In vitro activity of antibiotic combinations against multidrug-resistant strains of Acinetobacter baumannii and the effects of their antibiotic resistance determinants
Correspondence: Yoko Miyasaki, Department of Medicine, Cedars-Sinai Medical Center, 8700 Beverly Blvd. B-220, Los Angeles, CA 90048, USA. Tel.: +1 310 423 3896; fax: +1 310 423 4599; e-mail: email@example.com
Various combinations of antibiotics are reported to show synergy in treating nosocomial infections with multidrug-resistant (MDR) Acinetobacter baumannii (A. baumannii). Here, we studied hospital-acquired outbreak strains of MDR A. baumannii to evaluate optimal combinations of antibiotics. One hundred and twenty-one strains were grouped into one major and one minor clonal group based on repetitive PCR amplification. Twenty representative strains were tested for antibiotic synergy using Etest®. Five strains were further analyzed by analytical isoelectric focusing and PCR to identify β-lactamase genes or other antibiotic resistance determinants. Our investigation showed that the outbreak strains of MDR A. baumannii belonged to two dominant clones. A combination of colistin and doxycycline showed the best result, being additive or synergistic against 70% of tested strains. Antibiotic additivity was observed more frequently than synergy. Strains possessing the same clonality did not necessarily demonstrate the same response to antibiotic combinations in vitro. We conclude that the effect of antibiotic combinations on our outbreak strains of MDR A. baumannii seemed strain-specific. The bacterial response to antibiotic combinations is probably a result of complex interactions between multiple concomitant antibiotic resistance determinants in each strain.
Fully active antibiotic options available to treat nosocomial infections with multidrug-resistant (MDR) Acinetobacter baumannii (A. baumannii) are extremely limited (Perez et al., 2007). With its capacity to demonstrate multiple mechanisms of resistance to various antibiotic classes, clinicians predict that antibiotics will soon be unavailable to treat serious MDR A. baumannii infections (Boucher et al., 2009). Select antibiotic combinations reportedly show synergy, that is, significantly greater activity provided by two antibiotics combined than the sum of each antibiotic's activity, against MDR A. baumannii infections (Rahal, 2006). Examples of such combinations include imipenem (IMP)–rifampin (RIF) (Tripodi et al., 2007; Song et al., 2009; Panchón-Ibáñez et al., 2010), carbapenem–colistin (COL) (Timurkaynak et al., 2006), COL–RIF (Hogg et al., 1998; Giamarellos-Bourboulis et al., 2001; Timurkaynak et al., 2006; Li et al., 2007; Tripodi et al., 2007), and COL–doxycycline (DOX) (Timurkaynak et al., 2006). Two small clinical studies showed very good and limited efficacy of COL–RIF and IMP–RIF combinations, respectively, for patients with A. baumannii infections (Motaouakkil et al., 2006; Saballs et al., 2006).
The mechanism of synergy between antibiotic combinations against MDR A. baumannii, however, is undetermined. For example, A. baumannii is intrinsically resistant to RIF, and we hypothesized that the reported synergistic effect of combinations containing RIF comes from RIF potentiating the other antibiotic by interfering with mRNA production. Acinetobacter baumannii strains infamously carry a multitude of antibiotic resistance determinants, either on their chromosome or on their plasmids (Perez et al., 2007), and it is conceivable that not all strains or even strains within the same clone respond to antibiotic combinations equally.
During 2006 and 2007, Cedars-Sinai Medical Center in Los Angeles, CA, USA, experienced an outbreak of MDR A. baumannii. The outbreak was terminated by a ‘bundle approach’ of strict infection control measures (Murthy et al., 2008). To provide insight into approaches for treatment of MDR A. baumannii, we evaluated dual combinations of antibiotics for possible synergy against our outbreak strains of MDR A. baumannii using Etest. Although the correlation between the Etest and time-kill methods for in vitro testing of antimicrobial combinations on A. baumannii is reported as 72% (Bonapace et al., 2000), we chose Etest because it is less labor-intensive than time-kill assays and may facilitate rapid clinical decisions. Additionally, our study aimed to determine whether our clonal strains would respond to antibiotic combinations equally and to investigate the effects of β-lactamases (blas) and other antibiotic resistance determinants in select strains on their response to these antibiotic combinations. We screened for β-lactamase genes, including blaTEM, blaSHV, blaPER, blaADC, blaIMP, blaVIM, blaOXA-23,blaOXA-Ab (housekeeping gene belonging to the blaOXA-51/69 families), and blaOXA-58, and for the genes encoding aminoglycoside-modifying enzymes (AMEs) including aphA6, aadA1, aadB, aacC1, and aacC2 (Hujer et al., 2006). Additional determinants included integrase genes (intII, intI2, and intI3), to detect the presence of integrons that can express gene cassettes encoding multidrug resistance; disruptions in the outer membrane protein, CarO, associated with resistance to carbapenems; and the presence of adeR, the gene that encodes a regulator for the AdeABC efflux pump whose enhanced expression confers bacterial resistance to aminoglycosides, fluoroquinolones, tetracyclines, and trimethoprim. We also assayed the strains for the presence of mutations in the quinolone resistance–determining regions (QRDRs) of gyrA gene encoding GyrA subunit of DNA gyrase and parC gene encoding ParC subunit of topoisomerase IV.
Materials and methods
Clinical strains and clonal analysis
We prospectively collected 121 consecutive single-patient MDR A. baumannii clinical strains during 2006 and 2007 at Cedars-Sinai Medical Center. We considered a strain as MDR if it was resistant to two or more antibiotic classes that included anti-pseudomonal penicillin and its combination with β-lactamase inhibitor (e.g. piperacillin/tazobactam), anti-pseudomonal cephalosporins (e.g. ceftazidime or cefepime), carbapenems (e.g. IMP), aminoglycosides [e.g. tobramycin or amikacin (AN)], and fluoroquinolones (e.g. ciprofloxacin or levofloxacin) based on VITEK® 2 (bioMérieux, Inc.). All 121 strains were analyzed by repetitive PCR (rep-PCR) amplification using the DiversiLab®Acinetobacter Fingerprinting Kit according to manufacturer's instructions (bioMérieux, Inc.). Briefly, bacterial DNA was extracted using UltraClean™ Microbial DNA Isolation Kit (MO BIO Laboratories, Inc.). Amplification reactions were performed in the GeneAmp® PCR System 9700 under the following conditions: 2 min at 94 °C, 35 cycles of denaturation (30 s at 94 °C), annealing (30 s at 50 °C) and extension (90 s at 70 °C), and a final extension of 3 min at 70 °C. Rep-PCR products were separated by electrophoresis using the Agilent 2100 Bioanalyzer (Agilent Technologies, Inc.). Band patterns for each strain were aligned and interpreted with web-based DiversiLab software provided by the manufacturer (bioMérieux, Inc.). Strains were grouped by ≥ 95% similarity.
Medical record review identified an incident episode of nosocomial acquisition according to Centers for Disease Control surveillance definitions (Horan et al., 2008). Accordingly, 19 strains from patients with evidence of infection or colonization with A. baumannii prior to or at the time of admission to our institution during the study period were considered as having either a repeat episode or non-nosocomial A. baumannii infection, respectively, and their clinical strains were excluded from this study. Of the remaining strains, those belonging to the two prevalent clones, A and B, were selected for further analyses.
Etest (bioMérieux, Inc.) was performed on 33 representative strains that were resistant to at least three classes of antibiotics (26 of clone A and seven of clone B) for susceptibility to IMP, COL, AN, DOX, tigecycline (TGC), RIF, and azithromycin (AZT) as per manufacturer's recommendations. The minimum inhibitory concentrations (MICs) of IMP, COL, and AN were interpreted according to Clinical & Laboratory Standards Institute (CLSI) (2009) breakpoints. MICs of TGC (MICTGC) were interpreted as per the US Food and Drug Administration's breakpoint recommendations for Enterobacteriaceae. MICs of RIF (MICRIF) and AZT (MICAZT) were interpreted using CLSI breakpoints for Haemophilus influenzae (Clinical & Laboratory Standards Institute, 2009). The test was conducted in triplicate. Escherichia coli ATCC 25922 and Pseudomonas aeruginosa ATCC 27853 strains were used as controls.
In vitro testing of antimicrobial combinations
Of the 33 strains from the susceptibility testing, 13 clone A strains and seven clone B strains were randomly chosen for the in vitro testing of the following combinations using Etest: IMP–RIF, IMP–COL, IMP–DOX, IMP–AN, IMP–AZT, COL–RIF, COL–DOX, COL–AN, COL–TGC, TGC–AN, and TGC–AZT. Two strips, one containing antibiotic A and another containing antibiotic B, were aligned at 90° at the respective MIC (mg L−1) of two antibiotics, MICA and MICB, as previously described (White et al., 1996). To evaluate interactions between antibiotics, the fractional inhibitory concentration (FIC) and FIC index were calculated as previously described (White et al., 1996; Tascini et al., 1998; Timurkaynak et al., 2006; Tan et al., 2007). Antibiotic combinations were evaluated based on the FIC index, which ranges as follows: synergistic if ≤ 0.5; additive if > 0.5 but < 1; indifferent if ≥ 1 but < 4; and antagonistic if ≥ 4. The effect of adding RIF on the mean MICs of IMP (MICIMP) and of COL (MICCOL) for the same 20 strains was analyzed using two-tailed paired t-test with 95% confidence interval. The tests were performed in duplicate, with additional tests performed until two identical results were selected as confirmed. No fifth test was required.
Analytical isoelectric focusing and PCR of antibiotic resistance determinants
Three clone A strains and two clone B strains from the in vitro testing of antimicrobial combinations were analyzed using analytical isoelectric focusing (aIEF) as previously described (Paterson et al., 2001). These strains had different MICs of IMP, AN, AZT, COL, DOX, RIF, and TGC. Using PCR we screened for β-lactamase genes (blaTEM, blaSHV, blaPER, blaADC, blaIMP, blaVIM, blaOXA-23,blaOXA-Ab, and blaOXA-58) and genes encoding AMEs (aphA6, aadA1, aadB, aacC1, and aacC2). Additional determinants included integrase genes (intII, intI2 and intI3); disruptions in the outer membrane protein gene, carO; and the presence of adeR. We also examined the mutations in QRDRs of gyrA and parC by direct sequencing of the PCR products. All determinants were PCR-amplified using established primers and controls previously described (Hujer et al., 2006).
Clinical strains and clonal analysis
According to our medical record review, of 121 MDR A. baumannii strains, 102 were hospital-acquired. Based on rep-PCR, 76 belonged to a major clone A and eight to a minor clone B. The study strains’ clonality, sources, and antimicrobial susceptibility data based on VITEK 2 are summarized in Supporting Information, Table S1.
Based on Etest, non-susceptibility of the tested strains was 97% to IMP; 9% to COL; 85% to AN; 97% to DOX; and 58% to TGC. The results of VITEK 2 and Etest were disconcordant with respect to the susceptibility of the study strains to IMP and AN (data not shown). Strains belonging to the same clone had different MICs. For example, MICIMP of clone A strains ranged from 1.5 to > 32 mg L−1, and MICAN among clone A strains ranged from 2 to > 256 mg L−1 (complete data now shown). Three strains belonging to clone A were resistant to COL, with MICsCOL of 24, 128, and 256 mg L−1.
In vitro testing of antimicrobial combinations
The MIC values, both original and those resulting from combining two antibiotics, are presented in Table S2. The combination of COL–DOX showed the best result, being additive or synergistic to 70% of tested strains. Synergy was observed in four instances: the COL–DOX combination to strain 12 (clone A) and strain 19 (clone B); the IMP–COL combination to strain 12; and the COL–RIF combination to strain 12. The IMP–RIF, IMP–COL, and IMP–AZT combinations had different effects on tested strains depending on their clonality (Table 1). For example, the IMP–RIF combination was additive to five clone B strains, but not to any clone A strains. Conversely, the IMP–COL combination was additive to four clone A strains, but not to any clone B strains. For clone A, the addition of RIF did not result in a significant reduction in the mean MICIMP (P = 0.34) or the mean MICCOL (P = 0.24), while for clone B, the addition of RIF resulted in a significant reduction in MICIMP (P << 0.05) and the mean MICCOL (P << 0.005). Combinations containing COL showed the following results for two clone A, COL-resistant strains: for strain 12 (MICCOL = 128 mg L−1), the IMP–COL, COL–RIF, and COL–DOX combinations were synergistic, while the COL–AN and COL–TGC combinations were indifferent. For strain 13 (MICCOL = 24 mg L−1), the IMP–COL, COL–RIF, and COL–DOX combinations were additive, while the COL–AN and COL–TGC combinations were indifferent.
Table 1. Additive effects of antibiotic combinations observed
|IMP and RIF||5/25%a||0/0%||5/71%|
|IMP and COL||4/20%||4/31%||0/0%|
|IMP and AZT||8/40%||2/15%||6/86%|
|COL and RIF||8/40%||4/31%||4/57%|
|COL and DOX||12/60%||6/46%||6/86%|
|TGC and AZT||1/5%||0/0%||1/14%|
aIEF and PCR screening
The results of aIEF and PCR screening on five strains (three from clone A and two from clone B) are summarized in Table S3. Overall, the results of the aIEF and PCR screening were consistent with each other. To illustrate, the β-lactamase ‘band’ detected at isoelectric point (pI) of 5.0 (observed in strains 12 and 13) corresponds to the PER β-lactamase. The pIs of 5.3, 5.4, 5.5, and 5.6 may represent the TEM β-lactamases (seen in strains 11, 12, 13, and 15), and the pI of 6.3 most likely corresponds to the OXA-Ab β-lactamase, while the band at pI of > 9.0 corresponds to the ADC β-lactamase. PCR amplification confirmed the presence of the genes encoding these enzymes. The only exception is strain 16, which had a band at pI of 5.6 but was negative for the TEM β-lactamase.
Based on VITEK 2 and/or Etest, these strains were non-susceptible to the tested anti-pseudomonal penicillins in combination with a β-lactamase inhibitor, anti-pseudomonal cephalosporins, carbapenems, aminoglycosides, fluoroquinolones, and tetracyclines according to CLSI (2009) breakpoints (data not shown).
All strains were resistant to aminoglycosides owing to the presence of genes encoding AMEs and to fluoroquinolones owing to both Ser83Leu substitution in GyrA and Ser80Phe substitution in ParC. All strains had the adeR gene, indicating the possibility that there could be enhanced expression of the AdeABC efflux pump and accounting for non-susceptibility to fluoroquinolones, aminoglycosides, and tetracyclines. aIEF and PCR screening did not suggest a carbapenemase in tested strains although all were highly carbapenem-resistant. We suspect the presence of an insertion sequence (IS) upstream of the blaOXA-Ab gene, which can increase the expression of the OXA-Ab β-lactamase (Poirel & Nordmann, 2006). Similarly, non-susceptibility to anti-pseudomonal penicillins in combination with a β-lactamase inhibitor and to anti-pseudomonal cephalosporins could be due to the presence of an IS upstream of the blaADC gene, which increases the expression level of the ADC β-lactamase (Heritier et al., 2006).
In the context of carbapenem resistance and efflux pump, the IMP–DOX combination was indifferent to all given strains. On the other hand, the COL–DOX combination was additive or synergistic to four of five strains. The AN-containing antibiotic combinations, IMP–AN, COL–AN, and TGC–AN, were indifferent to tested strains, all of which had a single gene encoding AME and were resistant to AN. Strains exhibiting the same profile of bla genes on our aIEF and PCR screening did not show the same response to β-lactam-containing combinations. For example, strains 12 and 13 showed the same pattern of bla genes and were resistant to COL. In the presence of COL, MICIMP of strain 12 decreased by 81% (i.e. from 32 to 6 mg L−1), while that of strain 13 decreased by only 50% (i.e. from 32 to 16).
The effects of antibiotic combinations on our MDR A. baumannii strains appeared to be strain-specific, regardless of clonality. Even two strains belonging to the same clone could possess different antibiotic resistance determinants and hence demonstrate different responses to antibiotic combinations. In the presence of a gene encoding AME and conferring AN resistance, all AN-containing combinations were consistently indifferent. This observation renders an AN-containing combination a poor candidate for empirical treatment for AN-resistant, MDR A. baumannii. Combining IMP and DOX did not appear to modify the effect of carbapenem resistance or efflux pump. On the other hand, the COL–DOX combination was additive or synergistic to four of five strains. We speculated that COL might have attenuated the effect of efflux pump, reducing MICDOX.
Clinicians will consider combination antibiotic therapy against MDR A. baumannii, particularly if the strain is also resistant to COL. Three combinations (IMP–COL, COL–RIF, and COL–DOX) were additive or synergistic to the COL-resistant strains in this study; however, interpretation of these results requires caution. To illustrate, strain 12 to which the IMP–COL combination was synergistic was highly resistant to both IMP (MICIMP > 32 mg L−1) and COL (MICCOL = 128 mg L−1). Combining IMP and COL decreased MICIMP from 32 to 6 mg L−1 and MICCOL from 128 to 32 mg L−1. This result yielded an FIC index of 0.44, meeting the definition of synergy. However, as per CLSI breakpoint, MICIMP of 6 mg L−1 against A. baumannii indicates IMP non-susceptibility, while MICCOL of 32 mg L−1 against A. baumannii indicates COL resistance. Therefore, this combination was considered clinically insignificant. The same conclusion applies to the other synergistic combinations that were observed in this study.
We conclude that the effect of antibiotic combinations on our outbreak strains of MDR A. baumannii seemed highly strain-specific. The complex genetic background of each A. baumannii strain seems to exert differential effects on bacterial response to antibiotic combinations. The choice of antibiotic combinations should be dictated by results of susceptibility tests performed on each strain. Further investigations are warranted to ascertain the molecular basis of the COL-DOX synergy.
This project was supported by an investigator-initiated grant from Merck. We thank the Cedars-Sinai Microbiology Laboratory and Hospital Epidemiology Department staff for assistance in technical aspects and data collection, respectively. We thank Drs. Michael Jacobs, Andrea Endimiani, and Ms. Saralee Bajaksouzian of Case Western Reserve University for assistance with MICs.
A portion of this manuscript was presented at the 45th Annual Meeting of the Infectious Disease Society of America (2007, San Diego, CA). Y.M. is partially supported by the Cedars-Sinai Clinical Scholars’ Funding Award. R.A.B. is supported by the VISN 10 Geriatric Research Education and Clinical Care Center (GRECC), Merit Review Program of the Veterans Administration, and the National Institute of Health (R01AI072219-05). All other authors have no financial disclosures.