ABSTRACT
Pseudomonas aeruginosa is a prevalent and pernicious pathogen equipped with extraordinary capabilities both to infect the host and to develop antimicrobial resistance (AMR). Monitoring the emergence of AMR high-risk clones and understanding the interplay of their pathogenicity and antibiotic resistance is of paramount importance to avoid resistance dissemination and to control P. aeruginosa infections. In this study, we report the identification of a multidrug-resistant (MDR) P. aeruginosa strain PA154197 isolated from a blood stream infection in Hong Kong. PA154197 belongs to a distinctive MLST550 clonal complex shared by two other international P. aeruginosa isolates VW0289 and AUS544. Comparative genome and transcriptome analysis of PA154197 with the reference strain PAO1 led to the identification of a variety of genetic variations in antibiotic resistance genes and the hyperexpression of three multidrug efflux pumps MexAB-OprM, MexEF-OprN, and MexGHI-OpmD in PA154197. Unexpectedly, the strain does not display a metabolic cost and a compromised virulence compared to PAO1. Characterizing its various physiological and virulence traits demonstrated that PA154197 produces a substantially higher level of the P. aeruginosa major virulence factor pyocyanin (PYO) than PAO1, but it produces a decreased level of pyoverdine and displays decreased biofilm formation compared with PAO1. Further analysis revealed that the secondary quorum-sensing (QS) system Pqs that primarily controls the PYO production is hyperactive in PA154197 independent of the master QS systems Las and Rhl. Together, these investigations disclose a unique, uncoupled QS mediated pathoadaptation mechanism in clinical P. aeruginosa which may account for the high pathogenic potentials and antibiotic resistance in the MDR isolate PA154197.
INTRODUCTION
Pseudomonas aeruginosa is a ubiquitous Gram-negative pathogen that causes a variety of notorious infections in humans such as ventilator-associated pneumonia, lung infections of cystic fibrosis (CF) patients, burn wound infection, and various sepsis syndromes. It is the second leading cause of hospital-acquired infections and is especially problematic in intensive care units, where it is the leading cause of pneumonia among pediatric patients and is responsible for a large number of urinary tract (10% in the United States and 19% in Europe), bloodstream (3% in United States and 10% in Europe), eye, ear, nose, and throat infections (1–3). Compounding the burden of these infections is the extraordinary capability of the pathogen to develop antibiotic resistance and multidrug resistance (MDR) even during the course of antibiotic therapy. P. aeruginosa is one of the “ESKAPE” (Enterococcus spp., Staphylococcus aureus, Klebsiella spp., Acinetobacter baumannii, P. aeruginosa, and Enterobacter spp.) organisms that are recognized by the Infectious Diseases Society of America as an alarming threat to the global public heath associated with antimicrobial resistance (AMR). As a consequence, the diseases outcome of the P. aeruginosa infections is the complex interplay of the pathogen (its pathogenicity and virulence), hospital environments (antibiotic therapies and the emergence of AMR), and the patient’s conditions (host immune responses) (4–6).
P. aeruginosa is genetically equipped with outstanding intrinsic antibiotic resistance machineries. These include the inducible production of the AmpC cephalosporinase, presence of the housekeeping MexAB-OprM multidrug efflux pump, and the limited permeability of its outer membrane caused by the low expression and narrow substrate specificity of porin proteins (4, 7, 8). In addition, the pathogen has extraordinary capabilities of developing acquired antibiotic resistance. Mutational overexpression of one of at least four efflux pumps, MexAB-OprM, MexCD-OprJ, MexEF-OprN, and MexXY-OprM, encoded in the genome of P. aeruginosa often plays an important role in acquired resistance and can lead to clinically significant MDR (9, 10). Among them, the MexAB-OprM pump displays the broadest substrate profile, and mutational overexpression of this pump can lead to resistance to all β-lactams (except imipenem), (fluoro)quinolones, tetracyclines, and macrolides in clinics. Similar to the MexAB-OprM pump, overexpression of MexXY is common (10 to 30%) among clinical strains and causes decreased susceptibility to aminoglycosides and cefepime. Overexpression of MexEF-OprN and MexCD-OprJ is less common in clinical isolates (<5%), and these two efflux pumps mainly affect fluoroquinolone resistance (5, 11). In addition to efflux pump overexpression-mediated MDR, P. aeruginosa also readily develops resistance to specific class of antibiotics through genetic mutations and the acquisition of target or drug inactivation genes. These include overexpression of ampC caused by mutations in peptidoglycan-recycling genes ampD, dacB, or ampR that cause resistance to noncarbapenem β-lactams; the acquisition of aminoglycoside modification enzymes (AMEs) and ribosomal methyltransferase (Rmts) enzymes that are associated with aminoglycoside resistance; mutations in the DNA gyrase (gyrA and gyrB) and/or topoisomerase IV (parC and parE), which result in fluoroquinolone resistance; and mutational repression of the OprD porin, which leads to resistance to imipenem (12–15).
Regardless of the resistance mechanisms involved, the emergence and prevalence of MDR P. aeruginosa strains continue to rise rapidly worldwide. Although the bacterium displays an overall nonclonal epidemic population structure with most isolates represented by single multilocus sequence typing (MLST) genotypes, MDR or XDR (extensively drug-resistant) P. aeruginosa isolates display a much lower clonal diversity than do the susceptible isolates, and recent studies have reported the existence of MDR/XDR global clones disseminated in different hospitals worldwide (11, 16, 17). Interpatient spread of antibiotic-resistant mutations linked to the transmission of epidemic CF strains has also been reported. These clones are denominated as international high-risk clones. Among them, the most recognized successful clones are the high-risk MDR clones ST111, ST175, and ST235 and the Liverpool Epidemic Strain (LES) ST146, which is the most epidemic clone among CF patients (11). Further deeper analysis of the molecular details of their resistance development and close monitoring of the emerging of new international MDR/XDR high-risk clones are of paramount importance to avoid the worldwide dissemination of these clones and to control P. aeruginosa infections.
In addition to their broad and high level of antibiotic resistance, many MDR clones, especially the three major international high-risk clones ST111, ST175, and ST235, are associated with a defined set of biological markers that include defective motility (swimming, swarming, and twitching), reduced pigment production (pyocyanin and pyoverdine), and reduced in vitro fitness, but enhanced biofilm formation (5, 11). These suggest the presence of a fitness tradeoff that compromises the pathogenic potentials and virulence of MDR isolates. Several epidemiological survey and molecular evolutionary studies ascribed these phenotypes to quorum-sensing (QS) deficiency in these strains (18–21), which is not unexpected since the production of many virulence factors, such as exoproteases, elastase, rhamnolipids, lectin, pyoverdine, pyocyanin, and hydrogen cyanide, etc., is primarily regulated by QS systems in P. aeruginosa (1). On the other hand, recent studies also reported antibiotic-resistant strains displaying enhanced virulence (17, 22–24), such as P. aeruginosa strains lacking the oprD porin (22), implying that certain MDR clones may develop compensatory mutations, allowing them to recover their virulence without affecting the level of resistance.
Recently, we surveyed 84 P. aeruginosa clinical strains isolated from various types of infections, including wound (both superficial and deep), urine, ear, pus, drain fluid, and blood infections, at the Queen Mary Hospital, Hong Kong, and identified the MDR isolate PA154197 from a bloodstream infection. PA154197 displays an MDR level and profile comparable to the international high-risk clone ST175. Sequencing its genome reveals that it belongs to a distinctive genotype MLST550 that is shared by two additional clinical strains currently available in the database: VW0289 and AUS544. VW0289 and AUS544 were isolated from the sputum of a CF patient in The Netherlands and the bronchial lavage fluid of a CF patient in Australia, respectively. The clonal lineage of PA154197 with these two strains suggests a potential international transmission of this emerging clonal complex. Yet, the two CF isolates VW0289 and AUS544 are poorly characterized, and little is known about the ST550 clonal lineages. This prompted us to use PA154197 as a prototype to conduct comparative genomic and transcriptomic analysis and understand the resistance development and virulence of the MLST550. Our investigations led to the identification of the adaptive activation of a secondary QS system, Pqs (Pseudomonas quinolone system), independent of the primary QS systems Las and Rhl, which may account for the uncompromised virulence in the MDR strain PA154197. Together, our studies describe the emergence of a high-risk clone in clinics and its underlying pathoadapation mechanism mediated by uncoupled QS.
RESULTS
PA154197 displays extraordinary antibiotic resistance.We examined the susceptibility of PA154197 to various classes of antibiotics, including the common antipseudomonal drugs aztreonam, imipenem, ciprofloxacin, levofloxacin, and amikacin, etc., by measuring their MIC values and comparing them to those for PAO1 (Fig. S1A). We found that PA154197 displays resistance to almost all these different classes of antibiotics. The only two classes of antibiotics to which it remains susceptible are the aminoglycosides (amikacin and gentamicin) and the cyclic nonribosomal polypeptides (polymyxin B and E), which meet the XDR criteria (resistance to all but one or two of the eight classes of antipseudomonal drugs) according to Magiorakos et al. (16). To further evaluate its resistance level, we compared the antibiotics susceptibility profile of PA154197 with the three international high-risk clones using the published data (25) and found that PA154197 displays a comparable resistance profile with the high-risk clone ST175 (Fig. S1B), highlighting the high epidemic and risk potential of PA154197.
Sequencing type, phylogenetic position, and genomic islands of PA154197.We then sequenced the genome of PA154197 which resulted in one single circular chromosome with a length of 6,445,239 bp and a GC content of 66.38%. A total number of 5,923 genes are predicted in PA154197 genome; these include 5,816 coding gene sequences (CDS), 26 pseudogenes, 12 rRNAs, 65 tRNAs, and 4 ncRNAs (Table 1). Based on PubMLST, PA154197 is assigned to MLST 550, a clonal complex which is currently shared by two additional clinical isolates, VW0289 and AUS544 from the CF patients in the Netherland and Australia, respectively (Fig. S2A), but the complete genome sequences of the two strains are unavailable. Past analysis revealed that PA154197 belongs to serotype O9 (26) (Table 1). Analyzing its accessory genome identified 21 genome islands (GIs), including three prophages (Table S1). In addition to hypothetical proteins, a large number of metabolic enzymes, such as oxidoreductase, oxygenase, and halogenase, as well as the AraC and LysR type transcription regulators, are encoded in these GIs, implying the potential roles of the accessory genome of PA154197 in the adaptation and fitness of the bacterium.
Genomic features of P. aeruginosa PA154197
To examine the phylogenetic relationship of PA154197 with other P. aeruginosa strains, we constructed a phylogenetic tree based on 140,634 single nucleotide polymorphisms (SNPs) generated from 150 P. aeruginosa genomes, which include all publicly available complete genomes (78 as of October 2017) and representative noncomplete genomes. As shown in Fig. S2B, PA154197 exhibits a distinct phylogenetic position from other genomes available in the database. This may indicate its unique virulence and pathogenic potentials.
Antibiotic resistance genes and mutational changes therein.Using resistance gene identifiers (RGIs), we then identified antibiotic resistance genes in PA154197 against the Comprehensive Antibiotic Resistance Database (CARD) (27) and examined genomic variations, including mutations, deletions, and indels in these genes, using the genome of PAO1 as a reference (Table 2). It was revealed that PA154197 contains a large number of mutations associated with antibiotic resistance genes, including both previously reported genetic variations and those newly identified in PA154197. Several well-established mutations include an 8-bp deletion in the mexT gene, which leads to upregulation of the MexEF-OprN efflux pump and downregulation of the OprD porin, and consequently to resistance to fluoroquinolones and imipenem, respectively (28–30); the T105-A substitution in AmpC, which causes resistance to noncarbapenem β-lactams (31); and the T83-I substitution in the so-called quinolone resistance-determining region (QRDR) of GyrA, conferring quinolone resistance (32). Another genomic variation commonly found to cause resistance in many P. aeruginosa isolates is the nonsynonymous mutations in the TetR transcriptional regulator NalC, which often leads to aztreonam resistance (33). Three nonsynonymous mutations—G71-E, E153-Q, and S209-R—are identified in the nalC gene in PA154197 (Table 2).
Mutational changes of genes conferring antibiotic resistance and their expression levels in PA154197 relative to that in PAO1a
New genomic variations which have not been reported in other strains but potentially lead to antibiotic resistance is also identified in PA1541197, such as a 6-bp deletion (corresponding to the deletion of E959 and S969) in gyrA and an E76* (stop codon) mutation in the mexR gene, which will lead to premature termination of MexR at E76 and potentially to derepression of the MexAB-OprM major efflux pump, causing multidrug resistance (34). Mutations that potentially derepress or induce the expression of several other efflux pumps are also identified in the corresponding transcription regulators or the promoter region of the efflux genes. For example, a single amino acid substitution, H398R, is identified in the parS gene, which regulates the expression of the mexEF-oprN efflux pump (35). A 5-nucleotide alteration in the promoter region (bp: −250 to −1) of the mexGHI-opmD operon is present in PA154197 compared to that in PAO1, which may lead to the expression change of the MexGHI-OpmD efflux pump. A complete summary of the identified genomic variations in all the antibiotic resistance genes is provided in Table 2.
Expression of the resistance genes and the functional genome of P. aeruginosa PA154197.To further examine the resistance mechanisms of PA154197, we performed a comparative transcriptome analysis on PA154197 and the reference strain PAO1 using transcriptome sequencing (RNA-seq). Of the 5,543 orthologous genes identified in the two strains using progressiveMauve (36), differential expression of 3,148 genes was observed, and their expression levels were compared using DESeq (37) (Fig. 1 and Table S2). Notably, expression of four efflux pump gene operons, mexEF-oprN (PA2493-PA2495), mexRAB-oprM (PA0424-PA0427), mexGHI-opmD (PA4205-PA4208), and mexKJL (PA3676-PA3678), were found to be significantly higher in PA154197 than in PAO1 (Fig. 1 and Fig. 2A), consistent with the MDR profile of PA154197. Genes that confer resistance to specific antibiotics, such as ampC, gyrA, gyrB, and parC and parE, were not differentially expressed in the RNA-seq analysis, suggesting that the overexpression of several multidrug efflux pumps may play a major role in MDR development in PA154197. COG functional distribution analysis of all differentially expressed genes also revealed the enrichment of efflux pump genes expressed at a higher level in PA154197 than in PAO1 (Fig. S3).
Genomewide transcriptomic profile of P. aeruginosa PA154197 and PAO1. Orange dots indicate genes with higher relative expression levels in PA154197 than in PAO1, and blue dots represent genes with higher relative expression levels in PAO1 than in PA154197. The black dashed lines represent 4-fold (value of log2 > 2) changes in expression. Genes and operons with distinctive expression patterns in the two strains are indicated. Among them, antibiotic resistance genes are highlighted in red, and genes encoding virulence factors are highlighted in green (expressed at a higher level in PA154197 than in PAO1) or purple (expressed at a higher level in PAO1 than in PA154197). The x axis represents the gene locus with PAO1 genome as a reference.
Several multidrug efflux pump genes are overexpressed in P. aeruginosa PA154197. (A) RPKM abundance of the major efflux pump genes in PA154197 and PAO1 calculated from the RNA-seq data. (B) RT-qPCR analysis of the expression of selective efflux genes and other genes associated with resistance to specific classes of antibiotics.
To verify the relative expression levels of these genes in the two strains, we conducted reverse transcription-quantitative PCR (RT-qPCR) analysis. As shown in Fig. 2B, consistent with the RNA-seq data, the MexEF-OprN efflux system displayed significantly higher expression (265- to 574-fold) in PA154197 than in PAO1, followed by the MexGHI-OpmD (8.5- to 41.9-fold) and MexAB-OprM (4.6- to 11-fold) efflux systems. On the other hand, the MexXY-OprM efflux system, whose overexpression is associated with aminoglycoside resistance, did not show increased relative expression in PA154197 in either RNA-seq or RT-qPCR analyses, consistent with the fact that PA154197 remains susceptible to the aminoglycoside antipseudomonal drug amikacin and gentamicin (Fig. S1). RT-qPCR analysis also revealed the decreased expression of another resistance gene, oprD, in PA154197, which serves as the entry portal for imipenem (38), consistent with the observed imipenem resistance (MIC as 8 μg/ml) in this strain.
Virulence of P. aeruginosa PA154197.Our comparative transcriptome analysis also reveals a significantly higher expression of several virulence genes in PA154197 than in PAO1 (Fig. 1 and Table S2). These include the type III (psc genes) and type VI (tss and hcp-1 genes) secretion systems, pyocyanin production (phz genes), the Pqs quorum-sensing system (pqs genes), and a series of Fe and heme acquisition genes (fep and has genes). On the other hand, several genes involved in biofilm formation (such as psl, alg, and lecA), motility, pilus and fimbrial assembly (such as cup genes), and pyoverdine production (pvd genes) are expressed at a lower level in PA154197 than in PAO1.
To examine the virulence factors production and the virulence of PA154197, we measured pyocyanin (PYO) and pyoverdine production, biofilm formation, swimming and swarming motilities, and the infection virulence of PA154197 and compared these findings to PAO1. As shown in Fig. 3, PA154197 cells produce a significantly higher level of PYO than PAO1 cells (Fig. 3A), which may be responsible for the uncompromised virulence of the strain as observed in the Caenorhabditis elegans infection assay (Fig. 3E). In contrast, PA154197 displays reduced pyoverdine production compared to PAO1 (Fig. 3B), suggesting that the strain may utilize other components and pathways rather than pyoverdine for iron acquisition (39). Consistent with the lower expression of biofilm genes in PA154197 than in PAO1, PA154197 displays a reduced biofilm formation capability (Fig. 3C). In terms of the motility of the bacterium, PA154197 displays a swimming activity comparable to PAO1, but a defective swarming motility (Fig. 3D), which is consistent with the lower expression of the pilus and fimbrial assembly genes in this strain. To evaluate the in vivo virulence of PA154197, we conducted Caenorhabditis elegans fast-killing and slow-killing assays, which are established infection models to evaluate the cytotoxicity and pathogenicity of P. aeruginosa, respectively (40). We found that the two strains display a comparable cytotoxicity and pathogenicity (virulence) (Fig. 3E) to the C. elegans host. Together, these studies indicate that the MDR strain PA154197 may achieve an uncompromised virulence by overexpressing a subset of virulence factors.
Virulence and production of several virulence factors in PA154197 and PAO1. (A) Pyocyanin production in PA154197 and PAO1 cells cultured in both LB broth and agar. (B) Pyoverdine production examined under UV light. (C) Biofilm formation examined by the crystal violet stain. (D) Swimming and swarming motilities examined in the corresponding specific agar plates. (E) Virulence of PA154197 and PAO1 measured in a C. elegans infection model. N2 C. elegans grown to L4-stage adults were infected with PA154197 or PAO1 for 24 h. The survival of C. elegans was monitored under a light microscope and recorded.
The Pqs quorum-sensing system is activated in PA154197 independent of the primary QS systems Las and Rhl.It is known that three hierarchically organized QS systems, Las, Rhl, and Pqs, regulate the production of an arsenal of virulence factors in P. aeruginosa, with each being primarily associated with the production of the virulence factors elastase (encoded by lasB gene), rhamnolipid (encoded by rhlA gene), and pyocyanin (encoded by phz genes), respectively (41, 42) (Fig. 4A). Among them, the Las system proceeds the Rhl and Pqs systems and is the master regulator of the QS circuit. Several previous epidemiological survey and molecular evolutionary studies have identified lasR-null and mutant variants in clinical isolates and ascribed the attenuated virulence of the isolates to an overall QS deficiency caused by the inactivation of the master regulator Las (1, 42). However, PA154197 does not display an attenuated virulence as most of the resistant strains do, and comparative genomics did not reveal mutations in the QS regulatory genes lasR, rhlR, and pqsR. Nonetheless, to explore whether the uncompromised virulence in PA154197 is related to QS, we examined the RNA-seq transcription levels of the QS systems and genes regulated by each of the QS systems, especially virulence genes. Surprisingly, we found that while expression of the lasRI and rhlRI in PA154197 is undetectable, expression of the genes encoding the secondary QS system pqsA-E is very high and is among the top three most significantly expressed (log2 RPKM [reads per kilobase per million mapped reads] abundance in a range of 4.1 to 5.4) gene operon in PA154197 than in PAO1 (Fig. 4A). Consistently, genes primarily controlled by the Las and Rhl systems, such as toxA, lasB, lecA are expressed in a lower level in PA154197 than in PAO1, whereas genes controlled by Pqs, such as phZ1 and phZ2, pqsA-E, and phnAB, are expressed in a significantly higher level in PA154197 than in PAO1 (Fig. 4A and Table S2). RT-qPCR analysis confirmed this observation (Fig. 4B). These data suggest that the secondary QS system Pqs is activated and expressed independent of the primary QS systems Las and Rhl in PA154197, which may account for the hyperproduction of a subset of virulence factors and uncompromised virulence of the strain.
Hyperactivation of the Pqs quorum-sensing system independent of the master systems Las and Rhl in PA154197. (A) Schematic diagram of the three quorum-sensing systems in P. aeruginosa, Las, Rhl, and Pqs, as well as their selective regulons. The relative expression of the genes in the Las, Rhl, and Pqs regulons is depicted in a color scheme with the genes expressed at a lower level in PA154197 than in PAO1 in red and those expressed at a higher level in PA154197 than in PAO1 in blue. The RPKM abundance of the genes calculated from the RNA-seq data are indicated. (B) RT-qPCR analysis of selective genes belonging to the Las, Rhl, and Pqs regulons, respectively.
DISCUSSION
As a successful and ubiquitous pathogen, P. aeruginosa is equipped with extraordinary machineries to adapt to the host environments and antibiotic therapies to survive and disseminate. As a result, the disease development caused by P. aeruginosa infections is driven by several dynamic factors, including bacterial pathogenesis, selective forces resulting from the antimicrobial interventions, and the fitness costs of resistance development. It is generally recognized that while acquisition of antibiotic resistance mechanisms confers selective advantages in the presence of antimicrobial therapies, expressing and maintenance of the resistance determinants often incurs metabolic costs to the pathogen and consequently compromises its fitness and pathogenicity potentials (5). How P. aeruginosa reconciles this scenario and succeeds in the dynamic cycle of infection and dissemination is not known. In this study, we identify and utilize an MDR P. aeruginosa isolate PA154197 that displays an MDR profile comparable to the epidemic high-risk clone ST175 but does not exhibit compromised virulence as a paradigm for examining the underlying pathoadaptive mechanisms. Comparative transcriptome and RT-qPCR analysis reveals an uncoordinated, significant activation of the Pqs quorum-sensing system and its regulated virulence genes in PA154197 independent of the primary QS Las and Rhl systems, providing a compensatory mechanism for virulence factor production without the activation of the master and primary QS systems Las and Rhl, as well as the hundreds of genes regulated by them. A schematic diagram to illustrate this pathoadaptive mechanism is shown in Fig. 5.
Schematic model to show the pathoadaptation mechanism adopted by PA154197 to modulate its antibiotic resistance and virulence. In the high virulent, susceptible strains, the three quorum-sensing (QS) systems Las, Rhl, and Pqs mediate the production of a variety of virulence factors and a high virulence in the cell. Antibiotic-resistant strains with acquired resistance determinants often carry mutations in the master QS regulator Las to downregulate the QS system and virulence. Strains with a pathoadaptive strategy such as PA154197 uncouple the activation of the secondary QS system Pqs from Las and Rhl to maintain an uncompromised virulence with simultaneous production of its resistance traits.
Several previous studies reported QS deficiency in antibiotic resistant clinical isolates and attributed the attenuated virulence of the strains to this mechanism, since these mutants are proposed to be social cheaters that exploit shared QS products without incurring metabolic costs to themselves (18–21). However, majority of these studies were based on comparative genomic analysis to identify mutations in the QS system genes, and the identified variations were mainly found in the master QS regulator LasR, including both lasR-null and various lasR point mutations. Consistently, the identified strains often display an overall QS inactivation phenotype, such as decreased production of all the major QS-regulated virulence factors, including elastase, rhamnolipids, pyoverdine, and pyocyanin. In contrast to these lasR-defective isolates, PA154197 produces a significantly larger amount of PYO, a major virulence factor secreted by P. aeruginosa (43) and other Pqs-regulated gene products, but lower levels of Las- and Rhl-regulated virulence factors and genes than does PAO1. This uncoupled activation of the Pqs system ensures the production of a major virulence factor PYO with a minimal metabolic burden and thus is beneficial to the fitness and virulence of the pathogen. How P. aeruginosa achieves this differential activation is not known currently. We compared the promoter regions of the lasRI, rhlRI, pqsR, and pqsA-E operons with the corresponding sequences in PAO1 but did not find any nucleotide alternations that may account for the differential activation of pqs genes. Comparative genomics analysis revealed a point mutation Q98P in PA154197 LasR which is located at the α5 position of the N-3-oxo-dodecanoyl-homoserine lactone binding domain of the protein, but the residue is not located in the binding pocket (44). Whether this point mutation leads to differential activation of the Pqs system warrants further investigation. Notably, Las-independent activation of a downstream QS system was also reported in a collection of P. aeruginosa isolates from CF patients in which half of the lasR-null strains were found to retain the RhlR activity (20). Hence, an adaptable QS hierarchy that uncouples a downstream QS system from the master regulator LasR may represent an emerging compensatory mechanism that facilitates the fitness, virulence, and persistence of P. aeruginosa in a host setting. In PA154197, at least one downstream physiological process mediated by the uncoupled QS, the production of the virulence factor PYO, is observed. This process might be further facilitated in PA154197 by the overproduction of the MexGHI-OpmD efflux pump (Fig. 2B) in the strain, which has been shown to dedicate to the export of 5-methylphenazine-1-carboxylate, an intermediate in the PYO biosynthetic pathway, and promote PYO production (45).
With the increasing frequency of detecting the MDR P. aeruginosa isolates from various sources, it is recognized that the resistance traits of a strain are often the result of a complex interaction of several cellular processes, and no individual mutation or resistance gene is sufficient to confer clinically significant resistance (46, 47). In this study, we identified multiple genetic variations potentially associated with the resistance development in PA154197. These include both machineries which confer resistance to a diverse class of antibiotics and hence cause MDR (4), such as the overexpression of multidrug efflux pumps, and genes which, once mutated, result in resistance to specific classes of antibiotics, such as mutations in ampC and gyrA. Transcriptome and RT-qPCR confirmed the hyperexpression of three multidrug efflux pumps, MexAB, MexEF, and MexGHI, in PA154197, but the expression of ampC and gyrA was found to be at a level similar to that in PAO1. Whether these two genes contribute to the profile and level of resistance exhibited in PA154197 requires further molecular validation. Notably, both the T105A mutation in AmpC and the T83I mutation in GyrA are the well-characterized genetic variants associated with antibiotic resistance, which have been termed the PDC-3 type AmpC variant (31) and the QRDR hot spot mutation, respectively (14). However, these two mutations are also identified in other P. aeruginosa isolates we have surveyed, such as PA150577, which does not display resistance to β-lactams or fluoroquinolones (data not shown). Hence, the observed resistance to antipseudomonal aztreonam (monobactam) and imipenem (carbapenem) in PA154197 may not be due to the T105A mutation in AmpC, and the observed resistance to ciprofloxacin and levofloxacin (fluoroquinolone) may not be due to the T83I mutation in GyrA. In addition, acquiring additional and extended-spectrum β-lactamase on mobile genetic elements represents another common mechanism of β-lactam resistance in P. aeruginosa (48), but no plasmid encoding β-lactamases has been identified in PA154197. Indeed, it has been proposed that the genetic background of the drug-resistant strains influences the epistatic interactions of the various resistant determinants and the resistance readout (49). This highlights the intricacy of the mechanisms underlying the resistance development of MDR strains. Our ongoing targeted molecular investigations and evolutionary trajectory analysis may provide mechanistic insight into these processes.
MATERIALS AND METHODS
MIC measurements.MICs were measured according to the standard protocol from the American Society for Microbiology, with slight modification (50). Single fresh colonies of PA154197 and PAO1 were inoculated into Mueller-Hinton (MH) broth overnight at 37°C with 220-rpm agitation. The resulting cell culture was diluted and distributed to the wells of a 96-well plate at a final cell density of 5 × 105 CFU/ml. Selected antibiotics were added to the wells, with concentration, ranging from 0.25 to 128 μg/ml. Plates were incubated at 37°C for 16 to 20 h. MIC values were determined as the lowest concentration of antibiotics that completely inhibit growth of bacteria as detected by the unaided eye. Antibiotic susceptibility profiles of the strains indicated are displayed in color scheme with low susceptibility (high MIC values) in red color and high susceptibility (low MIC values) in blue color. The color scheme is constructed using background filling application in the Excel of MS Office.
Genomic DNA extraction.Extraction of the genomic DNA of PA154197 was performed following the description in a previous study (51). Briefly, PA154197 was cultivated in Luria-Bertani (LB) broth overnight with shaking (220 rpm) at 37°C. Bacterial cells were harvested from a 1-ml culture via centrifugation at 10,000 rpm for 10 min. Genomic DNA was extracted using the QIAamp DNA minikit according to the manufacturer’s instructions (Qiagen, Hilden, Germany). The concentration and quality of genomic DNA was determined by NanoDrop and agarose (0.8%) gel electrophoresis.
Genome sequencing and annotations.Genome sequencing of P. aeruginosa PA154197 was conducted on the Illumina NextSeq (300 Cycles) PE150 High Output Flow Cell platform in Georgia Genomics Facility at the University of Georgia. SPAdes (52) was used to assemble the reads after the removal of adapter, primers, and low-quality bases using Trimmomatic (53). The initial assembly of PA154197 genome yielded 21 contigs. The 21 contigs were aligned against the reference genomes of PAO1 and ATCC 27853, and gaps were filled using Sanger sequencing to generate the complete genome sequence of PA154197. Gene calling and annotation were carried out using the National Center for Biotechnology Information Prokaryotic Genome Annotation Pipeline 2.0 (PGAP) (54).
Sequence type and phylogenetic analyses.The MLST and serotype of PA154197 was predicted based on the PubMLST database (https://pubmlst.org/) and the Pseudomonas aeruginosa serotyper (PAst) tool (26).
To carry out the phylogenetic analysis, we downloaded all available complete genomes of P. aeruginosa (one representative per distinct phylogenetic group) from NCBI GenBank (78 genomes by October 2017) (see Table S3) and selected 72 representative noncomplete genomes that are distributed in the 32 phylogenetic groups as defined in the NCBI. SNPs were collected using Parsnp with default parameters (55), and that of PAO1 served as the reference. SNPs were then used to build a maximum-likelihood tree in MEGA (56) with the following parameters: Tamura-Nei substitution model, gamma rate distribution among sites, nearest neighbor interchange for tree inference options, a bootstrap value 100, and an initial tree was generated using the neighbor-joining method. Variants were called using the Harvest tools (55) and annotated using SnpEff (57).
Identifications of antibiotic resistance genes and mutations therein.RGIs from CARD (27) were used to identify antibiotic resistance genes in PA154197 using the “perfect and strict hits only” parameter. BLAST analyses of all genes from the present study against the antibiotic resistance gene database of ResFinder (v2.1) (58) and BLAST analyses against a database of antibiotic resistance genes curated by ourselves based on previous reports (59) were also conducted.
Variants of all predicted antibiotic resistance genes were collected from the annotation with SnpEff (57) using PAO1 as the reference. SNPs that cause nonsynonymous mutations and gaps of <6 bp were also identified and summarized.
RNA-seq.RNA extraction, quality control, and RNA-seq were performed in PA154197 and the reference strain PAO1 with three biological replicates according to our previous descriptions (51). Stranded libraries for all RNA samples were constructed using Kapa Biosystems RNA library preparation chemistry in Georgia Genomics Facility at University of Georgia.
Orthologous genes between PAO1 and PA154197 were obtained by comparison using progressiveMauve with default settings (36) and were used in the following RNA-seq analysis. RNA-seq reads were preprocessed for quality control using Trimmomatic (53) and then mapped to the reference genomes of PAO1 and PA154197, respectively, using Stampy (to speed up the alignment process, the alignment was initially conducted using BWA-MEM with default parameters [35] and then passed to Stampy with the flags –bamkeepgoodreads –M) (60). SAMtools and BamTools were used for format conversions, statistics, and quality assessment and control (61). The Integrative Genomics Viewer was used to visually inspect mapping quality (62). Fragments counting per genomic features (genes) were performed using featureCounts (featureCounts -R -M -Q 10 -p -P -s 2 -t gene -g locus_tag –largestOverlap) (63). Reads that mapped with MAPQ (map quality) scores below 10 were removed. Enforcing a MAPQ score below 10 also excludes multimapped reads, albeit the percentage of this category is low (data not shown). DESeq was used to analyze differentially expressed genes. Selective genes with high expression levels in PA154197 were verified using RT-qPCR. Primers used in the present study are listed in Table S4.
Motility assays.The assays were performed as previously described with slight modifications (64) for (i) swimming motility and (ii) swimming motility. For swimming motility, semisolid motility plates were prepared by mixing the LB broth with 0.25% (wt/vol) agarose. Overnight cultures of PAO1 and PA154197 were subcultured (1:200 dilution) in LB broth. When to optical density at 600 nm (OD600) reached 0.1 to 0.2, a 2-μl sample of the cells was spotted at the center of a freshly prepared semisolid swimming plate. The plates were dried at room temperature for 1 h and subsequently incubated at 37°C for 8 h. Swimming motility was evaluated by measuring the diameter of the covered areas. All assays were conducted in triplicates. For the swarming motility, motility plates were prepared by mixing M8 broth (21.1 mM Na2HPO4, 11 mM KH2PO4, 42.8 mM NaCl, 9.3 mM NH4Cl, 1 mM MgSO4, 0.2% glucose, 0.2% Casamino Acids [pH 6.5]) with 0.5% (wt/vol) agarose. Then, 2-μl portions of subcultures of PAO1 and PA154197 (OD600 of 0.1 to 0.2), as mentioned above, were spotted at the center of a predried swarming plate. The plates were incubated at 37°C, and images of the developed swarms were recorded at 16 to 48 h. All assays were conducted in triplicates.
Pyocyanin production assay.Pyocyanin production was examined according to a protocol described previously with modification (51). A 1-ml portion of the bacterial culture was subject to centrifugation at 13,000 rpm for 5 min. The supernatant was collected and extracted with 600 μl of chloroform following vortexing for 10 s twice. After centrifugation at 13,000 rpm for 5 min, the chloroform phase was transferred into a clean tube and subsequently mixed with 0.5 ml 0.2 M HCl, followed by gentle shaking to transfer the pyocyanin to the aqueous phase. The concentration of PYO was determined by measuring the absorbance of the aqueous phase at 510 nm.
Biofilm assay.A crystal violet staining method was used to determine the biofilm formation of PAO1 and PA154197, with slight modifications (65). Overnight cultures of P. aeruginosa cells were inoculated into 0.5 ml of LB broth in 5-ml round-bottom polypropylene Falcon tubes (final cell density, 0.1) and incubated statically at 37°C for 24 or 48 h to allow biofilm formation. After removing the medium, the biofilms formed at the bottom of the tubes were gently washed with phosphate-buffered saline three times to remove residual planktonic cells. The tubes were then air dried and stained with 0.1% crystal violet for 15 min at room temperature. After the tubes were washed three times with sterile distilled water to remove excess dye molecules, biofilms were dissolved in 150 μl of 30% acetic acid, and the OD595 was measured using a 96-well plate spectrometer reader. All experiments were performed in triplicates.
Nematode killing assay.To examine the virulence and pathogenesis of PA154197 compared to PAO1, fast- and slow-killing kinetics of Caenorhabditis elegans were performed with slight modifications (66). (1) For the fast-killing assay, the preparation of the killing plates can be divided into two steps. First, 100 μl of an overnight culture of PA154197 or PAO1 was spread onto 3.5-cm LB agar plates, followed by incubation at 25°C for 2 days to produce the toxins. Then, a lawn of PA154197 or PAO1 was scraped from the LB agar plates containing the toxins completely using a sterile L spreader. C. elegans strain N2 was synchronized by isolating eggs from gravid adults and plating the eggs onto lawns of Escherichia coli OP50 on NGM (0.25% peptone, 0.3% NaCl, 2% agar, 5 μg/ml cholesterol, 1 mM MgSO4, and 25 mM KH2PO4 [pH 6]) agar plates, followed by incubation at 25°C for 48 to 52 h to reach the L4 stage. Each fast-killing plate was seeded with 20 to 30 worms. Plates were incubated at 25°C and scored every 4 to 6 h. A worm was considered dead when it no longer responded to touch. For the slow-killing assay, the procedure was similar to that used for the fast-killing assay except for the preparation of the killing plates. For slow-killing plates, the lawn of PA154197 or PAO1 was not moved away. All experiments were performed in triplicates.
Pyoverdine production assay.The siderophore pyoverdine is observed under UV light as a fluorescent zone around the colony (67). Fresh single colonies of PA154197 and PAO1 were inoculated in LB medium overnight at 37°C with 220-rpm agitation. After diluting overnight culture to 1 × 109 CFU/ml with fresh LB medium, a 10-μl cell suspension was spotted onto an LB agar plate. The plate was incubated at 37°C for 16 to 20 h. Fluorescent pyoverdine was then visualized and imaged under a UV light.
ACKNOWLEDGMENTS
We thank Karen Yuen (School of Biological Sciences, HKU) for help in establishing the C. elegans killing assay.
This study was supported by the Hong Kong University Grants Council General Research Fund (17142316 to A.Y.), Seed Funding for Strategic Interdisciplinary Research Scheme (HKU 2017 to A.Y.), and the Shenzhen City Knowledge Innovation Plan (JCYJ20160530174441706 to A.Y.).
H.C. and A.Y. designed the studies. H.C., T.X., Y.L., Z.X., Y.K.L., S.B., and V.B.B. conducted the experiments. H.L. provided sequencing service and analysis. P.C.Y.W. provided clinical strains and conducted data analysis on ARGs. H.C., T.X., Y.L., Z.X., and A.Y. wrote the manuscript.
FOOTNOTES
- Received 11 September 2018.
- Returned for modification 20 November 2018.
- Accepted 13 January 2019.
- Accepted manuscript posted online 22 January 2019.
Supplemental material for this article may be found at https://doi.org/10.1128/AAC.01944-18.
- Copyright © 2019 American Society for Microbiology.
