The clone CHA strains CHA, 491 and PT22 were isolated from sites in Grenoble, Hannover and Mülheim in 1990, 2005 and 1992, respectively. Strain PT22 was isolated from a river, whereas strains CHA and 491 are CF airway isolates. Strain CHA was recovered from a critically ill CF patient with advanced lung disease and chronic P. aeruginosa infection 14 . Strain 491 was the first clone CHA isolate from respiratory secretions of a female CF patient with normal lung function 20 . The strain was successfully eradicated from the patient’s airways by antipseudomonal chemotherapy and no further clone CHA strain has since been identified in the patient’s respiratory secretions. The three clone TB strains were isolated from a burn wound (strain TB63741) and two unrelated CF patients (strains TBCF10839 and TBCF121838 16 ) during a local outbreak at Hannover Medical School in summer 1983.

Fragment libraries of CHA, 491, PT22 and TB63741 were sequenced with the Illumina Genome Analyser II generating 36 bp reads as previously reported for strains TBCF10839 and TBCF121838 16 . Reads passing quality criteria 10 were mapped to the PAO1 genome sequence ( 21 ; NCBI sequence NC_002516.2) in order to detect SNPs, indels and PAO1 loci absent in clones CHA and TB. Contigs representing the non-PAO1 loci of the accessory genome were de novo assembled from reads that could not be mapped to the PAO1 reference.

In contrast to the analysed clone CHA strains, little intraclonal genomic diversity was observed for the three clone TB strains that were sampled during a local outbreak at Hannover Medical School. As reported earlier, only five individual nucleotide exchanges and one deletion each in a pilus assembly gene could be detected in the two CF airways isolates TBCF10839 and TBCF121838 16 . Though many phenotypic differences were observed, also the accessory genome differed by only one 81 kb Ralstonia pickettii PAGI-2 like genomic island absent in the first but present in the latter isolate 16 .

Sequencing of a third clone TB isolate, the wound isolate TB63741, revealed some more intraclonal diversity, but still less than observed for the three clone CHA strains. TB63741 lacked six nucleotide exchanges that were detected for both TB CF isolates, but carried 22 individual SNPs not seen in any of the two CF isolates (Figure 1, Additional file 11). TB63741 did not harbour any deletion in a pil gene, but it had acquired a 9-bp in-frame deletion in a two component sensor gene and two frame-shift mutations in a phage gene and in oprD (see Additional file 11). The porin OprD transports basic amino acids and peptides but it also takes up the antipseudomonal agent imipenem. Loss-of-function mutations in oprD as seen in the clinical isolate TB63741 are a common mechanism of imipenem resistance 46 .

Similar to the clone CHA lineage, the conservation of described non-coding sRNA loci does not differ within the clone TB lineage apart from one exception. The sRNA phrD and 30 pant-sRNAs are absent in the three genomes, of another 10 pant-sRNA loci significant parts were lacking (see Additional file 6). The phage DNA-associated sRNA pant78, present in both CF-isolates but absent in TB63741 made up the only intraclonal difference regarding sRNAs in clone TB.

Comparison of the sRNA conservation in clonal lineages CHA and TB revealed clone-specific patterns. While phrD and 20 pant-sRNA loci from PAO1 were completely absent (and four more partially) in both lineages, clone CHA lacked 17 pant-sRNAs which were present in clone TB. Six pant-sRNAs, however, were absent in clone TB but fully conserved in clone CHA. For another 23 pant-sRNA loci conservation patterns were partially divergent in the two clonal lineages (see Additional file 6). According to that, varying spectra of small non-coding RNA genes in P. aeruginosa might contribute significantly to interclonal diversity but only to a small degree to diversity between clonal variants, if sRNA genes are parts of strain-specific acquisition of mobile DNA elements.

Clone TB is endowed with a large accessory genome including the genomic islands PAGI-1, PAGI-2, PAGI-5 and PAGI-6 16 . The wound isolate TB63741 lacks the 81 kb TBCF121838-specific R. pickettii genomic island and numerous phage-like ORFs of phage Pf1 and of genomic island LESGI-1 which were present in both CF isolates. Conversely, TB63741 has incorporated more than 300 kbp that are absent in the two CF strains. Virtually all this DNA is of phage origin including LES-prophage 2 and 3 sequence 37 , of which 67.3 or 76.2%, respectively, of the DNA were found in TBCF63741 with nucleotide identities ranging from 80 to 100%. The closest homologues of accessory genome ORFs were found in other P. aeruginosa clones, other Pseudomonas taxa or in ‘honorary’ pseudomonads (see Additional file 12). The shuffling of phage DNA apparently was the major driving force of microevolution of clone TB during the outbreak.

This study compared the intraclonal genome diversity of P. aeruginosa isolates derived from common and divergent sources. Consistent with our expectation higher genomic variation was found among the clonal isolates with a more diverse spatiotemporal origin.

Sequence variation was low among the three clone TB strains that had been sampled in summer 1983 during a local outbreak. The two CF isolates belong to a small epidemic that tripled the prevalence of P. aeruginosa – positive patients at the CF clinic 15 . Despite individual profiles of phenotype, strains TBCF10839 and TBCF121838 show only minute differences in their genome sequence 16 . Strain TB63741 was isolated from a patient with severe burns who had been treated at the intensive care unit for burns from which clone TB had initially spread to surgical wards and later to the CF clinic. The ancestors of the TB63741 strain had incorporated numerous phages into the clone TB genome that were absent in the isolates from the CF lungs indicating that highly colonised burn wounds themselves and/or the associated hospital environment had tolerated or favoured the uptake of phages.

The three clone TB isolates had descended from a common source and the individual clades had diverged from each other by at most two years. In contrast, the three sequenced clone CHA isolates were sampled from spatially and temporarily distinct habitats. Correspondingly, the sequence of the core genome and the composition of the accessory genome were significantly more diverse among the three clone CHA than among the three clone TB strains. In particular, the numerous strain specific SNPs in absence of pairwise shared SNPs demonstrate the distinct microevolution of the clone CHA strains (Figure 3). Conversely, shared de novo mutations and comparably very few individual de novo mutations highlight the close relatedness of the two clone TB CF isolates.

The environmental isolate PT22 was endowed with the largest accessory genome of the investigated strains. PT22 was collected from the river Ruhr at a site with substantial anthropogenic pollution and contamination with industrial sewage (Wasserqualität der Ruhr 1992 47 ). Consistent with its source, the genomic islands of PT22 encoded genes for the detoxification of xenobiotics and the efflux of heavy metal ions. PT22 carried a copy of PAGI-2 which also exists in CF isolates and Cupriavidus metallidurans CH34 that had been sampled from an industrial site polluted with heavy metal ions 48 49 .

The CF airways isolates 491 and CHA were retrieved from patients with the extremes of the general state of health that are feasible with CF as the underlying predisposing condition: The clinically highly pathogenic strain CHA was isolated from a CF patient with end-stage lung disease, whereas strain 491 was recovered from an individual with normal anthropometry and excellent lung function. Strain 491 was eradicated by antipseudomonal chemotherapy and no clone CHA strain has yet been re-isolated from the patient’s respiratory secretions in the last seven years. 491 had gained numerous elements of genomic mobility that may confer some global fitness to the strain, but only a few amino acid substitutions in traits that may facilitate the colonization of CF airways. In other words, the microevolution of the 491 clade does not point to any pronounced selection of the 491 ancestry to accommodate itself to the CF lung habitat.

Conversely, the ancestors of the strain CHA isolate had selected numerous non-conservative amino acid substitutions in elements of chemotaxis, exopolysaccharide biosynthesis, motility and virulence. In addition, the genes gacS and ldhA were destroyed by a deletion. The lactate dehydrogenase LdhA has recently been demonstrated in strains PA14 and PAO1 to be indispensible for microcolony formation in biofilms 35 . Hence deletion of the 3’ end of ldhA could alter biofilm formation although strain CHA displayed mucoid growth on agar plates (data not shown). The GacS/GacA two-component system controls the reciprocal expression of acute and chronic virulence determinants 34 50 . The deletion of gacS should abrogate this control. Consistent with this interpretation, strain CHA strongly expresses the pathways for alginate biosynthesis, a hallmark of a chronic infection, and the virulence effectors and structural elements of type III secretion, a hallmark of an acute infection (mRNA microarray data from bacteria grown to stationary phase, data not shown). Deletions and point mutations in key determinants of virulence and the control thereof thus established a genetic repertoire in the strain CHA isolate that is distinct from 491 and PT22 and should translate into the observed high pathogenic potential in the predisposed human host. This microevolution towards virulence seems to be quite specific for the inhabited CF lungs because strain CHA was inconspicuous in standard P. aeruginosa worm and fly infection models 51 . Strain CHA apparently acquired signatures of a host-specific pathogen, whereas the 491 and PT22 clades retained the balance between environmental organism and opportunistic pathogen.

The clone CHA and TB genomes share numerous prophages and genomic islands with the virulent and transmissible LES clone, which has caused substantial morbidity in the CF patient population in the UK 37 . The relatedness of their genomes may explain why these clones are prone to nosocomial spread among predisposed human hosts and why virulent clades with uncommon pathogenicity traits have evolved in these clonal complexes. Subsequent evolvement of pathogenicity arising from such genomic predisposition proceeded differently then in the highly virulent examples TBCF10839 and CHA.

In the case of TBCF10839 only few sequence variations clearly differentiated its genome from that of the other two less virulent TB strains, mainly a loss-of-function mutation in TBCF10839 52 . While lacking of type IV pili on the surface and being impaired in twitching motility, TBCF10839 was metabolically more active 16 , produced more outer membrane transporters and secreted more virulence effectors 53 than its clonal variants. Apparently the loss of PilQ induced a global response in the TB background that is far beyond pilus biogenesis. Any further mutations that are necessary to generate the unique ability of TBCF10839 to grow in neutrophils must have already existed in the clone TB lineage. Strain CHA, however, exhibits numerous strain-specific gain- or loss-of-function mutations in global regulators or key pathogenicity factors that should be involved in the specific virulence features of strain CHA like its capability to cause oncosis of neutrophils 17 18 19 . Evolvement of the specific pathogenicity traits likely occurred by a series of microevolution events in this case.

P. aeruginosa strains 491, TBCF10839, TBCF121838 and TB63741 were isolated from patients seen at the Medizinische Hochschule Hannover. Strain PT22 was retrieved from the river Ruhr close to Mülheim. Strain CHA was isolated from a patient seen at the CF clinic in Grenoble. First subcultures were maintained in LB supplemented with 15% (w/v) glycerol at −80°C until use.

P. aeruginosa strains were genotyped by a custom-made microarray following the protocol published previously 6 .

P. aeruginosa genomic DNA was prepared from cells grown in LB medium following a protocol optimized for Gram-negative bacteria 54 .

After preparing genomic DNA libraries according to the manufacturer’s instructions, sequencing-by-synthesis was performed at GATC-Biotech (Constance, Germany) for each library with an Illumina Genome Analyser II generating 36 bp sequence reads. Illumina Genome Analyser Pipeline Version 0.2 software was applied to qualify reads passing default signal quality filters. Obviously incorrect reads with homooligomers > 13 bases in length (not present in the P. aeruginosa genome) or an ‘N’-base call in at least three positions were excluded from the analysis 10 . All sequence data from this study have been submitted to the Sequence Read Archive (SRA) of the EBI (strain TB63741: study accession no. ERP001300; clone CHA strains CHA, PT22 and 491: study accession no. ERP001750).

36 bp reads data of the strains were individually mapped to the PAO1 reference genome (NC_002516.2) using the accurate alignment software Novoalign V2.07.00 (Novocraft Technologies, 2010). The command: novoalign –d Indexed_reference_genome –f Reads.fastq –o SAM > out.sam, was used during the mapping to create “sam” formatted alignment files. Two pools of data consisting of the PAO1 mapped and unmapped reads were then extracted directly from the three alignment files using a custom script. Unmapped reads representing non-PAO1 DNA and the mapped reads representing the PAO1 DNA were assigned to not-in-reference and in-reference read pools, respectively.

Clone CHA strains with genomic positions indicating single nucleotide variants relative to the PAO1 reference were extracted from the novoalign alignment files using SAMtools 55 . The variant call format (vcf) output files generated by SAMtools were further filtered for low quality variants. Variants with minimum coverage of six reads with minimum base calling quality (Q) of 30 at the respective position, a minimum SNP-call quality (QUAL) of 160 (QUAL = −10 log₁₀ (probability of wrong call) 56 ) and with more than 67% of all quality reads calling the SNP were retained. These variants were then compared against each other to identify sets of strain specific SNPs through the use of an in-house SNP filter pipeline.

The SAMtools derived sequence variants output files were further searched for predictions of small indels. The top candidates (QUAL ≥ 160) were verified by manual inspection of the alignment. Predicted indels were removed that did not pass the following criteria: minimum coverage of more than five high quality reads (Q ≥30 at the candidate position) and more than 95% of reads flag the indel. Predicted indels and SNPs were subsequently annotated using SNPeff version 1.9.5 57 to identify their effect on coding DNA sequences.

The not-in-reference pools of sequence reads characterized as Clone CHA accessory genome were assembled to larger contigs with the de novo assembler Velvet version 1.0.12 58 . Commands used during the assembly process are as follows: velveth 63741_cov5_23 23 63741_reads.fas; velvetg 63741_cov5_23 -cov_cutoff 5.0 -max_coverage 300. The assembler parameters were set for a minimum read coverage of 5 and kmer size of 23 to construct reliable contigs. These criteria were set for the analysis as they were demonstrated to maximise the tradeoff between base pairs incorporated and average and maximum contig size after thorough empirical testing. Assembled contigs of strain triplets were aligned against one another by blastn (1e-5 E-value threshold) to search for similarity between the sequences. Contigs that lacked similarity with others were designated as strain-specific DNA. These candidates were further validated using alignments of the short read data sets from both other strains using Novoalign. Contigs covered by reads were not considered to be strain-specific.

Validated strain-specific contigs were aligned using blastx against the UniProt database 59 to identify sets of known (present in other P. aeruginosa) and novel (not present in other P. aeruginosa) genes in their accessory genomes.

Assembled contigs of the three clone CHA strains were aligned against all known P. aeruginosa genomic islands and insertions in regions of genome plasticity using blastn (1e-10 E-value threshold). Alignment results for all the searches were then visualized by GenomeGraphs 60 , an integrated genomic data visualization package for R (http://www.r-project.org) to help determine which of the known horizontally transferred genomic elements are completely/partially present in the three clone CHA strains.

Uncovered regions of the reference were extracted from the alignment results for the individual strains and checked for intersection with the 557 sRNA loci described for the PAO1 reference 34 . Complete or partial absence (> 10% not conserved) was confirmed by visual inspection of alignment/coverage for these loci using the Integrative Genomics Viewer 61 .