Molecular Epidemiology and Risk Factors of Carbapenem-Resistant Klebsiella pneumoniae Infections in Eastern China

Background: The increasing prevalence of carbapenem-resistant Klebsiella pneumoniae (CRKP) poses an immediate threat to treatment worldwide. This retrospective study assessed the molecular epidemiology and determined the risk factors for and outcomes of CRKP infections in a general teaching hospital in Shanghai, China. Methods: From January 2013 to July 2015, 100 consecutive unique CRKP isolates isolated from hospitalized patients were collected. Isolates were screened for antibiotic resistance genes by polymerase chain reaction and molecular typing was performed by pulsed-field gel electrophoresis (PFGE). Patients infected with CRKP comprised the case group and were compared to the control group of patients infected with carbapenem-susceptible Klebsiella pneumoniae. Therapeutic effects were compared in the CRKP infection group. Results: Among the 100 CRKP isolates, the percentages of multidrug-resistant, extensively drug-resistant (XDR), and pandrug-resistant were 50.0, 50.0, and 0%, respectively. All the CRKP isolates produced KPC-2 and could be divided into 18 PFGE clusters (A–O) and 70 subtypes. No dominant intra-hospital PFGE type was detected using a cutoff of 80% similarity. The ratio of CRKP infection to colonization was 51 to 49. Risk factors correlated with CRKP infection included pulmonary disease (p = 0.038), ICU stay (p = 0.002), invasive ventilation (p = 0.009), blood transfusion (p = 0.028), parenteral nutrition (p = 0.004), sputum suction (p = 0.006), medical history of previous hospitalization (p = 0.022), exposure to antibiotics 90 days before infection (p = 0.030), and antibiotic exposure during hospital stay including carbapenems (p = 0.013), enzyme inhibitors (p = 0.021), nitroimidazoles (p = 0.029), and glycopeptides (p = 0.000). Multivariable analysis showed that sputum suction (odds ratio 3.090, 95% confidence intervals 1.004–9.518, p = 0.049) was an independent risk factor for CRKP infections. Patients infected with CRKP with longer carbapenems treatment course (p = 0.002) showed better outcome. Conclusion: This study showed the severity of CRKP infection in eastern China. Sputum suction was an independent risk factor for CRKP infection. Prolonged duration of treatment with carbapenems benefited the patients infected with CRKP.


INTRODUCTION
Klebsiella pneumoniae (KP), one of the most commonly isolated Gram-negative bacterial pathogens in nosocomial infections, has posed a major threat to clinical and public health because of its ability to produce carbapenemases, which hydrolyze carbapenem antibiotics. Carbapenems constitute the last line of defense against infections caused by multidrugresistant (MDR) Gram-negative organisms. The first case of carbapenem-resistant Klebsiella pneumoniae (CRKP) was reported in MacKenzie et al. (1997), after which it spread worldwide and outbreaks were reported in several countries (Meybeck et al., 2014;Zweigner et al., 2014;Matsumura et al., 2015;Stillwell et al., 2015). In China, the prevalence of CRPK has rapidly increased from 2.9% in 2005 to 13.4% in 2014 (Hu F.-P. et al., 2016) and the distribution differed greatly by region, with the lowest prevalence in the northeast and the highest prevalence in the eastern region of China .
Carbapenem-resistant Klebsiella pneumoniae infections have a poor prognosis, with a mortality rate of approximately 40-50% (Borer et al., 2009;Hoxha et al., 2016). Inferior outcomes are especially significant in high-risk immunocompromised patients including those with solid organ transplants, in intensive care units, and hematological malignancies (Kalpoe et al., 2012;Satlin et al., 2014;Simkins et al., 2014;Vardakas et al., 2015). Bloodstream infections caused by CRKP increase the risks of treatment failure and death (Tumbarello et al., 2012). As colistin is one of the few available antibiotics against CRKP, the therapeutic options for infections are limited. However, with the appearance of colistin resistant strains, treatment of CRKP infections is increasingly difficult (Capone et al., 2013).
Resistance to carbapenems is associated with several mechanisms. Carbapenemases, which are largely responsible for resistance, are classified into three functional groups: class A serine carbapenemases (for example, KPC), class B metalloβ-lactamases (for example, NDM), and class D β-lactamases (for example, OXA) (Bush, 2013). Among the above types of carbapenemases, KPC-producing KP has spread widely throughout the world in the past few years. Other factors contribute less to resistance, including the production of β-lactamases with weak carbapenemase activity, especially class A extended-spectrum β-lactamases (ESBLs) and class C AmpC lactamases, permeability defects and loss of porins (Ompk35 and Ompk36) (Tsai et al., 2013;Pitout et al., 2015).
Recently, outbreaks of CRKP have become increasingly common in China (Wang et al., 2014;Zhou et al., 2015) and have been recognized as a tremendous challenge. High mortality rates and lack of effective treatments place the patients into a perilous situation after infection with CRKP. Identification of risk factors for CRKP infection would improve the choice and efficacy of empirical therapy (Kofteridis et al., 2014), which could assist in early recognition and timely intervention. In addition, the majority of studies on the risk factors of CRKP are carried out in the United States, where the dominant clone of KPC-producing KP is ST256, in contrast to China, where the dominant strain is ST11 (Qi et al., 2011). Our study comprehensively described 100 CRKP clinical isolates collected in a tertiary teaching hospital in Shanghai, China, between January 2013 and July 2015. The study examined three dimensions: antibiotic resistance genes, molecular genotypes, and patient risk factors.

Study Design and Bacterial Isolates
The retrospective observational study was conducted in a comprehensive teaching hospital in Shanghai, China (Renji Hospital affiliated to Jiaotong University). This is a centrally located large teaching hospital in Shanghai (1,600 beds) with approximately 8,000 admissions/day. Consecutive non-duplicate strains from patients infected and/or colonized by CRKP between January 2013 and July 2015 were stored at −80 • C and included in this study. All the CRKP isolates were identified by matrix-assisted laser desorption ionization time-of-flight mass spectrometry (MALDI-TOF MS, Bruker Daltonics, Bremen, Germany).
To study the risk factors of patients infected by CRKP infection, a stepwise matching technique was used to identify appropriate control cases from patients infected with carbapenem-susceptible Klebsiella pneumoniae (CSKP). Controls were matched to cases who were in the same ward during the same period (within 30 days) and were within 5 years of their age. If several patients in the control group met the qualification, one was randomly selected. An independent researcher, blinded to the objectives of the study, selected the control group according to these inclusion criteria. Informed consent was waived by The Institutional Review Board of Renji Hospital as all the patients' personal information was de-identified and was unknown to the authors, and the collection of isolates was also approved.

Pulsed-Field Gel Electrophoresis (PFGE)
Molecular typing was performed by pulsed-field gel electrophoresis (PFGE). Genomic DNA was fragmented by restriction endonuclease ApaI (TaKaRa, Tokyo, Japan), and the DNA fragments were separated by electrophoresis on 1% SeaKem Gold agarose (Lonza, Rockland, ME, United States) in 0.5× TBE (45 mM Tris, 45 mM boric acid, 1.0 mM EDTA; pH 8.0) buffer using the CHEF Mapper XA PFGE system (Bio-Rad, Unietd States) at 6 V/cm and 14 • C, with alternating pulses at a 120 • angle in a 5-20 s pulse time gradient for 19 h. The PFGE results were visually assessed, and a dendrogram was generated by unweighted pair group method using arithmetic average (UPGMA) clustering in NTSYS software version 2.1 (State University of New York, Stony Brook, NY, United States). The isolates were placed into one PFGE type (i.e., clonally related) if the Dice coefficient was >80% (correlation coefficient > 0.80) (Endimiani et al., 2009;Chiu et al., 2013;Cheng et al., 2016), while patterns with no differences in fragments were considered to be of the same subtype (Tenover et al., 1995).

Data Collection
After patients in both the case and control groups had been confirmed, data were collected by two trained reviewers from the electronic medical records and clinical microbiology laboratory databases. Thirty-nine clinical variables were applied in the analysis of the risk factors and evaluated for an interval of 90 days prior to the occurrence of KP infection. These variables included (1) demographics (gender, age); (2) baseline diseases (hypertension, diabetes mellitus, heart disease, malignancy, chronic pulmonary disease, liver disease, hematologic disease, nervous system disease); (3) community or nosocomial acquired infection; (4) previous therapeutic processes during the hospital stay, including length of hospital stay prior to KP infection, immunosuppressant therapy (>14 days of corticoid drugs or other immunosuppressants, or receipt of >1 dose of chemotherapy or immunosuppressants), ICU stay, surgery, blood transfusion, invasive device use (invasive ventilation, parenteral nutrition, drainage tube, urinary catheter, vascular catheter, sputum suction, nasogastric feeding tube); (5) medical history (90 days before KP infection), including previous hospitalization and exposure to antibiotics; (6) antibiotic exposure during hospital stay (cephalosporins, carbapenems, quinolones, enzyme inhibitors, penicillins, aminoglycosides, macrolides, tetracyclines, nitroimidazoles, fosfomycins, glycopeptides, lincosamides, macrodantin, sulfanilamides). These variables were analyzed to identify risk factors for CRKP infection.
Patients in the CRKP infection group were enrolled and divided into effective and non-effective groups to evaluate the impact of clinical outcomes. The definition of effective group with a positive outcome (cure or improvement) of CRKP infection was based on clinical (defeverscense, resolution or partial resolution of presenting symptoms, and signs of pneumoniae decline in suctioning requirements), radiological (decrease or disappearance of the findings on chest X-ray), and laboratory findings (improvement of arterial blood gasses and normalization of white blood cell count, C-reactive protein, and procalcitonin) (Michalopoulos et al., 2008). The therapeutic effect was studied by comparing the course of antibiotic treatment (first-line drugs: 3rd or 4th generation cephalosporins, carbapenems, quinolones, enzyme inhibitors; second-line drugs: tigecycline, fosfomycin, sulfanilamides) in CRKP-infected patients.

Definitions
Carbapenem-resistant Klebsiella pneumoniae was defined according to the recommendation of the Center for Disease Control and Prevention 2 (March 1, 2016) as KP strains resistant to at least one of the carbapenem agents including imipenem, ertapenem, and meropenem. The percentages of MDR, extensively drug-resistant (XDR) and pandrug-resistant (PDR) strains in CRKP were calculated according to the pre-specified criteria (Magiorakos et al., 2012).
The definitions of the site of infection and the responsible pathogen were based on the clinical symptoms and signs in individual patients, imaging findings, and on the isolation of a potential pathogen from evaluable clinical specimens. In addition, patients must have had fever >38 • C with no other recognized cause, or abnormal white blood cell count [leukopenia (<4000 WBC/mm 3 ) or leukocytosis (≥12,000 WBC/mm 3 )], and at least two of the following: new onset of purulent sputum or change in the sputum characteristics, increased respiratory secretions or increased suctioning requirements, new onset or worsening of a cough or dyspnea or tachypnea, rales or bronchial breath sounds, or worsening gas exchange (Falagas et al., 2005).

Statistical Analysis
All the statistical analyses were performed by SPSS 18.0 (IBM Corp., Armonk, NY, United States). Continuous variables, summarized as means ± SD (x ± SD) or medians, were compared by Student's t-test or the Mann-Whitney U-test. Qualitative variables were compared by χ 2 or Fisher's exact test. The univariate analysis was used for each variable. Twosided p-values less than 0.05 were considered statistically significant. Variables with p < 0.05 were inserted into a binary logistic regression model for determining the potential independent risk factors associated with CRKP infection. Odds ratios (ORs), including their 95% confidence intervals (CIs), were calculated to evaluate the association between the factors and outcomes.

PFGE Analysis
Pulsed-field gel electrophoresis patterns of the 100 blaKPC-2 producing CRKP isolates identified 18 PFGE types (clusters A-R) (Figure 1). No single dominant intra-hospital PFGE type was detected by using a cutoff of 80% similarity. Seven PFGE subtypes, including C3, H3, M2, M5, N5, Q2, and Q3, each contained at least three CRKP strains. Among them, all three of the CRKP strains in the H3 subtype were isolated from CRKP infection patients according to electronic medical records. In addition, according to epidemiological data such as the isolation date, department and specimen types, there were two CRKP detection peak periods (September, 2014-October, 2014and April, 2015-June, 2015 involving three departments surgery, neurosurgery and geriatric ward. Details of PFGE clusters of CRKP strains isolated during those two detection peak periods were listed in Table 2.
The multivariate analyses of CRKP infection and control group models were based on the adjustment of the logistic regression model for 12 variables, wherein the risk factors differed significantly in the univariate analyses. As shown in Table 4, exposure to sputum suction was found to be the sole independent risk factor for CRKP infection.

CRKP Infection Outcome Analysis
In 51 cases of CRKP infection, 28 patients (54.9%) were effective and 23 patients (45.1%) were non-effective, which were identified according to the preset criteria described in Section "Materials and Methods." The usage and duration of various kinds of antibiotics were compared between the effective and non-effective treatment groups. As shown in Table 5, among all the antibiotics (cephalosporins, enzyme inhibitors, quinolones, carbapenems, fosfomycins, tigecycline, sulfanilamides), the treatment course (days) of carbapenems (p = 0.025) was a statistically significant factor.

DISCUSSION
Infections caused by carbapenem-resistant Enterobacteriaceae, especially by CRKP strains, are a pressing issue for public health. Although several attempts been made to control the spread of these infections at the local or national level, the rapid dissemination of CRKP has still posed an urgent threat globally (Girmenia et al., 2016;Lee et al., 2016). Because CRKP strains are either MDR, XDR or PDR, the available treatment protocols specific to the CRKP infected patients are limited. Although some combination therapy methods have been recommended for the treatment of CRKP infections, there is still a dearth of clinical evidence. Several studies have certified that CRKP, which is one of the most critical nosocomial pathogens, could increase the mortality rate, especially in those patients with severe diseases (Ducomble et al., 2015;Tumbarello et al., 2015). Therefore, our study combined the molecular epidemiology of clinical CRKP isolates and the study of risk factors for CRKP infection, discussed the epidemiological pattern, and evaluated different therapeutic effects in various treatments, to enable guidance in preventive measures and therapeutic strategies of CRKP infection in clinical applications.
In the current study, we used CRKP strains identified in patients between January 2013 and July 2015 and observed a distinct escalating trend during this period: CRKP-positive from 0.6% (4/656) in 2013 to 30.1% (142/472) in 2015. Focusing on the molecular detection and genetic characterization of CRKP, we found that the emergences of ESBL and carbapenemase genes were the primary resistance mechanisms. With respect to the ESBL genes, of which CTX-M type enzymes were reported as the most common type, 96% of CRKP isolates in our study harbored blaCTX-M-9, whereas only 1% contained the blaCTX-M-15 gene, which is by far the most prevalent CTX-M variant worldwide (Cantón et al., 2012). Much of this discrepancy is due to geographical area (Ranjbar et al., 2016), because blaCTX-M-9 and blaCTX-M-14, which belong to the CTX-M-9 group, were important CTX-M-positive strains in China (Cantón et al., 2012). On the other hand, among the carbapenemase genes, all the CRKP strains tested positive for blaKPC-2 while only one CRKP was positive for the metallo-carbapenemase gene blaIMP-1. The KPC enzyme was first reported in 1996 in a North Carolina (United States) hospital and was subsequently reported in several countries. Due to the emergence of diverse carbapenemase genes, KPC-2 is the most prevalent in China, with rare detection of metallo-carbapenemases, which was in agreement with our results. However, some other countries may have different dominant carbapenemases; for example, the United Kingdom is likely to have a mixed carbapenemase pattern with VIM and NDM, and NDM types followed by OXA-48-like types were prevalent in India (Munoz-Price et al., 2013). Moreover, the rate of positive detection of AmpC type β-lactamase was 14%, with only the DHA type being detected. DHA has been reported as the most frequent among AmpC β-lactamase-producing KP isolates in many studies (Shi et al., 2013;Liu and Liu, 2016), which was in line with our study. Therefore, DHA is speculated to be the most common type of AmpC β-lactamase in China.
We identified a total of 18 PFGE types (A-R) and 70 subtypes in 100 CRKP strains using PFGE analysis. Though no dominant intra-hospital PFGE type was identified using a cutoff of 80% similarity, two issues should be noted. Firstly, three PFGE H3 type CRKP strains were isolated from one surgical patient and from two patients in SICU later within 2 weeks. Although there   Furthermore, our study shows that among the 100 CRKP strains, 51% isolates were from patients infected with CRKP while the remaining 49% were isolated from colonized patients. Unlike other studies, the respiratory tract was the most common infection site (32/51, 62.7%) followed by bloodstream (5/51, 9.8%) and urinary tract (5/51, 9.8%) while many other studies presented bacteremia (Falagas et al., 2007) and urinary tract (Teo et al., 2012) as the leading sites of infection. This may be correlated with the distribution of the specimen types sent to the clinical microbiology laboratory.
Univariate analyses revealed that the factors associated with the CRKP infection included 12 variables significantly different between the case (CRKP) and control (CSKP) groups. With respect to the baseline disease, the chronic pulmonary disease showed a low incidence rate in the case group compared to that in the control group with the OR value 0.256 (95% CI 0.072-0.908). Interestingly, if the pulmonary disease emerged as the underlying illness, it would be a protective factor for CRKP infection while if it occurred as a comorbidity disease, the opposite effect would occur (Lucena et al., 2014). In addition, ICU stays, various invasive manipulations, and prior medical history including admission to the hospital or exposure to antibiotics 90 days before the outcome of interest, were strongly correlated with CKPN infections, which has also been identified by other studies (Lucena et al., 2014;Hoxha et al., 2016;Lee et al., 2016). Regarding antibiotic exposure during hospital stays, the usage of carbapenems, enzyme inhibitors, nitroimidazoles, and glycopeptides was associated with CRKP infections. This prompted us to apply antibiotics according to the results of antimicrobial susceptibility testing instead of using broad spectrum antibiotics to prevent the increasing CRKP infections from causing in-hospital mortality.
Unlike other studies, only use of sputum suction (p = 0.006) was shown to be an independent risk factor in this study. Jiao et al. (2015) reported that exposure to glycopeptides, cefoperazone plus sulbactam, and tracheostomy were independent risk factors for CRKP infection/colonization. Falagas et al. (2007) conducted a case-control study and found that exposures to fluoroquinolones and antipseudomonal penicillin were independent risk factors for CRKP infections. Thus, nosocomial transmission may play a more critical role in the spread of CRKP than the selective pressure imposed by the usage of antibiotics in our hospital.
In addition, we further investigated the outcome of different treatment options for all the patients with infections caused by CRKP. Carbapenems play a major role in therapies treating infections caused by Enterobacteriaceae. However, the emergence of carbapenem-resistant Enterobacteriaceae has compromised the application of carbapenems in clinical practice. Because of this, we investigated the effectiveness of different antibiotic regimes in treating CRKP infection. In our study, only continual carbapenems usage was associated with greater treatment success. Although the antimicrobial susceptibility test showed in vitro resistance of bacteria to carbapenems, these antibiotics may be effective if the treatment period is prolonged. The results of this study may be partially influenced by the retrospective nature of data collected from the medical records and the treatment options, and the continuous utility period of the antibiotics may have been affected by medical insurance or patient's financial capacity. However, consistent usage of carbapenems would be beneficial to CRKP-infected patients.
In summary, SICU, NSG, SUR, and GA were high-risk CRKP outbreak departments in our hospital, and epidemiological surveillance of resistance should be performed routinely. In the case-control study, sputum suction was an independent risk factor for CRKP infections. Prolonged utility period of carbapenems benefited patients infected with CRKP.

ETHICS STATEMENT
This study was approved by The Institutional Review Board of the Renji Hospital, School of Medicine, Shanghai Jiaotong University, Shanghai, China. No consent was needed for this study.

AUTHOR CONTRIBUTIONS
BZ, YD, WS, ED, and DZ performed experiments. YY, YL, and YH assisted in data collection from the case and control groups. WS assisted in antimicrobial susceptibility testing. ML, BZ, YY, and WS conceived the study and analyzed the results. ML and BZ supervised the study and prepared the manuscript. All authors read and approved the final manuscript.