In Vitro Activity of Rifabutin and Rifampin against Antibiotic-Resistant Acinetobacter baumannii, Escherichia coli, Staphylococcus aureus, Pseudomonas aeruginosa, and Klebsiella pneumoniae

ABSTRACT We recently reported that the antimicrobial activity of rifabutin against Acinetobacter baumannii is best modeled by the use of RPMI for in vitro susceptibility testing. Here, we define the effects of medium on the susceptibility and frequency of resistance emergence in a panel of A. baumannii, Escherichia coli, Staphylococcus aureus, Klebsiella pneumoniae, and Pseudomonas aeruginosa clinical isolates. Only A. baumannii was hypersusceptible to rifabutin in vitro and in vivo using a Galleria mellonella infection model. In vitro, the frequency of resistance emergence was greater when the bacteria were selected on RPMI versus tryptic soy agar (TSA) or Mueller-Hinton II (MHII) agar plates. However, the frequency of resistance emergence was lower in vivo than in the RPMI in vitro condition. IMPORTANCE Rifabutin has been recently described as a potential adjunctive therapy for antibiotic-resistant A. baumannii infections due to hypersensitivity in iron-depleted media, which may more closely mimic an in vivo environment. Here, we report that this hyperactivity is specific for A. baumannii, rather than being a general effect for other pathogens.

Untested in previous studies were the in vitro and in vivo activities of rifabutin and rifampin (RIF) against other Gram-positive and -negative bacteria, as well as the impact of the culture medium on the emergence of resistance to rifabutin. In the current study, we determined if other bacteria are also more susceptible to rifabutin under low-iron conditions by quantifying rifabutin and rifampin MICs against a panel of multidrug resistance (MDR) A. baumannii strains, extended-spectrum-beta-lactamase (ESBL)-producing Escherichia coli, methicillin-resistant Staphylococcus aureus (MRSA), carbapenem-resistant (CR) Klebsiella pneumoniae, and MDR Pseudomonas aeruginosa clinical isolates to determine if other bacteria are also observed to be more susceptible to rifabutin when the MIC assay is conducted under low-iron conditions. We also evaluated the efficacy of rifabutin and rifampin in vivo against representative isolates of the aforementioned bacterial species using a Galleria mellonella infection model. Finally, we sought to determine if the hyperactivity of rifabutin but not rifampin against A. baumannii altered the frequency of emergence of drug-resistant mutants in vitro and in vivo.

RESULTS
MIC distributions. Significant shifts in the distribution of rifabutin MICs in MHII versus RPMI were observed for A. baumannii, E. coli, P. aeruginosa, K. pneumoniae, and S. aureus clinical isolates (Mann-Whitney test, P , 0.0001) ( Table 1; Fig. 1 and 2). However, the shift in MIC was greatest for A. baumannii isolates, as evidenced by a .125-fold and 8-fold shift in the rifabutin MIC 50 and MIC 90 , respectively (Table 1; Fig. 1 and 2). In contrast, small rifabutin MIC 50 shifts were observed for E. coli (2-fold), P. aeruginosa (2-fold), and K. pneumoniae (4-fold) ( Table 1; Fig. 1 and 2). No significant shift in rifabutin or rifampin MICs for S. aureus isolates was observed when values obtained in MHII were compared to those obtained in RPMI broth (Table 1; Fig. 1 and 2).
In vivo rifampin and rifabutin efficacy. In vivo survival of Galleria mellonella challenged with A. baumannii, E. coli, P. aeruginosa, K. pneumoniae, or S. aureus and treated with rifampin or rifabutin was consistent with RPMI MIC testing results ( Fig. 3; Table 2). Specifically, rifabutin was more effective than rifampin at equivalent doses to rescue Galleria from A. baumannii infection (log rank test: P = 0.011 at 10 mg/kg; P = 0.0003 at 5 mg/kg; P , 0.0001 at 1 mg/kg) (Fig. 3). There was no significant difference between rifabutin's and rifampin's abilities at all tested doses to rescue Galleria infected with E. coli, P. aeruginosa, or K. pneumoniae. Though there was no significant difference between rifabutin's and rifampin's efficacies at 10 mg/kg or 5 mg/kg against S. aureus infection, rifampin was better able than rifabutin at 1 mg/kg (log rank test, P = 0.0001) to rescue Galleria (Fig. 3).

In Vitro Activity of Rifabutin and Rifampin
In vitro rifampin and rifabutin mutation frequency. We expanded on previous efforts to study mutation frequency (9) by comparing the frequency of mutant emergence under in vitro and in vivo conditions. Because only A. baumannii experienced a dramatic shift in rifabutin MICs in RPMI versus MHII, we focused on two specific A. baumannii clinical isolates (HUMC1 and ABNIH1) for studies on the effect of medium on resistance emergence frequencies. In MHII, both isolates exhibited the same sensitivity to both rifampin and rifabutin (MIC = 3.13 mg/L). In RPMI, both isolates were equally susceptible to just rifampin (MIC = 3.13 mg/L). However, in RPMI, HUMC1 was highly susceptible to rifabutin (MIC # 0.05 mg/L), while ABNIH1 was not (MIC = 3.13 mg/L). Serial passage allowed the selection of highly resistant mutants for both drugs in both isolates. Despite HUMC1 being more susceptible to rifabutin in RPMI media, it was still possible to select for resistant mutants (Fig. 4).
Using a high-inoculum plating strategy, the frequency of rifabutin resistance emergence was higher in RPMI (1E205 to 1E206) than MHII and tryptic soy agar (TSA) (1E208 to 1E210). The frequency of resistance emergence to both rifampin and rifabutin was similar when bacteria were cultured on the same media (Table 3) (Mann-Whitney, P . 0.05). However, the frequency of resistance emergence was higher when bacteria were cultured on RPMI agar than TSA for both rifampin (Kruskal-Wallis test; P = 0.014) and rifabutin (Kruskal-Wallis test; P = 0.028) ( Table 3).
In vivo rifampin and rifabutin mutation frequency. Because we found that the culture conditions affect A. baumannii susceptibility to rifabutin, as measured in vitro by MIC assay or frequency of mutant resistance emergence, we therefore sought to determine how exposure in an in vivo environment would affect the frequency of resistance emergence. We were unable to select for any rifampin-or rifabutin-resistant mutants when mice were infected and treated in vivo and blood or kidney homogenates were plated on antibiotic selective agar plates. The experiment was repeated, but mice were not treated with rifampin or rifabutin in vivo to allow harvesting of a greater number of bacteria in the blood and kidney homogenates. The mutation frequencies of HUMC1 collected from both blood and kidney were 0.5 to 2 log higher for rifampin than rifabutin on both RPMI agar and TSA. On the other hand, no difference in mutation emergence was observed under any of the conditions tested for LAC-4. This may be due to the generally lower mutation frequency observed in LAC-4 than HUMC1 (Table 3).

DISCUSSION
We have shown that A. baumannii is significantly more susceptible to rifabutin when cultured in RPMI medium (9,10). By screening a larger collection of A. baumannii, E. coli, S. aureus, K. pneumoniae, and P. aeruginosa clinical isolates, we demonstrated that hypersensitivity to rifabutin is commonly and uniquely observed in A. baumannii. Additionally, consistent with our previous study, we were unable to observe any hypersusceptible isolates in the non-Acinetobacter bacteria that were tested (9,10).
We also found that under any medium condition, rifabutin resistance emergence rates were lower than the rates of rifampin resistance, which supports the potential clinical utility of adjunctive rifabutin for such infections (12). However, we observed a higher frequency of resistant mutants in RPMI agar selection plates than in MHII. Fortunately, the low frequency of resistant mutants in vivo was consistent with that observed in vitro on the nutrient-rich media such as MHII and TSA rather than on RPMI agar. The bacteria grow more slowly in RPMI, and this lower growth rate may affect the frequency of resistance emergence. Thus, paradoxically, RPMI is more accurate at predicting in vivo rifabutin efficacy but less accurate at predicting in vivo emergence of resistance to rifabutin for treatment of A. baumannii infection.
Consistent with other published studies, we were unable to select for any rifampin-resistant mutants in less than 48 h when mice were infected with A. baumannii and treated in vivo (13)(14)(15). However, A. baumannii rifampin-resistant isolates were reported to have been selected for in a murine pneumonia infection model between 48 and 72 h postinfection (16).
Based on in vitro susceptibility, in vivo treatment efficacy, and in vitro and in vivo frequency of resistance emergence, rifabutin remains a potential useful therapeutic for the treatment of A. baumannii infections.

MATERIALS AND METHODS
Serial passage mutant selection. Two A. baumannii clinical isolates, HUMC1 and ABNIH1, were serially passaged for 20 days in broth supplemented with antibiotics as previously described (17). Briefly, bacteria were serially passaged 20 times with sub-MICs of rifabutin (R3530-25MG; Sigma) or rifampin (R3501-1G; Sigma) in MHII (90000-602; VWR) or RPMI (11875119; Gibco) media at 37°C and 200 rpm. Every day, 100 mL of culture was passaged into 10 mL fresh MHII or RPMI medium with antibiotics.  During the first 5 days of passage, bacteria were grown in media with 1/3 the MIC of rifabutin or rifampin. Antibiotic susceptibility was tested by plating bacteria on TSA (211825; VWR) plates containing 3Â, 10Â, or 100Â the passage concentration of antibiotic. The starting antibiotic concentration on day 6 was determined based on the agar plate result. The same testing process was performed on days 10, 15, and 20 to determine the starting antibiotic concentration of the subsequent 5 days.
High-inoculum mutant selection. There are no CLSI breakpoint interpretations for RIF or RFB against A. baumannii. We therefore referenced the RIF MIC breakpoint for staphylococci (4 mg/L). To increase the stringency of our conditions, we then then doubled that to reach the 8 mg/L that was used in our experiments. Bacteria were cultured overnight in tryptic soy broth (TSB) at 37°C and 200 rpm, and spontaneous rifampin-and rifabutin-resistant mutants were selected on TSA or MHII agar plates supplemented with 8 mg/L of rifampin or rifabutin. Alternatively, bacteria were cultured overnight in RPMI alone or RPMI plus amino acids (RPMI1AA) at 37°C and 200 rpm as previously described (9), and FIG 4 20-day serial passage mutant selection. For HUMC1, the starting MIC for both drugs was 3.13 mg/L in MHII. In RPMI, however, the starting MICs for rifampin and rifabutin were 3.13 mg/L and ,0.05 mg/L, respectively. For ABNIH1, the starting MIC for both drugs was 3.13 mg/L in both MHII and RPMI. Despite HUMC1 initially being more susceptible to rifabutin in RPMI, it was still possible to select high-level resistance mutants. Similarly, no difference in rifabutin resistance emergence was observed between the hypersusceptible strain HUMC1 and the susceptible strain ABNIH1 after serial passage.  (18)(19)(20). Mice were euthanized at 16.5 h postinfection, and blood and kidney samples were collected. Kidneys were homogenized in phosphate-buffered saline (PBS). Serial dilutions were plated on drug selection plates to select for mutants and on nondrug plates to determine total CFU. Drug selection plates were supplemented with 8 mg/L rifampin or rifabutin in TSA or RPMI agar. Plates were incubated at 37°C overnight, and colonies were enumerated.
MIC assay. Unless otherwise indicated, the broth microdilution method was used to determine MICs (9,21). The medium used for the MIC assays performed in the present study was either MHII alone or RPMI supplemented with 10% fetal bovine serum (PS-100; Phoenix Scientific). The final drug concentrations in the plate were 2-fold dilutions ranging from 0.05 to 25 mg/L. The inoculum concentration was confirmed by plating serial dilutions on TSA plates, incubating at 37°C overnight, and enumerating colonies. MIC plates were incubated at 35 6 2°C without shaking, and results were recorded at the recommended time points (21).
In vivo rifampin and rifabutin efficacy. Galleria mellonella larvae (WAXB500; Timberline) were stored at room temperature in the dark for no longer than 1 week before experimentation. Larvae weighing 240 to 350 mg (group mean standard deviation # 4.6) and grouped by tens were incubated at 4°C for up to 1 h before infection to reduce their movement during injection. Bacteria were prepared for infection by growing overnight in TSB at 37°C and 200 rpm and then subculturing in TSB until midlog phase. Bacterial suspensions were washed 3 times with PBS and adjusted to an optical density (OD) of 0.5 before dilution to desired inocula. Each larva was disinfected by brief rolling in a 70% ethanolsoaked KimWipe (470224-038; Kimtech Science) and infected subcutaneously using an NE-1000 fully programmable single-syringe pump (New Era Pump Systems, Inc.) with 10 mL A. baumannii HUMC1, E. coli JJ1886, P. aeruginosa PAO1, K. pneumoniae KPC-KP1, or S. aureus LAC to determine respective 100% lethal doses (LD 100 ). No-infection control larvae received 10 mL PBS. Larvae were then infected with previously determined LD 100 (in CFU) of each strain and treated with 10 mL PBS or 1, 5, or 10 mg/kg rifampin or rifabutin 1 h postinfection. No-infection control larvae received 2 doses of 10 mL PBS. For all experiments, larvae were incubated in 100-mm petri dishes (25384-302; VWR) at 37°C, and survival was monitored up to day 4 postinfection. CFU were calculated by plating serial dilutions on TSA plates, incubating them at 37°C overnight, and enumerating colonies.