The Effect of β-Lactam Antibiotics on the Evolution of Ceftazidime/Avibactam and Cefiderocol Resistance in KPC-Producing Klebsiella pneumoniae

ABSTRACT In this study, we aimed to clarify the evolutionary trajectory of a Klebsiella pneumoniae carbapenemase (KPC)-producing Klebsiella pneumoniae (KPC-Kp) population during β-lactam antibiotic therapy. Five KPC-Kp isolates were collected from a single patient. Whole-genome sequencing and a comparative genomics analysis were performed on the isolates and all blaKPC-2-containing plasmids to predict the population evolution process. Growth competition and experimental evolution assays were conducted to reconstruct the evolutionary trajectory of the KPC-Kp population in vitro. Five KPC-Kp isolates (KPJCL-1 to KPJCL-5) were highly homologous, and all harbor an IncFII blaKPC-containing plasmid (pJCL-1 to pJCL-5). Although the genetic structures of these plasmids were almost identical, distinct copy numbers of the blaKPC-2 gene were detected. A single copy of blaKPC-2 was presented in pJCL-1, pJCL-2, and pJCL-5, two copies of blaKPC (blaKPC-2 and blaKPC-33) were presented in pJCL-3, and three copies of blaKPC-2 were presented in pJCL-4. The blaKPC-33-harboring KPJCL-3 isolate presented resistance to ceftazidime-avibactam and cefiderocol. The blaKPC-2 multicopy strain KPJCL-4 had an elevated ceftazidime-avibactam MIC. The patient had been exposed to ceftazidime, meropenem, and moxalactam, after which KPJCL-3 and KPJCL-4 were isolated, which both showed a significant competitive advantage under antimicrobial pressure in vitro. Experimental evolution assays revealed that blaKPC-2 multicopy-containing cells were increased in the original single-copy blaKPC-2-harboring KPJCL-2 population under selection with ceftazidime, meropenem, or moxalactam, generating a low-level ceftazidime-avibactam resistance phenotype. Moreover, blaKPC-2 mutants with a G532T substitution, G820 to C825 duplication, G532A substitution, G721 to G726 deletion, and A802 to C816 duplication increased in the blaKPC-2 multicopy-containing KPJCL-4 population, generating high-level ceftazidime-avibactam resistance and reduced cefiderocol susceptibility. Ceftazidime-avibactam and cefiderocol resistance can be selected by β-lactam antibiotics other than ceftazidime-avibactam. Notably, blaKPC-2 gene amplification and mutation are important in KPC-Kp evolution under antibiotic selection.

Resistance to CAZ/AVI in Klebsiella pneumoniae carbapenemase (KPC)-producing Klebsiella pneumoniae (KPC-Kp) is associated mainly with bla KPC-2 or bla KPC-3 mutation and bla KPC overexpression (4). bla KPC overexpression is often accompanied by changes in membrane permeability or increased efflux pump expression to yield low-level CAZ/AVI resistance (5,6). High-level CAZ/AVI resistance is often related to bla KPC-2 and bla KPC-3 mutations. The most common mutation is the G532T substitution, which leads to the D179Y amino acid change in KPC-3 (named KPC-31) and KPC-2 (named KPC-33). Most KPC variants were selected after exposure to CAZ/AVI (7), and some of the KPC variants conferred cross-resistance to CAZ/AVI and CFDC (8,9), which is a clinical concern in the application of antibiotics.
Although CAZ/AVI-resistant isolates are associated with CAZ/AVI selection pressure, data showed that 33% of cases had no previous CAZ/AVI exposure (7). Here, we analyzed five homologous KPC-Kp isolates from a single patient. After exposure to ceftazidime, meropenem, and moxalactam, the KPC-Kp population showed a temporal pattern of evolution to high-level CAZ/AVI and CFDC resistance through bla KPC-2 gene amplification and mutation. We reconstructed the evolutionary pathway in vitro using experimental evolution assays.

RESULTS
Patient and isolates. KPJCL-1 to KPJCL-5 were sequentially isolated from a 56-year-old male patient who was admitted to the intensive care unit (ICU) due to cerebral hemorrhage. KPJCL-1 was isolated from urine on hospitalization day 55 and diagnosed as colonization. A scrotal abscess developed on hospitalization day 91, and KPJCL-2 was isolated. Although the patient underwent scrotal abscess incision and drainage, the abscess was sustained, and a fistulous tract appeared between the abscess and the urinary tract. KPJCL-3 was isolated from the abscess on hospitalization day 142, and KPJCL-4 was isolated from urine on hospitalization day 156. Before the isolation of KPJCL-3 and KPJCL-4, the patient was treated with moxalactam for 13 days (1.0 g once a day [q.d.], days 92 to 104), meropenem for 20 days (1.0 g every 12 h, days 104 to 106; 0.5 g every 12 h, days 106 to 123), and ceftazidime for 16 days (1.0 g every 12 h, days 123 to 138). The scrotal abscess persisted, and KPJCL-5 was sampled from the abscess on hospitalization day 196. The patient died on hospitalization day 220 due to septic shock (Fig. S1 in the supplemental material).
An elevated bla KPC-2 copy number contributes to low-level CAZ/AVI resistance. The relative bla KPC-2 copy number (copy number relative to the bla KPC-2 -containing plasmid) was 4.750 6 0.501 in the ceftazidime-selected subpopulation (KPCAZ-E), 5.089 6 0.860 in the meropenem-selected subpopulation (KPMEN-E), 5.598 6 0.373 in the moxalactamselected subpopulation (KPMOX-E), and 5.615 6 0.681 in the subpopulation exposed to all three antibiotics (KPMulti-E). The relative bla KPC -containing plasmid copy numbers were unchanged among the isolates ( Fig. 3c and d). The selected bla KPC-2 multicopy subpopulation   (Table S3). Experimental evolution of KPJCL-4 under ceftazidime pressure. KPJCL-4 has the closest genetic relationship with KPJCL-3, and we wanted to observe the distribution of mutations occurring in the three bla KPC-2 regions (the TraI, IS, and tandem regions); therefore, KPJCL-4 was chosen as the target population for evolution. The bla KPC-2 G532T mutation frequency in KPJCL-4 was 6.45 Â 10 210 , which was much lower than the bla KPC-2 gene amplification frequency (Fig. S7). Of the 18 lineages cultured with a ceftazidime concentration of 512 mg/L, six populations developed increased frequency of resistant colonies on antibiotic plates, from an average frequency of 2.38 Â 10 27 after 24 h of exposure to 9.57 Â 10 24 after 6 days of exposure. No increase in the frequency of resistance subgroups was observed in control lineages in CAMHB (Fig. 3b).

DISCUSSION
In the present study, we demonstrate that clinical commonly used b-lactam antibiotics can select for CAZ/AVI and CFDC resistance. Using clinical data, drug susceptibility phenotypes, KPC-Kp genotypes, growth competition assays, and experimental evolution assays, we reconstructed the evolutionary trajectory of CAZ/AVI and CFDC resistance in KPC-Kp observed in vivo.
Based on the genomic data and the temporal pattern of isolation, we hypothesized that the evolutionary trajectory of the bacterial population developed from a bla KPC-2 singlecopy population to a bla KPC-2 multicopy population and then a bla KPC G532T mutant population (as summarized in Fig. S8 in the supplemental material). Amplification of bla KPC-2 via an increased copy number on a plasmid facilitates the development of antibiotic resistance. First, the increased bla KPC-2 copy number on plasmids increases bacterial tolerance to b-lactam antibiotics. Second, it increases the opportunity for mutations to arise (10). Our results showed an up to 6-fold increase in the bla KPC-2 copy number on plasmids after antibiotic selection, which provides a substantial increase in the number of targets for the occurrence of random mutations within bla KPC-2 . Third, multiple copies of bla KPC-2 genes on the same plasmid with one of the bla KPC-2 mutations enabled the strain to compete against the trade- off effect of bla KPC mutation. For the majority of reported CAZ/AVI-resistant strains, the mutation of a single copy of bla KPC in the plasmid restores susceptibility to carbapenems (11). In this case, two copies of bla KPC (bla KPC-2 and bla KPC-33 ) present in the same plasmid in pJCL-3 enable KPJCL-3 to resist both carbapenems and CAZ/AVI. Fourth, amplification can occur at much higher rates than point mutations (12,13). The difference in the frequencies of amplification (5.96 Â 10 27 ) and mutation (6.45 Â 10 210 ) in this study likely explains the evolutionary trajectory of the KPC-Kp population, with gene amplification emerging first under antibiotic selection. After repeated exposure to ceftazidime at concentrations of 512 mg/L, 6 populations developed mutated colonies with reduced susceptibility to CAZ/AVI and CFDC, and half of these resistant subpopulations had the D179Y substitution in KPC-2. In addition, other mutations associated with CAZ/AVI and CFDC resistance were identified. One of the variants is a D179N substitution in KPC-2. The D179N substitution in KPC-3 has been reported in a previous study of bacteria derived under CAZ/AVI pressure in vitro (11). The D179N substitution might increase the affinity between ceftazidime and the variant KPC, thereby preventing the binding of avibactam (14). Another variant is KPC-14, which was reported in a case after prolonged exposure to CAZ/AVI (15). KPC-14 does not affect the inhibitory properties of avibactam; however, it possesses a higher ceftazidime affinity and increased ceftazidime hydrolysis (16). To the best of our knowledge, the KPC-2 variants E275-A276dup and K269-H273dup we observed in our experiments have not been described thus far. The KPC-2 variants that emerged during ceftazidime exposure overlap with those selected under CAZ/AVI pressure, with mutations clustering in three regions, that is, amino acids 164 to 179 (X-loop), 234 to 242 (b-strand), and 263 to 277 (in the vicinity of the X-loop and the hinge-loop) (11,17,18). The tolerance to amino acid substitutions, insertions, and deletions reflects the evolutionary diversity of bla KPC , which is a challenge for the clinical application of CAZ/AVI and CFDC.
The concentrations of ceftazidime, meropenem, and moxalactam that we tested in vitro are not readily achieved in most infection sites. However, KPC-Kp strains were isolated from urine or a scrotal abscess in this study, and a fistulous tract appeared between the abscess and the urinary tract in the patient; thus, the urinary tract and the scrotal abscess together formed a "pool" for these pathogen populations. Ceftazidime, meropenem, and moxalactam are excreted mainly via the kidneys, and high concentrations of ceftazidime (2 g intravenous [i.v.], 8,000 to 16,000 mg/L, 0 to 3 h; 110 to 555 mg/L, 6 to 12 h) (19), meropenem (1 g i.v., 45.4 to 1,141.6 mg/L, 0 to 8 h) (20), and moxalactam (1 g i.v., 594 to 2,094 mg/L, 0 to 8 h) (21) that reached the selection window of the bla KPC-2 multicopy and bla KPC-2 mutated isolates were detected in the urine, which provides selection pressure for the emergence of resistant isolates.
Taken together, we report the evolutionary trajectory of the KPC-Kp population under clinical antibiotic pressure. The evolution is initiated by an increase in the copy number of bla KPC and is then further enhanced by point mutations within the bla KPC gene. These findings broaden our understanding of antibiotic resistance development in clinical settings and hence will significantly benefit carbapenem-resistant Klebsiella pneumoniae (CRKP) infection treatment.

MATERIALS AND METHODS
Strains, ethics, and susceptibility testing. Strains KPJCL-1 to KPJCL-5 were sequentially isolated from a patient in the ICU. The study was approved by the ethics committee (20170301-3). MICs were determined using broth microdilution (levofloxacin, tigecycline, colistin, and cefiderocol) or agar dilution (imipenem, meropenem, ceftazidime, moxalactam, and CAZ/AVI) according to Clinical and Laboratory Standards Institute (CLSI) standards (22). The MICs of cefiderocol were tested in iron-depleted medium. MICs were interpreted using CLSI breakpoints, where available. The FDA resistant breakpoint of $8 mg/L was applied for tigecycline. Escherichia coli ATCC 25922 served as a quality control strain, and the mcr-1-positive strain E-FQ (23) served as an extra quality control strain in colistin MIC tests.
Transformation experiment. The bla KPC-33 and bla KPC-2 sequences with their putative promoters were amplified (primers are listed in Table S1 in the supplemental material) and cloned into plasmid pCR2.1 (Invitrogen, Carlsbad, CA, USA). K. pneumoniae ATCC 13883 was used for transformation.
Bacteria growth assay. Overnight cultures were diluted 1:100 in cation-adjusted Mueller-Hinton broth (CAMHB) containing a gradient of antibiotic concentrations. Bacterial growth was detected in three replicates using a Bioscreen C MBR machine (Oy Growth Curves Ab Ltd., Finland). Optical density at 600 nm (OD 600 ) values of the isolates were compared using a two-sided Mann-Whitney U test and are presented as the medians (maximum to minimum values). A P value of ,0.05 was considered a significant difference. The statistical software used in this study was Prism5.
Growth competition experiments. Competition experiments were conducted as described in a previous study (24) in the presence of antibiotic concentrations of significant differences in growth (medium con- The adjusted cultures of KPJCL-3 and KPJCL-4 were mixed at a 1:1 ratio (KPJCL-3:KPJCL-4) in 2 mL of CAMHB or ceftazidime (512 mg/L) medium. The mixed cultures were grown at 37°C for 24 h. The total number of isolates was determined by plating aliquots onto nonselective plates. The numbers of KPJCL-3 and KPJCL-4 colonies were calculated by plating aliquots onto plates containing 32/4 mg/L CAZ/AVI or 512 mg/L moxalactam, respectively. Each ratio of mixed culture was performed in three independent experiments. Competitive advantage was calculated as the competition index (CI), where CI = (isolate A/isolate B) t24 /(isolate A/isolate B) t0 . The ln (CI) values were compared using a two-sided Mann-Whitney U test and are presented as the means 6 standard deviation (SD).
Estimation of the mutation and amplification frequencies.
Overnight KPJCL-2 and KPJCL-4 populations were harvested. The total number of populations was determined by plating serial dilutions on Mueller-Hinton agar (MHA) plates. The bla KPC-2 G532T mutants in the KPJCL-4 population were selected on plates containing 32 mg/L/4 mg/L CAZ/AVI. The bla KPC-2 gene of colonies that grew on selective plates was amplified and sequenced to confirm the presence of bla KPC-2 G532T mutation. The bla KPC-2 multicopy strains in the KPJCL-2 population were selected through growth on plates containing 512 mg/L moxalactam and were confirmed by quantitative PCR (primers are listed in Table S1). The frequencies were determined by dividing the median number of mutants/multicopy strains by the average number of populations (25).
Experimental evolution assays. In KPJCL-2 evolution assays, clones of KPJCL-2 were passaged (1:100) for 6 days in CAMHB or in CAMHB containing ceftazidime (128 mg/L), meropenem (64 mg/L), or moxalactam (512 mg/L) medium or were passaged sequentially in medium containing the three antibiotics each for 2 days. In the KPJCL-4 evolution assay, clones of KPJCL-4 were passaged (1:100) for 6 days in CAMHB and ceftazidime (512 mg/L) medium. Before every transfer, the proportion of subgroups was calculated by dividing the number of colonies growing on plates containing 512 mg/L moxalactam (KPJCL-2 evolution) or 32 mg/L/4 mg/L CAZ/AVI (KPJCL-4 evolution) by the number of total isolates. PCR and quantitative PCR (qPCR) were used to confirm the subgroups.
qPCR. 2 -DDCT method was used for relative quantitation of the bla KPC gene normalized to the plasmid replication protein gene (rep) or rep normalized to the pgi housekeeping gene (primers are listed in Table S1). The mean C T value was calculated from three replicate reactions, and the DDC T value was calculated from three different DNA preparations. Data are presented as the means 6 SD.
Data availability. The sequencing data have been deposited in GenBank under the accession numbers JAKJSA000000000, JAKJRZ000000000, JAKJRY000000000, JAKJRX000000000, and JAKJRW000000000.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 1.7 MB.
analysis protocols. X.W. and A.M. contributed to writing the manuscript. L.S. contributed to the collection of specimens and data management. P.Z., H.H., and Q.S. were responsible for the lab work with all isolates.