High-Level Aminoglycoside Resistance in Human Clinical Klebsiella pneumoniae Complex Isolates and Characteristics of armA-Carrying IncHI5 Plasmids

Aminoglycosides are important options for treating life-threatening infections. However, high levels of aminoglycoside resistance (HLAR) among Klebsiella pneumoniae isolates have been observed to be increasing frequently. In this study, a total of 292 isolates of the K. pneumoniae complex from a teaching hospital in China were analyzed. Among these isolates, the percentage of HLAR strains was 13.7% (40/292), and 15 aminoglycoside resistance genes were identified among the HLAR strains, with rmtB being the most dominant resistance gene (70%, 28/40). We also described an armA-carrying Klebsiella variicola strain KP2757 that exhibited a high-level resistance to all aminoglycosides tested. Whole-genome sequencing of KP2757 demonstrated that the strain contained one chromosome and three plasmids, with all the aminoglycoside resistance genes (including two copies of armA and six AME genes) being located on a conjugative plasmid, p2757-346, belonging to type IncHI5. Comparative genomic analysis of eight IncHI5 plasmids showed that six of them carried two copies of the intact armA gene in the complete or truncated Tn1548 transposon. To the best of our knowledge, for the first time, we observed that two copies of armA together with six AME genes coexisted on the same plasmid in a strain of K. variicola with HLAR. Comparative genomic analysis of eight armA-carrying IncHI5 plasmids isolated from humans and sediment was performed, suggesting the potential for dissemination of these plasmids among bacteria from different sources. These results demonstrated the necessity of monitoring the prevalence of IncHI5 plasmids to restrict their worldwide dissemination.

Aminoglycosides are important options for treating life-threatening infections. However, high levels of aminoglycoside resistance (HLAR) among Klebsiella pneumoniae isolates have been observed to be increasing frequently. In this study, a total of 292 isolates of the K. pneumoniae complex from a teaching hospital in China were analyzed. Among these isolates, the percentage of HLAR strains was 13.7% (40/292), and 15 aminoglycoside resistance genes were identified among the HLAR strains, with rmtB being the most dominant resistance gene (70%, 28/40). We also described an armA-carrying Klebsiella variicola strain KP2757 that exhibited a high-level resistance to all aminoglycosides tested. Whole-genome sequencing of KP2757 demonstrated that the strain contained one chromosome and three plasmids, with all the aminoglycoside resistance genes (including two copies of armA and six AME genes) being located on a conjugative plasmid, p2757-346, belonging to type IncHI5. Comparative genomic analysis of eight IncHI5 plasmids showed that six of them carried two copies of the intact armA gene in the complete or truncated Tn1548 transposon. To the best of our knowledge, for the first time, we observed that two copies of armA together with six AME genes coexisted on the same plasmid in a strain of K. variicola with HLAR. Comparative genomic analysis of eight armA-carrying IncHI5 plasmids isolated from humans and sediment was performed, suggesting the potential for dissemination of these plasmids among bacteria from different sources. These results demonstrated the necessity of monitoring the prevalence of IncHI5 plasmids to restrict their worldwide dissemination.

INTRODUCTION
Klebsiella pneumoniae (K. pneumoniae) is one of the most common pathogens responsible for human infections. The emergence of highly antibiotic-resistant K. pneumoniae has become a major challenge facing clinical management and global public health (Kathryn et al., 2015).
Aminoglycosides are important options for treating lifethreatening infections and are generally administered in combination with β-lactam agents (Becker and Cooper, 2013). However, increasing rates of aminoglycoside resistance in K. pneumoniae have been reported in recent years (Nasiri et al., 2018). Moreover, an increase in the aminoglycoside resistance rate in Escherichia coli isolates has been reported in France, leading to global concerns (Caméléna et al., 2020). Studies of aminoglycoside resistance in different species or populations have subsequently been conducted. A high level of aminoglycoside resistance in Acinetobacter baumannii and Pseudomonas aeruginosa isolated from a hospital in Brazil was observed (Ballaben et al., 2018), and an alarmingly high resistance to aminoglycosides has also been reported in diseased animals and clinical isolates of Enterobacterales from China (Fang-you et al., 2010;Yu-Ting et al., 2013).
The most common mechanism of resistance to aminoglycosides involves aminoglycoside-modifying enzymes (AMEs). These enzymes include acetyltransferases (AACs), nucleotidyltransferases (ANTs), and phosphotransferases (APHs), which differ in their ability to modify aminoglycosides (Miró et al., 2013). Other mechanisms contributing to aminoglycoside resistance involve the upregulation of efflux pumps (Poole, 2004) and reduced intake of antibiotics into the bacterial cell (Nakamatsu et al., 2007); however, the production of 16S rRNA methyltransferase (16S-RMTase) is another important mechanism mediating resistance to nearly all clinically available aminoglycosides. Different types of 16S-RMTase genes, including rmtA, rmtB, rmtC, rmtD, rmtE, rmtF, rmtG, rmtH, armA, and npmA, have been identified (Doi et al., 2016). Of these genes, rmtB and armA represent the most widespread 16S rRNA methylase genes (Caméléna et al., 2020). Genes encoding AMEs or methylases can be carried in integrons or transposons located in a variety of plasmids, which may facilitate the rapid spread of 16S-RMTase genes and AME genes among bacterial species (Nagasawa et al., 2014).

Bacterial Strains
A total of 292 nonrepetitive isolates of the K. pneumonia complex were isolated from sputum, urine, blood, secretion, and pus samples for routine examination from 2015 to 2017 in an affiliated hospital of Wenzhou Medical University in Zhejiang Province, China. These strains were identified using a VITEK 2 automated microbiology analyzer (BioMerieux Corporate, Craponne, France). Given that the strains could not be distinguished into exact species on the basis of their biological phenotypes and multilocus sequence typing (MLST; Long et al., 2017;Yatao et al., 2018), we referred to them collectively as the K. pneumonia complex.

Antimicrobial Susceptibility Testing
The minimum inhibitory concentrations (MICs) were determined by the agar dilution method and interpreted in accordance with the Clinical and Laboratory Standards Institute (CLSI) guidelines. As no breakpoint criteria for streptomycin were established by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) or CLSI, the corresponding results were interpreted according to the criteria of the US Food and Drug Administration (Ya et al., 2017). High-level resistance of the bacterium to aminoglycosides was defined according to the MICs of certain antibiotics. For streptomycin (Pericàs et al., 2017) and kanamycin, MIC was ≥1,024 mg/L. For netilmicin, the MIC was ≥512 mg/L, while for gentamicin, amikacin, and tobramycin, the MIC was ≥256 mg/L (Caméléna et al., 2020). E. coli ATCC 25922 was used as a quality control strain.

Genome Sequencing, Assembly, Annotation, and Bioinformatic Analysis
In order to identify aminoglycoside resistance genes that were carried by all strains, we performed pooled genomic DNA sequencing for 292 isolates. Each strain was cultured in 5 ml of lysogeny broth (LB) at 37°C for 16 h, and the strains were subsequently pooled together. Whole-cell DNA was extracted using an AxyPrep Bacterial Genomic DNA Miniprep Kit (Axygen Scientific, Union City, CA, United States) and pooled DNA sequencing was performed at Shanghai Sunny Biotechnology Co., Ltd. (Shanghai, China). Megahit (Li et al., 2015) was used to assemble the pooled sequencing data. In addition, three strains (KP2757, KP2723, and KP1878) showing high-level resistance to aminoglycosides were selected for whole-genome sequencing (WGS). Genomic DNA was extracted using the AxyPrep Bacterial Genomic DNA Miniprep kit and sequenced with Illumina NovaSeq 6,000 with 2 × 150-bp paired-end libraries (Illumina, Inc., San Diego, CA, United States) and PacBio RS II (Pacific Biosciences) systems with a 20-kb library at Shanghai Sunny Biotechnology Co., Ltd. (Shanghai, China). The Illumina reads were subjected to adaptor trimming and quality filtering (Phred quality score >20). Reads from PacBio RS II sequencing were assembled using the Canu v1.8 pipeline (Koren et al., 2017), and the quality of the genomic sequence was further corrected by reads from Illumina NovaSeq 6,000 sequencing.
Open reading frame (ORF) prediction of the pooled sequence data was performed using Prodigal, with the default parameters. Antibiotic resistance genes were predicted using both the Comprehensive Antibiotic Resistance Database (CARD; Moura et al., 2009) and the ResFinder (Jia et al., 2017) database. The relative abundance (sequencing depth) of a certain gene was calculated using BBMap. 1 Potential ORFs of the whole genome of an isolate were predicted using the RAST pipeline (Overbeek et al., 2014) and annotated against the UniProt/SwissProt and nonredundant protein databases using the BLASTX program, with an e-value threshold of 1e-5. Annotation of mobile genetic elements and resistance genes was performed using ISfinder (Siguier et al., 2006), INTEGRALL, CARD, and ResFinder, with the default parameters being employed accordingly.
Basic genomic features were constructed using BLAST Ring Image Generator (BRIG; Nabil-Fareed et al., 2011). The comparison of nucleotide sequences was visualized using the genoPlotR package. 2 A SNP tree was reconstructed with the pipeline CSI phylogeny accessible from the Center for Genomic Epidemiology. 3

Detection of Aminoglycoside Resistance Genes by PCR
The distribution of 16 aminoglycoside resistance genes was screened via PCR. The primers used for the resistance genes and their expected amplicon sizes and annealing temperatures are shown in Supplementary Table S1. The PCR products were sequenced, and the sequences were compared with those in the NCBI nucleotide database using BLAST. 4

Conjugation Experiments
Conjugation experiments were performed with sodium azideresistant E. coli J53 as the recipient and the armA-positive isolate KP2757 as the donor. Briefly, a mixture of the donor and recipient strains was placed onto a sterile nitrocellulose filter, which was subsequently placed on a brain heart infusion (BHI) agar plate and incubated at 24°C for 24 h. The transconjugant was selected on a Muller-Hinton (MH) agar plate containing 256 mg/L sodium azide plus 64 mg/L amikacin. The existence of armA and additional AMEs in the transconjugant was verified by PCR and Sanger sequencing. Antimicrobial susceptibility testing was further performed for the transconjugant.

Detection of Aminoglycoside Resistance Genes
The pooled DNA sequencing of the 292 isolates predicted 16 aminoglycoside resistance genes, of which 15 were identified among the 40 HLAR strains [the most dominant resistance gene was rmtB (70%, 28/40)]. In addition, 55% (22/40) of the isolates carried two or more resistance genes. The most prevalent combination of the resistance genes was rmtB and aadA2, and 15% (6/40) of the strains had this combination of genes. Of the 29 16S-RMTase gene-positive strains, strain KP2757 carried the most aminoglycoside resistance genes, namely, aac (3) Supplementary Table S2.

Comparative Genomic Analysis of Antibiotic Resistance Island
Plasmid p2757-346 contained two antibiotic resistance islands (ARIs), designated ARI-A and ARI-B (Figures 1, 4, 5). ARI-A measured 36,950 bp in length and was bracketed by IS26 and IS5075 elements from positions 307,130 to 344,080, in which six resistance genes (dfrA1, aadA5, sul1, armA, msrE, and mphE) were encoded. ARI-B was 77,593 bp in size and started from the truncated transposase gene ΔtnpA of ISEC63 at nucleotide sequence position 136,625 and ended in merR at 214,218. ARI-B encoded resistance genes providing resistance against various classes of antibiotics (aminoglycosides, β-lactams, fosfomycin, macrolides, rifampicin, quinolones, and sulfonamides), mer (merRTPCADEY) genes, and ars (arsR1BCH) resistance genes (Figure 1). Pairwise sequence comparisons via BLASTN showed that significant differences between the eight plasmids were observed in the ARIs. In addition, ARI-B in p19051-IMP had high similarity with that in p2757-346 (approximately 89% coverage and 99% identity).
All eight plasmids carried two copies of the intact armA gene (one encoded in ARI-A and the other in ARI-B), with the exceptions of pSIM-1-BJ01, which did not have an armA gene in the ARI-B region, and p12208-IMP, which had a truncated armA in the ARI-B region. Among these plasmids, the armA gene was located in the Tn1548 or truncated Tn1548 transposon (Figures 4, 5).

Dissemination of armA-Carrying IncHI5 Plasmids
p2757-346 is a conjugative plasmid. The antibiotic susceptibilities of the transconjugant are shown in Table 2. Moreover, bacteria carrying the armA-encoding IncHI5 plasmids were isolated FIGURE 1 | Schematic diagram of the p2757-346 plasmid. The various colors indicate different gene functions, including replication (in purple), transfer (in green), maintenance (in pink), and resistance (in red), and mobile elements (in blue). The circles show (from outside to inside) predicted coding sequences, the GC content, the GC skew, and the scale (in kilobases).
from humans and sediment (no data about bacteria of animal origin available), suggesting the potential of this plasmid in mediating the dissemination of the 16S-RMTase-encoding genes among humans and throughout the environment (Supplementary Table S4).

DISCUSSION
In this work, 13.7% (40/292) of clinical K. pneumoniae complex isolates collected in the period of 2015-2017 were high-level aminoglycoside-resistant strains. Most of them showed high  MICs for aminoglycosides, especially kanamycin, with 97.2% (39/40) showing an MIC of >1,024 mg/L, and 9.9% (29/292) of strains showed high MICs for gentamicin, amikacin, and tobramycin. A previous study conducted in the same district (Wenzhou, China) found that only 6.2% (21/337) of the K. pneumoniae isolates collected from 2006 to 2007 showed high MICs to gentamicin, amikacin, and tobramycin (MIC ≥ 256 mg/L; Fangyou et al., 2009). The number of K. pneumoniae complex isolates resistant to all three antibiotics increased by approximately 3.7% (a relative increase of 59.7%) during the past 10 years. High-level aminoglycoside-resistant K. pneumoniae isolates have been reported in different districts and countries (Sabtcheva et al., 2008;Qiong et al., 2009). However, to the best of our knowledge, no other publication has reported the percentage of K. pneumoniae with HLAR, although high levels of aminoglycoside resistance in other species have been reported. The percentages of HLAR (MICs for both gentamicin and amikacin were higher than 512 mg/L) of the clinical A. baumannii isolates from Beijing, China, in 2006, 2007, 2008were 52.63, 65.22, 51.11, and 70.83%, respectively (Lu et al., 2014. Thirty of the 137 aminoglycoside-resistant E. coli collected from nine hospitals in France showed high-level resistance (MIC ≥ 256 mg/L) to gentamicin, amikacin, and tobramycin (Caméléna et al., 2020). The prevalence of high-level aminoglycoside-resistant bacteria is worrisome.
In this work, two copies of armA and multiple AME genes were found in a transferable IncHI5 plasmid. armA genes have previously been identified in IncL/M, IncA/C, IncX, IncHI1, and IncF plasmids (Dolejska et al., 2013;Caméléna et al., 2020). Compared with armA-positive IncHI5 plasmids, to the best of our knowledge, IncL/M, IncA/C, IncX, IncHI1, and IncF plasmids have never been reported to carry two copies of armA.
Comparative genomic analysis revealed that the genetic context of the armA gene was similar in IncHI5 plasmids from K. pneumoniae, K. variicola, and K. michiganensis. The module of armA-related transposable elements was highly similar to those of armA harboring IncFII plasmids (Xiang-Dang et al., 2012) and IncA/C plasmids (Patrick et al., 2012), suggesting that the armA module could disseminate among plasmids of different incompatibility groups.
Except for armA and multiple AME genes, extendedspectrum β-lactamase (ESBL) genes were colocalized with genes conferring resistance to fosfomycin, macrolides, rifampicin, quinolones, and sulfonamides on the IncHI5 plasmid p2757-346. Acquired 16S-RMTase genes were mainly identified in transferable plasmids and were always associated with ESBL and carbapenemase genes or other resistance genes (Laurent et al., 2014). Plasmids carrying resistance genes may cause the global spread of resistance determinants, constituting an emerging global threat. In recent years, in addition to the phenomenon of resistance genes of various antibiotic classes being identified in a single bacterium, bacteria with multiple copies of identical resistance genes or gene arrays have been detected more often (Naas et al., 2001;Ding-Qiang et al., 2016;Teng et al., 2018). This phenomenon makes the treatment of infectious diseases more difficult and threatens human health.

CONCLUSION
To the best of our knowledge, we first described that two copies of armA together with six AME genes coexisted on one plasmid in a HLAR K. variicola isolate. Comparative genomic analysis showed that most armA-positive IncHI5 plasmids harbored two copies of the armA gene, which had not been previously reported in IncL/M, IncA/C, IncX, IncHI1, and IncF plasmids. Eight armA-encoding IncHI5 plasmids were isolated from humans and sediment, suggesting the potential FIGURE 5 | Comparative genomic analysis of antibiotic resistance island B (ARI-B). Mobile elements are shown by green, brown, and blue arrows, respectively. The red arrows indicate antibiotic resistance genes, the pink arrows denote biocide resistance genes, and the gray arrows represent genes of other functions. The shading indicates 100% nucleotide sequence identity.

DATA AVAILABILITY STATEMENT
The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found in the article/Supplementary Material.