Background

Klebsiella pneumoniae (K. pneumoniae) is a common bacterium that can cause various diseases in both immunocompromised and otherwise healthy individuals, such as pneumonia, bacteremia, urinary tract infection, and pyogenic liver abscess (PLA) [1]. PLA is a life-threatening disease that is frequently observed worldwide and, in particular, is endemic to East Asia, showing a morbidity rate of 15.45 per 100,000 person-years in 2011 and a mortality rate of 4.7–9.8% [2,3,4]. With respect to the causative agents, bacteria resulted in 75.0% of all types of liver abscess [5]; K. pneumoniae in particular accounted for 52.4–81.7% of bacteria that cause liver abscesses worldwide[4, 6,7,8].

Typically, K. pneumoniae strains that cause PLA were susceptible to antimicrobials, exception of intrinsic resistance [9, 10]. However, recent studies have revealed the emergence of carbapenem-resistant K. pneumoniae (CRKP) [4, 7, 11]. In the past three decades, cases of hypervirulent K. pneumoniae (HvKP), which is more virulent than classical K. pneumoniae (cKP), have been increasingly documented. HvKP could be differentiated from cKP via mouse or Galleria mellonella (G. mellonella) lethality tests [12, 13]. HvKP is more common in the Asian side of the Pacific Rim, but is emerging globally [14], causing a variety of invasive infections, such as lung abscess, PLA, and meningitis [15]. The hypercapsule of HvKP itself could mask the fimbriae and hamper conjugation. Nevertheless, the capsule of cKP is slim and often has an impaired immune response against exocellular mobile elements [16]. Thus, the increasing multidrug-resistant (MDR) HvKP evolves more often from cKP than HvKP because of the acquisition of mobile elements carrying virulence determinants, thereby resulting in nosocomial infections [12, 14]. Among the various strains, hypervirulent carbapenem-resistant K. pneumoniae (Hv-CRKP) has gained notoriety as a highly infectious pathogen due to an increase in the number of severe infections and the increasing scarcity of effective treatments, broadening the number of people susceptible to all types of infections [14, 17].

Our hospital once reported 45 K. pneumoniae strains giving rise to PLA, which were collected during 2008 and 2012 [9]. With the passage of 5–9 years, the traits of such strains change remarkably. Here, another 43 K. pneumoniae strains were analyzed for drug resistance, virulence genes, serotypes, sequence types (ST), pulsed-field gel electrophoresis (PFGE)-based phylogenetic analysis, mobile genetic element (MGE) types, and lethality. The role of Hv-CRKP in PLA is intensively discussed.

Methods

K. pneumoniae strains

All 43 K. pneumoniae strains were isolated from patients with PLA at Department of Infectious Diseases, the First Affiliated Hospital of Zhejiang University in 2017. The specimens included abscess, drainage, and puncture fluid. K. pneumoniae strains were confirmed using a matrix-assisted laser desorption/ionization time-of-flight mass spectrometry system (Bruker Daltonics Inc., Fremont, CA, USA). The strains were stored at − 80 °C prior to use. Standard strains K. pneumoniae ATCC 700603 and Escherichia coli ATCC 25922, purchased from the National Centre for Medical Culture Collection of China, were used as controls for strain identification and antimicrobial susceptibility testing (AST).

NTUH-K2044 (Accession number: AP006725.1) is a hypervirulent K. pneumoniae strain typed as K1 and was isolated from Department of Internal Medicine, National Taiwan University Hospital, Taipei, Taiwan [18]. HS11286 (Accession number: CP003200.1) is a hypovirulent K. pneumoniae typed as K47 and containing blaKPC and was isolated from Department of Laboratory Medicine, Huashan Hospital, Fudan University, Shanghai, China [19]. Strains NTUH-K2044 and HS11286 were used as controls for string test and G. mellonella lethality test.

All the strains were non-repetitive. All patients were diagnosed with PLA based on pathological and imaging evidences (B-mode ultrasonography and computed X-ray tomography).

Determination of hypermucoviscous phenotype

The hypermucoviscous phenotype was determined by “string test” as described previously [20]. Formation of a viscous string > 5 mm in length was considered as a positive phenotype.

AST analyses

AST for the 43 K. pneumoniae strains was performed using a bioMérieux VITEK-2 analyzer (bioMérieux Co., Marcy-Etoile, France) and the Kirby-Bauer (K-B) method. The GN337 card included the antibiotics ticarcillin/clavulanate, piperacillin-tazobactam, ceftazidime, cefepime, cefoperazone/tazobactam, aztreonam, imipenem, meropenem, amikacin, tobramycin, ciprofloxacin, levofloxacin, doxycycline, minocycline, tigecycline, chloramphenicol, and trimethoprim-sulfamethoxazole. The K-B method included ampicillin, cefazolin, cefuroxime, cefotaxime, gentamycin, nitrofurantoin, and fosfomycin. AST results were elucidated based on the latest guidelines by the Clinical and Laboratory Standards Institute (CLSI; Pittsburgh, PA, USA), and the latest breakpoint by the European Committee on Antimicrobial Susceptibility Testing (EUCAST; Basel, Switzerland; for tigecycline). MDR strains were characterized as strains non-susceptible to three or more antimicrobial classes [21].

Multilocus sequence typing (MLST)

DNA of all 43 K. pneumoniae strains was extracted using the QIAamp DNA mini kit (QIAGEN Co., Venlo, Netherlands). Seven housekeeping genes (gapA, infB, mdh, pgi, phoE, rpoB, and tonB) [22] were sequenced for STs of the 43 strains according to the K. pneumoniae MLST database given at the website (http://www.pasteur.fr/recherche/genopole/PF8/mlst/Kpneumoniae.html). The primers are shown in Additional file 1.

Polymerase chain reaction (PCR) for serotypes, drug-resistance, and virulence genes

Serotypes (K1, K2, K5, K20, K54, and K57) [23, 24], drug-resistance genes (blaKPC, blaKPC-2, blaCTX-M1, blaCTX-M2, blaCTX-M8, blaCTX-M9, blaOXA48, blaNDM, blaIMP, and blaSHV) and virulence genes (wzy-K1, wzx, wzc, allS, entB, irp2, ybtS, iroB, iroN, iucA, kfu, fimH, mrkD, wabG, uge, rmpA, rmpA2, c-rmpA, p-rmpA, p-rmpA2, terB, peg-344, peg-589 and peg-1631) [12, 25, 26] were all determined by regular PCR using an Applied Biosystems Veriti PCR system (ABI, San Ramon, CA, USA). The primers used are described in Additional file 1. Sequencing of wzi loci was also used to determine serotypes [27] by comparison with the database of Pasteur Institute (https://bigsdb.pasteur.fr/klebsiella/klebsiella.html).

Definitions of putative HvKP, cKP, Hv-CRKP, and carbapenem-resistant HvKP strains

Hypercapsule-associated genes (wzy-K1, c-rmpA, p-rmpA and p-rmpA2) and siderophore genes (entB, irp2, iroB, and iroN) were included for screening HvKP and cKP. HvKP and cKP were putatively defined as described previously [28]. CRKP was defined as K. pneumoniae strains that are non-susceptible to imipenem or meropenem. Hv-CRKP was defined as CRKP (cKP) that acquires key virulence genes that confer hypervirulence. Carbapenem-resistant HvKP was defined as HvKP (serotypes K1, K2, K5, K10, K20, K25, K27, and K57) that acquires carbapenem resistance.

MGE and PFGE analyses

MGE and PFGE analyses were both performed as the reference [29].

G. mellonella lethality test

G. mellonella larvae were used to determine the lethality of K. pneumoniae strains [30]. G. mellonella larvae, weighing approximately 300 mg, were purchased from Tianjin Huiyude Biotech Company, Tianjin, China. Mid-log phase cultures of K. pneumoniae strains were washed with phosphate-buffered saline and further adjusted to a concentration of 1 × 107 CFU/mL. Ten G. mellonella larvae were used per test. Survival analysis was done to compare the lethality of K. pneumoniae strains. All experiments were performed in triplicates.

Statistical analysis

GraphPad Prism 8 (GraphPad Software Inc., USA) was used to perform Chi-square test, Fisher’s exact test and survival analysis. The value of p < 0.05 was regarded as statistically significant.

Results

General information of 43 K. pneumoniae strains

Characteristics of the 43 K. pneumoniae strains were shown in Table 1. Among the K. pneumoniae strains, the positive rate of “string test” was 27.9% (12/43). Serotypes K1 and K2 accounted for a total of 69.8% of the cases (30/43), with K1 and K2 accounting for 46.5% (20/43) and 23.3% (10/43), respectively. Clonal group (CG) 23 was predominant, with a share of 34.9% (15/43). Types D and M dominated MGE types with ratios of 62.8% (27/43) and 18.6% (8/43), respectively.

Table 1 Characteristics of the 43 K. pneumoniae strains
Full size table

Drug-resistance of 43 K. pneumoniae strains

Among the 24 kinds of antibiotics, the ones that showed susceptibility rates of over 80.0% included cefuroxime, cefotaxime, gentamycin, ticarcillin/clavulanate, ceftazidime, cefoperazone/tazobactam, cefepime, aztreonam, imipenem, meropenem, amikacin, tobramycin, ciprofloxacin, levofloxacin, doxycycline, minocycline, tigecycline, chloramphenicol, and trimethoprim-sulfamethoxazole according to Additional file 1.

Prevalence of drug-resistance and virulence-related genes

As shown in Fig. 1a, blaKPC and blaKPC-2 were positive in strains H15, H36, H38, and H42, showing a rate of 9.3% (4/43); blaSHV was positive in all the strains except strain H1.

Fig. 1
figure 1figure 1

a Prevalence of ten drug-resistance genes; b Prevalence of twenty-four virulence-related genes. The presence of drug-resistance and virulence-related genes is represented by a black box, and the absence of others is represented by a white box. Ten bla genes for the 43 K. pneumoniae strains are shown in (a). Twenty-four virulence-related genes for the 43 K. pneumoniae strains are shown in (b)

Full size image

The detection rates of virulence genes varied remarkably: c-rmpA (7/43, 16.3%), allS (16/43, 37.2%), wzy-K1(20/43, 46.5%), wzc (20/43, 46.5%), wzx (20/43, 46.5%), peg-1631 (24/43, 55.8%), kfu (27/43, 62.8%), terB (30/43, 69.8%), irp2 (32/43, 74.4%), ybts (32/43, 74.4%), rmpA2 (32/43, 74.4%), p-rmpA2 (32/43, 74.4%), iroN (34/43, 79.1%), peg-589 (35/43, 81.4%), iucA (36/43, 83.7%), p-rmpA (36/43, 83.7%), iroB (37/43, 86.0%), rmpA (39/43, 90.7%), peg-344 (39/43, 90.7%), uge (40/43, 93.0%), entB (43/43, 100.0%), fimH (43/43, 100.0%), mrkD (43/43, 100.0%), and wabG (43/43, 100.0%). The following virulence genes represent certain siderophores: entB, enterobactin; irp2 and ybtS, yersiniabactin; iroB and iroN, salmochelin; iucA, aerobactin. In addition, fimH and mrkD represent type 1 and type 3 fimbriae, respectively, and wabG and uge represent lipopolysaccharides. As shown in Fig. 1b, wzx, wzc, and allS coincided well with wzy-K1 at rates of 100.0%, 100.0%, and 75.0%, respectively. The positive rates of entB and irp2 were different: 100% (43/43) vs. 74.4% (32/43) (p = 0.0005). The positive rate of putative HvKP was 95.3% (41/43), except for H18 and H25. As shown in Fig. 1a, b, strains H15, H36, H38, and H42 were all putative Hv-CRKP.

PFGE dendrograms

Figure 2 shows a total of 29 clusters, indicating highly divergent origins for the 43 K. pneumoniae strains. However, putative Hv-CRKP strains H36 and H38 presented the same background for both PFGE dendrogram and MGE type, showing that they belonged to the same clone. The other putative Hv-CRKP strains, H15 and H42, belonged to different clones. Therefore, all the four putative Hv-CRKP strains originated from three distinct clones.

Fig. 2
figure 2

PFGE dendrogram of the 43 K. pneumoniae strains. ST, sequence type; MGE, mobile genetic element. CG, clonal group. Genetic relationships among the 43 K. pneumoniae strains are shown in Fig. 2. In addition, the serotype, sequence type and mobile genetic element type of each strain are together indicated

Full size image

Clinical traits of four patients infected with putative Hv-CRKP

The demographic and clinical traits of the four patients who were infected with putative Hv-CRKP strains are shown in Table 2. Patient 15 was with several underlying conditions, no severe syndromes, had underwent surgery, and eventually survived. The other 3 patients were all diagnosed with several underlying diseases, had surgeries, and eventually died.

Table 2 Clinical characteristics of four patients infected with putative Hv-CRKP
Full size table

G. mellonella lethality test

The four putative Hv-CRKP strains (H15, H36, H38, and H42) were analyzed for their lethality using the G. mellonella model. Log-rank (Mantel-Cox) test showed significant differences among six groups: χ2 = 40.5688 and p < 0.0001 (Fig. 3). Figure 3 also shows no significant difference among NTUH-K2044, H38, and H42 (χ2 = 5.0659 and p = 0.0794), and HS11286 and H15 (χ2 = 2.1096 and p = 0.1464). However, H36 was significantly different from NTUH-K2044 (χ2 = 16.4627 and p < 0.0001) and HS11286 (χ2 = 8.2092 and p = 0.0042). The overall survival rate of G. mellonella injected with H36 was 40.0%. Therefore, H36, H38, and H42 were denoted as Hv-CRKP, and H15 was confirmed as cKP.

Fig. 3
figure 3

Survival curves of G. mellonella infected by K. pneumoniae. The percent survival of G. mellonella is shown in Fig. 3, which was injected with 0.1 mL of K. pneumoniae suspension at the concentration of 1 × 107 CFU/mL. CFU colony forming unit

Full size image

Discussion

We analyzed 43 K. pneumoniae strains that induced PLA, disclosed their molecular epidemiological status, and explored the emerging trend of Hv-CRKP strains in causing PLA. Serotypes K1 and K2, clonal group 23, and MGE types D and M predominated the 43 strains with rates of 69.8%, 34.9%, and 81.4%, respectively. According to the susceptibilities of the 43 strains, aminoglycosides, generation 2nd-4th cephalosporins, β-lactamase/β-lactamase inhibitors, carbapenems, and fluoroquinolones could still be appropriate, alternative, and empirical treatment choices. The PFGE dendrogram confirmed the highly divergent origins of the 43 strains. These findings were in line with previous reports [9, 11]. However, in comparison with data obtained in a previous study [9], the incidence of serotype K1 decreased (χ2 = 4.5186 and p = 0.0335) and that of serotype K2 was equal (χ2 = 0.1377 and p = 0.7106), indicating a new trend in PLA.

K. pneumoniae can harbor many factors, such as capsule, siderophore, exopolysaccharide, fimbriae, of which the first three could determine whether it is hypervirulent or not [1, 14, 18]. In this study, 24 virulence-associated genes were identified. As shown in Fig. 1b, wzx, wzc, and allS coincided well with wzy-K1 with rates of 100.0%, 100.0%, and 75.0%, respectively, indicating that these three genes are associated with serotype K1. Although enterobactin and yersiniabactin are “basic” siderophores for K. pneumoniae, the positivity rate of entB was higher than that of irp2 (p = 0.0005). There are several markers for HvKP, such as string test, rmpA, and peg-344 [14, 26]. The positive rate of string test among HvKP ranged from 27.9% to 90.7% based on different criteria. The poor positive rate of string test in this study declared its antiquation. There is an inevitable bias if only 1–3 genes are termed as markers of HvKP. Detection of c-rmpA in combination with p-rmpA equaled that of rmpA only: 41/43 vs. 39/43 (p = 0.6761). Intriguingly, c-rmpA was only present in serotype K1 of K. pneumoniae with positivity rates of 16.3% (7/43) in 43 strains and 35.0% (7/20) in K1 K. pneumoniae, indicating that rmpA is transposed into the chromosome of K1 K. pneumoniae more readily than other serotypes of K. pneumoniae. Although peg-344 is not an exact virulence gene [31], it served as a better indicator of HvKP than peg-1631: 39/43 vs. 24/43 (χ2 = 15.0938 and p = 0.0001) and equaled peg-589: 39/43 vs. 35/43 (χ2 = 1.5496 and p = 0.2132).

In this study, three strains (H36, H38, and H42) were confirmed to be Hv-CRKP with a positivity rate of 7.0% (3/43) due to blaKPC-2, which was the predominant blaKPC in China [32]. According to ST (ST11 and ST660) and MGE (A and G) types, these three strains belonged to two clusters, suggesting their different origins: ST11-K64 strain versus ST660-K16, which is different from what was observed in a previous study [8]. The rate of blaKPC-2-producing ST11 in Hv-CRKP was 33.3%, similar to that reported previously [8]. However, ST660 also shared a 66.7% rate, which was zero in the previous study [8]. K. pneumoniae strains in the previous study [8] were collected from 15 centers located in 11 Chinese cities, and the data in it reflected the general prevalence of Hv-CRKP that causes PLA in mainland China from 2012 to 2016. All three strains were MDR (Additional file 1) [21], which brought therapeutic challenges clinically. Although there are several methods for treating PLA, such as various drainage techniques, antimicrobials are still essential.

The first Hv-CRKP in mainland China emerged in 2013 [13]. Thereafter, the rate of Hv-CRKP was thought to increase gradually to 7.4%-15.0% [33]. Another five Hv-CRKP strains were reported in 2018 [12], which showed extremely high virulence and resulted in 100.0% deaths; they were the same clone and belonged to ST11. For virulence, they had three siderophores (enterobactin, yersiniabactin, and aerobactin) and rmpA2. In our study, H42 possessed four siderophores (enterobactin, yersiniabactin, salmochelin, and aerobactin), rmpA, and rmpA2, whereas H36 and H38 harbored three siderophores (enterobactin, yersiniabactin, and aerobactin) and rmpA2. H36, H38, and H42 also caused deaths, which is of great concern. Furthermore, the 3 deaths were the only ones in this study, indicating the important role of Hv-CRKP in PLA. Hv-CRKP, armed with its hypercapsule, could effectively resist the phagocytosis of leukocytes and enable systemic tissue invasion as a “Trojan horse”, resulting in thrombophlebitis, meningitis, etc. [14]. With the increasing incidence of metastatic K. pneumoniae meningitis, secondary to PLA, K. pneumoniae has become the leading pathogen of adult community-acquired bacterial meningitis instead of Streptococcus pneumoniae in Taiwan [34]. Due to extreme drug resistance and hypervirulence, Hv-CRKP may be a notable superbug in the future.

This study had some limitations. First, the sample size was small. Second, H36 and H38 showed different virulence, although PFGE confirmed the same origin. It may result from some slight differences of the genomes between H36 and H38 for virulence is the overall outcome of a series of virulence genes.

Taken together, we report the molecular characteristics of 43 different K. pneumoniae strains that caused PLA in 2017, which differed from the strains described in a previous study conducted between 2008 and 2012 in a tertiary hospital in East China. Our study highlights the imperative need to note the role of Hv-CRKP in PLA.