Carbapenem Nonsusceptible Klebsiella pneumoniae in Taiwan: Dissemination and Increasing Resistance of Carbapenemase Producers During 2012–2015

Before 2011, the prevalence rates of carbapenemase-producing Klebsiella pneumoniae (CPKP) among carbapenem nonsusceptible K. pneumoniae (CnSKP) isolates were below 10% in Taiwan. The study presents the dissemination and increased antimicrobial resistance of CPKP from January 2012 to August 2015, as shown by Taiwanese multicenter surveillance. Isolates with minimum inhibitory concentrations (MICs) of >1 μg/mL for imipenem or meropenem were collected, screened for various carbapenemase genes by PCR, and tested for antimicrobial susceptibility. Among 1,457 CnSKP isolates, 1,250 were collected from medical centers. The CnSKP prevalence in medical centers increased by 1.7-fold during the study. Among all CnSKP isolates, 457 were CPKP. The CPKP rate among CnSKP increased by 1.5-fold and reached 36.8% in 2015. The CPKP nonsusceptibility rate to aztreonam, fluoroquinolones, and aminoglycosides increased yearly. Six CPKP isolates carried dual carbapenemase genes. Three Ambler classes were identified in 451 isolates with a single carbapenemase: classes A (315 blaKPC-2, 2 blaKPC-3, 28 blaKPC-17, 2 blaKPC-34), B (26 blaIMP-8, 2 blaNDM-1, 36 blaVIM-1), and D (40 blaOXA-48). The blaOXA-48 rate among CPKP increased by 6-fold over three years. Most KPC and OXA-48 producers were ST11. CnSKP was increasingly prevalent, owing to CPKP dissemination. Additionally, CPKP became more resistant during the study period.


Results
Prevalence of CnSKP and Specimen Sources. A total of 1,457 CnSKP isolates were collected during the study period. Among all CnSKP, 1,250 (85.8%) were collected from medical centers, and 207 (14.2%) from regional hospitals. In three major regional hospitals, 137 isolates nonsusceptible to carbapenem were identified from 6,446 clinical K. pneumoniae isolates during the study period, and the average prevalence of CnSKP was 2.13%.
In all medical centers, the average prevalence of CnSKP was 1.10%, with 114,065 clinical isolates screened during the study period. The temporal trends in prevalence were analyzed by geographic region (Fig. 1). In the North region, the average prevalence was 1.40%, with no significant increase during the study period (p = 0.380). In the West region, the prevalence significantly increased (p < 0.001) 6-fold from 0.50% (14/2778) in 2012 to 3.12% (91/2917) in 2015, which was the highest yearly prevalence observed by region in Taiwan. The prevalence also increased significantly in the South and East regions (p < 0.001 and p = 0.040, respectively). Overall, the prevalence of CnSKP in Taiwan significantly increased 1.7-fold (0.87% in 2012, 1.46% in 2015; p < 0.001).
Temporal Trends of CPKP. Of the 1,457 CnSKP isolates, 457 (31.4%) were positive for genes encoding carbapenemase (CPKP), while no carbapenemase genes were detected in 1,000 isolates (non-CP-CnSKP). The temporal trends for the rate of CPKP among CnSKP were analyzed by geographic region (Fig. 1). In the West region, the yearly rate of CPKP significantly increased nearly 2-fold from 35.7% to 70.3% in four years (p = 0.002). The rates in the North, South and East regions also increased, but not significantly (p = 0.695, 0.447, 0.570; respectively). Overall, the rate of CPKP in Taiwan significantly increased 1.5-fold from 24.1% to 36.8% during the study period (p = 0.003). The rate of non-CP-CnSKP decreased from 75.9% to 63.2%. Antimicrobial Nonsusceptibility Rates. All CnSKP isolates were tested with β-lactam and non-β-lactam agents ( Table 2). For the β-lactams, most (>99%) of the CnSKP isolates were nonsusceptible to cephalosporins, except for cefepime (nonsusceptible rate, 92.2%). The rate of nonsusceptibility to cefepime was lowest among non-carbapenem β-lactams, and the rates to meropenem and doripenem were the lowest among all β-lactams.
For the non-β-lactams, the nonsusceptibility rates of CnSKP to the fluoroquinolones (91.0% to ciprofloxacin and 88.3% to levofloxacin) were similar to that of cefepime. Among all tested agents, the rate of nonsusceptibility to tigecycline was the lowest, followed by colistin and the aminoglycosides.
The nonsusceptibility rates of CPKP and non-CP-CnSKP isolates were analyzed and compared, and different patterns were observed (Table 2). To β-lactams, CPKP were highly nonsusceptible, with rates >99% to all cephalosporins and >98% to piperacillin-tazobactam and all carbapenems. Aztreonam showed the lowest rate among all β-lactams. While non-CP-CnSKP isolates showed similar rates to most cephalosporins and aztreonam, they were significantly less nonsusceptible to piperacillin-tazobactam, cefepime, and carbapenems than their counterparts. The most obvious difference between these two groups were the rates of nonsusceptibility to doripenem and meropenem, which for non-CP-CnSKP were only 71.6% and 71.2%, respectively.
Among the non-β-lactams, CPKP isolates were significantly more nonsusceptible to fluoroquinolones than non-CP-CnSKP. No significant difference was seen between the two groups in terms of nonsusceptibility to colistin and tigecycline.

Discussion
In this surveillance study, the prevalence of CnSKP in Taiwanese medical centers averaged 1.10% and increased 1.7-fold during the study period. This prevalence was much lower than that reported in the U.S. (6%), according to a report from the 2007-2009 SENTRY Antimicrobial Surveillance Program 10 . A similar trend of increased CnSKP prevalence during the same period in Taiwan was observed by the Taiwan Nosocomial Infections Surveillance  (TNIS) which only included those isolates collected from patients in intensive care units. The prevalence from the TNIS was indeed much higher than that in this study, in which we collected isolates from all patients, regardless they were inpatients or outpatients. The TNIS showed that the prevalence of CnSKP in intensive care units in Taiwanese medical centers increased nearly 1.5-fold from 15.7% in 2012 to 22.5% in 2015 22 . The geographic differences of CnSKP prevalence of medical centers found in this study were also reported by the TNIS, with the highest prevalence being 28.1% in the West region in 2015 22 . Geographic differences in CnSKP prevalence were also reported in the U.S. 23 . While an increase of 1.7-fold over four years in this study is not so impressive, the prevalence of CnSKP in the TNIS increased 15.7-fold from 1.8% in 2005 to 28.2% in 2017 22 . Additionally, this study showed that the rates of CPKP and non-CP-CnSKP among CnSKP increased and decreased annually, respectively. The increasing prevalence of CnSKP is regarded as a result of CPKP dissemination.  This study found that the majority of CnSKP in Taiwan were isolated from sputum and urine, which is compatible with other surveys 24,25 . Our study also showed that the incidence of bacteremia increased nearly 2-fold within the four years studied. A similar increase in the incidence of bacteremia caused by CnSKP was also observed in a large teaching hospital in northern Italy 26 . Although K. pneumoniae is the main cause of liver abscess in Taiwan, those isolates are usually community-acquired and rarely reported to be resistant 1 . However, liver abscess caused by carbapenem-resistant Klebsiella pneumoniae was reported in a liver transplant patient in Italy 27 . Moreover, 1.2% (2/165) of K. pneumoniae isolates causing liver abscess were resistant to carbapenem in a China hospital during 2010-2014 28 . The barrier between community-acquired liver abscess caused by susceptible K. pneumoniae and hospital-acquired infection by resistant isolates is becoming blurred as resistance is becoming increasingly prevalent.
The rate of CPKP among CnSKP increased annually in Taiwan. A similar trend was also observed in other surveys, which showed a rate of 3.1% between 2010 and 2012 24 that increased to 20.0% in 2013 25 . While the highest rate in Taiwan was 70.3% in the West region in 2015, rates as high as 89.9% (113/126) and 86.7% (65/75) of CnSKP isolates were identified as CPKP in the U.S. 10 and Romania 29 , respectively.
Different profiles of antimicrobial nonsusceptibility were found between CPKP and non-CP-CnSKP in this study, as CPKP was more nonsusceptible to β-lactams than non-CP-CnSKP. Few studies have compared the resistance profiles between these two groups of clinical isolates; however, a study on carbapenem-resistant Enterobacteriaceae demonstrated that carbapenemase-producing isolates were more resistant to meropenem than their non-carbapenemase-producing counterparts 33 .
Our study showed that CPKP was significantly increasingly nonsusceptible to fluoroquinolones and aminoglycosides, while non-CP-CnSKP showed greater nonsusceptibility to doripenem and meropenem, during the study period. Few studies have examined this trend specifically in these resistant bacteria; however, a similar trend for increasing resistance of general K. pneumoniae isolates was found in national surveillance studies from Taiwan 34 and Korea 35 .
The coproduction of dual carbapenemases by K. pneumoniae has been reported sporadically 36,37 . Many combinations have been identified, including KPC/MLB and OXA/MLB 7 ; however, the KPC-2/OXA-48 coproducing isolates found in this study have not been reported before to the best of our knowledge. Furthermore, these dual carbapenemase producers belonged to the epidemic ST11 clone. Because of their high resistance profile and transferability, these resistant organisms warrant intensive monitoring.
Since it was first reported in 2001 8 , KPC has spread worldwide 38 . This study showed that KPC outnumbered previous endemic MBL and became the most common carbapenemase in Taiwan soon after its arrival, spreading to all regions of this island. Such rapid dissemination of KPC-KP was also observed in Greece 39 and Italy 40 . We previously sequenced the complete Taiwanese plasmid encoding KPC-2 and analyzed its high transferability 41 . In this study, we identified one novel bla KPC variant, namely, bla KPC-34 , and two that had not been reported in Taiwan before 2012. The first KPC-3 producer was isolated in 2013 and was geographically and genetically linked to those in the U.S. 42 , and the emergence of a KPC-17 producer in Taiwan was also previously observed 25,43 . Similar to the U.S. 10 and China 11 , KPC was the predominant carbapenemase in Taiwan; however, most of the Taiwanese KPC-producers were ST11, which is similar to China 44 , but different from the U.S., where the dominant clone is ST258 45 . ST11 is a single-locus variant of ST258, indicating their close relationship 46 . However, our previous study showed that K. pneumoniae ST23 strains are strongly associated with liver abscess in Asian countries 47 . Our study results indicated that these hospital-acquired resistant isolates are quite different from their susceptible counterparts causing community-acquired liver abscess.
After its emergence in 2013, OXA-48 became the second most prevalent type of carbapenemase two years later. Four OXA-48 producers were isolated in 2013 from patients without a travel history abroad 48 . Most of the OXA-48 producers in this study were the endemic ST11 clone, while ST101 was the most prevalent clone in Europe and North Africa 49 . The OXA-48 ST11 clone caused an outbreak in Spain in 2009 47 ; however, the plasmids of the Taiwanese and Spanish clones belonged to different incompatibility groups 48,50 .
While this extensive surveillance study involved all regions in Taiwan, not all carbapenemase types were identified during the study period. In a previous study, an isolate producing KPC-16 was reported in the South region in 2014 43 . Additionally, all isolates were screened by PCR in this study, which is consistent with many previous studies 4, 24,25 . However, some emerging yet unidentified carbapenemases might be missed by this method. Thus, phenotypic methods, such as Carba NP, the carbapenem inactivation method (CIM), and the modified Hodge test (MHT), are mandatory to identify novel carbapenemases 51 . Moreover, these CnSKP will be assessed using virulence assay, as hypervirulent strains have been reported in China recently 52 .
In conclusion, the current CnSKP, and especially CPKP, scenario in Taiwan is ominous. New types of CPKP have emerged, and these resistant bacteria have become increasingly nonsusceptible to particular antimicrobials, especially aztreonam, fluoroquinolones, trimethoprim-sulfamethoxazole, gentamicin, amikacin, and colistin over time. This increase would make the treatment of infections caused by CnSKP more difficult. Close monitoring and enhanced infection control measures are mandatory to curb the further spread of these highly resistant organisms. these hospitals are regional hospitals, while 12 are medical centers that provide tertiary care; each of the included hospitals contains more than 1,000 beds. This study was approved by the Institutional Review Boards of par-  : 100-076). The IRBs waived the need for informed consents from source patients of the enrolled bacterial isolates because the isolates were obtained as part of routine hospital care procedures, and involved very minimal risk to the source patients; this waiver does not adversely affect the rights and welfare of the source patients.

Study
Nonduplicate K. pneumoniae isolates collected from various sites in adult patients were tested for susceptibility to carbapenems at the participating hospitals, as part of routine laboratory procedures. Preliminary isolates that were nonsusceptible to imipenem or/and meropenem were sent to a reference laboratory at the National Health Research Institutes, Miaoli, Taiwan. After species identification was confirmed with a VITEK 2 automated system (bioMérieux, Marcy l' Etoile, France), the isolates were stored at −70 °C until further testing. All experimental procedures were performed in accordance with specified guidelines for the use of studied isolates, and were approved by the Institutional Biosafety Committee of Chi Mei Medical Center.
Detection of Carbapenemase Genes. All verified CnSKP isolates were subjected to polymerase chain reaction (PCR) detection of genes encoding carbapenemases, including class A (bla KPC , bla GES , bla IMI , bla NMC , bla SME ), class B (bla IMP , bla VIM , bla NDM , bla GIM , bla SIM , bla SPM ), and class D (bla OXA-48 ), using the primers described previously 4 . The amplicons were further sequenced to identify the molecular type. Isolates positive for the carbapenemase gene were defined as CPKP, and those that were negative were defined as non-CP-CnSKP. The chi-square test was applied for categorical variables, and linear-by-linear association was used to analyze trends (chi-square for trend). A p value of < 0.05 was considered statistically significant.