Klebsiella pneumoniae Isolates from Meningitis: Epidemiology, Virulence and Antibiotic Resistance

Klebsiella pneumoniae (KP) resistance to broad-spectrum cephalosporin (BSC) in meningitis is important because of limited therapeutic options. To investigate the antibiotic resistance, virulence and epidemiology of KP in meningitis, we conducted a retrospective study for 33 non-metastatic isolates, including primary meningitis (n = 20) and post-craniotomy meningitis (n = 13) collected from 1999 to 2013. BSC resistance was found in 9 (27.3%) isolates, all from post-craniotomy meningitis, harboring bla SHV-5 (n = 6), bla CMY-2 (n = 2), bla DHA-1 (n = 2), and bla TEM-1B (n = 1). Positive virulence factors were hypermucoviscosity (n = 22), larger bacterial size (n = 24), virulent capsule serotypes (n = 24, K2, 11; K1, 5; K57, 3; K5, 2; K20, 2 and K54, 1), rmpA (n = 23), rmpA 2 (n = 20), aerobactin gene (n = 22) and high-grade serum resistance (n = 23, 69.7%). Higher mouse lethality (LD50 < 106) was found in 16 isolates (48.5%). Post-craniotomy isolates were significantly less virulent than primary meningitis isolates, except for similar serum resistance capability. The pulsotype and sequence typing (ST) results were diverse. A minor cluster with pulsotype C and ST23 (n = 5) was identified in primary meningitis isolates. In conclusion, virulence factors and BSC resistance corresponded to about 70% and 30% of KP meningitis isolates respectively. BSC remains appropriate for treating primary meningitis, whereas meropenem is indicated for post-craniotomy meningitis empirically.

PCR detection of virulence factors. Genomic DNA as templates, PCR was used to amplify the K capsule serotype-specific genes (including 6 liver abscess-associated capsule serotypes K1, K2, K5, K20, K54, and K57), c-rmpA, kfu and allS genes. Plasmid DNA as templates, PCR was used to amplify the rmpA, rmpA2 and aerobactin gene. Specific primers were used as previously described 19-23 . Phenotypic detection of virulence factors with HV phenotype and capsule size. The HV phenotype was defined positive as a viscous string of >5 mm of the colony on blood agar plate 24 . We measured bacterial size to represent capsule size of the bacteria. The microscopic images of bacterial size in short transverse diameter randomly counted from 100 bacterial cells per single isolate were analyzed by using the image analysis software cellSens standard version 1.8 (Olympus Optical Co. Ltd., Tokyo, Japan).
Phenotypic detection of virulence factor with serum resistance. The susceptibility of the K. pneumoniae isolates to human serum was analyzed as described previously 25 . In brief, twenty-five microliters of the bacterial suspension (about 2 × 10 6 CFU) were mixed with 75 μl of pooled normal human serum in microtiter plates and then incubated at 37 °C for a period of 3 h. The test was performed in triplicate and the number of recovered bacteria was determined and graded. Resistance grading was defined from grade 1 to grade 6, with grade 6 (viable counts at 1, 2 and 3 h >100% and increasing throughout the 3-h period) considered to be the most serum resistant. Grades 5 and 6 were regarded as high-grade serum resistance.
Mouse lethality experiment. Determination of the virulence of K. pneumoniae in mouse lethality tests and the median lethal dose (LD 50 ) was performed as described previously 22 . In brief, female BALB/c mice (6-7 weeks old) were obtained from the National Laboratory Animal Center (NLAC) (Tainan, Taiwan). Mice were maintained under standard conditions of temperature, light and feeding according to NLAC guidelines and the Chi Mei Medical Center Animal Care and Use Committee approved-protocols (Permit Number: 100120771). Each dose was injected intraperitoneally with 0.1 mL of bacterial suspension into four mice. After 14 days, calculations were based on the number of survivors. The degree of virulence was read as highly virulent for an LD 50 of <10 3 CFU; moderate virulence for an LD 50 of 10 4 -10 5 CFU; low virulence for an LD 50 of 10 6 -10 7 CFU; and no virulence for an LD 50 of >10 8 CFU.

Molecular genotyping of isolates
Pulsed-field gel electrophoresis (PFGE). PFGE, as a major strain typing method, was used to confirm genetic relatedness of isolates as described by Pfaller et al. 26 . Whole chromosomal DNA in agarose was digested with XbaI (Bio-Rad Laboratories, CA., USA), and the restriction fragments were separated in a CHEF Mapper XA System (Bio-Rad Laboratories, CA., USA). Cluster analysis was performed by the Dice similarity coefficients and unweighted-pair group matching algorithm (UPGMA) with a tolerance of 1.0% and 1.0% optimization using the BioNumerics program. All bands had to match exactly to classify isolates as indistinguishable. Isolates were designated nontypeable if repeated attempts to prepare DNA failed.

Data collection.
The imaging studies and following main data relevant to patient characteristics were collected: gender, age, comorbidities, use of ventilator, CSF profiles, and random blood sugar obtained on the day of CSF collection. Outcome was described as in-hospital mortality within the same hospital episode.

MIC.
All 20 primary meningitis isolates were susceptible to cefotaxime and ceftazidime. Nine of 13 post-craniotomy meningitis isolates were resistant to cefotaxime and ceftazidime, but were all susceptible to meropenem, colistin and tigecycline. These 9 isolates were of multi-resistant phenotype. In contrast, all the 20 primary meningitis isolates were only resistant to ampicillin (Table 1).
ESBL and AmpC phenotypes. Resistance to cefotaxime and ceftazidime was found in 9 isolates. Among them, ESBL phenotype was detected in 7 isolates and AmpC phenotype was suspicious for 2 isolates (KP-004 and KP-042).
The HV phenotype was highly prevalent in the isolates with mean bacterial size of >2 mm (91.7%, 22/24), whereas none of the nine isolates with mean capsule size of <2 mm had the HV phenotype (p < 0.0001).
Serum resistance. In general, the distribution of isolates in the grading of serum resistance were 23 strains in grades 5 and 6; grade 4 (1 strain), grade 3 (3 strains); and grade 2 (6 strains) (Table 3). Unexpectedly, the prevalence of high serum resistance (grades 5 and 6) could not reach statistically significant difference between primary meningitis isolates and post-craniotomy isolates (16/20 vs. 7/13, p = 0.1393), which was distinct from the above-mentioned virulence factors between groups (Table 3).
MLST. From all isolates performing MLST, 6 strains were ST23 (including one ST23-like variant), 5 strains were ST373, 4 strains were ST65, 2 strains were ST86 (including one ST86-like variant), ST15, ST17 and ST218 respectively. Others were one each for ST29, ST54, ST107, ST185, ST268, ST592, ST661, ST1049, ST1544 and unidentified ST (strain KP-023). The association between genotyping, sequence type and K capsule serotypes of K. pneumoniae meningitis isolates were summarized in Fig. 1. Notably, the 5 primary meningitis isolates of pulsotype C (C1-C5) were capsule serotype K1, including ST23 in 4 strains and ST 23-like in 1 strain. Five ST373 strains belonged to pulsotypes J and P. Isolates of ST373, ST65 and ST86 belonged to capsule serotype K2. Two isolates of ST218 and 1 isolate of ST29 belonged to capsule serotype K57 and K54 respectively. Demographic data and clinical outcomes. The 33 K. pneumoniae isolates from CSF basically of primary meningitis or post-craniotomy meningitis were collected from 1999 to 2013. However, detailed clinical data could be retrieved for the 12 patients from 2009 to 2013 (Table 4). The reasons of craniotomy were mainly to remove hematoma of various causes. Five of six patients in this group underwent craniectomy and two patients developed brain abscess (Fig. 2). The main comorbidities of patients with primary meningitis were diabetes mellitus and alcoholic liver cirrhosis (Table 4). Two patients in this group developed brain abscess and two patients developed ventriculitis (Fig. 2). Only one patient died during the hospital episode. There were no statistically significant differences in the demographic data, CSF profiles and clinical outcomes between the two groups of K. pneumoniae meningitis.

Discussion
The study investigated the antibiotic resistance profiles, β-lactamase genes, various hypervirulence determinants, mouse lethality, as well as the epidemiology of genomic macrorestriction patterns and sequence types of   Table 3. Different virulence profiles between cerebrospinal fluid Klebsiella pneumoniae (KP) isolates from postcraniotomy and primary meningitis.   Table 4. Clinical profiles and outcome of Klebsiella pneumoniae isolates from CSF of post-craniotomy and primary meningitis. Note CSF, cerebrospinal fluid; M, male; F, female; SAH, subarachnoid hemorrhage related to a brain aneurysm; ICH, intracerebral hemorrhage; AVM, arteriovenous malformation; DM, diabetes mellitus; ALC, alcoholic liver cirrhosis; RBC, red blood cell count (normal, <5/μl); WBC, white blood cell count (normal, <5/μl); NA, not applicable; a from Chi-Mei Medical Center in Tainan City, southern Taiwan, 2009-2013 (n = 12); b random blood sugar obtained on the day of CSF collection.
33 K. pneumoniae isolates that were capable of invading the central nervous system to cause primary meningitis (n = 20) or post-craniotomy meningitis (n = 13). Generally consistent with our hypothesis, primary meningitis isolates were more susceptible to third-and fourth-generation cephalosporins, were more hypervirulent (hvKP) and had higher mouse lethality capability than post-craniotomy meningitis isolates (mostly cKP). Exceptionally, serum resistance did not reach statistically significant difference between groups, suggesting that high-grade serum resistance (n = 23) is the major common virulence determinant of K. pneumoniae isolates with capability to invade the central nervous system, regardless of primary or post-craniotomy meningitis. The post-craniotomy isolates had significantly higher carriage rate of ESBL/AmpC genes (69.2%) than primary meningitis isolates did (0%), suggesting that it is appropriate for meropenem but not for broad-spectrum cephalosporin in the empirical therapy of post-craniotomy meningitis. Chang et al. reported data of 49 adult K. pneumoniae meningitis cases (36 spontaneous meningitis and 13 post-neurosurgical meningitis), collected over a period of 11 years (2000-2010) in Taiwan 28 . Data on antibiotic resistance of these isolates also showed none of 36 spontaneous meningitis isolates but only 2 of 13 (15.4%) post-neurosurgical meningitis isolates were resistant to third-and fourth-generation cephalosporins and one of the strains was ESBL-producing 28 . Thus, it is important to monitor the trend of antibiotic resistance in K. pneumoniae CSF strains for the empirical antibiotic strategy of meningitis.
With regard to specific virulence factors, primary meningitis isolates were more virulent than post-craniotomy isolates in term of the expression of larger bacterial size, HV phenotype, rmpA, rmpA2, aerobactin gene, virulent capsule K serotypes (K1, K2, K5, K20, K54, and K57), and high mouse lethality (LD 50 < 10 6 ). Capsular K2 serotype was the most common serotype observed (33.3%). Most hvKP isolates were significantly associated with the above-mentioned virulence factors and exhibited high-grade serum resistance as well as high mouse lethality. Furthermore, aerobactin is one of the siderophore systems, which mediate acquisition of iron to help virulent bacteria to overcome iron starvation, while bacteria invade and proliferate in the human systems 14, 24, 25 . In general, the current data support our initial hypothesis. The presence of these virulent genes and characters of primary meningitis hvKP isolates might assist in their capability to invade CSF space through intact central nervous system barrier to cause infection, whereas the less virulent cKP might still be able to cross the CSF space through disruption of central nervous system barrier, probably with the aid of the mechanism of serum resistance. We found that high-grade serum resistance occurred in 7 of 13 (53.8%) post-craniotomy isolates and in 16 of 20 (80%) in primary meningitis isolates, without reaching significant difference (p = 0.139). We acknowledge that high serum resistance is the major common virulence determinant of hvKP and cKP isolates contributing to invade the central nervous system, regardless of primary or post-craniotomy status.
With regard to molecular epidemiology, no major genomic clone or sequence type broadly existed among the 33 K. pneumoniae CSF isolates. However, there were 5 primary meningitis isolates of pulsotype C (C1-C5) belonging to ST 23 or ST23-like sequence type. The ST23 has been recognized as the most prevalent MLST type of capsule serotype K1 K. pneumoniae isolates from liver abscess in Taiwan 29 . The ST65, ST86, ST373 and ST375 (not detected in the study) have been the major clones associated with K2 K. pneumoniae liver abscesses in Taiwan 29,30 . In addition, ST65, ST65-like and ST86-like MLST types were most predominant among K2 serotype isolates of the community-acquired infection cases from Singapore, Hong Kong and China 31,32 .
ST218 has been reported in capsule serotype K57 liver abscess isolates in Taiwan 30 . ST29 K54 ESBL-producing K. pneumoniae with HV phenotype was detected in Hunan, China 33 . We previously reported on a mycotic aneurysm caused by hypermucoviscous ST29 K54 K. pneumoniae in Taiwan 34 . Furthermore, the ST29 was the predominant ST of carbapenem-resistant K. pneumoniae isolates in Saudi Arabia 35 . The ST15, ST17, ST54, and ST107 isolates producing ESBL, DHA-1, KPC-2 or NDM-1metallo-β-lactamase have been found in Taiwan, China and other various geographic areas in the world [36][37][38][39][40][41][42] . Importantly, 5 of 9 ESBL/AmpC producers in our current study belonged to these international clones, which often exhibited high-grade serum resistance. In addition to our recent report on the bacteremic isolates 43 , we continuously highlight the importance of monitoring the emergence of hypervirulence in the ESBL/AmpC-producing K. pneumoniae, particularly in the CSF isolates.
Although the K. pneumoniae isolates had different virulence mechanisms between the groups of primary and post-craniotomy meningitis, they did not result in significantly different clinical features and outcomes among limited patients between groups. The reasons could be explained by the effectiveness of BSC and meropenem against the isolates causing primary and post-craniotomy meningitis respectively. However, continuous monitoring for the emerging resistance profiles of K. pneumoniae between different groups of meningitis is clinically important.

Conclusion
Primary K. pneumoniae meningitis isolates had hypervirulence profiles of virulent capsule K serotypes, larger capsule size, HV phenotype, rmpA, rmpA2, aerobactin gene, high serum resistance and high capability of mouse lethality. Post-craniotomy K. pneumoniae meningitis isolates had relatively low virulence profiles and exhibited low mouse lethality, but still had high serum resistance which supported their capabilities to invade the central nervous system. The MLST international clones (ST15, ST17, ST54, and ST107) have been found in the post-craniotomy meningitis isolates in Taiwan, which could produce ESBL/AmpC β-lactamases. Therefore, physicians should be aware the emerging trend of antibiotic resistance in the empirical treatment for K. pneumoniae, particularly in post-craniotomy meningitis.