Analysis of Paradoxical Efficacy of Carbapenems against Carbapenemase-Producing Escherichia coli in a Murine Model of Lethal Peritonitis

Bacterial strains, conjugation assays.Three clinical isolates were used as sources of carbapenemase genes: KPC-3-producing E. coli 13832, OXA-48-producing E. coli 13828, and NDM-1-producing E. coli UR20. E. coli DH5a was used for cloning experiments, and uropathogenic E. coli CFT073 (O6:K2:H1), previously used to produce pyelonephritis and peritonitis in mice, was used as the final recipient strain (26, 27). The plasmid pCR-Blunt II-TOPO (Life Technologies, Saint-Aubin, France), which carried a kanamycin resistance gene, was used for cloning experiments. In order to avoid any problem of expression of resistance in vitro, carbapenemase genes were cloned into pTOPO but with their own promoter. In order to avoid plasmid loss, all subcultures were done in the presence of kanamycin (400 mg/liter). PCR, cloning, and DNA sequencing have been described previously (27). For NDM-1, cloning into pTOPO was performed using the NDM-1-F (5′-AGAGAAATTTGCTCAGCTTGTTGA-3′) and NDM1-R (5′-GATGGCAGATTGGGGGTGAC-3′) primers. The four isogenic strains carrying the empty vector (pTOPO) and recombinant plasmids with the bla_KPC-3, bla_OXA-48, and bla_NDM-1 genes are referred to as TOPO, KPC-3, OXA-48 and NDM-1, respectively.

In vitro antibiotic activities.MIC values of imipenem (IMP) and ertapenem (ERT) were determined by the broth microdilution method in accordance with EUCAST guidelines (www.eucast.org). Time-kill curves were performed to test for the bactericidal effect of each drug. To assess the impact of protein binding on antibiotic activity, tests were also performed in the presence of 4% albumin (to approach plasma albumin concentrations in humans of around 40 g/liter). To assess the impact of human serum on IMP activity, time-kill curves were also performed in the presence of 50% normal human serum (S4190-100 EUROBIO). To assess the influence of zinc depletion, MICs were also assayed in zinc-depleted Mueller-Hinton broth (MHB) by addition of 300 mg/liter of EDTA, as described (11).

For time-kill curves, exponentially growing E. coli cells were diluted in glass tubes containing 10 ml MHB to obtain an inoculum of ca. 10⁶ CFU/ml and incubated with IMP or ERT in concentrations equal to 0.25 to 2 times the MIC value for the tested strain, or with no antibiotic. Viable counts were enumerated by plating 100-μL serial dilutions of cultures onto MH agar plates after 0, 1, 3, 6, and 24 h of incubation at 37°C, and expressed in log₁₀ CFU per milliliter. The lower limit of detection was 1 log₁₀ CFU/ml. A bactericidal effect was defined as a decrease of at least 3 log₁₀ in CFU counts compared to the initial inoculum.

Growth experiments.To measure growth rates of the different isolates with or without subinhibitory concentrations of ERT and IMP, bacterial cells were grown overnight in Luria-Bertani medium at 37°C with constant agitation at 250 rpm, and were transferred into fresh medium at a dilution of 1:100,00. On the same plate, wells contained no antibiotic or ERT or IMP at concentrations of 1/4, 1/8, or 1/16 of the MIC for each strain (from 0.0625 to 0.0156 mg/liter of IMP for TOPO, to 4 to 16 mg/liter IMP for the NDM-1 strain). Growth was recorded by an Infinite 200 Tecan spectrophotometer, which measured the optical density at 600 nm (OD₆₀₀) in each well every 5 min at 37°C for 24 h. OD₆₀₀ values were collected and log-transformed. The time in minutes to reach maximum growth rate was measured and the maximum growth rate was calculated and expressed in h⁻¹ (27). All in vitro experiments were repeated at least three times. Gram staining was performed on the bacteria growing at 24 h.

Experimental murine model.Swiss ICR-strain immunocompetent female mice aged 5 to 6 weeks and weighing 25 to 30 g were used in the experimental model of intraabdominal infection (27).

Animal experiments were performed in accordance with prevailing regulations regarding the care and use of laboratory animals and approved by the Departmental Direction of Veterinary Services (Paris, France, agreement no. 75-861). The peritonitis protocol (APAFIS number 4949-2016021215347422 v5) was approved by the French Ministry of Research and the ethics committee for animal experiments.

Antibiotic concentrations in mice.To determine the therapeutic regimen in mice that best reproduces the fT>MIC obtained in humans with standard regimens (1 g every 8 h for IMP and 1 g every 24 h for ERT), pharmacokinetic studies were performed on mice. A dose of 100 mg/kg was chosen for both IMP and ERT as they had previously been shown to reproduce the plasma peak concentrations that corresponded to human concentrations (28, 29). Free-drug concentrations of IMP and ERT had previously been determined in mice in our laboratory, and were found to be 37% for IMP and 1% for ERT (28).

Blood samples of a least 500 μl were obtained by intracardiac puncture from 4 anesthetized mice 30, 60, 90, and 120 min after a single subcutaneous injection of IMP (100 mg/kg) or 30, 60, 120, 240, 360, and 480 min after a single subcutaneous injection of ERT (100 mg/kg) (15). Concentrations of IMP and ERT were determined by high-performance liquid chromatography (HPLC) with UV detection at 237 nm, and the limit of detection was 0.5 mg/liter for both compounds (29).

Treatment experiments.Pellets of overnight E. coli cultures were mixed with porcine mucin 10% (Sigma-Aldrich, Saint-Quentin Fallavier, France). Mice, kept in single units during the experimental time, were inoculated with a 250-μl intraperitoneal injection of bacteria/mucin mix, corresponding to a final inoculum of approximately 10⁸ CFU/ml. Two hours after inoculation, 15 mice for each strain were sacrificed to determine start-of-treatment bacterial load. We had previously shown that the death rate of this model with the same bacterial strains was almost 100% in the absence of treatment and all mice were bacteremic; thus, for ethical reasons no untreated control group was used (27).

In the therapeutic groups, treatment was started 2 h postinfection and lasted 24 h. Mice received either IMP 100 mg/kg subcutaneously every 2 h or ERT 100 mg/kg subcutaneously every 6 h. Mice were sacrificed by intraperitoneal injection of 200 μl of sodium thiopental 4 h after the last antibiotic injection. Immediately after sacrifice, a peritoneal wash was performed by intraperitoneal injection of 2 ml of sterile saline solution followed by gentle massage of the abdomen and opening of the peritoneum to collect 1 ml of fluid. Spleen was extracted and homogenized in 1 ml of sterile saline solution. In mice that did not survive the infection, only spleen was extracted within 2 h after death to avoid sample contamination. Samples diluted by a factor of 10 were plated onto MH agar for quantitative culture, containing or not IMP and ERT at a concentration of 4×MIC to detect the selection of resistant mutants in vivo. Results were expressed as log₁₀ CFU/g for spleen and log₁₀ CFU/ml for peritoneal fluid. The lower limit of detection was 1.3 log₁₀ CFU/g or ml. Lack of plasmid loss was verified on surviving microorganisms by determining ERT and IMP MICs and by enumerating CFU on kanamycin-containing plates (400 mg/liter).

RNA extraction and qRT-PCR.To measure the expression of the bla_NDM-1 gene, total RNA was extracted from bacterial cells grown to the late exponential phase in vitro and from bacterial cells extracted from the peritoneal cavity 2 h after inoculation in vivo. Total RNA was extracted using the RNeasy PowerMicrobiome kit (Qiagen Hilden, Germany) and residual chromosomal DNA was removed by treating samples with the TURBO DNA-free kit (Life Technologies, Saint-Aubin, France). Samples were quantified using the NanoDrop One spectrophotometer (Thermo Fisher Scientific, Courtaboeuf, France). cDNA was synthesized from total RNA (approximately 1 μg) using the QuantiTect reverse transcription kit (Qiagen) according to the manufacturer’s instructions.

Since it is impossible to know if cells are in the same growth phase in vitro and in vivo, we performed an absolute quantification of the expression of bla_NDM-1 gene. The amounts of transcript levels were normalized using the rrsA gene as a housekeeping control gene. The number of transcript copies was determined by extrapolation from the linear regression of the standard curve obtained for each set of primers amplifying bla_NDM-1 and rrsA genes (Table 3). The ratio bla_NDM-1 copy number/rrsA copy number was calculated to compare the bla_NDM-1 expression in in vitro and in vivo samples.

Zinc concentrations.Levels of Zn²⁺ were measured in MH broth and in mouse serum by inductively coupled serum mass spectrometry (ICP-MS) on an X-Series II from Thermo Scientific (Platform AEM2, University of Rennes 1/Biochemistry Laboratory, University Rennes Hospital) (30).

Statistical analysis.Continuous variables are expressed as the median followed by the range (minimum to maximum) in brackets, and were compared using a nonparametric test (Kruskall-Wallis). Maximal growth rates and time to achieve maximal growth rate values were compared using the Kruskall-Wallis test. A P value of less than 0.05 was considered significant. All statistical analyses were conducted using the Prism software (version 7.0a).