Identification of WLIP biosynthetic genes in P. ‘reactans’ LMG 5329
Random mutagenesis of strain LMG 5329 was carried out with the Tn5-delivering suicide vector pLG221 (Boulnois et al., 1985). Kanamycin-resistant mutants were screened for a lack of WLR against P. tolaasii CH36 (Rokni-Zadeh et al., 2012). For seven WLR-negative mutants, single genomic insertion of Tn5 was confirmed by Southern hybridization (data not shown). Cloning of the DNA flanking the insertion sites by inverse PCR (Martin and Mohn, 2002) was achieved for four of these mutants and subsequent sequence analysis enabled identification of the inactivated genes in a draft genome sequence of strain LMG 5329. This was obtained by subjecting its genomic DNA, isolated using the Gentra Puregene Yeast/Bact. Kit (Qiagen), to 100-cycle pair-end sequencing with an Illumina HiSeq 2000 (Genomics Core Facility, KU Leuven). Assembly through Velvet (Zerbino and Birney, 2008) yielded 293 contigs (total size ∼6.7 Mb) that were annotated automatically using RAST (Aziz et al., 2008). For the contigs containing the genes disrupted in the mutants, annotations were verified by FramePlot protein-coding prediction (Ishikawa and Hotta, 1999), BLAST homology searches and Pfam domain analysis.
Analysis of the WLR-minus transposon mutants (CMPG2233-CMPG2236) enabled to locate the wip (WLIP production) genes (wipA and tandemly organized wipB-wipC) of strain LMG 5329, encoding its three WLIP-biosynthetic NRPSs, in two unlinked genomic regions [Fig. 1; GenBank accession numbers JQ974025 (wipA) and JQ974026 (wipBC)]. Assuming consecutive co-linear biosynthesis by WipA, WipB and WipC, the predicted lipopeptide product matches the nonapeptide sequence of WLIP. The phenotypes associated with WLIP production in P. putida RW10S2, namely anti-Xanthomonas activity, haemolysis, swarming and biofilm formation (Rokni-Zadeh et al., 2012), were also identified for strain LMG 5329 wild type and found to be abolished in the LMG 5329 mutants with disrupted WLIP-biosynthetic genes (Fig. 1). Biofilm formation capacity of strain LMG 5329 on a polystyrene surface, quantified as described previously (Rokni-Zadeh et al., 2012), was comparable with the one of P. putida RW10S2. This capacity was strongly reduced in the LMG 5329 NRPS mutants: CMPG2233 (10.09 ± 0.04), CMPG2234 (9.17 ± 0.03), CMPG2235 (11.21 ± 0.03) and CMPG2236 (9.27 ± 0.05), with values in percentage relative to the wild type (set to 100%). This residual biofilm formation of on average 10% is similar to the ∼10-fold reduction observed for P. putida RW10S2 upon inactivation of its WLIP production (Rokni-Zadeh et al., 2012).
Figure 1. Organization of the WLIP biosynthetic genes and WLIP-dependent phenotypes of P. ‘reactans’ LMG 5329. The predicted NRPS domains of WipA, WipB and WipC (labelled circles) are indicated: C, condensation; A, adenylation; T, thiolation; TE, thioesterase. C1 represents the condensation domain of the initiatory module. The amino acid specificity of the modules inferred from A-domain analysis (NRPSpredictor2; Röttig et al., 2011) is shown. Transposon insertion sites in NRPS mutants lacking WLIP production are marked with vertical bars. Mutant phenotypes are compared with those of the wild type (WT): (A) WLR upon confrontation with P. tolaasii CH36 (upper bacterial streak), (B) antagonism against Xanthomonas citri pv. malvacearum LMG 761 (growth-inhibitory halo formation by spotted LMG 5329 cells in a Xanthomonas overlay), (C) haemolysis (halo formation in blood agar plates), (D) swarming on 0.8% tryptic soy agar and (E) biofilm formation on polystyrene pegs visualized by staining of adherent cells (corresponding quantitative data in the text). By homology to other similarly organized lipopeptide-associated genes in Pseudomonas (Rokni-Zadeh et al., 2012), the linked wipR and wipD-wipE-wipF genes are predicted to encode a cognate LuxR-type regulator and a tripartite efflux transporter (composed of MacA, MacB and OprM homologues) respectively.
Download figure to PowerPoint
The modular structure and gene organization of wipA (with upstream divergent wipR-wipF genes) and of unlinked wipB-wipC (with downstream convergent wipD-wipE pair) are quite similar to those reported for the NRPS gene clusters of viscosin (P. fluorescens SBW25 visc genes; de Bruijn et al., 2007), massetolide (P. fluorescens SS101 mass genes; de Bruijn et al., 2008) and WLIP (P. putida RW10S2 wlp genes; Rokni-Zadeh et al., 2012). The relative positions of the respective WLIP biosynthetic clusters could not be determined in the draft genomes of strains LMG 5329 and RW10S2. The similarly organized clusters for viscosin and massetolide production are separated by ∼1.6 Mb in strain SBW25 (Silby et al., 2009) and by ∼1.4 Mb in strain SS101 (Loper et al., 2012).
The Wip NRPSs are very similar to the Visc enzymes (∼90% amino acid identity) and the Mass enzymes (82–84% amino acid identity) (Table S1). Such level of homology might be expected for these enzymes with only minor differences in catalytic activities, confined to the respective second NRPS enzyme: compared with ViscB, WlpB incorporates D-Leu instead of L-Leu (fifth module) and MassB attaches D-allo-Ile instead of D-Val (fourth module). Most remarkably, however, the wlp-encoded NRPSs share only about 60–65% amino acid identity with their wip-encoded counterparts (Table S1), although both systems synthesize the same lipopeptide. Comparison of the linked genes predicted to encode a LuxR-type regulator (WipR) and a tripartite export system (WipD-WipE-WipF; homologues of MacA, MacB and OprM respectively) reveals the same unexpected trend of much higher similarity between components of the Wip and Visc/Mass systems (producing structural variants) compared with the rather low similarity between those of the Wip and Wlp systems (producing the same secondary metabolite) (Table S1). Comparison of the amino acid identities of the predicted export proteins between the Wip and Wlp systems reveals a higher level of homology than observed for the NRPSs: 67%, 75% and 81% amino acid identity for the outer membrane proteins (WipF/WlpF), periplasmic adaptor proteins (WipD/WlpD) and inner membrane ABC transporters (WipE/WlpE) respectively. This better conservation may be due in part to structural requirements imposed by the assembly of the three export modules in the cell envelope. It is not known whether this type of lipopeptide-associated system, coexpressed with the cognate biosynthetic genes (Lim et al., 2009; Rokni-Zadeh et al., 2012), may also export other substrates or assist in efflux. Mutational analysis of the export genes downstream of the putisolvin (Dubern et al., 2008), arthrofactin (Lim et al., 2009) and syringopeptin (Cho and Kang, 2012) biosynthetic operons revealed a strongly reduced but not abolished lipopeptide production, indicating the existence of alternative export routes. For WLIP production by strain RW10S2 a functional linked LuxR-family regulator (WlpR) is essential (Rokni-Zadeh et al., 2012), which also applies to its counterparts involved in production of syringofactin (Berti et al., 2007), putisolvin (Dubern et al., 2008), viscosin (de Bruijn and Raaijmakers, 2009), arthrofactin (Washio et al., 2010) and entolysin (Vallet-Gely et al., 2010). The low degree of amino acid sequence conservation between the cognate LuxR-family regulators WipR and WlpR (55% identity; Table S1) suggests that some environmental cues controlling WLIP production may differ between the P. putida RW10S2 and P. fluorescens LMG 5329, possibly reflecting adaptation to the different niches of origin (wetland rice rhizosphere and cultivated mushroom respectively).
Inspection of the genomic DNA regions flanking the wip and visc clusters revealed a pronounced synteny of highly homologous genes (Fig. 2). This is most prominent for the genes down- and upstream of the wipBC/viscBC genes. In the latter regions several very similarly organized pyoverdine genes are located (Fig. S1). Homologues of some of these pvd genes are equally positioned upstream of the wlpBC cluster, but in a different organization and encoding gene products with a significantly lower pairwise amino acid sequence identity (Table S2). Whereas no pyoverdine NRPS is encoded by the wlpBC-upstream region, the wipBC-upstream region contains three genes for such enzymes (orf6, orf7, orf8) compared with two genes (pvdI, pvdJ) located upstream of viscBC. Orf6 is a homologue of SBW25 PvdI and Orf8 is similar to SBW25 PvdJ, but contains an extra N-terminal module. With involvement of the additional bimodular Orf7, the LMG 5329 pyoverdine would contain a peptidic backbone of 10 amino acids. Elucidation of the SBW25 pyoverdine structure has revealed seven amino acids: D-Ser–L-Lys–Gly–L-foOHOrn–L-Lys–D-foOHOrn–L-Ser (FoOHOrn = N5-formyl-N5-hydroxyornithine; Moon et al., 2008). For strain LMG 5329, the predicted peptide backbone contains three extra amino acids inserted between L-foOHOrn and L-Lys (Fig. S1). This matches the peptide chain of P. fluorescens strain 18.1 pyoverdine (Kilz et al., 1999). Munsch and colleagues (2000) assigned strain LMG 5329 to siderovar 4 among P. ‘reactans’ isolates. The capability of this siderovar type to incorporate iron mediated by the strain 18.1 pyoverdine supports the proposed LMG 5329 pyoverdine structure.
Figure 2. Genomic context of wlp, wip and visc regions in P. putida RW10S2, P. ‘reactans’ LMG 5329 and P. fluorescens SBW25 respectively. The two unlinked genomic regions are depicted separately (A with lipoinitiating NRPS; B with middle and terminating NRPSs). Dotted arrows represent the common genes assigned to the WLIP and viscosin systems. Flanking genes are not specified except for the pyoverdine biosynthesis (pvd) genes that are labelled with A, E, F, M, N, O and P. Organization and functions of the pyoverdine genes are further described in Table S2 and Fig. S1. Pairwise homology among flanking genes between strains is indicated by differential shading of the corresponding arrows: white (unrelated gene products), light grey (∼75% amino acid identity) or dark grey (> 90% amino acid identity). Synteny among flanking genes or gene clusters is delineated with dashed lines. Diagonal line-filled boxes correspond to parts of the C1 and TE domains for which PCR amplicons were generated from WLR-positive strains (Rokni-Zadeh et al., 2011). In C, these strains are grouped according to pairwise identity of amplicon-deduced amino acid sequences to either the RW10S2 Wlp proteins (Pseudomonas sp. RW3S1) or the LMG 5329 Wip proteins (P. ‘reactans’ LMG 2338, P. extremorientalis LMG 19695, Pseudomonas spp. PGSB3962, PGSB7828, PGSB8273).
Download figure to PowerPoint
No synteny is apparent for the wlpBC-downstream and wlpA-flanking regions with the equivalent wip/visc genomic regions. Also in the viscA and wipA flanking regions many homologues are shared by strains LMG 5329 and SBW25, although in the former several additional interspersed genes are present.
Taxonomic assignment of P. ‘reactans’ LMG 5329
The relatedness among the host strains carrying the visc, wlp and wip genes was further examined by phylogenetic analysis of 20 housekeeping genes (Fig. 3). WLIP producer LMG 5329, isolated from mushroom, grouped tightly with viscosin producer SBW25, representative of subclade 3 within the P. fluorescens branch (Loper et al., 2012). Conversely, the other WLIP producer, rice rhizosphere isolate RW10S2, is clearly associated with the P. putida–Pseudomonas entomophila cluster. The insect pathogen P. entomophila is a close relative of P. putida (Silby et al., 2011; Mulet et al., 2012). A similar clustering pattern was observed in the corresponding 16S rRNA phylogenetic tree (Fig. S2; 16S rRNA sequence of LMG 5329 submitted to GenBank under accession number JQ974027). These analyses identify P. ‘reactans’ LMG 5329 as an authentic P. fluorescens strain and further substantiate its phylogenetic distance from another WLIP producer, P. putida RW10S2, already inferred from comparison of the WLIP gene clusters and their genomic contexts.
Figure 3. Phylogenetic relationship of WLIP-producing strains to representative Pseudomonas species. From an alignment of 20 concatenated amino acid sequences, representing proteins with diverse housekeeping functions (listed in Supporting information), a maximum-likelihood tree using PhyML (JTT matrix; Guindon and Gascuel, 2003), as implemented in Geneious Pro 5.6.3 (Drummond et al., 2011), was constructed for WLIP producers LMG 5329 and RW10S2 (labelled in bold), Pseudomonas aeruginosa PAO1, Pseudomonas brassicacearum NFM421, P. entomophila L48, P. fluorescens strains F113, Pf0-1, Pf-5 and SBW25, Pseudomonas fulva 12-X, Pseudomonas mendocina ymp, P. putida strains BIRD1, F1, GB1, KT2440, S16 and W619, Pseudomonas stutzeri A1501 and Pseudomonas syringae pv. tomato DC3000. The filled triangles represent the tightly clustered strains F113-NFM421 and KT2440-BIRD1-F1 respectively. Bootstrap values (%) and substitutions per site (scale bar) are indicated.
Download figure to PowerPoint
Comparative analysis of genetic backbones for WLIP production
A sequence comparison of the entire multi-modular NRPS enzymes does not provide information about the extent of inter-strain similarity for individual modules and for the respective domains within these modules The existence of two apparently divergent biosynthetic systems for the same complex secondary metabolite produced by strains of two different species was therefore further scrutinized by comparative analysis of the NRPS domains [adenylation (A) and condensation (C) domains of each module and the tandem thioesterase (TE) domain]. This analysis, which was extended to the associated regulator and export system, included similarly organized Pseudomonas NRPS systems that produce lipopeptides from other groups: arthrofactin (lipoundecapeptide, Roongsawang et al., 2003), syringafactin (lipooctapeptide, Berti et al., 2007), orfamide (lipodecapeptide, Gross et al., 2007), putisolvin (lipododecapeptide, Dubern et al., 2008) and entolysin (lipotetradecapeptide, Vallet-Gely et al., 2010).
A coherent picture emerges from the phylotrees of the C-domains (Fig. 4A), as well as of the A-domains (Figs S3 and S4) and the TE domains (Fig. S5): the Wip sequences cluster closely with those of Visc and Mass, but separately from the Wlp sequences that consistently branch off together with those derived from the Pso (putisolvin) and Etl (entolysin) systems. The same observation is made when comparing the associated regulatory and export proteins (Fig. S6). The similarity of the Wlp/Pso/Etl biosynthetic, regulatory and export systems is remarkable as their products differ strongly in peptide sequence and length (Fig. 4B) and, hence, are classified into three different lipopeptide groups (Gross and Loper, 2009).
Figure 4. Comparative analysis of selected domains in the modules of the NRPS enzymes of the Wip and Wlp systems.
A. Cladogram of maximum-likelihood tree inferred from amino acid alignment of C-domains extracted from representative functionally characterized Pseudomonas NRPSs. NRPS enzymes are designated with lipopeptide-specific codes: Arf (arthrofactin, Pseudomonas sp. MIS38); Etl (entolysin, P. entomophila L48); Mass (massetolide, P. fluorescens SS101); Ofa (orfamide, P. fluorescens Pf-5); Pso (putisolvin, P. putida PCL1445); Syf (syringafactin, P. syringae DC3000); Visc (viscosin, P. fluorescens SBW25); Wip [WLIP, P. fluorescens (P. ‘reactans’) LMG 5329] and Wlp (WLIP, P. putida RW10S2). The tree was rooted with the divergent SyrE-C1 domain (syringomycin, P. syringae pv. syringae strain B301D). Clusters highlighted in red or green colour contain domains derived from Wip (in bold) or Wlp (in bold) respectively. The clustering based on the catalytic type of C-domain (lipoinitiation, non-epimerizing, epimerizing) is indicated in an expanded phylotree (Fig. S7).
B. Similarity of A-domains among Pseudomonas strains synthesizing the lipopeptides entolysin, massetolide, putisolvin, viscosin or WLIP. For each lipopeptide, the order of the three enzymes and of the modules therein reflects co-linear incorporation of the amino acid substrates indicated (positions corresponding to numbered boxes). The resulting peptide sequence of only the major massetolide of P. fluorescens SS101 is shown (de Bruijn et al., 2008). The respective initiating NRPS genes are located in a genomic region unlinked to the gene pairs encoding the middle and terminating NRPSs, except for the contiguous putisolvin genes (psoABC; Dubern et al., 2008). If known, the absolute configuration of the amino acids is indicated. Xle indicates that the residue's identity (either Leu or Ile) was not elucidated (Dubern et al., 2008; Vallet-Gely et al., 2010). Colour coding is based on co-clustering in a maximum-likelihood tree constructed from alignment of A-domain sequences (cladogram shown in Fig. S3). Clades with the same or similar amino acids (hydrophobic residues Ile/Val/Leu; hydroxylated amino acids Ser/Thr) are shown with different shades of a colour (colour assignment based on phylogenetic tree with correspondingly coloured branches shown in Fig. S4).
Download figure to PowerPoint
Balibar and colleagues (2005) showed that the D-configuration of amino acids in arthrofactin is due to the additional epimerization activity of a condensation domain with a distinctive primary sequence, acting on the amino acid attached to the previous module. WLIP differs from viscosin by a different stereochemistry at position 5. Hence, the catalytic activity of the respective C-domains of the following module (C6) is expected to cause this difference. Phylogenetic analysis assigns WipB-C6 and WlpB-C6 to the epimerizing branch among condensation domains, in line with the presence of a D-Leu at the fifth position (Fig. S7). However, ViscB-C6 co-clusters with WipB-C6, thus constituting a notable exception to the Balibar rule, along with the corresponding domains of massetolide (MassB-C6) and orfamide (OfaB-C6) (de Bruijn et al., 2007; 2008; Gross et al., 2007). Amino acid sequence alignment of these domains with the equivalent prototypical arthrofactin ArfB-C6 domain did not reveal a diagnostic sequence or motif potentially linked to the different activities of the viscosin and WLIP biosynthetic enzymes (Fig. S8).
To visualize the phylogenetic relatedness among A-domains, a colour code was assigned to the different branches in the corresponding phylotree (Fig. S4), based on patristic distances as a measure of evolutionary divergence (Table S3). This differentiation among A-domains was applied to both WLIP biosynthetic systems in comparison with related Pseudomonas lipopeptide NRPSs (Fig. 4B). This representation clearly highlights the dual nature of WLIP production by the Wip and Wlp systems: the Wlp-type WLIP system is related to those of entolysin (Etl) and putisolvin (Pso) in the P. putida–P. entomophila group, while the Wip-type WLIP system is more similar to the massetolide (Mass) and Visc (viscosin) ones of the P. fluorescens group.
WLIP biosynthesis: same product but distinct genes in different hosts
Previously, a PCR-based screening for lipopeptide-specific NRPS genes in Pseudomonas based on amplification and sequencing of part of the unique C1 and TE domains also yielded amplicons for a number of WLR-positive Pseudomonas strains from different origins (Rokni-Zadeh et al., 2011). In retrospect, among these isolates no additional or intermediate genotypes with respect to WLIP genes appear to be present as the diagnostic amplicon sequences (corresponding to stretches for the lipoinitiation domain and the terminal tandem thioesterase domain as indicated in Fig. 2) can be assigned to either the Wlp type (rice rhizosphere isolate RW3S1 from Sri Lanka) or to the Wip type (five strains). The latter type was identified for a mushroom isolate from the UK (LMG 2338), for three strains originating from the rhizosphere of maize in Belgium (PGSB3962, PGSB7828, PGSB8273), and for Pseudomonas extremorientalis LMG 19695, a strain affiliated with the P. fluorescens cluster that was isolated from a drinking water reservoir in Russia (Ivanova et al., 2002).
From an evolutionary viewpoint, the occurrence of two distinct NRPS systems generating the same secondary metabolite conferring similar phenotypes upon phylogenetically distinct hosts is of particular interest. The high similarity of the Wip and Visc genes and close relatedness of their respective representative hosts (P. fluorescens LMG 5329 and SBW25) suggests that these lipopeptide systems diverged relatively recently to produce two lipopeptides that merely differ by the D/L-configuration of one amino acid residue. Apparently, WLIP biosynthesis has evolved independently in another host of the P. putida clade. This is inferred from homology of the Wlp components with NRPS systems assembling the structurally unrelated lipopeptides putisolvin and entolysin, and from common affiliation of their host strains with the P. putida–P. entomophila clade. Analysis of the Wlp/Pso/Etl substrate-selecting domains (Fig. 4B) reveals a prominent example of patchwork assembly of similar NRPS modules recruited to generate lipopeptides of different amino acid sequences and lengths.
The identification of WLIP genes in strains of P. fluorescens and P. putida, both behaving as P. ‘reactans’, clearly urges the discontinuation of the use of this invalid species designation, originally based on a diagnostic phenotype linked to secondary metabolism, which is still propagating in research papers and reviews (Kobayashi and Crouch, 2009; Park et al., 2009; Largeteau and Savoie, 2010; Oksinska et al., 2011; Prashanth et al., 2011; Shirokova et al., 2012).