Klebsiella pneumoniae, is a major pathogen that can cause both nosocomial- and community-acquired infections, such as urinary tract infections, bacteremia, respiratory infections and soft tissue infections, especially in immunocompromised individuals (Ullmann, 1998; Holt et al., 2015). The emergence of carbapenem-resistant K. pneumoniae (CRKP) has limited antibiotic therapeutic choices and led to serious challenges in clinical treatment and infection control (Rice, 2008; Livermore et al., 2015a; Moradigaravand et al., 2017). The World Health Organization (WHO) considers carbapenem-resistant Enterobacteriaceae to be critical priority pathogens that require new antibiotics (Shrivastava et al., 2018). Ceftazidime/avibactam (CZA) and tigecycline (TGC) are last-line choices for the treatment of CRKP infections.

CZA is a novel β-lactam-β-lactamase inhibitor combination with the ability to inhibit the activity of AmpC, extended-spectrum β-lactamases (ESBLs), class A carbapenemases such as KPC and class D carbapenemases such as OXA-48 but not class B carbapenemases such as NDM, IMP and VIM (Shirley, 2018). CZA was approved for the treatment of complicated intra-abdominal infections and urinary tract infections by the FDA on February 25, 2015. A multicenter, observational study reported that CZA could be a reasonable substitute for colistin for the treatment of CRKP infections with reduced mortality (Van Duin et al., 2018). In the International Network For Optimal Resistance Monitoring (INFORM) surveillance programme, the in vitro susceptibility of carbapenemase-positive and MBL-negative isolates to CZA was 99.8% between 2015 and 2017 (Spiliopoulou et al., 2020). However, acquisition of CZA resistance has been reported after CZA therapy in recent years. Resistance to CZA has been observed in strains with mutations in AmpC, bla _KPC-2 and bla _KPC-3, and the mutation point was mostly in Ω-loop in bla _KPC genes (Shirley, 2018; Zhang et al., 2020). In addition, high expression of KPC-3 was reported to be associated with CZA resistance (Humphries and Hemarajata, 2017).

The emergence of TGC resistance after exposure to TGC has also been reported (Duin et al., 2014; Lin et al., 2016; Ye et al., 2017). Overall, the primary TGC resistance mechanisms in K. pneumoniae include overexpression of efflux pumps, such as AcrAB and OqxAB, which can be mediated by regulators (RamR-RamA and OqxR-RarA) and global activators (MarA, SoxS and Rob) (Veleba and Schneiders, 2012). There also exist other known mechanisms, conferred by the Lon protease, Tet(A) protein and ribosomal protein S10, which is encoded by rpsJ (Villa et al., 2014). In addition, Linkevicius M et al. reported that lipopolysaccharide (LPS) defects could lead to reduced susceptibility to TGC (Linkevicius et al., 2016).

Among all the antibiotics tested by the INFORM surveillance programme (2015–2017) against 1,460 carbapenem-nonsusceptible Enterobacteriaceae isolates, CZA and TGC showed the highest susceptibility rates (73.0% and 78.1%, respectively) (Spiliopoulou et al., 2020). Strains resistant to these last-line antibiotics could pose a great threat to public health.

Here, we report a series of CRKP strains isolated in 2017 that have evolved resistance to both CZA and TGC during exposure to β-lactam antibiotics and TGC. The CRKP isolates resistant to CZA were isolated before CZA was approved in China. The aim of this study was to describe the phenotypic and genotypic adaption characteristics of these strains during in-host evolution and explain the molecular mechanisms of CZA and TGC resistance.

Under antibiotic pressure, bacteria adapt to the host or environment with genomic and phenotypic changes (Linkevicius et al., 2016). Here, we described a series of closely related CRKP strains that evolved resistance to CZA and TGC in vivo in a hospitalized patient. ST analysis showed that the strains belonged to a new ST, ST4496, closely related to ST11. Before the CZA- and TGC-resistant strain XDX51 was isolated, TGC was used for 27 days, and β-lactam antibiotics (cefoperazone-sulbactam and piperacillin-tazobactam) were prescribed alternately, while there was no exposure to CZA. Compared to the index strains, three genomic differences were identified in XDX51, including duplication of bla _KPC-2, ramR and wbbL insertions caused by IS elements. Ye et al. reported that deletion of the ramR ribosomal binding site could cause in vivo development of TGC resistance during TGC treatment (Ye et al., 2017). Long-term usage of antibiotics such as β-lactams and TGC could have led to the evolution of the series of CRKP strains in our study with gene mutations related to antibiotic resistance.

Increased KPC expression could enhance ceftazidime hydrolysis, which could not be completely inhibited by avibactam (Zhang et al., 2020). Both XDX16 and XDX51 carried an IncFII plasmid containing bla _KPC-2 and rmtB. The MICs of CZA were 4-fold higher for XDX51 than for XDX16. There was duplication of bla _KPC-2 on the XDX51 plasmid due to unequal crossover of the IS26 composite transposon causing enhanced KPC expression levels, as identified by qRT-PCR. KPC-23-producing K. pneumoniae has been reported to be resistant to CZA due to increased ceftazidime hydrolysis without prior exposure to CZA (Galani et al., 2019). Similarly, Humphries RM reported CZA resistance due to increased expression of KPC-3 in clinical isolates previously unexposed to CZA (Humphries and Hemarajata, 2017). Both situations above were combined with membrane porin deficiency. In addition, Ω-loop alterations caused by KPC mutations can prevent avibactam binding and lead to CZA resistance, such as mutations at position 179 in bla _KPC-2 or bla _KPC-3 (Livermore et al., 2015b; Barnes et al., 2017). CZA resistance caused by KPC point mutations tends to occur after the strains are exposed to CZA either in vivo or in vitro (Livermore et al., 2015b; Barnes et al., 2017; Shields et al., 2017). Additionally, Antinori E. reported that deletion in bla _KPC-3 in clinical K. pneumoniae could also result in CZA resistance (Antinori et al., 2020). In our study, the CZA-resistant strain evolved double copies of bla _KPC-2 compared with the index strains. An increase in the MICs of carbapenem and CAZ for XDX51 indicated enhanced CAZ hydrolysis due to overexpression of bla _KPC-2. Our conclusion was confirmed by other studies as well. Shen Z. et al. reported that the expression level of bla _KPC-2 in CRKP with CZA MIC 4-8 μg/mL group was 4.2-4.8-fold higher than that in CZA MIC 1-2 μg/mL and ≤0.5 μg/mL group, and hydrolysis activities of CAZ was 4-4.6-fold higher in CZA MIC 4-8 group than the other two groups, indicating that the enhanced expression of bla _KPC-2 could result in the decrease of CZA susceptibility due to the increased hydrolysis activity of CAZ (Shen et al., 2017). Hence, we considered the enhanced KPC expression was responsible for CZA resistance in XDX51.

Nonfunctional RamR can lead to TGC resistance via regulation of efflux pumps. Mutations in ramR, including deletion, mutation and insertion, resulting in TGC resistance have been previously reported (Hentschke et al., 2010; Ye et al., 2017). In XDX51, ramR insertion caused by ISKpn14 was identified through whole-genome sequencing. It was interrupted at 12 bp, affecting the translation of N-terminal amino acids. The RamR N-terminus acts as a DNA-binding site (Yamasaki et al., 2013). Without its DNA binding function, RamR is not able to inhibit RamA expression, and as a result, the efflux pump AcrAB-TolC is upregulated. As we speculated, qRT-PCR showed upregulated expression levels of ramA, acrA and acrB compared to that in the baseline strains. Moreover, efflux pump inhibition experiments with PAβN demonstrated the role of AcrAB-TolC in the increase in the TGC MIC in XDX51. Briefly, in our study, the dysregulation of RamR caused by IS element insertion was a major molecular mechanism for TGC resistance in XDX51.

In addition to antibiotic resistance-related genomic differences, LPS phenotype-related gene mutations occurred during within-host evolution. The 307 bp wbbL gene in XDX51 was interrupted by ISKpn26, which shares 99% amino acid similarity with IS5. WbbL is a rhamnosyl transferase that transfers L-Rha residues to the O4 position of D-Glc or D-GlcNAc to obtain a complete O-antigen (Erbing et al., 1977; Izquierdo et al., 2003). O-antigen typing showed that XDX16 and XDX51 belonged to OL101 and that wbbL was an important component of the OL101 locus (Follador et al., 2016). LPS analysis suggested that the O-antigen in the XDX51 and wbbL knockout strains was shorter than that in the wbbL wild-type strain. In addition, the TGC MIC for the reconstructed wbbL defective mutant was 2-fold higher than that for the strain with the complete wbbl sequence. Leth K et al. reported that wbbL mutations could cause serotype shifts and MIC changes for mecillinam in combination with bla _CTX-M mutations in E. coli. A similar study also reported that mutations in LPS-related genes could cause antibiotic resistance (Anton, 1995; Leth et al., 2019). In our study, the wbbL mutation had a slight effect on TGC susceptibility, with a twofold change in the MIC. Our result was consistent with Linkevicius’s report showing that defects in LPS-related genes (rfaC, rfaE, lpcA) could cause low levels of TGC resistance (Linkevicius et al., 2016).

In our study, the Insertion sequences played an important role in phenotypic changes in antibiotic resistance and genomic variation during within-host evolution, causing gene devitalization and duplication. Antibiotic resistance caused by ISs has been previously reported (Yang et al., 2020). Under long-term antibiotic stress, genomic variation caused by ISs could help the host overcome environmental challenges. Compared to XDX16 and XDX31, longer-term usage of the β-lactam-β-lactamase inhibitor and TGC led to the ISKpn14 insertion in wbbl, ISKpn26 insertion in ramR and replicative IS26 transposition with the bla _KPC-2 transposon structure. The ISs above were widely distributed in the genomes of the baseline strains, and their movement in the genome led to key antibiotic resistance. Furthermore, composite transposons with two identical ISs, such as IS26 on the bla _KPC-2-bearing plasmid in our study, can become mobilized and transferable among different strains. The dissemination of antibiotic resistance genes could cause great challenges to public health (Partridge et al., 2018). In the future, more clinical isolates would be studied to investigate the characteristic of CRKP that are prone to develop last line antibiotic resistance.

In brief, we tracked how a CRKP strain developed TGC and CZA resistance phenotypes during within-host antibiotic treatment. Three different genomic mutations were identified in XDX51, and all of them were caused by Insertion sequences. Double copies of bla _KPC-2 contributed to the CZA MIC changes due to enhanced ceftazidime hydrolysis. CZA resistance occurred without previous exposure to CZA. In addition, ramR inactivation led to TGC resistance. Furthermore, inactivation of wbbL was identified as being associated with LPS O-antigen deficiency and slightly reduced the susceptibility to TGC. The ST4496 clone needs attention, as it has the potential to evolve TGC and CZA resistance rapidly under TGC and β-lactam antibiotic exposure.

The datasets presented in this study can be found in online repositories. This Whole Genome Shotgun project has been deposited at GenBank under the accession JAIWPV000000000-JAIWPW000000000.

Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

XHa and QS conceived the idea and designed the experiments. XHa and YM performed the experiments. XHa analyzed the data. JQ, PZ, and PL helped with materials and reagents. XHa wrote the manuscript. YJ, DZ, XW, XHu, and YY reviewed the manuscript. All authors contributed to the article and approved the submitted version.

This work was supported by the China National Natural Science Foundation grant (81830069).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.