PDF
ePub
Citation
Print
INTRODUCTION
Acinetobacter baumannii (AB) has become associated with increasing morbidity and mortality rates in hospitals in the last two decades, owing to the emergence of drug resistance and limited treatment options [1234]. Recently, an increase in carbapenem resistance among AB strains has been reported [56]. These carbapenem-resistant AB strains also frequently display resistance to other antibiotics, consequently posing an eminent clinical threat. Thus, interest in “old” antibiotics has been rekindled [78].
Colistin, introduced in the 1950s to treat infections caused by gram-negative bacteria (GNB), exerts bactericidal activity by displacing the membrane-stabilizing calcium and magnesium ions and targets the polyanionic lipopolysaccharide (LPS) components [910]. However, because it induced nephrotoxicity and neurotoxicity, it was replaced by safer antimicrobial agents (e.g., aminoglycosides) [1112]. Despite these potential side effects, worldwide dissemination of extensively drug-resistant GNB (XDR-GNB) has rekindled the usage of this drug in clinical settings as a last-resort treatment.
The clinical use of colistin for XDR-GNB infections has led to the development of colistin resistance (CR) in GNB species [131415], and reports on the occurrence of colistin-resistant AB (CR-AB) are increasing globally [2]. Previous in vivo studies have demonstrated that CR in AB is mediated by a complete loss of LPS production through mutations in LPS-producing genes (lpxA, lpxC, lpxD, and lpsB) [1617] or by modification of lipid A components of LPS through mutations in pmrA and pmrB genes. These genes regulate the expression of the downstream target pmrC, which encodes an inner membrane phosphoethanolamine (PE) transferase modifying the outer membrane lipid A [1819]. Recently, the emergence of a plasmid-mediated mobile CR gene, mcr-1, in Enterobacteriaceae has been reported [20].
Most CR-AB clinical strains are reported to acquire resistance by in vivo selection during colistin treatment [21]; however, clonal spreading of CR-AB strains causing infections or colonization in patients without colistin treatment history has also been reported recently [1]. Both in vitro and in vivo models have shown that mutations in the PmrAB system lead to decreased fitness and virulence compared with that of colistin-susceptible (CS) parental strains [222324]. However, these studies evaluated serially obtained CR isolates and their parental CS strains only and thus did not consider the characteristics of different clinical CR-AB isolates obtained from different patients with and without history of colistin administration [23242526].
We compared CR-AB isolates recovered from patients with and without prior colistin treatment to assess whether prior colistin treatment affects CR in CR-AB isolates, patient demographics, mortality rates, or genetic mutations. Additionally, mortality rate was assessed to determine clinical characteristics.
METHODS
1. Bacterial isolates
In total, 36 non-duplicate AB clinical isolates resistant to both carbapenems and colistin were collected from a tertiary care hospital in Seoul, Korea, from April 2012 to December 2014. At the time of sample collection, 18 patients had received previous colistin treatment (Group CT), and the rest had not (Group non-CT). For comparison, AB isolates (N=11) that were resistant to carbapenems but susceptible to colistin were also studied. Bacterial species were identified by partial rpoB gene sequences and PCR detection of blaOXA-51-like. Patient data, including acute physiology and chronic health evaluation (APACHE II) score, use of colistin treatment, and 30-day mortality from the day of AB recovery, were examined retrospectively using electronic medical records. Multivariate logistic regression analysis was performed to identify risk factors associated with 30-day mortality from the day of CR-AB recovery. This project was approved by the Institutional Review Board of Yonsei University Severance Hospital, Seoul, Korea (4-2017-0758).
2. Antimicrobial susceptibility testing
The susceptibility of the isolates to colistin, meropenem, imipenem, piperacillin-tazobactam, ceftazidime, cefepime, gentamicin, tobramycin, amikacin, tetracycline, ciprofloxacin, and trimethoprim/sulfamethoxazole was determined by the disk diffusion method following the CLSI guidelines [27]. Minimum inhibitory concentrations (MICs) of meropenem and imipenem were determined by using Etest (bioMérieux, Inc., Durham, NC, USA). Colistin MIC was determined by the broth microdilution method, following recommendations of the Joint CLSI-EUCAST Polymyxin Breakpoints Working Group [28]. Synergistic effects of drug combinations of colistin (32–4,096 µg/mL) either with meropenem (4–256 µg/mL) or with rifampicin (0.25–32 µg/mL) were evaluated by the checkerboard method [29] in microtiter plates. The fractional inhibitory concentration (FIC) index of each drug combination was determined by dividing the MIC of each drug when used in combination by the MIC of each drug when used alone. The effect of a drug combination was determined by the FIC index: ≤0.5, a synergistic effect; 0.5–4.0, neutrality; and >4.0 an antagonistic effect.
3. PCR analysis of drug-resistant genes
A series of PCR experiments (primer information available upon request) were conducted to detect the OXA carbapenemase genes blaOXA-23-like,
blaOXA-24-like,
blaOXA-48-like, and blaOXA-58-like [30]; the metallo-β-lactamase genes blaIMP,
blaVIM, and blaNDM; and the serine carbapenemase genes blaGES and blaKPC [31]. The presence of ISAba1 upstream from the blaOXA-51-like gene was detected by PCR [32].
4. Genomic analysis of genes associated with colistin resistance
Genes associated with CR in AB (pmrA, pmrB, pmrC, lpxA, lpxC, lpxD, and lpsB) were analyzed by Sanger sequencing [1733]. The AB ATCC 17978 strain and 10 randomly selected colistin-susceptible AB (CS-AB) isolates were used as controls to distinguish CR-inducing mutations from polymorphisms. The mcr-1 gene was also identified by PCR [20].
5. Analysis of lipid A structure
Lipopolysaccharides and lipid A components were extracted from whole bacterial cells using Tri-reagent and mild acid hydrolysis, and were subjected to negative-ion matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry (Bruker Daltonik GmbH, Leipzig, Germany) in negative reflection mode. For comparison, three randomly selected CS-AB isolates were used as controls.
6. Pulsed-field gel electrophoresis
Pulsed-field gel electrophoresis (PFGE) was conducted with SmaI-digested genomic DNA extracted from the AB clinical isolates using a CHEF-DRII device (Bio-Rad, Hercules, CA, USA). PFGE band patterns were analyzed with Molecular Analyst Fingerprinting Software Ver. 3.2 (Bio-Rad). Genetic relatedness of PFGE profiles was interpreted using the criteria of Tenover et al [34].
7. Multilocus sequence typing (MLST)
MLST experiments were performed following the Bartual scheme [13]. Sequences of seven housekeeping genes (cpn60, gdhB, gltA, gpi, gyrB, recA, and rpoD) were used to determine the sequence types (STs) of the AB clinical isolates. Each ST number was assigned by comparing the allele sequences with those in MLST databases (http://pubmlst.org/abaumannii). Clonal complex (CC) was defined as a group of STs that shared five or more of seven alleles and was determined by eBURSTv.3 (http://eburst.mlst.net).
8. Statistical analysis
All variables were evaluated for Gaussian distribution using the Shapiro-Wilk test. Differences were tested with the Fisher exact test for categorical data and with the Mann-Whitney U test for continuous data. Univariate and multivariate analyses were carried out using logistic regression to investigate the association between CR, mortality rate, and potential covariates. The presence of variance inflation factors was examined for all parameters of the multiple regression model. P<0.05 was considered statistically significant. All analyses were performed using R (Version 0.99.893, R Studio, Inc., Boston, MA, USA).
RESULTS
1. Clinical characteristics of patients
The characteristics of patients infected or colonized by CR-AB are presented in Table 1. To determine whether prior colistin treatment had any relevant effect on patient outcome, Groups CT and non-CT were compared.
To determine the characteristics associated with higher survival rates, the “within 30-days deceased group” (13/36) was compared with the “alive group” (23/36) (Table 2). Only bloodstream infection and APACHE II scores significantly differed between groups.
2. Strain typing
MLST showed that all CR-AB isolates belonged to CC92. Among these, 91.7% (33/36) were identified as ST191 (Fig. 1). All isolates from Group non-CT belonged to ST191, whereas the three non-ST191 isolates were retrieved from Group CT. The 36 CR-AB isolates were assigned to 13 PFGE types (pulsotypes A to N) on the basis of banding patterns. With the exception of pulsotype A1, isolates of pulsotypes A, F, and H did not show any CR-related genetic mutations, and the majority (16/18) of the host patients belonged to Group non-CT. However, isolates of pulsotypes B, C, E, and G showed genetic mutations, and most (8/10) of the host patients belonged to Group CT.
3. PCR analysis and antimicrobial susceptibility
Only the blaOXA-23-like carbapenemase gene was found in all CR-AB isolates. No other carbapenemase genes were detected by PCR. ISAba1, located upstream from the blaOXA-51-like gene, was not detected. All CR-AB isolates were resistant to more than three antimicrobial classes by a disk diffusion susceptibility test (data not shown). All colistin MICs determined by the broth microdilution method were >128 µg/mL.
4. Mutations in genes associated with colistin resistance
All isolates with genetic mutations had mutated pmrB gene, and pmrB gene was the most frequently mutated (16/36, 44.4%). Genetic mutations were more frequently observed in isolates from Group CT than in those from Group non-CT patients (72.2% [13/18] and 11.1% [2/18], respectively, P≤0.001). CR-AB 32, 34, 35, and 36 had the same mutation in pmrB (g.938_939insTTT) and belonged to the same strain type according to both PFGE (pulsotype G) and MLST (ST191). These patients stayed in the same isolated location (ICU-H) for more than 47 days. Based on the available microbiological and clinical information, we concluded that these isolates were of the same strain, and clonal spread was revealed. Interestingly, with the exception of two isolates (CR-AB35 and 37), no genetic mutations were detected in isolates from Group non-CT.
5. Lipid A structure
Lipid A MALDI-TOF mass spectra of CR-AB isolates showed distinct profiles (Fig. 2). Common intensity peaks at m/z 1729, 1912, and 2139 were found in both CS-AB and CR-AB, indicating normal hexa-, hepta-, and octa-acylated lipid A, whereas in some isolates, the hexa- and octa-acylated lipid A peaks were completely changed to phosphoethanolaminated lipid A (hexa-acylated: 1/36, octa-acylated: 9/36). In each isolate, intensities of other lipid A components were compared with that of the hepta-acylated lipid A peak (Table 3), which was set as 100% because some hexa- and octa-acylated peaks were completely lost, and therefore were not suitable as reference peaks.
Peak intensity was significantly higher at two phosphoethanolamine (PE)-modified hexa-, hepta- and octa-acylated lipid A (P=0.008, <0.001, and 0.002, respectively) in isolates from Group CT than in those from Group non-CT. Furthermore, original hexa-acylated lipid A intensity peaks were significantly lower (P=0.021), and two PE-modified octa-acylated lipid A with 4C-H2O was significantly higher in isolates from Group CT (P=0.01).
The lipid A composition determined by MALDI-TOF M/S divided the CR-AB strains into strains harboring pmrB mutation and those having normal pmrB (Table 4). When lipid A peaks were grouped and analyzed according to the number of substitutions occurring in pmrB, pmrC, and lpsB genes, only the octa-acylated lipid A peak and its PE-modified forms were statistically different.
6. In vitro synergy testing
In vitro synergistic resistance effects (ΣFIC ≤0.5) for CR-AB isolates were most frequently observed for the colistin-meropenem combination (94.4%, 34/36) followed by the colistin-rifampicin combination (83.3%, 30/36). Although the synergy testing did not show significant differences (colistin-meropenem, P=0.467; colistin-rifampicin, P=0.655), combinations of colistin with meropenem or rifampin lowered the colistin MICs by 16-fold (range, 4–128-fold) and 8-fold (range, 4–128-fold), respectively (data not shown).
DISCUSSION
In our comparison of the characteristics of CR-AB clinical isolates recovered from CT and non-CT patients, no specific patient trait was found relevant to the clinical outcome. As for the mortality rate, the APACHE II score and bloodstream infections were two noteworthy markers that should be taken into consideration when managing CR-AB-infected patients. These findings were expected because the APACHE II scoring system is designed to measure disease severity in patients admitted to ICUs, and because bloodstream infections have a negative impact on patient outcome [35]. Although there were no significant differences in terms of patient characteristics, the causative CR-AB isolates presented obvious differences associated with CR, such as altered lipid A components, as indicated by MALDI-TOF M/S and genetic mutations associated with outer membrane modification.
Most of the CR-AB isolates from Group non-CT did not show any genetic mutations, whereas the revised lipid A component was characterized by shifted lipid A component peaks in MALDI-TOF M/S. Two potential hypotheses explain these unexpected results. First, the isolates may be a hetero-population composed of subpopulations of CR-AB and CS-AB lacking any evident genetic mutation, thus presenting with so-called heteroresistance [3]. Heteroresistance may be the primary stage, which in the presence of colistin, results in the proliferation of resistant subpopulations, and may prolong the treatment period or even lead to mortality [3153637]. The major subpopulation of CS-AB possibly produces erroneous colistin susceptibility data when using commercially automated systems and disk diffusion tests [3], whereas multiplication of the minor CR-AB subpopulation results in at least little growth in the presence of high concentrations of colistin by broth dilution, resulting in high MICs [336]. The different density of subpopulations might mask genetic mutations in CR-AB strains analyzed by Sanger sequencing. Similar findings have been demonstrated in Mycobacterium tuberculosis [3839]. PCR-based detection was not sufficient to identify heteroresistance, because minor allele frequencies of less than 15% were below the detection threshold of the method [40]. In addition, the CS-AB population feasibly flourished owing to better fitness in a colistin-free environment compared with the CR-AB population. As a consequence, the proportion switch of the two subpopulations might have produced ambiguous results. Secondly, although less likely, a novel mechanism conferring resistance to colistin might be involved. Since this study focused on genetic mutations in known CR-associated genes, unknown mechanisms of resistance might have been missed. Future studies should conduct a complex functional whole-genome analysis.
Regardless of the population heterogeneity, CR-AB was detectable by MALDI-TOF M/S, based on distinct spectra of modified lipid A compositions. Modification of lipid A by the addition of PE to the hexa-, hepta-, and octa-acylated lipid A has been suggested as a major mechanism of CR in AB. Similarly, even though some isolates exhibited unmodified lipid A peaks in this study, CR-AB displayed shifted peaks of one or two PE additives to the three lipid A moieties. Interestingly, the relative peak levels of PE-modified compared with unmodified lipid A components were much more elevated in Group CT. Notably, the relative peak levels of the two PE modified hepta-acylated lipid A moieties clearly separated the two groups.
Most of the CR-AB isolates from both groups showed a synergistic effect of colistin upon addition of meropenem or rifampin: synergism of both combinations was observed for most isolates, without any noticeable difference between combinations or between groups. Thus, combination treatment with either meropenem or rifampin should be considered for both CT and non-CT patients.
Out study has some limitations. As its main scope was to determine characteristics of CR-AB in clinical isolates and did not entail confirmation of heteroresistance, we could not confirm heteroresistant AB. Our data were collected from a single center in Korea, so the findings may not be generalized to other institutions. The limited number of CR-AB isolates precludes definitive conclusions on heterogeneous AB populations and CR.
Our study demonstrated that although there were no differences in clinical characteristics between Groups CT and non-CT, there were pathological differences, including those involving characteristics useful in diagnosing CR-AB. Population heterogeneity masked resistance-causing genetic mutations, traditionally determined by Sanger sequencing, especially in Group non-CT; therefore, to identify CR, accurate testing methods reflecting physiological alteration of the bacteria, such as PE-modified lipid A identification by MALDI-TOF M/S, should be carried out. Since colistin heteroresistance is common in patients without prior drug treatment and can be caused by better bacterial fitness in the colistin-free environment, lipid A analysis shows clearer results for CR-AB isolates. Broth microdilution was found to accurately determine CR in AB regardless of population heterogeneity, which prevented exact susceptibility interpretation because of the subpopulations of CR-AB. Furthermore, combination treatment, specifically with meropenem and rifampicin, should be considered for the treatment of CR-AB infections.
XML Download