Molecular and Clinical Characterization of Multidrug-Resistant and Hypervirulent Klebsiella pneumoniae Strains from Liver Abscess in Taiwan

Hypervirulent Klebsiella pneumoniae strains are the major cause of liver abscesses throughout East Asia, and these strains are usually antibiotic susceptible. Recently, multidrug-resistant and hypervirulent (MDR-HV) K. pneumoniae strains have emerged due to hypervirulent strains acquiring antimicrobial resistance determinants or the transfer of a virulence plasmid into a classic MDR strain. In this study, we characterized the clinical and microbiological properties of K. pneumoniae liver abscess (KPLA) caused by MDR-HV strains in Taiwan.

ABSTRACT

Hypervirulent Klebsiella pneumoniae strains are the major cause of liver abscesses throughout East Asia, and these strains are usually antibiotic susceptible. Recently, multidrug-resistant and hypervirulent (MDR-HV) K. pneumoniae strains have emerged due to hypervirulent strains acquiring antimicrobial resistance determinants or the transfer of a virulence plasmid into a classic MDR strain. In this study, we characterized the clinical and microbiological properties of K. pneumoniae liver abscess (KPLA) caused by MDR-HV strains in Taiwan. Patients with community onset KPLA were retrospectively identified at Taipei Veterans General Hospital during January 2013 to May 2018. Antimicrobial resistance mechanisms, capsular types, and sequence types were determined. MDR-HV strains and their parental antimicrobial-susceptible strains further underwent whole-genome sequencing (WGS) and in vivo mice lethality tests. Thirteen MDR-HV strains were identified from a total of 218 KPLA episodes. MDR-HV strains resulted in similar outcomes to antimicrobial-susceptible strains. All MDR-HV strains were traditional hypervirulent clones carrying virulence capsular types. The major resistance mechanisms were the overexpression of efflux pumps and/or the acquisition of ESBL or AmpC β-lactamase genes. WGS revealed that two hypervirulent strains had evolved to an MDR phenotype due to mutation in the ramR gene and the acquisition of an SHV-12-bearing plasmid, respectively. Both these MDR-HV strains retained high virulence compared to their parental strains. The spread of MDR-HV K. pneumoniae strains in the community raises significant public concerns, and measures should be taken to prevent the further acquisition of carbapenemase and other resistance genes among these strains in order to avoid the occurrence of untreatable KPLA.

INTRODUCTION

Klebsiella pneumoniae liver abscess (KPLA) is endemic in Taiwan and some East Asian countries (1, 2), and cases are being increasingly reported worldwide (3). KPLA is usually caused by hypervirulent K. pneumoniae strains, which can also cause extrahepatic dissemination, such as ocular or central nervous system complications, associated with severe morbidity and poor long-term prognosis (1–4). Although a consensus phenotypic definition of hypervirulent K. pneumoniae strains is still lacking, the virulence traits of these strains have been linked to specific K. pneumoniae capsular polysaccharide types (K1, K2, K5, K16, K20, K54, and K57), plasmid-borne virulence genes (e.g., rmpA, regulator of the mucoid phenotype A), and hypermucoviscosity phenotypes (based upon positive string test results) (5, 6). On the host side, although diabetes is a typical risk factor for KPLA (7), these hypervirulent strains are able to cause severe infections in healthy hosts (3, 8).

Except for intrinsic resistance to ampicillin, hypervirulent K. pneumoniae strains are commonly susceptible to a variety of antibiotics, including cephalosporins and carbapenems (3); however, multidrug-resistant and hypervirulent (MDR-HV) strains have recently emerged (9). MDR-HV K. pneumoniae can arise by two different mechanisms (6). In one case, tentatively designated type I MDR-HV strains, hypervirulent K. pneumoniae strains acquire antimicrobial resistance genes or plasmids. For example, a capsular type K1 CTX-M-15 ESBL-producing K. pneumoniae strain causing a case of liver abscess in Taiwan (10), and two liver abscess cases caused by hypervirulent (capsular K1 and K20) and CTX-M type ESBL-producing K. pneumoniae strains in China have been reported (11). We also found a capsular type K2 K. pneumoniae with bla_CMY-2 causing a recurrent liver abscess in Taiwan (12). In the second mechanism, termed type II MDR-HV strains, resistance can arise from the transfer of a pLVPK-like virulence plasmid into a classic MDR K. pneumoniae strain. As an example, a recent study in China described a fatal outbreak caused by a KPC-producing ST11 strain acquiring a pLVPK-like virulence plasmid (13). We also reported an extensively drug-resistant and carbapenemase-producing hypervirulent ST11 K. pneumoniae strain occurring in our hospital recently (14). Patients infected by these MDR-HV strains commonly had severe outcomes (14, 15), but KPLA caused by MDR-HV has rarely been reported in the literature (10–12).

In this study, we characterized the clinical and microbiological properties of KPLA caused by MDR-HV K. pneumoniae strains in our hospital, and examined whether both types of MDR-HV strains are able to cause KPLA.

RESULTS

Comparison of clinical features in KPLA patients with MDR-HV versus antimicrobial-susceptible strains.

During the study period, 218 KPLA cases or patient episodes caused by hypervirulent strains were identified based on the enrollment criteria. Ninety-four episodes had bacteremia, including 30 episodes enrolled based on positive blood cultures with imaging of abscess formation. A total of 216 unique patients were identified in this study, and 2 patients had an initial episode caused by a hypervirulent strain and a breakthrough or recurrent episode caused by an MDR-HV strain (see below). All episodes were community onset infections. Seventeen patients (7.8%) were previously hospitalized in the last 6 to 12 months. Among the 218 KPLA cases or patient episodes, 200 antimicrobial-susceptible strains (resistant to ampicillin only), 13 MDR-HV K. pneumoniae strains, and 5 strains that were resistant to one or two antimicrobial categories were identified.

Table 1 shows the comparison of clinical features among the 13 cases infected with MDR-HV strains and 200 cases infected with antimicrobial-susceptible strains (resistant to ampicillin only). MDR-HV infections were associated with older age than antimicrobial-susceptible ones (mean age, 76.00 ± 9.40 years versus 64.38 ± 14.23 years; P = 0.004). In addition, MDR-HV strains were associated with a higher disease severity (APACHE II score) than antimicrobial-susceptible strains with borderline statistical significance (mean, 15.46 ± 7.30 versus 11.43 ± 7.20; P = 0.051). Only four patients had previous antibiotic exposure in the past month, including one case with an MDR-HV strain. Table 2 shows outcome comparisons between antimicrobial-susceptible and MDR-HV strains. Patients infected with MDR strains were less likely to receive appropriate antibiotics than those infected with antimicrobial-susceptible strains (84.6% versus 99.5%; P = 0.010). Other outcomes, including in-hospital mortality (7.7% versus 5.0%; P = 0.671) and infection-related mortality (6.3% versus 4.5%; P = 0.598), were similar between the two groups.

TABLE 1.

Clinical characteristics of KPLA patients infected with antimicrobial-susceptible and MDR-HV K. pneumoniae strains

Variabled	MDR strainsa	Antimicrobial-susceptible strainsb	P value
Demographics
Mean age in years ± SD	76.00 ± 9.40	64.38 ± 14.23	0.004
No. male (%)	8 (61.5)	124 (62.0)	0.974
Mean Charlson score ± SD	2.77 ± 2.32	2.28 ± 2.16	0.431
Comorbidity condition (no. [%])
Malignancy	2 (15.4)	21 (10.5)	0.582
Diabetes mellitus	5 (38.5)	93 (46.5)	0.573
Chronic kidney disease	2 (15.4)	14 (7.0)	0.266
Congestive heart failure	0 (0)	8 (4.0)	0.462
Liver cirrhosis	1 (7.7)	7 (3.5)	0.441
Cerebral vascular disease	0 (0)	14 (7.0)	0.324
Chronic obstructive lung disease	0 (0)	5 (2.5)	1.000
Collagen vascular disease	1 (7.7)	3 (1.5)	0.224
Originc (no. [%])			0.930
Cryptogenic	11 (84.6)	171 (85.5)
Biliary tract origin	2 (15.4)	29 (14.5)
Abscess location (no. [%])			0.840
Right lobe	9 (69.2)	135 (67.5)
Left lobe	3 (23.1)	38 (19.0)
Both lobes	1 (7.7)	27 (13.5)
Abscess size (no. [%])			0.280
<5 cm	3 (23.1)	76 (38.0)
≥5 cm	10 (76.9)	124 (62.0)
Gas-forming abscess (no. [%])	2 (15.4)	24 (12.0)	0.718
Multiple abscesses (no. [%])	1 (7.7)	44 (22.0)	0.221
Ocular and/or central nervous system infections (no. [%])	0 (0)	5 (2.5)	1.000
Capsular type K1 and K2 (no. [%])	11 (84.6)	150 (75.0)	0.434
Invasive procedures and devices at onset of infection (no. [%])
Central venous catheter	1 (7.7)	6 (3.0)	0.358
Nasogastric/Nasojejunal tube	1 (7.7)	8 (4.0)	0.521
Urinary catheter	0 (0)	13 (6.5)	0.343
Endotracheal tube	1 (7.7)	5 (2.5)	0.273
Surgical drainage	0 (0)	2 (1.0)	1.000
Prior any antibiotic exposure within one month (no. [%])	1 (7.7)	3 (1.5)	0.224
Mean APACHE II score ± SD	15.46 ± 7.30	11.43 ± 7.20	0.051
Clinical outcomes (no. [%])
Appropriate empirical antimicrobial therapy	11 (84.6)	199 (99.5)	0.010
Definitive treatment with cephalosporins	7 (53.8)	139 (69.5)	0.239
Definitive treatment with carbapenems	3 (23.1)	26 (13.0)	0.305
Intensive care unit admission	4 (30.8)	51 (25.5)	0.674
Hospital days, mean ± SD	26.62 ± 14.15	30.85 ± 30.43	0.619
Septic shock	3 (23.1)	44 (22.0)	0.928
In-hospital mortality	1 (7.7)	10 (5.0)	0.671
Infection related mortality	1 (7.7)	9 (4.5)	0.598

^an = 13.

^bn = 200.

^cCryptogenic KPLA was defined as those cases without obvious extrahepatic source of infection; biliary KPLA was defined when the clinical features of cholecystitis/cholangitis or extrahepatic biliary ductal abnormalities were identified by image studies.

^dSD, standard deviation; APACHE, acute physiology and chronic health evaluation.

TABLE 2.

Microbiological characteristics of MDR-HV K. pneumoniae strains

Strain	Capsular type	ST type	Antimicrobial resistance patterna	β-lactamase	Efflux pump overexpression and quinolone resistance mechanism
KP0175	K2	86	Non-susceptibility to CZ (MIC ≥ 64 μg/mL), CXM (MIC = 16 μg/mL), FOX (MIC = 32 μg/mL), CRO (MIC = 8 μg/mL), CAZ (MIC ≥ 16 μg/mL), and SAM (MIC ≥ 32 μg/mL)	SHV-1, CMY-2b
KP0430	K2	65	Non-susceptibility to CXM (MIC = 32 μg/mL), SAM (MIC = 16 μg/mL), and TGC (MIC = 4 μg/mL)	SHV-11	Overexpression of AcrAB efflux pump; missense mutations in ramR (G42V and I141T)
KP0971	K2	380	Non-susceptibility to CZ (MIC = 16 μg/mL), SAM (MIC ≥ 32 μg/mL), and TZP (MIC = 32 μg/mL)	SHV-1, TEM-1
KP1290	K5	1049	Non-susceptibility to CXM (MIC = 16 μg/mL); CIP (MIC = 1 μg/mL), and TGC (MIC = 2 μg/mL)	SHV-11	Overexpression of OqxAB efflux pump; missense mutations in oqxR (R5C)
KP1371	K2	86	Non-susceptibility to CXM (MIC ≥ 64 μg/mL), SAM (MIC ≥ 32 μg/mL), and TGC (MIC = 4 μg/mL)	SHV-1	Overexpression of AcrAB efflux pump; truncation of ramR
KP1562	K1	23	Non-susceptibility to CZ (MIC ≥ 64 μg/mL), CXM (MIC ≥ 64 μg/mL), CRO (MIC = 16 μg/mL), CAZ (MIC ≥ 64 μg/mL), CIP (MIC = 1 μg/mL), and GM (MIC ≥ 16 μg/mL)	SHV-5c	Presence of qnrS
KP1810	K16	660	Non-susceptibility to CXM (MIC = 16 μg/mL), FOX (MIC = 32 μg/mL) and TGC (MIC = 2 μg/mL)	SHV-1	Overexpression of AcrAB efflux pump; missense mutation in ramR (G180C)
KP1859	K1	23	Non-susceptibility to CXM (MIC = 16 μg/mL), CIP (MIC = 1 μg/mL), and TGC (MIC = 3 μg/mL)	SHV-11	Overexpression of OqxAB efflux pump; missense mutations in oqxR (Y80D)
KP2301	K5	1049	Non-susceptibility to CXM (MIC = 16 μg/mL), CIP (MIC = 1 μg/mL), and TGC (MIC = 2 μg/mL)	SHV-11	Overexpression of OqxAB efflux pump; missense mutations in oqxR (S82L, E118G and Q119G); and Ala insertion between amino acid positions 118 and 119
KP2371	K1	8	Non-susceptibility to CXM (MIC = 16 μg/mL), CIP (MIC = 1 μg/mL), and TGC (MIC = 3 μg/mL)	SHV-1	Overexpression of OqxAB efflux pump; putative mutation of OqxR-binding site of oqxAB promotor
KP2484	K1	23	Non-susceptibility to CZ (MIC ≥ 64 μg/mL), CXM (MIC = 16 μg/mL), FOX (MIC ≥ 64 μg/mL) and SAM (MIC ≥ 32 μg/mL)	SHV-11, DHA-1b
KP2485	K1	23	Non-susceptibility to CXM (MIC = 16 μg/mL), CIP (MIC ≥ 4 μg/mL), and TGC (MIC = 2 μg/mL)	SHV-11	Overexpression of OqxAB efflux pump; missense mutations in oqxR (L14F); amino acid substitution of GyrA in S83F; and the presence of qnrS
KP2611	K1	23	Non-susceptibility to CZ (MIC ≥ 64 μg/mL), CXM (MIC = 16 μg/mL), CRO (MIC ≥ 64 μg/mL), CAZ (MIC = 16 μg/mL), and CIP (MIC = 1 μg/mL)	SHV-11 SHV-12c	The presence of qnrS1 and aac(6’)-Ib

^aCZ, cefazolin; CXM, cefuroxime; FOX, cefoxitin; CRO, ceftriaxone; CAZ, ceftazidime; SAM, ampicillin-sulbactam; TZP, piperacillin-tazobactam; GM, gentamicin; TGC, tigecycline; SXT, trimethoprim-sulfamethoxazole; CIP, ciprofloxacin.

^bPlasmid-mediated AmpC β-lactamase.

^cExtended-spectrum β-lactamase.

Microbiological characteristics of K. pneumoniae strains.

Capsular type K1 (n = 117, 54.7%) and K2 (n = 47, 21.6%) were the most common capsular types among the 218 K. pneumoniae strains. Other capsular types included K5 (n = 13), K16 (n = 8), K20 (n = 3), K23 (n = 1), K54 (n = 8), K57 (n = 5), K62 (n = 5), K64 (n = 3), K68 (n = 1), and KN1 (n = 1), and 6 capsules were nontypeable. All K. pneumoniae strains showed hypermucoviscosity phenotypes and harbored rmpA/rmpA2 gene. We found that the in-hospital mortality in patients infected with non-K1/K2 strains was higher than that in patients infected with K1/K2 strains (11.1% versus 3.0%; P = 0.019). Other outcomes, including length of hospital stay and septic shock, were similar between the two groups. We identified 115 ST23, 27 ST65/375 and 9 ST86 clones using our previously described multiplex PCR assay, but no ST11 and ST258 clones were detected.

Table 3 shows the different resistance patterns and antimicrobial resistance mechanisms for the 13 MDR-HV strains. The results showed that all 13 strains had capsular types commonly associated with the hypervirulent K. pneumoniae background (type I strain), and the newly reported ST11 MDR-HV strains (type II strains) were not detected. The most common antibiotic resistance mechanism was the overexpression of efflux pumps, AcrAB/OqxAB (n = 8). High expression of AcrAB has been found to be associated with reduced susceptibility to tigecycline, while the overexpression of OqxAB could contribute to reduced susceptibility to ciprofloxacin. In addition, 2 strains had plasmid-mediated AmpC β-lactamase (bla_DHA-1 and bla_CMY-2) and two strains had ESBL genes (bla_SHV-5 and bla_SHV-12). One strain, KP0971, had a truncated OmpK35 protein, but the exact mechanism for β-lactam resistance was not clear.

TABLE 3.

Antimicrobial susceptibility testing data for the strains in this study

Strain	KP1762a	KP1810a	KP2329a	KP2611a	KP2611Sa	E. coli J53b	Transconjugantb
ST type	ST660	ST660	ST23	ST23	ST23
Antimicrobial resistance-related genes	blaSHV-1	blaSHV-1, ramR (G180C)	blaSHV-11	blaSHV-11, blaSHV-12, qnrS1 and aac(6’)-Ib	blaSHV-11, blaSHV-12, qnrS1 and aac(6’)-Ib		blaSHV-12, qnrS1 and aac(6’)-Ib
Antimicrobial	MIC (mg/liter)a
Cefazolin	≤4	≤4	≤4	≥64	≥64	≤4	≥64
Cefuroxime	2	16	2	16	16	4	32
Cefoxitin	2	32	3	3	3	NT	NT
Ceftriaxone	≤1	≤1	≤1	≥64	8	0.016	6
Ceftazidime	≤1	≤1	≤1	16	64	0.094	12
Cefepime	≤1	≤1	≤1	≤1	≤1	≤1	≤1
Piperacillin/tazobactam	≤4	16	≤4	≤4	8	≤4	≤4
Gentamicin	≤1	≤1	≤1	≤1	≤1	≤1	≤1
Amikacin	≤2	≤2	≤2	≤2	≤2	≤2	≤2
Ciprofloxacin	≤0.25	≤0.25	≤0.25	1	1	≤0.25	1
Ertapenem	≤0.5	≤0.5	≤0.5	≤0.5	≤0.5	≤0.5	≤0.5
Imipenem	≤0.25	≤0.25	≤0.25	≤0.25	≤0.25	≤0.25	≤0.25
Tigecycline	0.38	2	0.75	0.75	1	NT	NT
Trimethoprim-sulfamethoxazole	0.19	0.5	0.125	0.125	0.125	NT	NT

^aValues were MIC-correlates determined by the Vitek2 System, except for cefoxitin, trimethoprim-sulfamethoxazole, and tigecycline, which were determined by Etest.

^bValues were MIC-correlates determined by the Vitek2 System, except for ceftriaxone and ceftazidime, which were determined by Etest. NT, not tested.

Evolution of antimicrobial resistance in breakthrough KPLA.

One 84-year-old male had a biliary liver abscess caused by a capsular type K16 antimicrobial-susceptible K. pneumoniae strain (KP1762) in January 2016. He received intravenous cefmetazole (2 g every 8 h) and then cefuroxime (1,500 mg every 8 h) for a total of 26 days. He then continued on oral ciprofloxacin (500 mg twice daily), but an MDR-HV K. pneumoniae strain (KP1810) was isolated from a residual abscess 42 days after the isolation of KP1762 in February 2016. This patient was finally treated with ceftriaxone (2 g once daily) and subsequent cefmetazole (2 g every 8 h) for 6 weeks.

KP1762 and KP1810 showed an identical pulsed-field gel electrophoresis (PFGE) profile (data not shown), but KP1810 displayed increased MICs to cefuroxime, cefoxitin, piperacillin-tazobactam, and tigecycline (Table 3). WGS of KP1762 and KP1810 showed they were both ST660 type, with 16 core single nucleotide polymorphism (SNP) differences. Among them, we found that KP1810 harbors the missense mutation G180C in ramR compared to KP1762. The gene ramR is the negative regulator of ramA, which stimulates transcription of acrA/-B, encoding the multidrug efflux pump AcrAB-TolC. This mutation has been shown to be associated with reduced susceptibility to tigecycline in K. pneumoniae (16, 17). Reverse transcription-quantitative PCR (qRT-PCR) assays showed the overexpression of the AcrAB efflux pump (acrB) in KP1810 (5.9 ± 1.3-fold versus 1.4 ± 0.3-fold) and its regulatory gene ramA (18.5 ± 3.9-fold versus 1.7 ± 0.5-fold) compared to KP1762.

A murine model of septicemia generated by the intraperitoneal injection of K. pneumoniae was used to study the relative virulence of the KP1762 and KP1810 strains. The virulence of the two strains were comparable as the 50% lethal dose (LD₅₀) of MDR-HV strain KP1810 (3.8 × 10⁵ CFU) was similar to that of KP1762 (1.9 × 10⁶ CFU).

Evolution of antimicrobial resistance in a hypervirulent K. pneumoniae strain in recurrent KPLA.

Five patients had recurrent KPLA, and these strains were from capsular types K1 (n = 3) and K2 (n = 2). Among the five cases, only one patient had different antimicrobial resistance patterns between the initial and recurrent K. pneumoniae strain. This 73-year-old female with controlled diabetes suffered from the first episode of KPLA caused by an antimicrobial-susceptible capsular type K1 strain (KP2329) in April 2017, and was successfully treated with ceftriaxone (2 g once daily). She had the recurrent episode of KPLA caused by a MDR-HV capsular type K1 strain (KP2611) isolated from an abscess in February 2018, and was successfully treated with levofloxacin (750 mg once daily). WGS of both KP2329 and KP2611 showed 4 SNP differences between the 2 strains. KP2611 had an extra (compared to KP2329) IncN2-type plasmid (pKP2611-N, GenBank accession no. MN967025) harboring bla_SHV-12, aac(6′)-Ib and qnrS1. pKP2611-N is 60,970 bp in length, with a GC content of 51.1%, and harbors 71 open reading frames. Besides the antimicrobial resistance genes, pKP2611-N also contains a full complement of conjugative transfer genes (tra), which could explain its conjugability observed in the experiments described below. The presence of the ESBL gene bla_SHV-12 likely explains the resistance to ceftriaxone and ceftazidime, and the presence of qnrS1 and aac(6′)-Ib correlates with the increased MIC of ciprofloxacin (from <0.25 to 1 mg/liter). One K. pneumoniae strain (KP2611S) isolated from stool in this patient during the recurrent KPLA episode also belonged to capsular type K1 and carried bla_SHV-12. KP2329, KP2611, and KP2611S strains had an identical PFGE pattern (data not shown) and their antimicrobial susceptibility profiles are shown in Table 3.

The LD₅₀ values for the MDR-HV strain KP2611 and the antimicrobial-susceptible strain KP2329 were less than 500 CFU. Conjugation experiments showed that the IncN2 plasmid carrying bla_SHV-12 could be transferred successfully to recipient E. coli strain J53, with a frequency of 1.48 × 10⁻⁵/per donor. The MICs to ceftriaxone and ceftazidime comparing strain E. coli J53 to the transconjugant increased from 0.016 mg/liter and 0.094 mg/liter to 6 mg/liter and 12 mg/liter, respectively.

DISCUSSION

KPLA is a typical disease caused by antimicrobial-susceptible hypervirulent strains in East Asian countries. In the current study, 13 KPLA episodes (6.0%) were caused by hypervirulent K. pneumoniae isolates that also acquired multidrug resistance. In addition, patients infected with MDR-HV strains had similar outcomes to those with antimicrobial-susceptible strains, suggesting that the acquisition of multidrug resistance didn’t compromise the overall virulence, and consequently their associated disease burden in the community represents an emerging threat.

All 13 MDR-HV strains in the current study were from type I MDR-HV strains, and the major resistance mechanisms included the overexpression of efflux pumps and the acquisition of ESBL or AmpC β-lactamase genes (Table 3). They belonged to predominant clones associated with hypervirulent strains (ST23, ST65, and ST86) and common capsular types in the community, and, therefore, their evolution is associated with acquisition or genetic changes associated with resistance and distinct from the previously described ST11 carbapenem-resistant hypervirulent strains (13, 14). Although KPC-producing ST11 strains harboring pLVPK-like plasmids were previously reported in our hospital (14), further screening of all the KPLA strains did not find any ST11 or ST258 strains. To the best of our knowledge, ST11 carbapenem-resistant hypervirulent strains (type II MDR-HV strains) have not been found in community onset liver abscess, despite increasing reports of ST11 MDR-HV strains in East Asian countries (9). One possibility is that these type II MDR-HV strains have not been able to establish themselves in the community. Further, our previous study showed that although MDR-HV ST11 strain had increased virulence in the in vivo model, this strain had a higher LD₅₀ in comparison to the typical hypervirulent K1 or K2 strain (14). Thus, another explanation is that these type II MDR-HV strains may not be as virulent as the typical hypervirulent strains in causing KPLA. Nevertheless, it remains inconclusive whether the type II MDR-HV strains will cause life-threatening pyogenic infections in the community and, therefore, monitoring their spread would be a cautious and proactive public health action.

In this study, we identified a case with breakthrough KPLA caused by a capsular type K16 MDR-HV strain (KP1810), and showed it had similar virulence compared with its antimicrobial-susceptible parental strain (KP1762). The tigecycline resistance in KP1810 is likely the result of overexpression of the AcrAB efflux pump as a result of the upregulation of ramA, likely due to a mutation in the negative regulatory gene ramR. Unlike the acquisition of a plasmid bearing a resistance gene, we demonstrated that chromosome-encoded mutations can lead to the evolution of an MDR phenotype in hypervirulent strains. In addition, our in vivo animal model experiments demonstrated that acquisition of relevant mutations did not compromise virulence, which is consistent with our previous studies in tigecycline-resistant K. pneumoniae strains (16). Moreover, our observation that the overexpression of an efflux pump after exposure to β-lactams and fluoroquinolone gives further insight on how MDR phenotypes evolve in hypervirulent K. pneumoniae strains.

Recurrent KPLA has rarely been reported in the literature (17). We identified 5 cases with KPLA during this study period and one of them was caused by a capsular type K1 SHV-12-producing K. pneumoniae strain (KP2611). We also identified an SHV-12-producing hypervirulent K. pneumoniae strain (KP2611S) isolated from the patient’s stool. Notably, we found the virulence of KP2611 and KP2611S was not attenuated after the acquisition of the MDR phenotype compared to the susceptible strain (KP2329). We have previously shown that interspecies transfer of drug-resistance determinants into hypervirulent K. pneumoniae strains can occur in the intestine (12). Our current finding further supports the observation that the intestine could be the reservoir for the MDR-HV K. pneumoniae strains responsible for recurrent KPLA (12). Although none of the 13 MDR-HV strains were found to harbor carbapenemase genes, clinical K1 or K2 hypervirulent strains acquiring carbapenemase genes have been increasingly reported (9). Screening the intestinal carriage for MDR-HV strains may be of value in prevention for this disease.

Our study has some limitations. The MDR-HV case numbers in this study were relatively small and they were from a single center. However, the present study was conducted in a region of endemicity for KPLA and it represents one of the largest collections of MDR-HV K. pneumoniae in liver abscess disease in the literature. This is also one of the few studies to compare the clinical features of patients infected with and without antimicrobial-resistant strains in KPLA. Due to the nature of a retrospective study, there might be selection bias in this study. Finally, not all MDR-HV strains underwent in vivo virulence testing to evaluate their virulence. We selected two strains for in vivo virulence testing and compared the virulence with their susceptible counterparts, which showed that the virulence was not attenuated despite the evolution of resistance in the hypervirulent strains.

In conclusion, we first demonstrated the clinical and microbiological features of KPLA caused by MDR-HV strains in the community. These hypervirulent strains were able to evolve resistance and retain virulence, while classic MDR strains harboring virulent plasmids were not found to cause KPLA in this study. The spread of MDR-HV K. pneumoniae strains in the community raises significant public concerns, and measures should be taken to prevent the further acquisition of carbapenemase and other resistance genes among these strains in order to avoid the occurrence of untreatable KPLA.

MATERIALS AND METHODS

Study population and data collection.

Consecutive adult KPLA patients admitted to Taipei Veterans General Hospital (a 2,900-bed tertiary medical center) from January 2013 to May 2018 were retrospectively identified. Patients aged <20 years, with polymicrobial or nosocomial infections, or who developed a liver abscess after an invasive procedure, such as a transhepatic arterial chemoembolization or a stent replacement over a common bile duct, were excluded. Data regarding clinical characteristics, including underlying diseases, origin and nature of the liver abscesses, findings from imaging, and treatment outcomes were collected from medical charts. The study protocol was approved by the Institutional Review Board at Taipei Veterans General Hospital. The informed consent form was waived.

KPLA case definition.

An episode of KPLA was defined as a culture-confirmed K. pneumoniae isolated from an abscess or blood, plus the presence of a liver abscess identified by sonography or computed tomography. Recurrent KPLA was defined as a secondary liver abscess episode developed at least 3 months after the discontinuation of antibiotics from the first episode. Both the first episode of KPLA caused by a hypervirulent strain in an individual patient and the episodes of KPLA caused by MDR-HV K. pneumoniae strains were analyzed during the study period. An MDR K. pneumoniae strain is defined as nonsusceptible to at least one agent in three or more antimicrobial categories (18).

Bacterial identification and antimicrobial susceptibility testing.

Bacterial identification was determined using a Vitek2 System (bioMérieux, Marcy l’Etoile, France) and matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF) (bioMérieux SA, Marcy l’Etoile, France). Antimicrobial susceptibility was determined by the Vitek2 System (bioMérieux, Marcy l’Etoile, France) and interpreted following CLSI guidelines (19). The MIC of tigecycline was determined using the Etest method (bioMérieux) and interpreted by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) breakpoints (susceptible MIC, ≤1 mg/liter; intermediate MIC, 2 mg/liter; and resistant MIC, >2 mg/liter) in 2018.

Microbiological characteristics of K. pneumoniae strains.

Capsular genotyping, detection of the rmpA/rmpA2 genes, and colony mucoviscosity testing were performed as previously described (16, 20). PCR amplification and DNA sequencing were used to identify ESBL genes (encoding CTX-M, SHV, and TEM), plasmid-borne AmpC-like genes (encoding CMY and DHA), and mutations in ompK35 and ompK36 in each of the MDR-HV K. pneumoniae strains (16, 20, 21). For quinolone resistance, the presence of mutations in the quinolone resistance-determining regions (QRDR) of gyrA and parC, and PCR detection of plasmid-mediated quinolone resistance genes, including qnrA, qnrB, qnrC, qnrS, aac(6′)-Ib, and qepA, were performed as described previously (16). For tigecycline resistance, real-time quantitative reverse transcription PCR was used to evaluate the mRNA expression of pump genes, and the presence of mutations in the ramR, acrR, oqxR and rpsJ genes as previously described (16, 21). Hypervirulent K. pneumoniae strains were defined as those with a hypermucovisity phenotype and containing rmpA or rmpA2 (22). The information of two strains (KP0175 and KP1562) have been published before (12, 15).

Pulsed-field gel electrophoresis and multilocus sequence typing.

Pulsed-field gel electrophoresis (PFGE) was performed on selected MDR-HV strains and their parental strains as described in previous reports (14, 15). All strains were subjected to a multiplex PCR analysis for the rapid detection of common clonal types (ST23, 65, 86, 375, 11/258) (23). All MDR-HV strains were also characterized by multilocus sequence typing (MLST) using the method described previously (14), and the results were analyzed using the international K. pneumoniae MLST database at the Pasteur Institute (Paris, France) (24).

Whole-genome sequencing and analysis.

We aimed to study the evolution of antimicrobial resistance in hypervirulent strains. Therefore, two MDR-HV strains (KP1810 and KP2611) and their parental antimicrobial-sensitive counterparts (KP1762 and KP2329) were selected for whole-genome sequencing (WGS). The selected K. pneumoniae strains and the IncN2 plasmid pKP2611-N were sequenced using the Illumina MiSeq (Illumina, Inc., San Diego, CA) platform, followed by assembly using Spades 3.13.0 (25). Comparative genomic analysis between resistant and susceptible pairs were conducted using Mauve 2.4.0 (26). Core SNP analysis was conducted by mapping the raw reads of MDR strains to the respective antimicrobial-susceptible parental genome assemblies using the Burrows-Wheeler Aligner, and SNPs were called using SAMtools and VarScan, followed by filtering the SNPs in the repeated regions using VCFtools, as described previously (27). (See below for data availability.)

Virulence assessment.

MDR-HV K. pneumoniae strains and their antimicrobial counterparts were evaluated in a mouse lethality assay to determine the LD₅₀. In brief, female 6- to 8-week-old C57BL/6 mice were administered with an intraperitoneal injection of K. pneumoniae at various concentrations as previously described. Six mice were used for each bacterial concentration and the LD₅₀ was calculated using the Reed-Muench method (28). All animal care procedures and protocols were approved by the institutional animal care and use committee at the National Yang-Ming University.

Conjugation.

Plasmid conjugation was performed using the MDR-HV K. pneumoniae strain (KP2611) as the donor strain (ceftriaxone-resistant) and Escherichia coli J53 (sodium azide-resistant) as the recipient strain. The experiment was conducted following a previously described filter mating method (14).

Data analysis.

Data were analyzed using SPSS software (version 17; SPSS Inc., Chicago, IL, USA). Chi-square and Fisher’s exact tests were used to analyze categorical variables. Student’s t test and the Mann–Whitney rank sum test (Wilcoxon rank-sum test) were used to analyze continuous variables. Differences with a P value of <0.05 were considered statistically significant.

Data availability.

The whole-genome shotgun project has been deposited at DDBJ/ENA/GenBank under the accessions VSML00000000 to VSMO00000000. The sequence of pKP2611-N has been deposited at GenBank under the accession MN967025.