Development of a Multilocus Sequence Typing Scheme for the Opportunistic Pathogen Pseudomonas aeruginosa

P. aeruginosa culture collection.P. aeruginosa strains were obtained from the Hajo Grundmann collection (10), deposited at the Health Protection Agency, Colindale, London, United Kingdom. Six isolates were from mushroom compost at one experimental mushroom farm and were provided by Alun Morgan Horticulture Research International. In addition, ∼100 isolates were collected from hospitals across the United Kingdom. See Table 1 for a summary of all 143 isolates. Strains were identified as P. aeruginosa as described previously (10), and DNA fingerprinting of toxA and serotyping were also performed.

Culture of isolates and preparation of chromosomal DNA.Bacterial strains were maintained at −80°C in 12% (vol/vol) glycerol in brain heart infusion (BHI) broth, streaked to single colonies, and cultured on BHI agar at 37°C under aerobic conditions. Chromosomal DNA was extracted from these purified strains with a DNeasy kit (Qiagen).

Locus selection.Several potential loci were identified by using the P. aeruginosa PAO1 genome database (http://www.pseudomonas.com/) (27). Criteria governing locus selection included biological role (e.g., a diverse range of different central housekeeping roles, such as mismatch repair, DNA replication, and amino acid biosynthesis), size (>600 bp), location (i.e., a minimum of 6 kbp upstream or downstream from known virulence factors, lysogenic phage, or insertion sequence elements), and suitability for nested primer design and sequence diversity (ideally, the possession of conserved domains flanking a variable central core). The seven genes finally selected for use with the MLST scheme were acsA, aroE, guaA, mutL, nuoD, ppsA, and trpE (Table 2).

Amplification and sequencing of loci.PCR primers were designed for the loci listed above by using the published P. aeruginosa sequences (27). The primers used, all of which had a common melting temperature_, are shown in Table 3. The 50-μl amplification reaction mixture comprised ∼10 ng of chromosomal DNA, 1 μM each primer, 1× PCR buffer (Qiagen), 1.5 mM MgCl₂, 2 mM each deoxynucleoside triphosphate, and 2.5 U of Taq DNA polymerase (Qiagen). The reaction conditions were denaturation at 96°C for 1 min, primer annealing at 55°C for 1 min, and extension at 72°C for 1 min for 35 cycles. The amplification product was purified with MinElute UF (Qiagen), according to the protocol of the manufacturer. The nucleotide sequences were determined by using internal nested primers and 2 μl of BigDye Terminator Ready Reaction Mix (version 3.1) with standard sequencing conditions, according to the protocol of the manufacturer. Unincorporated dye terminators were removed by precipitation with 95% alcohol. The reaction products were separated and detected on an ABI PRISM 3100 genetic analyzer by using a standard sequencing module with Performance Optimized Polymer Applera UK, Warrington, United Kingdom) and a 50-cm array.

Allele and ST assignment.An arbitrary number was given to each distinct allele within a locus. Each isolate was therefore given seven numbers that represented its sequence type (ST). Each sequence type was numbered in order of appearance (ST1, ST2, etc.). Allele profiles and STs can be found at http://pubmlst.org/paeruginosa.

Phylogenetic analysis.The number of polymorphic nucleotide sites, calculation of the ratio of the number of nonsynonymous substitutions to the number of synonymous substitutions (the d_N/d_S ratio), and construction of a dendrogram by the unweighted pair group method with arithmetic averages (UPGMA) were performed with the START program (http://www.mlst.net) (11).

Allelic variation in P. aeruginosa.Among the 143 isolates investigated, the number of housekeeping gene alleles ranged from 21 for nuoD to 43 for acsA (Table 4). There were between 23 and 37 variable sites within each locus; i.e., 5 to 8% of base pairs represented variable sites.

The d_N/d_S ratio indicates the presence or absence of a selection pressure on the locus. Usually, most nonsynonymous changes would be expected to be eliminated by purifying selection, but under certain conditions Darwinian selection may lead to their retention. Investigation of the number of synonymous and nonsynonymous substitutions may therefore provide information about the degree of selection operating on a system. The low d_N/d_S ratios in Table 4 indicate the absence of a strong positive selective pressure at these loci and the suitability of these loci for population genetic studies.

Relatedness of P. aeruginosa isolates.A total of 139 different STs were assigned to the 143 isolates investigated (Table 1). The rank order of isolates within Table 1 was derived from a UPGMA dendrogram of ST allelic profiles. Ten lineages or clonal complexes were identified among these isolates, and these were composed of strains with either identical STs or STs that varied at one or two loci (single- or double-locus variants), with founder strains indicated below with an asterisk. A founder strain has the ST to which all other STs in the clonal group are related (at least for that sample of strains examined). The compositions of these groups were as follows: group 1, 2 isolates; STs 82 and 83; group 2, 12 isolates, STs 7, 11, 15, 27*, 119, 122, 128, and 129; group 3, 5 isolates, STs 14, 17*, 115, 117, and 142; group 4, 6 isolates; STs 53, 97, 102, 111, 113, and 124; group 5, 5 isolates, STs 30, 31, 38, 39, and 46; group 6, 3 isolates, STs 104, 107, and 109; group 7, 2 isolates, STs 61 and 69; group 8, 5 isolates, STs 41, 45, 49, 52, and 57; group 9, 2 isolates, STs 9 and 118; group 10, 2 isolates; STs 5 and 23.

The environmental isolates (from mushroom compost, soil, and an oil-contaminated aquifer) were unrelated to each other but did cluster among the clinical isolates. In fact, mushroom compost isolates 1349 M (ST5, group 10) and 1346 M (ST41, group 8) clustered with clinical isolates from around the United Kingdom; and isolate 2359 (ST7, group 2), which was from a Canadian oil-contaminated aquifer, clustered with 11 clinical isolates from hospitals around the United Kingdom.

Isolates previously identified as members of the Liverpool, Manchester, Midlands, and Melbourne epidemic clones (Table 1) were found to be unrelated, sharing few if any alleles. Previously reported clone C was also unrelated to the United Kingdom epidemic isolates, although two isolates from the United Kingdom, isolates 8277 (ST14) and 8735 (ST17), from Durham and Birmingham, respectively, were identified here by MLST as belonging to clone C. In fact, from this small data set, ST17, the Birmingham isolate, was identified by BURST (based upon related STs) as the founder member of this small group of clone C isolates. BURST is a novel clustering algorithm designed for use with microbial MLST data. The approach specifically examines the relationships within clonal complexes.

The index of association (23) for all 143 isolates was found to be 0.288, and that for the 139 individual STs was found to be 0.17, indicating that P. aeruginosa has a nonclonal population structure. This statistical test attempts to measure the extent of linkage equilibrium within a population by quantifying the amount of recombination among a set of sequences and detecting associations between alleles at different loci. Comparisons of the topologies of neighbor-joining trees for the nucleotide sequences of individual loci (data not presented) revealed that there was little, if any, congruence between the trees. This is further evidence of the importance of recombination in the evolution of P. aeruginosa and indicates that the long-term phylogenetic inference of interstrain relationships, beyond the closely related groups identified, is relatively meaningless. For this reason we have not presented a dendrogram; however, as mentioned previously, the order of strains in Table 1 is the same as that derived from a UPGMA tree of allelic profiles (STs).

Relationship between ST, serotype, and toxA type.The serotypes of 118 of the isolates included in this study had been determined previously, and the toxA types had been determined for 38 isolates. Individual serotypes were found to be widely distributed across the dendrogram generated from the allelic profiles rather than solely associated with closely related clusters of strains (Table 1). Within the BURST groups of closely related isolates, 7 of 10 BURST groups possessed more than one serotype. Group 2 had five different serotypes among the nine isolates that had previously been serotyped, and the ST27 isolates from group 2 had four different serotypes.

Although fewer data were available for toxA types, a picture similar to that for serotypes was also found for toxA types. BURST group 2 contained multiple toxA types, and each of the five ST27 isolates possessed a different toxA type.

Both of these data sets reveal that there is a weak linkage between ST, serotype, and toxA type, which could be expected from the population structure. There is evidence that strains possess identical serotypes and toxA types but different STs (e.g., ST8277 and ST8735 members of clone C BURST group 2 and ST102, ST111, and ST113 members of BURST group 4), and there are examples of strains with identical STs but different serotypes and toxA types (e.g., members of ST27 in BURST group 2).