Genomic epidemiology and ceftazidime-avibactam high-level resistance mechanisms of Pseudomonas aeruginosa in China from 2010 to 2022

ABSTRACT

Ceftazidime-avibactam (CZA) resistance is a huge threat in the clinic; however, the underlying mechanism responsible for high-level CZA resistance in Pseudomonas aeruginosa (PA) isolates remains unknown. In this study, a total of 5,763 P. aeruginosa isolates were collected from 2010 to 2022 to investigate the ceftazidime-avibactam (CZA) high-level resistance mechanisms of Pseudomonas aeruginosa (PA) isolates in China. Fifty-six PER-producing isolates were identified, including 50 isolates carrying bla_PER-1 in PA, and 6 isolates carrying bla_PER-4. Of these, 82.1% (46/56) were classified as DTR-PA isolates, and 76.79% (43/56) were resistant to CZA. Importantly, bla_PER-1 and bla_PER-4 overexpression led to 16-fold and >1024-fold increases in the MICs of CZA, respectively. WGS revealed that the bla_PER-1 gene was located in two different transferable IncP-2-type plasmids and chromosomes, whereas bla_PER-4 was found only on chromosomes and was carried by a class 1 integron embedded in a Tn6485-like transposon. Overexpression of efflux pumps may be associated with high-level CZA resistance in bla_PER-1-positive strains. Kinetic parameter analysis revealed that PER-4 exhibited a similar kcat/Km with ceftazidime and a high (∼3359-fold) IC50 value with avibactam compared to PER-1. Our study found that overexpression of PER-1 combined with enhanced efflux pump expression and the low affinity of PER-4 for avibactam contributes to high-level resistance to CZA. Additionally, the Tn6485-like transposon plays a significant role in disseminating bla_PER. Urgent active surveillance is required to prevent the further spread of high-level CZA resistance in DTR-PA isolates.

KEYWORDS:

• DTR-PA
• Ceftazidime-avibactam resistance
• PER
• Efflux pump
• Tn6485

Introduction

Pseudomonas aeruginosa (PA) is an important nosocomial pathogen frequently found in health care settings that has become a major burden on health care due to its association with a variety of infections [1]. Over the past decade, the prevalence of PA isolates worldwide has escalated dramatically, primarily due to their resistance to numerous antibiotics. In fact, there has even been a rise in the emergence of “difficult-to-treat” resistant (DTR) strains [2], which pose a grave threat to human health [1–3]. DTR-PA evolved as a result of an interplay of multiple complex resistance mechanisms, including the production of β-lactamase enzymes, decreased expression of outer membrane proteins, overexpression of efflux pumps and mutations in penicillin-binding protein targets [4–6]. Among these mechanisms, the production of broad-spectrum β-lactamases is the most significant contributor to β-lactam resistance in DTR-PA isolates [6].

PER-like enzymes, one specific group of Ambler class A β-lactamases, confer resistance to penicillins, cephalosporins and aztreonam. The first detection of the PER-1 enzyme in PA isolates occurred in 1993 [7], and since then, fifteen additional PER-like variants have been reported. Among these PER-coding genes, bla_PER-1 is particularly prevalent in PA isolates. Furthermore, genes encoding PER-1 enzymes have been identified in a wide range of gram-negative bacteria, such as Enterobacteriaceae [8], Acinetobacter baumannii [9], Aeromonas spp. [10], and Vibrio cholerae [11]. In addition, the distribution of PER-producing PA was geographically different. In the ASPIRE-ICU study from 11 different countries in Europe, of the 402 PA isolates, 21.4% showed acquired β-lactamase, and 48.6% of them carried bla_PER [12]. Two Spanish nation-wide surveys performed in 2017 and 2022 reported that among the 342 XDR-PA, bla_PER was detected in only two strains isolated from 2017 and could no longer be detected in 2022 [13]. However, in a study conducted in Iran showed that among the 43 PA isolates that were resistant to at least one of the tested β-lactams, 35 (81.39%) isolates carried the bla_PER [14]. Currently, information on the prevalence of PA isolates carrying bla_PER is unknown in China.

In recent years, a novel antipseudomonal agent, ceftazidime-avibactam (CZA), has been introduced in clinics to combat classical β-lactam resistance mechanisms in PA isolates. It shows activity against Ambler class A β-lactamases (including ESBLs and KPCs), class C β-lactamases, and some class D β-lactamases [15]. While initially demonstrating significant clinical efficacy, numerous studies, including our own recently published research, have revealed an alarming rise in CZA resistance among PA isolates, even the emergence of high-level CZA-resistant PA isolates [16–19]. Notably, CZA resistance in PA isolates, even the emergence of high-level CZA-resistant PA isolates, primarily arises from mutations or the overexpression of class A β-lactam enzyme-encoding genes, alongside the presence of metallo-beta-lactamases (MBLs) in strains [15–19]. These factors commonly result in low levels of CZA resistance. However, the underlying mechanism responsible for high-level CZA resistance in PA isolates remains elusive.

To understand epidemiology and the high-level CZA resistance mechanism in PA isolates, 56 PER-producing PA isolates were collected from 67 hospitals. Antimicrobial susceptibility, whole-genome sequencing (WGS), cloning, RNA expression assays and biochemical assays were performed to decipher the molecular mechanisms responsible for the high-level CZA resistance phenotype.

Materials and methods

Bacterial isolates and antimicrobial susceptibility testing

Fifty-six PER-producing isolates were identified from 5763 PA isolates from 67 hospitals in 22 regions of China from 2010 to 2022. Isolate identification was performed using MALDI-TOF MS technology (bioMérieux, Marcy l'Etoile, France).

Antimicrobial susceptibility was determined by the broth microdilution method according to the Clinical and Laboratory Standards Institute (CLSI) guideline [20]. The MICs results were interpreted according to the CLSI M100 33th Edition, [20] except for colistin, which was interpreted according to the European Committee on Antimicrobial Susceptibility Testing (EUCAST) guidelines [21]. P. aeruginosa ATCC 27853 was used as the quality control strain.

Cloning experiments

Cloning experiments were carried out as previously described [16]. Plasmids carrying bla_PER-1 and bla_PER-4 were constructed and introduced into PA PAO1 isolate by electroporation, respectively. Briefly, the region including the wild promoter and open reading frame of bla_PER, derived from the PA1045 and PA2500 genome, were both cloned into the pGK-1900 vector. The recombinant plasmids were then transferred into PAO1 and selected on Mueller-Hinton agar plates containing 50 µg/mL gentamicin. The empty vector (pGK-1900 vector) was also introduced into PAO1 as a control. The MICs of the transformants were determined by the broth microdilution method. The primers are listed in Table S1.

Conjugation experiments

Conjugation experiments were carried out using a spontaneous rifampicin-resistant mutant of PAO1 as the recipient strain [16]. PA268, PA272, PA282, PA975, PA1017 and PA1018 isolates (different plasmid types with CZA resistance) were used as the donor strains. Transconjugants were selected on Mueller-Hinton agar plates which containing rifampicin (800 μg/mL) and carbenicillin (500 μg/mL). The presence of the bla_PER-1 gene in the transconjugants was identified by PCR and Sanger sequencing.

Quantitative real-time polymerase chain reaction (qRT-PCR)

The expression level of efflux pumps (MexAB-OprM, MexCD-OprJ, MexEF-OprN and MexXY-OprM) and PER-1 encoding genes were performed as previously described [19]. Total RNA was extracted from bacteria using an E.Z.N.A. total RNA kit I (Omega Bio-Tek, GA, USA). The relative expression of mexA, mexD, mexE, mexY and bla_PER-1 were determined by qRT-PCR using TB Green Premix Ex Taq II (TaKaRa Bio) in a LightCycler 480 system (Roche, Switzerland). Results were normalized to an endogenous reference gene (rpsL). The primers are listed in Table S1.

Efflux pump inhibition

Efflux pump inhibition was performed as previously described [16]. Briefly, the MICs of CZA were measured in the presence and absence of PAβN (Takara Bio Inc., Otsu, Shiga, Japan) at a concentration of 50 μg/mL [22]. An isolate was confirmed to over-express efflux pumps when the MICs in the absence of PAβN were determined at least 4-fold higher than the MICs in the presence of PaβN [22]. The wild-type PAO1 strain was used as the reference strain.

Whole-genome sequencing and bioinformatics analysis

Genomic DNA was sequenced using an Illumina NovaSeq (Illumina Inc., San Diego, U.S.A.) instrument generating 2 × 150 bp paired-end reads. The raw reads were assembled into draft genomes using the CLC Genomics Workbench v10.0.

Complete genome sequences of four PER-1-producing isolates (PA272, PA1045, PA2818, PA2209) and two PER-4-producing isolates (PA2500 and PA19-3047) were further obtained with an Oxford Nanopore MinION and Illumina platforms. Hybrid assembly was conducted using Unicycler version v0.4.8 [23]. The genome sequence was annotated using the RAST annotation website server (http://rast.nmpdr.org/rast/cgi) and Prokka 1.14.6 [24]. Resistance genes and multilocus sequence type (MLST) were obtained by using the ResFinder 4.1 and MLST 2.1 servers from the Center for Genomic Epidemiology (http://www.genomicepidemiology.org/).

The phylogenetic tree of PER-producing isolates based on single nucleotide polymorphisms (SNPs) was constructed with Snippy-multi (https://github.com/tseemann/snippy) and Fasttree. The tree was further visualized by iTOL (https://itol.embl.de/). BWA-MEM was used to align the assembled contigs of bla_PER-4-carrying isolates with the bla_PER-4-bearing fragment of PA2500 [25], and IGV_2.8.2 was used to visualize the alignments. The alignments of bla_PER-1-carrying strains against pPA272 were conducted by Proksee [26]. The genetic sequence comparison was performed and visualized by using Easy Figure 2.2.5 [27].

Cloning, expression, and purification of PER

Cloning, expression and purification of PER were performed [28]. The bla_PER-1 and bla_PER-4 genes were cloned into the pET-28a vector as an N-terminal His-tagged fusion protein, respectively. The products were then introduced into E. coli BL21 (DE3) competent cells for protein expression. His-tagged PER was purified using nickel affinity chromatography and gel filtration. Protein concentration was measured by absorbance at 280 nm using an extinction coefficient of 39,545 M-1 cm-1. The protein containing fractions were pooled and concentrated to 10-20 mg/mL and was stored at −80 °C.

Steady-state enzyme kinetic measurements

Kinetic parameters of purified enzymes PER-1 and PER-4 were determined using a spectrophotometer (Runqee, Shanghai, China) at room temperature in the phosphate-buffered saline (1×PBS, pH 7.2). The substrates include nitrocefin, ceftazidime and imipenem. The parameters adopted were listed in Table S2. The data were fitted to the Michaelis–Menten equation using GraphPad Prism v9.0.0 to obtain K_m and k_cat values [29].

For the inhibitor avibactam, tazobactam and clavulanic acid, the Ki and IC₅₀ values were determined in direct competition assays. Briefly, a fixed concentration of PER enzyme was made to mix with different concentrations of inhibitors (avibactam requires a 5-min incubation at room temperature). Then 10 μM of nitrocefin was added and the initial reaction velocities (v) was determined. Ki values were obtained by fitting the data to equation [29], and IC₅₀ values were calculated using the “dose–response-inhibition, variable slope (four parameters)” equation.

Structure modelling and molecular docking

Retrieving the RCSB Protein Data Bank (PDB), the crystal structures of β-lactamases PER-1 (PDB entry: 1E25) was applied as the structure template in the present study. After PSI-BLAST [30] and MUSCLE alignment [31], the homology models of PER-1 and PER-4 were both built in Schrödinger. The protein structure was refined and preprocessed using the OPLS3e force field [32], following the ligands preparing for diverse ionization states and isomers. In order to make certain of no steric clashes or other deviations, we further checked the protein systems after all preparation. The residue number followed the Ambler numbering scheme.

Taking Ser70 as the centroid of grid box and H-bond constraint atoms, induced fit docking [33] was used to identify and localize the accurate ligand binding pocket. According to the high-resolution structure analysis of the previous literature [34], correct docking poses where the aromatic carbonyl group of ligands forms hydrogen bonds with Ser70 and Thr237 were selected. The remaining poses were sorted and the top one acted as the initial structure of covalent docking [35]. In the covalent docking, the ligands bound in the selected poses of the last step were set as the centroid of grid box, and the β-lactam and self-defined diazabicyclooctane were defined as two reaction types for pose prediction. The refinement minimization radius was set as 5.0A. After completion of docking, the Prime MM-GBSA scores of the top one output poses for each ligand was calculated for comparing the binding affinities of the ligands [36].

Results

Antimicrobial susceptibility of PER-producing isolates

Fifty-six PER-producing isolates (0.97%, 56/5763) were identified, including 89.3% (n = 50) PER-1-producing isolates and 10.7% (n = 6) PER-4-producing isolates. Antimicrobial susceptibility testing showed that all of these PA isolates were resistant to ceftazidime (100%), cefepime (100%) and aztreonam (100%), but susceptible to colistin (Table 1). Most of them were resistant to imipenem (92.86%), meropenem (91.07%), amikacin (94.64%), ciprofloxacin (89.29%), piperacillin (98.21%), piperacillin/tazobactam (91.07%), levofloxacin (82.14%) and CZA (76.79%) (Table 1). Among all the PER-producing isolates, 82.1% (46/56) were classified as DTR-PA. In addition, the CZA MIC ranged between 4 and 2048 μg/mL in 56 PER-producing PA isolates. The MIC₅₀ and MIC₉₀ of these isolates were 16 and 512 μg/mL, respectively. Among these PA isolates, 74% (37/50) of PER-1- and 100% (6/6) of PER-4-producing isolates were resistant to CZA. Notably, 14.3% (n = 8) of isolates had high CZA MICs (>512 μg/mL) (Table S3).

Table 1. In vitro antimicrobial susceptibility tests of PER-producing 56 PA strains.

Antibiotics	MIC (mg/liter)			% resistance	CLSI cutoff (mg/liter)
	MIC5O	MIC9O	MIC range
PIP	>128	>128	32 to >128	98.21	64
CAZ	>128	>128	>128	100.00	32
FEP	>128	>128	32 to >128	100.00	32
PTZ	>128	>128	16 to >128	91.07	64/4
CZA	16	512	4–2048	76.79	16/4
IPM	16	32	0.5–64	92.86	8
MEM	16	64	2 to >128	91.07	8
AZT	>128	>128	64 to >128	100.00	32
AMK	>128	>128	2 to >128	94.64	64
CIP	4	16	0.25–16	89.29	2
LEV	16	32	0.5–64	82.14	4
CST	2	2	0.25–2	0.00	4

PIP, piperacillin; CAZ, ceftazidime; FEP, cefepime; PTZ, piperacillin/tazobactam; CZA, ceftazidime-avibactam; IPM, imipenem; MEM, meropenem; AZT, aztreonam; AMK, amikacin; CIP, ciprofloxacin; LEV,levofloxacin; CST, colistin.

Molecular epidemiology

A core-genome phylogenetic tree was constructed, and the data showed that the 56 isolates were grouped into 19 distinct sequence types (STs) (Figure 1). The top three STs were ST244, ST298 and ST235, which are among the top 10 high-risk clones worldwide [37]. The MIC distributions of CZA varied for different STs (Figure 1). Notably, all PER-4-producing isolates belonged to the ST235 clonal lineage (Figure 1).

Figure 1. The phylogenetic tree of bla_PER-carrying P. aeruginosa isolates. The rings visible on the exterior of the tree indicate sequence types, bla_PER genes, MICs of ceftazidime/avibactam and the contexts of bla_PER genes.

Overexpression of bla_PER-1 and mutant bla_PER-4 are responsible for the CZA resistance phenotype

To further demonstrate that the bla_PER-1 and bla_PER-4 genes are involved in CZA resistance, they were both cloned and expressed in PAO1. Antimicrobial susceptibility results of CZA showed that the MIC of the wild-type PAO1 strain and the PAO1 control strain carrying the pGK-1900 empty vector were 1 and 2 μg/mL of CZA, respectively (Table 2). In contrast, the MICs of the PAO1/pGK-1900-PER-1 strain and the PAO1/pGK-1900-PER-4 strain, carrying the bla_PER-1 gene from PA1045 and the bla_PER-4 gene from PA2500, were 32 μg/mL and >2048 μg/mL CZA (Table 2), respectively. Notably, PER-4 led to a CZA MIC value 64-fold higher than that of PER-1.

Table 2. Antimicrobial susceptibility of P. aeruginosa clinical isolates, transconjugants, and recombinant PAO1 strains carrying bla_PER.

	MICs (mg/L)
Isolates	PIP	MEM	IPM	FEP	CAZ	CFDC	AZT	PTZ	IR	CZA
Clinical isolates
PA272	>128	32	16	>128	>128	2	>128	>128	4	1024
PA282	>128	32	16	>128	>128	2	>128	>128	4	1024
PA268	>128	16	16	>128	>128	2	>128	>128	4	1024
PA975	128	16	16	128	>128	1	>128	64	2	16
PA1017	>128	16	32	>128	>128	2	>128	>128	4	128
PA1018	>128	32	32	128	>128	1	>128	>128	8	128
Transconjugants
PAO1-R^rif	4	0.5	1	2	1	0.5	2	4	0.25	1
PAO1/pPA272	>128	1	2	128	>128	2	>128	128	0.5	4
PAO1/pPA282	>128	1	1	128	>128	2	>128	128	0.5	8
PAO1/pPA268	>128	1	1	128	>128	2	>128	128	0.5	4
PAO1/pPA975	128	2	1	64	128	2	>128	64	0.25	8
PAO1/pPA1017	128	1	1	128	>128	2	>128	64	0.25	16
PAO1/pPA1018	>128	32	32	128	>128	1	>128	128	0.25	32
Recombinant PAO1 strains
PAO1	4	0.5	1	1	1	1	4	4	0.25	1
PAO1-pGK1900	4	1	1	4	2	0.5	8	4	0.125	2
PAO1-pGK1900-PER4	16	1	1	>128	>128	2	>128	16	0.5	>2048
PAO1-pGK1900-PER1	32	1	1	>128	>128	2	>128	16	0.5	32

PIP, piperacillin; MEM, meropenem; IPM, imipenem; FEP, cefepime; CAZ, ceftazidime; CFDC, Cefiderocol; AZT, aztreonam; PTZ, piperacillin-tazobactam; IR, imipenem-ralebactam; CZA, ceftazidime-avibactam

To further confirm CZA resistance caused by overexpression of bla_PER-1, we determined both the relative expression levels and copy number of the bla_PER-1 gene in these strains. We selected the 3 CZA resistance isolates (CZA-R strains: PA272, PA256, PA282) with the highest CZA MICs (MIC = 1024 μg/mL) and randomly selected 3 CZA susceptible isolates (CZA-S strains: PA0616, PA2217, PA1814; MIC = 4 μg/mL) (Table S3) for comparison of bla_PER-1 expression. The results showed that bla_PER-1 expression was significantly increased in CZA high-level resistance isolates compared with CZA susceptible isolates (Figure 2). The effect of bla_PER-1 copy number was excluded because there were no significant changes among these PER-1-producing PA isolates (Figure S1). We also performed RT-qPCR on another 31 PA strains of different ST types to determine the expression level of bla_PER-1 and the results also indicate that there was a positive correlation between the expression level of bla_PER-1 and the MIC of CZA in PA strains (Figure S2).

Figure 2. Relative ratios of the expression of efflux pumps and bla_PER-1. Relative ratio of efflux pump and bla_PER-1 expression in the CZA-resistant and CZA-sensitive isolates, in which the expression levels of the efflux pump are indicated as a ratio to the expression level in strain PAO-1 (expression = 1), and the expression level of bla_PER-1 is indicated as a ratio to the expression level in strain PA1814 (expression = 1). Gene expression was normalized to that of the rpsL housekeeping gene. Statistical analysis was performed through unpaired t tests. *, P < 0.05. CZA, ceftazidime avibactam. CZA-R, CZA-resistance group; CZA-R strains: PA272, PA256, PA282. CZA-S, CZA-sensitive group; CZA-S strains: PA0616, PA2217, PA1814.

Genetic contexts of bla_PER-1

Genetic environment analysis showed that 8 isolates had the bla_PER-1 gene located in the chromosome, while in the remaining 42 isolates, the gene was in two different types of IncP-2 plasmids (Figure 1). Mating experiments showed that two PER-1-producing isolates possessing different plasmid types could transfer their resistance phenotype to the recipient strain (Table 2). Of the two plasmid types, the type 1 plasmids (n = 4) were exclusively carried by the ST132 clone. Type 2 plasmids (n = 38) were most common and found in various ST strains. To gain further insights, representatives (PA272 and PA1045) from each plasmid pattern were selected for Nanopore long-read sequencing. Analysis revealed that the type 1 IncP-2 plasmid, named pPA272, was a 517,764 bp megaplasmid with a backbone similar to the IncP-2 plasmids reported previously [38]. Additionally, the type 2 IncP-2 plasmid, pPA1045, was 432,097 bp in size and shared a highly similar backbone (80% coverage and 99.97% identity) to pPA272 (Figure 3).

Figure 3. A: Comparative analysis of the PER-1-producing plasmid pPA272 and other plasmids. The rings of PA256 and PA667 were aligned with pPA272 by Illumina assembled sequences. The rings of PA1045 were aligned with Nanopore assembled sequences against pPA272. The outermost circle shows the open reading frames of pPA272. B: Comparison of the genetic contexts of bla_PER-1 in pPA272, pPA1045, PA2818 and PA2209. Arrows indicate open reading frames, with arrowheads indicating the direction of transcription as follows: red, antibiotic resistance-encoding genes; green, mobile elements; other genes are shown by yellow arrows. Abbreviation: IR, inverted repeat.

The genetic analysis of bla_PER-1 in both chromosome-borne and plasmid-borne forms revealed its association with three transposable elements (Figure 3). The primary location of bla_PER-1 was within a multidrug-resistant (MDR) composite transposon referred to as Tn6485-like. This transposon was bounded by left and right inverted repeats (IRL/IRR: CGGAAAAAATCGTACGCTAA/TTAGCGTACGATTTTTTCCG). In the case of PA2818, the downstream backbone structure (uspA-dksA-yjiK) was lost. Notably, bioinformatics analysis showed a Tn6485-like transposon located in the chromosome of the ST244 clone inserted between paaL and lpdA compared with the chromosome of another ST244 PA strain, KB-PA_FI9 (GenBank accession no. CP086010) (Figure S3(A)). The majority of the Tn6485-like transposons in PA strains contained a class 1 integron flanked by IRi (TGTCGTTTTCAGAAGACG) and IRt (CGTCTTCTGAAAATGACA). However, in the type 1 IncP-2 plasmid, the IRt was deleted along with the tnpR-TnAs1-sulP segment. Additionally, an inversion event occurred in PA2818, resulting in the change of IRt to its complementary sequence (IRt’). All the class 1 integron derivatives in Tn6485-like transposons contained the core ISCR1-bla_PER-1 unit that was organized as a 5’-conserved segment (5’-CS) (ISCR1), variable region (VR) (bla_PER-1-orf-msbA-orf-orf-orf), and 3’-CS (qacEΔ1/sul1). Notably, only the ISCR1 element of the type 1 IncP-2 plasmid, such as pPA272, showed a different VR, where the insertion of ISPst8 was found between the ISCR1 and bla_PER-1 genes (Figure 3).

Genetic contexts of bla_PER-4

To investigate the genetic context of bla_PER-4, PA19-3047 and PA2500 were sequenced by a hybrid assembly strategy combining Illumina and MinION. The alignments of Illumina and Nanopore assembled sequences showed the bla_PER-4 gene, which in all 6 ST235 PER-4-producing PA isolates had basically the same genetic contexts (Figure 4). The bla_PER-4 gene was present in a basic ISCR1-bla_PER-4 element (ISCR1-bla_PER-4-orf-msbA-orf-orf-orf-sul1) within the chromosome of all PER-4-producing PA isolates. The upstream region of ISCR1 corresponds to a classical class 1 integron (integron A), with the intI1 gene followed by a cassette array containing aadA6, ermE and sul1. Both the ISCR1-bla_PER-4 element and integron A were integrated into a large class 1 integron (integron B) flanked by inverted repeats (IRi and IRt), where the 5'-CS was interrupted by ISPa7. Moreover, there were three copies of bla_PER-4 embedded in two class 1 inverted integrons in PA19-3047. One integron showed almost identical characteristics to integron B in PA2500. The other integron featured two adjacent direct repeats of the ISCR1-bla_PER-4 element, with the IRi being disrupted by dadA (Figure 4). This indication can also be supplemented by another result; compared with the ST235 PA strain M27432 (GenBank accession no. CP101885), the ISCR1-bla_PER-4 element of PA2500 may move into the 3’-CS (qacEΔ1/sul1) of a class 1 integron by rolling-circle transposition, or integron B of PA2500 may replace the above class 1 integron by homologous recombination, resulting in the transmission of bla_PER-4 (Figure S3(B)). Furthermore, when comparing PA2500 with R12-04, the only ST235 PA strain not carrying bla_PER-4, it was discovered that bla_PER-4 was also located within an ∼22 kb MDR transposon, which has a downstream Tn6485-like backbone structure (uspA-dksA-yjiK) and is flanked by IRL with one base mutation (CGGAAAAAATCGTACGTTAA) and IRR. Furthermore, this transposon can transmit between ST235 PA strains, as evidenced by its insertion in the downstream region of a 1,206-bp hypothetical protein-encoding gene in R12-04 (Figure S3(B)).

Figure 4. Comparison of the genetic contexts of bla_PER-4. The grey bars in boxes represent the aligned contigs of strains against the bla_PER-4-carrying fragment of PA2500. Schematic illustration comparing the structural features of the genetic context of bla_PER-1 in PA2500 with PA19-3047. Arrows indicate open reading frames, with arrowheads indicating the direction of transcription as follows: red, antibiotic resistance-encoding genes; green, mobile elements; other genes are shown by yellow arrows. Abbreviation: IR, inverted repeat.

The overexpression of efflux pumps is associated with CZA resistance

Three PER-1-producing PA isolates showed high levels of resistance to CZA (MIC > 512 μg/mL), but their transformants showed lower levels of resistance to CZA (MIC, 4–32 μg/mL), indicating that mechanisms other than overexpression of bla_PER-1 are responsible for CZA resistance. Thus, the relative expression levels of efflux pumps (MexAB-OprM, MexCD-OprJ, MexEF-OprN and MexXY-OprM) were determined in CZA-resistant and CZA-susceptible isolates. Our results showed that mexD expression was not significantly increased, but mexA, mexE and mexY expression was significantly increased in the CZA high-level resistance isolates (Figure 2). In addition, the relative expression levels of efflux pumps in PER-4-producing PA isolates were not determined because the transformants of PER-4-producing PA isolates displayed high levels of resistance to CZA (Table 2).

This change in efflux pump expression was also demonstrated using efflux pump inhibition assays. The MICs of CZA of the CZA-resistant isolates were decreased by 32∼64-fold in the presence of PaβN compared with the absence of PaβN (Table S4). These results confirmed that overexpression of the MexAB-OprM, MexEF-OprN and MexXY-OprM efflux pumps contributes to the CZA high-level resistance phenotype in PER-1-producing isolates.

Enzyme kinetic data

To understand the CZA resistance mechanism, the enzyme kinetics of PER-1 and PER-4 were measured. The k_cat/K_m of the wild-type PER-1 enzyme with ceftazidime was similar to that of PER-4. The wild-type PER-1 with ceftazidime exhibited a lower K_m than PER-4 (Table 3). In addition, the wild-type PER-1 enzyme displayed lower nitrocefin hydrolysis activity (∼1.7-fold) than PER-4 (Table 3). Notably, the two enzymes showed hardly any detectable imipenem hydrolysis activity (Table 3).

Table 3. Kinetic parameters of purified β-lactamases PER-1 and PER-4.

β-Lactam(s)			PER-1		PER-4
	Km (µM)	kcat (s-1)	kcat/Km (μM-1.s-1)	Ki (μM)	Km (µM)	kcat (s-1)	kcat/Km (μM-1.s-1)	Ki (μM)
Nitrocefin	72.92	511.58	7.01 ± 0.157	ND	18.48	224.03	12.178 ± 0.311	ND
Ceftazidime	668.75	537.44	0.807 ± 0.014	ND	726.44	607.66	0.839 ± 0.008	ND
Imipenem	ND	ND	ND	ND	ND	ND	ND	ND
Avibactam	ND	ND	ND	0.191 ± 0.004	ND	ND	ND	473 ± 29.19
Tazobactam	ND	ND	ND	0.32 ± 0.0131	ND	ND	ND	707.23 ± 97.7
Clavulanic acid	ND	ND	ND	0.024 ± 0.0004	ND	ND	ND	0.099 ± 0.0006
Inhibitor					IC50 (μM)
			PER-1				PER-4
Avibactam			0.217				728.9
Tazobactam			0.363				1089.8
Clavulanic acid			0.027				0.152

ND: Not determinatable due to a low initial rate of hydrolysis

Km, kcat, and Ki values are shown as the means ± standard deviation from three independent experiments.

IC₅₀ represents the concentration of a drug that is required for 50% inhibition of the enzymatic activity.

To evaluate the inhibitory activity of the inhibitor against the PER-1 and PER-4 enzymes, the IC₅₀ values of avibactam, tazobactam and clavulanic acid were further determined. Avibactam exhibited a ∼3359-fold higher IC₅₀ value against PER-4 than against wild-type PER-1 (Table 3). Likewise, tazobactam exhibited an ∼3002-fold higher IC₅₀ value against PER-4 than against wild-type PER-1 (Table 3). Clavulanic acid exhibited ∼5.6-fold higher IC₅₀ values against PER-4 compared to PER-1 (Table 3).

Structural comparison of the ligand binding ability of PER-1 and PER-4

To study the binding modes of beta-lactams and inhibitors to PER-1 and PER-4, we prepared homology structures and conducted covalent docking with MM/GBSA scoring. All the scoring results is listed in Table S5. Among them, more negative binding free energy (ΔG_MM/GBSA) reveals more tight binding, more negative covalent docking affinity (cdock) implies more easy forward reaction, while positive cdock tilts to reverse reaction or even unattainable reaction. It is clear that the ranks of the binding scores especially cdock accord with the enzyme kinetic data. The ligand poses of the docking results, as shown in Figure S4(A and B), were in accordance with the previous high-resolution PDB structures. Ser70 is the reaction centre acting as a nucleophilic reagent for acylation (Figure S4(C and D)). We found that several amino acid residues, such as Ser130 (or Thr130 in PER-4), Asn132, Glu166, Thr235, Thr237, and Arg220, always tend to form hydrogen bonds with ligands and contribute to substrate recognition. In addition, Trp105 and Arg220 sometimes develop π interactions or salt bridges with ligands. The inhibitors bind less tightly than β-lactams due to their small volume and fewer interactions with enzymes. Among them, Thr235 is a part of the KTG motif, which could stabilize the oxyanion of the tetrahedral complex and activate the SXN triad motif [39–41]. Glu166 could activate catalytic water as a general base as a part of the EXXLN motif [39–41]. As another residue of the SXN motif, Ser130 is the proton donor that opens the β-lactam rings upon acylation and often develops hydrogen bonds with the carboxyl group of the β-lactam rings [39–41]. We also observed that Thr130 rotates away from ligands in PER-4 when compared to PER-1. The Ser120(S130T) of PER-4 could strengthen acidity and weaken the protonation ability of the β-lactam nitrogen leaving group, impairing the carbamylation and decarbamylation capability of the reaction intermediate. Hence, the covalent reaction is somehow obstructed, as shown by the positive docking affinity of avibactam and tazobactam in Figure S4(A–D). These effects are relatively obvious with the inhibitors and relatively obscure upon cephalosporin hydrolysis. Structure analysis showed that PER4 exhibits slightly weaker hydrolysis of cephalosporins and monobactams and shows less susceptibility to inhibition by clavulanic acid than PER-1. However, it shows strong resistance to inhibition by avibactam and tazobactam.

Fitness effects of bla_PER

To investigate the influence of the bla_PER gene on growth, the growth curves of the PAO1, PAO1/pGK-1900, PAO1/pGK-1900-PER-1 and PAO1/pGK-1900-PER-4 strains in the absence of antibiotics were measured. The results showed that the growth curves of these isolates exhibited no significant difference (Figure S5(A)). To further evaluate the effect of the bla_PER gene, two other bacterial growth parameters (the area under the growth curve (AUC) and the relative growth rate) of these constructed strains were compared in the absence of antibiotics. The relative growth rates of the PAO1/pGK-1900-PER-1 and PAO1/pGK-1900-PER-4 strains showed no significant difference compared to that of the vector-carrying strain PAO1/pGK-1900 (Figure S5(B)). However, the relative growth rate of the PAO1/pGK-1900-PER-4 isolate was lower than that of PAO1/pGK-1900-PER-1 by approximately 2.1% (1.0007 ± 0.0154 vs. 0.9796 ± 0.01, P < 0.05) (Figure S5(B)). The AUC results also showed that there was no significant difference among these isolates (Figure S5(C)). Overall, our results showed that the bla_PER gene has no negative impact on fitness, but the bla_PER-4 gene has a negative impact on fitness compared with bla_PER-1, mainly inhibition of the bacterial growth rate.

Discussion

Numerous studies have demonstrated the substantial efficacy of CZA against a wide range of PA isolates, including DTR-PA [42], indicating its effectiveness. However, it is worth noting that CZA-resistant strains of PA continue to emerge. In PA strains, resistance to CZA is primarily associated with β-lactam enzyme mutations, structural modifications, and increased efflux pump expression, aside from MBL-positive strains [43]. Notably, β-lactam enzyme mutations have become the primary contributing factor to CZA resistance in PA. Several reports, including two from our group, have shown that different β-lactam enzyme mutant genes, such as bla_OXA-10 [44], bla_GES-15 [45], bla_KPC-33[18] and bla_KPC-90[16], confer resistance to CZA in PA. Additionally, the overexpression of β-lactam enzymes such as KPC-2, AmpC, PER-1 and GES-1, has been associated with reduced susceptibility to CZA in PA [15,17,19,46,47]. It is important to note that previous studies have shown that these factors contribute to a low level of CZA resistance. However, in our study, we found that bla_PER-4 played a significant role in high-level CZA resistance due to its extremely low affinity to avibactam. This discovery is the first demonstration of bla_PER-4 involvement in high-level CZA resistance.

In contrast to the single mechanism of CZA resistance in Enterobacteriaceae, PA simultaneously uses multiple mechanisms and their interplay, such as ESBL carriage and overexpression of RND efflux pumps (such as MexAB-OprM), to achieve CZA resistance [48]. As the main efflux transporter of avibactam in PA isolates, over-expression of MexAB-OprM or mutations in nalD (coding the MexAB-oprM regulator) could lead to decreased susceptibility to CZA as previously reported [49,50]. In addition, it is worth noting that different PA strains may employ different efflux systems to achieve the same goal. We previously found that overexpression of KPC-90 and MexXY-OprM efflux pumps [16] as well as overexpression of the efflux pumps MexAB-OprM and GES-1 both conferred in directing CZA resistance [19]. Our results showed that MexAB-OprM, MexEF-OprN and MexXY-OprM expression were significantly increased in the CZA high-level resistance isolates. Thus, the overexpression of the multiple efflux pumps combined with PER-1 contributes to the high-level CZA resistance phenotype in these DTR-PA isolates, which is no doubt that effective treatment options for DTR-PA strains will be further reduced.

There is only one different amino acid (Ser120(S130T) between PER-1 and PER-4. However, there was a significant difference in their mediation of the CZA resistance phenotype. In previous studies, Ser130 was identified as a key catalytic residue for avibactam inhibition, which provides protons for the sulfate nitrogen (N6) of avibactam to achieve acylation and may act as a hydrogen acceptor to initiate avibactam recyclization [51]. The mutation to threonine will cause the sulfate nitrogen to leave, thereby impairing the proton donating ability of enzymes. Finally, the kinetic parameters of avibactam and tazobactam will increase significantly compared to those of other ligands. This effect not only exists with CTX-M type enzymes but also with SHV and KPC enzymes, as validated by steady-state kinetic experiments [52]. Previous studies have highlighted the S130G variant because this mutation completely abolishes the acylation and deacylation of avibactam [51]. However, this mutation also considerably affects the hydrolysis of cephalosporins, including ceftazidime; Thus, the susceptibility to CZA would not change much. In addition to the S130G variant, the S130C (GES-10) substitution is a conservative alternative of the nucleophilic catalyst and does not impair inhibitor kinetics. Unlike the first two mutations, the S130 T (TEM-211 and PER-4) substitution impairs the inhibitor kinetics but does not significantly affect the hydrolysis of cephalosporins. As a result, it causes high-level resistance to cephalosporin/β-lactamase inhibitor combinations. In addition to CZA and PTZ found in our study, Enterobacterales carrying PER-4 was previously found to exhibit high-level resistance to aztreonam-avibactam (ATM-AVI) [53]. Considering the structural similarity, PER-4-producing strains may also show high-level resistance to durlobactam; verification of this fact will require further study to clarify the treatment scope.

Currently, PA isolate treatment is difficult due to multiple resistance genes, especially for genes encoding β-lactamases (class A and B enzymes) and their variants. Over the past decade, there was still a low prevalence (0.97%, 56/5763) of these strains in China, but most of them were classified as DTR-PA. Moreover, these strains had various STs, especially international high-risk clones, which were involved in poor clinical outcomes due to multi and extensive resistance [35,36]. There is no doubt that these strains are a concern for public health due to their extensive drug resistance and high virulence.

The expression of acquired resistance genes, along with transposable element dissemination and intrinsic resistance mechanisms, considerably reduces the therapeutic options for treating infections caused by PA. bla_PER can be transferred by mobile genetic elements, including plasmids, transposons, and insertion sequences (ISs) [38]. The most striking finding in our study is that bla_PER is carried in a transposable element, called ISCR1-bla_PER, and has become part of a complex class 1 integron. The presence of transposons and variations in integrons highlights the flexibility and adaptability of bla_PER-1 in its genetic context. ISCR1 element played a major role for the host bacteria in the mobilization and accumulation of antibiotic resistance genes (ARGs), such as bla_PER, bla_NDM and bla_OXA-1041 [38,54,55]. This transposable element has been identified in a wide range of bacteria, including Pseudomonas, Acinetobacter, Vibrio, Aeromonas, Proteus, Providencia and Morganella [38]. ISCR1 element can transfer ARGs via the rolling-circle transposition and may result in tandem copies of the ISCR1-ARG by homologous recombination [54,55]. In our study, ISCR1 was located both upstream of bla_PER-1 and bla_PER-4, and tandem copies of the ISCR1-bla_PER-4 elements were observed. Furthermore, the ISCR1-bla_PER-1 element was identified in both plasmids and chromosomes of DTR-PA isolates, indicating that these DTR-PA isolates have the capacity to acquire resistance elements both vertically and horizontally by ISCR1-mediated transposition. In order to prevent further dissemination of ISCR1-bla_PER-1, rational application and effective supervision of antibiotics should be performed to decrease antibiotic pressure, such as β-lactam and CZA. In addition, it is more important to strengthen infection prevention controls (IPCs) for the staff, patients and environment to essentially reduce the incidence of PA infection.

In addition, bla_PER-1 is carried in the IncP-2 megaplasmid, which was recognized as one of the Pseudomonas antimicrobial resistant megaplasmids [56]. Recently, multiple studies have reported that this type of plasmid accumulated diverse resistance genes, including bla_AFM-1 and bla_IMP-45 [16,57]. Expanding gene content is a way to make larger plasmids inherit and spread more efficiently [56]. In the present study, bla_PER-1 was carried in an IncP-2 megaplasmid of an international high-risk clone, and its overexpression further contributed to CZA resistance. Furthermore, our results showed that the bla_PER gene has no negative impact on fitness. Research has found that cost-free resistance in bacteria can influence the dynamics of bacterial spread through populations, sometimes allowing resistance mechanisms to become established in bacterial populations [58]. Thus, the absence of fitness costs may cause a greater survival advantage for such strains in the environment as well as the widespread dissemination of resistance genes, and clinical use of drugs might induce the production of additional mutants, resulting in the emergence of more highly resistant clinical strains. All these findings indicate that these factors will undoubtedly lead to the further spread of PER-producing DTR-PA strains due to the development of CZA resistance, which brings challenges to clinical anti-infection treatment for this type of DTR-PA strain. It is imperative to perform active surveillance to prevent the further dissemination of CZA-resistant DTR-PA isolates.

Conclusion

Our study shows that the overexpression of bla_PER-1 combined with enhanced efflux pump activity, and the low affinity of bla_PER-4 to avibactam, contributes to high-level CZA resistance in clinical isolates of DTR-PA. Moreover, bla_PER is carried by the transposable element ISCR1-bla_PER, which could result in the further spread of high-level CZA resistance. Active surveillance is urgently needed to prevent further dissemination of high-level CZA resistance in DTR-PA isolates.

Author contributors

XL: data curation, formal analysis, visualization, writing – original draft. LZ: sample collection, sample processing, data curation, formal analysis, writing – original draft. TL and XZ: data curation, formal analysis, visualization, writing – original draft, writing – review and editing. JY: supervision, formal analysis, project administration. JH: formal analysis. HL: formal analysis. HC: formal analysis. JJ: samples collection. YZ: samples collection. YT: funding acquisition, conceptualization, writing – review and editing. YY: funding acquisition, conceptualization, writing – review and editing. HZ: resources, funding acquisition, conceptualization, writing – review and editing.

Consent and research ethics

This study was conducted in accordance with the Declaration of Helsinki and had been reviewed and approved by the Research Ethics Committee of Zhejiang Provincial People's Hospital (Approval no.: 2023-284).

Acknowledgement

We thank Professor Dazhi Jin (Hangzhou Medical College) for his help with revising the manuscript.

Data availability

Sequences mentioned in the present study were submitted to the GenBank nucleotide database with accession numbers CP129995, CP129997, CP129685, CP129686, CP129406, CP129405, CP129688, CP129687, CP068239, JAUKHA000000000-JAUKHV000000000, JAUKGA000000000-JAUKGZ000000000, JAUKFZ000000000 and JAPZMG000000000.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This study was supported by the National Natural Science Foundation of China (82172306 and 82272338), the Public Technology Research Projects of Zhejiang Province, China (LGD21H190001), the Research and Development Program of Zhejiang Province (2023C03068) and the National Health Commission Scientific Research Fund-Zhejiang Provincial Major Health Science and Technology Plan Project (WKJ-ZJ-2414).