|Home | About | Journals | Submit | Contact Us | Français|
Phenazines are versatile secondary metabolites of bacterial origin that function in biological control of plant pathogens and contribute to the ecological fitness and pathogenicity of the producing strains. In this study, we employed a collection of 94 strains having various geographic, environmental, and clinical origins to study the distribution and evolution of phenazine genes in members of the genera Pseudomonas, Burkholderia, Pectobacterium, Brevibacterium, and Streptomyces. Our results confirmed the diversity of phenazine producers and revealed that most of them appear to be soil-dwelling and/or plant-associated species. Genome analyses and comparisons of phylogenies inferred from sequences of the key phenazine biosynthesis (phzF) and housekeeping (rrs, recA, rpoB, atpD, and gyrB) genes revealed that the evolution and dispersal of phenazine genes are driven by mechanisms ranging from conservation in Pseudomonas spp. to horizontal gene transfer in Burkholderia spp. and Pectobacterium spp. DNA extracted from cereal crop rhizospheres and screened for the presence of phzF contained sequences consistent with the presence of a diverse population of phenazine producers in commercial farm fields located in central Washington state, which provided the first evidence of United States soils enriched in indigenous phenazine-producing bacteria.
The naturally occurring phenazines include more than 50 nitrogen-containing heterocyclic pigments of bacterial origin (36). They have characteristic absorption spectra with two peaks in the UV range and at least one peak in the visible range that determines their colors (42). Almost all phenazines are broadly inhibitory to the growth of bacteria and fungi due to their ability to undergo cellular redox cycling in the presence of oxygen and reducing agents (including NADH and NADPH) and cause the accumulation of toxic superoxide and hydrogen peroxide (42). Phenazine-1-carboxylic acid (PCA), 2-hydroxyphenazine-1-carboxylic acid, and phenazine-1-carboxamide (PCN) produced in the rhizosphere by Pseudomonas fluorescens and Pseudomonas chlororaphis inhibit soilborne phytopathogenic fungi (13, 62) and contribute to the natural suppression of Fusarium wilt disease in certain soils in France (45). Phenazines produced by Pantoea agglomerans on apple flowers contribute to suppression of phytopathogenic Erwinia amylovora, which causes fire blight disease (22). Production of pyocyanin (PYO) by Pseudomonas aeruginosa is required for generation of disease symptoms in plants and killing of the nematode Caenorhabditis elegans and the fruit fly Drosophila melanogaster (40, 54), and it is also critical for lung infection by P. aeruginosa in mice (35). In contrast, some phenazines produced by Streptomyces spp. are not cytotoxic in eukaryotes and have promise as anticancer or anti-infective drugs (36).
In addition to the effects of phenazines on other organisms, recent studies have indicated that these compounds directly activate certain transcription factors and act as intercellular signals in P. aeruginosa (17, 18, 53). Because of their redox properties, phenazines also can function in the physiology of the strains that produce them by mediating the reoxidation of NADH under oxygen-limiting conditions, such as those found in mature biofilms (39, 53). In soil, phenazines can promote microbial mineral reduction and may function as electron shuttles, facilitating bacterial and plant access to iron and nutrients such as phosphate, trace minerals, and organic compounds associated with mineral phases (26). Several studies have demonstrated that phenazines are beneficial for the competitiveness and long-term survival of the producers in natural habitats. Strains of Pseudomonas that synthesized phenazines were more competitive and survived longer on the roots of wheat than non-phenazine-producing mutants (46). Similarly, PYO-deficient mutants were less competitive and less virulent than wild-type P. aeruginosa in mouse acute and chronic pneumonia infection models (35).
Over the past decade, significant progress has been made toward understanding the enzymatic steps leading to the assembly of the phenazine scaffold, but our knowledge of the diversity of phenazine-producing bacteria and the evolution of phenazine (Phz) biosynthesis pathways is limited. Except for the archaebacterium Methanosarcina mazei, which utilizes an unusual membrane-bound phenazine as an electron carrier in methanogenesis (1), phenazine biosynthesis is limited to bacteria (36, 64), and the ability to produce phenazines occurs in the Actinobacteria and two clades (Betaproteobacteria and Gammaproteobacteria) of Gram-negative Proteobacteria (42, 64). Previously, the evolution of the phenazine pathway was investigated in a single study that focused only on sequenced microbial genomes (19). This study found that phenazine genes have a complex evolutionary history and that horizontal gene transfer likely occurs. In the present study, we utilized improved phenazine gene-specific probes and a large collection of strains having diverse geographic, environmental, and clinical origins to assess the distribution of phenazine genes in economically important groups of bacteria and to gain insight into habitats that support phenazine producers. We also studied the evolution of the phenazine pathway in different groups of bacteria by carrying out genome analyses and comparing phylogenies inferred from sequences of the key phenazine biosynthesis (phzF) and housekeeping genes. Finally, we collected samples of cereal crops grown in central Washington state and screened them for the presence of phenazine producers. Our results (i) confirmed the extensive diversity of phenazine-producing bacteria; (ii) suggested that most phenazine producers are soil-dwelling and/or plant-associated species; (iii) revealed that different molecular mechanisms are involved in the evolution and dispersal of phenazine pathways in different lineages of bacteria; and (iv) revealed diversity among phenazine producers in commercial farm fields located in central Washington, which provided evidence of a United States soil enriched in indigenous phenazine-producing bacteria.
Bacterial strains used in this study are described in Table Table1.1. These bacteria included strains belonging to the genera Pseudomonas (n = 51), Burkholderia (n = 27), Brevibacterium (n = 3), Streptomyces (n = 1), and Pectobacterium (n = 12) and having environmental and clinical origins, as well as strains received from the American Type Culture Collection (ATCC) (Manassas, VA). All strains were grown at 28°C. Pseudomonas spp., Burkholderia spp., and Pectobacterium spp. were grown in Luria-Bertani (LB) broth (5). Brevibacterium spp. and Streptomyces cinnamonensis were cultured in nutrient broth (BD Biosciences, Franklin Lakes, NJ) and yeast extract-malt extract broth (BD Biosciences), respectively.
Total DNA was extracted from bacteria by using a cetyltrimethylammonium bromide (CTAB) miniprep procedure (5). For Brevibacterium spp. and S. cinnamonensis, the protocol was modified by pretreating the biomass for 30 min at room temperature with lysozyme (final concentration, 1 mg·ml−1) prior to addition of SDS. DNA concentrations were measured by using a DNA quantitation kit (Bio-Rad, Hercules, CA).
Oligonucleotide primers targeting conserved regions in genes of interest were designed with the Oligo v. 6.71 software (Molecular Biology Insights, Inc., West Cascade, CO). Amplification was performed with a PTC-200 gradient thermal cycler (Bio-Rad) using GoTaq DNA polymerase (Promega, Inc., Madison, WI). The annealing temperature (Table (Table2)2) was optimized for each primer pair and, if necessary, 5% dimethyl sulfoxide was added to PCR mixtures. Amplification products were cleaned with QIAquick PCR purification spin columns (Qiagen, Valencia, CA) and sequenced with a BigDye Terminator v. 3.1 cycle sequencing kit (Applied Biosystems, Foster City, CA) used according to the manufacturer's recommendations.
The following combinations of oligonucleotide primers were used to target phzF in different groups of bacteria: primers Ps_up1 and Ps_low1 for Pseudomonas spp., primers Bcep_up and Ps_low2 for Burkholderia spp., primers Ecar_up and Ecar_low and primers Paggl_up and Paggl_low for Pectobacterium spp., primers Br_up and Br_low for Brevibacterium spp., and primers Ps_up1 and Ps_low2 for Streptomyces spp. (Table (Table22).
Housekeeping genes for phylogenetic inference were amplified as follows. A single lower primer, recArps, was used in combination with upper primers recAf1, recAfcep, and recAfcoli to amplify recA in Pseudomonas, Burkholderia, and Pectobacterium, respectively. The homologous gene in Brevibacterium iodinum and S. cinnamonensis was targeted with primers recAfbr and recArbr (Table (Table2).2). Primers rpoBup and rpoBlow were used to amplify the 740- to 749-bp fragment of rpoB in P. aeruginosa, P. chlororaphis, Burkholderia spp., Pectobacterium carotovorum, B. iodinum, and S. cinnamonensis, while in the rest of the strains this gene was amplified with primers rpoBup1 and rpoBlow1. To amplify atpD, the lower primer atpDlow was used in combination with the upper primer atpDup1 for all species except Pectobacterium species, for which amplification was performed with the upper primer atpDup2 (Table (Table2).2). Finally, gyrB was amplified from the Gram-negative species included in this study (i.e., Pseudomonas spp., Burkholderia spp., and Pectobacterium spp.) with primers Up-1G− and Up-2G−, while PCRs with Gram-positive Brevibacterium and Streptomyces species were performed with primers Up-1G+ and Up-2G+ (Table (Table22).
Due to the length of the rrs amplicons, four internal primers were used for sequencing in addition to the upper and lower primers 8F and 1492R (66). The rrs amplicons from Pseudomonas spp. were sequenced with internal primers 16Sps1, r16Sps1, 16Sps2, and r16Sps2. The rrs PCR products from Burkholderia spp., Brevibacterium spp., P. carotovorum, and S. cinnamonensis were sequenced with internal primers 16Sps3, r16Sps3, 16Sps4, and r16Sps4 (Table (Table2).2). Sequence data were assembled and analyzed by using Vector NTI Advance v. 10 (Invitrogen Corp., Carlsbad, CA) and Omiga v. 2.0 (Accelrys, Inc., San Diego, CA).
Plants with adhering soil were collected in triplicate from global positioning system (GPS)-tagged sites (Table (Table3),3), placed in plastic bags, transported to the laboratory, and stored at 4°C for no more than 24 h before processing as described below. DNA was extracted from plant root washes with a PowerSoil DNA isolation kit (MO BIO Laboratories, Carlsbad, CA) using the alternative protocol for wet soil. Briefly, the root system with adhering rhizosphere soil was placed in 10 ml of sterile distilled water, vortexed, and sonicated, and then 2 ml of the soil suspension was used for DNA extraction as described in the manufacturer's protocol. Rhizosphere DNA was quantified and screened for the presence of indigenous phenazine-producing bacteria by performing PCR with phzF-specific primers essentially as described above. For cloning and sequencing, PCR products were separated by gel electrophoresis, extracted from agarose by using a QIAEX II kit (Qiagen, Valencia, CA), and cloned using a pGEM-T Easy cloning system (Promega), and randomly chosen recombinant clones were sequenced with the M13 forward primer.
Selected plant root washes were also used for isolation of indigenous phenazine-producing bacteria. Briefly, 100 μl of a root wash was serially diluted and plated on 0.1× tryptic soy (BD Biosciences) and 0.33× King's B (31) media amended with cycloheximide (100 μg ml−1). Plates were incubated at room temperature, and after 48 h representative morphotypes were replated and screened by performing PCR with phzF-specific primers. Positive isolates were stored at −80°C as glycerol stocks.
Genomic DNA samples from Burkholderia spp. were transferred onto BrightStar-Plus positively charged nylon (Ambion, Austin, TX) with a dot blot vacuum manifold, and the bound DNA was denatured, neutralized, and immobilized by baking essentially as described by Ausubel et al. (5). Three Burkholderia-specific phz hybridization probes were prepared from Burkholderia lata 383 by performing PCR with primer pairs cep11/phzAcelow, cep14/phzFcep_low, and cep17/cep10 (Table (Table2).2). Prior to labeling, PCR products were cleaned by using QIAquick PCR purification spin columns (Qiagen). Probe labeling, prehybridization and hybridization at 42°C, and subsequent detection of DNA-DNA hybrids were carried out with a DIG-High Prime II DNA labeling and detection starter kit (Roche Applied Science, Indianapolis, IN) used according to the manufacturer's recommendations.
Bacterial cultures were grown with shaking in LB or King's B broth for 48 h at 27°C, and phenazines were extracted with ethyl acetate as described previously (44). PCA extracted from P. fluorescens 2-79 was used as a standard. The extracts were dried, suspended in methanol, and spotted on silica gel plates (GHLF Uniplate; Analtech, Newark, DE). Chromatography was carried out as described by Thomashow et al. (63) using benzene-acetic acid (95:5), and chromatograms were examined at 254 nm.
The deduced protein sequences were aligned using Clustal X v.2.0.9 (33), and the resultant alignments were further edited with reference to protein folds predicted by the fold recognition program Phyre (8). The refined amino acid alignments were then used to align the corresponding nucleotide sequences using CodonAlign v.2 (25).
Molecular Evolutionary Genetics Analysis software (MEGA) version 4.0.2 (61) was used to infer trees based on neighbor joining of genetic distances (NJ) and maximum parsimony (MP). The NJ method was used with DNA and protein distances corrected by using the Kimura two-parameter (30) and Jones-Taylor-Thornton (29) models of evolution, respectively. For MP analyses, the close-neighbor interchange algorithm was used, and an initial tree was generated by random addition of the sequences. The reproducibility of clades in the inferred NJ and MP trees was assessed by bootstrap resampling with 1,000 replicates.
For maximum likelihood (ML) and Bayesian analyses, sequences for each locus were aligned and collapsed into haplotypes by removing indels and infinite-site violations with the Map program and phylogenetically incompatible sites with the Clade, Matrix, and CladeEx programs implemented in SNAP Workbench (52). Models of sequence evolution were evaluated for each data set and the combined data set using Modeltest v.3.7 implemented in PAUP v. 4.10b10 (60) and the Modeltest server (51). The Akaike information criterion (3) was used to select the most appropriate evolutionary model for each data set. ML phylogenies were estimated independently for each data set (phzF, rrs, atpD, gyrB, recA, and rpoB) and for the combined housekeeping gene data set (rrs, atpD, gyrB, recA, and rpoB) by using heuristic searches in PAUP. The reproducibility of clades was assessed by using heuristic searches of 1,000 bootstrapped data sets with “fast” stepwise addition of taxa and no branch swapping. To assess the suitability of combining the rrs, atpD, gyrB, recA, and rpoB data sets for phylogenetic analysis, a “conditional data combination” approach (27) was employed. Incongruence among partitions was assessed with a Shimodaira-Hasegawa (SH) test (57) implemented in PAUP using 1,000 bootstrapped replicates with RELL approximation of likelihood. The tests compared the topologies of the ML consensus trees estimated for each partition to the topology of the combined (concatenated) ML consensus tree.
ML phylogenies were also estimated for the combined and phzF data sets in a Bayesian framework using MrBayes version 3.1 (28). Flat Dirichlet probability densities were used as priors for the substitution rate parameters, and stationary nucleotide frequencies and uniform priors were used for the shape and topology parameters. An exponential unconstrained prior was used for the branch length parameter. Each analysis consisted of two runs of the Markov chain started from a random tree with 1,000,000 generations per run, six chains employed in each run, and random chains swapped three times per generation. Trees were sampled every 100 generations, and the first 50,000 generations of each analysis were discarded as burn-in. Average posterior probabilities were estimated for each node of the phylogeny across the two runs (19,000 trees from 2,000,000 total generations of the Markov Chain Monte Carlo).
Genome-scale analyses of phenazine biosynthesis locus organization were performed using the Pseudomonas (http://v2.pseudomonas.com/index.jsp) and Burkholderia (http://burkholderia.com/index.jsp) online genome databases, as well as the microbial genome resources section of GenBank (http://www.ncbi.nlm.nih.gov/genomes/lproks.cgi). Comparisons of genome regions flanking phenazine biosynthesis operons were performed with web-based implementation (WebACT) of the Artemis comparison tool (12).
Sequence data have been deposited in the GenBank database under the following accession numbers: FJ652597 through FJ652623 for rrs; FJ652624 through FJ652650 for recA; FJ652651 through FJ652677 for atpD; FJ652678 through FJ652704 for rpoB; FJ652705 through FJ652731 for gyrB; and FJ652732 through FJ652786 for phzF. Other GenBank entries used in this study include (i) the phzF gene from Streptomyces anulatus LU9663 (accession no. FN178498); (ii) rrs genes from P. chlororaphis ATCC 13985 (accession no. AF094722) and P. chlororaphis ATCC 9446 (accession no. AF094723); (iii) phenazine biosynthesis operons from P. fluorescens 2-79 (accession no. L48616), P. chlororaphis PCL1391 (accession no. AF195615), P. chlororaphis 30-84 (accession no. L48339 and AF007801), P. agglomerans Eh1087 (accession no. AF451953), and S. cinnamonensis DSM1042 (accession no. AM384985); (iv) sequenced microbial genomes of P. aeruginosa PAO1 (accession no. AE004091), P. aeruginosa PA7 (accession no. CP000744), P. aeruginosa UCBPP-PA14 (accession no. CP000438), P. aeruginosa 2192 (accession no. AAKW00000000), P. aeruginosa C3719 (accession no. AAKV00000000), P. aeruginosa LESB58 (accession no. FM209186), P. fluorescens Pf-5 (accession no. CP000076), B. lata 383 (accession no. CP000152), Burkholderia glumae BGR-1 (accession no. CP001504), Pectobacterium atrosepticum SCRI1043 (accession no. BX950851), Brevibacterium linens BL2 (accession no. AAGP00000000), and Nocardiopsis dassonvillei subsp. dassonvillei DSM 43111 (accession no. ABUI00000000).
A set of phenazine-specific PCR primers was developed based on a comparison of the phzF gene sequences from P. fluorescens 2-79, P. aeruginosa PAO1, B. lata 383, P. agglomerans Eh1087, P. atrosepticum SCRI1043, and B. linens BL2. The phzF gene encodes an enzyme involved in isomerization of the phenazine precursor trans-2,3-dihydro-3-hydroxyanthranilic acid to a highly reactive ketone (9). phzF is required for phenazine biosynthesis and is highly conserved in all known phenazine producers (42). PhzF is a member of a large protein superfamily that also includes diaminopimelate epimerases, proline racemases, and the 2-methylaconitate isomerase PrpF. However, because the constraints at the nucleotide level are much more relaxed than those at the amino acid level, it was possible to design primers specifically targeting enzymes involved in synthesis of phenazines. Initial efforts to develop “universal” oligonucleotide primers that would efficiently amplify phzF in all known phenazine-producing genera were unsuccessful due to the high degree of nucleotide variation among phzF alleles (data not shown). Instead, we designed genus-specific primer sets (Table (Table2)2) that have different sequences but target the same region of phzF. The 5′ ends of these primers complement positions 184 and 611 in the phzF DNA sequence of P. fluorescens 2-79. At the protein level the primers encompass two short stretches of invariant amino acids that are located adjacent to the active center residues His74 and Asp208 and form parts of the active site and intermonomer cavity.
The phzF-specific primers were utilized to screen DNA extracted from the rhizosphere of crops sampled in late 2007 and in 2008 and 2009 at 11 GPS-tagged sites located near Ritzville and Lind, WA (Table (Table3).3). The sites selected for sampling are wheat-producing commercial farm fields typical of central Washington state and are characterized by fine, light-colored, sandy silt loam loess-derived soils with very low humic content and annual precipitation of less than 300 mm. A winter wheat-summer fallow rotation is the dominant cropping system in this region. The results of the PCR-based screening revealed the presence of phenazine sequences in all of the locations and all of the crops sampled, including winter wheat, alfalfa, spring wheat, and spring barley (Table (Table33).
The new phzF-specific primers were also used to screen 94 known and presumptive phenazine-producing strains representative of economically important groups of bacteria from diverse environmental and clinical habitats. Primers Ps_up1 and Ps_low1 (Table (Table2)2) amplified the predicted 427-bp phzF fragment from all 51 pseudomonads tested, including strains of P. aeruginosa, P. chlororaphis subsp. chlororaphis, P. chlororaphis subsp. aureofaciens, and P. fluorescens and 24 Pseudomonas strains of uncertain taxonomic affiliation. The latter group included 15 new phenazine-producing isolates from the rhizosphere of wheat collected near Ritzville and Lind, WA (Table (Table1).1). Of 27 Burkholderia strains, 15 were positive for phzF (Table (Table1)1) as determined by PCR with the Burkholderia-specific primers Bcep_up and Ps_low2 (Table (Table2).2). Unexpectedly, Burkholderia phenazinium ATCC 33666, described as a strain that produces the phenazine iodinin (64), was negative when a PCR was performed with these primers. Because this strain is described as a phenazine producer, we also screened genomic DNA from all 27 Burkholderia strains by performing high-stringency dot blot hybridization with three DNA probes spanning different parts of the phz operon from B. lata 383. In all cases, strains that were negative when PCR was used (including ATCC 33666) also were negative when hybridization was used (data not shown). The phzF-negative Burkholderia strains were also tested for production of PCA by thin-layer chromatography. Except for B. phenazinium ATCC 33666, the only strain that was positive for PCA, the results of the thin-layer chromatography- and PCR-based screens were in complete agreement (data not shown).
Eleven Pectobacterium strains representing three species were screened by performing PCR with two primer sets based on phzF sequences from P. atrosepticum and P. agglomerans. Only P. atrosepticum SCRI 1043 and P. carotovorum cc303 were positive when PCR was performed with the Pectobacterium-specific primers Ecar_up and Ecar_low, and amplification with the Pantoea-specific primers Paggl_up and Paggl_low identified only one positive strain, Pectobacterium betavasculorum Ecb168 (Table (Table1).1). All phzF-negative Pectobacterium strains also were negative for production of PCA when thin-layer chromatography was used (data not shown). Finally, genomic DNA samples from Brevibacterium and Streptomyces strains were screened with the oligonucleotide primer pairs Br_up/Br_low and Ps_up1/Ps_low2, respectively (Table (Table2).2). The Br_up/Br_low primer set yielded amplified phzF fragments for all strains of B. iodinum tested, and primers Ps_up1 and Ps_low2 amplified DNA from S. cinnamonensis DSM1042 (Table (Table1).1). The amplified DNA fragments obtained from the strains were sequenced and used in phylogenetic analyses, as outlined below.
A phzF-based phylogeny was derived by using a comprehensive data set that was strictly limited to functional homologues (i.e., homologues that were present in sequenced genomes and experimentally characterized loci and encoded enzymes that form part of a complete phenazine biosynthesis pathway). The data set contained 57 phzF nucleotide sequences generated in this study together with alleles previously deposited in the GenBank databases. The nonredundant and environmental GenBank databases included 15 phenazine biosynthesis operons, 10 of which originated from partially or completely sequenced bacterial genomes. The DNA sequence of the phenazine operon from P. fluorescens S2P5 (GenBank accession no. AY960782) was identical to that of P. fluorescens 2-79 (44), and it was not included. We also identified a 2.4-kb contig (GenBank accession no. AACY020165889) in the Sargasso Sea metagenome that contained a phzF-like open reading frame. The gene, however, appeared not to be a part of the phenazine biosynthesis pathway and was not included in further analyses. The resultant combined phzF data set (n = 83) was subjected to NJ and MP analyses, and based on this initial evaluation of diversity, we chose representative strains from each cluster (Table (Table1,1, bold type) for further phylogenetic analyses and topological comparisons.
Phylograms inferred from Pseudomonas phzF sequences revealed four clades designated clades A1 through A4 (Fig. (Fig.1A).1A). Two of these clades, clades A1 and A4, were supported by high bootstrap values and contained phzF genes from strains of P. chlororaphis and P. aeruginosa, respectively. The A2 clade contained phzF genes from the model biocontrol strain P. fluorescens 2-79 together with phzF genes from phenazine-producing strains that we recently isolated from wheat grown in central Washington. Compared to clades A1 and A4, clade A2 was more diffuse, and it contained at least two subclades: one subclade formed by strains 2-79, R14-24-07, and R5-90-07 and a second subclade formed by strains R11-23-07, R4-X-07, and R2-X-07 (Fig. (Fig.1A).1A). The fourth clade, clade A3, included only sequences obtained from strains CMR12a and CMR5c, which were isolated from cocoyam (49).
We also correlated the four groups defined by phzF with the established systematics of Pseudomonas species by comparing the 16S rRNA gene sequences of the phenazine producers to those of type strains representing species complexes as defined by Anzai et al. (4). The results indicated that strains that formed clades A1, A3, and A4 fell into the “P. chlororaphis,” “Pseudomonas putida,” and “P. aeruginosa” species complexes, respectively (Fig. (Fig.1B).1B). The A2 clade corresponded to the “P. fluorescens” complex and again formed two subgroups, one corresponding to Pseudomonas synxantha, Pseudomonas libanensis, and Pseudomonas gessardii and the other corresponding to Pseudomonas orientalis.
Phylograms inferred from rrs (Fig. (Fig.2A)2A) and phzF (Fig. (Fig.3A)3A) data revealed that phenazine genes are carried by strains in three distinct rrs clades and are highly conserved in Burkholderia spp. According to 16S rRNA gene analyses, the phenazine-producing strains sequenced in this study fell into clade A and clustered with species of the Burkholderia cepacia complex (Fig. (Fig.2).2). This complex includes at least 17 species that occupy diverse ecological niches and can cause human, animal, or plant diseases (65). These species are closely related and are commonly identified based on recA sequence and multilocus sequence typing analyses. recA-based dissection of clade A revealed that the newly sequenced strains 5.5B and 2424 (together with very similar strains PC4, PC8, PC17, PC22, PC28, and PC39) are closely related to Burkholderia pyrrocinia and B. cepacia, respectively (Fig. (Fig.2B),2B), whereas strain 383, which was recently separated and placed in a new species called B. lata (65), is similar to Burkholderia contaminans and Burkholderia ambifaria. Clade B contained Burkholderia plantarii, Burkholderia gladioli, and B. glumae CIP106418, as well as the phenazine-producing strain B. glumae BGR-1 (Fig. (Fig.2A).2A). Clade C was represented by a single phenazine-producing species, B. phenazinium ATCC 33666, for which we confirmed phenazine production but were unable to isolate and sequence core phenazine genes (Fig. (Fig.2A2A).
Analysis of genome regions flanking phz operons revealed that phenazine genes in Burkholderia spp. are associated with unique insertion sequence (IS) composite transposons. In B. glumae BGR-1, the phenazine genes are on chromosome 2 and are flanked by 32-bp imperfect inverted repeats and direct copies of the IS402 transposase gene, one of which contains an internal frameshift (Fig. (Fig.4B).4B). The putative transposon also carries a full-length ISRsp9 transposase gene and a truncated gene encoding a transposase of the IS1421 family. The phz operon in B. lata 383 also is on chromosome 2 and is flanked by transposase genes of IS1202 (Fig. (Fig.4B).4B). The putative composite transposon is probably anchored in the genome since only the first transposase gene is intact and it is preceded by a 32-bp stretch of nucleotides that form the left end of IS1202 (http://www-is.biotoul.fr), whereas the second transposase gene is truncated and lacks the right transposon end. The putative “phenazine transposon” in B. lata 383 forms part of a unique 49-kb DNA segment that is not present in the genomes of the closely related phenazine-nonproducing strains B. ambifaria AMMD (accession no. CP000441) and B. ambifaria MC40-6 (accession no. CP001026) (see Fig. S1 in the supplemental material). The phz operon occupies the center of the segment and is flanked by regions with low G+C contents and at least two putative operons with unknown functions.
We also screened regions flanking the phz operons in other groups of bacteria to look for evidence of recent horizontal gene transfer events. Comparison of the regions flanking phenazine operons of P. aeruginosa to sequences in phenazine-nonproducing Pseudomonas spp. revealed that there were no extended stretches with significant similarity. Similar results were obtained when sequences of P. atrosepticum SCRI1043 and the closely related phenazine-nonproducing strain Erwinia tasmaniensis Et1/99 were compared (32).
To compare the phylogeny of phzF to that of the core bacterial genome, we estimated phylogeny using five highly conserved housekeeping genes: rrs, recA, rpoB, atpD, and gyrB. These genes function in unrelated metabolic processes and encode the RNA component of the 30S ribosome subunit, the homologous recombination and DNA repair protein recombinase A, the β subunit of bacterial DNA-directed RNA polymerase, the catalytic β subunit of the F0F1-ATP synthase complex, and the B subunit of DNA gyrase (topoisomerase II), respectively.
The topology of the phylogram inferred using the concatenated housekeeping genes agreed with the accepted phylogeny of the taxa included in our analysis (Fig. (Fig.3B).3B). Phylogenies estimated by different methods were highly congruent and revealed three major clades corresponding to members of the Gammaproteobacteria (Pseudomonas spp. and Pectobacterium spp.), Betaproteobacteria (Burkholderia spp.), and Actinomycetales (S. cinnamonensis, N. dassonvillei, and Brevibacterium spp.). In contrast, phylogenies inferred from phzF data were only partially congruent with the concatenated phylogeny (Fig. (Fig.3A).3A). Most notably, phzF genes from S. cinnamonensis DSM1042 and N. dassonvillei DSM43111 clustered with the Betaproteobacteria genes, whereas the homologues from P. carotovorum cc303 and P. atrosepticum SCRI1043 formed separate nodes.
Data sets with a reduced number of taxa (n = 13) were constructed to facilitate more detailed ML phylogenetic analyses and topological comparisons of phylogenies estimated using different genomic regions. Isolates indicated by bold type in Fig. Fig.33 were selected to represent each of the major clades revealed during preliminary distance analyses of all of the taxa sampled. The topologies of the combined housekeeping gene phylogenies estimated by the ML, Bayesian, parsimony, and distance methods were not significantly different (0.277 < P < 1.000). The topologies of the phzF trees estimated by the same four methods also were not significantly different (0.207 < P < 0.266). The topology of the combined ML phylogeny was compared to the topology of the phzF ML phylogeny by using SH tests with null distributions constructed in two ways. First, the phzF consensus topology was tested against the combined consensus topology with bootstrap resampling performed with the combined data set (885 characters). The second test involved comparison of the combined consensus topology with the phzF consensus topology with bootstrap resampling performed using the phzF data set (67 characters). Both tests demonstrated that the topology of the phzF phylogeny was significantly different (P = 0.000 and P < 0.013, respectively) from that of the combined phylogeny (see Fig. S2 in the supplemental material). The phylograms inferred from phzF data were also congruent with the phylograms inferred from the data for the phzD, phzE, and phzG core phenazine genes shared by all phenazine pathways and with the phylograms inferred from a concatenated data set for the four phz genes (data not shown).
Finally, we analyzed polymorphisms in phzF and the housekeeping genes in different groups of phenazine-producing bacteria. For all genera, the diversity of phzF sequences varied between 0.03 (Burkholderia) and 0.162 (Pseudomonas) and on average was slightly greater than the diversity of the corresponding housekeeping genes (see Table S1 in the supplemental material). The nucleotide diversity values for the housekeeping genes ranged from 0.003 (rrs from Burkholderia) to 0.173 (gyrB from Brevibacterium), and on average rrs and gyrB were the most and least conserved genes, respectively, for all taxa. The ratios of nonsynonymous substitutions to synonymous substitutions (dN/dS) exhibited a similar trend; the values obtained for phzF were higher (0.056 to 0.248) than those obtained for the housekeeping genes, for which the values were between 0.025 (atpD from Brevibacterium) and 0.085 (gyrB from Brevibacterium). All dN/dS ratios were <1, indicating that all genes, including phzF, were subjected to purifying selection (i.e., a form of natural selection that lowers the frequency of alleles with reduced fitness or viability in a population).
A comparison of the sequences for phz operons clearly revealed that in all species the core biosynthesis genes are clustered and highly conserved (Fig. (Fig.4A).4A). Minor variations in the organization of the core pathway include the absence of the 3-deoxy-d-arabino-heptulosonate-7-phosphate (DAHP) synthase-encoding gene phzC in the pathways of enteric bacteria and duplication of the phzA/B gene in pseudomonads and S. cinnamonensis DSM1042. Enzymes encoded by phzB, phzD, phzE, phzF, and phzG appear to be indispensable for synthesis of the phenazine scaffold and are present in all phenazine-producing species. Some bacteria, including S. cinnamonensis DSM1042 and strains of P. aeruginosa, contain two copies of the core pathway (23, 43). The core genes are often physically linked to various auxiliary genes encoding phenazine-modifying enzymes and efflux and regulatory proteins (42) (Fig. (Fig.4A).4A). Compared to the parameters for the core genes, the distribution, orientation, proximity, and location of phenazine-modifying genes are highly variable, which means that they are of little value for phylogenetic inference. The putative phenazine efflux transporters include members of the ABC (B. linens BL2 and P. atrosepticum SCRI1043), proton motive force-dependent RND (P. aeruginosa PAO1), and major facilitator (P. agglomerans Eh1087) superfamilies. Finally, in most Pseudomonas spp., phenazine operons also are coupled with dedicated regulatory genes. In P. chlororaphis, P. aeruginosa, Pseudomonas sp. CMR12a, and strains of the P. fluorescens complex, the regulator genes phzI and phzR form a quorum-sensing circuit that regulates production of phenazines in relation to population density (42; M. Höfte, personal communication). The phylogeny of the phzI and phzR genes has recently been reported (37) and is not described here.
The work described in this paper required development of new probes targeting phenazine biosynthesis genes because the probes described previously (43) were based on sequences only from Pseudomonas and thus had a limited detection range. Using the new primers targeting phzF, a core biosynthesis gene common to all known phenazine-producing species, we screened 94 strains representing five genera and generated 57 new phzF sequences that, together with previous data, confirmed the remarkable diversity of phenazine-producing bacteria and at the same time revealed that most of them appear to be soil-dwelling and/or plant-associated species (Table (Table1).1). The latter conclusion is supported by our discovery of indigenous phenazine-producing bacteria in the rhizospheres of cereals collected near Lind and Ritzville, WA (Table (Table3).3). This is the first example of soils from the United States enriched in indigenous phenazine-producing bacteria and only the second such example in the world after the Fusarium wilt-suppressive soil of the Châteaurenard area in Dijon, France (45). Preliminary characterization suggested that phenazine-producing bacteria belonging to the P. fluorescens species complex (Fig. (Fig.11 and and3)3) are abundant in the Lind and Ritzville soils (105 to 106 CFU/g of root [data not shown]). The analysis of DNA extracted from the rhizosphere of winter wheat also revealed the presence of a novel group of phenazine producers whose phzF genes form two distinct clades clustering with Actinobacteria and Burkholderia spp. (Fig. (Fig.3A).3A). Taken together, our findings suggest that phenazine producers may be ubiquitous and far more diverse in the environment than previously recognized.
Comparison of the various phenazine operons revealed high levels of similarity in the overall organization of the core biosynthesis genes and conservation of functions encoded by phzD, phzE, phzF, and phzG (Fig. (Fig.4A).4A). A few notable differences include the lack of a PhzC homologue in pathways in P. atrosepticum (based on interrogation of the SCRI1043 genome) and P. agglomerans and the presence in core operons of B. lata, B. glumae, B. linens, N. dassonvillei, P. atrosepticum, and P. agglomerans of only one copy of the phzA/B homologue. PhzC is a putative DAHP synthase that is thought to act at the branch point of the shikimic acid pathway, probably to ensure abundant chorismate for phenazine synthesis (42) (see Fig. S3 in the supplemental material). The fact that phzC is not present in the operons of P. atrosepticum and P. agglomerans may reflect alternative routes leading to chorismate pools in these species or may indicate a level of phenazine synthesis lower than that in Pseudomonas. The presence or absence of a second copy of the phzB gene (designated phzA) is more intriguing and may be linked to production of different primary phenazine products.
All natural phenazines are derived from PCA and/or phenazine-1,6-dicarboxylic acid (PDC) (see Fig. S3 in the supplemental material), and evidence obtained previously suggests that phenazine-producing bacteria can be grouped into organisms that produce only PCA and organisms that produce PCA plus PDC (21, 48). A key step in formation of the phenazine scaffold involves the condensation of two molecules of amino-cyclohexenone to form a tricyclic phenazine precursor via a reaction catalyzed by the small dimeric protein PhzB (2, 9) (see Fig. S3 in the supplemental material). Interestingly, all pseudomonads carry the phzA gene, which is very similar to phzB and encodes an enzymatically inactive paralogue. It has been suggested that this may lead to formation in vivo of a PhzA/B heterodimer with only one catalytically competent active center. Such an enzyme might allow rapid spontaneous oxidative decarboxylation of a transient tricyclic intermediate, resulting in formation of PCA instead of PDC. We hypothesize that the presence or absence of phzA is a global feature that separates pathways that form only PCA and pathways that form PCA plus PDC and ultimately predetermines the identity of phenazine compounds produced by various bacterial genera (Fig. (Fig.4A;4A; see Fig. S3 in the supplemental material). Indeed, the comparison of core biosynthesis genes and naturally occurring phenazines revealed that both confirmed (P. agglomerans Eh1087) and proposed (P. atrosepticum, Brevibacterium spp., and Burkholderia spp.) PDC producers lack the phzA homologue and produce 1,6-disubstituted phenazines (21, 64), whereas S. cinnamonensis and all Pseudomonas spp. carry both phzA and phzB and produce phenazines that are derived from PCA (42, 64).
In Streptomyces and some closely related taxa, enzymes encoded by certain core phenazine genes are “recycled” in other biosynthesis pathways. Through interrogation of sequenced bacterial genomes we discovered homologues of phzC, phzD, and phzE in Mycobacterium abscessus (GenBank accession no. CU458896), Streptomyces sp. Mg1 (GenBank accession no. DS570398), and Frankia sp. EAN1pec (GenBank accession no. CP000820). Because these species lack orthologues of phzB and phzF, we suggest that these genes are not involved in phenazine biosynthesis but instead function in the synthesis of uncharacterized secondary metabolites or derivatives of shikimic and trans-2,3-dihydro-3-hydroxyanthranilic acids. One such pathway was recently described for the marine actinomycete Micromonospora sp. 046Eco-11, in which homologues of phzC, phzD, and phzE were implicated in the production of a novel antimicrobial alkaloid, diazepinomicin (47).
Our investigation of the phylogeny of the core phenazine biosynthesis genes focused first on fluorescent pseudomonads, the best-characterized group of phenazine producers. Two trends were immediately apparent when the evolutionary history of phzF was compared to that of the corresponding housekeeping genes. First, the topological congruence among the inferred evolutionary trees (Fig. (Fig.11 and and3)3) suggests that acquisition of the phenazine genes by pseudomonads was an early event, possibly preceding speciation in this group of bacteria. Second, our results underscore the emerging taxonomic diversity of phenazine-producing pseudomonads. It has long been recognized that phenazine synthesis occurs predominantly in strains of P. chlororaphis, Pseudomonas aureofaciens (now classified as P. chlororaphis subsp. aureofaciens), and P. aeruginosa. However, our analyses of housekeeping genes suggest that newly isolated phenazine-producing members of the P. fluorescens and P. putida species complexes (clusters PF and PP in Fig. Fig.1)1) may belong to new species, raising questions about the actual number of Pseudomonas spp. capable of phenazine production. Further genetic and bioanalytical analyses are needed to clarify this issue.
Results of our analyses suggest that phz pathways may have evolved via horizontal gene transfer in some bacterial lineages. At least one event is ancient and best exemplified by the unequivocal clustering of the phzF genes from Streptomyces and Nocardiopsis (Gram-positive Actinobacteria) with their homologues from Pseudomonas spp. (Gram-negative Gammaproteobacteria) and Burkholderia spp. (Gram-negative Betaproteobacteria) (Fig. (Fig.3A;3A; see Fig. S2 in the supplemental material). This pattern agrees with the findings of Fitzpatrick (19) but is not consistent with the generally accepted systematics of these genera (15) or with the phylogeny inferred from data for housekeeping genes in this study (Fig. (Fig.3B;3B; see Fig. S2 in the supplemental material). In contrast to the situation in Actinobacteria, horizontal transfer of phenazine biosynthesis genes among species of Burkholderia, Pectobacterium, and Pantoea is probably still occurring and mediated by mobile genetic elements. Our results suggest that in Burkholderia spp. phenazine genes are dispersed by unique composite transposons similar to those found in the genomes of B. lata 383 and B. glumae BGR-1 (Fig. (Fig.4B).4B). This hypothesis is supported by the unusually high degree of sequence conservation and the patchy distribution of phz genes among different Burkholderia spp. Interestingly, in members of the B. cepacia complex, the phenazine genes seem to be carried predominantly by species that include environmental isolates that are not highly prevalent in cystic fibrosis patients (i.e., B. cepacia, B. pyrrocinia, and B. lata) (38). We did not detect phenazine genes in strains of Burkholderia cenocepacia, Burkholderia multivorans, and Burkholderia vietnamiensis, which represent the most common groups of cystic fibrosis pathogens (Table (Table1).1). In contrast to the situation in Burkholderia spp., analysis of phzF from phenazine-producing Pectobacterium spp. revealed two very divergent gene variants exemplified by the genes found in P. atrosepticum SCRI1043 and P. carotovorum cc303. The estimated genetic distance between the phzF genes from these strains is far greater than one would expect in such close relatives (Fig. (Fig.3).3). In another closely related strain, P. agglomerans Eh1087, the phenazine biosynthesis genes are carried by a large indigenous plasmid (21), and in P. atrosepticum SCRI1043 the phenazine operon was identified as a genomic island (7).
We speculate that the conservation of phenazine genes in some bacteria, versus the active dissemination via horizontal transfer in other bacteria, reflects differences in the roles that phenazines play in microorganisms. Bacteria with highly conserved phz genes are exemplified by Pseudomonas spp., in which the functions of the genes likely include not only signaling but also antibiosis and extracellular electron shuttling, which can mediate mineral acquisition and the reoxidation of NADH under high-density oxygen-limited conditions, such as those that occur in mature biofilms (17, 18, 26, 39, 53). Consequently, large quantities of phenazines (milligrams to grams per liter in vitro) are produced by pseudomonads, and the biosynthesis pathways are enmeshed in the metabolism and tightly regulated (42). At the opposite end of the spectrum are phenazine-producing species like P. agglomerans Eh1087, in which the phz genes are in a plasmid. The pathway in Eh1087 does not include a dedicated DAHP synthase, which probably results in lower phenazine yields. The main phenazine product, d-alanylgriseoluteic acid (AGA), acts in P. agglomerans like a typical antibiotic and is employed in competition with closely related bacterial species in its ecological niche (22). Interestingly, AGA itself is toxic to P. agglomerans Eh1087, and the phz pathway encodes a special phenazine-binding protein, EhpR, that prevents self-poisoning of the producer (21). The diversity and patchy distribution of phenazine-modifying genes may also reflect the connection between the types of phenazine compounds produced by individual species and their habitats.
We are grateful to Joyce E. Loper (USDA-ARS, Corvallis, OR), Monica Höfte (Ghent University, Ghent, Belgium), Gee W. Lau (University of Illinois at Urbana-Champaign, Urbana, IL), Kenichi Tsuchiya (Kyushu University, Fukuoka, Japan), and Shaji Philip (Rubber Research Institute of India, Kottayam, India) for providing bacterial strains used in this study and to Karen J. Adams for excellent technical assistance.
Published ahead of print on 11 December 2009.
†Supplemental material for this article may be found at http://aem.asm.org/.