In Vitro Synergistic Activity of Antimicrobial Agents in Combination against Clinical Isolates of Colistin-Resistant Acinetobacter baumannii

Emerging resistance to colistin in clinical Acinetobacter baumannii isolates is of growing concern. Since current treatment options for these strains are extremely limited, we investigated the in vitro activities of various antimicrobial combinations against colistin-resistant A. baumannii. Nine clinical isolates (8 from bacteremia cases and 1 from a pneumonia case) of colistin-resistant A. baumannii were collected in Asan Medical Center, Seoul, South Korea, between January 2010 and December 2012. To screen for potential synergistic effects, multiple combinations of two antimicrobials among 12 commercially available agents were tested using the multiple-combination bactericidal test (MCBT). Checkerboard tests were performed to validate these results. Among the 9 colistin-resistant strains, 6 were pandrug resistant and 3 were extensively drug resistant. With MCBT, the most effective combinations were colistin-rifampin and colistin-teicoplanin; both combinations showed synergistic effect against 8 of 9 strains. Colistin-aztreonam, colistin-meropenem, and colistin-vancomycin combinations showed synergy against seven strains. Colistin was the most common constituent of antimicrobial combinations that were active against colistin-resistant A. baumannii. Checkerboard tests were then conducted in colistin-based combinations. Notably, colistin-rifampin showed synergism against all nine strains (100%). Both colistin-vancomycin and colistin-teicoplanin showed either synergy or partial synergy. Colistin combined with another β-lactam agent (aztreonam, ceftazidime, or meropenem) showed a relatively moderate effect. Colistin combined with ampicillin-sulbactam, tigecycline, amikacin, azithromycin, or trimethoprim-sulfamethoxazole demonstrated limited synergism. Using MCBT and checkerboard tests, we found that only colistin-based combinations, particularly those with rifampin, glycopeptides, or β-lactams, may confer therapeutic benefits against colistin-resistant A. baumannii.

A cinetobacter baumannii is regarded as an important nosocomial pathogen causing various infections, including ventilator-associated pneumonia, bloodstream infections, surgical site infections, and urinary tract infections (1). It has become more problematic by developing resistance to a wide range of antimicrobials, including carbapenems (2)(3)(4)(5). Colistin, the most active agent against multidrug-resistant (MDR) Gram-negative pathogens in vitro, has been reintroduced for the treatment of carbapenem-resistant A. baumannii (6). Unfortunately, colistin-resistant A. baumannii strains have been reported recently (7). As these strains are simultaneously resistant to most antimicrobial agents, treatment options for them are extremely limited (8). A few previous studies evaluated the in vitro synergism of antimicrobial combinations against colistin-resistant A. baumannii (9)(10)(11). In those studies, however, the number of antimicrobial agents tested did not exceed four, and only colistin-based combinations were tested. In real clinical practice, colistin-associated nephrotoxicity occurs in about 40% of treated patients, and colistin therapy is frequently stopped because of this (8,12,13). Therefore, the in vitro efficacy of non-colistin-based combinations against colistinresistant A. baumannii strains should also be evaluated. The aim of this study was to assess the in vitro efficacy of antimicrobial combinations, among 12 commercially available antimicrobial agents, against clinical isolates of colistin-resistant A. baumannii using the multiple-combination bactericidal test (MCBT) and checkerboard method.
Susceptibility testing and interpretation. In vitro antimicrobial susceptibility testing was performed in triplicate using the broth microdilution method according to the Clinical and Laboratory Standards Institute (CLSI) guidelines (14). Fresh Mueller-Hinton broth was used for all susceptibility testing. CLSI susceptibility criteria were used, except with azi-thromycin, aztreonam, vancomycin, teicoplanin, tigecycline, and rifampin. No susceptibility breakpoints for rifampin and tigecycline are given in the CLSI guidelines; therefore, CLSI criteria recommended for staphylococci were applied to rifampin (MIC Ն 4 mg/liter as resistance), and European Committee on Antimicrobial Susceptibility Testing (EUCAST) criteria for Enterobacteriaceae were used for tigecycline (MIC Ͼ 2 mg/liter as resistance) (28). Escherichia coli ATCC 25922 was used as a reference strain, and all results determined with this strain were within the CLSI quality control ranges. The category of extensively drug-resistant (XDR) strains was defined as nonsusceptibility to at least one agent in all but two or fewer antimicrobial categories, and pandrug-resistant (PDR) was defined as nonsusceptibility to all antimicrobial agents (29).
MCBT. The multiple-combination bactericidal test (MCBT) was performed to test combinations of two antimicrobials as previously described (32)(33)(34)(35). Combinations of two antimicrobials were placed in 96-well, round-bottomed microtiter plates (Nunc Inc., Roskilde, Denmark). The antimicrobial agents were prepared in Mueller-Hinton II cation-adjusted broth (MHB II; Becton, Dickinson Microbiology Systems, Cockeysville, MD) at 10 times the required concentrations. One or two antimicrobial agents were added, each in 10-l volumes, to the wells. The necessary volume of MHB II was then added to the wells containing antimicrobial agents. The A. baumannii inocula consisted of 70 l of a 100-fold dilution of a 0.5 McFarland turbidity standard prepared during the growth phase of culture in tryptone soya broth (Oxoid Laboratories, Basingstoke, United Kingdom). The final inoculum concentration was 5 ϫ 10 5 CFU/ml in each well. Growth and sterility control wells (no antibiotic and no inoculum, respectively) were included in all plates. Plates were incubated at 35°C for 48 h. At 24 and 48 h, the wells were examined for turbidity. Each well with no visible growth at 48 h was subcultured to establish whether 99.9% killing was achieved. Reproducibility of the MCBT results was confirmed in triplicate. For the purposes of the MCBT analysis, combinations were considered synergistic if bactericidal activity (99.9% killing) was achieved when the two agents were tested in combination.
The final concentrations of antimicrobials selected for MCBT corresponded to the criteria for resistance (35). The antimicrobial agents were used in MCBT at the following fixed concentrations: colistin at 2 mg/liter, ampicillin-sulbactam at 16/8 mg/liter, amikacin at 16 mg/liter, azithro-mycin at 4 mg/liter, aztreonam at 16 mg/liter, ceftazidime at 16 mg/liter, meropenem at 8 mg/liter, rifampin at 2 mg/liter, tigecycline at 2 mg/liter, trimethoprim-sulfamethoxazole at 4/76 mg/liter, vancomycin at 4 mg/ liter, and teicoplanin at 16 mg/liter. Synergy testing of colistin combinations with the checkerboard method. To identify synergistic effects, the checkerboard synergy test was performed in triplicate in 96-well microtiter plates containing colistin and 1 of 11 other antimicrobials. Each antimicrobial was diluted using an automated dilutor, with concentrations ranging from 0.031ϫ MIC to 4ϫ MIC. The initial inoculum was approximately 5 ϫ 10 5 CFU/ml. Microtiter trays were incubated at 35°C for 48 h under aerobic conditions (36).
After incubation, any well showing turbidity was considered to exhibit microbiological growth. The fractional inhibitory concentration index (FICI) was calculated for each antibiotic in each combination. The mean FICI of all nonturbid wells, along the turbidity/nonturbidity interface, was then calculated (37). The FICI results for each combination against each test isolate were interpreted as follows: FICI of Յ0.5, synergism; FICI of between 0.5 and 1, partial synergism; FICI of Ն1 but Ͻ4, indifference; FICI of Ն4, antagonism (38,39).

Microbiological and genotypic characteristics of colistin-resistant A. baumannii.
Of nine colistin-resistant A. baumannii strains, eight were blood isolates and one was a sputum isolate. All of the strains were also resistant to carbapenems. Results of MLST, carbapenemase types, and MICs of antimicrobials against each strain are summarized in Table 1 and in Table S1 in the supplemental material. All of the tested strains carried the OXA-51 gene, and OXA-23 was detected in seven strains (78%). Eight of nine strains had the IMP-1 gene encoding a metallo-␤-lactamase. By MLST, 7 strains were found to belong to ST191, while the remaining two were ST357. Six of nine strains were resistant to all classes of antimicrobials (PDR), and the remaining three A. baumannii strains were XDR.
MCBT. Using the MCBT method, each two-drug combination was tested ( Table 2). The most effective combination regimens were colistin-rifampin and colistin-teicoplanin, both of which showed synergy against eight of nine strains. The colistin-aztreonam, colistin-meropenem, and colistin-vancomycin combinations were synergistic against seven strains. All of the regimens exhibiting synergistic effect against at least four strains included colistin. Other combinations were active against two or fewer strains. Among the colistin-based combinations, only colistintigecycline was not synergistic against any of the strains tested.
Checkerboard synergy test. Since only colistin-based regi- mens were highly effective in the MCBT, checkerboard tests were performed to validate presence of synergism among these combination regimens. As shown in Table 3, results of the checkerboard synergy analysis of colistin-resistant A. baumannii were similar to those of MCBT. The colistin-rifampin combination was fully synergistic against nine of the A. baumannii strains tested. The combinations of colistin-vancomycin and colistin-teicoplanin showed either synergy or partial synergy against all strains. However, colistin-vancomycin (6/9, 67%) was more frequently synergistic than colistin-teicoplanin (4/9, 45%). With colistin-aztreonam and colistin-ceftazidime, and with colistin-meropenem, 7 (78%) strains exhibited synergy and partial synergy, respectively. Colistin combinations with ampicillin-sulbactam, tigecycline, azithromycin, and trimethoprim-sulfamethoxazole were synergistic against only one strain. Colistin-tigecycline and colistin-azithromycin showed indifference against seven and eight strains, respectively. No antagonistic interactions were observed with any of the combinations evaluated. Clinical characteristics and treatment outcomes. The clinical characteristics and treatment outcomes of patients with colistinresistant A. baumannii infections are summarized in Table 4. Most patients had severe underlying diseases, such as malignancy, hematologic disease, liver transplantation, and acute liver failure related to a hepatitis B virus (HBV) flare-up. All nine patients were nosocomially infected with A. baumannii, and 7 of 9 patients experienced an intensive care unit (ICU) stay. Four of the nine patients had a history of prior colistin use, and all of the patients had previously used carbapenems. Antibiotic regimens and empirical treatment outcomes varied by patient. Three patients were treated with colistin-based combinations, and microbiological eradication was achieved in two patients. The mortality rate was high, and most patients (67%) died within 14 days.

DISCUSSION
The main purpose of this study was to assess the in vitro synergistic effects of antimicrobial combinations against colistin-resistant A. baumannii. Combinations of commonly used antimicrobial agents were tested by MCBT, and synergistic results were confirmed using the checkerboard method. By MCBT, colistin was determined to be the most common constituent of antimicrobial combinations that were active against colistin-resistant A. baumannii. Non-colistin-based combinations were not active against these strains.   a Other antimicrobial combinations that are not shown (e.g., CST ϩ TGC) were not synergistic against any of the strains tested. b If an XDR strain (c, e, or h) was killed because the drug MIC for the strain was equal to or lower than the tested concentration of an antimicrobial agent, in an antimicrobial combination that included this agent, the strain was not listed.   studies reported that colistin-resistant A. baumannii strains had higher susceptibility rates for the majority of antimicrobial agents than colistin-susceptible strains (40,41). In contrast, antimicrobial agents showed high MICs against colistin-resistant strains in the current study and the recent study by Qureshi et al. (8). These differences were probably due to frequent simultaneous exposure to carbapenems, vancomycin, and colistin.
Colistin with rifampin has been the most frequently studied combination in vitro (7). Although a recent randomized clinical trial failed to show a difference in outcomes between colistinrifampin and colistin monotherapies against XDR A. baumannii, the microbiological eradication rate was significantly higher in the combination arm (42). In the present study, a strong synergistic effect from colistin combined with rifampin was shown in both the MCBT and the checkerboard test. Notably, with the checkerboard test, colistin-rifampin was found to be fully synergistic (FICI Յ 0.5) against all nine (100%) A. baumannii strains. Therefore, the clinical efficacy of colistin-rifampin should be further evaluated in colistin-resistant A. baumannii infections.
Glycopeptide MICs of tested strains were higher than those of two previous studies indicating relatively low MICs of glycopeptides against colistin-resistant A. baumannii (43,44). Albeit with high MICs against our strains, vancomycin and teicoplanin consistently showed synergism in combination with colistin, in accordance with previous in vitro and in vivo studies (27,43,44). We conjectured that glycopeptides might be effective in combination with colistin, regardless of its MIC, because of an adjuvant permeabilizing effect of colistin on the A. baumannii outer membrane. In this regard, other cell wall-active agents such as ceftazidime, aztreonam, and meropenem also tended to show synergistic effects in our tests.
Tigecycline, regarded as an effective treatment option for MDR A. baumannii infections, showed low antimicrobial activity against colistin-resistant strains in the present study. Tigecyclinecontaining combinations did not show synergistic effect against any of the strains in MCBT, even in combination with colistin. Colistin-tigecycline showed only limited synergistic effects by the checkerboard test. Cheng et al. reported a higher adjusted 14-day mortality rate in the colistin-tigecycline combination treatment group than in the colistin-carbapenem treatment group in one prospective, observational study of XDR A. baumannii bacteremia (45). They deduced that tigecycline was less effective because this agent targets the 30S ribosomal subunit, not the cell wall.
Our study had several limitations. All tested strains were collected from a single tertiary center, and the mechanism of colistin resistance was not evaluated, which limits our ability to generalize from these results. However, results of the synergy tests performed on study strains were similar to those of previous colistin-based studies. In addition, FICIs from the checkerboard test can differ, depending on the various methods used for interpretation (46). Finally, this was an in vitro study that did not test clinical outcomes; clinical studies are needed to confirm our findings.
In conclusion, using MCBT and checkerboard testing, we found that only colistin-based combinations, particularly combinations with rifampin, glycopeptides, or ␤-lactams, should be expected to confer therapeutic benefits in colistin-resistant A. baumannii infections. The development of new antimicrobial agents is urgently needed to treat infections by this pathogen.