Susceptibility Testing of Colistin for Acinetobacter baumannii: How Far Are We from the Truth?

Acinetobacter baumannii is involved in life-threatening nosocomial infections, mainly in the intensive care units (ICUs), and often colistin may represent the last therapeutic opportunity. The susceptibility to colistin of 51 epidemiologically typed A. baumannii strains isolated in 2017 from clinical samples of patients hospitalized in the ICU of a tertiary care academic hospital was investigated. All isolates were carbapenem-resistant due to the presence of the blaOXA-23 gene in sequence group 1 (international clonal lineage II) and sequence group 4 (related to international clonal lineage II) isolates, and to the blaOXA-24/40 gene in sequence group 2 (international clonal lineage I) isolates. Vitek®2, agar diffusion, and broth microdilution tests showed major discordancy (≥2 dilution factors) in the minimum inhibitory concentration (MIC) values for colistin in 24 out of 51 isolates, resulting in erroneous reporting of qualitative susceptibility data for eight isolates. In growth kinetics experiments in the presence of colistin, five isolates grew with drug concentrations above the susceptibility breakpoint when incubated for >12 h, and three isolates showed the presence of heteroresistant subpopulations. This study highlights that the high frequency of isolation of carbapenem-resistant A. baumannii strains in high-risk infectious wards requires an accurate application of methods for detecting susceptibility to antibiotics, in particular to colistin, so as to ensure a correct therapeutic approach.


Introduction
Acinetobacter baumannii and species of the A. baumannii complex are Gram-negative coccobacilli involved in the etiology of nosocomial infections and manifesting an important antibiotic resistance. A. baumannii is the cause of a wide range of clinical manifestations, including hospital-acquired pneumonia (HAP), followed by urinary tract infections (UTIs), bloodstream infections (BSIs), and surgical site infections (SSIs), which occur preferentially in patients admitted to intensive care units (ICUs), where they have five to ten times higher risk of acquiring hospital infections [1].
Several predisposing factors have been identified, including previous antibiotic therapy, which can alter normal bacterial flora and select resistant strains. Furthermore, crosstransmission among patients and the hospital environment are the most likely sources of infection. Colonization or infection sustained by A. baumannii has been invoked as an independent risk factor of mortality [2], though the debate around the direct attribution of mortality to infections from A. baumannii is still open.
A number of A. baumannii isolates are highly resistant to most of the clinically available antibiotics [3]. Carbapenems have been used for decades as the antibiotics of choice in the clinical management of the infections caused by this organism. Yet, resistance to carbapenems emerged, leaving very few therapeutic options [4]. Class D (OXA-type) carbapenemases are the main enzymes responsible for this resistance. The OXA carbapenemases found in A. baumannii complex species belong to four subgroups: OXA-23-like, OXA-24-like, OXA-51-like, and OXA-58. The OXA-51-like subgroup is intrinsic to A. baumannii, occurring in most or all of the strains [5,6]. Other OXA enzymes have been detected with variable frequency in the major epidemic lineages, including the international clones (ICL) I, II, and III [7].
Recently, colistin has become one of the major therapeutic options in the management of carbapenem-resistant A. baumannii infections. However, colistin resistance has rapidly emerged in A. baumannii after reintroduction of this drug into the clinical practice [4]. To complicate the picture, observations have recently been made about the congruity of the results of the susceptibility profiles and the values of the minimum inhibitory concentration (MIC) obtained by automated diagnostic devices and diffusion methods [8]. These observations were assumed on the basis of the discontinuity of the diffusion of colistin in agar, which often renders untrustworthy the reading of the inhibition halos and of all the interpretations related to the diffusion method (for example E-test ® and various semiautomatic systems). As confirmation of this problem, a caveat for the interpretation of the susceptibility tests obtained through these techniques has been issued by the European Committee on Antimicrobial Susceptibility Testing (EUCAST), with the indication to perform sensitivity tests using the microdilution method, considered as a reference method [9][10][11][12].
One mechanism of colistin resistance involves reduced colistin binding to the bacterial outer membrane due to the charge modification of the lipopolysaccharide (LPS) moiety [13]. A second mechanism is the complete loss of lipid A, which also prevents the interaction of the antibiotic with the bacterial outer membrane [14]. In A. baumannii infections, another problem that may cause therapeutic failure is the emergence of resistance to colistin in subpopulations from an otherwise susceptible (MIC ≤ 2 mg/L) population, a phenomenon defined as colistin heteroresistance, often associated with previous use of the drug [15][16][17].
In this work, a collection of well-characterized, multidrug-resistant A. baumannii isolates obtained from clinical specimens of infected or colonized patients was used to compare the performances of both commercial and in-house prepared colistin susceptibility tests. Substantial discrepancies between an automated testing method and both agar diffusion and in-house broth microdilution methods were observed. Our findings highlight the need for confirmative colistin susceptibility testing by the broth microdilution method, especially in the case of borderline MICs, which are increasingly frequent among multidrugresistant A. baumannii, so as to ensure a consistent therapeutic approach.

Characteristics of A. baumannii Isolates
In the observational period of February to July 2017, 51 consecutive isolates of A. baumannii were obtained from different clinical samples of 32 patients (Table 1; Table S1) hospitalized in the ICU of a large teaching hospital in Rome, Italy. The A. baumannii isolates were responsible for 15 bloodstream infections (29.41%), including two cases of central venous catheter (CVC)-related bacteremia, 19 cases of ventilator-associated pneumonia (VAP) (37.25%), and 5 infections of other body sites (9.81%). Twelve isolates (23.53%) were involved in colonization of various body sites of patients without clinical signs of infection (Table S2). All 51 isolates were unambiguously identified as A. baumannii by the Vitek ® 2 system (bioMérieux, Marcy l'Etoile, France) and MALDI-TOF (Bruker Daltonics, Bremen, Germany) with full agreement between the two methods.
To determine if there were any inaccuracies in reporting the colistin MIC values by Vitek ® 2, due to causes strictly referable to the instrument, colistin susceptibility was tested for all 51 A. baumannii isolates by the broth microdilution method, considered as the gold standard [8][9][10][11][12]. The broth microdilution test showed that 12/51 isolates (23.53%) were resistant to colistin. Furthermore, 11.76% (6/51) of the isolates showed discordancy between the values obtained with Vitek ® 2 and the microdilution method with discrepant MIC values of more than two dilution factors. By comparing colistin susceptibility results between the MIC Test Strips and broth microdilution, it emerged that 5.88% (3/51) of the strains showed a discrepancy between the two techniques with values greater than two dilution factors (Table S4). Relative to broth microdilution (golden standard), the essential agreement values for Vitek ® 2 and MIC Test Strips for colistin were 88.23% (45/51 MIC results) and 66.67% (34/51 MIC results), respectively (Table S4), thus below the acceptability threshold in both comparisons.
A plot of the MIC values according to the test method provides evidence that colistin resistance is largely underestimated according to Vitek ® 2 results (Figure 2). Of note, all isolates that presented discordant MIC values between Vitek ® 2 and the microdilution test belonged to ICL II (not shown).

Growth Kinetics of A. baumannii in the Presence of Colistin
Spectrophotometric readings (OD 600 ) provide a more objective assessment of bacterial growth compared with visual inspection of microplates. On this basis, six A. baumannii isolates (#3, 6, 7, 25, 28, 46) that resulted colistin-sensitive in the Vitek ® 2 assay but resistant in the broth microdilution test (Table S4) were selected to investigate their growth behavior in the presence of scalar concentrations of colistin (0.50-4 µg/mL). Colistin caused a dosedependent reduction of the bacterial growth rate for all the tested strains, although the dose dependance and the extent of growth reduction differed between strains ( Figure 3). However, all isolates showed appreciable growth in the presence of 2 µg/mL colistin, thus above the susceptibility threshold (≤2 µg/mL), and one isolate (#3) in the presence of 4 µg/mL colistin. These results are in agreement with those obtained in the broth microdilution test and provide evidence of its superiority with respect to the Vitek ® 2 assay, which categorized these isolates as colistin-sensitive (MIC ≤ 2 µg/mL).

Heteroresistance to Colistin in A. baumannii
The heteroresistance to colistin was evaluated in six A. baumannii isolates that displayed colistin susceptibility by Vitek ® 2 and resistance by broth microdilution (#3, 6, 7, 25, 28, 46; Table S4). Only three isolates, monitored for the growth at 24 and 48 h, formed colonies inside the inhibition halo of colistin MIC Test Strips. These colonies were tested for colistin susceptibility by all three methods and disclosed a higher colistin MIC (>16 µg/mL) than that previously observed in all tests, thus demonstrating the presence of an A. baumannii subpopulation that turned out to be resistant to colistin.

Discussion
The persistence in hospital environments and the extraordinary ability to rapidly acquire resistance to new antibiotics rendered problematic the treatment and eradication of infections caused by multidrug-resistant (MDR) A. baumannii strains [20]. In this dramatic situation, colistin often represents the last-resort (lifesaving) drug for the treatment of infections sustained by A. baumannii [21,22]. However, since 1999 there has been evidence of the growing frequency of nosocomial A. baumannii isolates that showed resistance to colistin [23]. In this work, the vast majority of A. baumannii isolates obtained from ICU patients were classified as XDR (86.27%) or PDR (11.76%) [18], witnessing the evolution of A. baumannii towards panresistance and causing serious alarm about the increasing prevalence of colistin-resistant A. baumannii, especially in ICU settings [18,19].
A progressive decline in susceptibility to β-lactam antibiotics in A. baumannii has been documented over the past several decades [24]. Carbapenem resistance was mainly attributed to the spread of genes encoding for oxacillinase (bla OXA-like ). In the early 2000s, OXA-58 oxacillinases were the most widespread in Italy and neighboring countries, but their prevalence has drastically decreased over the years, with the onset of OXA-23, which gradually replaced OXA-58 [25]. Therefore, the epidemiological profiles of the carbapenemresistant A. baumannii isolates analyzed in the present work are largely consistent with the traits of MDR A. baumannii circulating in Central Italy since 2005 [7,[26][27][28], as documented by the high prevalence (86.28%) of isolates belonging or related to ICL II, carrying the bla OXA-23 carbapenemase gene. By comparison with previous epidemiological data, major differences are observed in the emergence of a new clone in Central Italy, consisting in ICL I strains carrying the bla OXA-24/40 gene (13.72% of the isolates). It seems therefore that ICL I re-emerged in Central Italy after a decade predominated by ICL II [26] in association with the bla OXA-24/40 gene that has anecdotally been documented in Italy, invariably associated with ICL II [29]. These findings provide useful updates about the prevalence of XDR and PDR A. baumannii in Central Italy, and highlight the emergence of novel, potentially epidemic, clones.
The problem of the clinical management of patients affected by PDR A. baumannii strains poses a number of questions about the correct execution and interpretation of colistin susceptibility tests, especially agar diffusion tests (e.g., E-test ® , disk-diffusion) and automated systems (e.g., Vitek ® 2). The relative instability of the maintenance of concentration gradients of colistin in agar plates has raised a caveat against the use of such systems, both manual and automated. In fact, isolates that were considered susceptible to the action of the antibiotic resulted resistant when tested with reference tests, such as broth dilution methods [11]. Inappropriate colistin susceptibility testing can lead to misinterpretation of the results and, consequently, to an inadequate antibiotic therapy [11]. For this reason, the susceptibility to colistin of all of the isolates of our collection was compared between broadly used commercial methods, namely Vitek ® 2 and MIC Test Strips, and the broth microdilution method, considered as the gold standard. Data revealed from comparison of the three methods indicate poor concordance between Vitek ® 2 or MIC Test Strips and the broth microdilution method (<90% essential agreement). The MIC values obtained through both commercial methods were discordant by ca. 10% when compared with the MIC values obtained by the broth microdilution, even setting the MIC discordancy threshold at more than two dilution factors. These results corroborate the conclusion that diffusion methods should not be used to test colistin susceptibility in A. baumannii [11] and make questionable colistin susceptibility testing using Vitek ® 2 [30]. Intriguingly, all of the 13 strains that showed discordant values between Vitek ® 2 and the reference broth microdilution method were found to belong to ICL II. Of course, this observation is far from a general conclusion and further studies are needed to confirm this result.
In time-course experiments, we observed that the growth in the presence of colistin of a subset of A. baumannii isolates whose MIC at Vitek ® 2 showed values close to the sensitivity breakpoint (MIC ≤ 2 µg/mL) became spectrophotometrically detectable only after at least 12 h of incubation. To the best of our knowledge, no other sound studies attempted to correlate the kinetic of growth of A. baumannii in the presence of colistin in a microdilution method at concentrations that are critical in discriminating resistance or susceptibility of a given strain. In our opinion, if discontinuity of the concentration gradient of colistin diffusion in the agar explains the inaccuracy of the diffusion tests, a different explanation should be applied to the results obtained with the Vitek ® 2 system. In fact, this is an instrument with a temperature-controlled incubation chamber (35.5 • C), which, to provide early information on the sensitivity of an antibiotic, signals bacterial growth as soon as it is detected (at approximately four hours). Therefore, the duration of the incubation time could represent an important limiting factor in determining the accuracy of colistin MIC, since microbial growth, which starts after about 10-12 h at >2 µg/mL (breakpoint) antibiotic concentrations, could be undetectable at shorter incubation times. This could explain why in our comparative tests the Vitek ® 2 system underestimated MIC levels for colistin, resulting in problematic therapeutic implications.
To make the picture more complicated, we observed the presence of heteroresistance to colistin in several isolates of A. baumannii. Although the population analysis profile (PAP) seems to show a better accuracy in detecting heteroresistance by quantifying the proportion of resistant cells existing within a culture, in this work we adopted the E-test ® approach that has been shown to correctly identify heteroresistant subpopulations in a faster and less time-consuming manner [31]. Our results show the presence of heteroresistance in three out of the six strains that presented discordance in the susceptibility to colistin. Therefore, it is our opinion that heteroresistance cannot be considered a confunding factor in determining the extent of susceptibility to colistin in our strain collection.
In conclusion, this study highlights that the high frequency of isolation of XDR and PDR A. baumannii isolates in high-risk infectious wards requires an accurate application of methods for detecting susceptibility to antibiotics, in particular to last-resort antibiotics like colistin, so to ensure a correct therapeutic approach.

Materials and Methods
Strains of A. baumannii (n = 51) were isolated from clinical samples of 32 patients admitted from February 2017 until July 2017 to the ICU of a 1300-bed tertiary care academic hospital (Policlinico "Umberto I", Rome, Italy) with an average acceptance of 38,000 patients/year. The clinical samples considered in this study were respiratory secretions, blood cultures, urine, central venous catheters, and samples from other sites (i.e.,cerebrospinal fluid (CSF), nasal and wound swabs, wound drainage, abscess). The samples were cultured on agar media (blood agar and MacConkey Agar, bioMérieux, Marcy l'Etoile, France) and incubated for 24 h at 37 • C. The bacterial colonies were identified by a semiautomatic biochemical method (ID-GN cards, Vitek ® 2, bioMérieux) and MALDI-TOF mass spectrophotometry (Bruker Daltonics, Bremen, Germany), with a discriminatory score >2300 [32].
Multiplex PCR was performed to identify resistance genes belonging to oxacillinases (bla OXA-like ). The sequences of bla OXA-like alleles encoding carbapenemases were aligned and group-specific regions were identified using Bioedit software [33]. Primer  Table 2). The amplification conditions were initial denaturation at 94 • C for 5 min, 30 cycles of 94 • C for 25 s, 52 • C for 40 s, and 72 • C for 50 s, and a final elongation at 72 • C for 6 min [34,35].
Several molecular methods with different resolutions have been used to type A. baumannii strains [9]. A typing scheme based on two multiplex PCRs targeting three genes under selective pressure (ompA, csuE, and bla OXA-51-like ) has been a convenient method for rapid assignment of A. baumannii isolates into three major PCR-based groups (Gs) corresponding to international clones I (G2), II (G1), and III (G3) [19]. Multiplex PCRs for identification of the ompA, csuE, and bla OXA-51-like sequence groups defined as Group 1 and Group 2 were performed using the primers listed in Table 3. Primers for ompA and csuE were designed from sequences available at GenBank (AY485227, DQ093960; AY241696) or from sequences determined during initial studies (DQ289014-DQ289019). These were aligned and consensus primers were designed from common sequence areas. The amplification conditions were: 94 • C for 3 min, followed by 30 cycles of 94 • C for 45 s, 57 • C for 45 s, and 72 • C for 1 min, followed by a final extension at 72 • C for 5 min [19,36]. Identification of a strain as a member of Group 1 or Group 2 required the amplification of all three fragments in the corresponding multiplex PCR, and an absence of any amplification by the other multiplex PCR. Group 3 isolates were defined by the amplification of only the ompA fragment in the Group 2 PCR, and the amplification of only the csuE and bla OXA-51-like fragments in the Group 1 PCR. The latest EUCAST interpretative criteria of the MICs for colistin were used [9], considering resistant the strains with an MIC ≥ 4 µg/mL. In all the experiments, Escherichia coli ATCC 25922 was used as the susceptible control strain. Furthermore, A. baumannii type strain ATCC 19606 was used as colistin-susceptible control strain.
The growth dynamics of 13 A. baumannii strains in the presence of colistin scalar concentrations were kinetically tested in sterile microplates using a microtiter plate reader (Spark, Tecan Group, Mannedorf, Switzerland). After checking the 0.50 MF inoculum by the reference standard, the sample absorbance was automatically detected at OD 600 at two-hour intervals for 24 h and at a temperature of 37 • C. The results were used to construct the bacterial growth curves, correlating the absorbance in function of time.
The presence of colistin heteroresistance in six isolates of A. baumannii was checked by plating the bacterial suspension at 0.50 MF and 2 MF on Mueller-Hinton Agar II plates and tested with MIC Test Strips at 37 • C for 48 h.