This study was approved by the Ethics Committee of Southwest Hospital, Third Military Medical University. No written informed consent was required because we received anonymized isolate samples with all the personal information removed.

Clinical P. aeruginosa isolates were collected from 2011 to 2016 at the Institute of Burn Research at Southwest Hospital in Chongqing, one of the largest burn center in China, with 150 beds that takes a mass of patients mainly from southwest China. Strains were isolated from the Bl and wound surface of burn patients and identified using the API 20 NE system (BioMerieux) and 16S rRNA gene sequence analysis (AbdulWahab et al., 2015). The isolates were grouped depending on whether they were obtained from the Bl, the wound surface of a patient without bacteremia (WN), or the wound surface of a patient with bacteremia (WB). Every strain was the first isolate of a series of sample collection during hospitalization that isolated from the specific source of the patient. P. aeruginosa PAO1 (Jacobs et al., 2003) was used as a reference strain in the quantification of ampC and the efflux pump gene expression.

Multilocus sequence typing was performed as described in the PubMLST database of P. aeruginosa¹. Briefly, genomic DNA from each isolate was extracted from cultures grown to the late exponential phase by using a Genomic DNA Purification Kit (Promega). The resulted genomic DNA was used as template to amplify seven housekeeping genes (acsA, aroE, guaA, mutL, nuoD, ppsA, and trpE) by PCR as described in the PubMLST database² with a few modifications. 2 × Taq premix (Takara) was used in PCR according to the recommended conditions. The amplification reaction for aroE required the addition of 5% DMSO (dimethyl sulfoxide). The PCR products were purified with Gel Extraction Kit (OMEGA) and then underwent bidirectional sequencing using the Applied Biosystems (ABI) 3730 DNA analyzer. Gene sequences were used to query the PubMLST database to identify matches to known (numbered) alleles. The seven allele numbers were combined to construct an identifier for a ST. The types that could not match any known types were deposited to obtain new STs. BioNumerics (version 7.6) was used to analyze the clonal relationships between the STs and create a minimum spanning tree. A clonal complex (CC) was defined to contain at least two STs sharing any six of the seven alleles.

Susceptibility to piperacillin, piperacillin/tazobactam, ceftazidime, cefepime, sulbactam/cefoperazone, amikacin, gentamicin, netilmicin, ciprofloxacin, and levofloxacin was tested for all isolates using the K-B agar diffusion method (CLSI 2011–2017, M100-S21-M100-S27). MICs of imipenem, meropenem, and polymyxin B were determined by the microdilution method. Susceptibility was categorized using the breakpoints defined by the Clinical and Laboratory Standards Institute guidelines (CLSI 2011–2017, M100-S21-M100-S27) and European Committee on Antimicrobial Susceptibility Testing (EUCAST, version 7.1, http://www.eucast.org, for polymyxin B only).

Multiplex PCR was used to detect acquired carbapenemase genes in carbapenem-non-susceptible isolates as described previously (Ellington et al., 2007; Poirel et al., 2011). Eleven genes were divided into 3 groups as follows: group 1 bla_IMP, bla_{V IM}, bla_NDM, and bla_GES, group 2 bla_SPM, bla_BIC, and bla_KPC, group 3 bla_AIM, bla_GIM, bla_SIM, and bla_DIM. One microliter of DNA (50 ng) was subjected to multiplex PCR in a 20 μL reaction mixture with 10 μmol/L of each primer shown in Supplementary Table S1 (Ellington et al., 2007; Poirel et al., 2011) and 10 μL 2 × Taq premix (Takara). When necessary, 5% DMSO was added to the reaction. Amplified products detected by agarose gel electrophoresis were sequenced.

The full-length oprD gene from each isolate was amplified and sequenced using the primers in Supplementary Table S1 (Rodriguez-Martinez et al., 2009). DNA sequences were compared with the oprD sequence from the reference strain PAO1 using MEGA7 (Kumar et al., 2016).

Expression of ampC, mexB, and oprD was measured using quantitative real-time PCR. Total RNA was extracted using TriPure RNA isolation reagent (Roche) and reverse transcribed to cDNA with a first-strand cDNA synthesis kit (Thermo Fisher Scientific). RT-qPCR was then performed in a CFX Connect Real-Time PCR System (BIO-RAD) using SYBR green real-time PCR master mix (TOYOBO). The rpsL gene was used as an internal reference. Primers for all the genes in RT-qPCR are shown in Supplementary Table S1 (Juan et al., 2006; Bubonja-Sonje et al., 2015). Three independent experiments were performed and the mean values of relative expression for each gene were compared with their corresponding expression levels in PAO1. The evaluation criteria which are used widely in many other studies were as previously described (Oh et al., 2003; Cabot et al., 2011; Vatcheva-Dobrevska et al., 2013; Bubonja-Sonje et al., 2015). For ampC overexpression is defined as a level at least 10-fold higher than the corresponding level in PAO1, negative if less than 5-fold higher, and borderline if between 5- and 10-fold. For mexB, overexpression indicates a level at least 3-fold higher, negative indicates the level is less than 2-fold, and borderline indicates the level is between 2- and 3-fold higher.

Data were analyzed using the GraphPad Prism analysis package. The antimicrobial susceptibility of the isolates in different groups was compared using Pearson’s chi-square test. P < 0.05 was considered statistically significant.

A total of 196 isolates from different sources (Bl = 73, WN = 80, WB = 43) were analyzed by MLST to investigate their clonal relationships. The results revealed high clonal diversity, with 42 known STs identified among 171 isolates, and 16 new STs among 25 isolates (Supplementary Table S2). ST316, ST111, ST360, ST244, and ST1158 were the dominant STs with at least 15 isolates in each type, and accounted for 51.53% of the total isolates. Six CCs were identified, in which CC111, CC360, and CC244 were considered as large CCs (Figure 1A). The top 5 STs accounted for 68.49 or 72.09% of the isolates collected from the Bl or wound surface of patients WB, respectively. However, among strains isolated from wound samples of patients WN, only 25 percent belonged to these 5 STs (Figure 1B). This result indicates that P. aeruginosa isolated from WN were more diverse than those isolated from Bl or WB. Among isolates from Bl, the main STs varied depending on the year of collection. In 2011 and 2012, ST316 and ST360 were the dominant types. Beginning in 2013, ST111 emerged and became the primary type in 2014 along with ST2483. In 2015, both lost their position to ST244 and ST1158 (Figure 1C). As there was only one isolate from Bl in 2016 which belonged to ST270, it was not presented in Figure 1C. In June and July 2014, 6 strains of ST111 caused an outbreak of Bl infections in our center (Supplementary Table S3).

Notably, in the 43 patients WB accompanied by wound infections, only 23 patients were infected with identical STs in their Bl and wound surface, while the other 20 patients were infected with two different STs simultaneously (Table 1).

The susceptibility test (raw data are shown in Supplementary File 1) revealed that 107 (54.59%) of the isolates exhibited an MDR phenotype. 42 (21.42%) strains were XDR and sensitive only to polymyxin B, which is rarely used in China. More seriously, these isolates showed high resistance to carbapenems, with only 70 (35.71%) of them susceptible to imipenem and 63 (32.14%) to meropenem. Since carbapenems are currently the most important therapeutic option to treat infections caused by MDR P. aeruginosa, the 141 carbapenem-non-susceptible isolates (either non-susceptible to imipenem or meropenem) were selected for further study to investigate their resistance mechanisms.

Compared with isolates collected from WN, the samples from Bl and WB showed higher resistance rate to the vast majority of the antibiotics tested including piperacillin (PIP), sulbactam/cefoperazone (CSL), piperacillin/tazobactam (TZP), ceftazidime (CAZ), cefepime (FEP), imipenem (IPM), amikacin (AMK), gentamicin (GEN), netilmicin (NET), ciprofloxacin (CIP), and levofloxacin (LVX) (Figures 2A–C). This suggests that P. aeruginosa isolated from the patients WB is more resistant than that isolated from patients with only wound infections.

In the top 5 STs, although no ST was associated with any specific resistance profile, ST360 and ST1158 showed higher resistance rate to most of the commonly used antibiotics in burn centers such as sulbactam/cefoperazone (90.91 and 93.33%), piperacillin/tazobactam (81.28 and 93.33%), ceftazidime (77.27 and 86.67%), cefepime (100 and 100%), imipenem (81.82 and 93.33%), and meropenem (81.82 and 93.33%) (Figure 2D).

Of the 141 carbapenem-non-susceptible P. aeruginosa, 11 (7.80%) isolates were positive for bla_IMP and 9 (6.38%) isolates harbored bla_{V IM}. Sequence analysis showed that these two genes encode IMP-9 and VIM-2, respectively (Table 2). No other carbapenemase genes were detected by PCR screening. Among these carbapenemase-positive isolates, ST499 (6 isolates) and ST111 (5 isolates) were the main types. One ST499 strain was positive for both bla_IMP and bla_{V IM}. Among the strains producing IMP-9, apart from 6 strains of ST499 and 2 strains of ST111, there was only 1 strain of ST244, ST316, and ST1028, respectively. While, in the strains harboring VIM-2, there were 3 strains of ST111 and ST260, respectively, 2 of ST1158, and only 1 of ST499. None of these STs that producing IMP-9 or VIM-2 belonged to the same CC.

Expression levels for ampC β-lactamase gene and the mexB multidrug efflux gene among carbapenem-non-susceptible or meropenem-non-susceptible isolates are shown in Table 2. AmpC hyperproduction is most common in these isolates. Ninety-six (68.09%) isolates were found to meet criteria for overexpression, and 12 (8.51%) isolates showed a borderline increase. The efflux pump gene mexB was overexpressed in 12.78% of the meropenem-non-susceptible isolates.

Mutational inactivation of the oprD gene is the major reason for OprD loss in P. aeruginosa (Rodriguez-Martinez et al., 2009; Fang et al., 2014). The oprD gene sequence was therefore determined in all 141 carbapenem-non-susceptible isolates (Table 3). 125 (88.65%) isolates were found to have various inactivating mutations, potentially resulting in truncated or aberrant proteins. 95 strains contained frameshift mutations due to deletions or insertions of 1 or several base pairs. Premature stop codons caused by point mutations were detected in 12 isolates. In addition, two insertion sequence (IS) elements were found to have interrupted oprD. In 13 ST1158 isolates, IS1411 inserted into the coding sequence of oprD, while in 5 ST499 isolates, IS1394 inserted upstream of the start codon in a location that could influence transcription initiation or translation. In the 16 isolates without inactivating mutations, 8 contained amino acid substitutions and no mutations were identified in 6 strains. In the 8 strains containing amino acid substitutions, 5 of them were resistant to imipenem, and the other 3 were sensitive. While in meropenem susceptibility assay, 5 were resistant, 2 were intermediate, and 1 was sensitive. This result suggests that there is no necessary connection between these substitutions and carbapenem resistance. Two isolates were negative by PCR detection. We speculate that these may have suffered more serious changes affecting the oprD gene, such as large deletions (Fang et al., 2014). We further measured oprD expression in these 14 isolates without inactivating mutations, and all of them showed reduced oprD expression (Supplementary Table S4).