Virulence Factors and Molecular Epidemiology of Pyogenic Liver Abscess Causing Multidrug Resistant Klebsiella pneumoniae in Wenzhou, China

Background: To date, little is known about virulence characteristics of pyogenic liver abscess (PLA) causing multidrug resistant (MDR) Klebsiella pneumoniae (K. pneumoniae). It may be that these strains are rare. The aim of this study was to analyze the virulence characteristics and molecular epidemiology of 12 MDR strains from 163 PLA cases in a tertiary teaching hospital from the perspective of clinical characteristics, virulence phenotypes and genotypes. Results: The virulence phenotypes of the 12 PLA-causing MDR K. pneumoniae were similar or even more obvious than those of typical hypervirulent Klebsiella pneumoniae control strains according to the results of growth curves, string test, capsular quantication, serum killing test, biolm formation assay, and infection model. These MDR strains were mainly non-K1/K2 serotypes and carried multiple virulence genes. Multilocus sequence typing (MLST) illustrated that the MDR strains were categorized into nine sequence types. Conclusions: This study is the rst analysis of the virulence factors in PLA-causing MDR strains. Our data exhibited the coexistence of hypervirulence and multidrug resistance in PLA-causing MDR K. pneumoniae strains, and the clones of those PLA-causing MDR strains were diverse and scattered. The study was rstly found one ST11 carbapenem-resistant hypervirulent strain in PLA.


Serum killing test
All MDR strains and hypervirulent control strains isolated from KP-PLA were susceptible to serum. The anti-serum ability was no signi cant difference between the two groups (P > 0.05) (Figure 3).

Bio lm formation assay
As shown in Figure 4, the OD values of bio lms formed by MDR strains ranged from 0.31 to 0.80, with an average value of 0.58 ± 0.19; while the OD values of bio lms formed by hypervirulent control strains ranged from 0.06 to 0.39, with an average value of 0.27 ± 0.10. The bio lm formation ability of MDR strains was signi cantly higher than that of hypervirulent control strains (P < 0.05).
Infection model of Galleria mellonella larvae As shown in Figure 5, mortality of larvae depended on inoculation concentration and action time of the three MDR strains and three hypervirulent control strains (P < 0.05) ( Figure 5A, B, C, D, E, F). In addition, the lethality of MDR strains and hypervirulent control strains was similar when using 10 6 CFU/mL bacterial suspensions to infect larvae, but both were signi cantly higher than that of the standard strains ATCC 700603 and PBS controls (P < 0.05) ( Figure 5G).

Polymerase chain reaction for capsular serotypes and virulence genes
As shown in Figure 6, among 12 PLA-causing MDR strains, there were four of K1 serotype, one of K2 serotype, one of K20 serotype and six of non-type. Except for magA, iroN and kfuBC (ranging from 33.3% to 50.0%), all remaining virulence genes were presented in more than half of MDR strains (ranging from 75.0% to 100%). The prevalence of rmpA and aerobactin was 83.3% and 85.3%, respectively. Among 12 hypervirulent control strains, all were K1 or K2 serotype, and all virulence genes were presented in most strains (ranging from 75.0% to 100%) with the exception for iroN (33.3%). Additionally, the prevalence of numerous virulence genes in MDR strains was not signi cantly different with hypervirulent control strains.

Discussion
As well documented, MDR K. pneumoniae usually causes infections of patients with basic diseases and was considered as cKP with a high resistance rate but hypovirulence [2,12,14]. However, K. pneumoniae isolates from KP-PLA converged hypervirulence and high antibiotic resistance limited the clinical treatment options largely [16]. To date, little is known concerning virulence characteristics of PLA-causing MDR strains. Therefore, 12 MDR K. pneumoniae strains were collected from 163 KP-PLA cases and virulence and molecular epidemiology were further analysed. To the best of our knowledge, this study is the rst analysis of the virulence factors in PLA-causing MDR strains.
Numerous studies have reported that antibiotic resistance rates were low in KP-PLA [8,11,12]. Moreover, MDR strains were rare, and patients infected with them were more likely to accompany with hepatobiliary diseases compared to non-MDR ones (Table S1). Importantly, the uncontrollable infections and ineffective prognosis in patients with hepatobiliary diseases may be associated with recurrent bacteremia due to MDR bacteria, suggesting that these MDR isolates may not be the traditional cKP and the acquisition of MDR may not compromise the overall virulence which needs further veri cation. However, the actual virulence of these MDR strains has not been well evaluated so far.
The results of growth ability suggested there was no tness cost of the strains with resistant phenotype.
In addition, the hypermucoviscosity was considered as a surrogate marker for hvKP [5]. Here, we found the percentage of hypermucoviscous MDR strains was slightly lower than that of hypervirulent control strains. However, it may not suitable to consider hypermucoviscosity as the only indicator of hypervirulence, the polysaccharide capsule can protect K. pneumoniae from phagocytosis of immune cells and bactericidal action of complement or antimicrobial peptides and could act as a major virulence factor for hvKP [17]. Based on the data of capsular quanti cation, results showed the capsular content of PLA-causing MDR strains was higher than that of the standard strain and lower than that of the hypervirulent control strains, which was consistent with the results of string test. Although the MDR strains and hypervirulent control strains were all sensitive to serum, the antiserum killing ability of these PLA-causing strains was signi cantly higher than that of the hypovirulent standard strains, which may be related to the content of capsular polysaccharide. Furthermore, bacteria attach to the surface of host during the infectious process and are coated with polymers such as extracellular polysaccharides and DNA to form bio lms. The physical barrier formed by bio lms can protect bacteria from attacking by phagocytes and enzymes, which improves the bacterial defenses against host and resistance to antimicrobials [18]. It also appears that bio lm formation ability of MDR strains was signi cantly higher than that of hypervirulent control strains, which may be one of the reasons for the MDR strains to exhibit resistant phenotype. Moreover, G. mellonella larvae, as a good model of invertebrate host infection, has been applied to explore the virulence and pathogenicity of PLA-causing MDR K. pneumoniae strains [19]. The consistency between the clinical data and the results of the phenotypic assays supported the notion that the PLA-causing MDR K. pneumoniae strains were hypervirulent.
Analysis of virulence genotypes can further validate our hypothesis. K. pneumoniae strains are presented in at least 78 capsular serotypes, in which K1 and K2 are related to hvKP, as well as being pathogenic to humans strongly [17,20]. In the present study, K1 or K2 serotypes accounted for less than half of the PLAcausing MDR strains, while the hypervirulent control strains were all K1 or K2 serotypes. Although K1 or K2 serotypes can regulate the virulence of K. pneumoniae, hypervirulence is not unique to these capsular serotypes [21]. In addition, rmpA and aerobactin are the most important genes for hypervirulence [1]. rmpA regulates the synthesis of extracellular polysaccharide capsule to enhance virulence [22][23]. aerobactin is essential for the growth and virulence of K. pneumoniae via regulation of iron supply [1]. In the present study, the prevalence of rmpA and aerobactin in the MDR strains was slightly lower than that of hypervirulent control strains, re ecting that PLA-causing MDR strains may be combined with hypervirulence from the perspective of virulence genes. Importantly, wcaG, magA, and uge genes related to capsule synthesis were also prevalent in PLA-causing MDR strains [24][25]. Moreover, the high prevalence of siderophores genes ybtA, entB, and kfuBC in PLA-causing MDR strains suggested that the ability to uptake iron may be equivalent to that of hypervirulent control strains. Furthermore, almost all PLA-causing MDR strains carried mH (related to type 1 mbriae), mrkD (related to type 3 mbriae), and ureA (an α-subunit of the urease, associated with invasion) [24,26], genetically corroborated virulence phenotype results. Therefore, clinicians should be advised to pay more attention to the MDR strains, and also carefully select appropriate managements to treat KP-PLA to reduce bacterial adhesion and colonization.
MLST analysis uncovered the molecular epidemiology of PLA-causing MDR strains. The clones of these MDR strains were diverse and scattered, while the clones of hypervirulent control strains were all belong to hypervirulent clones, and ST23 was the predominated type consistenting with previous reports [27]. It has previously been common for ST11-type K pneumoniae to be resistant to carbapenems, but not hypervirulent. However, the new ST11-type strain that has emerged in recent years is simultaneously hypervirulent, multidrug resistant, and transmissible, and this kind of real superbug could pose a serious threat to public health [9,15]. Upon previous literatures, ST11 carbapenem-resistant hypervirulent strains have not been found in KP-PLA. To the best of our knowledge, this is the rst time that one ST11 carbapenem-resistant strain which may be MDR-hypervirulent K. pneumoniae has been described in KP-PLA. Importantly, further surveillance and implementation are needed to implement to control the dissemination in hospital settings and the community.

Conclusion
Combining the virulence phenotypes and genotypes, the convergence of hypervirulence and multidrug resistance in PLA-causing MDR K. pneumoniae strains was observed, which might lead to further emergence of a "post-antibiotic" scenario. Importantly, it reminded that clinicians should be highly prudent in prescribing antibiotics on such KP-PLA patients due to severe antibiotic resistance and take inspection measures timely in view of hypervirulence-induced invasive infections, and supervisors should implement stricter control measures to prevent such real superbug from further disseminating in patients and hospitals. Moreover, further research is needed to elucidate the mechanisms between host, pathogen, and host-pathogen interactions, which will lay a foundation to raise the awareness of MDR-hvKP and provide effective treatments for KP-PLA patients. Initial strains were isolated from sterile uid (including pus, blood, and drainage uid) of KP-PLA patients and identi ed as K. pneumoniae by matrix-assisted laser desorption/ionization time-of-ight mass spectrometry (MALDI-TOF/MS; bioMérieux, Lyons, France). Antimicrobial susceptibility testing of K. pneumoniae isolates was conducted by bioMerieux VITEK-2 (BioMérieux, Marcy-l'Étoile, France) initially. Multidrug resistant strains were de ned as non-susceptible to three or more different antimicrobial categories [29]. A total of 12 MDR K. pneumoniae were found in 163 KP-PLA cases. Meanwhile, an equal number of antimicrobial-susceptible hypervirulent strains were selected as the experimental hypervirulent control strains (isolated from healthy, ambulatory patients with KP-PLA and carried both aerobactin and rmpA genes) and the standard strain ATCC 700603 as the hypovirulent standard strain [1,30].

Methods
The minimum inhibitory concentrations (MICs) of ampicillin, aztreonam, ceftriaxone, ceftazidime, cefepime, imipenem, cipro oxacin, levo oxacin, gentamicin, tobramycin, sulfamethoxazole/trimethoprim, nitrofurantoin and colistin were con rmed by the agar dilution method and microdilution broth method. The results were interpreted by the latest guidelines published by the Clinical and Laboratory Standards Institute (CLSI; Pittsburgh, PA, USA) and the European Committee on Antimicrobial Susceptibility Testing clinical breakpoints (http://www.eucast.org). Escherichia coli ATCC25922 and Pseudomonas aeruginosa ATCC27853 were served as the quality control strains.

Growth curves
The growth curves of 12 PLA-causing MDR K. pneumoniae isolates were measured by following previous methods [31]. In brief, overnight cultures of selected K. pneumoniae clinical isolates from KP-PLA and K. pneumoniae ATCC 700603 were diluted 1:100 by Luria-Bertani (LB) broth. The cultures were incubated at 37 °C with constant shaking at 200 rpm. Bacteria suspensions were collected at 0, 2, 4, 6, 8, 10, 12, 18, 24 h and the absorbance at 600 nm was determined. Each suspension was measured in triplicates and averages of absorbance values were used for analysis. The growth of PLA-causing MDR K. pneumoniae was evaluated by plotting the values of OD 600 against time.

String test and quanti cation of capsule
The bacterial colonies of K. pneumoniae strain on an agar plate were stretched by an inoculation loop. The string test was considered positive when the strain generates a viscous string with a length of >5 mm, and this strain was also considered hypermucoviscous [30].
Capsule was quanti ed as described previously with some modi cations [10,32]. Brie y, 500 μL of cultured bacteria suspensions were resuspended and adjusted to 10 8 CFU/mL, and 1.2 mL sodium tetraborate in sulfuric acid were added in the resuspensions that placed in ice bath and incubated for 5 min at 100°C, and then left on ice for 10 min. A 20 μL volume of 1.5 mg/mL m-hydroxyphenyl was then added and mixed. After a 5 min incubation at room temperature, the absorbance at 590 nm was measured. The glucuronic acid content was determined from a standard curve of glucuronic acid and expressed as μg/10 8 CFU. Results were presented as the mean of the data of three independent experiments.

Serum killing test
Serum bactericidal activity was measured using the method as described previously [6]. Bacteria suspensions in nutrient broth were collected during logarithmic phase and adjusted to 10 6 CFU/mL of concentration. 25 μL of bacteria suspension was added to 75 μL of pooled human sera in the tube. Tubes were shaken and incubated for 0, 1, 2, or 3 h. An aliquot of each bacterial suspension was removed at the designated time point and diluted corresponding fold by adding Mueller-Hinton broth, and then cultured to determine the number of viable bacteria after exposure to serum. Results were expressed as a percentage of the inoculum and graded, then a strain was considered serum resistant or serum sensitive according to the standards, and each strain was tested at least three times.
Bio lm formation assay The bio lm formation assay was measured using the method of Wilksch et al. [33]. Brie y, clinical isolates were grown to logarithmic phase in LB broth and diluted 1:100 with fresh LB broth. A total of 200 μL of each dilution were added to a 96-well polystyrene microtiter plate and blank controls were set at the same time, and per strain was set three duplicate wells. Then, the plate was incubated at 37℃ for 24 h. Planktonic cells were removed, and the wells were washed three times with sterile water, and then stained with 250 μL 0.1% crystal violet for 10 min and rinsed three times with sterile water. Stained bio lms were solubilized with 95% ethanol and quanti ed by measuring the OD 600 . Each sample was measured in triplicates, and the averages of absorbance values were used for analysis.

Infection model of Galleria mellonella larvae
The model of G. mellonella larvae was carried out on the three PLA-causing MDR isolates (FK3068, FK3228, FK4603) and three control isolates (FK3112, FK3837, FK3914) that were randomly selected and standard strain ATCC 700603 to investigate the virulence and pathogenicity of the strains [34][35]. A serial concentration gradient bacterium suspension of each strain (10 7 ,10 6 ,10 5 ,10 4 CFU/mL) was prepared in advance. Eight larvae weighing of 200 mg -250 mg were randomly selected for each strain and each concentration. A 10 μL of bacterial suspension was injected into the last left proleg by using a 25 μL Hamilton precision syringe. Larvae injected with 10 μL phosphate-buffered saline were used as control. And then, the insects were incubated at 37℃ in the dark and observed after 24 h, 48 h, 72 h and 96 h. Larvae were considered dead when they repeatedly failed to respond to physical stimuli.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and analyzed during the study are available from the corresponding author on reasonable request.