Emergence of the Novel Aminoglycoside Acetyltransferase Variant aac(6′)-Ib-D179Y and Acquisition of Colistin Heteroresistance in Carbapenem-Resistant Klebsiella pneumoniae Due to a Disrupting Mutation in the DNA Repair Enzyme MutS

ABSTRACT Amikacin and colistin are effective against carbapenem-resistant Klebsiella pneumoniae. In 2017, we successively isolated three carbapenem-resistant K. pneumoniae isolates (ST967) from a patient with chronic renal failure in Japan. The first (SMKP01, sputum, day 0) and second (SMKP02, blood, day 14) strains were resistant to most antimicrobials tested but still susceptible to amikacin (MICs of 4 and 0.5 mg/liter, respectively) and colistin (MIC of 0.5 mg/liter for both). The third strain (SMKP03, blood, day 51) was not susceptible to amikacin (MIC, 32 mg/liter), and its MIC for colistin varied (0.5 to 8 mg/liter). Whole-genome sequencing of SMKP01 revealed that 17 of 20 antimicrobial resistance genes, including qnrB91 (a novel qnrB2 variant) and aac(6′)-Ib-cr, were located on an 86.9-kb IncFII-IncQ plasmid. The qnrB91 conferred greater fluoroquinolone resistance than qnrB2. SMKP03 aac(6′)-Ib-cr that possessed a gene mutation that resulted in an R102W substitution, namely, aac(6′)-Ib-D179Y, made a greater contribution to amikacin resistance than did aac(6′)-Ib-cr. SMKP03 harbored a nonsense mutation in mutS, which encodes a DNA repair enzyme. Introduction of this mutation into SMKP01 (SMKP01mutSA307T) resulted in a dramatic increase (>58-fold) in the frequency of spontaneous amikacin-resistant mutants relative to SMKP01, and the substantial mutants possessed aac(6′)-Ib-D179Y. SMKP01mutSA307T exhibited an unstable MIC for colistin (0.5 to 8 mg/liter). The results demonstrate that a disruptive mutation in MutS, arising during the clinical course of an infection, created a platform for the acquisition of amikacin nonsusceptibility and colistin heteroresistance in multidrug-resistant K. pneumoniae, mediated by the elevated frequency of spontaneous mutations.

course of an infection, created a context for the acquisition of colistin resistance and the emergence of a novel variant of the amikacin resistance gene in multidrug-resistant K. pneumoniae via an increase in the frequency of spontaneous mutation. This observation is important for understanding how K. pneumoniae develops multidrug resistance during infection and could potentially lead to new antimicrobial treatments for high-risk pathological microbes.
KEYWORDS antimicrobial resistance, mutS, qnrB91, aac(69)-Ib-D179Y, amikacin resistance, colistin resistance, Klebsiella pneumoniae K lebsiella pneumoniae, an important nosocomial pathogen, can cause infections of the respiratory and urinary tracts, as well as bacteremia (1). Antimicrobial-resistant K. pneumoniae is more prone to cause such infections. Accordingly, the emergence of carbapenem-resistant K. pneumoniae is of serious clinical concern. Amikacin is the antimicrobial of choice for treating carbapenem-resistant K. pneumoniae because most isolates continue to be susceptible to amikacin (2,3). Colistin and tigecycline are last-line drugs for fighting multidrug-resistant Gram-negative bacteria (4,5).
In this study, we successively isolated three carbapenem-resistant K. pneumoniae isolates during the clinical course of a patient in Japan. Although the first and second strains were susceptible to amikacin and colistin, the third strain was no longer susceptible to amikacin and exhibited unstable susceptibility to colistin (ranging from a susceptible to a resistant phenotype). We analyzed the resistance mechanisms observed in the third strain.

RESULTS
Isolation and antimicrobial susceptibility of three carbapenem-resistant K. pneumoniae isolates. Among three carbapenem-resistant K. pneumoniae isolates obtained from the patient with chronic renal failure, the first (SMKP01) and second (SMKP02) were resistant to most antimicrobials tested but still susceptible to amikacin (MIC of 2 and 0.5 mg/liter, respectively) and colistin (MIC of 0.5 mg/liter for both) ( Table 1). The third strain (SMKP03) was not susceptible to amikacin (MIC of 32 mg/liter), and the MIC for colistin was unstable (MICs ranging from 0.5 to 8 mg/liter, by multiple repetition of the susceptibility test). These three strains were the same multilocus sequence type (MLST), ST967.
We identified a novel qnrB2-like variant with the R87C amino acid substitution due to a single-nucleotide mutation at nucleotide position 259 (CGC!GGC). We registered this variant as qnrB91 according to the recommendation of Jacoby et al. (https:// externalwebapps.lahey.org/studies/QNR_Studies.aspx) (10).
Comparison of qnrB91 and qnrB2 with respect to fluoroquinolone susceptibility. We constructed expression vectors for qnrB91 or qnrB2. Introduction of either vector into K. pneumoniae ATCC 13883 increased the fluoroquinolone MIC, although the effect of qnrB91 was .8-fold greater than that of qnrB2 ( Table 2).
Frequencies of amikacin-resistant mutants arising from SMKP01 harboring a mutation in mutS. We used pORTMAGE to construct a SMKP01 derivative harboring the mutS mutation from SMKP03, SMKP01mutS A307T . SMKP01mutS A307T had a higher frequency of spontaneous amikacin-resistant mutants (9.8 Â 10 27 ), as determined by selection with amikacin in vitro, relative to the parental strain (1.7 Â 10 28 ) (57.7-fold; P , 0.01) ( Fig. 4A and B). The R102W mutation in AAC(69)-Ib-cr [AAC(69)-Ib-D179Y], which was detected in SMKP03, was only observed in selected mutants derived from SMKP01mutS A307T (Fig. 4C). These results suggested that the nonsusceptibility of SMKP03 to amikacin was acquired due to the R102W point mutation in AAC(69)-Ib-cr and that this mutation arose due to the disruption of MutS.
Occurrence of colistin-resistant mutants in SMKP03. SMKP03 exhibited an unstable MIC for colistin that varied among examinations (Table 1). In the MIC measurement, "skip-well" phenomena were observed in SMKP03 but not in SMKP01 and SMKP02 (Fig. 5D). Hence, we examined the emergence of colistin-resistant mutants from these strains and SMKP01mutS A307T in the presence of colistin (2 mg/liter). SMKP03 and SMKP01mutS A307T had a markedly higher frequency of colistin-resistant colonies (3.3 Â 10 25 and 3.2 Â 10 25 , respectively) than SMKP01 (4.1 Â 10 27 ) (80.0-fold  [P = 0.02] and 78.8-fold [P = 0.01], respectively) ( Fig. 5A). In the broth culture, the colistin-resistant mutants were emerged in the presence of 2 mg/liter colistin in several culture batches of SMKP03 and SMKP01mutS A307T but not in all batches of SMKP01 and SMKP02 ( Fig. 5B and C). The colistin-resistant mutants derived from SMKP01mutS A307T had colistin MICs from 16 to .64 mg/liter, and these values did not alter after several passages in plain Mueller-Hinton II broth.
Prevalence of K. pneumoniae mutS-null isolates. To evaluate the dissemination of K. pneumoniae isolates with putative disruptions of functional MutS, we sought to identify mutS-null mutations (such as nonsense mutations, frameshift mutations, gene deletions, and gene insertions) from a database of complete genome sequences of 454 K. pneumoniae isolates. A total of 3.1% (14/454 isolates) of K. pneumoniae isolates possessed the mutS-null mutation. These 14 isolates were composed of different MLSTs (see Table S4).

DISCUSSION
In this study, we observed mutS-mediated acquisition of resistance (or nonsusceptibility) to amikacin and colistin in K. pneumoniae during the clinical course in a patient (Fig. 6). MutS is a DNA mismatch repair enzyme that is essential for combating the adverse effects of damage on the genome. Functional disruption of MutS by frameshift or gene deletion leads to a hypermutable phenotype (14,15). Six regions of MutS, N-19   (16). These functional regions were truncated by a nonsense mutation in MutS of K. pneumoniae isolate SMKP03, and the strain exhibited a hypermutable phenotype.
The hypermutable phenotype acquired by disruption of MutS confers a selective advantage despite the fact that inactive MutS is inferior to the wild type in a nonselective environment (15). Consistent with this, SMKP03 developed amikacin nonsusceptibility and colistin resistance during exposure to those agents in vitro more frequently than the two other isolates, SMKP01 and SMKP02, which lacked mutations in MutS. These observations suggested that acquisition of a hypermutable phenotype caused by loss of MutS function was an adaptive strategy for overcoming antimicrobial selection pressure. In other words, functional disruption of MutS offers a platform for acquisition of novel resistance mechanisms via novel mutations. aac(69)-Ib-D179Y is a previously unreported chimeric variant of aac(69)-Ib and aac(69)-Ib-cr (Fig. 1). AAC(69)-Ib-cr is a variant of aminoglycoside acetyltransferase AAC(69)-Ib, differing by two amino acid substitutions, W102R and D179Y (17). These substitutions confer N-acetylation at the amino group piperazine ring of several fluoroquinolones (17), such as ciprofloxacin and norfloxacin, thereby increasing resistance to these compounds; in contrast, they provide a lower level of amikacin resistance than AAC(69)-Ib (Table 3). We found that the novel variant, AAC(69)-Ib-D179Y, provided levels of amikacin and fluoroquinolone resistance comparable to those conferred by AAC(69)-Ib and AAC(69)-Ib-cr, respectively ( Table 3). The patient had been treated with amikacin (from day 15 to day 70), suggesting that AAC(69)-Ib-D179Y was generated by spontaneous mutation in a MutS-disrupted strain SMKP03 and subsequently selected during amikacin treatment. The instability of the colistin MIC in SMKP03 was also due to disruption of functional MutS. MutS disruption dramatically increased the frequency of colistin-resistant mutants and led to "skip-well" phenomena in MIC tests of colistin (Fig. 5). The frequency of such mutants in SMKP03 was 2.5 Â 10 25 , suggesting that spontaneous colistin-resistant mutants might have sometimes emerged during the MIC tests. Indeed, PAP test demonstrated that SMKP03 exhibited colistin-heteroresistant phenotype. Although previous studies reported the contribution of pmrAB or phoPQ mutation to colistin heteroresistance in K. pneumoniae (18,19), the identity of the bacterial factor(s) that generate this mutation was still unclear. This study first demonstrated that the mutS mutation contributes this mechanism. These might cause difficulty with colistin susceptibility testing in clinical laboratories. The patient had no clinical history of colistin treatment. Frequent emergence of spontaneous colistin resistance and the ability to evoke stable colistin resistance by a mutation occurring in phoP or pmrB in multidrug-resistant K. pneumoniae has a clinical impact because colistin is a last-line drug for treatment of this type of infection. Thus, our observations are a reminder that colistin administration during the course of an infection can easily select for colistin-resistant K. pneumoniae possessing an mutS mutation.
In this case, none of the genes encoding carbapenemases were detected in any of the three strains. Resistant and intermediate phenotypes to imipenem, meropenem, and doripenem may be explained by the combination of b-lactamase, CTX-M-3, and the absence of porins, OmpK35, and OmpK36, in K. pneumoniae, as seen in the present study and in previous studies (20,21). We also detected mutations of genes that encode multidrug efflux pumps (such as mdtA, mdtM, acrB, and oqxAB) in SMKP03 (see  Table S2). Among them, the mRNA expression of acrB was higher in SMKP03 than in SMKP01 and SMKP02 (P , 0.05). AcrAB is a major multidrug efflux pump that excretes several antimicrobials such as fluoroquinolones, minocycline, and tigecycline (22)(23)(24). The expression levels were regulated by SoxS, MarR, AcrR, and RamR (25)(26)(27). Thus, these mutations may contribute to the enhancement of moxifloxacin, sitafloxacin, and minocycline MICs in SMKP03, since soxS mutation was detected only in SMKP03 (see Table S2). SMKP01, SMKP02, and SMKP03 exhibited resistance to tigecycline. This may be explained by the higher acrB expression levels in SMKP01, SMKP02, and SMKP03 than that in the tigecycline-susceptible strain, ATCC 13383, due to several nonsynonymous mutations in acrR (P161R, G164A, F172S, R173G, L195V, F197I, and K201M) and RamR (T43A, K194Stop). Tigecycline is another option against multidrug-resistant Gramnegative bacteria with colistin. Therefore, the acquisition of colistin resistance by multidrug-resistant K. pneumoniae isolates with tigecycline resistance (as seen in SMKP03) is a serious problem in a clinical setting.
SMKP01 and SMKP03 carried a plasmid that harbored numerous resistance genes against multiple classes of antimicrobial agents. A BLAST search suggested that this plasmid was generated through combination of two plasmids, pQnrB_LL34 and p2, both previously reported in K. pneumoniae. This observation alerts us to the emergence of a plasmid capable of conferring a high degree of multidrug resistance. Among the resistance genes located on this plasmid, we identified a novel qnrB variant, qnrB91, which conferred a higher level of resistance to fluoroquinolones than qnrB2. Dissemination of such a plasmid, which causes the development and accumulation of multidrug resistance, among Enterobacteriaceae clinical isolates could make it more difficult to choose effective antimicrobial agents to combat infections in the future.
In conclusion, our results demonstrate that MutS disruption that arose during the clinical course in a patient led to the acquisition of amikacin nonsusceptibility and colistin resistance in multidrug-resistant K. pneumoniae. A previous study reported the frequent isolation of hypermutable Haemophilus influenzae with mutations in mutS from cystic fibrosis patients, as well as the clinical advantages of these mutants (28). Given that a substantial fraction of K. pneumoniae isolates (3.1% of entire K. pneumoniae isolates available in BLASTn search) harbor mutS-null mutations, these mutations probably affect the clinical condition, the effectiveness of antimicrobial treatment, and the diagnosis of antimicrobial susceptibilities. Therefore, mutS mutation should be taken into account when considering novel emergence mechanisms associated with bacterial survival and antimicrobial resistance in clinical settings.

MATERIALS AND METHODS
Bacterial isolation and antimicrobial susceptibility testing. We successively isolated three carbapenem-resistant K. pneumoniae isolates from a patient with chronic renal failure at Sapporo Medical University Hospital in 2017. The first strain (SMKP01) was isolated from sputum at day 0. The second (SMKP02) and third strains (SMKP03) were isolated blood at day 13 and day 50, respectively. Susceptibility to antimicrobials was determined by measuring MICs according to CLSI guidelines (29). This study was approved by Sapporo Medical University School of Medicine Ethics Committee (no. 302-1031).
After the sequence of each expression vector was confirmed, the constructs were electroporated into K. pneumoniae ATCC 13883. Three clones harboring each vector were selected and used to determine aminoglycoside and/or fluoroquinolone MICs in triplicate.
Construction of mutS-mutated SMKP01 (SMKP01mutS A307T ). The nonsense mutation of mutS (A307T) found in SMKP03 was introduced into SMKP01 by using pORTMAGE, as described previously (30). Hygromycin-integrated pORTMAGE (pORTMAGE-hyg) was generated by replacement of ampR of pORTMAGE-2 (Addgene, Watertown, MA) with a hygromycin resistance cassette (Gene Bridges, Heidelberg, Germany). ampR-deleted pORTMAGE2 was amplified by inverse PCR using primers PORTinvF and PORTinvR (see Table S5). The hygromycin resistance cassette with 15-bp regions overlapping the 59 and 39 sequences of the ampR-deleted region of pPORTMAGE2 was amplified using the primers PORTHyg-F and PORTHyg-R. These two fragments were assembled using NEBuilder HiFi DNA Assembly master mix according to the manufacturer's instructions and chemically transformed into E. coli DH5a. Transformants were selected on LB agar containing 100 mg/liter of hygromycin. Then, pORTMAGE-Hyg was introduced into SMKP01 by electroporation. Oligonucleotides (90 bp) for mutS containing A307T were designed using the Mage Oligo Design Tool (MODEST [http://modest.biosustain .dtu.dk]; MAGEmutS, ACTGGTCAATCAGGGCGAGTCGGTGGCCATTTGCGAGCAGATTGGCGACCCGGCCACC ACCtAGGGGCCGGTCGAGCGTAAAGTGGTGCG; the underlining indicates the stop codon introduced into SMKP01; bold lowercase "t" indicates the introduced nucleotide). Gene replacement in the mutants was confirmed by direct DNA sequencing.
Expression profile of OmpK35 and OmpK36. Outer membrane proteins of SMKP01, SMKP02, SMKP03, and ATCC 13833 were prepared as described previously (31). Protein concentrations were determined by using a BCA protein assay kit (Thermo Fisher Scientific K.K., Tokyo, Japan). Portions (2 mg) of total proteins were boiled for 5 min in Laemmli sample buffer and loaded onto 12% polyacrylamide gels. The gels were then stained with Coomassie brilliant blue.
Measurement of acrB, oqxB, bla SHV-27 , and fosA mRNA expression by real-time reversetranscription PCR. SMKP01, SMKP02, SMKP03, and ATCC 13383 were cultivated in 0.5 ml of tryptic soy broth overnight. Then, 150-ml portions of the cultures were inoculated into 10 ml of Muller-Hinton II broth (cation adjusted; Becton Dickinson, Franklin, Lakes, NJ), and the cells were grown for 1.5 h at 37°C. RNA was isolated using an RNeasy Plus minikit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. The concentration of RNA was measured spectrophotometrically using an Infinite M200 PRO instrument (Tecan, Kawasaki, Japan), and 0.5 mg was used to synthesize cDNA using ReverTra Ace qPCR RT Master Mix with gDNA remover (Toyobo, Tokyo, Japan). The expression of genes encoding acrB, oqxB, bla SHV-27 , and fosA was estimated using KOD SYBR qPCR mix (Toyobo). The primer pairs for acrB (acrB-F and acrB-R) (32) and oqxB (oqxB1806F and oqxB1806) (33) were used in previous studies. The primer pairs for bla SHV-27 and fosA are listed in Table S5 in the supplemental material. The PCR cycling conditions were as follows: initial activation at 95°C for 5 min, followed by 40 cycles at 95°C for 10 s and 60°C for 30 s. Reactions were performed in a LightCycler 480 II (Roche, Mannheim, Germany). rpoB genes served as an endogenous reference for normalizing expression levels (33). Data were calibrated against the baseline expression level of ATCC 13383, and fold changes in expression were calculated using the comparative C T method. The data are expressed as means 6 the standard deviations from three independent experiments.
Whole-genome sequencing. Genomic DNA was isolated by using a Wizard Genomic DNA purification kit (Promega, Madison, WI). The complete genome sequence was determined using the PacBio RS SMRT Portal (Pacific Biosciences, Menlo Park, CA). The filtered subreads (in fastq format) were assembled with 300-bp paired-end reads obtained by MiSeq sequencing (Illumina, San Diego, CA) using Unicycler (version 0.4.3) (34). This yielded a large circular contig (;5.15 Mb) and three smaller circular contigs (;111, ;86.9, and ;54.5 kb). To increase the accuracy of the final sequence, the sequences were curated by mapping analysis with short reads using the CLC Genomics Workbench (Qiagen). The finalized contig lengths were 5,155,320 bp (the SMKP03 genome), 111,317 bp (pSMKP03L, the large plasmid of SMKP03), 86,902 bp (pSMKP03M, the middle plasmid of SMKP03), and 54,548 bp (pSMKP03S, the small plasmid of SMKP03). Before registration, the sequences were automatically annotated by DFAST (legacy version) based on PROKKA (35).
The SNPs of SMKP02 and SMKP03 were detected with CLC Genomic Workbench using the basic variant detection method by sorting the obtained variant using a minimum frequency of 70%. The genome sequence of SMKP01 was used as a reference. Comprehensive genetic comparison between SMKP03 and its colistin-resistant mutants was performed, and nonsynonymous mutated genes were detected by the same method used for SNP detection.
Determination of MLST and detection of antimicrobial resistance genes were performed using MLST analysis and ResFinder, respectively, at the Center for Genomic Epidemiology (http://www .genomicepidemiology.org).
Measurement of the frequency of mitomycin C-resistant, amikacin-nonsusceptible, or colistinresistant mutants. Overnight cultures of strains in tryptic soy broth were spread on Mueller-Hinton II agar (Becton Dickinson)-with or without mitomycin C (8 mg/liter), amikacin (16 mg/liter), or colistin MutS Disruptive Mutation in K. pneumoniae (2 mg/liter)-and incubated at 37°C for 48 h. The frequency of the mitomycin C-resistant, amikacin-nonsusceptible, or colistin-resistant mutants was calculated by dividing the number of colonies that grew on Mueller-Hinton II agar containing the corresponding agent by the number of colonies that grew on plain Mueller-Hinton II agar. Experiments were performed in triplicate.
The frequency of amikacin-resistant mutants harboring aac(69)-Ib-D179Y was measured in triplicate. Appropriately 100 colonies grew on Mueller-Hinton II agar containing 16 mg/liter of amikacin; these clones were selected, and aac(69)-Ib-D179Y was detected by SNP PCR (see Table S5). Sanger DNA sequencing was performed for confirmation (see Table S5).
Bacterial growth determination. Bacterial growth was monitored by measuring the turbidity (that is, the optical density at 600 nm [OD 600 ]) using an Infinite M200 PRO multimode microplate reader (Tecan, Kawasaki, Japan). Strains (eight clones per strain) were grown in 0.5 ml of tryptic soy broth (Becton Dickinson) overnight at 37°C, and 1 Â 10 5 CFU/ml bacteria were cultured in 0.1 ml Mueller-Hinton II broth (Becton Dickinson) in a 96-well plate at 37°C with shaking at 140 rpm for 16 h. Bacterial growth curves were created based on measurements made every 10 min for 16 h at least in triplicate.
Defining colistin heteroresistance in SMKP03 and SMKP03-derived colistin-resistant mutants. We performed PAP test to determine whether SMKP03-derived colistin-resistant mutants exhibited heteroresistance or not according to a previous study, with slight modification (36). SMKP03 and the colistin-resistant mutants were cultured in tryptic soy broth overnight at 37°C (these strains were assessed in this test after growth at least 50 generations on Mueller-Hinton II agar without colistin). Then, 100-ml portions of an overnight culture (ca. 1 Â 10 9 CFU/ml) were spread on Mueller-Hinton II agar containing colistin (from 0.03125 to 128 mg/liter) and cultured 24 h at 37°C. The CFU were counted, and the colistin-resistant population was calculated based on the CFU that grew on Mueller-Hinton II agar containing colistin divided by the CFU that grew on Mueller-Hinton II agar without colistin. We determined "colistin heteroresistance" based on the following criteria: (i) the main population was clinically susceptible to colistin (,2 mg/liter) and (ii) the colistin concentration appearing in the colistin-resistant subpopulation (with a frequency greater than 1 Â 10 27 ) was at least 8-fold greater than the highest colistin concentration that does not affect growth of the main colistin-susceptible population (37).
Prevalence of mutS-null K. pneumoniae isolates in the NCBI database. We collected the mutS DNA sequences of 454 complete genome sequences from K. pneumoniae isolates in the NCBI database; these represented all complete genome sequences available as of 25 April 2020. Amino acid translation from the mutS DNA sequence was performed using the Sequence Manipulation Suite (https://www .bioinformatics.org/sms2/translate.html). The translated sequences were aligned by MAFFT (38) to identify each nonsynonymous mutation, including likely null mutations (such as nonsense mutations, frameshifts, gene deletions, or gene insertions). K. pneumoniae SMKP01 was used as a reference. Determination of MLST was performed using MLST analysis at the Center for Genomic Epidemiology (http://www.genomicepidemiology.org) with the genome data.
Statistical analysis. One-way analysis of variance and Dunn's multiple-comparison test were used for multiple comparisons (P , 0.05). A paired t test was used for comparisons of two groups (SMKP01 versus SMKP01mutS A307T ) (P , 0.05).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.

ACKNOWLEDGMENTS
This study was supported in part by the Japan Initiative for Global Research Network on Infectious Diseases (J-GRID) from the Ministry of Education, Culture, Sport, Science, and Technology in Japan; by the Japan Agency for Medical Research and Development (AMED; JP19ak0101118h0001 and 20jk0210004h0001); and by a grant from MEXT for the Joint Research Program of the Research Center for Zoonosis Control, Hokkaido University. This study was also partly supported by JSPS KAKENHI grant JP19K16648.
We thank Csaba Pál for giving us permission to construct pPORTMAGE-hyg.