isolates harbored bla_KPC-2. Additionally, half of the bla_KPC-2-positive isolates co-harbored bla_CTX-M-65 and bla_SHV-11, while all bla_NDM-1-positive isolates co-harbored bla_SHV-71 (Fig. 1).

Five PTs and two STs were identified among the CR-HMKP isolates. All bla_NDM-1-positive isolates belonged to PT5 and were assigned to ST1764 (gapA_5, infB_3, mdh_1, pgi_1, phoE_9, rpoB_4, tonB_283), which is a three-locus variant of ST23 (gapA_2, infB_1, mdh_1, pgi_1, phoE_9, rpoB_4, tonB_12). The bla_KPC-2-positive isolates belonged to ST11 and were divided into four PTs. Nine of the isolates (69.2%) belonged to capsular serotype K64 (wzi-64) and four (30.8%) to capsular serotype K47 (wzi-209). Of the virulence genes tested, entB, iucA, kfuBC, mrkD, rmpA2, ureA, wabG, ybtS, and ycfM were detected in all CR-HMKP isolates, and the detection rates for aerobactin, allS, iroB, and rmpA were 53.8% (7/13), 46.2% (6/13), 53.8% (7/13), and 69.2% (9/13), respectively. The magA gene was not detected in any of the isolates. Notably, all NDM-1-producing isolates were positive for the aerobactin gene (Fig. 1).

Five representative CR-HMKP isolates [KPN18(PT1), KPN11(PT2), KPN100(PT3), KPN34(PT4), and KPN53(PT5)] were selected based on PT, and their virulence phenotypes were established. The serum killing assay results demonstrated that KPN53 was highly resistant (Grade 5), KPN11 and KPN18 were intermediately sensitive (Grade 4), and KPN34 and KPN100 were serum-sensitive (Grade 1). Of the two reference strains, KPN54798^H-control was highly resistant (Grade 6), and KPN49^L-control was serum-sensitive (Grade 1; Fig. 2).

In the G. mellonella infection model, time- and dose-dependent deaths were observed for all tested isolates. At the dose of 1×10⁷ CFU/larvae, 100.0% mortality was observed in all tested isolates except for the KPN49^L-control. At the dose of 1×10⁶ CFU/larvae, only KPN53 and KPN54798^H-control resulted in 100.0% larval mortality at 24 hours post-infection, whereas KPN11, KPN18, KPN34, KPN100, and KPN49^L-control killed 70.0%, 65.0%, 75.0%, 60.0%, and 40.0% of the larvae at this dose, respectively. The observed larval lethality rates due to KPN53 were higher than those of the other CR-HMKP isolates and KPN49^L-control (P<0.01 by the log-rank test) (Fig. 3). Additionally, KPN53 had a lower lgLD₅₀ than the other isolates except KPN54798^H-control (P<0.05 by the one-sample t-test) (Table 3).