A total of 250 unique K. pneumoniae bacteremia isolates were collected from patients at Shenzhen Nanshan People's Hospital, Shenzhen University and Huashan Hospital, Fudan University in China between January 2010 and August 2017. The strains were identified with a Phoenix 100 automated microbiology system (BD, Franklin Lakes, NJ, USA), and then after two subcultured generations re-identified with matrix-assisted laser desorption ionization-time of flight mass spectrometry (IVD MALDI Biotyper, Bruker, Bremen, Germany). All procedures involving human participants were performed in accordance with the ethical standards of Shenzhen University and Fudan University, and the 1964 Helsinki declaration and its later amendments. For this type of study, formal consent is not required.

Clinical data were collected for each case from an electronic medical records database. Bacteremia origin was determined, based on bacteriological sampling at the suspected origin and clinical examination reports (Picot-Guéraud et al., 2015). In cases that were unclear, a second physician was consulted. We sought to determine the infection acquisition setting in each case. If this could not be confirmed and bacteremia occurred >48 h after hospital admission, hospital-acquired infection was presumed (Picot-Guéraud et al., 2015). If the bacteremia began <48 h after admission and at least one of Friedman's criteria (i.e., intravenous chemotherapy or hemodialysis in the last 30 days; home intravenous therapy or wound care in the last 30 days; hospitalization for ≥2 days in the last 90 days; or residence in a long-term care facility) was met, then the bacteremia was considered to be healthcare associated (Culshaw et al., 2014; Picot-Guéraud et al., 2015). In all other cases, the bacteremia was considered to be community acquired.

The susceptibilities of the isolates to clinically relevant antibiotics (amikacin, cefotaxime, ceftazidime, cefepime, cefoperazone-sulbactam, chloramphenicol, ciprofloxacin, levofloxacin, gentamicin, piperacillin-tazobactam, tetracycline, imipenem, and meropenem) and extended-spectrum β-lactamase (ESBL) production were detected with a Phoenix 100 automated microbiology system (BD) (Saffert et al., 2011) or by the disk diffusion test (Chang et al., 2015). The minimal inhibitory concentrations (MICs) of imipenem, meropenem, tigecycline, and eravacycline were determined by the agar dilution method according to Clinical and Laboratory Standards Institute guidelines (Michail et al., 2013).

Strains with the hypermucoviscosity phenotype (as revealed by a positive string test result) were defined as hypervirulent K. pneumoniae; those with a negative result were defined as classic K. pneumoniae. A positive string test result was defined as the formation of a mucoviscous string of >5 mm on a bacteriology inoculation loop used to stretch a colony grown overnight on an agar plate at 37°C (Shon et al., 2013).

Bacterial DNA was extracted from isolates and purified with a DNeasy Blood and Tissue Kit (Qiagen China Co., Ltd, Shanghai, China) according to the manufacturer's protocol. MLST was conducted by the method of Diancourt et al. (2005). All primer sequences used in MLST are listed in Table S1. Briefly, seven housekeeping genes (gapA, infB, mdh, pgi, phoE, rpoB, and tonB) were PCR-amplified and sequenced from all isolates according to the K. pneumoniae MLST protocol (www.pasteur.fr/mlst). Alleles and sequence types (STs) were assigned by the MLST database (www.pasteur.fr/mlst/Kpneumoniae.html). K. pneumoniae CCs were identified by the eBURST v 3.0 program as described by Feil et al. (2004). CCs were defined as groups of two or more independent isolates that shared identical alleles at six or more loci; each complex was named after the putative founder ST (Wang et al., 2013).

K. pneumoniae BF was detected by the method of Wu et al. (2011). Briefly, 1 μl of an overnight culture was inoculated into 100 μl of fresh Luria-Bertani (LB) broth in each well of a 96-well polystyrene plate (Costar3599, Corning, New York, USA). After 5 h of static incubation at 37°C, bacteria were stained with 25 μl of 0.5% crystal violet for 20 min. The supernatant was discarded, and plates were washed three times with deionized water to remove unattached cells. The biofilm-bound dye was then eluted with 95% ethanol, and the optical density at 550 nm (OD₅₅₀) was determined. The NTUH-K2044 strain, which exhibits strong BF, served as a positive control (Wu et al., 2011). Each assay was performed in triplicate on at least three occasions.

K. pneumoniae virulence genes were detected by PCR. The extracted genomic DNA from planktonic isolates served as a template for the amplification of virulence genes and for determining capsular serotypes. All PCR primer sequences and corresponding references are listed in Table S2. PCR amplification was performed in a total volume of 50 μl, containing 2 × PCR Master Mix (Tiangen Biotech Beijing Co., Ltd, Beijing, China), 0.5 mM of each primer, and 1 μl template DNA. The cycling conditions were as follows: 95°C for 3 min, followed by 30 cycles at 95°C for 30 s, 49°C−58°C for 30 s, 72°C for 60 s, and a final 10-min extension step at 72°C. Each PCR set included a no-template control and a positive control [NTUH-K2044 strain for magA(K1) and wcaG]. Amplification products were analyzed by electrophoresis in 1.0% agarose gels.

We used the antisense RNA gene silencing technique to construct wcaG antisense RNA silencing K. pneumoniae strains (Nakashima et al., 2006). The wcaG antisense RNA, which included the predicted Shine-Dalgarno sequence plus ~100 bp following the start codon, was amplified by PCR. PCR fragments were purified and digested with HindIII and BamHI endonucleases and then inserted into the hairpin structure of plasmid pHN680 for gene silencing. Correct cloning was verified by PCR and sequencing. Verified loaded plasmids were introduced into three K. pneumoniae bacteremia isolates: Kp25, Kp57, and Kp63. All strains, plasmids, and primers used for RNA silencing are listed in Tables S3, S4. RNA silencing was induced with 1.0 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). BF was detected by crystal violet as mentioned above. Growth of planktonic bacteria was measured with an OD₆₀₀ assay. All assays were performed at least in triplicate.

The expression levels of the wcaG were determined by quantitative real-time PCR (qRT-PCR). The primers used for qRT-PCR are listed in Table S4. Briefly, the total bacterial RNA was extracted by using an RNeasy Mini Kit (Qiagen China Co., Ltd, Shanghai, China). cDNA was then synthesized using a PrimeScript RT Reagent Kit (TaKaRa, Dalian, China). Finally, qRT-PCR was performed with a SYBR Premix Ex Taq II Kit (TaKaRa, Dalian, China) on the Mastercycler ep realplex System (Eppendorf, Hamburg, Germany), with an initial incubation at 95°C for 2 min, followed by 40 cycles of 15 s at 95°C and 60 s at 60°C. Each reaction was carried out in triplicate.

For all samples, a housekeeping gene (16S rRNA gene, rrsE) was used to normalize the expression of wcaG. The threshold cycle (CT) numbers were confirmed by the detection system software, and the data were analyzed based on the 2^−ΔΔCt method. The expression levels of wcaG were determined and compared with those of wildtype strains (expression = 1).

The data, which are reported as means ± standard deviations (SDs), were analyzed with Student's t-test, one-way factorial analysis of variance, or the nonparametric Mann–Whitney U-test. Virulence factor association with BF was determined with a multivariate logistic regression model constructed based on the Wald statistic. Odds ratios (ORs) are reported with 95% confidence intervals (CIs). P < 0.05 were regarded as statistically significant. All data were analyzed in SPSS version 14.0 (Chicago, IL, USA).

OD₅₅₀ microplate readings of isolates after crystal violet staining ranged from 0.05 to 3.5. The median OD₅₅₀ value for NTUH-K2044 (control strain) was 2.180, a value indicative of strong BF as expected (Wu et al., 2011). As reported in Table 1, the average OD₅₅₀ value of the 250 examined K. pneumoniae bacteremia isolates was 1.23 ± 0.52. Biofilm-forming K. pneumoniae prevalence was similar between male and female patients and was unrelated to various coexisting conditions, but was greater among young adults under 40 years old than among seniors over 65 years old (P = 0.002, Table 1). Biofilm-forming K. pneumoniae prevalence was unrelated to infection acquisition source (community, hospital, intensive care unit, medical unit, or surgical unit), but was more common among isolates from bacteremia cases that persisted beyond 72 h despite treatment than from bacteremia cases that were cleared within 72 h of commencing treatment (P < 0.001, Table 1).

BF prevalence did not differ between antibiotic-resistant and -susceptible K. pneumoniae isolates for any of the tested antibiotic drugs (Table 2). MIC levels (MIC < 1 mg/L, 1 ≤ MIC mg/L < 4 or MIC ≥ 4 mg/L) for imipenem, meropenem, tigecycline, and eravacycline were not significantly related to BF among K. pneumoniae isolates (Table 2). The capacity for ESBL production was also unrelated to BF.

MLST of the 250 K. pneumoniae bacteremia strains yielded clear results for 187 isolates, including 65 different STs. The biofilm characteristics and numbers of isolates assigned to each ST are shown in Table S5. Because the STs widely variation among these K. pneumoniae isolates, CCs were analyzed based on STs in eBURST (Figure 1). The main CCs of the K. pneumoniae bacteremia isolates were CC23 and CC65. This study also identified ST11 as the main ST among the K. pneumoniae bacteremia isolates. Thus, further analysis of the BF characteristics of CC23, CC65, and ST11 isolates (Figure 2) indicated that BF tended to be stronger among CC23 isolates than among CC65 isolates (P < 0.001) or ST11 isolates (P < 0.001).

As reported in Table 3, analysis of PCR-amplified virulence factors showed that BF was more pronounced among magA(K1)-positive (+) isolates than among magA(K1)-negative (–) isolates. Additionally, BF was more pronounced among aero+, rmpA+, rmpA2+, allS+, wcaG+, and iutA+ isolates than among isolates that were negative (–) for these virulence factors. However, BF was not increased in hypermucoviscous or K2A(K2) serotype positive (+) isolates compared with isolates not expressing each of the two virulence factors.

In order to determine the independent contribution of each virulence factor to the K. pneumoniae bacteremia BF, multiple logistic regression analysis was conducted. As shown in Table 4, of eight independent variables [magA(K1), K2A(K2), wcaG, aero, rmpA, rmpA2, alls, and iutA] submitted to a logistic regression model constructed by a backward selection approach based on the Wald statistic (dependent variable: OD₅₅₀-value ≥ 1.25), only one factor, wcaG, was found to be an independent risk factor for BF of K. pneumoniae bacteremia isolates (OR 11.426, P < 0.001). Notably, magA(K1) was not associated with the BF of K. pneumoniae bacteremia isolates.

In order to confirm the role of wcaG in the BF of K. pneumoniae bacteremia, wcaG was silenced by antisense RNA. As Figure 3A indicates, the expression levels of wcaG decreased by half following IPTG induction. RNA silencing of wcaG expression with pHN680 reduced growth, as indexed by OD₆₀₀, of Kp25, Kp27, and Kp63 isolates relative to wildtype (no plasmid) controls, and the inducer IPTG (1.0 mM) had no apparent effect on the growth of the three pHN680-containing isolates (Figure 3B). Interestingly, the BF, as indexed by OD₅₅₀, of Kp25 and Kp63 strains was decreased following IPTG induction, relative to controls not exposed to IPTG (Figure 3C).

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.