Introduction

The genus Burkholderia represents a rapidly expanding group of Gram-negative non-fermenting bacteria that occur worldwide in virtually all possible environments. Some species occur in plain soil or in planktonic form in fresh water, but most occur in association with an ever-increasing number of hosts including humans, animals (both vertebrates and invertebrates), plants and fungi. The type of interaction with these hosts is often not known, but a growing body of literature demonstrates that these interactions can be beneficial, harmful or both. Within the genus Burkholderia, a cluster of closely related species is known as the Burkholderia cepacia complex (Bcc) and presents particular challenges. Bcc bacteria are indeed well-known as rare but potentially life-threatening pathogens in patients with cystic fibrosis (CF) but simultaneously have been studied intensively for their biotechnological applications in plant growth promotion, biological control of plant pests and bioremediation.

The present mini-review provides an update on the taxonomy of these bacteria and addresses recent developments in terms of their capacity for secondary metabolism and antibiotic biosynthesis. The interested reader is referred to recent reviews on Burkholderia and Bcc that focused on extracellular products (Vial et al. 2007), relevance as contaminants in the pharmaceutical industry (Torbeck et al. 2011), conflicting life styles (Vial et al. 2011), potential for aromatic compound degradation (Pérez-Pantoja et al. 2012), common characteristics of plant-associated species (Suarez-Moreno et al. 2012), melioidosis (Currie 2015) and to the book entitled ‘Burkholderia: from genomes to function’ (2014).

Taxonomy

The genus Burkholderia belongs to the Betaproteobacteria class within the phylum of the Proteobacteria. When first described in 1992, it comprised seven species, two of which were subsequently reclassified into another novel genus, Ralstonia (Yabuuchi et al. 1992; Yabuuchi et al. 1995). During the past two decades, a large number of novel Burkholderia species have been reported and validly named. Yet, several species proved poorly characterised and needed further reclassification (Coenye et al. 1999; Coenye et al. 2000); at present (January 2016), the genus consists of 90 validly named species (Parte 2014) and a large number of uncultivated candidate species (van Oevelen et al. 2004; Verstraete et al. 2011; Lemaire et al. 2011; Lemaire et al. 2012). However, literature data and an analysis of publicly available 16S rRNA gene sequences suggest that many additional Burkholderia species await formal description. In addition, mining the Bcc PubMLST database (Jolley and Maiden 2010) using the 3 % threshold value of average concatenated allele sequence divergence for species delineation (Vanlaere et al. 2009; Peeters et al. 2013) demonstrated that also within the Bcc, a substantial number of additional Burkholderia species awaits formal naming (Vandamme and Peeters 2014).

The genus Burkholderia is phylogenetically diverse and consists of multiple deep-branching 16S rRNA lineages (Fig. 1). The first deep-branching Burkholderia clade comprises the type species, Burkholderia cepacia, and consists of all Bcc species, a group of plant-pathogenic species that includes Burkholderia gladioli, Burkholderia plantarii and Burkholderia glumae and a group of species closely related to the risk class 3 pathogens Burkholderia mallei and Burkholderia pseudomallei, the causative agents of glanders in Equidae and melioidosis in humans, respectively (Fig. 1). This first deep-branching Burkholderia clade comprises the majority of well-known pathogens in this genus but also includes many strains that have been used for plant growth promotion or biological control, such as Burkholderia vietnamiensis TVV74 and Burkholderia ambifaria AMMDT, respectively (Parke and Gurian-Sherman 2001). The former is a rice isolate which, when inoculated in field studies on rice, increased grain yield by 13 to 22 % (Tran Van et al. 2000). The latter was isolated from the rhizosphere of peas and has activity against Pythium aphanidermatum (responsible for pre- and post-emergence damping-off in peas) and Aphanomyces euteiches (responsible for root rot in peas) (Parke 1990; Bowers and Parke 1993; Heungens and Parke 2000; Heungens and Parke 2001; Parke and Gurian-Sherman 2001). In addition, although Burkholderia cenocepacia is generally considered the most problematic Bcc species in patients with CF (Lipuma 2010), recently, a genome sequence of a plant-beneficial endophytic B. cenocepacia strain with both biocontrol and plant growth-promoting characteristics was reported (Ho and Huang 2015).

Fig 1
figure 1

Phylogenetic tree based on partial 16S rRNA gene sequences of Burkholderia species. Sequences (1125–1610 bp) were aligned against the SILVA SSU reference database using SINA v1.2.11 (http://www.arb-silva.de/aligner/) (Pruesse et al. 2012). Phylogenetic analysis was conducted using MEGA6 (Tamura et al. 2013). All positions containing gaps and missing data were eliminated, resulting in a total of 1087 positions in the final dataset. The optimal tree (highest log likelihood) was constructed using the maximum likelihood method and Tamura-Nei model (Tamura and Nei 1993). A discrete Gamma distribution was used to model evolutionary rate differences among sites (five categories [+G, parameter = 0.3498]) and allowed for some sites to be evolutionarily invariable ([+I], 68.6154 % sites). The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1000 replicates) are shown next to the branches if greater than 50 %. The sequence of Ralstonia solanacearum LMG 2299T was used as outgroup. The scale bar indicates the number of substitutions per site

Full size image

The second deep-branching Burkholderia clade comprises Burkholderia glathei (Zolg and Ottow 1975) and 11 recently named Burkholderia species (Fig. 1). Several of these species were isolated from polluted soils. For instance, Burkholderia udeis comprises naphthalene-degrading isolates from a polycyclic aromatic hydrocarbon-contaminated hillside soil (Wilson et al. 2003; Vandamme et al. 2013), while Burkholderia jiangsuensis and Burkholderia zhejiangensis are methyl parathion-degrading bacteria isolated from methyl parathion-contaminated soil and a wastewater treatment system, respectively (Lu et al. 2012; Liu et al. 2014). Other species in this clade have been isolated from less studied sources such as fungal mycelia and mosses (Burkholderia sordidicola and Burkholderia grimmiae, respectively) (Lim et al. 2003; Tian et al. 2013). This clade also includes isolates from insect guts (Kikuchi et al. 2011; Shibata et al. 2013) and several candidate species with an endophytic lifestyle in plants (Carlier and Eberl 2012; Verstraete et al. 2013). Although these bacteria show a remarkable diversity in terms of ecological niches, to our knowledge, none of the present species within this group has been involved in human or animal infections. However, at least two novel B. glathei group species have been isolated from human sources including pleural fluid and blood, and await formal classification (own unpublished data).

The third deep-branching Burkholderia clade comprises more than 40 primarily environmental and plant-associated species, many of which are diazotrophic and have been documented as beneficial to their host (Suarez-Moreno et al. 2012) (Fig. 1). Among these species, Burkholderia fungorum is a most striking exception, as it has been isolated from a wide range of human and veterinary samples including human blood, cerebrospinal fluid, vaginal secretions, sputum and lavage samples of CF patients, the brain of a pig with neurological deficit, the brain stem of an injured deer and the nose of mice (Coenye et al. 2001b; Coenye et al. 2002; Gerrits et al. 2005) (own unpublished data). In addition, Burkholderia tropica has been isolated from a neonatal patient with necrotizing enterocolitis and bowel perforation, who developed septicaemia (Deris et al. 2010).

In addition to these three main clades, several Burkholderia species represent unique deep-branching 16S rRNA lineages (Fig. 1). These include Burkholderia rhizoxinica and Burkholderia endofungorum (two endosymbionts of the plant-pathogenic fungus Rhizopus microsporus) (Partida-Martinez et al. 2007) and a group consisting of Burkholderia caryophylli (a pathogen of carnations and onions) (Ballard et al. 1970), Burkholderia symbiotica (a root nodule endosymbiont of Mimosa species) (Sheu et al. 2012) and Burkholderia soli (a soil bacterium) (Yoo et al. 2007). These species do not cluster closely with any other Burkholderia species and their 16S rRNA sequence-based phylogenetic position appears variable and dependent on the other taxa included in a phylogenetic analysis. Finally, Burkholderia andropogonis, a pathogen causing stripe disease of sorghum and leaf spot of velvet bean (Coenye et al. 2001a), clustered within the B. glathei group clade in the present analysis (Fig. 1); yet it often occupies a distinct position in 16S rRNA-based phylogenetic trees as well (see e.g. Sawana et al. 2014; Estrada-de los Santos et al. 2015).

Today, the availability of whole genome sequences enables phylogenetic analyses based on the entire part of the genome that is shared between organisms, a discipline referred to as phylogenomics (Yutin et al. 2012; Wang and Wu 2013). Although the number of Burkholderia species for which whole genome sequences are available is still limited, phylogenomics only partially reveals the same major subdivisions as the 16S rRNA tree (Fig. 1). Analysing the diversity across the Burkholderia genus using complete genome comparison is difficult given their inherent genetic diversity. Nevertheless, comparison of the 53 ribosomal protein-encoding genes within the current ribosomal multilocus sequence typing (rMLST) scheme provides a robust and biologically meaningful representation of the core Burkholderia genome (Fig. 2) (Jolley et al. 2012). A first well-supported rMLST Burkholderia clade also comprises the Bcc, plant-pathogenic and B. pseudomallei species groups, though with greater support for their divergence (Fig. 2). The second 16S rRNA sequence-based branch including B. glathei group species is also distinct when analysed by rMLST (Fig. 2). However, it clusters among species belonging to the third 16S rRNA-based lineage. In the rMLST analysis too, B. rhizoxinica and B. symbiotica occupy very distinct positions each. In contrast to what is observed in the 16S rRNA-based phylogeny, B. andropogonis represents a unique, very deep-branching lineage in the rMLST tree, illustrating its isolated taxonomic position. It will be interesting to see how other Burkholderia species will group when additional whole genome sequences become available.

Fig 2
figure 2

Burkholderia phylogeny reconstructed from concatenated ribosomal protein gene sequences. Aligned, concatenated gene sequences from defined ribosomal multilocus sequence typing (rMLST) loci were downloaded from the rMLST database at http://pubMLST.org (Jolley et al. 2012). Low confidence regions of the alignment were removed with Gblocks (Talavera and Castresana 2007), resulting in a total of 18,490 positions in the final dataset. A phylogeny was reconstructed with FastTree (Price et al. 2010) using the generalised time-reversible (GTR) model of nucleotide evolution, and the resulting tree was visualised with FigTree (http://tree.bio.ed.ac.uk/software/figtree). Type strains are indicated in bold type. Node confidence is shown if less than 80 %. The sequence of Ralstonia solanacearum PSI07 was used as outgroup. The scale bar indicates the number of substitutions per site

Full size image

The phylogenetic heterogeneity described above inspired several researchers to suggest (Gyaneshwar et al. 2011; Suarez-Moreno et al. 2012; Estrada-de los Santos et al. 2013; Angus et al. 2014) and eventually propose (Sawana et al. 2014) a taxonomic subdivision of the genus Burkholderia. Although it is clear that one can distinguish beneficial and harmful interactions of Burkholderia strains, there is no such phylogenetic subdivision in this genus. In a study of whole genome sequences of 45 Burkholderia strains representing some 25 formally named species and several unclassified strains, species belonging to the first Burkholderia clade were characterised by a percentage guanine plus cytosine content in their genomes of 65 to 69 %, while all other Burkholderia strains examined had a percentage guanine plus cytosine content in their genomes of 61 to 65 %. In addition, species belonging to the first Burkholderia clade shared six conserved sequence indels. The remaining Burkholderia strains represented species belonging to 16S rRNA clades 2 and 3 described above and one of the ungrouped species (B. rhizoxinica), and shared two conserved sequence indels. However, the phylogenetic diversity among the clade 2 and 3 species and B. rhizoxinica as revealed by 16S rRNA-based divergence and by differences in the distribution of 22 additional conserved sequence indels was ignored, as the authors proposed to restrict the name Burkholderia to 16S rRNA clade 1 species while reclassifying all other species into a single novel genus, Paraburkholderia (Sawana et al. 2014). These novel names were subsequently validated (Garrity and Oren 2015). Clearly, rMLST analysis is also supportive of greater evolutionary divergence between Burkholderia senso strictu (i.e. 16S rRNA clade 1 species) and Paraburkholderia but excludes B. rhizoxinica, B. symbiotica and B. andropogonis from the latter (Figs. 1 and 2). It will be up to the scientific community to adopt these novel names or not. According to the International Code of Nomenclature of Bacteria, researchers who are convinced that these name changes are ill-founded can continue to work with the original species names as these all were validly published.

Capacity for secondary metabolism: quorum sensing and antibiotic biosynthesis

Burkholderia bacteria are incredibly versatile organisms, with a phenomenal capacity for secondary metabolite production (Cimermancic et al. 2014), among which many with antifungal, antibacterial, herbicidal or insecticidal properties. Strains of Burkholderia are known to produce pyrrolnitrin (El-Banna and Winkelmann 1998), xylocandins (Meyers et al. 1987), cepafungins/glidobactins (Schellenberg et al. 2007), altericidins (Kirinuki et al. 1984), cepacins (Parker et al. 1984), cepaciamides (Jiao et al. 1996), phenazines (Cartwright et al. 1995) and quinoline derivatives (Moon et al. 1996). Although historical interest in Burkholderia secondary metabolites was largely focused on these primarily antifungal compounds, there is now growing evidence that Burkholderia also produce a range of potent antibacterial antibiotics, such as enacyloxin IIa (Mahenthiralingam et al. 2011). The following discussion will provide an overview of the state of the art on antimicrobial products from Burkholderia, as an update to the literature review by Vial et al. (2007) and genome-driven analysis of Liu and Cheng (2014). Biosynthesis of multiple Burkholderia antibiotics is controlled by quorum sensing (QS). For example, production of enacyloxin IIa and the resultant bioactivity of B. ambifaria AMMDT against B. multivorans is lost when the QS system is genetically disrupted (Mahenthiralingam et al. 2011) and the production of multiple other Burkholderia antibiotics is regulated in a similar way (Schmidt et al. 2009; Seyedsayamdost et al. 2010). Since QS and the signalling molecules produced by this process represent a major class of Burkholderia secondary metabolites, this process will be discussed first.

Quorum sensing in the genus Burkholderia

The expression of extracellular products is tightly regulated in bacteria. The immediate environment of the organism has a major influence on the production and secretion of extracellular products, but bacteria themselves also have a global regulation system in place to coordinate their behaviour. This QS system, as it is known, allows bacteria to alter their gene expression according to population density, as a form of cell-to-cell communication. In Gram-negative bacteria, N-acyl homoserine lactones (AHLs) are the most commonly used signal molecules, usually produced by an autoinducer synthase of the LuxI protein family and perceived by a transcriptional regulator belonging to the LuxR family (Whitehead et al. 2001). Bcc bacteria contain a LuxI/R type QS system, known as CepI/R, which was first discovered in B. cenocepacia K56-2 (Lewenza et al. 1999). The CepI AHL synthase is responsible for the production of N-octanoyl homoserine lactone and, as a minor by-product, N-hexanoyl homoserine lactone, whereas CepR acts as a transcriptional regulator. This CepI/R QS system is highly conserved among members of the Bcc (Sokol et al. 2007) and has been shown to regulate the production of a variety of extracellular products, including siderophores, fungicides and proteases (Lewenza et al. 1999; Zhou et al. 2003; Malott et al. 2005), as well numerous antibiotics, as discussed below.

Certain Bcc members harbour additional QS systems, such as BviI/R in B. vietnamiensis (Conway and Greenberg 2002) and CciI/R in B. cenocepacia strains belonging to the epidemic ET12 lineage (Malott et al. 2005). In addition, Boon et al. (2008) described another QS system in B. cenocepacia, which uses cis-2-dodecenoic acid (also known as Burkholderia diffusible signal factor or BDSF) as a signal molecule. BDSF is structurally similar to the diffusible signal factor (DSF) produced by the plant pathogen Xanthomonas campestris pv. campestris in order to regulate virulence. A homologue of the rpfF gene, the key enzyme in DSF biosynthesis, has been identified in B. cenocepacia and appears to be conserved throughout the Bcc (Deng et al. 2010). RpfF deletion mutants show reduced motility and adherence to porcine mucin, decreased extracellular polysaccharide production and diminished biofilm formation (Ryan et al. 2009). BDSF thus acts as an intraspecies signal in B. cenocepacia, yet it is also involved in interspecies and interkingdom communication, as antagonistic effects on Candida albicans have been observed (Boon et al. 2008).

Although QS has been less explored outside the Bcc, it appears to be widespread in the genus Burkholderia. For example, the plant pathogens B. glumae and B. plantarii are known to have QS systems similar to CepI/R, known as TofI/R and PlaI/R, respectively (Kim et al. 2004; Solis et al. 2006). Another distinct AHL-based QS system, designated BraI/R, is present in Burkholderia kururiensis and other members of 16S rRNA clade 3 (Suarez-Moreno et al. 2008). This QS system is similar to the LasI/R system found in Pseudomonas aeruginosa and relies on N-3-oxo-dodecanoyl homoserine lactone as main signal molecule. Despite the high conservation of the BraI/R system in this Burkholderia clade, most phenotypes (including biofilm formation, plant colonisation and degradation of aromatic compounds) seem to be regulated in a species-specific manner, suggesting that its role has evolved to suit the niche-specific needs of each species (Coutinho et al. 2013). Finally, several highly conserved and complex QS systems are present in members of the B. pseudomallei group. Both B. pseudomallei and Burkholderia thailandensis contain three complete QS circuits (QS-1, QS-2 and QS-3) and at least two orphan luxR homologues, whereas B. mallei has lost a large genomic region containing the QS-2 system through reductive evolution (Majerczyk et al. 2014).

Antibiotic biosynthesis by members of the Bcc

Members of the Bcc are known to produce multiple antimicrobial products, as described below. Pyrrolnitrin is a potent antifungal and antibacterial metabolite produced by strains of Burkholderia, Pseudomonas, Myxococcus, Serratia and Enterobacter (El-Banna and Winkelmann 1998), and it plays a role in the biocontrol activity of Bcc strains against phytopathogenic fungi such as Rhizoctonia solani and Fusarium spp. (Burkhead et al. 1994; Hwang et al. 2002). Pyrrolnitrin also inhibits growth of C. albicans and several Gram-positive bacteria, whereas Gram-negative organisms, except Proteus vulgaris, are not affected (El-Banna and Winkelmann 1998). A recent study determined the distribution of prnD, the gene responsible for the last step of pyrrolnitrin biosynthesis, within the genus Burkholderia (Schmidt et al. 2009). The pyrrolnitrin operon was found in strains belonging to eight Bcc species (Burkholderia pyrrocinia, B. cepacia, B. cenocepacia, B. ambifaria, Burkholderia ubonensis, Burkholderia lata and two novel Bcc groups) and in three species from the B. pseudomallei group (B. pseudomallei, B. thailandensis and Burkholderia oklahomensis). In addition, pyrrolnitrin production was shown to be under the control of the CepI/R QS system in B. lata 383T, as both cepI and cepR mutants lost inhibitory activity against R. solani, which could be restored in the cepI mutant through the addition of exogenous AHLs.

El-Banna and Winkelmann (1998) previously reported that glycerol strongly enhanced the production of pyrrolnitrin by B. cepacia NB-1. This finding was later confirmed in Burkholderia sp. O33, which produced increased amounts of pyrrolnitrin and polyhydroxyalkanoates in the presence of glycerol (Keum 2009). Based on these observations, Mahenthiralingam and colleagues used a basal salt medium, supplemented with glycerol as the only carbon source, to screen members of the Bcc for antimicrobial production (Mahenthiralingam et al. 2011). This led to the discovery that several B. ambifaria isolates show strong antimicrobial activity against pan-resistant Gram-negative pathogens, including Acinetobacter baumannii and two closely related Bcc species, B. multivorans and Burkholderia dolosa. The compounds responsible for this activity, enacyloxin IIa and the novel isomer cis-enacyloxin IIa, are produced by an unusual hybrid modular polyketide synthase (PKS) gene cluster. The fact that this cluster contains two orphan luxR-type homologues, disruption of which abolishes enacyloxin production, combined with the observation that cepI mutants no longer produce the enacyloxins, also indicates that QS plays a key role in regulation of this antibiotic biosynthesis cluster.

Another group of potent antifungals, named occidiofungins, was recently isolated from cultures of Burkholderia contaminans MS14 (Lu et al. 2009). These compounds display potent antifungal activity against a range of animal- and plant-pathogenic fungi, including Alternaria alternata, Aspergillus fumigatus, R. solani and several Phytium species. Occidiofungins are cyclic glycosylated oligopeptides, synthesised by a nonribosomal peptide synthetase (NRPS) and are structurally similar to the fungicidal xylocandins (Meyers et al. 1987) and the newly described burkholdines from B. ambifaria 2.2N (Tawfik et al. 2010). An NRPS gene cluster with close homology to the occidiofungin biosynthetic cluster was recently identified in B. ambifaria AMMDT and B. vietnamiensis DBO1. This gene cluster was responsible for hemolytic and insecticidal activity in B. vietnamiensis DBO1, suggesting that occidiofungins/burkholdines are not strictly fungicidal, but rather cytotoxic (Thomson and Dennis 2012). The synthesis of this group of bioactive oligopeptides is likely also QS-regulated, as the occidiofungin/burkholdine gene cluster was identified among several QS-controlled loci in B. ambifaria (Chapalain et al. 2013).

Finally, B. ambifaria is known to produce a range of bioactive volatile compounds that inhibit growth of the phytopathogenic fungi A. alternata and R. solani (Groenhagen et al. 2013). In addition to exhibiting fungicidal activity, the volatiles also increased biomass in the model plant Arabidopsis thaliana and were able to induce increased levels of antibiotic resistance in Escherichia coli.

An unidentified Burkholderia species related to the Bcc produces cepafungins, also known as glidobactins, which exhibit broad-spectrum antifungal and antitumor activities (Schellenberg et al. 2007). Similar gene clusters were found in B. pseudomallei, and glidobactins were recently identified as strong inhibitors of the eukaryotic proteasome (Groll et al. 2008).

Antibiotic biosynthesis by non-Bcc Burkholderia

Although a wide variety of antimicrobial compounds have been isolated from Bcc bacteria, several non-Bcc members of the genus Burkholderia are also known to exhibit antimicrobial activity. Species belonging to the B. pseudomallei group in particular appear to be excellent sources of antimicrobial natural products such as betulinans (Biggins et al. 2011), malleilactone/burkholderic acid (Franke et al. 2012; Biggins et al. 2012), thailandamides (Nguyen et al. 2008), thailandepsins (Wang et al. 2011), capistruin (Knappe et al. 2008) and bactobolins (Seyedsayamdost et al. 2010). Three of these bioactive compounds, capistruin, bactobolin and thailandamide, were discovered through genome mining of B. thailandensis E264T, which indicated the presence of a gene cluster involved in the synthesis of a lasso peptide, a type of ribosomally assembled bioactive peptide frequently isolated from Actinobacteria. The predicted molecule, capistruin, could be isolated from culture supernatant and demonstrated antibacterial activity against closely related Burkholderia and Pseudomonas strains (Knappe et al. 2008). Indication for the production of another antibacterial molecule came from a study investigating the role of the second QS system (QS-2) in B. thailandensis, encoded by BtaI2/R2, as these genes were found in clusters predicted to be involved in antibiotic biosynthesis. Supernatants from stationary-phase cultures of B. thailandensis E264T, but not a btaR2 mutant or a strain defective in AHL production, showed inhibitory activity against Gram-positive bacteria such as Bacillus subtilis, Staphylococcus aureus and Streptococcus pyogenes (Duerkop et al. 2009). This inhibitory activity was later attributed to a mixture of polar antibiotic compounds, known as bactobolins A–D, synthesised by a NRPS/PKS hybrid gene cluster (Seyedsayamdost et al. 2010).

The link between Burkholderia antibiotic production and QS was confirmed by Ishida et al. (2010). In this study, induction of thailandamide lactone A production in B. thailandensis was achieved by genetic manipulation of QS. Bioinformatic mining of the B. thailandensis E264T genome identified biosynthetic gene clusters which were congruent with the biosynthesis of polyketide metabolites. One of the metabolites, thailandamide A, was inconsistently detected in minute quantities within the supernatants of early-stage B. thailandensis cultures. Examination of the putative thailandamide biosynthetic cluster demonstrated the presence of an orphan luxR gene, encoding a transcriptional regulator (designated pthaA), which was phylogenetically distinct from other Burkholderia LuxR homologues. Disruption of the pthaA transcription factor binding region by mutagenesis resulted in a strain that accumulated large quantities of yellow pigment. This pigment was subsequently identified as thailandamide A and displayed antiproliferative, cytotoxic and anticancer activities (Ishida et al. 2010). Contrastingly, production of the cytotoxin malleilactone is regulated by an orphan luxR homologue, known as malR, which is not responsive to AHLs (Truong et al. 2015).

Identification and characterisation of bioactive natural products outside of the Bcc and B. pseudomallei group has not been as extensive. The plant pathogen B. plantarii produces tropolone, a compound with antibacterial, antifungal and phytotoxic properties (Azegami et al. 1987). The observation that a sesquiterpene signal molecule from Trichoderma virens PS1–7, a biocontrol agent of B. plantarii, represses tropolone production via transcriptional suppression of the AHL synthase plaI suggests that tropolone production is at least partially QS-regulated (Wang et al. 2013). Another plant pathogen, B. glumae, is known for the production of the phytotoxin toxoflavin (Jeong et al. 2003), which is regulated by the TofI/R QS system (Kim et al. 2004). Recently, another TofI/R-independent regulatory mechanism for toxoflavin production was discovered. Deletion mutants of this regulatory factor, tofM, produced lower levels of toxoflavin and showed reduced virulence, suggesting that tofM is a positive regulator of toxoflavin production (Chen et al. 2012). Besides this well-known phytotoxin, B. glumae also produces a bioactive pyrazole with antibacterial activity against Erwinia amylovora and several other Erwinia and Pseudomonas species (Mitchell et al. 2008). B. gladioli, a third plant pathogen in the genus Burkholderia, is known to produce several toxic and antimicrobial metabolites. The respiratory toxin, bongkrekic acid, associated with the fermented coconut-based Indonesian food, tempe bongkrek, was identified as the product of a polyketide biosynthesis cluster in B. gladioli pv. cocovenenans (Moebius et al. 2012). Tempe bongkrek is produced via fermentation by the mould Rhizopus oligosporus, which was initially thought to be the source of bongkrekic acid. Later, it was shown that B. gladioli pv. cocovenans, which was found as a contaminant of the fungal cultures and fermentations, was responsible for the production of this toxin (Moebius et al. 2012). Further evidence that B. gladioli can secrete bioactive molecules has been observed by Bharti et al. (2012), who noted that strain OR1 produces a range of, as yet uncharacterised, antimicrobial compounds with activity against Staphylococcus and Candida species. Finally, B. gladioli was shown to produce a polyketide of the enacyloxin family with antibacterial and antifungal activities when grown in co-culture with R. microsporus (Ross et al. 2014). B. rhizoxinica, an endosymbiont of the plant-pathogenic fungus R. microsporus, was also shown to be the source of a polyketide toxin, rhizoxin (Partida-Martinez and Hertweck 2005; Scherlach et al. 2012). Interkingdom interactions, in particular with plants and fungi, are key aspects of the natural biology of Burkholderia and it will be interesting to see if further bioactive natural products will be discovered as these complex environmental lifestyles are explored.

A genomic perspective on the capacity of Burkholderia for antibiotic biosynthesis

Greater access to complete microbial genome sequences facilitates the discovery of novel antibiotics via genome mining (Zerikly and Challis 2009). Several genomic approaches have been used to identify multiple antibiotics within Burkholderia, and particularly within the B. pseudomallei group (Fig. 1), as recently reviewed by Liu and Cheng (2014). In the last decade, hundreds of genomes were obtained for this group of potential bioterrorism agents, facilitating the application of genome mining for nondefence-related research. With the recent availability of genome sequences of other B. pseudomallei-related Burkholderia strains (Figs. 1 and 2), spanning the phylogenetic diversity of this group, it appears that the genomic capacity for antibiotic biosynthesis is an intrinsic feature of this group of organisms. Several bioinformatics tools are available for identifying secondary metabolite pathways within microbial genomes, the antibiotics Secondary Metabolite Analysis Shell (antiSMASH) being among the most advanced and well-curated (Blin et al. 2013; Weber et al. 2015). To provide a perspective on the genomic capacity for antibiotic biosynthesis within the genus Burkholderia, 15 complete genomes of strains representative of current diversity (Fig. 2) were analysed using antiSMASH 2.2.1 (Blin et al. 2013). The metrics for overall secondary metabolism and for the encoded PKS, NRPS, terpene and homoserine lactone biosynthetic capacities encoded in these 15 genomes are summarised in Table 1. The use of predictive biology software such as antiSMASH greatly accelerates the potential for novel secondary metabolite pathway discovery and was able to identify known PKS and NRPS pathways within the Burkholderia genomes analysed (Table 1). However, structure-pathway correlation and conventional chemical characterisation will still be required to characterise the wealth of potential bioactive molecules encoded by Burkholderia.

Table 1 Genomic capacity for antibiotic biosynthesis in Burkholderia
Full size table

A large genome size is a well-recognised feature of Burkholderia bacteria, which stand out in terms of their general functional versatility (Suarez-Moreno et al. 2012): the mean genome size of the strains included in the analysis is nearly 7 Mb (Table 1). The number of secondary metabolite clusters encoded within Burkholderia genomes varies greatly: for the 15 genomes analysed, a mean of 13 clusters was observed (range 7 to 22). On average, this equates to more than 7 % of the Burkholderia genome (>450 kb) being devoted to secondary metabolism. Another key feature of this genomic capacity for secondary metabolism is the substantial size of the encoded pathways, at averages of 49.5, 58.7, 21.8 and 18.3 kb for PKS, NRPS, terpene and homoserine lactone biosynthetic loci, respectively (Table 1). Given the size of the identified clusters and the current knowledge about large multimodular pathways such as those found in PKS and NRPS operons, the majority of pathways are likely to be complete, functional and able to synthesise complex compounds, provided their expression is activated. With complex regulatory controls involving QS (Schmidt et al. 2009; Ishida et al. 2010), inducing carbon sources (El-Banna and Winkelmann 1998; Mahenthiralingam et al. 2011) and as yet uncharacterised environmental interactions as potential stimulants, understanding how to activate cryptic antibiotic biosynthetic pathways within Burkholderia will be key to unlocking their biotechnological potential as a source of fine pharmaceuticals.

Liu and Cheng (2014) suggested that B. thailandensis E264T could be considered a champion Burkholderia in terms of its encoded capacity for natural product biosynthesis. From our preliminary analysis, the species within the B. pseudomallei group collectively encode the largest capacity for secondary metabolite biosynthesis (>11 % of their genomes). Evidence for significant antibiotic biosynthetic capacity is also present within the Bcc, with B. ambifaria currently leading the group, devoting over 9 % of its genome to secondary metabolism (Table 1). B. gladioli and B. glumae also dedicate 10 % or more of their genomes to antibiotic biosynthesis (Table 1). The potential of Burkholderia species belonging to the B. glathei and B. xenovorans clades (Fig. 1 and 2) appears less impressive, with fewer pathways and less than 5 % of their genomes in general being dedicated to antibiotic production (Table 1). However, within the latter group, fewer genomes have been characterised to date and the genomic distance between species is substantial as demonstrated by their deep-branching phylogenies, suggesting that greater diversity within these groups still remains to be characterised. B. rhizoxinica is an outlier, both in terms of its phylogenetic position (Figs. 1 and 2), as well as its capacity for antibiotic biosynthesis, which is substantial at 12.9 % of its relatively small endosymbiotic-adapted genome (Table 1).

In addition to the potential for the biosynthesis of antimicrobial compounds such as polyketide antibiotics and NRPS products, another highly conserved feature, observed across all Burkholderia genomes, is the significant potential for terpene production (Table 1). The well-known interactions of Burkholderia with plants (Suarez-Moreno et al. 2012) and the intrinsic ability of these bacteria to colonise the rhizosphere (Vidal-Quist et al. 2014) suggest that terpene biosynthesis and the potential interplay of these molecules during bacteria-plant interactions are areas worthy of future biotechnological exploration. Finally, another interesting feature of Burkholderia genomes is their organisation in a multireplicon structure. Species within the Bcc have a three replicon genome (Table 1) and can tolerate deletion of the smallest replicon (Agnoli et al. 2012). This results in strains which are highly attenuated in virulence as well as antibiotic production, especially in the case of B. ambifaria AMMDT, where this deletion results in the loss of the enacyloxin pathway, encoded on the third replicon (Mahenthiralingam et al. 2011). The ability to colonise the rhizosphere is not affected by deletion of this third chromosome in Bcc strains (Vidal-Quist et al. 2014), suggesting that in the future, it may be possible to engineer biological control strains that contain the antimicrobial pathways required for biopesticidal activity but lack the virulence pathways associated with Burkholderia pathogenicity.

Conclusion

Burkholderia continue to fascinate as a diverse group of Gram-negative bacteria. They are arguably better known as primary pathogens such as B. pseudomallei, opportunistic pathogens such as members of the B. cepacia complex and plant pathogens such as B. glumae. However, the diverse environmental interactions of these bacteria are now pointing towards multiple beneficial properties extending beyond their known capacities for bioremediation and biological control, towards the significant biotechnological potential as antibiotic producers, as illustrated within this review. For most antimicrobial metabolites, the function in the natural environment is still not known. However, with insights into the ecological relevance of antibiotics, we could take advantage of the multifunctionality of these natural products, although positive exploitation of Burkholderia as biotechnological agents will have to balance against their potential pathogenicity. With the ability to rapidly define the entire functional content of Burkholderia strains using genomics, future exploitation of these organisms for biotechnological purposes will be greatly accelerated.