Analyses of a Ceftazidime-Avibactam-Resistant Citrobacter freundii Isolate Carrying blaKPC-2 Reveals a Heterogenous Population and Reversible Genotype

The development of ceftazidime-avibactam resistance among KPC-producing isolates during treatment with this agent has been reported. Usually isolates that become resistant have a mutated blaKPC gene that confers resistance to ceftazidime-avibactam and susceptibility to meropenem. We report a Citrobacter freundii isolate that developed ceftazidime-avibactam resistance due to mutations within the coding region of the blaKPC-2 Ω-loop previously reported; however, in this case, only 11% of the whole-genome sequencing reads had mutations, making this alteration difficult to detect and the treatment of these isolates more challenging. In addition to blaKPC, the initial and the follow-up patient isolates displayed hyperexpression of the AcrAB-TolC efflux system and disruption of the outer membrane protein (OMP) OmpF, which contribute to carbapenem resistance. Experiments performed to confirm our findings included generating mutant isolates from the initial patient isolate, passaging the isolates for purity in drug-free medium, resulting in a reversible phenotype, and cloning the mutations to demonstrate the resistance conferred.

KEYWORDS Citrobacter freundii, ␤-lactams, ceftazidime-avibactam, resistance mechanism I solates producing Klebsiella pneumoniae carbapenemase (KPC) enzymes are widespread in the United States and other countries. Until recently, the treatment options for organisms producing this carbapenemase were limited to antimicrobial agents that display toxicity issues or have a suboptimal distribution in certain infection sites or the scarce agents active against the organisms carrying these enzymes, which are often multidrug resistant.
Ceftazidime-avibactam has been approved in the United States since 2015, and this combination demonstrated good activity against isolates producing KPC enzymes and other serine ␤-lactamases. Ceftazidime-avibactam was superior to other regimens for treating carbapenem-resistant Klebsiella pneumoniae (1), Enterobacteriaceae (2), and carbapenemase-producing isolates (3) collected in U.S. hospitals.
Shortly after the introduction of this combination in clinical use, reports of isolates developing ceftazidime-avibactam resistance were published (4). The initial report was from a K. pneumoniae isolate producing KPC-3 that recently had been shown to have alterations in outer membrane proteins (OMPs) and hyperexpression of the AcrAB-TolC efflux system (5). Later studies demonstrated that mutations in the KPC-encoding gene leading to the D179Y substitution codified resistance to ceftazidime-avibactam (6). In most cases, the isolates displaying these mutations are susceptible to meropenem and the mutations described increase ceftazidime hydrolysis by creating a deeper pocket that traps the ceftazidime molecule (7,8). After reading these reports, many hypothesized if ceftazidime-avibactam resistance developed during treatment that changing to meropenem would be an option.
In this study, we report a clinical case of ceftazidime-avibactam resistance developing in a KPC-2-producing Citrobacter freundii isolate during treatment with this combination and the characterization of the ceftazidime-avibactam-resistant isolate. Due to the initial lack of resistance mechanisms that could explain the resistance developed during therapy, we submitted the initial clinical isolates to passaging experiments in ceftazidime-avibactam, and we observed a heterogenous population of ceftazidimeavibactam-resistant bla KPC-2 mutants and wild-type (WT) sequences. Constructs were created to show the effect of each mutation on the susceptibility to ␤-lactams alone and in the presence of avibactam.

RESULTS
Clinical case. A 44-year-old woman with a failed renal transplant who was on peritoneal dialysis presented to a tertiary-care hospital in South Carolina in September 2015 with epigastric pain, nausea, and vomiting for 3 weeks. On hospital day 3 (HD 3), the patient was placed on piperacillin-tazobactam for a perforated duodenum. After multiple surgeries and washouts, the patient was no longer a candidate for operative intervention due to hemodynamic instability and was being managed conservatively with drainage. Piperacillin-tazobactam was discontinued on HD 17. On HD 24, a percutaneous left-upper-quadrant drain and a biliary drain were placed. On HD 26, blood cultures (initial isolate; Fig. 1) and abdominal fluid drain cultures were positive for C. freundii. The patient was started on ceftazidime-avibactam and tigecycline. The initial ceftazidime-avibactam dosing was 2.5 g every 12 h. The dose was increased to 2.5 g every 8 h on HD 33 based on the patient receiving aggressive continuous renal replacement therapy (CRRT) (9). Repeat blood cultures collected on HD 29 and HD 33 were negative, but the patient minimally improved from a clinical standpoint. Amikacin was added on HD 29. The infectious source was feculent peritonitis with signs and symptoms suggesting persistent intra-abdominal infection, but the patient was not a candidate for operative intervention due to hemodynamic instability. On HD 37, repeat abdominal fluid cultures grew C. freundii (follow-up isolate). The patient died on HD 39 due to multiorgan failure, untreatable intra-abdominal bleeding, coagulopathy, and liver failure.
Characterization of C. freundii clinical isolates. The initial C. freundii isolate and the follow-up isolate from the patient displayed MIC results within Ϯ1 log 2 dilution for most antimicrobial agents tested, except for ceftazidime-avibactam. The isolates were resistant to ceftazidime alone, cefepime, meropenem, piperacillin-tazobactam, aztreonam, levofloxacin, and tobramycin (Table 1). Both isolates were susceptible to amikacin and tigecycline when the Clinical and Laboratory Standards Institute (CLSI) or U.S. Food and Drug Administration breakpoints were applied, and colistin MIC values were wild type when the CLSI epidemiologic cutoff value was applied (10). The initial isolate displayed a ceftazidime-avibactam MIC of 4 g/ml, and the follow-up isolate had an  MIC of 64 g/ml for this combination. The decreased ceftazidime-avibactam susceptibility of the follow-up isolate was confirmed by Etest (initial isolate MIC of 3 g/ml and follow-up isolate MIC of 128 g/ml) and disk diffusion (initial isolate zone diameter of 24 mm and follow-up isolate zone diameter of 16 mm). Pulsed-field gel electrophoresis analysis showed that the 2 C. freundii clinical isolates were indistinguishable (data not shown). Both isolates carried bla KPC-2 identical to the reference sequence (GenBank accession no. AY034847.1) when PCR amplification followed by Sanger sequencing was performed. The investigation of additional ␤-lactam resistance mechanisms in the follow-up isolate revealed highly elevated expression of acrA in the follow-up isolate that was Ͼ1,000-fold (relative quantification value [RQ] of 1,189.0 and minimum to maximum RQ of 795.6 to 1,776.9) that of the susceptible C. freundii control strain, but this increase was only 2.5-fold greater than the expression of this gene in the initial isolate from the same patient. The expression of ompC was similar to that of the susceptible control strain, and the initial isolate from the same patient and ompF could not be amplified in the expression assay.
Due to the lack of resistance mechanisms that could explain the increase in the ceftazidime-avibactam MIC values, the 2 clinical isolates were submitted for wholegenome sequencing (WGS). OMP sequences were analyzed, and OmpF was found to be disrupted in both clinical isolates. OmpC had no alterations compared to a susceptible control. Single nucleotide polymorphism (SNP) analysis showed minor alterations in genes that have not been previously associated with ␤-lactam resistance (Table 2).
Passaging, cloning, and characterization of passaged isolates. The initial patient isolate was submitted to ceftazidime-avibactam passaging experiments in an attempt to recreate the genotype observed in the follow-up patient isolate. After being passaged for 7 days, colonies growing in 32 g/ml ceftazidime with avibactam at a fixed concentration of 4 g/ml or at a 4:1 ratio were tested for susceptibility against a panel of ␤-lactams and were submitted for WGS. The isolates obtained after passaging were highly resistant to ␤-lactam agents, including carbapenems ( Table 1). The sequence analysis and SNP comparison with the initial clinical isolate demonstrated that approximately 78% of the sequence reads of the bla KPC-2 had the alteration D176Y (bla KPC-2 -D176Y) for the isolated passaged on a fixed concentration of 4 g/ml, and 82% of the bla KPC-2 reads had R164S plus P147L substitutions (bla KPC-2 -R164S ϩ P147L) for the 4:1 ratio passaging. These isolates were highly resistant to all ␤-lactams tested. Prompted by this finding, a thorough analysis of the WGS data of the clinical follow-up isolate was performed and revealed 11% of bla KPC-2 reads with the alteration D179Y (bla KPC-2 -D179Y, currently named bla KPC-33 ; GenBank accession no. NG_056170), which has been associated with elevated ceftazidime-avibactam MIC values in bla KPC-2 and bla KPC-3 (6,8).
To evaluate the stability of the genotypes, 20 colonies obtained from the C. freundii follow-up patient isolate and the isolate passaged with ceftazidime-avibactam at a fixed concentration of 4 g/ml and purity streaked were streaked on medium without selection, and the colonies were susceptibility tested. For the clinical isolate, 13 of the 20 colonies displayed ceftazidime-avibactam MIC values of Ն16 g/ml, but 7 had MIC values of 1 to 4 g/ml for this combination. Sequencing confirmed that ceftazidimeavibactam-susceptible colonies had only the wild-type bla KPC-2 , whereas the colonies exhibiting high ceftazidime-avibactam MIC results had reads displaying the mutation D179Y. For the passaged isolates, 12 of the 20 colonies were resistant to ceftazidimeavibactam, but 8 had reverted to susceptible MIC values for this combination and wild-type bla KPC-2 sequences. The reversion of phenotype confirms that these colonies might harbor mutated bla KPC-2 encoding ceftazidime-avibactam resistance and the wild-type version of the gene encoding resistance to meropenem, as seen in the WGS reads.
The impact of the mutations observed was assessed by creating 6 plasmid constructs carrying the bla KPC-2 gene and genes encoding KPC-2 with alterations D179Y, D176Y, R164S, P147L, and R164S plus P147L. Escherichia coli TOP10 strains carrying the recombinant plasmids were susceptibility tested ( Table 3). The MIC values for ceftazidime increased 2-to 8-fold, and those for ceftazidime-avibactam increased 8-to 32-fold for KPC-2 variants. The higher increases were observed for D179Y, D176Y, and P164S plus P174L. The constructs carrying the latter substitutions alone (R164S or P147L) had lower MIC values for ceftazidime-avibactam (2 g/ml), but the same or slightly higher MIC values compared to the construct producing wild-type KPC-2. The MIC values for ceftriaxone and cefepime alone were 4-to 8-fold lower for the KPC-2 mutants than that of the construct producing wild-type KPC-2. When these cephalosporins were tested with a fixed concentration of 4 g/ml of avibactam, the MIC values were similar to that for the wild-type KPC-2. A similar effect was noted for aztreonam, which had activity restored for all constructs carrying KPC-2 from 2 to Ͼ16 g/ml to 0.12 and 0.25 g/ml by adding 4 g/ml of avibactam (Table 3). As observed by other authors, the MIC values for imipenem and meropenem alone were lower for the mutated KPC-2, and the MIC values were lowered by adding avibactam. Piperacillin MIC values were unchanged with KPC-2 constructs, and avibactam had a similar effect for all recombinant isolates when tested with this agent.

DISCUSSION
Similar to other literature reports about ceftazidime-avibactam resistance developing during treatment, we observed the increase in ceftazidime-avibactam MIC values from 4 to 64 g/ml in a C. freundii isolate carrying bla KPC-2 . Our investigation showed that both clinical isolates had a highly elevated expression of AcrAB-TolC compared to a C. freundii control strain susceptible to all ␤-lactams. This mechanism has been recently associated with an increased ceftazidime-avibactam MIC value of 16 g/ml (5); however, since both patient isolates had similar expression of this efflux system, the hyperexpression of this gene was unlikely to cause the increase in the ceftazidimeavibactam MIC during treatment. The elevated expression of AcrAB-TolC and OMP alterations could, however, explain the elevated MIC result of 4 g/ml in the initial C. freundii isolate.
The initial results from the patient follow-up C. freundii isolate did not indicate a mutation on bla KPC-2 since the gene carrying these alterations usually encodes an extended-spectrum ␤-lactamase-like phenotype and susceptibility to carbapenems, and the follow-up isolate evaluated in this study was resistant to all carbapenems tested. WGS analysis of the isolates obtained after ceftazidime-avibactam passaging experiments exhibited a percentage of the bla KPC-2 sequence reads with mutations leading to amino acid substitutions D176Y or R164S plus P174L, which are located in the ⍀-loop (11) and were demonstrated to encode ceftazidime and ceftazidimeavibactam resistance. Further investigations on the follow-up C. freundii patient isolate showed that a small percentage of the bla KPC-2 sequencing reads had mutations leading to the D179Y alteration, which has been previously reported to increase ceftazidime hydrolysis and decrease inhibition by avibactam (8). The low percentage (11%) of reads was in the initial WGS analysis but was unnoticed.
Sequencing reads of the wild-type and mutated bla KPC-2 genes were present in the clinical and passaged isolate experiments that generated high MIC values to all ␤-lactam agents. Subculturing these isolates in the absence of selective pressure demonstrated that part of the population reverted to wild-type bla KPC-2 sequences to the detriment of the mutant reads. However, we could not separate colonies carrying the wild-type and the D179Y mutant from the C. freundii clinical isolate, despite multiple attempts.
Mutations of the ⍀-loop of various ␤-lactamase enzymes increased the hydrolysis for ceftazidime, and the KPC-2 variants carrying D179Y alterations were able to trap the ceftazidime molecule for longer periods and avoid binding of avibactam (7,8). This effect was not observed with other ␤-lactams tested, and our results demonstrate that avibactam was still able to inhibit the KPC-2 mutants harboring D179Y, D176Y, and R164S plus P174L when paired with ceftriaxone, cefepime, aztreonam, imipenem, and meropenem.
Our findings implicate that once the ceftazidime-avibactam selective pressure is removed, the population might revert to a majority of the original genotype and again display elevated MIC values for other ␤-lactams, including carbapenems, which means that ␤-lactams alone might not be used, as suggested by others. Instead, addition of another ␤-lactam such as cefepime, aztreonam, or a carbapenem might be more prudent to make sure that coverage is provided for both populations.
Other antimicrobial agents displaying activity against KPC-producing isolates have been approved or are in late-stage development, but as shown with the experience of ceftazidime-avibactam, only clinical use will reveal the advantages and limitations of each of them; however, at least now more options exist to treat these troublesome organisms.

MATERIALS AND METHODS
Bacterial strains. Two C. freundii clinical isolates recovered from blood and abdominal fluid from a patient hospitalized in South Carolina in September 2015 were submitted to a reference laboratory for confirmation of susceptibility results. A timeline of the infection and the antibiotics used is presented in Fig. 1. Bacterial identification was confirmed by matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS [Biotyper; Bruker Daltonics, Billerica, MA]). Isolates were susceptibility