The Acquisition of Resistance to Carbapenem and Macrolide-mediated Quorum Sensing Inhibition by Pseudomonas aeruginosa via a Novel Integrative and Conjugative Element ICETn43716385

Pseudomonas aeruginosa can cause persistant and life-threatening infections in immunocompromised patients. Carbapenems are the first-line agents to treat P. aeruginosa infections; therefore, the emergence of carbapenem-resistant P. aeruginosa strains has greatly challenged effective antibiotic therapy. In this study, we characterised the full-length genomes of two carbapenem resistant P. aeruginosa clinical isolates that produce the carbapebemase New Delhi metallo-β-lactamase-1 (NDM-1). We found that the blaNDM-1 gene is encoded by a novel intergrative and conjugative element (ICE) ICETn43716385, which also carries the macrolide resistance gene msr(E) and the florfenicol resistance gene floR. The msr(E) gene has rarely been described in P. aeruginosa genomes. To investigate the functional roles of msr(E) in P. aeruginosa, we exogeneously expressed this gene in P. aeruginosa laboratory strains and found that the acquisition of msr(E) could abolish the azithromycin-mediated quorum sensing inhibition in vitro and the anti-Pseudomonas effect of azithromycin in vivo. In addition, the expression of msr(E) almost completely restored the azithromycin-affected P. aeruginosa transcriptome, as shown by our RNA sequencing analysis. We present the first evidence of blaNDM-1 to be carried by intergrative and conjugative elements, and the first evidence of co-transfer of carbapenem resistance and the resistance to macrolide-mediated quorum sensing inhibition into P. aeruginosa genomes. Importance Carbapenem resistant P. aeruginosa has recently been listed as the top three most dangerous superbugs by World Health Organisation. The transmission of blaNDM-1 gene into P. aeruginosa can cause extreme resistance to carbapenems and fourth generation cephalosporins, which greatly compromises the effectiveness of these antibiotics against Pseudomonas infections. However, the lack of complete genome sequence of NDM-1-producing P. aeruginosa has limited our understanding of the transmisibility of blaNDM-1 in this organism. Here we showed the co-transfer of blaNDM-1 and msr(E) into P. aeruginosa genome by a novel integrative and conjugative element (ICE). The acquisition of these two genes confers P. aeruginosa with resistance to carbapenem and macrolide-mediated quorum sensing inhibition, both of which are important treatment stretagies for P. aeruginosa infections. Our findings highlight the potential of ICEs in transmitting carbapenem resistance, and that the anti-virulence treatment of P. aeruginosa infections by macrolides can be challenged by horizontal gene transfer.


Introduction
Pseudomonas aeruginosa is an opportunistic pathogen which can cause life-threatening infections in immunocompromised patients (1). It is also responsible for nosocomial infections such as chronic wound infections and ventilator-associated pneumonia, and chronic airway infections in cystic fibrosis and chronic obstructive pulmonary disease patients (2)(3)(4)(5). These infections are usually very difficult to eradicate and associated with high mortality rates (3,6). Worse still, P. aeruginosa is also notorious for its ability to develop multidrug resistance, which leaves only a handful of antibiotics remain effective to treat its infections in clinical practice (7).
Carbapenems such as imipenem and meropenem are the first-line agents for treating P.
aeruginosa infections and the last-ressort drugs in severe infections caused by Gram-negative bacteria (8). However, the clinical efficacy of carbapenems has been greatly compromised by the spreading of the carbapebemase New Delhi metallo-β-lactamase-1 (NDM-1), which is usually encoded and transmitted by broad-host self-conjugative plasmids in Enterobacteriacea spp. and Acinetobacter baumannii (9). The blaNDM-1 gene was first described in P. aeruginosa in 2011 and has ever since been identified in P. aeruginosa isolated from North America, Europe and Asia (10)(11)(12)(13)(14)(15)(16). The NDM-1-producing P. aeruginosa strains usually have extremely high resistance to carbapenems and many other classes of antibiotics, which makes infections caused by these superbugs even more difficult to treat (10)(11)(12)(13)(14)(15)(16). Therefore, it is important to develop novel treatment strategies for P. aeruginosa infections.
One alternative strategy for treating P. aeruginosa infections is to inhibit the production of virulence factors, which are essential for the pathogenesis of this bacterium. For instance, the type III secretion system and the secreted products such as elastase and exotoxin A have been shown to play important roles during the colonization of P. aeruginosa in human airways (17,18). In particular the secreted detergent rhamnolipid causes rapid necrosis of the host immune cells and protects bacterial cells from immune attack and facilitate the establishment of infections (19). In P. aeruginosa, the expression of many virulence factors is under tight control of its quorum sensing (QS) systems, which can be potential targets for the design of anti-virulence drugs (20,21). However, although several anti-QS compounds have been identified in the past, none of them has so far entered clinical trial (21)(22)(23). On the other hand, the macrolide antibiotics such as erythromycin, azithromycin (AZM) and clarithromycin were shown to have a promising anti-QS activity, which makes them ideal anti-virulence drugs for treating P. aeruginosa infections (21). It was suggested that macrolides repress the synthesis of QS signaling molecules by interfering with the signaling pathways of RsmZ and RsmY through yet-to-be identified targets, resulting in the downregulation of the QSregulated virulence products such as rhamnolipids and elastase in P. aeruginosa (24). The use of macrolides as QS inhibitor may therefore expand our antimicrobial arsenal against P. aeruginosa infections until new and more efficient anti-virulence drugs become clinically available.
Although NDM-1-producing P. aeruginosa strains have been prevalent worldwide, the complete genome information and transmission mechanisms is still lacking (10)(11)(12)(13)(14)(15)(16). To improve the understanding of the transmission of blaNDM-1 among P. aeruginosa strains, we sequenced the complete genomes of two local clinical NDM-producing P. aeruginosa isolates (14). Comparative genomic analysis showed that the blaNDM-1 gene is encoded by a novel Tn4371 family integrative and conjugative element (ICE), which is a class of mobile genetic element present in the genomes of a broad range of β-and γ-proteobacteria (25). In addition to blaNDM-1, this element also carries the macrolide resistance gene msr(E) and the florfenicol resistance gene floR. We found that the acquisition of msr(E) by P. aeruginosa abolished the AZM-mediated QS inhibition in vitro and the anti-Pseudomonas effect of AZM in vivo. To our knowledge, this is the first description of blaNDM-1 encoded by ICEs and the first evidence on the co-acquisition of carbapenem resistance and the resistance to AZMmediated QS inhibition by P. aeruginosa.

Identification of a novel NDM-1-producing P. aeruginosa group
A nosocomial outbreak of NDM-1-producing P. aeruginosa was previously reported in a local Singapore hospital and was the first case report of NDM-1-producing P. aeruginosa in Southeast Asia (14). The strains isolated during this outbreak exhibited multidrug resistance to carbapenems, cephalosporins, aminoglycosides, and fluoroquinolones, whereas remained sensitive to polymyxin B (Table S1). To identify the origin of these NDM-1-producing P.
aeruginosa isolates, we sequenced the draft genomes of 11 isolated strains on an Illumina MiSeq platform and found that all these isolates belong to multilocus sequence type ST308 and have harbored the same sets of antibiotic resistance genes (Fig. S1). In addition, the 12 genomes are highly similar to each other as shown by multiple genome alignment using Progressive Mauve (26) (Fig. S2). These results suggested that the 12 NDM-1-producing P. aeruginosa strains isolated in this outbreak belong to the same phylogenetic group. Therefore, we named this closely related P. aeruginosa group PASGNDM and selected two representative isolates, PASGNDM345 and PASGNDM699 for further investigation.
To better understand the detailed features of the PASGNDM genomes and the transmission of blaNDM-1 into P. aeruginosa genome, we further sequenced PASGNDM345 and PASGNDM699 genomes on a Pacific Biosciences RSII platform. The PacBio sequencing reads have achieved 163-and 161-fold coverage for PASGNDM345 and PASGNDM699 genomes, respectively, and were successfully assembled into two full-length genomes.
Construction of a phylogenetic tree using PASGNDM345 and PASGNDM699 genomes together with other 21 P. aeruginosa full-length genomes showed that the two PASGNDM strains formed a monophyletic group (Fig. 1). Interestingly, the closest genome to the PASGNDM group in the phylogenetic tree is PA_D1 (NZ_CP012585.1), which is an endemic P. aeruginosa strain causing ventilator-associated pneumonia in China as identified in previous work by us (manuscript submitted).

Characterisation of PASGNDM345 and PASGNDM699 genomes
The genome of PASGNDM345 consists of a circular chromosome of 6,893,164 bp with an average GC content of 66.1%, whereas the PASGNDM699 genome is 6,985,102 bp with an average GC content of 66.0%. In total, 6,503 and 6,589 genes were predicted from PASGNDM345 and PASGNDM699 genomes, respectively. General features of these predicted genes can be found in Table S2.
To identify the strain specific genomic regions in the PASGNDM genomes, the complete genomes of PASGNDM699 and PASGNDM345 were compared with six other strains clustered in the same clade in the phylogenetic tree (Fig. 1). Genome alignment result showed that the two PASGNDM genomes possess several regions with low sequence identity to the other six strains (Fig. 2). Furthermore, we also predicted the genomic islands (GIs) in the two PASGNDM genomes using Island Viewer 3 server (27). A total of 41 and 47 GIs were predicted from the PASGNDM345 and PASGNDM699 genomes, respectively, which correlate well with the strain specific regions in the PASGNDM genome ( Fig. 2 and Table S3, S4). The genes located in the GIs are mostly enriched in generating transposons, efflux pumps and multidrug resistance (Table S3, S4), which may be important for the survival and nosocomial spread of the PASGNDM strains. In addition, eight acquired antibiotic resistance genes including blaNDM-1 are embedded in the GIs of PASGNDM699 and PASGNDM345 genomes (Fig. 2), suggesting the importance of mobile genetic elements in the acquisition of antibiotic resistance by the PASGNDM isolates.

Identification of a novel ICETn43716385 encoding blaNDM-1, msr(E) and floR
It was noted that the blaNDM-1 is clustered together with msr(E) and floR in a GI region, suggesting the three antibiotic resistance genes might be co-transferred into the PASGNDM genomes (Fig. 2). Further sequence analysis revealed that the three genes are embedded in a 74.2 kb ICE-like element located between the exoY (PA2191) and hcnA (PA2193) genes in the PASGNDM345 and PASGNDM699 genomes (Fig. 3). Both ends of this element are flanked by a 5'-TTTTTTGT-3' sequence, which resembles the conserved attB site of almost all Tn4371 family ICEs (25). It also contains the core genes conserved among the Tn4371 family ICEs, including a int integrase gene, the parB, repA and parA genes of the ICE stabilisation system, and homologues to the DNA conjugative transfer machineries such as traI and traG (25) (Fig. 3). In addition, the integrase encoded by the int gene shared 71% identity with IntTn4371 (AJ536756), which further proved that the ICE identified here should be considered as a member of the Tn4371 family, as suggested by a previous study (25). We therefore named this element ICETn43716385 following the nomenclature system proposed by Roberts et al. (28).
To track the origin of ICETn43716385, we searched its entire sequence against Genbank and found that ICETn43716385 is similar to a 56.4 kb ICE-like element present in Pseudomonas fluorescens UK4 genome (CP008896.1). Comparative sequence analysis between the two elements showed that they shared a common Tn4371 ICE core gene scaffolds (25), whereas their major differences are the unique accessory gene cluster downstream of the int gene, and a 13.7 kb segment encoding the three antibiotic resistance genes harbored by ICETn43716385 (Fig. 3). This 13.7 kb segment is immediately flanked by 695 bp direct repeats (termed DR), which share 99.87% identity (694/695) (Fig. 3). Interestingly, only a single copy of this DR sequence was present in the ICE-like element of the P. fluorescens UK4 genome (Fig. 3). It is possible that the duplicated DR sequences in ICETn43716385 were the result of homologous recombination, which might lead to the acquisition of the 13.7 kb segment by ICETn43716385.
Therefore, we present a novel ICETn43716385 element identified from PASGNDM genomes, which have acquired three antibiotic resistance genes namely blaNDM-1, msr(E) and floR. The acquisition of the blaNDM-1 gene by PASGNDM699 is probably responsible for its extreme resistance to carbapenems (Table S1), whereas msr(E) and floR are associated with macrolides and florfenicol resistance, respectively (29, 30).

Acquisition of msr(E) protects P. aeruginosa from AZM-mediated QS inhibition
Previous studies have reported that sub-MIC concentrations of AZM could supress the expression of several QS-regulated virulence factors such as elastase and rhamnolipids and the swarming motility of P. aeruginosa, probably by inhibiting its QS systems (24,31,32).
We therefore hypothesized that the acquisition of msr(E) by P. aeruginosa could counteract the QS-inhibition effect of AZM. However, the investigation on the functions of msr(E) gene in the PASGNDM strains by a reductionist approach is difficult due to their multidrug resistance; hence, an alternative approach by exogenously expressing msr(E) gene in laboratory strains of P. aeruginosa was adopted for this purpose. We first amplified the msr(E) gene together with its putative promoter from PASGNDM699 genome and have the entire fragment inserted into a pUCP18 vector to construct pUCP18::msr(E) plasmid, in which the expression of msr(E) gene is sorely controlled under its putative promoter.
Transformation of pUCP18::msr(E) into the P. aeruginosa strains PAO1 and PA14 increased their MICs to the AZM by more than 8 fold (from 256 μg/ml to >2048 μg/ml), indicating that the acquired msr(E) in PASGNDM genomes can be expressed from its own promoter to enhance resistance to AZM in P. aeruginosa. Both the PASGNDM699 and PASGNDM345 have comparable AZM resistance levels to the two transformed strains, suggesting that the encoded msr(E) is expressed in the PASGNDM strains.
We then performed elastase and rhamnolipid quantification and swarming motility assays with PAO1/pUCP18::msr(E) and PA14/pUCP18::msr(E) strains. The results showed that 8 μg/ml of AZM (1/32 of the MICs of PAO1 and PAO1/pUCP18) have reduced elastase production in both wildtype PAO1 and vector-carrying strain PAO1/pUCP18 by at-least 40%, whereas the elastase produced by PAO1/pUCP18::msr(E) was not significantly affected upon AZM treatment (Fig. 4a). We also found that AZM have reduced the amount of rhamnolipid in the supernatant of both PAO1 and PAO1/pUCP18 overnight culture by at-least 60%, whereas PAO1/pUCP18::msr(E) produced similar levels of rhamnolipid in the presence and absence of AZM (Fig. 4b). In addition, the swarming motilities of PAO1 and PAO1/pUCP18 were significantly inhibited by 8 μg/ml of AZM, under which PAO1/pUCP18::msr(E) exhibited a normal swarming phenotype (Fig. 5). Similar results were also obtained using the PA14 strains ( Fig. 4c and 4d, Fig. 5), suggesting that the anti-AZM-mediated QS-inhibition effect of Msr(E) is not strain-specific. Taken together, these results clearly showed that the acquisition of msr(E) could protect P. aeruginosa from AZM-mediated QS inhibition.

Expression of msr(E) in P. aeruginosa restores AZM-affected transcriptome
AZM was previously reported to affect the transcriptome of P. aeruginosa in a microarray analysis (32). To better illustrate the effect of Msr(E) on the AZM-affected transcriptome of P. aeruginosa, we compared the gene expression profiles of PAO1/pUCP18::msr(E) in the presence and absence of 8 μg/ml AZM by total RNA sequencing. In PAO1, a total of 550 genes (~10% of PAO1 genome) were differentially expressed upon AZM treatment, of which 305 were upregulated and 245 were downregulated (Table S4, Fig. 6). We found that the upregulated genes are enriched in genes encoding, the type III secretion pathway apparatus, and other possible defense mechanisms against AZM (infA and efp) (Table S5). Whereas the downregulated genes include rsmY and rsmZ, the non-coding RNA products of which positively regulates QS by sequestering the translational inhibitor RsmA, pslE-I of the Psl synthesis operon, lasI of the LasI/LasR QS system, and the flK gene the expression of which is essential for swarming motility (Table S5). These results are consistent with findings in previous studies (21,24,(31)(32)(33) and could explain the QS-inhibition effect of AZM on PAO1 as shown in this study (Fig. 4, Fig.5, and Fig.6). Surprisingly, the transcriptome of PAO1/pUCP18::msr(E) was almost not affected by AZM, for which only 6 genes were differentially expressed upon AZM treatment (Table S6, Fig. 6). Therefore, Msr(E) probably protected P. aeruginosa from AZM-mediated QS inhibition by restoring AZM-affected transcriptome.

Msr(E) abolished anti-Pseudomonas effect of AZM in vivo
Previous studies showed that macrolide antibiotics such as erythromycin and clarithromycin could enhance the clearance of P. aeruginosa at the infection sites in a murine implant infection model and a murine lung infection model (34,35). To investigate if the acquisition of msr(E) by P. aeruginosa can affect the anti-Pseudomonas activity of AZM in vivo, we used the murine implant model to compare the effectiveness of AZM treatment on PAO1 and PAO1/pUCP18::msr(E) infections. Briefly, silicone implants pre-colonized with PAO1 or PAO1/pUCP18::msr(E) strains were inserted into the peritoneal cavity of mice. After 12 hours of incubation, mice were treated with AZM by injecting AZM solution into the peritoneal cavity, whereas the control mice were injected with the same amount of saline.
The silicone implants and mice spleens were harvested at 24-hour post infection to enumerate colony forming units (CFU) of P. aeruginosa, which were used to indicate the clearance of bacterial at infection site and the spreading of infection to other organs according to previous studies (35, 36). We found that AZM treatment could reduce the CFU of PAO1 residing in the silicone implants and the mice spleens by 2.0-log and 3.8-log, respectively (Fig. 7). By contrast, AZM treatment did not significantly affect the CFU of PAO1/pUCP18::msr(E) recovered from either the silicone implants or the mice spleens, as compared to the saline-treated control group (Fig. 7). These results indicated that Msr(E) could confer P. aeruginosa resistance against the anti-Pseudomonas activity of AZM in vivo.

Discussion
Tn4371 family ICEs have been previously described in a broad range of β-and γproteobacteria isolated from both environmental and clinical settings, which could confer their hosts adaptive functions such as multidrug/heavy metal resistance and the ability to metabolize specific carbon sources (25). Furthermore, a Tn4371-like element from P. aeruginosa genomes has recently been characterized to carry blaSPM-1 and bcr1, suggesting that the Tn4371 family ICEs is an important source for the acquisition of carbapenem resistance by P. aeruginosa (37). In the current work, we identified a novel ICETn43716385 from PASGNDM699 isolated from clinical sources. This element carries three antibiotic resistance genes including blaNDM-1, msr(E) and floR, and retains the intact machinery for conjugative transfer (Fig. 3), suggesting that it is responsible for the transmission of these resistance genes into the PASGNDM genomes (25,38). However, the conjugative transfer of ICETn43716385 was not observed in mating experiments between PASGNDM699 and several recipient P. aeruginosa strains under laboratory conditions (data not shown). It is possible that the conjugation has extremely low efficiencies that is below our detection limit or requires specific inductive conditions (38, 39). Our findings provide the first evidence that blaNDM-1 is encoded and potentially disseminated by ICEs, together with msr(E) and floR.
The macrolide resistance gene msr(E) and its variants were mainly identified from Grampositive bacteria and several Gram-negative bacterial species such as Enterobacteriaceae spp., Pasteurella multocida and A. baumannii, but has been rarely described in P. aeruginosa (29, 40-42). Hence, the transmission of msr(E) by ICETn43716385 into the clinical isolates characterized in this study has led us to investigate its possible functional roles in P. aeruginosa. Our in vitro assays clearly showed that the exogenous expression of msr(E) in P. aeruginosa could abolish the QS inhibition activity of AZM in vitro, which in turn rescued the inhibition of elastase production, rhamnolipid production and swarming motility by AZM treatment (Fig. 4, 5). In addition, the acquisition of msr(E) by PAO1 almost completely restored the AZM-affected transcriptome and abolished the anti-Pseudomonas activity of AZM in a murine infection model (Fig. 6, Fig. 7, Table S5, Table S6). These results demonstrate that Msr(E) could confer P. aeruginosa with resistance to AZM-mediated QS inhibition and that the transmission of msr(E) into this organism will greatly challenge the use of AZM in treating infections caused by P. aeruginosa.
The third antibiotic resistance gene carried by ICETn43716385 is floR, which encodes an efflux pump and is the determinant for florfenicol resistance (43). This gene was first identified in Salmonella typhimurium DT104 in 1999, and has ever since been detected in E. coli and P. multocida isolated from livestock and aquaculture settings (44-47). It is possible that the emergence and spreading of floR is due to the selection by florfenicol, since this antibiotic has been extensively used in veterinary medicine, especially in aquaculture settings (46, 48). Therefore, the carriage of floR by ICETn43716385 may confer its hosts with advantages under environmental conditions where florfenicol may be present (48, 49), and hence facilitate the transmission and dissemination of blaNDM-1 and msr(E).
In conclusion, we have characterized the full-length genomes of two NDM-1-producing P.
aeruginosa clinical isolates, from which we identified a novel ICETn43716385 element encoding three antibiotic resistance genes. Among them, the blaNDM-1 gene is responsible for extreme resistance to anti-Pseudomonal carbapenems, whereas msr(E) can quench AZMmediated QS inhibition, and floR may enhance the survival of host bacteria under florfenicol exposure. We present the first evidence of blaNDM-1 to be carried by ICEs, and the first evidence of co-transfer of carbapenem resistance and the resistance to macrolide-mediated QS inhibition into P. aeruginosa. Our findings highlight the importance of ICEs in transmitting antibiotic resistance, and that anti-virulence treatment of P. aeruginosa infections by targeting QS can be challenged by horizontal transfer of resistance mechanisms.

Materials and Methods
Bacterial strains and plasmids All the strains were routinely grown in Lysogeny broth (LB) or on LB plates with 1.5% agar at 37℃. The msr(E) gene with its putative promoter were amplified from PASGNDM699 genome using primers 5'ACCGGCCAAGATAGTTGACG3' and 5'AGGAAGTTCAACCGCCCTTT3' and ligated into the smaI site of pUCP18 vector (50) to construct pUCP18::msr(E) plasmid, where the msr(E) gene is upstream to the lac promoter. Both pUCP18 and pUCP18::msr(E) were transformed into PAO1 and PA14 by electroporation, and the strains carrying these two plasmids were grown in LB supplemented with 300 μg/ml of carbenicillin (MP Bio) to maintain the plasmids.
Sequencing, assembly, and annotation Genomic DNA was purified using Blood and Cell Culture DNA Midi Kit (Qiagen) and sequenced on an Illumina MiSeq platform or a PacBio RS II system. The full-length genomes of PASGNDM345 and PASGNDM699 were assembled from long reads obtained from the PacBio RS II system using HGAP2 pipeline assisted with manual curation. The ICETn43716385 sequence and ICE-like element from P. fluorescens UK4 were uploaded to the Rapid Annotations using Subsystem Technology (RAST) server for gene prediction and annotation, assisted with manual curation (51).
Phylogenetic and comparative genomic analysis Phylogenetic analysis was carried out using Parsnp v1.1.2 (53). The phylogenetic tree was constructed using the approximate maximum likelihood algorithm based on core genome alignment, with filtration of recombination sites. Genomic islands of PASGNDM699 and PASGNDM345 genomes were predicted by the IslandViewer 3 server (54), and the encoded antibiotic resistance genes were predicted using the ResFinder 2.1 server (55). Comparison of PASGNDM699 and PASGNDM345 genomes with six other full-length P. aeruginosa genomes was done by BLASTn search using BLAST Ring Image Generator v0.95 (56), with genomic islands and antibiotic resistance genes plotted in the same image.

Quantification of elastase, rhamnolipid and swarming motility
Bacterial strains were grown in ABT minimal medium supplemented with 5 g l -1 glucose and 2 g l -1 casamino acids (ABTGC), with or without addition of 8 μg/ml AZM (Spectrum). The supernatants of overnight culture were filter sterilised before quantification of elastase and rhamnolipid production. Elastase activity was measured using EnzCheck Elastase Assay Kit (Thermo Fisher Scientific) as instructed by the manual. The quantification of rhamnolipid was performed as previously described with modifications (57). Briefly, rhamnolipid was extracted from supernatants with diethyl esther and dissolved in deionised water. The solution was added with 0.19% (w/v) orcinol dissolved in 50% H2SO4, followed by heating at 80℃ for 30 min. The elastase activity and rhamnolipid were quantified by measuring emission at 530 nm and absorbance at 460 nm by using an Infinite 200 PRO system (Tecan), respectively. Both experiments were performed in at least triplicates, and the results were normalised with optical density at 600 nm (OD600) and shown as mean ± standard deviation in the figures. The swarming motility of each strain was measured on 0.5% agar plates containing 8 g/L nutrient broth (Oxoid) and 5 g/L glucose (58). One μl of bacterial overnight culture (adjusted to OD600 = 1) was inoculated to the center of the plate, followed by incubation at 37℃ for 12 hours. The images were taken by a Gel Doc XR+ System (Bio-Rad).
Transcriptomic analysis Strains were grown in ABTGC medium at 37℃ with 200 rpm shaking. Cells were harvested at mid-log phase (OD600 = 0.6), and the total RNA was extracted using RNeasy Mini Kit (Qiagen). RNA sequencing was performed on an Illumina HiSeq 2000 platform to generate paired-end reads of 100 nt, which were mapped to PAO1 genome (NC_002516.2) by using CLC Genomics Workbench 9.0 (Qiagen). The transcript count table was analysed by DESeq 2 package of the R/Bioconductor by performing negative binomial test (59). Hierarchical clustering analysis was performed to produce the heatmap for the differentially expressed genes with statistical significance (fold change > 4, P <0.05) using heatmap.2 package of the R/Bioconductor (60).
Animal model A murine silicone implant model was used to evaluate effects of AZM treatment in vivo on the two strains as previously described with modifications (36). Briefly, washed bacterial cells were re-suspended in 0.9% NaCl to an OD600 of 0.01. Bacteria cells were allowed to attach onto sterilize silicone tubes (length, 3 mm; inner diameter, 4 mm; outer diameter, 6 mm; Ole Dich) by incubation at 37℃, with 110 rpm shaking for 18 hours, after which the silicone tubes coated with bacteria were implanted into mouse peritoneal cavity by surgery. Treatment with AZM (10 mg per mouse, dissolved in 0.2 ml of 0.9% saline) or saline was performed by injection into the mouse peritoneal cavity 12 hours after surgery. Mice were sacraficed 24 after surgery and the bacterial cells residing on the silicone tube and in the mouse spleen were dispersed into the 0.9% NaCl solution by sonication using an Elmasonic P120H (Elma, Germany; power=50% and frequency=37 KHz) and homogenisation using a Bio-Gen PRO200 Homogenizer (Pro Scientific), respectively. The CFU was quantified by serial dilution and plating on LB agar plates, and the results were shown in log CFU with mean ± standard deviation.    Table S7.  Table S5, and GIs are listed in Table S3 and S4.    Table S4 and Table S5.  PASGNDM strains using Progressive Mauve (26). Alignment results showed the 11 strains are highly similar.