Enhanced Spread of Carbapenem-Resistant Pseudomonas aeruginosa in ICU Environment During a COVID-19 Upsurge Period in China

Introduction

Pseudomonas aeruginosa is one of the most important nosocomial pathogens and is widespread in both community and hospital settings. The World Health Organization (WHO) has listed carbapenem-resistant Pseudomonas aeruginosa (CRPA) as a high-priority pathogen that urgently requires antibiotic development.¹ China Antimicrobial Surveillance Network (CHINET) reported the resistance rates of P. aeruginosa to imipenem (IPM) and meropenem (MEM) were 21.9% and 17.4%, respectively.² Multiple factors contribute to carbapenem resistance in P. aeruginosa, including the acquisition of exogenous carbapenemases (eg, bla_KPC, bla_VIM, bla_GES), membrane porin modification, efflux pump overexpression, and endogenous carbapenemase expression. Carbapenemase mutation, for example, the conversion from bla_KPC-2 to bla_KPC-33 or bla_KPC-71, was reported in clinic as one important mechanism of bacterial resistance acquisition.^3,4 Meanwhile, multiple virulence factors including type VI effectors, type III effectors, and adherence factors have been shown to facilitate P. aeruginosa infection and contribute to increased mortality in clinical settings.^5,6 In southeast China, sequence type (ST) 463 P. aeruginosa has been described as a representative high-risk clone with high resistance and virulence, causing a significant burden on public health.^7,8

Patients in hospital wards, especially intensive care units (ICU), are susceptible to P. aeruginosa infection originating from other patients or hospital environments. Environmental elements such as hospital water systems, ventilators, electrocardiograph monitors, and healthcare workers provide transmission routes via cross-contamination.⁹ Transmission may occur as either an outbreak or a sporadic event. Tracking the transmission process using whole-genome sequencing (WGS) over a specific time and location will improve hospital infection management.¹⁰

The COVID-19 pandemic emerged as a public health crisis in 2019, and the impact of this unique context on the epidemiology of P. aeruginosa remained unclear. Large-scale retrospective studies using diverse settings and sampling methods have revealed that approximately 60% of COVID-19 patients with bacterial coinfection exhibit antimicrobial resistance (AMR).¹¹ Recorded CRPA isolation rate through COVID-19 pandemic varied in different medical centers, and the discrepancy remains inadequately explained by antibiotic stewardship practices or public health policy.^12–16 Changes in nosocomial transmission dynamics can significantly influence CRPA isolation rates.¹⁷ However, limited research has focused on the precise transmission route of P. aeruginosa using comprehensive and continuous sampling methods before and during a public health crisis, given its inherent unpredictability.¹⁰

Here, we conducted a prospective observational study of P. aeruginosa in a 21-bed ICU at a tertiary teaching hospital in Hangzhou, China during a COVID-19 upsurge period. CRPA isolates from patients and the hospital environment were screened actively and periodically, and then subjected to phenotypic and genotypic analyses. This study aimed to describe the transmission and genetic characteristics of CRPA, focusing on the prevalence of ST463 strains and their corresponding mobile carbapenemases (bla_KPC-2, bla_KPC-33, and bla_KPC-71). We also explored the evolution process from bla_KPC-2 to bla_KPC-71 within one patient.

Materials and Methods

Study Design

This prospective study was conducted over a span of 6 months, from October 21, 2022, to April 27, 2023, in a newly established ICU at a tertiary teaching hospital. The sampling paused from December 20, 2022, to January 5, 2023 due to COVID-19 upsurge. ICU received its first patient on October 22, 2022 (one day after the start of screening).

Patients and Samples

Patients were assigned numbers based on their orders in our research registration, and those with a stay length of less than two days were excluded from the study. For example, Patient No. 001 denotes the first patient recorded in our patient’s data list. Rectal and nasopharyngeal swab screening was performed on patients within one day of ICU admission and discharge. Rectal and nasopharyngeal swabs from all ICU patients were screened at fixed intervals (14 days). Samples from the same patient in the same clinical site of infection within 7 days were considered repeated samples and were not included in future analyses. Clinical samples were included in the study if needed. Informed consent was obtained from all related patients before study commencement.

Environmental Surveillance

The ICU has 19 rooms, and each room has a single patient bed, except Room 10, which has three beds. Environmental sampling was conducted at fixed intervals (14 days), except for sampling on October 21, 2022 (one day before the ICU received patients). In the patient rooms, samples were collected from a ventilator, bedside instruments (ECG monitor, infusion pump, vaporizer, and hanging tower), bed rails, and an operation tabletop. In corridor areas outside the patient rooms, we sampled nurse hand washing sinks, computer keyboards, and mice adjacent to the patient room doors. Water sinks were sampled from the patient room toilet. At the nurse station, treatment room, and clinical examination room, samples were collected from the entire surface of computer keyboards and mice, as well as tabletops. The handwashing sink located at the nursing station was also sampled.

Bacteriological Methods

Patients and environmental sites were sampled using sterile swabs (i-Quip; Shanghai, China) moistened with tryptic soy broth (TSB; Hopebio, Qingdao, China). All swabs were immediately placed into 15 mL sterile tubes containing 2 mL of TSB after sampling. After overnight incubation at 37°C, 10 μL fluid medium was spread on Pseudomonas selective plates (Hopebio, Qingdao, China) to isolate P. aeruginosa (presence/absence only). The species of the collected strains was later confirmed using VITEK^® MS (BioMérieux, Lyon, France).

Antimicrobial Susceptibility Testing

For 484 collected P. aeruginosa strains, The minimum inhibitory concentrations (MICs) for a series of antibiotics, including MEM, IMP, ceftazidime (CAZ), ceftazidime/avibactam (CZA), cefepime(FEP), aztreonam (AZT), piperacillin/tazobactam (P/T4), levofloxacin (LEV), ciprofloxacin (CIP), polymyxin B (POL), and amikacin (AMK), were determined using the broth microdilution method in accordance with the guidelines of the Clinical and Laboratory Standards Institute (CLSI).¹⁸ The tested concentration range for each drug was listed in Supplementary Table 1. Strains with IMP or MEM MIC ≥ 8 mg/L were defined as CRPA.¹⁹ P. aeruginosa strain ATCC 27853 was used as a control. In addition, we compared the overall resistance rate of P. aeruginosa to all first-line anti-Pseudomonas drugs before/after the start of COVID-19 upsurge period (Dec 16, 2022 was set as boundary). The statistical analysis was based on Chi-square test.

Whole-Genome Sequencing

WGS was performed in all 148 CRPA strains. Genomic DNA was extracted using the QIAamp DNA Mini Kit (Qiagen; Hilden, Germany) and genome sequencing was conducted on the Illumina HiSeq X10 platform (Illumina, San Diego, CA, USA) after DNA library preparation. The CRPA strains P150F1 (Patient No. 180), P145F (Patient No. 136), and P167F (Patient No. 150) were subjected to additional long-read genome sequencing using a MinION Sequencer (Nanopore, Oxford, UK).

Genome Assembly and Profiling

Short-read sequencing data were assembled using Shovill version 1.1.0 (https://github.com/tseemann/shovill). Long-read data were first assembled with Raven version 2.10.3²⁰ and then polished using Pilon version 1.19.²¹ We performed sequence typing (ST) with assembled genomes queried in the PubMLST database²² (https://github.com/tseemann/mlst). The serotype of each P. aeruginosa strain was determined using PAST version 2.0.²³ The acquired antimicrobial resistance genes and virulence factors were identified using abricate version 1.0.1 (https://github.com/tseemann/abricate) from the resfinder²⁴ and vfdb²⁵ databases, respectively.

SNP Calling and Phylogenetics

Core genome single nucleotide polymorphisms (cgSNPs) were identified among CRPA isolates using genome alignment generated by snippy version 4.4.5 (https://github.com/tseemann/snippy) and recombination sites were filtered by gubbins version 3.3.5.²⁶ A maximum-likelihood phylogenetic tree of CRPA was constructed using RAxML version 8.2.²⁷

Divergence Time Estimation with Bayesian Analysis

The substitution rate and the most recent common ancestor were then estimated using BEAST version 1.10.4²⁸ with a strict molecular clock type and HKY nucleotide substitution mode. We downloaded all ST463 strain sequence data with available sampling dates before May 8, 2024, from the NCBI database to estimate the divergence time of the ST463 strains collected in the current study. The outcome files with 80 million MCMC chain lengths showed that all parameters were greater than 200 ESS values with convergence. TreeAnnotator version 1.10.4 was used to generate a maximum clade credibility tree, based on the resulting log files.

Plasmid Conjugation

Based on bla_KPC-71 plasmid carried by P167F strain (from Patient No. 150), we performed conjugation experiments using the filter-mating method with a rifampicin-resistant P. aeruginosa PAO1 strain derivative as the recipient. Mueller-Hinton agar plates loaded with rifampicin (300 μg/mL) and CZA (8/4 μg/mL) were used for selection. Polymerase chain reaction (PCR) and Sanger sequencing were used to verify the putative transconjugants. The experiments were independently repeated three times.

Cloning Experiments

In the P167F strain (from Patient No. 150), we amplified bla_KPC-71 gene with an upstream promoter region, whereas bla_KPC-2 gene in the P145F strain (from Patient No. 136) with the same promoter sequences was used as a reference. They were cloned into the plasmid vector pGK1900 and transformed into E. coli DH5a and P. aeruginosa PAO1.²⁹

Fitness Evaluation

As previously described,³⁰ we used the growth curve, area under the curve (AUC), and relative growth rate as fitness indicators. The target strains were grown in MH broth overnight, and nine replicates of the diluted culture (1:100 ratio) were placed in a 100-well plate at 37°C with shaking. The optical density of each culture was measured at 600 nm (OD600) every 5 min for 20 hours using a Bioscreen C Analyzer (Oy Growth Curves Ab. Ltd., Finland), and the growth rate was calculated based on OD600 curves using R script.³¹ Using GraphPad Prism 8 software, we plotted growth curves and calculated the corresponding AUC. Two-way analysis of variance (ANOVA) and Tukey’s honestly significant difference (HSD) analyses were used to evaluate differences between all tested strains (significant probability: p ≤ 0.05).

Antimicrobial Consumption Surveillance

Data on antibiotic consumption in the ICU during the study period were provided by the Pharmacy Department of Sir Run Run Shaw Hospital. We applied linear regression analysis to estimate temporal trends. Antibiotic consumption (defined as daily doses/100 patient days) was set as the dependent variable, and surveillance time was set as the independent variable.

Results

CRPA Isolated from Environmental and Patient Samples

353 patients were included in the study with the study flow chart displayed by Figure 1 and the ICU layout shown in Figure 2. The first patient was admitted to a newly opened ICU ward on October 22, 2022. Initial sampling on October 21 and October 24, 2022, revealed no CRPA contamination in the new ICU. The first CRPA strain isolated from this environment was detected on November 7, 2022. During the six-month study period, we collected 1694 samples from the sampling sites. Among all samples, 262 (15.47%) were P. aeruginosa positive including 40 CRPA strains (39 in patient room areas and 1 in the nurse’s station). In the immediate surroundings of patient beds, nine CRPA isolates were found on bedside instruments (ECG monitor, infusion pump, vaporizer, and hanging tower), five on bed rails, three on ward operation tabletops, and two on ventilators. Among the shared areas near patient rooms, sinks were the most severely CRPA-contaminated sites, with 10 strains from corridor sinks and eight from patient toilet room sinks. CRPA was detected in 52.6% of the patient rooms (10 out of 19), with the longest continuous isolation period of 14 days in Room 2 (Table 1, Supplementary Table 2, Figure 3A and B).

Table 1 ICU Environment Sampling Results

Figure 1 Flow chart for CRPA strain selection.

Figure 2 The ICU layout. There were 19 rooms with 21 beds (all rooms were one-bed except room 10, which comprised three beds). The different sampling sites were labeled with different colors and shapes. Abbreviation: ECG, Electrocardiograph.

Figure 3 CRPA surveillance results. (A) Isolation number of environmental PA in different timepoints. (B) Isolation number of environmental CRPA in different timepoints. (C) Isolation number of patient PA in different periods. (D) Isolation number of patient CRPA in different periods. (E) The overall resistance rate of P. aeruginosa to all first-line anti-Pseudomonas drugs before/after the start of COVID-19 upsurge period (Dec 16, 2022 was set as boundary). Statistical analysis was performed through Chi-square test. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

We collected 108 CRPA isolates from 47 patients. The mean age of the patients was 66 years (standard deviation [SD] = ± 20.48 years), ranging from 18 to 96 years. Most patients were male (74.5%; n = 35). The median length of hospital stay was 34 days (interquartile range [IQR], 10–44 days), and 59.6% (n = 28) of the patients experienced poor outcomes, such as deterioration or death. After the COVID-19 upsurge, eight patients with CRPA were diagnosed with COVID-19 infection. We considered 15 patients to be CRPA carriers upon admission, whereas 32 patients acquired CRPA during their ICU stay. Of all the CRPA strains isolated from patient samples, 52.8% (n = 57) were from nasopharyngeal swabs and 47.2% (n= 53) were from rectal swabs (Supplementary Table 3 and Figure 3C and D). The overall P. aeruginosa resistance rate to all first-line anti-Pseudomonas drugs increased after the start of COVID-19 upsurge period (Figure 3E).

Phylogenetic Analysis for CRPA and ST463 Strains

We performed phylogenetic analysis of all 148 CRPA isolates acquired in the current study based on the maximum-likelihood method. Our sampling yielded ST463 as the most prevalent CRPA sequence type in the environment (54.8%, n = 23) and patients (45.8%, n = 49). Based on pairwise SNP distances (SNP < 26 as the threshold), ST463 was assigned to three clones and one monophyletic group. Clone 1, comprising of 34 isolates, was associated with prolonged and widespread transmission within the ICU. Clones 2 and 3 were relatively small in size, and the monophyletic group consisted of two strains from Patient No. 097 (Figure 4). The average intra- and inter-clonal SNP distances among the three clones are shown in Figure 4. In contrast, ST357 and ST1596 formed two distinct clades. Environmental contamination by these sequence types was minimal, with only one ST1596 sample isolated from a bedside instrument in Room 5 on February 27, 2023 (Figure 4).

Figure 4 Maximum-likelihood phylogenetic tree for all CRPA. The predominant ST463 were assigned in three clades and one monophyletic group with different colors according to their pairwise SNP distance. The embedded table displayed the average intra-/inter-clonal SNP distance for three main clades. The annotation bars of binary data described the resistance gene profiles (blue circle if present) and results of antimicrobial resistance tests (red circle if resistant). Abbreviations: MEM, meropenem; IMP, imipenem; CAZ, ceftazidime; CZA, ceftazidime/avibactam; FEP, cefepime; AZT, aztreonam; P/T4, piperacillin/tazobactam; LEV, levofloxacin; CIP, ciprofloxacin; POL, polymyxin B; AMK, amikacin.

We also performed Bayesian analysis for all 432 Chinese ST463 isolates with available sampling dates in the NCBI database (uploaded prior to May 8, 2024), together with 72 ST463 strains acquired in our study, to investigate the ancestor divergence time and nucleotide substitution rate. The analysis estimated that the common ancestor of ST463 strains in China occurred in 1942 (95% highest posterior density [HPD]: 1930–1952). The substitution rate of the ST463 clone core genome was estimated to be approximately 1.984 single nucleotide polymorphisms (SNPs) per genome/year. For the ST463 strains identified in the current ICU study, the common ancestor was estimated to have existed in 2005 (95% HPD: 2003–2006), which was approximately 16 years earlier than the oldest isolate from the present collection (2022). Clone 1 was located in a distinct cluster 1 with three sub-clusters, clones 2 and 3 formed an adjacent cluster 2, and two isolates from Patient No. 097 formed a monophyletic group (cluster 3). The long phylogenetic distances between the clusters suggest a significant divergence from the common ancestor (Figure 5).

Figure 5 Bayesian analysis for ST463 strains. Three clusters from the current study are labeled with different colors.

Increased CRPA Transmission During COVID-19 Upsurge Period

Based on the literature^32,33 and data from the China Disease Control Center website (https://www.chinacdc.cn/), we defined December 16, 2022–January 19, 2023, as the COVID-19 upsurge period in Zhejiang, China. The curves on the right side of Figure 6 reflect the number of COVID-19 pneumonia patients in the ICU and the daily-reported COVID-19 nucleic acid test-positive cases in China during the study period. Spatiotemporal analysis with sequencing data revealed clonal spread of the ST463, ST1596, and ST357 strains in ICU settings. Sporadic transmission of ST463 clone 1 was detected before the COVID-19 upsurge period in Rooms 10 and 13 without environmental contamination. Patient No. 045 (Room 13), although not screened upon admission, was considered the initial introducer of clone 1 based on the earliest isolation time and prior CRPA infection diagnosis. During the COVID-19 upsurge period, ST463 clone 1 transmission events involved six patients across six different rooms. A re-introduction event by carrier Patient No. 136 on December 10, 2022, was deduced to be the contamination source for the clone 1 outbreak. Within two months of the COVID-19 upsurge period, clone 1 carriers caused prolonged contamination in the ICU and led to novel CRPA acquisition events in five patients (four related patients had been hospitalized in the ICU during the COVID-19 upsurge period). After March 20, 2023, transmission frequency declined significantly, with only one acquisition event detected in Patient No. 338 on April 26, 2023. Transmission is related to environmental contamination, and sink traps are important reservoirs. ST463 clone 2 was detected from February 2023 to April 2023. Clone 2 was first isolated in the bedrail of Room 17 on February 27, 2023; however, the primary patient sample was isolated 25 days later. ST463 clone 3 spread sporadically during the pre-pandemic period in Patient No. 012 (room 1), Patient No. 027 (room 5), and Patient No. 051 (room 18) with a narrow time window. We first acquired ST357 strains in patients staying in adjacent rooms 6 and 7 simultaneously (January 16, 2023); thus, the direction of transmission was unclear. However, the ST357 strain in Patient No. 150 (room 7) was not detected later but was replaced by ST463 clone 1. ST1596 was introduced by Patient No. 211 in room 8 on February 10, 2023, and subsequently transmitted to four patients within three adjacent patient rooms (Figure 4 and 6).

Figure 6 Bubble diagram of patient-related CRPA transmission. The horizontal axis illustrates the room number, while the vertical axis indicates the dates. Patients are labeled with numbers and the longitudinal lines connecting patient labels represent the ICU staying length. Strains of different STs are drawn in different colors. For ST463, isolates with less than 26 SNP are assigned in the same clone with a unique color. Environmental isolates are marked by red frames. The radium of each bubble is proportional to the number of isolates originated in the same room and the same timepoint, and the scale of the radium is shown. Multi-colored bubbles are shown as pie charts to reflect the different strains isolated in the same room and the same timepoint. The COVID-19 upsurge period is represented by a special background color. Curves in the right part of the bubble diagram demonstrate the shifts of COVID-19 patient numbers in the ICU (black curve) and the reported national COVID-19 nucleic acid test positive cases (red curve) during the study. aThe data was published on the website of Chinese Center for Disease Control and Prevention (https://www.chinacdc.cn/jksj/xgbdyq/202411/t20241112_302588.html). bRoom 10 has three beds, and another room has a single bed. cSampling paused from December 20, 2022, to January 5, 2023, due to limited manpower during COVID-19 upsurge.

Resistance Profile of CRPA

Among all CRPA, the resistance rates to MEM were 88.4% and 92.3%, respectively. All strains were susceptible to AMK and POL. In addition to AMK and POL, CZA was effective against a large proportion of CRPA (resistance rate: 28.4%). In contrast, the resistance rate to quinolones was notably higher (LEV: 71.8%, CIP: 72.4%) than that reported in the CHINET project.² The overall antibiotic resistance rate elevation of P. aeruginosa was driven by the higher isolation rates of ST463, ST1596, and ST357 after COVID-19 upsurge period (Figure 3E). These three nosocomial ST types exhibited high-level resistance to different classes of antibiotics and bore multiple resistance genes, including aac(6′)-IIa, ant(2″)-Ia, aph(3′)-IIb, bla_CARB, bla_KPC, bla_OXA, bla_PDC, catB7, crpP, fosA, sul1, and tetA (Figure 4).

We identified three variants of bla_KPC (bla_KPC-2, bla_KPC-33, bla_KPC-71) among the 84 CRPA strains. bla_KPC-33 and bla_KPC-71 conferred higher resistance to CZA, but lower resistance to carbapenem and AZT than bla_KPC-2. bla_KPC-2 was the predominant KPC variant (88.1%, n = 74), and was related to four sequence types: ST463 (n = 61), ST1596 (n = 10), ST1212 (n = 1), and ST316 (n = 2). ST1212 was detected in only one patient, whereas ST316 was found exclusively on bedside instruments and bedrails in room 6 at the early stage of the study (December 19, 2022). All bla_KPC-33 strains belonging to ST463 were identified in two patients with associated environmental contamination, whereas bla_KPC-71 strains (n = 4) were isolated from a single patient (Figure 4). Three different strains with respective bla_KPC gene (bla_KPC-2, bla_KPC-33, and bla_KPC-71) from three patients went through Nanopore long-read sequencing. The results revealed highly similar background of bla_KPC genes in each strain. All genes were located in single plasmid and the plasmid identity was higher than 99.9%.

Virulence Genes in CRPA

Multiple virulence genes were detected, including exotoxin-encoding genes (ToxA), T3SS effectors (exoU, exoS, exoT, exoY), T6SS effectors (tse1, tse2, and tse3), elastase (lasA and lasB), alkaline protease (aprA), adherence factors (fliC and fleP), siderophore metabolism (pvd and pch), pyocyanin secretion (phzA1), hemolysin secretion (plcH), and a two-component sensor (fleS). Specifically, exoU and exoS double positivity (exoU/S⁺) was observed in all ST463 CRPA strains, but not in other ST types (ST-specific). The exoU/S⁺ pattern was previously reported to be significantly more virulent than the single-positive strains.³⁴

A ST463 CRPA Strain Carrying bla_KPC-71

A 57-year-old male (Patient No. 150) was admitted to our ICU on December 16, 2022 (day 1) because of traumatic brain injury caused by a car accident and secondary pulmonary infection. The patient was administered empiric anti-infective therapy with P/T4 and Vancomycin from days 1 to 9. Owing to the rash, the treatment was switched to MEM and Vancomycin from day 9 to day 25. On day 25, the patient was diagnosed with an intracranial CRPA infection based on cerebrospinal fluid culture. The treatment was changed to intravenous CZA from day 25 to day 51 combined with intraventricular POL administration, following external ventricular drainage from day 25 to day 33. The CRPA-carrying bla_KPC-71 was detected on day 30. On day 35, the external ventricular drainage fluid was clear, indicating good control of intracranial PA infection, and the ventricular drainage tube was removed. However, on day 54, the patient had fever and elevated inflammatory markers and was treated with MEM and Vancomycin from day 54 to day 60. On day 60, the patient’s condition improved and he was discharged from the hospital. Detailed information on the infection and antibiotic application was shown in Figure 7A.

Figure 7 We acquired the information of (A) infection and antibiotic usage timelines in Patient No. 150 with blaKPC-71 CRPA isolation and constructed (B) a phylogenetic tree for all related strains from Patient No. 150. * CRPA strain P145F carrying blaKPC-2 from Patient No. 153 was also included in the tree (isolated on January 30, 2023, from a rectal swab). The isolation time after admission was calculated from the date of Patient No. 150 admission (December 16, 2022). Later, we compared (C) blaKPC genetic background in different plasmids after BLAST alignment. Blue shade between plasmids represented nucleotide identity (80–100%) in BLASTn. Colored arrows indicate predicted ORFs (red, resistance genes; yellow, other CDS; blue, transfer related gene; purple, replicative protein). Cloning experiments were performed to construct PAO1-pGK1900-KPC2/71 strains and then their growth conditions were assessed by (D) relative growth rate, (E) area under curve (AUC), and (F) growth curve using PAO1-pGK1900 as a reference. In figure D, E, F, error bars showed the standard deviations. Statistical analysis was based on unpaired t tests with two-way analysis of variance (ANOVA) and Tukey’s honestly significant difference (HSD) analysis for unequal variances. ns, P > 0.05, ***P < 0.001; ****P < 0.0001.

Characterization of bla_KPC-71 Strains

From Patient No. 150, we isolated a series of CRPA from multiple types of samples including nasopharyngeal swab, rectal swab, CSF, eye lid pus, PICC periphery blood, and sputum (Figure 7A). All ST463 strains isolated from Patient No. 150 were resistant to all the antibiotics tested in our study, except AMK and POL. Phylogenetic analysis revealed a close relationship among all ST463 strains in Patient No. 150 (Figure 7B).

P167F carrying bla_KPC-71 (isolated on February 15, 2023, from Patient No. 150) and P145F carrying bla_KPC-2 (isolated on January 30, 2023, from Patient No. 153) were subjected to long-read sequencing. We detected a single plasmid in each strain harboring the corresponding bla_KPC gene with a high identity (coverage > 99.9%, identity > 99.9%). Average SNP distance analysis (SNP = 8.95 between bla_KPC-2 and bla_KPC-71 strains within Patient No. 150) and spatial-temporal information demonstrated within-host evolution from bla_KPC-2 to bla_KPC-71 after nosocomial infection with bla_KPC-2 CRPA and CZA treatment in Patient No. 150. Both bla_KPC genes were located within the conserved platform (ISKpn27-bla_KPC-ISKpn6) flanked by two copies of IS26, indicating potential horizontal transmission. Subsequent BLAST in the NCBI database showed that the two plasmids shared high identity (> 98%) with pZYPA01 (GenBank accession number MZ050803), a bla_KPC-2 plasmid from an ST463 P. aeruginosa strain reported in 2021, Hangzhou, China. The main difference was that the two plasmids in our study lacked parts of Tn3 elements upstream of bla_KPC-2 upstream. Inter-species comparison of P. aeruginosa bla_KPC-71 plasmids with plasmids reported in K. pneumonia and S. marcescens^3,35 revealed the importance of a conserved genetic platform (ISKpn27-bla_KPC-ISKpn6) (Figure 7C).

Conjugation experiments with pbla_KPC-71-P167F failed, indicating that the plasmid was not transferable. Zhang et al failed to conjugate a similar plasmid in a previous study.⁴ By cloning experiments using the plasmid vectors pGK1900 and P. aeruginosa PAO1, we constructed PAO1-pGK1900-KPC2 and PAO1-pGK1900-KPC71 strains. We observed high resistance to CZA, decreased resistance to AZT and P/T4, and restored susceptibility to IMI and MEM in PAO1-pGK1900-KPC71 compared to PAO1-pGK1900-KPC2 (Table 2). Using the vector-carrying strain as a reference, we found a decreased relative growth rate (p < 0.001), demonstrating the fitness cost of bla_KPC-2 and bla_KPC-71 strains (Figure 7D–F).

Table 2 MICs (mg/L) Determined by Broth Microdilution Methods

Antibiotic Consumption in ICU

According to data from the Pharmacy Department of Sir Run Run Shaw Hospital, the overall consumption of antibiotics peaked from February 2023 to March 2023, with a one-month lag after the CRPA outbreak in January 2023. An increase in the use of β-lactam/β-lactamase inhibitor (BLBLI) combinations and polymyxin B was also observed during this period, whereas the consumption of other antibiotic classes remained relatively stable (Figure 8).

Figure 8 Main-class antibiotic consumption changes in the ICU during the study period, in defined daily doses (DDD) per 100 bed-days. Y axes show DDD per 100 bed-days. The last picture represents the overall antibiotic consumption. Abbreviation: BLBLI, beta-lactam/beta-lactamase inhibitor combination.

Discussion

The COVID-19 pandemic, which saw an unprecedented upsurge between December 2022 and February 2023, posed a significant threat to public health in China owing to a rapid policy transition from a zero-COVID approach to a more open stance. Within approximately 35 days, the COVID-19 infection rate among residents soared to nearly 90%, offering a unique context for investigating bacterial transmission and epidemiology within an ICU located in a densely populated city, during a rapid and extensive respiratory virus outbreak.³³ We hypothesized that changes in hygiene policies, infections among patients and healthcare workers, and the misuse of antibiotics during the COVID-19 period may affect bacterial transmission.¹⁷ However, compared to retrospective studies, there is a scarcity of prospective and comprehensive research on the transmission of P. aeruginosa during the COVID-19 period, with existing studies being both limited and fragmented. A few studies prospectively described P. aeruginosa epidemiology in the environment³⁶ and in patients^37,38 during the COVID-19 period, but none covered both aspects in a single study. For example, Sukhum et al studied P. aeruginosa epidemiology with WGS in hospital environment in Missouri, USA, but missed corresponding patient screening.⁹

In this study, we conducted prospective and ongoing CRPA screening in a newly established ICU during a COVID-19 upsurge period. The complex and dynamic patterns of CRPA introduction, re-introduction, and transmission were elucidated using a combination of whole-genome sequencing and spatial-temporal information. The CRPA isolates identified in the ICU were predominantly ST463 and belonged to O4 serotype. ST463 CRPA, known to carry bla_KPC-2 plasmid, has been reported to be prevalent in the southeast China region, including Zhejiang and Jiangsu province.^39–41 This strain exhibited high-level resistance to empirical carbapenems and cephalosporins, along with high virulence owing to its ExoU/S⁺ genetic profile.^34,39

Environmental contamination by CRPA has been documented in various regions.^9,10,42,43 Our study found that nearly 50% of CRPA transiently persist in the environment, contaminating a few patients. This underscores the necessity for stringent ward disinfection protocols and the strategic layout of ICUs with a single room for each patient. However, ST463 clone 1 led to sustained environmental contamination after January 15, 2023, which was primarily attributed to the longitudinal hospitalization of patients with ST463 CRPA. Conversely, the environmental colonization of CRPA, especially in water sinks and various surfaces, serves as a reservoir that facilitates patient acquisition.⁹ The transmission pathway from the CRPA source to the final receptors (susceptible patients) involves intermediaries such as medical devices, cleaning equipment, and ICU staff, with the detailed transmission process being intricate. The concurrent transmission routes across various rooms indicate that these incidents might have been facilitated by a single contaminating staff member. Ineffective infection control measures have led to the colonization of multiple patients by the CRPA ST463 clone during the COVID-19 upsurge period, in contrast to the disappearance or limited transmission of other clones. The CRPA outbreak emphasized the critical need to adhere to basic hygiene practices, including hand washing, routine environmental disinfection, and potential contaminator patient monitoring, particularly during public health emergencies.⁴⁴ Drug surveillance data indicated that antibiotic usage was not the sole contributing factor to the CRPA outbreak. Further studies are required to identify the factors contributing to the outbreak.

Pairwise SNP distance comparison is the golden standard for exploring the relatedness of bacterial strains during transmission events. Core genomes utilize the same genetic content for SNP calling across different bacterial lineages, allowing the cgSNP cutoff to be applied uniformly across all strains in the alignment.⁴⁵ While whole-genome SNP offer higher SNP coverage, cgSNPs are more efficient and suitable for identifying nucleotide substitution rates and analyzing inter-population diversity.⁴⁶ As suggested previously, we chose 26 as an appropriate cgSNP cutoff for accurate epidemiological analysis,⁴⁷ and the cutoff value selection was mainly rooted in the nucleotide substitution rate and within-host SNP diversity.⁴⁵ Focusing on ST463, Bayesian analysis of all ST463 strains from our surveillance along with 427 ST463 sequences downloaded from the NCBI database indicated a substitution rate of approximately 1.984 SNPs per genome per year in the current study. This evolutionary rate suggests a steady population development process, which is consistent with previously published data.⁴⁸ Within-host SNP diversity, particularly in long-term carriers, cannot be neglected in nosocomial infection studies. For instance, up to 611 SNP distances have been observed among different ST870 P. aeruginosa strains isolated from the same sample from a single patient.⁴⁹

In this study, we presented the first report of a bla_KPC-71-producing P. aeruginosa. bla_KPC-71, a variant of bla_KPC-2 with Ser 182 duplication, was initially identified in K. pneumoniae.³ In southeast China, a similar bla_KPC-harboring plasmid has been previously reported,⁴ indicating a possible clonal spread in the same region. The key ISKpn27-bla_KPC-ISKpn6 unit with IS26-ΔTn6296 mobile genetic element mode has been described in P. aeruginosa and other species.^3,35,50 Conservation of the genetic platform suggests the potential for the horizontal transfer of resistance genes.

Cloning experiments using bla_KPC-2 as a reference in the PAO1 and DH5α strains showed that bla_KPC-71 acquired increased resistance to CZA while restoring susceptibility to carbapenem. From previous enzyme kinetic data, 182S insertion in the KPC-71 variant sequence resulted in higher affinity toward ceftazidime, lower sensitivity to avibactam and reduced hydrolytic activity to carbapenems compared to wild-type KPC-2. KPC enzyme structure contained 4 loops surrounding the active site core. It is noteworthy that KPC-71 mutation located outside the loop region of KPC sequence, while a large number of CZA-resistance-related mutations were reported within the omega-loop of the protein region.^51–53 Globally, KPC-producing P. aeruginosa isolates are prevalent, with bla_KPC-2 being the predominant bla_KPC gene type.⁵⁴ Following the use of CZA to treat bla_KPC-2-producing CRPA, bla_KPC-2 variations that confer CZA resistance and restore carbapenem susceptibility present a new challenge.^29,55 Several bla_KPC-2 variants, including bla_KPC-71, were undetectable using the CARBA 5 test, a commercial rapid detection method for KPC-producing bacteria.³

In contrast to the cloning experiments, the bla_KPC-71 clinical strains from Patient No. 150 exhibited moderate carbapenem resistance, which was lower than that observed for bla_KPC-2 strains (Table 2). The incomplete carbapenem susceptibility restoration after bla_KPC gene type switch may be associated with additional factors such as MexB overexpression and OprD deactivation.⁸ The heterogeneity between bla_KPC-71 and bla_KPC-2 strains in Patient No. 150 could also compromise the efficacy of treatment regimens involving carbapenems and CZA.

Our study has some limitations, largely due to limited human resources. A prolonged sampling interval and a pause in sampling from December 20, 2022, to January 5, 2023, coupled with the absence of surveillance of ICU staff and cleaning tools, may have resulted in undetected transmission pathways. Furthermore, as this was a single-center study, it may not accurately represent the regional CRPA transmission dynamics outside the ICU. Despite these limitations, this study offers insight into CRPA transmission, which can enhance hospital infection control measures during public health emergencies. Additionally, we examined the bla_KPC-71 mutation that emerged in CRPA following CZA treatment, thereby providing a reference for optimizing the clinical antibiotic stewardship.

Conclusion

The prospective study showed that COVID-19 upsurge is associated with increased P. aeruginosa transmission. We described CRPA contamination in environments and patients, in which ST463 strains carrying KPC were recognized as high-risk clone. We also reported for the first time carbapenemase mutation from bla_KPC-2 to bla_KPC-71 in P. aeruginosa with elevated CZA resistance. These findings will improve nosocomial infection management and promote better antibiotic stewardship in future public health crisis.