Taxon K, a complex within the Burkholderia cepacia complex, comprises at least two novel species, Burkholderia contaminans sp. nov. and Burkholderia lata sp. nov.

The Burkholderia cepacia complex (Bcc) comprises a group of closely related organisms that are very versatile (Coenye & Vandamme, 2003). The organisms can be exploited for biocontrol, bioremediation and plant-growth-promotion purposes, but their capacity as opportunistic bacteria to cause human infections, in particular in cystic fibrosis patients, hampers their use in these biotechnological applications (Parke & Gurian-Sherman, 2001). The taxonomy of the Bcc has evolved dramatically in the last few years (Coenye et al., 2001). Until recently, the Bcc consisted of nine species with validly published names, but the recognition of Burkholderia ubonensis and the description of an additional five novel species (Vanlaere et al., 2008) has increased the number of Bcc species to 15. Recent multicentre studies revealed several clusters of isolates that do not group with one of these 15 Bcc species. One diverse group of unclassified Bcc isolates is known as taxon K or group K (Baldwin et al., 2005; Dalmastri et al., 2005, 2007; Mahenthiralingam et al., 2006; Payne et al., 2005; Vermis et al., 2002) and includes bacteria from human and environmental sources isolated worldwide. These sources include sputum and blood samples of debilitated patients such as cystic fibrosis patients, reservoir water, river water and sediments, soil, roots, animals, industrial and pharmaceutical product contaminants and domestic and personal care products (Mahenthiralingam et al., 2006, 2008; P. Vandamme and A. Baldwin, unpublished data). Moreover, several cases of epidemic transmission of taxon K strains among cystic fibrosis patients have been reported (Campana et al., 2005; Cunha et al., 2007). The whole-genome sequences of two taxon K strains, namely Burkholderia sp. 383 (=LMG 22485) (GenBank accession numbers NC_007509, NC_007510 and NC_007511 for the three chromosomes) and the Burkholderia sp. metagenome SAR-1 (Venter et al., 2004) (NC_000028), are publicly available. In an earlier study, two taxon K isolates were classified as B. cepacia, based on DNA–DNA hybridization experiments (Vermis et al., 2002). However, recA gene and multilocus sequence analyses, which are known to correlate well with Bcc species status (Baldwin et al., 2005; Mahenthiralingam et al., 2000b), allocated these and other taxon K strains in a distinct clade, suggesting that they may represent a novel Bcc taxonomic group.

The aim of the present study was to re-examine the taxonomic position and structure of taxon K. For this purpose, a set of strains was examined by a traditional polyphasic taxonomic approach, by multilocus sequence typing (MLST) analysis and by analyses of the available whole-genome sequences.

Bacterial strains and growth conditions.
In our ongoing diversity studies of Bcc bacteria, about 150 isolates have been identified as taxon K by means of HaeIII-based recA gene restriction fragment length polymorphism (RFLP) PAGE or recA gene sequence analysis. Twenty of these isolates were selected and used in the present study (Table 1). All strains were routinely cultured on trypticase soy agar (TSA) and incubated at 28 °C unless indicated otherwise.

Table 1. Bcc taxon K isolates examined ATCC, American Type Culture Collection, Manassas, VA, USA; LMG, BCCM/LMG Bacteria Collection, Laboratorium voor Microbiologie, Universiteit Gent, Gent, Belgium; CCUG, Culture Collection, University of Göteborg, Department of Clinical Bacteriology, Göteborg, Sweden; NCIMB, National Collections of Industrial, Marine and Food Bacteria, Bucksburn, Aberdeen, Scotland, UK; NCTC, National Collection of Type Cultures, London, UK; R, Research Collection, Laboratorium voor Microbiologie, Universiteit Gent; other strain numbers are personal designations of the depositor. CF, Cystic fibrosis; ND, not determined.

DNA preparation.
For PCR and RFLP experiments, DNA was prepared by alkaline lysis as described previously (Storms et al., 2004). For DNA–DNA hybridization experiments and determination of the DNA base composition, DNA was prepared as described by Pitcher et al. (1989).

Bcc-specific recA gene PCR.
The recA gene (1040 bp) was amplified using primers BCR1 and BCR2 as described by Mahenthiralingam et al. (2000b).

recA RFLP and recA gene sequence analysis.
Amplified recA fragments were subjected to HaeIII-RFLP analysis (Mahenthiralingam et al., 2000b). Electrophoretic separation of the restriction fragments was performed by PAGE as described by Vanlaere et al. (2005). The restriction patterns were analysed using the BioNumerics 4.5 software package (Applied Maths) and compared with those of Bcc reference strains. recA sequence analysis was performed as described previously (Mahenthiralingam et al., 2000b). Multiple alignment was performed by using the CLUSTAL_X program (Thompson et al., 1997). The aligned sequences were analysed phylogenetically using the BioNumerics 4.5 software. Distances were calculated using the Jukes–Cantor algorithm. Phylogenetic trees based on the neighbour-joining method were constructed with bootstrap values of 1000 replications.

16S rRNA gene sequence analysis.
16S rRNA gene sequence analysis was performed as described previously (Coenye et al., 2001).

Determination of the DNA base composition.
DNA was enzymically degraded into nucleosides as described by Mesbah & Whitman (1989). The nucleoside mixture obtained was separated using a Waters Breeze HPLC system and XBridge Shield RP18 column thermostabilized at 37 °C. The solvent was 0.02 M NH₄H₂PO₄ (pH 4.0) with 1.5 % (v/v) acetonitrile. Non-methylated lambda phage DNA (Sigma) and Escherichia coli LMG 2093 DNA were used as calibration reference and control, respectively.

DNA–DNA hybridization experiments.
Hybridization reactions were performed with photobiotin-labelled probes in microplate wells as described before (Ezaki et al., 1989; Goris et al., 1998), using an HTS7000 Bio Assay Reader (Perkin-Elmer) for the fluorescence measurements. The hybridization temperature was 50 °C.

MLST analysis.
MLST analysis was performed as described before (Baldwin et al., 2005). A phylogenetic tree of concatenated sequences (2773 bp), including fragments of seven genes [atpD (443 bp), gltB (400 bp), gyrB (454 bp), recA (393 bp), lepA (397 bp), phaC (385 bp) and trpB (301 bp)] from each isolate was constructed based on the neighbour-joining method using MEGA software package version 3 (Baldwin et al., 2005; Kumar et al., 2004). The significance of branching within the trees was evaluated by bootstrap analysis of 1000 computer-generated trees. The program DnaSP version 4.1 () was used to calculate the mean number of nucleotide substitutions per site (i.e. the percentage divergence of concatenated allele sequences) between populations based on a Jukes–Cantor method (Naser et al., 2007; Rozas et al., 2003). The standard deviation was calculated to show how widely the values are spread in the dataset. Evidence for clonal or recombining populations was estimated by measuring the extent of linkage [using the standardized index of association (sI_A)] between alleles at different loci around the chromosome as described previously (Baldwin et al., 2005; Haubold & Hudson, 2000; Maynard Smith et al., 1993). An sI_A not significantly greater than 0 after 1000 computer randomizations suggests that a single species population (monophyletic) is in linkage equilibrium (freely recombining), while a population with an sI_A significantly greater than 0 (P<0.001) is considered to be in linkage disequilibrium (clonal) (Baldwin et al., 2005; Haubold & Hudson, 2000). Nucleotide sequences of each allele and allelic profile and sequence types for all strains analysed in this study are available from the Bcc MLST website (http://pubmlst.org/bcc/) developed by Keith Jolley and sited at the University of Oxford (Jolley et al., 2004).

Determination of shared gene content and average nucleotide identity (ANI) values.
At the time of writing, 12 Bcc strains have been included in whole-genome sequencing projects, of which seven whole genomes were used in the present study (unfinished genomes are indicated by asterisks): Burkholderia multivorans ATCC 17616* (=LMG 17588), Burkholderia cenocepacia IIIA J2315^T (=LMG 16656^T), Burkholderia vietnamiensis G4 (=LMG 22486), Burkholderia dolosa AU0158* (=R-3200), Burkholderia ambifaria AMMD^T (=LMG 19182^T) and Burkholderia sp. LMG 22485 and the Burkholderia sp. metagenome SAR-1 (Supplementary Table S1, available in IJSEM Online). In addition, six genomes of other Burkholderia strains and the genome of Cupriavidus metallidurans CH34^T were included for comparison with Bcc genomes. All genomes are publicly available and were obtained from the FTP site of the National Center for Biotechnology Information (NCBI) in their version of 25 August 2007, except for J2315^T, which was produced at the Sanger Institute and which was obtained from ftp://ftp.sanger.ac.uk/pub/pathogens/bc (Supplementary Table S1). The number of protein-coding (annotated) genes in each Bcc genome ranged from 5441 to 7717, with a mean gene length of 992.16 bp.

ANI values were calculated as described by Konstantinidis & Tiedje (2005a). Whole-genome sequences present an opportunity to produce a more robust and resolved species tree using concatenation of core gene sequences. The set of single-copy orthologous core genes was determined by building gene families using OrthoMCL (Li et al., 2003). Protein sequence alignments were generated in MUSCLE (Edgar, 2004), and poorly conserved regions were automatically trimmed using Gblocks (Castresana, 2000). Maximum-likelihood trees with 100 bootstrap replicates were obtained for a subset of 20 000 randomly selected amino acid positions from the concatenated single-copy core gene alignment, using the PhyML algorithm (Guindon & Gascuel, 2003) with a JTT amino acid substitution matrix (Jones et al., 1992) and a discrete gamma model.

Biochemical characterization.
A biochemical characterization was performed as described previously (Henry et al., 1997, 2001).

recA RFLP and recA gene sequence analysis
Different HaeIII recA RFLP patterns, designated K, AT, Se27, AY, Se35 and J, were found among the 20 isolates studied (Table 1). Pattern J has also been observed in B. cenocepacia, Burkholderia stabilis and Burkholderia arboris strains (Mahenthiralingam et al., 2000a; Vanlaere et al., 2008). Therefore, these isolates were further examined by MnlI recA RFLP PAGE, generating unique profiles for each species (data not shown). The 20 strains examined were subjected to recA sequence analysis. Similarity of the recA sequences towards established Bcc species was greater than 94 %, and the recA sequences of taxon K isolates constituted one cluster, including Burkholderia metallica, within the Bcc. Within taxon K, several distinct clusters, referred to as recA-I, recA-II and recA-III, were observed that showed sequence similarities of more than 97 % (Fig. 1).

(50K): Fig. 1. Phylogenetic tree derived from analysis of recA gene sequences of established Bcc species and taxon K strains. Different B. cenocepacia recA lineages are designated IIIA, IIIB, IIIC and IIID (Vandamme et al., 2003). The tree was constructed using the neighbour-joining method (Jukes & Cantor, 1969). Accession numbers are given in parentheses. Bootstrap values (≥50 %) are shown for 1000 replicates. The sequence of Burkholderia xenovorans LMG 21463T was used as outgroup. Bar, 1 % sequence dissimilarity.

16S rRNA gene sequencing and phylogenetic analysis
The 16S rRNA gene sequences of a representative strain from each recA cluster were examined (LMG 16227, LMG 22485 and R-15816 for clusters recA-I, recA-II and recA-III, respectively). Comparison of these 16S rRNA gene sequences with those of other Bcc strains revealed similarity levels above 98 % (data not shown). These values were similar to those obtained between strains of other Bcc species. Similarity levels of less than 97.5 % were calculated towards representatives of other Burkholderia species (data not shown).

DNA–DNA hybridization and determination of the DNA base composition
DNA–DNA hybridization experiments were performed between representatives of the three recA clusters, B. metallica LMG 24274 and B. cepacia LMG 1222^T (Supplementary Table S2). DNA–DNA hybridization experiments were also performed between recA-I strain LMG 23253 and type strains of the established members of the Bcc, including B. ubonensis. The mean value obtained among recA-I strains LMG 23361^T, LMG 23253 and LMG 23255 was 88 % and the value between recA-II strains LMG 22485^T and LMG 6863 was 79 %, whereas the value obtained between the recA-III strains LMG 14095 and LMG 24274 was 68 %. Moderate to high values, in the range of 59 to 64 %, were obtained between recA-I and recA-II strains. Moderate to high values of 60 and 66 % were obtained between recA-I and recA-III strains. Somewhat higher values, in the range 66–68 %, were obtained between recA-II and recA-III strains. Values in the range of 61 to 66 % were found between strains of the different recA clusters and B. cepacia LMG 1222^T and B. metallica LMG 24274 (Supplementary Table S2). Moderate values (42–62 %) were obtained between recA-I strain LMG 23253 and established Bcc members (data not shown). All strains investigated had G+C contents of 66–68 mol%, which is typical for Bcc species (Coenye et al., 2001) (Supplementary Table S2). The DNA–DNA hybridization values demonstrate that strains of the three recA clusters are closely related, but may constitute distinct novel species within the Bcc (Wayne et al., 1987). However, cluster recA-II and recA-III strains showed DNA–DNA hybridization values at the threshold for species delineation. Additionally, the two cluster recA-III strains examined also exhibited a DNA–DNA hybridization value near the 70 % threshold.

MLST analysis
Eighteen of 20 taxon K strains selected for the present study were analysed by MLST. Three of these strains consistently yielded incomplete MLST profiles (five genes for R-15816 and R-18442; six genes for LMG 6993), probably due to a sequence variation disrupting the binding site of the amplification primers. Nevertheless, analysis of the remaining loci identified them as taxon K strains (Supplementary Table S3). Data for an additional 21 taxon K strains examined previously and for the Burkholderia sp. metagenome SAR-1 (Baldwin et al., 2005; Mahenthiralingam et al., 2006; A. Baldwin, unpublished results) strain were also included in the present analyses (Supplementary Table S3). Phylogenetic analysis of concatenated sequences demonstrated that the isolates occupied a single clade in the MLST tree which consisted of two clusters, designated MLST-I and MLST-II, supported by bootstrap values of 100 and 92 %, respectively (Fig. 2). When these results were compared with results from the recA-based sequence analysis, cluster MLST-I comprised all recA-I strains examined, while cluster MLST-II comprised all recA-II and recA-III strains examined. Although the recA-II strains formed a distinct lineage supported by a 100 % bootstrap value within MLST-II, they grouped among recA-III strains. The intra- and intertaxon divergence of concatenated nucleotide sequences was calculated for both MLST clusters and for the other members of the Bcc (Fig. 3). Comparison of the percentage divergence of concatenated allele sequences between established Bcc members demonstrated that an average divergence of 3 % can clearly delineate Bcc species. The diversity within Bcc species varies from 0.43 % among B. vietnamiensis strains (based on analysis of 29 strains) to 2.89 % within Burkholderia anthina strains (based on analysis of nine strains). Although part of these differences can be attributed to the number of isolates examined for each species, it is clear that the genomes of different Bcc species do not evolve at the same rate. The internal diversity observed within cluster MLST-I was 1.4 %, while the value obtained within MLST-II was 2.7 %. A divergence value of 3.5 % was obtained when MLST-I was compared with MLST-II.

(17K): Fig. 2. Phylogenetic analysis of concatenated nucleotide sequences from seven loci, using the neighbour-joining method (performed using MEGA version 3 software). Bootstrap values greater than 70 % are shown for 1000 replicates. Species names of the established species are given and the numbers of STs included in the analysis are indicated in parentheses. Different B. cenocepacia lineages are designated IIIA, IIIB, IIIC and IIID (Vandamme et al., 2003). The different STs observed among the taxon K isolates examined are shown. Burkholderia pseudomallei was used as outgroup.

(28K): Fig. 3. Concatenated nucleotide sequence divergence of MLST loci. Shaded bars represent the divergence within a species. Open bars represent the divergence between a Bcc species and its closest neighbour (listed to the right) as defined by percentage similarity between concatenated sequences (similar to taxon gap as described by Naser et al., 2007). The number of strains included in the calculations for each species is given in parentheses. B. cenocepacia strains represent recA lineages IIIA (44), IIIB (57), IIIC (4) and IIID (2). For each bar, the SD is indicated by error bars. The cut-off value of 3 % divergence for species demarcation is marked.

The MLST data were also used to assign allele types to sequences and allelic profiles and sequence types (STs) to all isolates (Baldwin et al., 2005). Isolates with an identical allelic profile were assigned the same ST identifier and considered to be isogenic, as they were indistinguishable at all seven loci (Baldwin et al., 2005). The 37 taxon K isolates examined for which a full MLST dataset was obtained were resolved into 27 STs, with ST-102 occurring seven times, ST-98 occurring three times and ST-119 and ST-362 occurring twice. MLST-I and MLST-II strains shared no identical allele types; identical alleles were mainly found within a species and have been observed between the different B. cenocepacia recA subgroups (Baldwin et al., 2005). However, interspecies recombination events were recently reported (Waine et al., 2007). The sI_A value estimating clonality within the clusters MLST-I and MLST-II was 0.180 (P<0.01) and 0.440 (P<0.01), respectively (Baldwin et al., 2005; Haubold & Hudson, 2000). The sI_A values were thus significantly greater than zero, and MLST-I and MLST-II are therefore considered clonal populations. Nevertheless, the values are quite low, which may indicate that some recombination is occurring. The sI_A values found were comparable with values observed for B. cepacia strains (Baldwin et al., 2005). Studies of the Bcc have revealed that some species (e.g. B. vietnamiensis and B. cenocepacia) recombine more than others (e.g. B. ambifaria) (Waine et al., 2007). Nevertheless, the number of STs for each species is small and, therefore, a much larger sample is required for a more accurate comparison of mechanisms of evolution for each Bcc species.

Determination of shared gene content and ANI values
The availability of complete genome sequences of many bacteria provides new possibilities for comprehensive demarcation of species (Konstantinidis & Tiedje, 2005b). In the present study, two genome sequences of taxon K strains were available. Mahenthiralingam et al. (2006) reported previously that the Burkholderia sp. metagenome SAR-1 was essentially complete and represented taxon K, with the closest sequenced relative being Burkholderia sp. LMG 22485, a strain of cluster MLST-II. The SAR-1 metagenome belongs to ST102 and was genetically identical at multiple random loci when compared to cultivable ST102 isolates (Mahenthiralingam et al., 2006) which reside in MLST-I (Supplementary Table S3). The 100 % nucleotide identity indicated that the genomic sequence of the ST102 SAR-1 metagenome may be used as a surrogate whole-genome sequence for cultivable ST102 isolates such as Burkholderia sp. strain LMG 23361. A whole-genome-based comparison between the cluster MLST-I and MLST-II strains was therefore carried out and compared with genomes of other available Bcc species, non-Bcc Burkholderia species and members of related genera (listed in Supplementary Table S1).

Goris et al. (2007) demonstrated that the recommended cut-off of 70 % DNA–DNA reassociation for species delineation corresponds to 95 % ANI, a measure for evolutionary relatedness based on sequence similarity between orthologous genes (Goris et al., 2007). The similarity matrix of ANI values of the strains studied is shown in Supplementary Table S1. The ANI values obtained between different Bcc species ranged from 85.04 to 89.92 %, whereas the ANI values between Bcc and other Burkholderia species ranged from 75.33 to 81.68 %. The ANI value between the Burkholderia sp. metagenome SAR-1 and the whole-genome sequence of strain Burkholderia sp. LMG 22485 was 92.71 %. This high value confirms the similarity between the two sequences, but is clearly lower than 95 %, which corresponds to the established threshold for species delineation (Goris et al., 2007).

A set of 420 single-copy core genes, i.e. orthologous genes present in exactly one copy (no duplicates) in each studied genome, was used to estimate the phylogeny of these bacteria by maximum-likelihood (Supplementary Table S4). The phylogenetic analysis (Fig. 4) revealed that the metagenome SAR-1 and Burkholderia sp. strain 383 were closest neighbours, showing distances that were comparable with those observed between other Bcc species.

(33K): Fig. 4. Whole-genome-based maximum-likelihood tree for Bcc species and other Burkholderia species. The tree was derived from analysis of concatenated single-copy core genes, with C. metallidurans CH34T as an outgroup. Bootstrap analysis indicated that the tree was completely resolved.

Biochemical characterization
In general, species of the Bcc are phenotypically nearly indistinguishable, making their differentiation quite difficult, even with an extended panel of biochemical tests (Henry et al., 2001). The variable nature of biochemical results for Bcc strains may be due to their unusual genomes, which are among the largest seen in Gram-negative bacteria (6–9 Mb) and which are divided into a minimum of two chromosomal replicons (Lessie et al., 1996; Mahenthiralingam & Drevinek, 2007). The genomes are rich in insertion sequences, allowing the recruitment of foreign genes (Kenna et al., 2006). Mobile elements such as plasmids, transposons and bacteriophages are also frequently observed (Mahenthiralingam & Drevinek, 2007). The high foreign and mobile DNA content, in combination with the multireplicon structure, contributes to considerable plasticity and diversity in their genomes which, when differentially expressed in isolates, probably results in their variable biochemical phenotypic profiles.

Biochemical characteristics were determined for seven MLST-I strains and 11 MLST-II strains. A comparison of the biochemical characteristics revealed that these taxa could be distinguished by testing growth at 42 °C and haemolysis, for which strains of cluster MLST-I were mainly positive. Supplementary Table S5 lists the characteristics that are helpful in distinguishing the two taxa from each other and from established Bcc species. The biochemical characteristics of the novel species are given in the species descriptions below.

Conclusions
The aim of the present study was to re-examine the taxonomic position and structure of Bcc taxon K. For this purpose, a set of strains was examined by a traditional polyphasic taxonomic approach, by MLST analysis and by analysis of available whole-genome sequences.

Analysis of the recA gene sequences of 20 isolates selected from a collection of over 150 revealed that they represented three different lineages, designated recA-I, recA-II and recA-III (Fig. 1). Analysis of 16S rRNA genes and G+C contents of isolates representing the three groups confirmed that these bacteria are typical members of the Bcc, but was not helpful in discriminating the different lineages or in distinguishing them from other Bcc bacteria. DNA–DNA hybridization experiments among and between representatives of each lineage demonstrated that recA-I isolates represented a single novel species (Supplementary Table S2). recA-II strains represented a second novel species, but these strains showed DNA–DNA hybridization values towards recA-III strains that were at the threshold level for species delineation, i.e. around 70 % (Wayne et al., 1987). Similarly, the two recA-III strains examined exhibited DNA–DNA hybridization near the same threshold (Supplementary Table S2).

MLST analysis of 40 isolates, including 18 that were also analysed by recA sequence analysis, proved very helpful. recA-I isolates were again clearly distinguished from the others and represented a distinct lineage referred to as MLST-I (Fig. 2). recA-II and recA-III isolates formed a second MLST lineage referred to as MLST-II (Fig. 2). The level of concatenated sequence divergence within MLST-I and MLST-II was below 3 % (Fig. 3); by comparison with the level of concatenated sequence divergence in established Bcc species, these data demonstrate that the MLST-I and MLST-II lineages represent two distinct species within the Bcc. The latter conclusion was supported by comparison of the whole-genome ANI of MLST-I and MLST-II strains and of strains of established Bcc species (Supplementary Table S1) and by a whole-genome-based phylogenetic analysis (Fig. 4).

It is clear that the taxonomy and the genomic structure of these bacteria, which typically have genomes in the range of 8–9 Mb organized in multiple replicons, is extremely complex and that the classification of bacteria such as taxon K requires care. Nevertheless, they represent some of the most common members of the Bcc and require a formal classification and nomenclature. We feel that the data from the present study indicate that these bacteria are best classified into two novel species corresponding with the two MLST lineages. As for most Bcc bacteria, species-level identification is cumbersome by biochemical characteristics but can be accomplished readily through DNA-based procedures. In particular, sequence analysis of the recA gene, or of individual or all genes of the MLST scheme, generally yields a straightforward identification result. Nevertheless, in spite of the current 17 established species within the Bcc, several unidentified strains form additional clusters which probably represent novel Bcc species (unpublished data). Pending the availability of a reasonable set of such isolates and a full characterization thereof, the identification of some of these strains may be problematic. We propose to classify taxon K bacteria from the MLST-I and MLST-II lineages as Burkholderia contaminans sp. nov. and Burkholderia lata sp. nov., respectively. The MLST approach was confirmed as a valuable instrument in polyphasic taxonomic studies. More importantly, the cumulative data for some 971 Bcc isolates analysed (Mahenthiralingam et al., 2008; A. Baldwin, unpublished data) demonstrate that the 3 % concatenated sequence divergence correlates with the 70 % DNA–DNA hybridization or 95 % whole-genome ANI thresholds for species delineation.

Description of Burkholderia contaminans sp. nov.
Burkholderia contaminans [con.ta'mi.nans. L. part. adj. contaminans contaminating, polluting, referring to the metagenome that was recovered from the Sargasso Sea, but which probably represented a sample contaminant; ST102 isolates and other Bcc isolates grew very poorly in seawater, suggesting that the open ocean is not a natural habitat of Bcc species (Mahenthiralingam et al., 2006)].

Cells are Gram-negative, aerobic, non-sporulating rods. Colonies are moist, showing a metallic sheen. All known strains grow on MacConkey agar. Known strains grow on B. cepacia-selective agar and some strains turn that medium alkaline. Growth is observed at 30, 37 and 42 °C. Most strains are yellow pigmented and are haemolytic, a characteristic not commonly observed among Bcc species. The strains assimilate glucose, L-arabinose, D-mannose, D-mannitol, N-acetylglucosamine, D-gluconate, caprate, adipate, L-malate, citrate and phenylacetate, but not maltose. Acidification of glucose, maltose, lactose, xylose, sucrose and adonitol is observed. Nitrate reduction is strain dependent, but mostly not present. Activities of oxidase, β-galactosidase, aesculin hydrolase and lysine decarboxylase and gelatin liquefaction are present, but no ornithine decarboxylase, tryptophanase, arginine dihydrolase or urease activity. The G+C content is 67 mol%. Strains have mostly been isolated from clinical samples, but also from environmental samples. Strains have been involved in a widespread outbreak in the USA due to a contaminated nasal spray (Mahenthiralingam et al., 2006) and as contaminants in a water reservoir supplying a renal dialysis machine in Brazil (Souza et al., 2004).

The type strain is strain J2956^T (=LMG 23361^T =CCUG 55526^T) and was recovered from milk of a sheep with mastitis in Spain during an outbreak in 1999–2000 (Berriatua et al., 2001). The recA RFLP type and a false-positive PCR test first suggested that this isolate belonged to B. cenocepacia. The type strain is identical in 14 allele sequences to the metagenome SAR-1 and can be used as a cultured surrogate for research using the SAR-1 metagenome (Mahenthiralingam et al., 2006). Phenotypic and biochemical characteristics are as described above for the species.

Description of Burkholderia lata sp. nov.
Burkholderia lata (la'ta. L. fem. adj. lata broad, wide, because it is a common Bcc species worldwide).

Cells are Gram-negative, aerobic, non-sporulating rods. Colonies are moist. All known strains grow on MacConkey agar. Strains grow on B. cepacia-selective agar and some strains turn that medium alkaline. Growth is observed at 30 and 37 °C but not at 42 °C (except R-18628). Some strains are yellow pigmented or yellow–purple pigmented. No haemolysis (except R-15816). The strains assimilate D-glucose, D-mannose, D-mannitol, N-acetylglucosamine, D-gluconate, adipate, L-malate and citrate, whereas the assimilation of maltose, L-arabinose, caprate and phenylacetate is strain dependent. Acidification of D-glucose, maltose, lactose and xylose is observed. Acidification of sucrose and adonitol is strain dependent. Nitrate reduction is strain dependent. Activities of oxidase and lysine decarboxylase are present, but no tryptophanase, arginine dihydrolase or urease activity. Gelatin liquefaction, ornithine decarboxylase, aesculin hydrolase and β-galactosidase are strain dependent. The variability observed in the β-galactosidase test is rather uncommon among Bcc species, which mostly show β-galactosidase activity. The G+C content is 67 mol%. Most strains have been isolated from diverse environmental and industrial samples, with some isolated from clinical samples. Remarkably, strains identified as recA cluster K-II have only been recovered from environmental samples in Trinidad.

The type strain is strain 383^T (=ATCC 17760^T =LMG 22485^T =CCUG 55525^T) and was originally recovered from forest soil in Trinidad in 1958 (Stanier et al., 1966). Phenotypic and biochemical characteristics are as described above for the species.

E. V. is indebted to the Special Research Council of Ghent University. P. V. and D. G. are indebted to the Fund for Scientific Research Flanders (Belgium) for research grants and a post-doctoral fellowship, respectively. A. B., C. D. and E. M. are indebted to the Cystic Fibrosis Trust UK (grant PJ535) for funding their research. J. J. L. is supported by the Cystic Fibrosis Foundation (USA). D. S. and D. H. are supported by the Canadian Cystic Fibrosis Foundation. We are grateful to all depositors of strains who contributed to this study. We thank Jean Euzéby for his advice on the nomenclature of the novel species B. lata.