Risk Factors, Antimicrobial Susceptibility Patterns and Carbapenem Resistance Mechanisms of Elizabethkingia anophelis Clinical Isolates From a University Tertiary Hospital in Southwest China

Purpose: Elizabethkingia anophelis (E. anophelis) is an important extensively drug-resistant (XDR) pathogen to which there are limited therapeutic options. E. anophelis is perpetually misidentied as Elizabethkingia meningoseptica (E. meningoseptica) by conventional methods. Consequently, this study reassessed risk factors for infection and mortality, antimicrobial susceptibility patterns and carbapenem resistance mechanisms of E. anophelis. Methods: This retrospective case–control study was conducted to reveal the risk factors associated with E. anophelis infection and in-hospital mortality from 2015–2019 in a university tertiary hospital in southwest China using univariable and multivariable logistic-regression analysis. Case patients infected with E. anophelis isolates and controls patients with non-E. anophelis infections were compared at a ratio of 1:3 during the same time period. We employed the broth microdilution method to investigate the antimicrobial susceptibility proles of 39 E. anophelis strains. PCR amplication, DNA sequencing and gene cloning were used to investigate the mechanisms of carbapenem resistance in E. anophelis. Results: We collected 39 non-repetitive E. anophelis isolates over the study period. Multivariate analysis indicated that coronary artery disease, chronic obstructive pulmonary disease, surgery in the past 6 months, anemia and systemic steroid use were independent risk factors for the acquisition of E. anophelis. Additionally, anemia was the only independent risk factor associated with in-hospital mortality in patients with E. anophelis infections. E. anophelis isolates showed high in-vitro susceptibility towards minocycline (100%) and piperacillin/tazobactam (71.8%), but were resistant to colistin, fosfomycin, ceftazidime/avibactam and aztreonam/avibactam. Additionally, we show that two metallo-β-lactamases (MBLs) BlaB and GOB, are responsible for carbapenem resistance and the serine-β-lactamase, CME, is functionally involved in resistance to cephalosporins and monobactams. Importantly, the various putative eux pumps in E. anophelis are not responsible for resistance. Conclusion: Our ndings will help clinicians identify high-risk patients and suggest that minocycline should be considered as an antibiotic therapeutic option for infections caused by E. anophelis. Additionally, carbapenem resistance in E. anophelis isolates is mainly associated with the MBLs, BlaB and GOB. cefepime, aztreonam, ceftazidime/clavulanic acid, cefepime/clavulanic acid, colistin and fosfomycin according to the breakpoints used. Most of the antimicrobial susceptibility results in this study are consistent with those of previous studies performed using the broth microdilution test. These results suggest that antimicrobial therapy for E. anophelis should prioritize minocycline or piperacillin-tazobactam. However, in this study, patients in the non-survival group were treated with signicantly more β-lactam/lactamase inhibitor antibiotics as compared to the survival group (P = 0.041). This observation indicate that piperacillin-tazobactam is not an effective treatment for E. anophelis infections. In this study, we screened for the presence of carbapenemase genes (blaKPC, blaIMP, blaVIM, blaNDM, blaBlaB, blaGOB and blaOXA-48-like) and ESBLs genes (blaTEM, blaCME and blaCTX-M) in all 39 E. anophelis isolates. No carbapenemase genes (blaKPC, blaIMP, blaVIM, blaNDM and blaOXA-48-like) or ESBLs genes (blaTEM, blaSHV and blaCTX-M) were detected in any isolates. However, we identied 35 isolates co-harboring blaGOB, blaBlaB and blaCME β-lactamase genes. The detected combination of β-lactamases was CME-1, BlaB-29 and GOB-38 = pET28a(+) vector construct. This suggests that the ESBL CME is functionally involved in resistance to cephalosporins and monobactams. The MIC of imipenem increased 32-fold (from 0.125 to 4 µg/mL) in the presence of the BlaB-expressing construct and 16-fold (from 0.125 to 2 µg/mL) in the presence of the GOB-expressing construct. This suggests that the metallo-β-lactamases BlaB and GOB are responsible for the observed carbapenem resistance. the function of these putative eux pumps. Our data show that all recombinant eux pump strains including: pET-ABC, pET-MFS, pET-MATE, pET-SMR and pET-RND did not result in increased MICs for β-lactam and non-β-lactam antibiotics. These results suggest that the putative eux pump genes from E. anophelis are not eux pumps which mediate antimicrobial drug resistance. Similarly, meningoseptica; ORs: odds ratio; CI: condence interval; IQR: interquartile range; CLSI: clinical and Laboratory Standards Institute; MICs: minimum inhibitory concentrations; ESBL: extended-spectrum betalactamase; MBLs: metallo-betalactamases; WHO: World Health Organization; FDA: Food and Drug Administration; EUCAST: European Committee on Antimicrobial Susceptibility Testing; CDSs: coding sequences.


Introduction
Elizabethkingia anophelis (E. anophelis) is an aerobic, immotile, oxidase-positive, indole-positive, Gram-negative, nonfermenting bacillus belonging to the genus Elizabethkingia and the family Flavobacteriaceae which was rst isolated from midgut of the mosquito Anopheles gambiae in 2011 [1]. The rst clinically signi cant E. anophelis infection was associated with a case of neonatal meningitis in Bangui, Central African Republic in 2011 [2]. Thereafter, the prevalence of E. anophelis infections has been rampant; it is the primary species of Elizabethkingia genus infections that have recently occurred in Singapore [3], Hong Kong [4], the United States [5,6,7,8], South Korea [9] and Taiwan [10]. E. anophelis is an important opportunistic pathogen associated with various invasive infections in adults with underlying diseases and infants (premature and neonates) [11,12]. Several outbreaks have occurred in Singapore and in the Midwestern United States of Wisconsin, Illinois, and Michigan, where the fatality rate of patients with E. anophelis infections ranged from 30-60% [5,6,7,8]. A recent study indicated that E. anophelis has been perpetually misidenti ed as Elizabethkingia meningoseptica (E. meningoseptica) using conventional methods (API/ID32, Phoenix 100 ID/AST, Vitek 2 and Vitek MS) [13]. Therefore, most of the previously reported data regarding the clinical characteristics, antimicrobial susceptibility patterns and antimicrobial resistance mechanisms of E. anophelis could be incorrect.
Extensive research has shown that E. anophelis isolates are resistant to most β-lactams, carbapenems and aminoglycosides [4,5,14,15,16,17]. Extensively drug-resistant pathogens pose a therapeutic dilemma for clinicians and are therefore associated with a high mortality rate and poor prognosis. Interestingly, different testing methods and geographic variation lead to obvious differences in patterns of susceptibility. So far however, the susceptibility patterns of E. anophelis isolates have not been reported in China, especial with data collated using the more robust broth dilution method. Genome-wide analysis reveals that this multidrug-resistant pathogen carries a class A serine-β-lactamase, CME, 2 metallo-β-lactamases, GOB and BlaB, in addition to numerous putative e ux pump genes [5,18,19,20]. However, no studies have focused on the functions of these βlactamases and putative e ux pumps in E. anophelis isolates. In addition, data reporting on the risk factors for infection and mortality of patients with an E. anophelis infection may potentially help clinicians identify high risk patients and inform on future therapeutic strategy.
The present study was initiated to: (i) identify the risk factors associated with E. anophelis infection and in-hospital mortality, (ii) investigate the antimicrobial susceptibility patterns and carbapenem resistance mechanisms of E. anophelis isolates and (iii) characterize the functions of β-lactamases and putative e ux pumps expressed in E. anophelis isolates.

Bacterial Strains
This study used the database of clinical isolates provided by the clinical laboratory of a 3200-bed university-a liated medical center (Chongqing, China) to collect the strains that were identi ed as Elizabethkingia species between January 2015 and December 2019. Microbial identi cation was performed in the microbiology laboratory using the VITEK2 compact (bioMérieux, Inc., NC, USA) system and the VITEK MS (bioMérieux, MO, USA) system. All Elizabethkingia species strains were stored at − 80 °C in 15% glycerol until use. Complete 16S rRNA gene sequencing was used to recon rm the identity of all isolates. The primers used for ampli cation and sequencing of the 16S rRNA gene are listed in Additional le 1. The sequences were assembled using Seqman (DNAStar) and compared with publicly available sequences in the NCBI (http://www.ncbi.nlm.nih.gov) using the BLAST algorithm. Strains were considered to be accurately identi ed when a strain shared > 99.0% 16S rRNA sequence with a type of strain in GenBank.

Data collection and Clinical De nitions
The electronic medical records of the patients were collected retrospectively. We excluded subjects with the following characteristics: patients with polymicrobial infection and patients admitted for < 48 hours. Only the rst episode was taken into account for patients with more than one positive E. anophelis culture. To evaluate the risk factors of E. anophelis infection, controls were de ned as randomly selected patients with non-E. anophelis infections during the same time period (at a 3:1 ratio to the case group). Selected epidemiological, demographic, clinical, laboratory, treatment and outcome data were obtained from the electronic medical records. Empirically administering agents to isolates which were not susceptible was de ned as inappropriate empirical antimicrobial therapy. Shock was de ned as the coexistence of systolic pressure of < 90 mm Hg and organ dysfunction of the respiratory system, liver or kidneys. Serum total protein content of < 60 g/L or albumin content < 25 g/L were the criteria used to de ne hypoproteinemia. Hypokalemia was diagnosed at a serum potassium level < 3.5 mmol/L. Systemic steroid use was de ned as oral or intravenous administration of at least 20 mg/day of a steroid (prednisone, hydrocortisone, methylprednisolone or dexamethasone) within 1 month of infection. We de ned anemia as hemoglobin < 130 g/L in men and < 120 g/L in women according to the WHO (World Health Organization) guidelines. The primary clinical outcome was in-hospital mortality.

Antimicrobial Susceptibility Testing
The reference broth microdilution method was used to evaluate the minimum inhibitory concentrations (MICs) of all antibiotics in E. anophelis and recombinant strains according to the Clinical and Laboratory Standards Institute (CLSI) M07-Ed11 (2019). The criterion suggested by the CLSI for "other non-Enterobacteriaceae" was used to determine the susceptibility of isolates to antibiotics with the exception of ceftazidime/avibactam, aztreonam/avibactam, vancomycin, tigecycline, rifampicin, colistin and fosfomycin. The US Food and Drug Administration (FDA) Enterobacteriaceae criteria was used to interpret isolate susceptibility to tigecycline (resistant MIC ≥ 8 µg/mL, susceptible MIC ≤ 2 µg/mL and intermediate MIC = 4 µg/mL). The MIC breakpoint applied to vancomycin and rifampicin was adapted from the CLSI criteria for Staphylococcus spp.. A MIC of ≥ 16/4 µg/mL was considered resistant for the combination ceftazidime/avibactam and aztreonam/avibactam. Colistin MICs were interpreted at susceptible breakpoints of ≤ 2 µg/mL and resistant breakpoints of > 2 µg/mL according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) Enterobacteriaceae criteria. Likewise, for fosfomycin, we elected to use the susceptible breakpoint of ≤ 32 µg/mL and resistant breakpoint of > 32 µg/mL based on EUCAST Enterobacteriaceae criteria. The reference strains Escherichia coli (ATCC 25922) and Pseudomonas aeruginosa (ATCC 27853) were used as negative controls and quality controls for antibiotic susceptibility testing.
We ampli ed the full length coding sequences (CDSs) of genes using speci c primers anked by restriction sites (EcoRI, XhoI or BamHI). These included genes encoding: blaCME, blaBlaB, blaGOB, CzcABC family e ux RND transporter, E ux ABC transporter (ATP-binding protein), MATE family of MDR e ux pump, small multidrug resistance family (SMR) protein and MFStype transporter. Ampli ed PCR fragments were puri ed and cloned into the corresponding sites of the pET-28a plasmid and were electroporated into BL21 competent E. coli. Successful cloning was veri ed by PCR and sequencing analysis. The sets of primers used for ampli cation and sequencing of target genes are provided in Additional le 1.

Statistical Analysis
The data were evaluated using SPSS statistical software (version 22.0, IBM). Data are presented as counts (proportions) for categorical variables. Direct comparisons between two groups were analyzed by the χ2 test or Fisher's exact test. Mean (± standard deviation) was used to express normally distributed continuous variables and median (± inter-quartile range) was calculated for non-normally distributed variables. Comparisons between two groups were conducted using a Student's t-test for normally distributed variables or a Mann-Whitney U test for non-normally distributed variables. To evaluate independent risk factors for the infection and in-hospital mortality of E. anophelis isolates we examined all plausible variables using a univariate analysis. Risk factors with a P value < 0.1 as detected by the univariate analysis were included in a multivariate logistic regression model with the enter method. Odds ratios (ORs) and 95% con dence intervals (CIs) were determined for each risk factor of infection and in-hospital mortality. A two-tailed P value < 0.05 (P < 0.05) was considered statistically signi cant.

Clinical Characteristics
A total of 59 non duplicated Elizabethkingia isolates were collected from the clinical microbiology laboratory to investigate microbial traits. Full 16S rRNA gene sequencing was performed on 59 Elizabethkingia clinical strains. After comparison to the available sequences in the NCBI using BLAST, 16S rRNA sequencing showed that 39 isolates were identi ed as Elizabethkingia anophelis R26.
Of the E. anophelis isolates with a documented site of infection, 19 (48.7%) were from the respiratory tract, 8 (20.5%) were from the urinary tract, 6 (15.3%) were from blood, 3 (7.6%) were from cerebrospinal uid, 2 (5.1%) were from the tip of the venous catheter and 1 (2.6%) was from peritoneal uid (Table 4). From January 2015 to December 2019, 39 E. anophelis samples were isolated from 39 consecutive patients which were enrolled in the study. These patients consisted of 21 males (53.8%) and 18 females (46.2%) with a median age of 61.
Univariate and multivariate logistic regression analysis results for the factors assosiated with in-hospital mortality are shown in Table 2. Univariate analysis showed that cerebrovascular disease (P = 0.035), chronic obstructive pulmonary disease (P = 0.020), nasogastric tube insertion (P = 0.008) and anemia (P = 0.002) were associated with a higher mortality rate. Using further multivariate analysis, anemia (OR 86.38, 95% CI: 1.42-5251.29; P = 0.033) was identi ed to be the only independent risk factor for in-hospital mortality in patients with E. anophelis infections.

MICs of Recombinant Strains
To further evaluate the function of β-lactamases, the most prevalent versions of blaCME, blaBlaB and blaGOB genes from E. anophelis isolates were cloned into a pET28a(+) plasmid vector. We also ampli ed and cloned genes encoding putative e ux pump proteins including: CzcABC family e ux RND transporter, E ux ABC transporter (ATP-binding protein), MATE family of MDR e ux pump, small multidrug resistance family (SMR) protein and MFS-type transporter. These plasmids were transformed into BL21 (DE3) E. coli and the MICs of the common antibiotics were tested in the resultant strains. The strain transformed with pET-CME displayed an increased MIC of ampicillin, piperacillin, cefazolin, cefuroxime, ceftazidime, ceftriaxone and aztreonam compared with the pET28a(+) vector construct. This suggests that the extended-spectrum serine-β-lactamase CME is functionally involved in cephalosporins and monobactams resistance ( Table 5). The MIC of imipenem increased 32-fold (from 0.125 to 4 µg/mL) in the presence of the pET-BlaB construct and 16-fold (from 0.125 to 2 µg/mL) in the presence of the pET-GOB construct. This suggests that the MBLs BlaB and GOB are responsible for increased imipenem resistance (Table 5). Along with the increase in imipenem resistance, the pET-BlaB and pET-GOB constructs also conferred an increased MICs of ampicillin, piperacillin, cefazolin, cefuroxime and ceftazidime. This indicates that the MBLs BlaB and GOB also have the ability to degrade ampicillin, piperacillin, cefazolin, cefuroxime and ceftazidime (Table 5). All the e ux pump transformants tested, including pET-ABC, pET-MFS, pET-MATE, pET-SMR and pET-RND did not result in increased MICs of any antibiotics tested (Table 5).

Discussion
Infection with E. anophelis in humans is increasing in many countries [3][4][5][6][7][8][9][10]. Furthermore, there have been several reports of E. anophelis outbreaks in the community and nosocomial environment in Singapore and the Midwestern United States of Wisconsin, Illinois and Michigan [3,[5][6][7][8]. However, as previously mentioned, E. anophelis is commonly misidenti ed as E. meningoseptica by biochemical identi cation methods or automated identi cation systems in clinical settings. Therefore, data regarding the clinical features, clinical prognosis and the antimicrobial susceptibility pro les of E. anophelis could be skewed. For the rst time this study identi ed the risk factors associated with the acquisition of E. anophelis. We show that anemia is an independent risk factor for in-hospital mortality in patients with E. anophelis infections. Moreover, we demonstrate for the rst time that various putative e ux pumps in E. anophelis do not alter antimicrobial resistance, thus seemingly do not to possess a drug e ux function. β-lactamases were commonly found in E. anophelis isolates. The MBLs BlaB and GOB are responsible for carbapenem resistance, whereas the ESBL CME is functionally involved in resistance to cephalosporins and monobactams.
In the present study we initially explored risk factors leading to E. anophelis infection. Using multivariate analysis we found that coronary artery diseases, chronic obstructive pulmonary disease, surgery in the past 6 months, anemia and systemic steroid use were independently associated with E. anophelis infection. Previous studies have suggested that patients with E. anophelis infection could have more underlying comorbidities [3][4][5][6][7][8][9][10]. This study is the rst to provide a statistical analysis to support this hypothesis. Moreover, surgery in the past 6 months, anemia and systemic steroid are known to compromise the immune status of the patients. Therefore, these patients are more vulnerable to acquiring E. anophelis infection in the same hospital environment.
In previous reports, the case fatality rate of patients with E. anophelis infection ranged from 24-60% in different countries [4][5][6][7][8][9][10][11][12][13][14]. In line with this, our study showed that the in-hospital mortality rate of patients with an E. anophelis infection was 51.3%. We further explored the factors in uencing mortality. Compared with the only previous study which investigated risk factors for mortality [10], our study demonstrated that anemia was the only independent predictor for mortality in patients infected with E. anophelis, a factor which has not been reported prior to the present study. Patients may present as anemic because of the hemolytic activity of E. anophelis. Several studies have reported that this bacterium lyse animal erythrocytes to access essential nutrients (such as amino acids) using hemolysins and heme-degrading proteins [20][21][22]. The process may alter the host physiological status and compromise the immune system, thus worsening the prognosis of patients infected with E. anophelis. Therefore, anemic patients with a con rmed E. anophelis infection should be considered as higher risk and should be given more attention and special care. Unexpectedly, inadequate antibiotic therapy was not associated with mortality probably because of the limited number of the infected patients in our study.
There are also obvious differences in the susceptibility of E. anophelis to antibacterial agents when the standard broth microdilution test is used. A possible explanation for this inconsistency is that natural geographical differences cause variation in the susceptibility patterns observed in previous studies. It is therefore necessary to investigate the antimicrobial susceptibility of E. anophelis in local area for to give effective guidance on antibiotic selection. To the best of our knowledge, this is the rst study to investigate the antibiotic susceptibility of E. anophelis in China. Our study showed the following susceptibility of E. anophelis to minocycline (100%), piperacillin-tazobactam (71.8%), levo oxacin (38.5%), cipro oxacin (30.8%), piperacillin (17.9%), rifampicin (20.5%) and tigecycline (10.3%). All of the isolates displayed resistance to ceftazidime, cefepime, aztreonam, ceftazidime/clavulanic acid, cefepime/clavulanic acid, colistin and fosfomycin according to the breakpoints used. Most of the antimicrobial susceptibility results in this study are consistent with those of previous studies performed using the broth microdilution test. These results suggest that antimicrobial therapy for E. anophelis should prioritize minocycline or piperacillin-tazobactam. However, in this study, patients in the non-survival group were treated with signi cantly more β-lactam/lactamase inhibitor antibiotics as compared to the survival group (P = 0.041). This observation indicate that piperacillin-tazobactam is not an effective treatment for E. anophelis infections.
To further evaluate the function of β-lactamases, recombinant strains harboring either blaCME, blaBlaB or blaGOB were constructed. The transformed strain expressing CME displayed an increased MIC of ampicillin, piperacillin, cefazolin, cefuroxime, ceftazidime, ceftriaxone and aztreonam as compared to the pET28a(+) vector construct. This suggests that the ESBL CME is functionally involved in resistance to cephalosporins and monobactams. The MIC of imipenem increased 32-fold (from 0.125 to 4 µg/mL) in the presence of the BlaB-expressing construct and 16-fold (from 0.125 to 2 µg/mL) in the presence of the GOB-expressing construct. This suggests that the metallo-β-lactamases BlaB and GOB are responsible for the observed carbapenem resistance.
Aztreonam/avibactam is a novel class of combination β-lactamase-inhibitor designed to treat serious infections of metallo-βlactamase (MBL)-producing, Gram-negative bacteria, which currently in phase I clinical trials (NCT01689207). Aztreonam is relatively stable against MBL hydrolysis, however, it is easily inactivated by class A (e.g., KPC), class C (e.g., AmpC) and certain class D (e.g., OXA-48) serine-β-lactamase enzymes [23]. Avibactam potently inhibits class A, class C and certain class D serineβ-lactamase enzymes, which displays a broader β-lactamase inhibition pro le than other β-lactamase inhibitors [23]. When combined, aztreonam/avibactam is effective against isolates co-producing ESBL and MBLs with porin loss/de ciency [24]. However, it is quite unexpected that E. anophelis is resistant to aztreonam/avibactam according to our experimental results.
Genomic annotation of all Elizabethkingia spp. reveals that besides β-lactamases, there are also numerous putative e ux pump proteins including: CzcABC family e ux RND transporter, E ux ABC transporter (ATP-binding protein), MATE family of MDR e ux pump, small multidrug resistance family (SMR) protein and MFS-type transporter. Interestingly however, none of these transporters have been phenotypically characterized [5,18,19,20]. It was therefore critical to investigate the function of these putative e ux pumps. Our data show that all recombinant e ux pump strains including: pET-ABC, pET-MFS, pET-MATE, pET-SMR and pET-RND did not result in increased MICs for β-lactam and non-β-lactam antibiotics. These results suggest that the putative e ux pump genes from E. anophelis are not e ux pumps which mediate antimicrobial drug resistance. Similarly, Schindler et al. cloned and expressed 21 putative e ux pump genes in Staphylococcus aureus which had no effect on any of the antibiotics tested [25]. In summary, we demonstrate for the rst time that the various putative e ux pumps in E. anophelis do not to possess antimicrobial drug e ux function.
There were some limitations to our study. Firstly, the limited sample size from the single-center study limited its translation to the wider population. However, the identi cation of both a carbapenem resistance mechanism and the susceptibility pro le of the extensively drug-resistant E. anophelis are of great clinical importance and warrant an urgent, wider, in-depth study. Secondly, no further investigation into the clonality of these isolates was performed, so the possibility of infection outbreaks cannot be ruled out.

Conclusions
Some conclusions of this study are signi cant. First, patients with anemia, coronary artery diseases, chronic obstructive pulmonary disease or patients who have received systemic steroids or surgery in the past 6 months are more likely to acquire an E. anophelis infection. A check list of the identi ed risk factors at hospital admission will help the clinicians identify high risk patients. Second, patients with anemia have a worse prognosis and therefore require more attention and special care from physicians. Third, our research reports the most prevalent rate of E. anophelis resistance to colistin (100%), fosfomycin (100%), aztreonam/avibactam (100%) and tigecycline (89.7%) which are regarded as last-resort treatments for carbapenem-resistant Enterobacteriaceae (CRE). Our results also showed that minocycline is the most effective antibiotic against E. anophelis. Fourth, mechanistic analysis reveals that carbapenem resistance is associated with the hydrolytic activity of the MBLs BlaB and GOB, and is not associated with various putative e ux pumps expressed in E. anophelis.