, Volume 26, Issue 4, pp 561–575

A combinatorial approach to the structure elucidation of a pyoverdine siderophore produced by a Pseudomonas putida isolate and the use of pyoverdine as a taxonomic marker for typing P. putida subspecies


  • Lumeng Ye
    • Research Group Microbiology, Department of Bioengineering Sciences, VIB Department of Structural BiologyVrije Universiteit Brussel
  • Steven Ballet
    • Research Group of Organic Chemistry, Department of ChemistryVrije Universiteit Brussel
  • Falk Hildebrand
    • Research Group Microbiology, Department of Bioengineering Sciences, VIB Department of Structural BiologyVrije Universiteit Brussel
  • Georges Laus
    • Research Group of Organic Chemistry, Department of ChemistryVrije Universiteit Brussel
  • Karel Guillemyn
    • Research Group of Organic Chemistry, Department of ChemistryVrije Universiteit Brussel
  • Jeroen Raes
    • Research Group Microbiology, Department of Bioengineering Sciences, VIB Department of Structural BiologyVrije Universiteit Brussel
  • Sandra Matthijs
    • Institut de Recherches Microbiologiques—Wiame
  • José Martins
    • NMR and Structure Analysis Unit, Department of Organic ChemistryUniversiteit Gent
    • Research Group Microbiology, Department of Bioengineering Sciences, VIB Department of Structural BiologyVrije Universiteit Brussel

DOI: 10.1007/s10534-013-9653-z

Cite this article as:
Ye, L., Ballet, S., Hildebrand, F. et al. Biometals (2013) 26: 561. doi:10.1007/s10534-013-9653-z


The structure of a pyoverdine produced by Pseudomonas putida, W15Oct28, was elucidated by combining mass spectrometric methods and bioinformatics by the analysis of non-ribosomal peptide synthetase genes present in the newly sequenced genome. The only form of pyoverdine produced by P. putida W15Oct28 is characterized to contain α-ketoglutaric acid as acyl side chain, a dihydropyoverdine chromophore, and a 12 amino acid peptide chain. The peptide chain is unique among all pyoverdines produced by Pseudomonas subspecies strains. It was characterized as –l-Asp-l-Ala-d-AOHOrn-l-Thr-Gly-c[l-Thr(O-)-l-Hse-d-Hya-l-Ser-l-Orn-l-Hse-l-Ser-O-]. The chemical formula and the detected and calculated molecular weight of this pyoverdine are: C65H93N17O32, detected mass 1624.6404 Da, calculated mass 1624.6245. Additionally, pyoverdine structures from both literature reports and bioinformatics prediction of the genome sequenced P. putida strains are summarized allowing us to propose a scheme based on pyoverdines structures as tool for the phylogeny of P. putida. This study shows the strength of the combination of in silico analysis together with analytical data and literature mining in determining the structure of secondary metabolites such as peptidic siderophores.


Pseudomonas putidaPyoverdineStructure elucidationBioinformatic predictionPhylogenetic marker


Pseudomonas is a genus of non-fermentative Gram-negative Gammaproteobacteria with wide distribution, such as surfaces of virtually all plant tissues, soil, organic matter and water (Goldberg 2000). The ubiquitous character of these microorganisms explains the diversity of secondary metabolites produced by Pseudomonas species, especially the fluorescent Pseudomonas (rRNA group I). Fluorescent Pseudomonas strains produce pyoverdines or other siderophores to chelate iron(III). So far, more than 70 pyoverdines have been well characterized, and they all consist of three distinct structural parts: a quinoline-1-carboxylic acid containing a chromophore responsible for the green fluorescence, a dicarboxylic acid or its monoamide, bound amidically to the 5-amino group of the chromophore and a peptide chain comprising 6–14 amino acids bound to its 1-carboxyl group (Meyer 2000; Visca et al. 2007; Cornelis 2010). The dihydroxyquinoline structure of the chromophore is identical for all pyoverdines due to the conservation of the biosynthetic gene pvdL among different species of Pseudomonas strains (Mossialos et al. 2002; Ravel and Cornelis 2003). However, slight differences in the chromophore structure have been reported for some strains (Jacques et al. 1995; Budzikiewicz et al. 2007). Different acyl side chains can be observed in isoforms of pyoverdine isolated from one single strain (Meyer 2000). The primary difference among different pyoverdines is to find in their peptide chain organization. Both the chromophore and the peptide chain of pyoverdines are synthesized by a thiotemplate mechanism catalyzed by non-ribosomal peptide synthetases (NRPS) (Ravel and Cornelis 2003; Visca et al. 2007). NRPSs are modularly assembled mega-enzymes, and each module is organized with domains with different catalysis functions. The most essential domains are: a condensation domain (C domain) which catalyzes the formation of the peptide bond, thiolation (T domain) and peptide carrier protein (PCP), and adenylation domain (A domain) which activates an amino acid substrate (Challis et al. 2000; Crosa and Walsh 2002). The resolution of the crystal structure of the phenylalanine-activating adenylation domain PheA illustrated that there are signature residues in the binding pocket of the A domain, controlling substrate recognition (Konz and Marahiel 1999; Stachelhaus et al. 1999). The substrate specificity of different A domains makes the prediction of amino acid sequence of the synthesized peptide possible by in silico studies, using publicly available softwares such as the PKS–NRPS (Bachmann and Ravel 2009) and antiSMASH (Rottig et al. 2011). The final configuration of the amino acid (l- or d-form) can also be predicted by the detection of an epimerization domain (E domain) downstream of the A domain. Thanks to these bioinformatic tools, researchers in the natural product field have already discovered a great number of novel non-ribosomal peptides, especially in the cases of orphan gene clusters and silent gene clusters in sequenced microorganisms (Gross et al. 2007; Rosconi et al. 2013). Different types of pyoverdines can be observed at species and even at sub-species levels as visualized by different band patterns in isoelectric focusing (IEF) gels, a method termed “siderotyping”, thus suggesting that this technique can be used as a taxonomic tool (Koedam et al. 1994; Fuchs et al. 2001; Meyer et al. 2007, 2008). Siderotyping confirmed the existence of 3 types of pyoverdines produced by different strains of Pseudomonas aeruginosa (Cornelis et al. 1989; Meyer et al. 2007), while conversely different Pseudomonas syringae strains corresponding to different pathovars were found to all produce the same pyoverdine (Bultreys et al. 2001). However, in the case of Pseudomonas fluorescens and Pseudomonas putida, which are less clearly defined species and rather form complexes of species, more information is still needed to be collected and the diversity of pyoverdines further investigated. Recent examples are the originally named P. fluorescens Pf-5 and P. fluorescens CHA0, which were proposed to belong to a new species, Pseudomonas protegens (Ramette et al. 2011). Interestingly, the pyoverdines produced by these two strains are almost identical and differ from the pyoverdines produced by other P. fluorescens strains. They share the same amino acid sequence in the peptide chain, but the sixth residue is d-Ala in Pf-5, and l-Ala in CHA0 (Hartney et al. 2013). In this study, we define the structure of an unusual pyoverdine produced by P. putida W15Oct28, isolated from the water surface of the Woluwe river in Belgium (Pirnay et al. 2005) by combining the data of a series of chemical analyses and a bioinformatic interpretation of the biosynthetic genes. Furthermore, we review all the pyoverdine structures produced by P. putida both from the literature and made a comparison of bioinformatic analysis of the whole genome sequenced strains. The correlation of the pyoverdine structures with the phylogeny of the producing strain by whole genome comparison is further validated, which confirms the potential of using pyoverdine as taxonomic and phylogenetic markers to distinguish P. putida subspecies (Meyer et al. 2007).

Materials and methods

Large scale pyoverdine purification

Pseudomonas putida W15Oct28 was grown at 28 °C in 1 l of iron-poor CAA medium (BactoCasamino Acid, BD, 5 g l−1; K2HPO4 1.18 g l−1; MgSO4·7H2O 0.25 g l−1) in 5 l Erlenmeyer flasks, at a shaking speed of 160 rpm for 48 h. Bacterial cells were removed by centrifugation at 10,000×g during 15 min. After filtration the supernatant was passed on a C-18 column that was activated with methanol and washed with distilled water. Elution was done with acetonitrile/H2O (70/30 %). Samples were lyophilized after most of the acetonitrile was evaporated (Matthijs et al. 2009).

High-performance liquid chromatography (HPLC), gas chromatography (GC) and mass spectrometry (MS)

HPLC analysis was carried out with an Agilent 1100 Series HPLC System with auto-sampler and degasser UV detector and thermostated column compartment with an operating temperature range from ambient to 105 °C. Preparative-scale purification was done on Gilson 712 semi-preparative HPLC system with a 322 pump, an UV–Vis 156 detector, a manual injector and 206 Fraction Collector. ESI–LC/MS was done by means of a Waters 600E HPLC Pump, a Waters 2487 Dual Absorbance Wavelength Detector and a Fisons VG II Quattro Mass Spectrometer (ESP ionisation). Operating temperature is from ambient to 80 °C. The mobile phase consists of a water/acetonitrile/TFA mixture with a gradient going from a mixture water/AcN (97:3) containing 0.1 % TFA to a mixture water/AcN (0:100) containing 0.1 % TFA in 30 min followed by 10 min isocratic run at these conditions, and with a flow rate of 20 ml min−1. High resolution MS and collision-induced dissociation tandem fragmentation MS were done on a QTof Micro MS (positive ion mode). Under standard measurement conditions the sample was dissolved in CH3CN/H2O (1:1) containing 0.1 % TFA. GC/MS spectra were recorded on a Trace MS Plus (Thermo). Separation was done on a J&W Scientific DBxXLB (30 m, 0.25 mm ID, 0.25 μm film thickness).

Amino acid analysis

Partial hydrolysis was realized by subjecting 4.2 mg of pure pyoverdine sample to 1.5 ml of 6 M HCl at 110 °C for 2 h, followed by ESI–LC/MS analysis. Full hydrolysis was done by using 4.8 mg of pyoverdine pure sample in 1.5 ml of 6 M HCl at 110 °C for 24 h. The amino acids were derivatized as N-TFA-n-butylesters according to Leimer et al. (1977). More precisely, the HCl solution was evaporated by gas N2, then n-butanol·3 N HCl is added and incubated at 100 °C for 15 min for esterification. After allowing the flask to cool down, the n-butanol·3 N HCl is removed by evaporation. Two ml of dichloromethane is added to azeotrope any remaining water and then removed by evaporation. Two ml of dichloromethane and 1 ml of TFAA are added to acylate the sample at 150 °C for 5 min; the derivatized sample is kept at 4 °C for GC/MS analysis.

Genome sequencing and bioinformatic analysis of the pyoverdine biosynthetic genes

The genome of P. putida W15Oct28 was sequenced at the VIB nucleomics core by the IlluminaMiseq system. The library was constructed using the Nextera kit, read length as 150 bp paired end. The coverage is about 62 times, and the sequencing reads were assembled by Velvet (Zerbino and Birney 2008), with 99 contigs representing the draft genome. Each contig was uploaded to antiSMASH to detect the pyoverdine biosynthetic genes cluster. The pyoverdine biosynthesis genes clusters were annotated using the RAST website (Aziz et al. 2008). To predict the substrate of the different adenylation domains, antiSMASH gives 4 predictions based on NRPSPredictor2 SVM, Stachelhaus code, Minowa, and consensus. When the 4 predictions are identical, then the proposed activated amino acid is accepted. When the 4 predictions are different, the amino acids sequence of the adenylation domain will be further analyzed by PKS–NRPS to compare with the Stachelhaus code. The three dimensional model of pyoverdine–Fe3+ complex is generated by drawing the peptide structure in the program Molecular Operating Environment (MOE) (2012). The complexation is proposed, not based on any experimental results. Restrictions are applied and the structure is energy minimized with the MMFF94x force field to obtain the presented structure. All structures were visualized using Pymol


Characterization of the α-ketoglutaric acid side chain, dihidropyoverdinechromophore and aspartic acid as the first amino acid of the peptide chain

Unlike most of fluorescent-Pseudomonas sp. strains which produce several isoforms of pyoverdine (mainly due to the difference of dicarboxylic acid side chain) when they grow in iron limited condition, only one form of pyoverdine was detected from the extraction of CAA culture of P. putida W15Oct28. The mass of the pyoverdine was found to be 1624.6 Da. ([M+H]+, 1625.6404 m/z) by high resolution MS. When fragmentation was attempted by CID tandem MS, just a few signals were however detected, which only allowed us to interpret part of the structure. The positively charged ion m/z signals 70 and 115 indicate the presence of ornithine in the peptide chain, but not a cyclo-hydroxy-ornithine at the end of peptide chain (Meyer et al. 2008). MS signals 439 and 501 m/z, indicate that the chromophore is linked with a α-ketoglutaric acid side chain on the 5-amino group, and that an aspartic acid is present as the first amino acid coupled to its 1-carboxyl group, as is the case of pyoverdines produced by P. monteilii Lille 1 and P. putida GS37 (Meyer et al. 2008). The 62 Da difference between signal 501 and 439 m/z confirms that the side chain is α-ketoglutaric acid (loss of [H2O + CO2] for a Kgl side chain, 62 Da) (Budzikiewicz et al. 2007). LC/MS analysis of partially hydrolyzed pyoverdineas described in “Materials and methods” gave two major products. The first product has a positively charged signal of 386 m/z, which corresponds to the α-ketoglutaric acid plus the chromophore. The second gives a signal of 501 m/z, which represents the α-ketoglutaric acid plus the chromophore plus Asp. The 115 Da difference between signal 501 and signal 386 m/z confirms that the first amino acid of the peptide chain is aspartic acid. Surprisingly, the chromophore of the P. putida W15Oct28 pyoverdine is a dihydropyoverdine, for 2 more protons were calculated based on signal 386 m/z ([M+H]+) compared to the common chromophore structure (1 S)-5-amino-2,3-dihydro-8,9-dihydroxy-1H-pyrimido-[1,2-a)quinoline-1-carboxylic acid (Budzikiewicz 2004). The positively charged signal 189 m/z is found as a result of CID tandem spectrums analysis. This signal is characteristic for the ketoglutaric acid-dihydropyoverdine composition (Budzikiewicz et al. 2007) (Table 1).
Table 1

Signal assignment of the different pyoverdine fragments

Signal, m/z



Ornithine signature


Dihydropyoverdine chromophore signature


[kgl + dihydropyoverdinechromophore + Asp]-CO2-H2O


kgl + dihydropyoverdinechromophore + Asp


[kgl + dihydropyoverdinechromophore + Asp + Ala]-CO2-H2O


kgl + dihydropyoverdinechromophore + Asp + Ala


[kgl + dihydropyoverdinechromophore + Asp + Ala + AOHOrn]-CO2-H2O


[kgl + dihydropyoverdinechromophore + Asp + Ala + AOHOrn]-CH3


kgl + dihydropyoverdinechromophore + Asp + Ala + AOHOrn


[pyoverdine + 2H]2+


pyoverdine-CO2-H2O-[Orn + Ser]-OHAsp-Hse-Ser-Hse


pyoverdine-CO2-H2O-[Orn + Ser]-OHAsp-Hse-Ser


pyoverdine-[kgl + dihydropyoverdinechromophore + Asp]


pyoverdine-CO2-H2O-[Orn + Ser]-OHAsp-Hse


pyoverdine-CO2-H2O-[Orn + Ser]-OHAsp


pyoverdine-CO2-H2O-[Orn + Ser]





Kgl α-ketoglutarate, AOHOrnNδ-acetyl-Nδ-hydroxy-ornithine, OHAsp hydroxyl aspartic acid, Hse homoserine

Determination of the amino acid composition of the peptide chain

Amino acid analysis was done by subjecting the N-TFA-n-butylesters of the amino acid derivates of the fully hydrolyzed pyoverdine to GC/MS analysis. Spectra show characteristic signals for alanine, glycine, threonine, serine, aspartic acid, ornithine, and one fraction with ion signals as 126, 140, 152, and 188 m/z, at retention time 21’76 (supplementary Fig. 2). Twelve adenylation domains were detected by bioinformatic analysis of the pyoverdine biosynthesis cluster found in scaffold 32 (Fig. 1). Bioinformatic analysis using the antiSMASH website provides a more detailed amino acid composition of the peptide chain. The result confirms that the first amino acid of the peptide chain is l-aspartic acid (signature residues: DLTKVGHV). The second amino acid is predicted to be l-alanine (signature residues: DLWNNALT), the third amino acid is predicted to be an ornithine (signature residues: DGEICGGV) and it is in d-configuration due to the presence downstream of an epimerization domain. The fourth to sixth amino acids are l-threonine-l-glycine-l-threonine (the signature residues of the two l-threonines are both: DFWNVGMV, whereas the signature residues of the glycine is DILQLGMI). The seventh amino acid could not be clearly annotated neither by antiSMASH nor by the PKS–NRPS software, while the eighth amino acid is an aspartic acid in d-configuration (signature residues: DLTKVGHVGK, epimerization domain and d to l type condensation domain downstream). The ninth amino acid and the last (12th) amino acid are predicted to be l-serine (signature residues: DVWHVSLI). The tenth amino acid gave no clear annotation by antiSMASH, however the signature residues in the binding pocket indicates that the substrate is l-ornithine and the signature residues DGEDHGTV of l-ornithine are unique to some P. putida strains like KT2440 and BIRD-1 (Table 2). The eleventh amino acid has no definite annotation by antiSMASH, however the signature residues in the binding pocket are the same as the seventh adenylation domain (signature residues: DLKNLGSD) which indicates that the seventh and the eleventh amino acid are identical. When comparing the pyoverdine biosynthesis genes of P. putida W15Oct28 using the BLAST algorithm, a non-ribosomal peptide synthetase for azotobactin synthesis from the genome of Azotobacter vinelandii DJ was noticed (Setubal et al. 2009). A. vinelandii should be considered to belong to the Pseudomonas genus and produces the siderophore azotobactin D, which is in fact a pyoverdine (Rediers et al. 2004; Ozen and Ussery 2012). Azotobactin D peptide chain has been determined as being l-Asp-d-Ser-l-Hse-l-Gly-d-β-threo-OH Asp-l-Ser-d-Cit-l-Hse-d-AOHOrn-l-Hse (Demange et al. 1988). Bioinformatic analysis of the azotobactin biosynthesis genes by antiSMASH gave a similar amino acid sequence of the peptide chain, but 3 adenylation domains (the third, eighth, and tenth) did not yield a clear annotation. From further analysis of these three adenylation domains by PKS–NRPS it turns out that the signature residues in the binding pocket of the eighth A domain (DLKNLGTD) are very similar to those predicted in the seventh and the eleventh A domains (DLKNLGSD) in the pyoverdine synthetase of P. putida W15Oct28 while the signature residues of the third and eighth A domains of azotobactin synthetase are identical (DLKNLGT-). The third, eighth, and tenth amino acids of the azotobactin peptide chain are all homoserine, which indicates that the corresponding A domains of azotobactin synthetase all recognize and activate homoserine as a substrate. Given the high similarity between the signature residues in the binding pocket of the two unclear A domains of the W15Oct28 pyoverdine and the three A domains of azotobactin synthetase which activate homoserine, we assumed that the two undetermined amino acids in the peptide chain of the P. putida pyoverdine are in both cases homoserine. An undetermined fraction in GC/MS analysis of the N-TFA-n-butylester derivatives of fully hydrolyzed pyoverdine at 21’76 which matches the relative retention time and double peak pattern of derivatized homoserine (Siezen and Mague 1977). A positively charged signal of 1,124 m/z was detected by tandem CID MS and corresponds to the full mass of pyoverdine when calculated together with the fragment 501 m/z, which representing the other part of the pyoverdine (peptide chain without the first amino acid). Positively charged signals 510 m/z (kgl plus dihydro-chromophore plus Asp plusAla-[H2O+CO2]) and 572 m/z (kgl plus dihydrochromophore plus Asp plus Ala) were detected, which confirm the second amino acid to be an alanine. Signals 682 and 744 m/z indicated that the third element of the peptide chain is a Nδ-acetyl-Nδ-hydroxy-ornithine (AOHOrn). Additionally, a series of fragments with signals at 1,405 m/z (Δ 219 Da compared to 1,624 m/z, -[Orn+Ser]), 1,361 m/z (Δ 44 Da compared to previous signal, -CO2), 1,230 m/z (Δ 131 Da compared to previous signal, -OHAsp), 1,129 m/z (Δ 101 Da compared to previous signal, -Thr or Hse), 1,042 m/z (Δ 87 Da compared to previous signal, -Ser), and 941 m/z (Δ 101 Da compare to previous signal, -Thr or Hse), indicate an amino acid sequence as Thr/Hse-Ser-Thr/Hse-OHAsp-Ser-Orn. However, this sequence did not match the prediction of amino acid sequence by the bioinformatic analysis. These results could however be rationalized by the formation of acyclic peptide structure between the last amino acid (serine) and the sixth amino acid (threonine) (Table 1). In summary, we conclude that the structure of the P. putida W15Oct28 pyoverdine is α-ketoglutaricacid-dihydropyoverdine chromophore–l-Asp-l-Ala-d-AOHOrn-l-Thr-Gly-c[l-Thr(O-)-l-Hse-d-Hya-l-Ser-l-Orn-l-Hse-l-Ser-O-].The predicted chemical formula and the calculated molecular weight of this compound are: C65H93N17O32, detected mass 1624.6404 Da, calculated mass 1624.6245 (Fig. 2 and supplementary Fig. 3).
Fig. 1

Pyoverdine non-ribosomal peptide chain synthetases of P. putida W15Oct28 and A. videlandii DJ. The different domains are predicted by antiSMASH. The letter C represents the condensation domain, the letter A an adenylation domain, the grey ellipse represents the peptidyl carrier protein, the letter E designates an epimerization domain, and letters TE correspond to the thioesterase domain. The eight capital letters above each module correspond to the signature amino acid residues in the binding pocket of each adenylation domain. The signature residues were extracted from the PKS–NRPS website. The amino acid abbreviations under each adenylation domains represent the substrate of each adenylation domain. The adenylation domain substrates of P. putida W15Oct28 pyoverdine synthetase were predicted by antiSMASH and PKS–NRPS. The adenylation domain substrates of A. videlandii DJ pyoverdine synthetase were annotated by antiSMASH and according to the published structure of azotobactin D (Demange et al. 1988). In the middle of the picture, the pathway of l-ornithine hydroxylation by PvdA is shown, which can be further acetylated by PvdYII to form Nδ-acetyl-Nδ-hydroxy-ornithine

Table 2

Peptide sequences of different fluorescent pseudomonads pyoverdines, including all P. putida pyoverdines

Bioinformatic analysis: Prediction by the 8 signature residues of the docking domain. Number indicates the aa position in the peptide chain

aAmino acid composition from Matthijs et al. (2009), stereology from bioinformatic analysis of this study

bPartial structure predicted from genome sequence/d-configuration
Fig. 2

The strategy used to elucidate the structure the pyoverdine produced by P. putida W15Oct28. The structure is presented at the bottom of the figure. The whole procedure is described in the “Results and Discussion” section

Pyoverdine biosynthesis genes clusters of P. putida W15Oct28

Like in most of fluorescent Pseudomonas strains, the pyoverdine biosynthesis genes of P. putida W15Oct28 are found in separated clusters distributed over 3 contigs (Fig. 3) (Ravel and Cornelis 2003). The pvdL, pvdS, and pvdY genes are adjacent in contig63 (peg4496-4498). PvdL is a NRPS encoding the chromophore synthetase (Mossialos et al. 2002), and PvdS is an extracytoplasmic sigma factor (ECF σ) involved in the transcription of pyoverdine biosynthesis genes (Visca et al. 2007). The P. putida W15Oct28 pvdY gene is determined to bepvdYII by BLAST, a gene which was reported as encoding an acetylase that facilitates the acetyl modification of Nδ-hydroxyornithine in type II pyoverdine producing P. aeruginosa strains (Lamont et al. 2006). Most type II pyoverdine producing P. aeruginosa strains (Pa4, PA7, and M18) possess both pvdF (encodes Nδ-hydroxyornithine formyltransferase) and pvdYII genes (Lamont et al. 2006). However, no AOHOrnis found in type II pyoverdine, suggesting that PvdYII may be responsible to the formation of cOHOrn at the end of the peptide chain (Lamont et al. 2006). In most of P. putida strains, such as P. putida KT2440, the pvdY gene is determined as pvdYII and pvdF is not found. Thus, the ornithine at the second position in the peptide chain of the pyoverdine produced by P. putida KT2440 has no modification, nor is there the presence of a cOHOrn at the end of the peptide chain. However, in P. fluorescens Pf0-1 and A. vinelandii DJ, which similarly have pvdYII and no pvdF either, acetyl-modified ornithines (AOHOrn) are found in their pyoverdines. There is no pvdF gene in the genome of P. putida W15Oct28, therefore the third amino acid of the pyoverdine peptide chain is AOHOrn which is modified by PvdYII, and the tenth amino acid is an unmodified ornithine. One major group of pyoverdine synthesis genes is clustered together in contig 32 of the draft genomic sequence of P. putida W15Oct28. This group starts with the 2-oxoglutarate aminotransferase gene pvdH, a gene encoding the SyrP protein, mbtH coding for a hypothetical MbtH-like protein, the thioesterase gene, and four nonribosomal peptide synthetase genes (pvdI/J/K/D) for the peptide chain synthesis (Fig. 3). After the NRPS genes follow in the opposite orientation the fpvA gene encoding the TonB-dependent ferripyoverdine receptor, pvdE coding for a pyoverdine ABC export protein, and the pvdN/O/M/P genes encoding periplasmic enzymes involved in the pyoverdine chromophore maturation (Cornelis 2010; Yeterian et al. 2010), the last gene being in opposite orientation One gene potentially involved in pyoverdine efflux terminates this cluster, however, we could not detect the genes encoding the tripartite efflux system PvdTR-OpmQ described in P. aeruginosa to be involved in the recycling of the apo-pyoverdine (Hannauer et al. 2012; Schalk and Guillon 2013b). This genes clustering is conserved in the pyoverdine synthesis genes of other P. putida strains, with the exception of one non-ribosomal peptide synthetase which has higher similarity with the peptide synthetase of azotobactin D, which in turn activates homoserine as substrate. In addition to the NRPS, the pvdN/O/M/P genes show much higher identities with the corresponding genes of Pseudomonas sp. TJI-51 (about 77 %) than with the genes of other P. putida strains (<40 %, Fig. 3). The periplasmic enzymes PvdN, PvdO, and PvdP were reported to be involved in the formation of the chromophore after transport of the pyoverdine precursor ferribactinby PvdE from the cytoplasm to the periplasm (Yeterian et al. 2010; Schalk and Guillon 2012). In the case of P. aeruginosa PAO1, a mutation in one of the genes pvdN, pvdO, and pvdP resulted in the accumulation of the pyoverdine precursor in the periplasm and prevented the secretion of pyoverdine into the growth medium (Yeterian et al. 2010). In the case of P. putida W15Oct28, the differences observed in the sequence of these periplasmic enzymes may be the reason why only the dihydropyoverdine chromophore could be identified. A phylogenetic tree of the W15Oct28 PvdN protein is presented in Fig. 4, which clearly shows that it does not cluster with the other PvdN sequences from P. putida strains. The iron transport genes (fpvCDEF) (Schalk and Guillon 2013a) are clustered in contig 13, together with the gene pvdA downstream, which encodes the l-ornithine 5-monooxygenase. The gene encoding the other periplasmic enzyme, the PvdQ acylase(Jimenez et al. 2010), is found singly in another contig (peg 2528) and shares about 91 % identity with the pvdQ gene of other P. putida strains.
Fig. 3

A schematic representation of the pyoverdine biosynthesis genes clusters of P. putida W15Oct28. The colored arrows represent genes involved in pyoverdine biosynthesis and transport. The numbers above the arrows represent their position in the draft genome. The function of the genes, when known, is annotated above them. Under each gene, the most similar homolog with highest coverage and identity by protein BLAST is mentioned. (Color figure online)
Fig. 4

Phylogenetic analysis of the chromophore maturation protein PvdN by amino acid alignment obtained by Clustal W (Larkin et al. 2007). The compiled best hits from PSI_BLAST of the PvdN gene of P. putida W15Oct28 were aligned. Abbreviations for strains represented in the tree are as follows: CKveEllin345 (Candidatus Koribacterversatilis Ellin345), CSusEllin6076 (Candidatus Solibacterusitatus Ellin6076), CseATCC21756 (Caulobacter segnis ATCC 21756), PaeATCC700888 (P. aeruginosa ATCC700888), PaeE2 (P. aeruginosa E2), PaePA14 (P. aeruginosa PA14), PaePAO1 (P. aeruginosa PAO1), PenL48 (P. entomophila L48), PflA506 (P. fluorescens A506), PflF113 (P. fluorescens F113), PflSBW25 (P. fluorescens SBW25), PflSS101 (P. fluorescens SS101), PfuUPB0736 (P. fuscovaginae UPB0736), PpoaeRE*1-1-14 (P. poae RE 1-1-14), PprPF-5 (P. protegens Pf-5), PpsL19 (P. psychrotolerans L19), PpuBIRD-1 (P. putida BIRD-1), PpuDOT-T1E (P. putida DOT-T1E), PpuF1 (P. putida F1), PpuGB-1 (P. putida GB-1), PpuKT2440 (P. putida KT2440), PpuW619 (P. putida W619), Psp.GM24 (Pseudomonas sp. GM24), Psp.TJI-51 (Pseudomonas sp. TJI-51), PsyBG33R (P. synxantha BG33R), Pex14-3substr. 14-3b (P. extremaustralis substr. 14-3b), PsyL1448A (P. syringae pv. phaseolicola 1448A), PsyD3000 (P. syringae pv. tomato str. DC3000), PspadixBD-a59 (Pseudoxanthomonas spadix BD-a59), PpuS16 (P. putida S16), TroseusDSM18391 (Terriglobus roseus DSM 18391), TsaSP1PR4 (Terriglobus saanensis SP1PR4), XalGPEPC73 (Xanthomonas albilineans GPE PC73), XgaATCC19865 (X. gardneri ATCC 19865), Xpe91-118 (X. perforans 91-118), Xtrpv.grART-Xtg29 (X. translucens pv. graminis ART-Xtg29), Xtrpv.trDSM18974 (X. translucens pv. translucens DSM 18974), Xcapv.vaNCPPB702 (X. campestris pv. vasculorum NCPPB 702), PpuW15Oct28 (P. putida W15Oct28)

Pyoverdines structure of P. putida strains and confirmation of the usefulness of pyoverdines as phylogenetic markers

In contrast to strains of the species P. syringae which all produce the same pyoverdine, most representatives of fluorescent Pseudomonas species produce several types of pyoverdines. In the well-studied species P. aeruginosa, three types of pyoverdines were determined and are associated with three types of receptors, respectively (Cornelis et al. 1989; Meyer et al. 1997; de Chial et al. 2003; Smith et al. 2005; Bodilis et al. 2009). In the case of the species P. fluorescens and P. putida, a large number of pyoverdines structures was reported (Table 2). With the structure of the P. putida W15Oct28 pyoverdine in hand, we decided to compare the different known P. putida pyoverdines. Furthermore, the pyoverdine structures of available whole genome sequenced P. putida strains were predicted by bioinformatics analysis (Table 2). Altogether, the pyoverdines can be divided into several groups, which provide more information for siderotyping of P. putida strains and support the proposal to use pyoverdines as phylogenetic markers for fluorescent Pseudomonas strains, especially within the species P. putida (Meyer et al. 2007). Thirteen genomes of P. putida strains, two genomes of Pseudomonas sp. strains (UW4 and GM78), the genome of P. entomophila L48 (a close relative of P. putida), and the genome of A. vinelandii DJ were analyzed by antiSMASH to predict the amino acid composition of the peptide chains of their respective pyoverdines. The pyoverdine of P. putida KT2440 is predicted to be the same as reported by Matthijs et al. (2009), with the third amino acid being a hydroxyl-aspartic acid in d-configuration. The pyoverdines of P. putida KT2440, P. putida BIRD-1, and P. putida S12 were predicted to be identical by antiSMASH and correspond to the pyoverdine of P. putida G4R (Salah-el-Din et al. 1997). The pyoverdine of P. putida GB-1 was predicted to be similar to the pyoverdine of P. putida L1, CFBP2461, and WCS358, with the fifth allothreonine amino acid missing. The pyoverdine of P. putida S16 was predicted to belong to this type, but with four amino acids missing (from 5th to 8th position, aThr-Ala-Thr-Lys). The third major type of pyoverdines produced by P. putida strains share similar structure with the pyoverdine produced by P. putida PutC. The pyoverdines of P. putida SJTE-1 and P. putida F1 have the same amino acid sequence and configuration as the pyoverdine of P. putida PutC. The modification of the second amino acid (OHbutOHOrn) is very rare among the pyoverdine structures, thus the precise structure of pyoverdines produced by SJTE-1 and F1 should be further confirmed by MS. There are more pyoverdines produced by genome-sequenced P. putida strains which are assumed to belong to this group, however, with one or more amino acid missing (P. putida ND-6, B6-2, and LS46).There exists a bias for the amino acid substrates incorporated in pyoverdines as compared to lipopeptides, which are also synthesized by NRPS, an observation made by Caboche et al. (2010) who found that different classes of bioactive peptides have a preference for certain amino acids. For example, leucine, isoleucine, proline, tryptophan, cysteine, and methionine are commonly found in lipopeptides, but have never been described to be present in pyoverdines. Valine and histidine were reported to be included in the structure of some pyoverdines, but with really rare records. Some non-proteinogenic amino acids including diamino-butanoic acid (Dab), ornithine, homoserine, and citrulline were also found in pyoverdines. As discussed before, the structural and genetic information of azotobactin D gave crucial clue of the existence of homoserines in the pyoverdine produced by P. putida W15Oct28. When analyzing the genomes of P. entomophila L48 and Pseudomonas sp. UW4, the third adenylation domains of the pyoverdine peptide synthetases of both strains gave no clear annotation, neither by antiSMASH nor by PKS–NRPS websites. The signature residues of these two adenylation domains are quite similar (DSAAIAEV for P. entomophila L48, and DSALIAEV for Pseudomonas sp. UW4). Adenylation domains with similar Stachelhaus codes were found inthe Norine and SBSPKS databases (Caboche et al. 2008; Anand et al. 2010), which are the tenth adenylation domain of bacitracin synthetase and the fourth adenylation domain of bleomycin synthetase both of them activating histidine as substrate. Only one pyoverdine containing histidine was reported with the sequence: (d)Ser-(l)eLys-(l)OHHis-(d)aThr-(l)Ser-(l)cOHOrn, produced by both P. fluorescens 9AW and P. putida 9BW (Budzikiewicz et al. 1997). The pyoverdine of Pseudomonas sp. UW4 is designated as the same as the pyoverdine produced by both P. fluorescens 9AW and P. putida 9BW by bioinformatics analysis. Strain UW4 was reassigned as being a Pseudomonas sp. strain which belongs to the fluorescens group, jessenii subgroup recently after being first designated as a P. putida strain (Duan et al. 2013). This result may suggest us to reinvestigate the taxonomy of both P. fluorescens 9AW and P. putida 9BW. The pyoverdine of P. entomophila L48 is predicted to be different from the structure reported by Matthijs et al. (2009) and in which the second amino acid still remains undetermined.


Pyoverdines are peptidic siderophores synthesized by NRPS. The biosynthesis of pyoverdine is activated when the producing strain encounters iron limited environment, in order to use this siderophore as a high-affinity ferric iron chelator. To decrease the chance that self-produced life support goods be utilized by other strains or species, in Pseudomonas, the specificity of uptake of ferri-pyoverdines is conferred by a specific interaction with a cognate receptor in the outer membrane. The diversity in peptide chains of pyoverdines implies that each ferri-siderophore is recognized by a specific receptor, suggesting a co-evolution between the modular NRPS enzyme and the receptor (Smith et al. 2005; Bodilis et al. 2009). Thus, the peptide chains of pyoverdines differ among species and even among strains in the same species (Meyer 2000; Meyer et al. 1997). The structure of the pyoverdine produced by P. putida W15Oct28 could finally be elucidated by using a combination of chemical and bioinformatics methods. Classical analytic chemical methods such as collision-induced dissociation MS and nuclear-magnetic resonance (NMR) spectroscopy are powerful tools for the structure elucidation of most pyoverdines. However, in the case of W15Oct28pyoverdine, results of CID MS tandem fragmentation could only provide part of the information. By combining whole genome sequencing and bioinformatics analysis, we finally could supplement the information of amino acid composition and amino acid sequence to the results of CID MS. The pyoverdine produced by P. putida W15Oct28 contains 2 threonines and 2 homoserines in the 12 amino acid peptide chain, making it one of the largest pyoverdine molecule described so far. The presence of homoserine in pyoverdine was only reported in azotobactin D (Demange et al. 1988), and its molar mass being identical to threonine (both are 119.12 g/mol) made it difficult to interpret the amino acid sequence from tandem fragments. With the fast development of next generation sequencing and the dramatically increasing bacterial whole genome sequences available in public database, bioinformatic methods to predict the structure of secondary metabolites represent a convenient and reliable complement to analytical methods. The basic algorithm to predict the structure of non-ribosomal peptide is to compare the signature residues in the binding pocket (Stachelhaus code) of the query adenylation domain to the A domains with known structure. The signature residues decide which substrate will be recognized and activated, thus the comparison will look for the statistic evidence for the highest identity. The disadvantage of this method is that wrong prediction or no prediction at all will be made when there is no template to compare to in the database. This was the case when using antiSMASH to analyze the pyoverdine NRPS in the genome of A. vinelandii DJ where the third, eighth, and tenth amino acid could not be annotated although they were experimentally proved to be homoserines. Another example is the seventh amino acid, which is predicted with reasonable certainty to be an arginine based on Stachelhaus code while the amino acid was proven to be a citrulline which was only reported in azotobactin D. Due to this fact, we also could extend the Stachelhaus codes of abnormal amino acids such as diaminobutyric acid (Dab), histidine, ornithine and its different modified derivates, homoserine, arginine and citrulline, which could not be differentiated by a bioinformatics method so far (Fig. 5). Like mentioned in the beginning of this discussion, the structure of pyoverdines is even more diverse within the species P. putida. By collecting the pyoverdines with correct structures reported experimentally and the pyoverdines with bioinformatic predicted backbones from available biosynthesis gene clusters sequences, we found out that the pyoverdines produced by P. putida strains could be classified into four major groups. This result supports the usefulness of pyoverdines as phylogenetic markers for the P. putida species representatives, with the potential to be used subspecies definition. The species of P. putida representing a large complex, the upcoming genomic information of P. putida strains will not only provide more knowledge about this species, but also will make bioinformatic prediction of pyoverdine structures as an alternative method for siderophore typing together with IEF gel fingerprinting, siderophore utilization and MS.
Fig. 5

Cladogram of N-J tree inferred from adenylation domains known to activate unusual amino acids. In the label of each branch, the number means the amino acid position in the peptide chain, the abbreviation stands for the amino acid followed by the strain name of the pyoverdine producing strain or the non-ribosomal peptide product. The curve covers the cluster of adenylation domains with similar or the same signature residues in the docking domain and which activate the same amino acid substrate


This study shows the power of the combination of analytical techniques, bioinformatic analysis and literature survey to determine the structure of a novel pyoverdine with interesting features. It contains a dihydropyoverdine type of chromophore and has an unusually long peptide chain containing two residues of homoserine, an amino acid rarely found in pyoverdines. This study also confirms the diversity of pyoverdines structures within the species P. putida.

Supplementary material

10534_2013_9653_MOESM1_ESM.ppt (1.2 mb)
Supplementary Figure 1IEF gel image of pyoverdines produced by P. putida W15Oct28 and P. putida L1. Pyoverdine of P. putida W15Oct28 (left lane) shows white fluorescence under UV light due to the un-matured dihydropyoverdine chromophore, while the pyoverdine L1 shows blue fluorescence and two isoforms. (PPT 1276 kb)
10534_2013_9653_MOESM2_ESM.doc (1.1 mb)
Supplementary Figure 2Amino acid analysis of fully hydrolyzed N-TFA-n-butylester derivates by GC/MS. (DOC 1155 kb)
10534_2013_9653_MOESM3_ESM.ppt (300 kb)
Supplementary Fig. 3Three dimensional structure model of P. putida W15Oct28 pyoverdine–Fe3+ complex based on the hexadentate character. Green: carbons, Red: oxygens, Blue: nitrogen, Brown: Iron ion, non-polar hydrogens not shown. (PPT 302 kb)

Copyright information

© Springer Science+Business Media New York 2013