|Home | About | Journals | Submit | Contact Us | Français|
Most known virulence determinants of Pseudomonas aeruginosa are remarkably conserved in this bacterium's core genome, yet individual strains differ significantly in virulence. One explanation for this discrepancy is that pathogenicity islands, regions of DNA found in some strains but not in others, contribute to the overall virulence of P. aeruginosa. Here we employed a strategy in which the virulence of a panel of P. aeruginosa isolates was tested in mouse and plant models of disease, and a highly virulent isolate, PSE9, was chosen for comparison by subtractive hybridization to a less virulent strain, PAO1. The resulting subtractive hybridization sequences were used as tags to identify genomic islands found in PSE9 but absent in PAO1. One 99-kb island, designated P. aeruginosa genomic island 5 (PAGI-5), was a hybrid of the known P. aeruginosa island PAPI-1 and novel sequences. Whereas the PAPI-1-like sequences were found in most tested isolates, the novel sequences were found only in the most virulent isolates. Deletional analysis confirmed that some of these novel sequences contributed to the highly virulent phenotype of PSE9. These results indicate that targeting highly virulent strains of P. aeruginosa may be a useful strategy for identifying pathogenicity islands and novel virulence determinants.
Pseudomonas aeruginosa is a medically important opportunistic pathogen that causes serious disease in hospitalized patients and individuals with cystic fibrosis (9, 56). In the environment, it naturally inhabits lakes, streams, moist soil, and plant matter (15, 20, 47) and has pathogenic activity against a wide spectrum of hosts, including mammals, worms, insects, fungi, amoebae, and plants (1, 14, 22, 24, 36, 42).
Observations from clinical experience and a number of infectious models indicate that the virulence of P. aeruginosa varies from strain to strain (32, 48, 51, 60), although the mechanisms accounting for this variation are not completely understood. The genes of most of the characterized P. aeruginosa virulence determinants are located in the core genome and therefore present in all strains (59). Thus, it is conceivable that varying expression of these conserved pathogenic factors is responsible for differences in virulence between P. aeruginosa strains. Alternatively, P. aeruginosa's accessory genome may contribute to the heterogeneity of virulence. The accessory genome consists of bacteriophages, plasmids, and genomic islands found in some strains but not in others. Genomic islands in particular have been the focus of much recent attention. These horizontally transferred segments of DNA are often integrated into tRNA genes, have G+C contents divergent from that of the host core chromosome, and include components of mobile genetic elements (3, 5, 31, 46). When they encode virulence determinants, genomic islands are referred to as pathogenicity islands (5).
One well-described example of a pathogenicity island contributing to strain-to-strain variation in P. aeruginosa virulence is the family of islands that carry the exoU gene (29), which encodes the type III secretion effector protein ExoU (8, 17). The exoU gene is present in approximately one-third of isolates obtained from acute infections, and secretion of the ExoU toxin is a marker for strains with enhanced virulence (51). It is likely that additional pathogenicity islands contribute to the especially virulent phenotypes of some P. aeruginosa strains. If this is indeed the case, then highly virulent strains should prove to be rich sources of these islands. The identification of novel pathogenicity islands is important because they likely encode novel virulence determinants that would increase our understanding of P. aeruginosa pathogenesis.
In the present study, we utilized a strategy in which we identified and targeted an especially virulent strain of P. aeruginosa for discovery of novel genomic islands. Seven novel islands were identified, one of which was designated P. aeruginosa genomic island 5 (PAGI-5) and examined further. This 99-kb island was shown to contain regions that were associated with highly virulent P. aeruginosa strains. Mutational analysis of these regions confirmed that they contributed to the highly virulent phenotype of the source strain. Targeting of highly virulent bacterial strains may be a useful strategy for identifying novel genomic islands and virulence determinants.
P. aeruginosa PSE strains PSE1 to PSE35 were previously obtained by culture of bronchoscopic fluid from patients who met strict criteria for ventilator-associated pneumonia (16). PAO1 is a laboratory strain of P. aeruginosa (23), and PA14 is a human clinical isolate known to be pathogenic in both plants and mammals (41). Escherichia coli strains JM109 (Promega, Madison, WI), EPI300-T1R (Epicentre, Madison, WI), and S17.1 (53) were used for cloning and conjugation experiments. Antibiotic concentrations and growth conditions are described below.
Data from experiments in which mice were infected with PSE strains were published previously (51) and are reproduced here with permission to facilitate comparison with data from plant virulence studies. The mice were infected intranasally as previously described (51).
Mouse survival studies were performed as previously described by Comolli et al. (4). Briefly, bacteria grown for 17 h in MINS medium (39) at 37°C with shaking (250 rpm) were diluted, regrown to exponential phase, and then were washed and resuspended in phosphate-buffered saline (PBS) (Invitrogen). Six- to eight-week-old female BALB/c mice were anesthetized by intraperitoneal injection of a mixture of ketamine (100 mg/ml) and xylazine (20 mg/ml). A bacterial dose that was approximately equal to the 50% lethal dose (LD50) of PSE9 in 50 ml PBS, as determined by measuring the optical density and confirmed by plating serial dilutions onto Vogel-Bonner medium (VBM) agar, was instilled into the noses of anesthetized mice. The mice were monitored for survival or severe illness over the next 7 days. Severely ill mice, as determined by the presence of matted fur, labored breathing, and decreased activity, were euthanized and scored as dead. The experiments were performed twice, and the results were pooled.
For competition experiments, mice were inoculated as described above for the survival experiments. Inoculation was performed using approximately equal numbers (as determined by measuring the optical density and by plating to obtain viable counts) of parental strain PSE9 and a deletion mutant strain or approximately equal numbers of wild-type strain PAO1 and a PSE9 strain tagged with a gentamicin resistance cassette to allow discrimination between PSE9 and PAO1. Mice were reanesthetized and sacrificed at 22 h postinfection. Lungs and spleens were aseptically removed prior to homogenization in 5 ml PBS. The bacterial load in each organ was determined following plating of serial dilutions on Luria-Bertani (LB) agar and LB agar supplemented with 100 μg/ml of gentamicin to distinguish PSE9 from the second bacterial strain. Colonies were counted following incubation at 37°C for 24 h. The following formula was used to calculate the competitive index (CI) (34): CI = (mutant/wild-type output ratio)/(mutant/wild-type input ratio).
All experiments were approved by and performed in accordance with the guidelines of the Northwestern University Animal Care and Use Committee.
The lettuce infection model was adapted from the model described by Rahme and colleagues (43). Briefly, P. aeruginosa strains were grown to saturation in LB broth at 37°C. Cultures were then diluted 1:200 in fresh LB broth and grown for an additional 3 to 4 h. The resulting log-phase cultures were diluted in 10 mM MgSO4 to obtain an optical density at 600 nm of 0.2. Romaine lettuce leaves were purchased from a local supermarket, washed in 0.1% bleach, rinsed with water, and then placed in a plastic container lined with Whatman paper impregnated with MgSO4. A pipette tip was used to puncture the lettuce midrib and inoculate 10 μl of a diluted culture. The leaves were incubated at 30°C in a humid environment for 4 days, after which the length and width of the region of soft rot were measured. The area of soft rot was estimated using the following formula: A = 0.25π × l × w, where A is area of tissue damage, l is the length, and w is the width. Each strain was inoculated in triplicate. The area of soft rot caused by each P. aeruginosa isolate inoculated was compared to the area of soft rot caused by PA14 inoculated adjacently to control for leaf-to-leaf variation. In certain experiments, the number of CFU present within a lettuce lesion was determined by a method adapted from the method of Dong et al. (6). Briefly, after 4 days the infected region of a lettuce leaf was cut from the midrib and macerated in 5 ml of 10 mM MgSO4 with a mortar and pestle. Serial dilutions were plated on LB agar for enumeration of bacterial CFU following incubation at 37°C for 24 h.
Bacterial genomic DNA was purified from P. aeruginosa strains PSE9 and PAO1 using Genomic-Tip 500/G columns (Qiagen, Valencia, CA) by following the manufacturer's instructions. Subtractive hybridization was then performed using the PCR-Select bacterial genome subtractive hybridization approach (Clontech, Mountain View, CA). Subtractive hybridization was performed as directed by the manufacturer except for the following changes. Genomic DNA was ethanol precipitated with a linear acrylamide carrier (Bio-Rad, Hercules, CA) (12). The primary PCR mixture was incubated at 72°C for 5 min to allow filling of the adapter overhangs before incubation at 94°C for 30 s, at 56°C for 30 s, and at 72°C for 90 s for 25 cycles. The secondary PCR mixture was heated to 72°C before addition of Taq polymerase. The sample was then incubated at 94°C for 30 s, at 58°C for 30 s, and at 72°C for 90 s for 15 cycles. PCR products were purified using the QIAquick PCR purification approach (Qiagen).
Subtractive hybridization products were ligated to the pGEM-T T/A cloning vector (Promega) at 4°C overnight (49). Transformation was performed by adding 2 μl of a ligation mixture to JM109 competent cells (49), and transformants were selected for by growth on LB agar supplemented with ampicillin (50 μg/μl), isopropyl-β-d-thiogalactopyranoside (IPTG) (50 μg/μl), and 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) (50 μg/μl; Sigma-Aldrich, St. Louis, MO). Following growth in LB broth supplemented with ampicillin (50 μg/μl), plasmid DNA was purified from selected transformants using a spin column technique (Qiagen). Plasmid DNA was digested with BglI for 1 h at 37°C and then screened to determine the presence of an insert and the insert size following electrophoresis through a 0.8% agarose gel. Plasmids containing inserts were sequenced by the University of Chicago Cancer Research Center DNA Sequencing Facility (Chicago, IL).
To generate a PSE9 genomic library, the fosmid vector pSB100 was first constructed as follows. The 4.35-kb DrdI fragment of plasmid mini-CTX1 (21), which encodes tetracycline resistance, has an oriT site for mating into P. aeruginosa, and has an attP site and integrase gene for integration into an intergenic chromosomal attB site on the P. aeruginosa chromosome, was purified. This fragment was treated with the DNA polymerase Klenow fragment (New England Biolabs, Beverly, MA) along with each deoxynucleoside triphosphate at a concentration of 33 μM to generate blunt ends and was ligated into the blunt Eco72 I site of the fosmid pCC1FOS (Epicentre) to generate pSB100. The pCC1FOS vector contributed chloramphenicol resistance and cos, ori2, and oriV sites to pSB100. ori2 is the E. coli F-factor single-copy origin of replication, and oriV is an inducible high-copy-number origin of replication. These ori sites allowed pSB100 to be maintained as a low-copy-number fosmid yet to be induced to high copy numbers to facilitate fosmid DNA purification.
To construct a fosmid library of PSE9 genomic DNA fragments, the vector pSB100 was digested with XhoI, and overhangs were partially filled with dTTP and dCTP to generate ends with 5′ TC overhangs. PSE9 genomic DNA was purified as described above, and 1 μg of DNA was partially digested with 0.3 U of Sau3AI (New England Biolabs) at 37°C for 1 h, which was followed by heat inactivation for 20 min at 60°C. To generate DNA fragments with GA 5′ overhangs compatible with the TC 5′ overhangs of the modified pSB100 fosmid vectors, 2.5 U of the DNA polymerase Klenow fragment was added, and the reaction mixture was incubated at 25°C for 15 min in the presence of 33 mM dATP and dGTP. The reaction was terminated by addition of 1.5 μl of 0.2 M EDTA and by heat inactivation at 75°C for 15 min. Following electrophoresis of eight of the reaction mixtures described above through a 0.6% low-melting-point agarose gel, DNA fragments that ranged from 25 to 40 kb long were extracted from the gel using the GELase enzyme preparation (Invitrogen, Carlsbad, CA). Extracted DNA was precipitated with ethanol.
The digested vector and insert were ligated by incubation with Fast-Link DNA ligase (Epicentre) at room temperature for 2 h. The ligation reaction mixture was then incubated with MaxPlax lambda packaging extract (Epicentre) and transduced into E. coli strain EPI300-T1R (Epicentre). Bacteria were plated on LB agar supplemented with chloramphenicol (12.5 μg/μl). A total of 960 colonies were individually inoculated into the wells of 96-well plates containing 150 μl/well LB broth supplemented with glycerol (7.5%) and chloramphenicol (12.5 μg/μl). Ten 96-well plates were incubated at 37°C overnight in a nonshaking incubator and then stored at −80°C.
To assess the quality of the fosmid library, fosmid DNA was isolated from 25 randomly selected clones. The DNA was digested with HindIII, and the restriction digestion patterns were examined following electrophoresis through an agarose gel (0.8%). From the restriction pattern of these fosmid clones, it was estimated that the fosmid insert sizes were between 30 kb and 40 kb. Conservatively assuming an average insert size of 30 kb, a library of this size would be predicted to have a 99% probability of containing any particular genomic sequence.
Fosmid library clones containing subtractive hybridization sequences were detected using a three-tiered PCR-based screening approach (see Fig. S1 in the supplemental material). In the first screening step, PCR amplification using primers specific for subtractive hybridization products (see Table S2 in the supplemental material) was performed with pools of fosmid clones; each pool consisted of all the fosmid clones from a 96-well plate. Fosmid pools were created by replica plating the stocks onto LB agar supplemented with chloramphenicol (12.5 μg/ml), followed by incubation overnight at 37°C. Pools of bacteria were collected directly from the plates by washing with STET buffer (0.1 M NaCl, 10 mM Tris, 1 mM EDTA, 5% Triton X-100) and centrifugation at 18,000 × g. Pools of DNA were then isolated using a column spin technique (Qiagen). Thus, the entire 960-member fosmid library was represented by 10 pools of DNA, which were used as templates for the PCRs. The reaction mixtures were incubated at 94°C for 30 s, at 52°C for 30 s, and at 72°C for 45 s for 25 cycles. In the second screening step, 12 fosmid pools were created using the eight clones from each column of wells in each 96-well plate that had tested positive in the first step of the screen. PCR amplification was then performed to identify which column pool contained the clone of interest. Thus, the location of the library clone containing a specific subtractive hybridization sequence was narrowed to a column of a 96-well plate by screening 12 pools. In the final step of the screening process, each of the eight individual clones from the identified column pool was individually screened by PCR amplification. In this way, the entire fosmid library was rapidly screened for the presence of subtractive hybridization sequences.
To obtain the sequences of the inserts in fosmids containing subtractive hybridization sequences, the EZ::TN<KAN-2> transposon-mediated sequencing approach was used (Epicentre). Briefly, 0.05 pmol of transposon EZ::TN<KAN-2> (provided by the manufacturer) and 2 μg of fosmid DNA were incubated with EZ::TN transposase (provided by the manufacturer) at 37°C for 2 h. The reaction was terminated with stop solution (provided by the manufacturer), and 1 μl of the reaction mixture was electroporated into electrocompetent E. coli EPI300-T1R cells (Epicentre). Electroporated E. coli cells were plated onto LB agar supplemented with kanamycin (50 μg/ml). Colonies were inoculated into 1 ml LB broth supplemented with kanamycin (50 μg/ml) and grown overnight. Cultures were added to 9 ml of LB medium supplemented with chloramphenicol (12.5 μg/ml) and 10 μl CopyControl induction solution (provided by manufacturer), which induces fosmids to high copy numbers, and shaken at 37°C for 5 h. Fosmid DNA was purified using a spin column approach (Qiagen). Primers hybridizing to the borders of the transposon were used to sequence the DNA flanking the transposon insertion site (primer KAN-2 FP-1, ACCTACAACAAAGCTCTCATCAACC; primer KAN-2 RP-1, GCAATGTAACATCAGAGATTTTGAG), and primer walking was used to fill in sequence gaps. Sequencing was performed by SeqWright (Dallas, TX) and by the University of Chicago Cancer Research Center DNA Sequencing Facility. Sequences not present in the fosmid library were obtained by PCR amplification of chromosomal DNA using the Advantage-GC genomic polymerase mixture (Clontech). Each strand of DNA was sequenced one or two times.
Contiguous sequences were assembled using Vector NTI Contig Express (InforMax, Inc., Frederick, MD). Open reading frames (ORFs) in genomic island sequences were predicted using GenDB (37) and GeneMark (35), and the G+C content was calculated by using Vector NTI BioPlot (InforMax, Inc.) from a sliding 100-bp window. Nucleotide and amino acid sequence similarities were identified using BLASTN and BLASTP, respectively (2), and sequences were aligned using Vector NTI AlignX (Informax, Inc.).
PSE9 mutants with deletion of novel region I (NR-I) of PAGI-5 (PSE9ΔNR-I) or novel region II (NR-II) of PAGI-5 (PSE9ΔNR-II) were created by homologous recombination using a variation of the method of Schweizer and Hoang (52). PCR primers were designed to amplify 500- to 700-bp fragments of the 5′ and 3′ ends of PAGI-5 NR-I and NR-II. These PCR fragments were engineered to have NgoMIV restriction sites on the exterior side and XmaI sites on the interior side. These PCR products were digested with NgoMIV and XmaI and sequentially cloned into the XmaI site of pEX100T (52). After PCR was used to confirm the correct orientation of the cloned fragments, the 2.3-kb XmaI digestion product of pX1918G (52), which contained a gentamicin resistance cassette, was cloned into the XmaI site between the two fragments, creating deletion vectors pPG5NRI-5G3 and pPG5NRII-5G3. The deletion vectors were transformed into E. coli S17.1 and then mated into PSE9 (52). Selection for vector integration into the PSE9 genome was obtained by growth on VBM (57) agar supplemented with 100 μg/ml gentamicin. Gentamicin-resistant colonies were transferred to VBM agar supplemented with 100 μg/ml gentamicin and 5% sucrose to induce a second recombination event that resulted in deletion of the targeted region as well as the vector backbone, which included the sacB sucrose sensitivity gene. PCR was used to screen gentamicin- and sucrose-resistant colonies for the presence of the gentamicin resistance cassette (X1918G-GentF, CGCAGCAGCAACGATGTTACGC; X1918G-GentR, CGCGTTGGCCTCATGCTTGA; X1918G-XylF, TCGAATTCCTCCGCGAGAGC; X1918G-XylR, AAATCCATGCCCGGCTCGTC) and deletion of NR-I (PAGI5-5Gupstream, GCACGTTGCCAGATGTTCTCC; PAGI5-5Gdownstream, GGCAGAAATGGCTGCGTTCG) and NR-II (PAGI5-MGupstream, CGATTCAAGCGAGCCAGGATC; PAGI5MGdownstream, GCCACCACGTTGACAACAAGCT).
To distinguish PSE9 from PAO1 in competition experiments, it was necessary to tag PSE9 with a chromosomal copy of a gentamicin resistance cassette. The 2.3-kb XmaI fragment from pX1918G (52) containing the gentamicin resistance cassette was cloned into the XmaI site of mini-CTX1 (21). The resulting mini-CTX-Gent construct was transformed into E. coli S17.1 and mated into wild-type strain PSE9, in which it integrated into the chromosomal attB site. The vector backbone was then excised by mating pFlp2 (21) into the strain, resulting in expression of Flp recombinase. Integration and vector excision were confirmed by PCR as described above for the PSE9ΔNR-I and PSE9ΔNR-II mutants. The absence of a virulence defect in the tagged PSE9 strain was confirmed by performing competition experiments with tagged PSE9 and parental strain PSE9 (data not shown).
The sequence of PAGI-5 has been deposited in the National Center for Biotechnology Information GenBank database under accession number EF611301.
We first investigated strain-to-strain variation in the virulence of P. aeruginosa. For this purpose, a set of 35 previously collected P. aeruginosa clinical isolates designated PSE1, PSE2, PSE3, etc., was used (16). Each of these isolates was originally cultured from patients with ventilator-associated pneumonia. We previously quantified the virulence of these 35 isolates in a mouse model of acute pneumonia by calculating the LD50 (51). As an aid, the data are shown in Fig. Fig.1A.1A. Significant strain-to-strain variation in the levels of virulence was observed, and the LD50s of the most and least virulent strains differed by almost 100-fold. The most virulent strain was PSE9, which had an LD50 of 1.3 × 106 CFU, while the least virulent strain was PSE7, which had an LD50 of 8.8 × 107 CFU. The laboratory strain PAO1 was used as a control and was found to have an intermediate level of virulence (LD50, 4.2 × 107 CFU). These results confirm that strains of P. aeruginosa differ in virulence in an animal model of infection.
The difference in pathogenicity of P. aeruginosa strains suggested that some strains might possess virulence factors that other strains lack. Although the genes encoding most known P. aeruginosa virulence factors are conserved in nearly all strains (59), the exoU and exoS genes, which encode effector proteins of the P. aeruginosa type III secretion system, are variable traits (7, 10). For this reason, the type III secretion profile of each of the 35 strains was determined previously (51); this analysis showed that ExoU-secreting strains as a group were indeed more virulent (Fig. (Fig.1A),1A), but neither ExoU nor ExoS secretion explained all the differences in virulence between these strains (51). Therefore, it was postulated that the remaining differences were due to either differential regulation of conserved virulence determinants or the variable presence of virulence-encoding genes in the accessory genomes of the strains. Subsequent experiments focused on the latter set of genes.
Since P. aeruginosa is also a pathogen of plants (19, 42), the virulence of the 35 isolates was quantified using a plant model of disease. The lettuce leaf infection system developed by Rahme and colleagues was used for this purpose (43). P. aeruginosa was inoculated into the spines of lettuce leaves, and the areas of tissue damage that developed over the ensuing 4 days were determined and used to quantify virulence (Fig. (Fig.2A).2A). Strain PA14, a clinical isolate known to be highly virulent in plants (42), was used as a positive control. Again, the 35 isolates differed in virulence (Fig. (Fig.2B).2B). Whereas some strains had no apparent effect on the lettuce, other strains caused areas of tissue damage larger than those caused by strain PA14. PAO1 was relatively avirulent in this model system, producing an area of damage that was only 15% of the area of damage produced by PA14. Some of the strains that were highly virulent in the mouse model exhibited low levels of virulence in the lettuce model (e.g., PSE41), while other strains were highly virulent in the lettuce model but only slightly virulent in the mouse model (e.g., PSE7 and PSE27). Still other strains were highly virulent in both models (e.g., PSE39 and PSE4). The most virulent isolate in the mouse model, PSE9, was the 17th most virulent strain in the plant model and caused an area of damage that was just under one-half the area of damage caused by PA14. These findings confirm those of Rahme et al. (41) and demonstrate that strains of P. aeruginosa differ in virulence in a plant model of infection. Furthermore, they are consistent with a model in which the accessory genetic material of some strains enhances virulence in animals, the accessory genetic material of other strains enhances virulence in plants, and the accessory genetic material of still other strains enhances virulence in both animals and plants.
Since PSE9 exhibited elevated levels of virulence in both the animal and plant models, it was reasoned that this strain had a high likelihood of containing a number of interesting genomic islands encoding virulence factors. For this reason, a PCR-based subtractive hybridization approach was used to identify genetic regions present in PSE9 but absent in the less virulent PAO1 strain. PAO1 was chosen as the reference strain for these experiments because of its relatively low virulence, the availability of its genomic sequence (55), and its growth rate, which was equivalent to that of PSE9 in LB medium (data not shown). A subtractive hybridization library consisting of 75 fragments of PSE9 DNA was generated, cloned, sequenced, and compared to the GenBank database (Fig. (Fig.3).3). One clone was found to have no insert and was removed from the library. Of the remaining 74 subtractive hybridization products, 13 (18%) were found to be nearly identical to PAO1 sequences, indicating that they were false positives. Of the 61 sequences that were not present in the PAO1 genome, 35 were similar to known sequences, whereas 26 had no significant similarity to known sequences. Based on sequence alignments, 26 sequences were determined to overlap and were therefore removed from the analysis, leaving 21 products with similarity to known sequences and 14 products without similarity to known sequences.
Of the 21 subtractive hybridization products with similarity to known sequences, 13 were nearly identical to previously sequenced P. aeruginosa genomic islands (Fig. (Fig.33 and Table Table1).1). Twelve of these sequences were sequences from a putative serotype 01 O-antigen biosynthesis gene cluster. This cluster is 1 of 11 distinct genetic elements found in P. aeruginosa that encode the biosynthetic enzymes necessary for serotype specificity (45). Thus, PSE9 is related to serotype 01 strains. Strain PAO1, on the other hand, carries the serotype 05 gene cluster, which is divergent from the serotype 01 cluster, explaining why the O-antigen biosynthesis gene cluster sequences were detected by subtractive hybridization. A single insert was similar to a region of the pilA gene, which encodes the pilin subunit of P. aeruginosa type IV pili (50). The type IV pilin genes from different strains of P. aeruginosa segregate into five subclusters that are dispersed among the type IV pilin genes of gram-negative bacteria; pilA genes from different subclusters share less than 30% nucleotide identity (54). For these reasons, it has been postulated that the pilA gene of P. aeruginosa was acquired by horizontal transfer. Thus, the identification of the O-antigen biosynthetic cluster and the pilA gene validated the ability of the subtractive hybridization approach to identify genetic elements that were different in different P. aeruginosa strains. Since the O-antigen biosynthetic cluster and the pilA gene have been characterized previously (45, 54), we did not evaluate them further.
The eight remaining sequences that were similar to known genes had characteristics that suggested that they were parts of novel genomic islands (Table (Table1).1). One sequence was similar to an ORF found in CTX, a cytotoxin-converting phage previously isolated from P. aeruginosa strain PA158 (18, 38). This subtractive hybridization product, however, also contained a novel sequence, suggesting that it was from a related but distinct phage. Another sequence had similarity to P. aeruginosa pathogenicity island 1 (PAPI-1) (19). Note that neither the CTX phage nor PAPI-1 is present in PAO1. The remaining sequences were similar to sequences encoding site-specific recombinases, a zinc-binding transcriptional regulator, a putative phage-related DNA binding protein, and Rhs family elements.
In several cases, the subtractive hybridization products appeared to have identified a small portion of a larger genomic island. Therefore, these sequences were used as tags to identify and characterize the entire genomic island, as well as the DNA flanking the island. To accomplish this, a genomic library of strain PSE9 was constructed using the fosmid vector pSB100, and PCR primers designed to amplify subtractive hybridization products were used to screen the library for individual fosmid clones that contained these sequences (see Materials and Methods). Using this approach, the fosmid library was screened for the presence of the 14 subtractive hybridization products with no similarity to known genes, as well as the eight sequences with similarity to genes not found in PAO1 (Fig. (Fig.3).3). Of these 22 sequences, 20 were represented in the library. The two sequences that were not detected were a clone with similarity to an Rhs family gene and a clone with similarity to PAPI-1. Overall, 25 fosmid clones that contained at least 1 of the 20 subtractive hybridization sequences were identified. A subset of nine fosmid clones cumulatively contained all 20 of the subtractive hybridization sequences and was used in subsequent analyses.
To further characterize the PSE9-associated genomic islands, the complete nucleotide sequence of the subset of nine fosmid clones containing all 20 subtractive hybridization products was obtained. Overall, this analysis suggested that the set of fosmid inserts analyzed represented seven distinct genomic islands located at different sites in the P. aeruginosa genome (data not shown). Here we describe characterization of the largest of these novel islands. The remaining six novel genomic islands will be described elsewhere.
The inserts of three fosmids were determined to contain portions of a single large genomic island with similarity to PAPI-1. The complete sequence of this island was obtained (see the supplemental material). Since this island differed substantially from PAPI-1 (see below), we elected to give it a unique name. Using the nomenclature system of Liang et al. (33), Larbig et al. (30), and Klockgether et al. (27), who identified PAGI-1, PAGI-2, PAGI-3, and PAGI-4, we designated this large island PAGI-5.
PAGI-5 is the largest of the genomic islands identified in PSE9; it is 99,276 bp long. Its G+C content is 59.6%, which is lower than the PAO1 overall genome G+C content, 66.6% (55). This island is predicted to contain 121 ORFs and is integrated into the genome immediately adjacent to a tRNALys gene (PA0976.1) at bp 1,061,197 in the core chromosome. (PAO1 gene designations are used throughout this paper .) tRNA genes frequently serve as integration sites for prokaryotic genomic islands (58), and this P. aeruginosa tRNA gene is no exception. It serves as the insertion site for PAPI-1, PAPI-2, pKLK106, and PAGI-4 (26, 27, 40).
Based on sequence comparisons, PAGI-5 is related to a known family of P. aeruginosa genomic islands that includes PAPI-1, PAPI-2, ExoU islands A, B, and C, and an unnamed 8.9-kb tRNALys-associated island in strain PAO1 (19, 27, 29). These islands themselves comprise a subset of a large family of pKLC102-related genomic islands prevalent in beta- and gammaproteobacteria (28). The members of the pKLC102 family of genomic islands are plasmid-phage hybrids that consist of two parts: a relatively conserved core set of genes involved in propagation, replication, and partitioning, and variable “cargo” gene cassettes (27, 28, 61). Kulasekara and colleagues (29) have proposed that the PAPI-1-related islands evolved from an ancestral integrative plasmid similar to pKLC102. According to their model, during evolution these related elements diverged into two clades, which can be distinguished by the presence of the genes encoding the type III effector protein ExoU and its chaperone SpcU. In one clade, consisting of PAPI-2 and ExoU islands A, B, and C, the exoU and spcU genes are present, but additional rearrangements during or following integration led to loss of the partitioning factor gene of the pKLC102-like plasmid (29). The loss of this plasmid feature may have fixed the island into the chromosome (29). Consistent with this model is the finding that each of these islands is integrated into the same tRNALys gene (PA0976.1). The second clade consists of PAPI-1 and an 8.9-kb tRNALys-associated island of strain PAO1, which evolved from a lineage of the ancestral plasmid that did not acquire (or lost) the exoU and spcU genes. PAPI-1 has maintained the features of the integrated plasmid and has been shown to be transferable (40). As a result, PAPI-1 can integrate into either of the two tRNALys genes (PA4541.1 and PA0976.1) present in the P. aeruginosa genome (40). It is not surprising that some members of this clade can integrate into either of these sites, since the pKLC102-like plasmid pKLK106 has been shown to integrate into either site (26, 27). PAGI-5 appears to be another member of the second clade, since it also does not contain the exoU and spcU genes. PAGI-5 is integrated into the tRNALys gene PA0976.1 in PSE9; additional studies are necessary to determine whether this island is also transferable and can be found in the tRNALys gene PA4541.1 in other strains. PCR analysis indicated that the integration site in the PA4541.1 tRNALys gene of PSE9 is unoccupied (data not shown). Interestingly, like PAPI-1, PAGI-5 contains an intact partitioning factor gene (5PG121), suggesting that it may be transferable.
Of this group of related genomic islands, PAGI-5 is most similar to PAPI-1; 79 of the 121 predicted PAGI-5 ORFs share similarity to PAPI-1 ORFs (Fig. (Fig.44 and and5;5; see Table S1 in the supplemental material). Yet PAGI-5 carries a substantial amount of genetic information that is not present in PAPI-1 and also lacks a number of PAPI-1 ORFs. Since PAPI-1 has been thoroughly described previously (19), here we highlight the features of PAGI-5 that differ from features of PAPI-1 (Fig. (Fig.44 and and5).5). A 14.8-kb region of PAPI-1 containing 11 ORFs (RL036 to RL046) is not present in PAGI-5. Two other large regions of PAPI-1 are also missing from PAGI-5, but both of these regions are replaced in PAGI-5 with novel DNA sequences. A 6.2-kb region containing seven PAPI-1 ORFs (RL004 to RL010) is replaced by an 8.5-kb sequence, which is referred to as NR-I. NR-I contains five ORFs (5PG3 to 5PG7), four without similarity to previously characterized genes and one (5PG4) that shares 25% identity with a putative methylase gene from Bacillus cereus. This region of PAGI-5 has a G+C content of 50.5%, which is considerably lower than the G+C content of PAGI-5 as a whole (59.6%), suggesting that its origins are distinct from those of the remainder of PAGI-5 (Fig. (Fig.4).4). A central 7.8-kb PAPI-1 region containing ORFs RL053 to RL062 is replaced in PAGI-5 by a 17.9-kb sequence, which is referred to as NR-II. NR-II contains 23 predicted ORFs (5PG40 to 5PG62) and has a G+C content of 56.6%. The first part of this region contains nine predicted ORFs that lack similarity to any previously characterized sequences. These ORFs are followed by an ORF (5PG49) that shares 99% identity with an ORF from P. aeruginosa strain C3719 (http://www.ncbi.nlm.nih.gov/sites/entrez?Db=genome&Cmd=ShowDetailView&TermToSearch=5397), which is predicted to encode an SOS response transcriptional repressor due to the presence of a region similar to a LexA domain (COG1974) (11). The 5PG49 and C3719 ORFs both share only 46% similarity and 32% identity overall with the ORF encoding PAO1 LexA (PA3007), the canonical SOS response repressor (13). The next ORF has apparently served as the integration site for a small genetic element (Fig. (Fig.4).4). This ORF, whose product is similar to a nucleotidyltransferase, is split into two separate ORFs (5PG50 and 5PG62) that flank the putative genetic element. Consistent with this interpretation is the finding that a region similar to 5PG50 and 5PG62 (along with 5PG43 to 5PG49) but lacking the inserted genetic element is present in strain C3719. The element itself consists of 11 ORFs, including ORFs encoding a predicted recombinase and two predicted integrases that are between 28% and 40% identical to the products of three predicted ORFs of Burkholderia multivorans. The presence of multiple ORFs associated with mobile elements suggests that these 11 ORFs had an evolutionarily distinct origin, which is supported by the presence of a region 99% identical to 5PG51 to 5PG61 in P. aeruginosa strain PA7 (http://www.ncbi.nlm.nih.gov/sites/entrez?Db=genome&Cmd=ShowDetailView&TermToSearch=21213). Also present in this genetic element is an ORF similar to ORFs encoding members of the TetR family of transcriptional regulators, which typically repress gene expression in response to environmental cues (44). The TetR regulator gene and an adjacent integrase gene are similar to a regulator-integrase gene pair found in P. aeruginosa plasmid pKLC102. In pKLC102 these ORFs are followed by ORFs encoding a putative short-chain dehydrogenase/reductase protein and another TetR regulator protein; in PAGI-5 they are instead adjacent to a cluster of ORFs (5PG56 to 5PG61) similar to a 4-kb fragment of the Pseudomonas mercury resistance transposon Tn5041 (25, 30; http://www.ncbi.nlm.nih.gov/sites/entrez?Db=genome&Cmd=ShowDetailView&TermToSearch=5398). This cluster includes ORFs encoding homologs of the following proteins: MerR (transcriptional regulator of the CueR family), MerT (mercuric ion transport), MerP (periplasmic mercuric ion binding), MerC [inner membrane Hg(II) uptake], MerA (mercuric ion reductase), and ORFY (hypothetical protein). Thus, PAGI-5 has the potential to confer mercury resistance.
There are other minor differences between PAGI-5 and PAPI-1. For example, in place of RL013 of PAPI-1, PAGI-5 carries IS407, an insertion sequence that contains two ORFs (5PG10 and 5PG11) predicted to encode transposases. Interestingly, all or portions of IS407 are also found in ExoU islands A, B and C, where the sequence is adjacent to the exoU and spcU genes. In contrast, in PAGI-5 and the 8.9-kb genomic island associated with the PAO1 PA0976.1 tRNALys gene, the IS407 sequences are not associated with the exoU and spcU genes, which are not present in these islands. PAPI-1 lacks both IS407 and the exoU and spcU genes (19). The close association between IS407, this group of genomic islands, and the exoU and spcU genes suggests that this insertion sequence played a role in either the acquisition or loss of the exoU and spcU genes from the ancestor of these elements (29).
Despite these differences, the majority of PAGI-5 is similar to PAPI-1 and even to the less closely related ExoU island A (Fig. (Fig.5).5). For example, ORFs 5PG12 to 5PG39 are conserved in all three islands and appear to involve plasmid-related functions. A subset of these ORFs (5PG21 to 5PG29) is similar to a cluster of genes from pKLC102 (CP73 to CP81) that are conserved in other tRNALys-associated islands in multiple species (27). Likewise, ORFs putatively encoding an integrase (5PG1), a plasmid stabilization protein (5PG8), a transcriptional regulator (5PG9), a helicase (5PG63), and a methyltransferase (5PG64) are present in all three islands. These conserved regions are consistent with a common ancestry.
As mentioned above, PAGI-5 appears to be a chimeric genomic island consisting of three PAPI-1-related regions and two novel regions (designated NR-I and NR-II) (Fig. (Fig.5).5). To determine the frequency and distribution of these regions among P. aeruginosa strains, we screened for the presence of NR-I, NR-II, and the large PAPI-1-like regions in the center and in the 3′ end of PAGI-5 in a collection of 35 clinical isolates (Fig. (Fig.1B).1B). PCR was used to amplify a sequence within each of these regions in each isolate (Fig. (Fig.5).5). Amplified products from the central and 3′ PAPI-1 conserved regions were observed in 34 (97%) of the 35 clinical isolates (Fig. (Fig.1B),1B), consistent with previous reports indicating that PAPI-1-related islands are common among P. aeruginosa strains (28). In contrast, amplification of sequences within the novel NR-I and NR-II regions was observed only in highly virulent isolates. Specifically, the NR-I sequence was observed only in PSE9 itself, the most virulent of the 35 isolates, and the NR-II sequence was observed only in seven of the most virulent isolates (Fig. (Fig.1B).1B). The presence of NR-II exclusively in highly virulent isolates suggested that it contributed directly to the increased pathogenicity of these isolates or was genetically associated with other factors that contributed.
To examine the role of NR-I and NR-II of PAGI-5 in virulence, two PSE9 deletion strains were created by homologous recombination (see Materials and Methods). The first mutant strain, PSE9ΔNR-I, had a deletion of bp 3712 to 9342 within NR-I, disrupting or deleting ORFs 5PG4 to 5PG7. The second mutant strain, PSE9ΔNR-II, had a deletion of bp 37,564 to 54,397 of NR-II, disrupting or deleting ORFs 5PG40 to 5PG62. In both strains, the deleted sequences were replaced with gentamicin resistance cassettes. Neither PSE9ΔNR-I nor PSE9ΔNR-II exhibited a growth defect in minimal medium (data not shown).
The importance of NR-I and NR-II to the virulence of PSE9 was then determined using the deletion mutants in the mouse model of acute pneumonia. Mice were inoculated by nasal aspiration with PSE9ΔNR-I, PSE9ΔNR-II, parental strain PSE9, or PAO1, and survival was monitored over the subsequent 7 days. Nearly all mice inoculated with parental strain PSE9 died during the course of the experiment, whereas all of the PAO1-infected mice survived (Fig. (Fig.6).6). The survival curve for mice inoculated with PSE9ΔNR-I resembled that for mice inoculated with parental strain PSE9, indicating that NR-I did not have a major effect on the survival of mice in the acute pneumonia model. In contrast, the mice infected with PSE9ΔNR-II had significantly improved survival compared to the mice infected with parental strain PSE9 (P = 0.0036). These results indicate that NR-II of PAGI-5 contributes to the highly virulent phenotype of PSE9.
Next, the virulence of the NR-I and NR-II mutants was measured using competition assays, which can detect small differences in virulence between two strains. Mice were inoculated by nasal aspiration with a mixed dose of PSE9ΔNR-I and parental strain PSE9 or with a mixed dose of PSE9ΔNR-II and parental strain PSE9, and the amounts of viable bacteria present in the lungs and spleen were determined after 22 h of infection. Deletion of either NR-I or NR-II resulted in modest but statistically significant decreases in competitive fitness; the mean CIs were 0.56 and 0.37 in the lungs and 0.35 and 0.33 in the spleens, respectively (Fig. (Fig.7).7). In comparison, wild-type strain PAO1 had mean CIs of 0.15 in the lungs and 0.16 in the spleens when it competed against PSE9 (Fig. (Fig.7).7). The finding that there was a substantial difference in virulence between PSE9ΔNR-I and PSE9ΔNR-II in survival assays but there was only a small difference in competition assays may reflect a threshold below which virulence is undetectable in survival assays but apparent in competition assays. Alternatively, the true virulence defect of PSE9ΔNR-II may be masked in competition assays by the “complementing” effect of coinoculated parental strain PSE9. Together with the results of the survival experiments, these results indicate that NR-II of PAGI-5 makes a substantial contribution to the virulence of PSE9, whereas NR-I makes a more modest contribution.
The NR-I and NR-II mutants were also tested using the lettuce leaf model. After 4 days, no difference in either the area of tissue damage or bacterial survival was detected between parental strain PSE9 and either of the mutants (data not shown). Thus, factors other than PAGI-5 NR-I and NR-II must contribute to the virulent phenotype of PSE9 in the lettuce leaf model.
The approach of targeting a highly virulent strain as a source of novel pathogenicity islands in P. aeruginosa has led to identification of seven novel genomic islands, at least one of which is a pathogenicity island. PAGI-5 is a 99-kb hybrid island that is related to the PAPI-1 family of islands but has two large regions with novel sequences, NR-I and NR-II. Deletion of NR-II resulted in a marked decrease in the virulence of parental strain PSE9, and deletion of NR-1 resulted in a modest decrease in virulence. Thus, both these regions encode novel virulence determinants that enhance the pathogenicity of PSE9 and are examples of factors responsible for strain-to-strain variation in P. aeruginosa virulence. Examination of other highly virulent strains may lead to identification of additional novel pathogenicity islands in P. aeruginosa, as well as in other bacteria. The advent of relatively inexpensive whole-genome sequencing should greatly facilitate these studies and enable more complete identification of the full arsenal of virulence factors available for use by P. aeruginosa.
This work was supported by an American Heart Association predoctoral fellowship to S.E.B. and by NIH grants K02 AI065615 and RO1 AI053674 to A.R.H.
We thank Peter Agron for advice and technical assistance with the subtractive hybridization technique, Laurence Rahme for assistance with the lettuce model of infection, Kathryn Bieging for experimental assistance, and Herbert Schweizer for providing the mini-CTX, pX1918G, pFlp2, and pEX100T vectors.
Published ahead of print on 29 August 2008.
†Supplemental material for this article may be found at http://jb.asm.org/.