Present address: Imperial College of Science, Technology and Medicine, Department of Biology, SAFB, Imperial College Road, SW7 2A2 London, UK.
Pyoverdines, the main siderophores of fluorescent pseudomonads, contain a peptide moiety, different for each pyoverdine, and an identical chromophore. While it has been shown that non-ribosomal peptide synthetases (NRPSs) are involved in the biosynthesis of the peptide chain of pyoverdines, this was not demonstrated for the biosynthesis of the chromo-phore part. We found that PvsA, from Pseudomonas fluorescens ATCC 17400, and PvdL (PA2424), from Pseudomonas aeruginosa are similar NRPSs and functional homologues, necessary for the production of pyoverdine. Transcriptional lacZ fusions showed that pvdL is co-transcribed with the upstream PA2425 gene, encoding a putative thioesterase, and is iron-regulated via PvdS. Similarly, RT-PCR analysis revealed that expression of pvsA is repressed by iron. Analysis of the adenylation domains of PvsA, PvdL and their homologues, revealed that their N-terminus starts with an acyl-CoA ligase module, followed by three amino acid activation domains. Computer modelling of these domains suggests that PvsA in P. fluorescens and PvdL in P. aeruginosa are orthologues involved in the biosynthesis of the pyoverdine chromophore.
Iron is an essential nutrient for nearly all microorganisms. It is needed as cofactor in a variety of proteins participating in many important cellular functions such as respiration (cytochromes, ferredoxines) and DNA synthesis (Neilands, 1991). Although iron is the fourth most abundant metal in Earth's crust, at neutral pH it forms insoluble Fe (III) oxyde hydrates, which are not available to microorganisms (Neilands, 1995; Braun and Killmann, 1999). Therefore, many microorganisms excrete iron-chelating molecules, termed siderophores (Neilands, 1995), which bind iron (III) before being recognized as ferrisiderophores by specific receptors (Guerinot, 1994). Several species of rRNA group I pseudomonads (Pseudomonas sensu stricto), produce and excrete, under conditions of iron limitation, fluorescent yellow-green siderophores, named pyoverdines or pseudobactins. They are composed of a dihydroxyquinoline chromophore and a variable peptide chain, comprising six to 12 amino acids, depending on the producing strain (Budzikiewicz, 1993, 1997; Meyer, 2000). The peptide chain comprises l- and d-amino acids, some of them unusual, such as N5-hydroxycycloornithine or N5-formyl-N5-hydroxyornithine (Meyer, 2000). A number of genes involved in the biosynthesis of pyoverdine in P. aeruginosa, P. fluorescens, and Pseudomonas putida have been identified and are summarized in Table 1. For P. aeruginosa, these include the pvcABCD genes potentially involved in the synthesis of the pyoverdine chromophore in P. aeruginosa (Stintzi et al., 1996, 1999), pvdA (Visca et al., 1994) that encodes an l-ornithine N5-oxygenase, pvdF that codes for a transformylase (McMorran et al., 2001), and the non-ribosomal peptide synthetases genes pvdD (Merriman et al., 1995), and pvdIJK (Lehoux et al., 2000).
Table 1. . Characterized genes known to be involved in the biosynthesis of pyoverdines in P. aeruginosa , P. fluorescens , and P. putida.
Non-ribosomal peptide synthetases (NRPSs) are large multimodular enzymes, which perform non-ribosomal peptide synthesis according to multiple carrier thiotemplate mechanism (Kleinkauf and von Döhren, 1996; von Döhren et al., 1999; Challis et al., 2000). A NRPS basic module consists of an adenylation domain, a thiolation domain and a condensation domain. The adenylation domain (comprising 10 conserved motifs) recognizes and activates a specific amino acid as its acyl adenylate by reaction with ATP (Konz and Marahiel, 1999; Stackelhaus et al., 1999). This activated ester is then covalently linked as its thioester on the thiolation domain. The condensation domain, comprising seven conserved motifs, catalyses the direct transfer to another acylamino acid intermediate on the adjacent downstream module to form a peptide bond. In some cases, epimerization of amino acid from the l- to the d-configurations is catalysed by an extra domain (comprising seven conserved motifs) within the module. A terminal thioesterase is often present to release the peptide from the enzyme by cyclization or hydrolysis.
NRPSs have been shown to participate in peptide antibiotic- (Hancock and Chapple, 1999), and siderophore biosynthesis in different bacterial species (reviewed by Quadri, 2000), including in P. aeruginosa where NRPSs are not only involved in the biosynthesis of pyoverdine, but also of the second siderophore, pyochelin (Quadri et al., 1999). By aligning adenylation domain protein sequences of different NRPSs with the corresponding domain of the structurally known gramicidin-synthetase, it is often possible to predict the amino acid that is to be activated (Conti et al., 1997; Stackelhaus et al., 1999; Challis et al., 2000).
In this study, we present data concerning a gene cluster, comprising one large iron-repressed pyoverdine synthetases, PvsA, from P. fluorescens ATCC 17400, together with its P. aeruginosa functional ortholog, PvdL, encoded by the PA2424 gene. We also present evidence that genes encoding homologues of PvsA are also present in the genomes of other fluorescent pseu-domonads, and that these NRPSs are involved in the biosynthesis of the pyoverdine chromophore.
Functional complementation of the pyoverdine-negative P. fluorescens ATCC17400 pyoverdine-negative 3G6 mutant
Pseudomonas fluorescens ATCC 17400 mutant 3G6 was isolated, after Tn 5 mutagenesis, as nonfluorescent, and was found to be unable to grow on CAA plates contain-ing 1 mg ml −1 ethylenediaminedihydroxyphenylacetic acid (EDDHA), a strong Fe(III) chelator ( Cornelis et al., 1992 ). One cosmid, from wild-type P. fluorescens ATCC 17400, with a 26 kb insert restored pyoverdine production in 3G6. The complemented 3G6 clone was designated as 3G6C and the cosmid bearing the DNA fragment responsible for the complementation, p3G6C.
DNA analysis of the complementing DNA fragment on p3G6C
The DNA flanking one of the two Tn5 insertions in 3G6 (Cornelis et al., 1992) was used as a probe and found to hybridize with p3G6C DNA. Southern blot analysis revealed that the Tn5-flanking DNA hybridized with one 3.8 kb-PstI and one 5.3 kb-BamHI fragment re-spectively (results not shown). Figure 1A shows the map that corresponds to 17.3 kb of p3G6C sequenced DNA. Computer-assisted analysis of the sequence from p3G6C revealed one, large open reading frame: pvsA. In silico analysis using the blast algorithm (Altschul et al., 1997) revealed that it encoded a protein of 4313 amino acids, with high similarity to NRPSs (see below). Downstream of pvsA, there is an inverted repeat that could form a Rho-independent terminator, followed by an ORF encoding a hypothetical protein. Upstream of pvsA, and transcribed in the same orientation, there is one open reading frame the product of which shows a similarity with thioesterases. Upstream of the thioesterase gene, and transcribed in the opposite orientation, is another ORF with high similarity with the pvdS gene from P. aeruginosa (Cunliffe et al., 1995; Miyazaki et al., 1995). This gene encodes a sigma factor that specifically activates the transcription of pyoverdine biosynthesis genes in P. aeruginosa (Wilson and Lamont, 2000; Wilson et al., 2001). Upstream of pvdS, and transcribed in the same direction as pvsA, is a gene encoding a putative transacetylase. Between the pvdS and the thioesterase gene, a typical Fur binding site was found (GATAATGATTGTCAAATGA), indicating that the expression of pvdS might be regulated by Fur as it is the case in P. aeruginosa (Vasil and Ochsner, 1999). A similar organization pvdS-thioesterase-NRPS is found in P. aeruginosa and P. syringae DC3000. In P. putida KT2440 and P. fluorescens Pf0, the thioesterase gene is present in front of one of the NRPS genes responsible for the biosynthesis of the pyoverdine peptide chain (J. Ravel, personal communication). The DNA sequence has been deposited in GenBank under accession number AF237701.
RT-PCR analysis of pvsA transcription
RT-PCR was performed using different pairs of primers (the positions of the primers are given in Fig. 1A), described in Experimental procedures. The primers pvsA1 and pvsA2 were designed in order to amplify a 352 bp region at the 3′ end of pvsA RNA. A second pair of primers (pvsA3 and pvsA4) was designed to amplify a 838 bp central region. A 633 bp fragment, closer to the 5′ end of pvsA could be amplified using two other primers, pvsA5 and pvsA6. Primer pvsA1 was also used in combination with antisense primer term1 located after the potential terminator in order to verify that transcription indeed was stopped after the terminator. In the last case, if the transcription would continue past the inverted repeat, we would expect a fragment of 507 bp. The results are presented in Fig. 1B. For the wild-type P. fluorescens ATCC 17400 and 3G6 mutant RNA’s, grown in CAA without added iron, the expected fragments were obtained, namely 633 bp, for amplifications done with pvsA5 and pvsA6 primers, and 838 bp for the amplifications done with pvsA1 and pvsA4 primers respectively (Fig. 1B). This was expected since the Tn5 insertion is situated downstream (Fig. 1A). Amplification of the pvsA region downstream of the Tn5 insertion gave a fragment for the wild type only (Fig. 1B, lane 3, compare with lane 7). The combination of primers pvsA1 and term1 did not result in amplification (Fig. 1B, lanes 5 for wild type, and lane 9 for 3G6).
Finally, no amplified product could be observed when the RNA was prepared from cells grown in the presence of iron (Fig. 1B, lanes 11–19), indicating that this cluster is repressed by iron.
Comparison of P. fluorescens ATCC 17400 PvsA with other NRPSs from fluorescent pseudomonads
A similarity search was done with the translation product of pvsA using the blast-p algorithm. PvsA (4313 aa) aligns almost perfectly (72% identity, 80% similarity, 1% gaps) with the product (4342 residues) of P. aeruginosa PAO1 gene PA2424 (http:www.pseudomonas.com; GenBank accession: AE0046699).
A tblastn search was done against the unfinished P. fluorescens Pf01 genome (http:www.jgi.doe.gov), which revealed a very high level of identity (75%) between PvsA (residues 1–3332) and a translation product (nucleotides 32480–22437) of an ORF present on contig 262. Another very good alignment (85% identity) was found in the same contig, between residues 3322 and 4313 of PvsA and the translation product of nucleotides 22474–19484 (frame−1), suggesting a possible frameshift in the Pf0 genome. Highly similar putative peptide synthetases were also detected in the genome of P. putida KT2440 (67% identity, gnl/TIGR 160488/13538) and P. syringae DC3000 (72% identity, gnl/TIGR 317/5559) (http:www.tigr.org).
PA2424 encodes a NRPS, PvdL, essential for pyoverdine synthesis in P. aeruginosa PAO1
Given the high similarities between PvsA and the product of PA2424, we decided to disrupt the PA2424 gene (renamed pvdL) in PAO1 as described in Experimental procedures. The pvdL mutant was non-fluorescent in CAA medium and could not grow on CAA medium in the presence of 0.5 mg ml−1 EDDHA (results not shown). This confirms that PvdL is a new pyoverdine synthetase. The start of the pvdL coding sequence is separated by 72 bp from the stop of the upstream PA2425 gene, but no promoter could be detected upstream of pvdL. Indeed, plasmid pPZ2424 containing this intergenic region fused to the lacZ gene did not result in any β-galactosidase reporter activity (data not shown), suggesting that pvdL forms an operon with the upstream gene (PA2425) encoding a putative thioesterase (Fig. 2A). Using plasmid pPZ2425 to monitor transcription from the PA2425 promoter, expression of this operon was 100-fold regulated by iron and strictly depended on the alternative sigma factor PvdS (Fig. 2B). Interestingly, the pvdS gene is located immediately upstream of PA2425, and transcribed in the opposite orientation (Fig. 2A). A motif resembling the PvdS recognition site (Wilson et al., 2001) was detected within the promoter region of PA2425, suggesting direct activation of the PA2425-PA2424 operon by PvdS (Fig. 2A).
Complementation of the P. aeruginosa pvdL mutant
The p3G6C cosmid was introduced by tri-parental mating into the P. aeruginosa pvdL mutant. The resulting trans-conjugants had restored fluorescence, indicative of pyoverdine production, and could grow in the presence of EDDHA, confirming the production of a functional pyoverdine siderophore (results not shown). Interestingly, the complementation could only be observed at 28°C, not at 37°C.
The pyoverdine produced by the complemented P. aeruginosa pvdL mutant was analysed by isoelectric focusing (IEF) (Koedam et al., 1994; Meyer et al., 1997), together with the pyoverdines produced by the wild types P. fluorescens and P. aeruginosa, and mutant 3G6 of P. fluorescens complemented in trans by p36C. As can be seen from Fig. 3A, the pattern of pyoverdine isoforms present in the culture supernatant of the P. aeruginosa pvdL mutant complemented in trans by the p3G6C cosmid differs both from the PAO1 wild type, and from the P. fluorescens patterns. For the P. fluorescens ATCC 17400 wild type and the mutant 3G6 complemented, two pyoverdine bands are visible, these two bands represent probably two forms of pyoverdine with a different side chain (succinate or glutamate) attached to the chromophore (Budzikiewicz, 1993, 1997; Meyer, 2000).
Structure determination of the pyoverdines produced by P. aeruginosa pvdL complemented by p3G6C
The pyoverdines produced by the P. aeruginosa pvdL mutant, complemented by cosmid p3G6C, were purified and analysed by mass spectrometry and NMR. After electrospray ionization three molecular species were observed with mass differences of 14 Da. The molecular mass 1333 Da corresponds to the original PAO1 pyoverdine with a succinic acid side chain (Briskot et al., 1989). By collision induced fragmentation of the three species in the octapole region and in the ion trap (Fuchs and Budzikiewicz, 2000, 2001), ions characteristic for the amino acid sequence of the peptide part (Roepstorff and Fohlman, 1984) were registered (Table 2). As shown in Fig. 3B, the three species with succinic acid as side chain differ only in the way of a modification of either one or both N-formyl-N-hydroxy-ornithines by exchange of the formyl with an acetyl substituent: all ions containing one AcOHOrn unit are shifted by 14 Da, those containing two AcOHOrn units by 28 Da. Hence, what had been obtained is a mixture of PAO1 pyoverdines with two FoOHOrn, with one FoOHOrn and one AcOHOrn (the shift data show that either FoOHOrn1 or FoOHOrn2 is replaced by AcOHOrn), and with two AcOHOrn residues in the peptide chain in a ratio of about 1:3:5. The presence of the acetyl groups was confirmed by an 1H-NMR spectrum of the mixture which in addition to the signals otherwise characteristic for the pyoverdine showed acetyl signals at 1.98 and 2.00 p.p.m. Three other pyoverdine species were found, identical to those described above, but with succinamide as side-chain (Briskot et al., 1989), bringing the total number of forms to six, as observed on the IEF gel (Fig. 3A).
Table 2. . Ion masses found for the three types of pyoverdines produced by P. aeruginosa pvdL mutant complemented by p3G6C (shown only for the forms with succinic acid as side chain).
Ions characteristic for the amino acid sequence: N-terminal: A…XNHCHR+, B…XNHCHRCO+, C-terminal: Y′…XCORCHNH3+, italic: one AcOHOrn instead of FoOHOrn, bold: two AcOHOrn instead of FoOHOrn, Ac(Fo)OHOrn … N5-acetyl(formyl)-N5-hydroxy-Orn
B4 – NH3 + lys
B4 – NH3 + lys + Thr
B4 – NH3 + lys + Fo/AcOHOrn
[M + 2H]2+
Analysis of NRPSs domains in PvsA and its homologues
Domains characteristic of NRPSs have been identified in PvsA, PvdL and their orthologues from P. syringae DC3000, P. putida KT2440, and P. fluorescens Pf0 (Konz and Marahiel, 1999) and consist of three adenylation do-mains (A-domains), four thiolation domains (T), three condensation domains (C-domains) and one epimerization domain (E), as represented in Fig. 4A. No thioesterase domain was found at the C-terminus of these NRPSs. In addition, the first module of these pyoverdine synthetases starts with an unusual domain with high homology to the acyl CoA ligases domains (ALs) found in the first module of the saframycin synthetase B, SafB (Pospiech et al., 1996), and the first module of the bleomycin synthetase, BlmVI-NRPS-5 (Du et al., 2000). The function of such domains in these synthetases has not yet been elucidated. We attempted to predict the amino acid activated by each adenylation domains of PvsA-M2, -M3 and -M4 of PvsA orthologs. By aligning the sequence between the A3 and A6 sequence motifs and the corresponding region in the gramicidin NRPS GrsA, the eight residues lining the specificity pocket in each domain were predicted (Challis et al., 2000; http:raynam.chm.jhu.edunrps). For domains PvsA-M2, -M3, and –M4, the following re-sidues were identified: -M2-DVWHFGRI, -M3-DAEFIGAV and -M4-DIWELTAD. Identical results were obtained for PvsA, PvdL and their homologues in P. putida and P. syringae, indicating that they activate the same amino acids, and in the same order. PvsA-M2 is predicted to activate l-glutamate. Its binding pocket motif is highly similar to the three l-glutamate M-domains of fengycin synthetase (FenA-M2, FenC-M1 and FenE-M1, Steller et al. 1999). A noteworthy observation is that the A domain of PvsA-M2 contains the first binding pocket with an arginine residue (R301). No reliable predictions were obtained for PvsA-M3 and PvsA-M4, using the above-mentioned web-site.
In most NRPSs, the organization and the order of the modules maps in a 1:1 manner to the amino acid sequence of the peptide products (co-linearity rule). If this paradigm applies to PvsA and its orthologues, and according to the proposed published mechanism for the biosynthesis of the pyoverdine chromophore (Böckmann et al., 1997), PvsA-M3 would recognize and activate either l-tyrosine or l-tri-hydroxyphenylalanine (TOPA). The architecture of the binding pocket of the PvsA-M3 A domain is not in agreement with those of known l-tyrosine A-domains. The purified wild-type fengycin synthetases FenA and FenD are reported to activate specifically l-tyrosine (Steller et al., 1999). This is reflected by the presence of three residues in the binding pocket capable of hydrogen bonding to the aromatic hydroxyl group of tyrosine (T239, T299 and E322). In contrast, only E239 would be capable of hydrogen bonding to the substrate in the binding pocket of PvsA-M3 domain. E239 could hydrogen bond with the para- and meta-hydroxyl groups of TOPA, as well as the hydroxyl group of tyrosine. In order to confirm this prediction, a three-dimensional model of PvsA-M3-domain binding pocket was generated based on the co-ordinate of the GrsA A-domain (Conti et al., 1997). The results (Fig. 5) show that TOPA can sterically fit inside the binding pocket. Moreover, residue E239 can form hydrogen bonds with both the para- and meta-hydroxyl groups of TOPA (predicted distance of 2.84 Å and 2.54 Å respectively). The presence of a phenylalanine residue at the bottom of the binding pocket appears to be responsible to position the substrate for hydrogen bonding with E239. We therefore predict that PvsA-M3 can activate both l-TOPA and l-tyrosine due to loose binding specificity.
An epimerization domain is found in PvsA-M3 after the thiolation domain, indicating that the amino acid incorporated into the peptide product by this module should be in its d-form. Tyrosine is indeed incorporated into the chromophore in its d-configuration (Böckmann et al., 1997). According to the co-linearity rule and the proposed biosynthetic scheme for pyoverdine chromophore, PvsA-M4 is predicted to activate l-2, 4-diaminobutyric acid (Dab). The A-domain of the 3rd and 4th module of the syringomycin synthetase SyrE has been proposed to activate l-Dab (Guenzi et al., 1998). The PvsA-M4 A domain binding pocket structure differs from those of SyrE-M3-4. One common feature is the presence of a glutamate and a tryptophan residue at the bottom of the pocket, allowing for hydrogen bonding with the terminal amino group of Dab (see Supplementary material).
Several NRPSs involved in pyoverdine biosynthesis have already been identified by analysis of DNA sequences flanking transposon insertions that result in a pyoverdine-negative phenotype, including in three recent reports (Adams et al., 1994; Merriman et al., 1995; Lehoux et al., 2000; Mirleau et al., 2000; Rossbach et al., 2000). Next to this molecular approach, Georges and Meyer (1995) identified high-molecular-mass, iron-repressed, cytoplasmic proteins (IRCPs) in different fluorescent pseudo-monads, which they postulated to be NRPSs involved in the biosynthesis of pyoverdines. Sequence analysis of a cosmid DNA insert complementing a pyoverdine-null phenotype revealed one large ORF, pvsA, encoding a protein of 4313 amino acids, with a predicted mass of 477 kDa. Reverse transcriptase-PCR experiments also demonstrated that repression by iron occurred at transcrip-tional level, as previously observed for pvdD expression (Merriman et al., 1995). Interestingly, PvsA showed high identity (higher than 70%) with a 4342-amino acid putative NRPS from P. aeruginosa, encoded by gene PA2424 (GenBank AF237701). The PA2424 gene is organized in an operon with PA2425, a gene that encodes a putative thioesterase, which represents an essential function for non-ribosomal peptide synthesis (Konz and Marahiel, 1999). Secondly, the expression of this P. aeruginosa operon depends on the alternative PvdS s factor that is needed for the transcription of various other pyoverdine biosynthesis genes (Vasil and Ochsner, 1999). Furthermore, we have shown that this gene is needed for pyoverdine biosynthesis, and therefore suggest that PA2424 should be renamed pvdL.
The pattern observed on IEF gels for the pyoverdine isoforms of pvdL (p3G6C) differs strongly from the wild-type P. aeruginosa pattern. These apparent differences do not seem to affect the biological activity of the pyoverdine produced since it allows the P. aeruginosa pvdL mutant to grow in the presence of EDDHA. Indeed, a mass spectrometric and NMR analysis revealed that the peptide chain of PAO1 pyoverdine is unchanged, except for a modi-fication (partial or complete) of the two FoOHOrn by exchange of the formyl to an acetyl group (Fig. 3B). The presence of a transacetylase gene upstream of the pvdS gene on p3G6C is likely to be at the origin of this modification. This gene shows similarity with iucB gene of E. coli for aerobactin synthesis (Martinez et al., 1994). We do not know, at the moment, whether these modifications affect the biological activity of the pyoverdines or not, since a small proportion of the pyoverdine produced is unmodified. Interestingly, the gene upstream of pvdS in PAO1, PA2427 shows only limited similarity with the transacetylase gene.
Structures of different pyoverdines have now been elucidated (Meyer, 2000; Fuchs and Budzikiewicz, 2001). The number of amino acids varies from six to 12, and some of them are unusual. In the case of P. fluorescens ATCC 17400 pyoverdine, the sequence of the peptide arm is: Ala-Lys-Gly-Gly-OHAsp-Gln/Dab-Ser-Ala-c-OHOrn where the amino acids in d-configuration are in bold (Demange et al., 1990). In this case, the unusual amino acids are β-hydroxy aspartic acid (OHAsp), cyclo-hydroxyornithine (3-amino-1-hydroxypiperidone), and Gln/Dab, a condensation product of these two amino acids, yielding a tetrahydropyrimidine ring. This sequence differs from that of the PAO1 pyoverdine, with Ser-Arg-Ser-FoOHOrn-c(Lys-FoOHOrn-Thr-Thr) having a cyclopeptidic C-terminal substructure as indicated by the parentheses (Briskot et al., 1989). According to Jülich et al. (2001), all P. syringae produce the same pyoverdine with the peptide chain ɛ-Lys-OHAsp-Thr-Thr-Ser-OHAsp-Ser. The structures of the pyoverdines of P. fluorescens Pf0 and P. putida KT2440 are not yet known, but probably differ from those of the pyoverdines mentioned above. The new pyoverdine synthetases PvsA and PvdL described in this work, are highly similar to putative NRPSs from P. fluorescens Pf0, P. putida KT2440, and P. syringae DC3000. This, together with the fact that these pyoverdines differ by their peptide chain, strongly suggests that PvsA, and its conserved orthologues might be implicated in the biosyn-thesis of the chromophore part that is conserved in all pyoverdines. It is also interesting to notice that, of all NRPSs involved in pyoverdine biosynthesis identified so far in PAO1, PvdL is the first to start with an activation domain (although it does not likely activate an amino acid). This means that the synthesis of pyoverdine starts with this domain, as it is the case for aryl-capped siderophores or other chromopeptides (Quadri, 2000; Keller and Schauwecker, 2001).
PvsA and its orthologues do not have a terminal thioesterase domain (TE). Moreover, the other pyoverdine NRPSs of P. aeruginosa (PA2402-PA2399), downstream from PvdL, start with a C domain, indicating a possible interaction between the two NRPSs, which would form the amide bond between the carboxy of Dab and the amino group of the first serine of the peptide chain.
The pvcABCD cluster in P. aeruginosa has been proposed to be involved in the biosynthesis of the chromophore (Stintzi et al., 1996, 1999). The following scheme has been proposed. The amino acid sequence γ-l-Glu-d-Tyr-l-Dab is attached to the pyoverdine peptide chain and Dab and tyrosine are condensed to a tetrahydropyrimidine ring (Böckmann et al., 1997; Hohlneicher et al., 2001). Metabolites containing this structure, referred to as ferribactins, have been found to accompany the pyoverdines (Hohlneicher et al., 2001; Baysse et al., 2002). Transformation of the ferribactin to the pyoverdine chromophore involves probably oxidation of tyrosine to 2,4,5-trihydroxy-phenylalanine (TOPA) (Longerich et al., 1993; Böckmann et al., 1997). Ring closure leads to dihydropyoverdines, which are then oxidized to the pyoverdines. The glutamate attached to the original tyrosine by its γ-carboxyl group can be transformed to α-ketoglutaric acid, succinamide and malamide; the two amides may be hydrolyzed to the free acids (Schäfer et al., 1991). The modification of tyrosine was proposed to be the result of the action of PvcB and PvcC, respectively, an oxygenase and an hydroxylase (Stintzi et al., 1999). Based on protein sequence analyses and modelling studies, we can predict that PvsA-M2 activates l-glutamate, while l-TOPA or l-Tyr are likely to be activated by PvsA-M3 and l-Dab by PvsA-M4 respectively. According to these results and physiolo-gical evidence, we propose the following chromophore biosynthetic scheme for PvsA and its orthologues. PvsA could incorporate tyrosine first, before it becomes modified in trans by PvcB and PvcC. A similar reaction, involving a haem-dependent P450 enzyme, has been proposed for the production of β-hydroxytyrosine in the case of novobiocin biosynthesis (Chen and Walsh, 2001; Walsh et al., 2001). Recently, we have shown that a ccmC mutant of P. fluorescens ATCC 17400 grown in the presence of l-cysteine produces almost exclusively ferribactin, suggesting that the conversion of Tyr to TOPA occurs after incorporation of d-Tyr (Baysse et al., 2002). The pvc genes are however, not present in the genomes of the other fluorescent pseudomonads (results not shown), suggesting that these genes play a role in pyoverdine biosynthesis only in the case of P. aeruginosa strains. However, genes with similar functions are present in the genomes of the other pseudomonads, and, in the case of P. putida, are located between pvsA and the genes for other NRPSs (J. Ravel, personal communication). In the first article describing the identification of the pvc cluster, a pvcC mutant was observed to produce no pyoverdine on succinate medium, but this phenotype was found to revert partially to pyoverdine production, suggesting that the pvc genes are indeed not indispensable (Stintzi et al., 1996).
Biochemical studies of PvsA and its orthologues will now be needed to elucidate the biosynthetic route to the pyoverdine chromophore, as well as to establish a role for PvcABCD.
Bacterial strains and growth conditions
Bacterial strains and plasmids used in this study are de-scribed in Table 3. Pseudomonas fluorescens ATCC 17400 and the pyoverdine-negative mutant 3G6 were grown at 28°C in casamino acids medium (CAA) (Cornelis et al., 1992) or in succinate medium (Meyer and Abdallah, 1978), in flasks at 200 r.p.m. (New Brunswick Innova shaker). All E. coli strains were grown at 37°C at 200 r.p.m. in Luria–Bertani (LB) medium (Difco Laboratories). Antibiotics, when required, were added at the following concentrations: ampicillin, 100 µg ml−1, kanamycin, 100 µg ml−1, streptomycin, 25 µg ml−1, spectinomycin 50 µg ml−1 (200 for P. aeruginosa), Gm 100 µg ml−1. All antibiotics and chemicals were purchased from Sigma Chemicals. For growth in conditions of iron sufficiency, FeCl3 was added at 100 µM. Growth in the presence of 1 mg ml−1 ethylenediaminedihydroxyphenylacetic acid (EDDHA) was done on solid CAA medium (15 g l−1 bacteriological agar, Difco Laboratories).
Table 3. . Strains and plasmids used in this study.
304 bp fragment of the intergenic region between PA2424 and PA2425 cloned in front of lacZ
Promoter region of PA2425 cloned in front of lacZ
Construction of the P. fluorescens genomic library
Genomic DNA from P. fluorescens ATCC 17400 was prepared according to Ausubel et al. (1994), and was partially restricted with PstI (Roche), using increasing dilutions of the restriction enzyme, in 100 µl volume. After 30 min, the reaction was stopped by heat inactivation, and 5 µl aliquots from each tube analysed by agarose gel electrophoresis (0.7% w/v). Tubes containing fragments in the range of 25–30 kb were pooled and ligated to PstI-restricted cosmid pRG930 (Van den Eede et al., 1992). Packaging of ligated DNA and transformation of E. coli HB101 was done by using Gigapack II Gold-4 kit, following the recommendations of the manu-facturer (Stratagene). Approximately 3000 colonies were selected and transferred with sterile toothpicks on LB agar plates containing streptomycin and spectinomycin, and in 96-well microtitre plates (Costar) containing 200 µl of LB, anti-biotics and 20% glycerol. These plates were stored at −80°C.
Functional complementation of pyoverdine-negative mutants
The cosmid library was mobilized by triparental conjugation to mutant 3G6 as described previously (Cornelis et al., 1992), using E. coli CM404 containing plasmid pRK2013 as helper strain. Screening for complementation was done on CAA plates containing streptomycin, spectinomycin (200 µg ml−1 each) and EDDHA (1 mg ml−1). Colonies growing on these plates and producing pyoverdine, as judged by the yellow-green colour and the fluorescence under UV light, were considered as complemented for pyoverdine production.
Detection of pyoverdine by isoelectric focusing (IEF) and pyoverdine purification
Pyoverdine isoforms were detected by isoelectric focusing of CAA liquid culture supernatants as described previously (Koedam et al., 1994).
Pyoverdines were purified from 500 ml of glutamate minimal medium supernatant of a 48 h culture at 28°C by C-18 chromatography. Briefly, 500 ml of sample were poured on 1 × 4 cm C-18 column, washed twice with 10 volumes of distilled water and eluted with 1 ml of 50% methanol and evaporated in a Speed-Vac.
The pyoverdine isoforms were separated as ferri-complexes on Nucleosil-100 C18, 250 × 4mm (Knauer), detection 254 nm, gradient in 50 mM CH3COOH with CH3OH (3% to 30%), during 30 min, at a rate of 0.7 ml min−1. Afterwards ferripyoverdines were de-complexed by passage on a Sep – Pak RP18 cartridge. They were first adsorbed on the column and rinsed with 6.5% (w/v) oxalate solution (pH 4.3), washed with water and de-sorbed with CH3OH/H2O 1:1 v/v, and dried. The pyoverdines were dissolved in CH3OH/H2O/CH3COOH (50:50:0.1 v/v) and introduced into the electrospray ionization (ESI) source of a Finnigan MAT (Bremen, Germany) 900ST mass spectrometer with an electrostatic/magnetic analyzer-quadrupole-ion trap geometry. Fragmentation of the protonated molecular ions was achieved by collision activation in the quadrupole unit and in the ion trap. 1H-NMR data were obtained with a Bruker (Karlsruhe, Germany) DRX 300 instrument, solvent D2O.
Plasmid DNA was purified and manipulated by using standard techniques (Sambrook et al., 1989). All enzymes were purchased from Roche. Escherichia coli DH5α was used as a host strain for most molecular cloning. As cloning vectors, pUC19 and pBluescript KS+ were used. DNA fragments were excised from agarose gels and purified by using the Concert Rapid Gel Extraction System (Life Technologies). Exonuclease III deletions were generated by using the Erase-a-Base system (Promega), according to the manufacturer's recommendations. Southern blot hybridization was done as described previously (Cornelis et al., 1992) using digoxigenin-labelled probes (Roche). Nylon membranes, positively charged, were purchased from Roche.
DNA sequence analysis
Nucleotide sequencing (carried out on DNA subclones in plasmid pBluescript KS+, using a Licor Long Reader 4200 automated sequencer (Westburg, the Netherlands), or ABI Prism Model 377 automated sequencer (Applied Biosystems). Universal M13 primers (forward and reverse) were used, as well as internal primers (Eurogentec, Seraing, Belgium). Direct, double-strand DNA, sequencing on the original cosmid clone was performed by Eurogentec. DNA sequences were analysed with the gene compare software package (Applied Maths, Gent, Belgium), and the Genetics Computer Group (University of Wisconsin) package of programs.
Total RNA was prepared from P. fluorescens ATCC 17400 and mutant 3G6 cultures grown in succinate with and without added FeCl3 (100 µM) using the High Pure RNA Isolation kit following the recommendations of the manufacturer (Roche). Total RNA was quantified by measuring the absorbance at 260 nm. cDNA synthesis was performed using the First Strand cDNA Synthesis kit following the recommendations of the manufacturer (Amersham Pharmacia Biotech) in 15 µl reaction volume starting with 5 µg of total RNA. Reverse transcription-PCR was performed using different pairs of primers. Primers were designed in order to amplify a 351-bp region at the 3′ end of pvsA RNA (pvsA1 sense 5′-GTTGAA CGCCAACGGTAAACTC-3′ starting at position 15964, pvsA2 antisense 5′-CAAATCACTCAACCGATCCACC-3′ starting at position 16315). A second pair of primers was designed to amplify an 838-bp central region (pvsA3 sense 5′-GACTA CTGGACCACGCAACTGG-3′, starting at position 9122, and pvsA4 antisense 5′-TGGACCTGCTGGCTGAACAGAT-3′, starting at position 9960). A 633 bp fragment, closer to the 5′ end of pvsA was expected to be amplified using two other primers (pvsA5 sense 5′-CGATCAGGAGCACTACCTGGTG-3′, starting at position 8914, and pvsA6 antisense 5′-GGCT TCCACCAGATGGTCGAAT-3′, starting at position 9547). Primer pvsA1 was also used in combination with antisense primer term1 (term1 antisense 5′-CCTGCAATACGGTCCAT CACAT-3′, starting at position 16475). Polymerase chain reaction was performed with different combinations of oligonucleotide pairs (40 pmol each) in a total volume of 50 µl with 2.5 units of Taq polymerase (Roche). Initial denaturation was for 2 min at 94°C and synthesis was performed during 30 cycles (1 min at 94°C, 30 s at 55°C and 1 min at 72°C). Samples were analysed to identify the size of the amplified products by electrophoresis on a 1.2% agarose gel.
Construction of a PA2424 P. aeruginosa mutant strain
The 5′ region of PA2424 was amplified by PCR from genomic DNA prepared from P. aeruginosa PAO1. The primers PA2424-forw (5′-GAGCGGACTTTCTCGTCAC) and PA2424-rev (5′-GGACGTTCGAGGAAGTAGC) and Taq DNA polymerase (Bethesda Research Laboratories, Rockville MD) were used in 30 cycles of denaturing (1.5 min, 95°C), annealing (1.5 min, 54°C), and extending (1.5 min, 72°C). The resulting 1105 bp PCR product was cloned into pCRII-2.1 (Invitrogen, San Diego CA), and a 762 bp EcoRI subfragment was then ligated into pEX18Tc (Hoang et al., 1998) linearised with EcoRI and HindIII after end-polishing with Klenow. A 223 bp PstI fragment of the PA2424 coding region was replaced by a 1.2 kb PstI gentamicin-resistance cassette that was obtained from pPS856 (Hoang et al., 1998). Escherichia coli SM10 harbouring the resulting plasmid pEX18Tc-ΔPA2424::Gm was used as the donor strain in a mating with P. aeruginosa PAO1. Transconjugants were selected on BHI agar containing gentamicin (75 µg ml−1) and irgasan (50 µg ml−1) and then plated on LB agar containing gentamicin and 5% sucrose to select for loss of the plasmid-encoded sacB gene. Single sucrose-resistant colonies were also checked for loss of the plasmid-encoded tetracyclin-resistance marker and verified by PCR using the above primers to generate a roughly 2.1 kb product, indicating a successful gene replacement.
Construction and analysis of gene fusions to lacZ
A 340 bp EcoRI fragment containing the intergenic region between PA2424 and PA2425 was excised from the cloned 1105 bp PA2424 PCR fragment (see above). Ligation into the EcoRI site of pPZTC, a modified version of pPZ30 (Schweizer, 1991), yielded a transcriptional reporter fusion of PA2424 to the lacZ gene (pPZ2424). Similarly, the PA2425 promoter region was amplified by PCR as described above, using the primers PA2425-forw (5′TGCGGAGGGGGCTG CAGAG and PA2425-rev (5′gtCGACCTCCTTGCGCCGCTG, SalI-site underlined, non-matching bases lowercase) and cloned into pCRII-2.1. This DNA fragment was then directionally cloned into pPZTC using the EcoRI and SalI sites, yielding a transcriptional PA2425-lacZ fusion (pPZ2425). Plasmids pPZ2424, pPZ2425 and the pPZTC control vector were transformed into P. aeruginosa PAO1 and an isogenic ΔpvdS mutant (Ochsner et al., 1996). Three individual colonies of each were grown for 16 h in low iron or high iron CAA medium at 32°C, and β-galactosidase activities were determined in soluble cell extracts (Sambrook et al., 1989).
Structural modelling for prediction of PvsA-M3 A-domain
A three-dimensional model of PvsA-M3 A-domain (aa 2203–2670 of PvsA) was generated using the comparative protein modelling server SWISS-MODEL (Guex and Peitsch, 1997), based on the coordinates of Bacillus brevis GrsA phenylalanine A-domain, which was solved at 1.9 Å resolution (Conti et al., 1997, and accession 1AMU in the Protein DataBank Brookhaven National Laboratory, Upton, N.Y). PvsA-M3 showed 32% identity with the corresponding sequence of GrsA. The model was submitted with L-TOPA docked in the binding pocket maintaining the positions of the α-amino and carboxy group of the amino acid substrate phenylalanine crystallized with GrsA A-domain. The quality of the model was assessed by using a 3D-1D profile verification model (Luthy et al., 1992) and Prosa II (Sippl, 1993), originally built within SWISS-MODEL. The results showed that the model was comparable to the GrsA template structure, indicating that PvsA-M3 can fold like GrsA. The resulting model was then manipulated with WebLab Viewer Pro software (Accelrys, Princeton N.J).
This work has been partially funded by the ‘Jean and Alphonse Forton’ fund against cystic fibrosis. Sequencing was realized thanks of the support of FWO (Fonds voor Wetenschappelijk Onderzoek). We thank Dr Iain Lamont for interesting discussions.
The GenBank accession number for the sequence reported in this manuscript is AF237701.
Fig. S1. Comparison of AL (acyl CoA ligase) motifs of PvsA ( P. fluorescens ATCC 17400), PvdL ( P. aeruginosa ), and homologs from P. putida KT2440 and P. syringae DC3000.
Fig. S2. A. Determination of the eight amino acids lining the binding pockets of adenylation domains for modules M2, M3 and M4 of PvsA, compared to known activation domains for l -Glu, l -Tyr, and l -Dab.
B. Predicted structures for the specificity pockets of the l-tyrosine activating domain from the fengycin synthetases FenA-M3 and FenD-M1 (left), and of PvsA-M3 with TOPA as substrate (center) or l-tyrosine (right).