Pyoverdine biosynthesis and secretion in Pseudomonas aeruginosa: implications for metal homeostasis

Authors


For correspondence. E-mail isabelle.schalk@unistra.fr; Tel. (+33) 3 68 85 47 19; Fax (+33) 3 68 85 48 29.

Summary

Pyoverdines are siderophores produced by fluorescent Pseudomonads to acquire iron. At least 60 different pyoverdines produced by diverse strains have been chemically characterized. They all consist of a dihydroquinoline-type chromophore linked to a peptide. These peptides are of various lengths and the sequences are strain specific. Pyoverdine biosynthesis in Pseudomonas aeruginosa and fluorescent Pseudomonads is a complex process involving at least 12 different proteins, starting in the cytoplasm and ending in the periplasm. The cellular localization of pyoverdine precursors was recently shown to be consistent with their biosynthetic enzymes. In the cytoplasm, pyoverdine appears to be assembled at the inner membrane and particularly at the old cell pole of the bacterium. Mature pyoverdine is uniformly distributed throughout the periplasm, like the periplasmic enzyme PvdQ. Secretion of pyoverdine involves a recently identified ATP-dependent efflux pump, PvdRT-OpmQ. This efflux system does not only secrete newly synthesized pyoverdine but also pyoverdine that already transported iron into the bacterial periplasm and any pyoverdine–metal complex other than ferri-pyoverdine present in the periplasm. This review considers how these new insights into pyoverdine biosynthesis and secretion contribute to our understanding of the role of pyoverdine in iron and metal homeostasis in fluorescent Pseudomonads.

Introduction

Pyoverdines were discovered 120 years ago, in 1892, and have been given various different names: fluorescins, pseudobactins and finally pyoverdins or pyoverdines. In 1952, J. Totter and F. Moseley observed that the iron concentration affected the production of fluorescin by Pseudomonas aeruginosa (Totter and Moseley, 1952). The role of pyoverdines in iron acquisition by fluorescent Pseudomonads was established at the end of the 1970s by J-M Meyer and M. Abdallah (Meyer and Abdallah, 1978; Meyer and Hornspreger, 1978). The first chemical structures of pyoverdines were solved in the 1980s and early 1990s by various groups, but the major contribution was made by H. Budzikiewicz using mass spectrometry. Today, more than 60 pyoverdines from different strains and species of Pseudomonas have been chemically identified (Table 1) (Demange et al., 1990; Budzikiewicz, 1997; Fuchs and Budzikiewicz, 2001; Budzikiewicz, 2004; Budzikiewicz et al., 2007). All these siderophores are composed of three parts (Table 1): a dihydroquinoline-type chromophore responsible for their fluorescence, a strain-specific peptide comprising 6–12 amino acids; and a side-chain bound to the nitrogen atom at position C-3 of the chromophore. In most cases, this side-chain is a diacid of the Krebs cycle, such as succinic, malic or α-ketoglutaric acid or one of their amide derivatives. The peptide chain and the side-chain are connected to the carboxyl group and to the NH2 group of the chromophore respectively (Table 1). The sequence of the peptide moiety differs substantially between species and even between strains of the same species. This peptide chain may contain unusual amino acids, such as d-isomers, and amino acids which are not usually found in biomolecules; the peptide in some pyoverdines is cyclic. For example, three distinct pyoverdine types can be produced by P. aeruginosa – PVDI, PVDII and PVDIII – each characterized by a different peptide chain (Meyer et al., 1997) (Table 1). Pyoverdines can be easily distinguished from each other by isoelectric focusing (Meyer et al., 1997).

Table 1. Selected pyoverdines produced by fluorescent Pseudomonads

A.

image

B.

R : Succinate

 Succinamide

 Malate

 Malamide

 Glutamate

 α-Ketoglutarate

C.
StrainsSpeciesPeptide sequencesReferences
  1. A. Dihydroquinoline-type chromophore present in all pyoverdines.
  2. B. Possible side-chains bound to the NH2 group of the chromophore.
  3. C. Examples of peptide moieties of pyoverdines. The peptide is bound to the carboxyl group of the chromophore.
  4. OHAsp, threo-β-hydroxy aspartate; Dab, 2,4-diaminobutyrate; Orn, ornithine; OHOrn, Nδ-hydroxyornithine; AcOHOrn, N4-acetyl-N4-hydroxyornithine; fOHOrn, Nδ-formyl-Nδ-hydroxyornithine. [ ] are used to show cyclic peptides.
PAO1P. aeruginosaSer–Arg–Ser–FoOHOrn–[Lys–FoOHOrn–Thr–Thr]Briskot et al. (1989)
ATCC27853P. aeruginosaSer–FoOHOrn–Orn–Gly–Thr–Ser–cOHOrnTappe et al. (1993)
Pa6P. aeruginosaSer–Dab–FoOHOrn–Gln–Gln–FoOHOrn–GlyGipp et al. (1991)
G173P. fluorescensSer–Ala–AcOHOrn–[Orn–Asp–AcOHOrn–Ser]Uría-Fernández et al. (2003a)
DSM50106 P. fluorescensSer–Lys–Gly–FoOHOrn–Ser–Ser–Gly–[Orn–FoOHOrn–Ser]Greenwald et al. (2009)
PL 9P. fluorescensSer–AcOHOrn–Ala–Gly–[Ser–Ser–OHAsp–Thr]Uría-Fernández et al. (2003b)
ATCC13525P. fluorescensSer–Lys–Gly–FoOHOrn–[Lys–FoOHOrn–Ser]Hohlneicher et al. (1995)
Pfl 18.1P. fluorescensSer–Lys–Gly–FoOHOrn–Ser–Ser–Gly–[Lys–FoOHOrn–Ser]Amann et al. (2000)
Pfl 12P. fluorescensSer–Lys–Gly–FoOHOrn–Ser–Ser–Gly–[Lys–FoOHOrn–Glu–Ser]Geisen et al. (1992)
ATCC 39167P. putidaSer–AcOHOrn–Ala–Gly–[Ser–Ala–OHAsp–Thr]Uría-Fernández et al. (2003b)
G76Pseudomonas sp.Ser–Ser–FoOHOrn–Ser–Ser–[Lys–Ser–FoOHOrn]Meyer et al. (2008)
D-TR133P. chlororaphisAsp–FoOHOrn–Lys–[Thr–Ala–Ala–FoOHOrn–Ala]Barelmann et al. (2003)

This review mostly focuses on PAO1, the archetype PVDI produced by P. aeruginosa, and describes new insights into its biosynthesis and secretion. The pyoverdine pathway of P. aeruginosa strain PAO1, producing PVDI, is the best characterized but genetic evidence clearly indicates that the biosynthetic pathways of other pyoverdines in other strains and species are analogous.

Pyoverdine biosynthesis

The synthesis of pyoverdines begins in the cytoplasm and ends in the periplasm, from where they are secreted into the extracellular medium. Numerous different enzymes are involved, including in particular non-ribosomal peptide synthetases (NRPSs) (Hohlneicher et al., 2001; Visca et al., 2007). NRPSs are very large enzymes that have a multimodular architecture, with modules composed of about 1000 amino acid residues. Each module catalyses the incorporation of one specific amino acid substrate into the peptide product and the formation of peptide bonds between the amino acids. The peptide is always synthesized in a N-terminal to C-terminal direction through a carrier thiotemplate mechanism. Such enzymes are responsible for the synthesis of a very wide range of peptide-like secondary metabolites including siderophores and antibiotics (Crosa and Walsh, 2002; Finking and Marahiel, 2004). NRPSs synthesize both the chromophore backbone and the peptide moiety of pyoverdines in a multistep reaction. The diversity of the peptides sequences of the different pyoverdines is a consequence of the different substrate specificities of pyoverdine-synthesizing NRPSs in different strains and species (Ravel and Cornelis, 2003). Indeed, all fluorescent Pseudomonads have orthologues NRPS genes that synthesize the peptide chain of pyoverdines (Ravel and Cornelis, 2003; Gross and Loper, 2009). Additional enzymes provide the unusual amino acid substrates for the NRPSs and modify the peptide precursor to yield the mature pyoverdines.

Cytoplasmic assembly of the PVDI backbone in P. aeruginosa PAO1

In the case of P. aeruginosa PAO1, producing PVDI (Table 1), it has been suggested that the assembly of PVDI backbone starts in the cytoplasm with PvdL, the only common PVD-synthesis NRPS in the genomes of all fluorescent Pseudomonads analysed so far (Mossialos et al., 2002; Ravel and Cornelis, 2003; Gross and Loper, 2009). PvdL is atypical among PVD-synthesis NRPSs because it lacks an initial C-terminal domain. Instead, module 1 (M1, Fig. 1) of PvdL – predicted to function as a starter module – contains an unusual domain which is very similar to acyl coenzyme A ligases (Mossialos et al., 2002). The binding pocket of this PvdL-M1 module can bind myristate (Gulick and Drake, 2011) and analysis of the structure of a cytoplasmic PVDI precursor indicated the presence of a myristic or myristoleic acid group (Hannauer et al., 2012b), not present in the final PVDI molecule. Therefore, it was proposed that PvdL couples coenzyme A to a myristate fatty acid in an ATP-dependent reaction, and delivers the complex to l-Glu carried by a second module of PvdL (Gulick and Drake, 2011). Following this step, d-Tyr and l-Dab are incorporated by the two other modules of PvdL into the nascent pyoverdine backbone, forming a tetrahydropyrimidine ring, which is the precursor of the dihydroxyquinoline chromophore (Mossialos et al., 2002) (Fig. 1). The C-terminal domain of module 1 of PvdL then catalyses the attachment of this precursor to a d-Ser residue, the first amino acid of the peptide moiety of PVDI. The peptide is further elongated by the linear progression of the PvdI, PvdJ and PvdD modules, and the peptide is then released by the activity of the PvdD thioester domain. Indeed, in P. aeruginosa PvdD is the only NRPS that ends with a thioesterase domain, arguing for a function in siderophore release at the end of the assembly step (Ackerley et al., 2003). After incorporation of all residues into the peptide backbone by NRPS, the cytoplasmic non-fluorescent precursor is probably transported across the inner membrane into the periplasm by PvdE, which is an ‘export’ ABC transporter essential for PVDI production (Yeterian et al., 2010b). Mutation of pvdE in P. aeruginosa PAO1 abolishes the production and secretion of PVDI (Yeterian et al., 2010b).

Figure 1.

PVDI biosynthesis in P. aeruginosa. PVDI backbone is assembled in the cytoplasm by four NRPSs: PvdL, PvdI, PvdJ and PvdD. These enzymes have a multimodular organization consisting of two to four modules (M1 to M4). The biosynthesis also involves three other cytoplasmic enzymes – PvdA, PvdF and PvdH – participating in the synthesis of atypical amino acids present in the peptide moiety of PVDI. The cytoplasmic PVDI precursor has an unformed chromophore and a myristic or myristoleic acid chain bound to the first residue of the peptide backbone (Glu). This fatty acid chain probably keeps the cytoplasmic PVDI precursors bound to the inner membrane. This precursor species is probably transported across the inner membrane by the export ABC transporter PvdE. In the periplasm, the fatty acid chain is removed by the periplasmic acylase PvdQ. Then, the chromophore is formed from the condensation of Tyr and Dab. This process probably involves the periplasmic enzymes PvdN, PvdO and PvdP. Non-fluorescent ferribactin and dihydropyoverdine are periplasmic PVDI precursors with a partially formed chromophore. According to Dorrestein et al. hydroxylation of ferribactin would give a cathechol derivative and its oxidation to the o-quinone followed by a conjugate addition would give dihydropyoverdine (the differences between ferribactin and dihydropyoverdine are shown in red on the figure) (Dorrestein et al., 2003). Newly synthesized PVDI is secreted by the ATP-dependent efflux pump PvdRT-OpmQ. Different isoforms of PVDI can be found in P. aeruginosa cultures, having as an acyl chain (R) either a succinate or its amide form, a malate, or its amide form, α-ketoglutarate, or a glutamate. OM, outer membrane; IM, inner membrane; ASA, aspartate β-semialdehyde; Dab, 2,4-diaminobutyrate; Orn, ornithine, OHOrn, Nδ-hydroxyornithine; fOHOrn, Nδ-formyl-Nδ-hydroxyornithine; R, acyl chain.

Pyoverdines contain unusual amino acids, for example diamino butyric acid and ornithines in the case of PVDI. Therefore, pyoverdine biosynthesis also involves a variety of other cytoplasmic enzymes providing the NRPSs with these atypical residues (Visca et al., 2007). For PVDI synthesis in P. aeruginosa PAO1, PvdH catalyses an aminotransferase reaction interconverting l-aspartate β-semialdehyde (l-ASA) and l-2,4-diaminobutyrate (l-Dab) (Vandenende et al., 2004). l-Dab is predicted to be one of the substrates for PvdL, the NRPS required for synthesis of the chromophore moiety of PVDI precursor. PvdA, an l-ornithine-N5-oxygenase, is an essential enzyme for PVDI biosynthesis, necessary for conversion of l-ornithine (l-Orn) to l-N5-hydroxyornithine (l-OHOrn) (Ge and Seah, 2006; Meneely et al., 2009). Downstream, PvdF catalyses formylation of l-OH-Orn to produce l-N5-hydroxyornithine (l-fOHOrn), which is then incorporated into the peptide moiety of PVDI by the NRPSs PvdI and PvdJ (McMorran et al., 2001). l-fOHOrn is unstable at neutral pH (Akers and Neilands, 1973), so cytoplasmic OHOrn must be promptly formylated and sequestrated into the nascent pyoverdine backbone by the pyoverdine biosynthesis machinery, suggesting a close interaction between these enzymes and the NRPSs. Neither pvdA nor pvdF mutants produce any PVDI or analogue suggesting that modifications of l-ASA and l-Orn occur before the amino acid incorporation into the peptide chain (Visca et al., 1994; McMorran et al., 2001; Guillon et al., 2012).

The cytoplasmic precursor synthesized by the different NRPS in P. aeruginosa PAO1 has been isolated and the structure determined (Hannauer et al., 2012b). This compound, when compared with the secreted PVDI has an unformed chromophore and a myristate or myristoleate chain bound to the first residue at the N-terminal (Glu), which is subsequently removed in the periplasm by an acylase (Hannauer et al., 2012b). The presence of a myristic or myristoleic acid group probably retains the PVDI precursor at the membrane during its assembly by the NRPSs, preventing both diffusion of the siderophore throughout the cytoplasm and, consequently, chelation of metals present in this cell compartment. Consistent with this idea, fractionation of P. aeruginosa cells producing fluorescent eYFP-labelled PvdA, the ornithine Nδ-oxygenase catalysing l-ornithine hydroxylation (Fig. 1), showed accumulation of the eYFP labels in the membrane fractions (Guillon et al., 2012). Cell fractionation and proteinase K accessibility experiments with P. aeruginosa confirmed the membrane-bound nature of PvdA, but excluded a transmembrane topology for its N-terminal hydrophobic region (Imperi et al., 2008). The N-terminal hydrophobic domain probably interacts with the lipid bilayer by forming a U-shaped or re-entrant loop aided by contiguous G-P residues without fully crossing the membrane. The bulk of the protein containing the binding site is presumably on the cytoplasmic side and, as result, accessible to the ornithine substrate. Epifluorescence imaging revealed that a PvdA–YFP fusion protein was not randomly distributed along the inner membrane but was clustered as a spot at the old cell pole of the bacterium, with a subset remaining in the cytoplasm (Guillon et al., 2012) (Fig. 2A). This clustering of PvdA seems to be a dynamic process: the number of spotted P. aeruginosa cells increases during growth in iron-limited media, whereas after addition of iron to the culture, the proportion of bacteria with spots decreased (Guillon et al., 2012). The PvdA patches may also contain other enzymes required for PVDI biosynthesis in the cytoplasm, such as NRPSs, PvdF and PvdH. This spotted localization of PvdA may reflect the need of bacteria to compartmentalize PVDI synthesis in a crowded environment such as the cytoplasm. As a consequence, PVDI precursors would not diffuse throughout the cytoplasm, limiting the risk of disrupting metal homeostasis elsewhere in this cell compartment.

Figure 2.

Fluorescent micrographs of P. aeruginosa cells showing C-terminally labelled PvdA (A) (Guillon et al., 2012), C-terminally labelled PvdQ mCherry (B) (Guillon et al., 2012) and periplasmic PVDI (C) (Yeterian et al., 2010b). Left images correspond to bright field and right images to epifluorescence of eYFP, mCHERRY and pyoverdine fluorophores for (A), (B) and (C) respectively (scale bars 2 μm).

Periplasmic maturation of PVDI in P. aeruginosa PAO1

The myristic or myristoleic chain is removed in the periplasm by PvdQ (Yeterian et al., 2010b; Gulick and Drake, 2011; Schalk et al., 2011), a member of the NTN hydrolase family (Bokhove et al., 2010). This enzyme was first identified as a periplasmic quorum-quenching protein that cleaves acyl homoserine lactones (Sio et al., 2006). The periplasmic PvdQ acylase recognizes tetradecanoic acid (Gulick and Drake, 2011). When PvdQ crystals were incubated with the cytoplasmic PVDI precursor possessing a myritic side-chain, the fatty acid chain was removed from the siderophore and found in the enzyme's binding pocket (Gulick and Drake, 2011). These elements confirm that this cytoplasmic siderophore precursor is the PvdQ substrate. A pvdQ mutant produces only the non-fluorescent cytoplasmic PVDI precursor with a myristic or myristolic chain (Hannauer et al., 2012b) (Fig. 1), indicating that fatty acid removal by PvdQ precedes chromophore cyclization. Unlike PvdA that is bound to the membranes and concentrated at the old cell pole, PvdQ is uniformly distributed throughout the periplasm (Guillon et al., 2012) (Fig. 2B). This implies that, after removal of this fatty acid chain, the PVDI precursor diffuses across the entire periplasm.

The PVDI chromophore is derived from Tyr and l-Dab (Fig. 1) (Dorrestein et al., 2003). Pyoverdines are fluorescent under ultraviolet (UV) light, a characteristic conferred by the quinolinic chromophore. Therefore, cyclization of the chromophore can be monitored in vivo using these spectral properties (Yeterian et al., 2010b). Ferribactin is a non-fluorescent PVDI precursor with no myristic chain but in which, instead of the dihydroxyquinoline chromophoric group, there is still the tripeptide γ-l-Glu–d-Tyr–l-Dab (Hohlneicher et al., 2001). This compound is probably the product of PvdQ, and the last PVDI precursor before chromophore cyclization (Fig. 1). The PVDI chromophore appears to be cyclized in the periplasm (Yeterian et al., 2010b) by a multistep oxidative process followed by tautomerization leading to dihydroxyquinoline ring formation (Mockmann et al., 1997; Dorrestein et al., 2003). Consistent with this oxidative cyclization, dihydropyoverdine can be isolated from cultures of P. aeruginosa. This precursor is identical to PVDI except that there is one less unsaturation within the future chromophore moiety. Consequently, this form of PVDI is not fluorescent and does not exhibit the absorbance typical of PVDI at 400 nm (Budzikiewicz, 2004). In turn, it must be a PVDI precursor following ferribactin formation. Dihydropyoverdine can be non-enzymatically converted to PVDI by oxidation at high pH (Teintze and Leong, 1981; Bultreys et al., 2001) and possesses a lower affinity than PVDI for iron (Teintze and Leong, 1981). The enzymes involved in the chromophore cyclization are yet to be identified. Mutation of the periplasmic enzymes PvdN (Voulhoux et al., 2006), PvdO and PvdP abolish production of fluorescent PVDI (Yeterian et al., 2010b), indicating that these proteins are involved either in chromophore formation or in a step preceding it. The exact roles of PvdN, M, O and P are still unknown. The sequence of PvdM is most similar to those of mammalian dipeptidases; PvdN is most similar to an isopenicillin N epimerase from Streptomyces clavuligerus (Kovacevic et al., 1990) and PvdO has been suggested to be a reductase (Ochsner et al., 2002). The periplasmic localization of PvdN has been demonstrated experimentally (Voulhoux et al., 2006) and the presence of a signal sequence on each of the other proteins strongly suggests that they are also localized in this cell compartment.

PVDI isoforms

Pseudomonas aeruginosa and other fluorescent pseudomonads produce multiple pyoverdine isoforms that have been identified by mass spectrometry or can be separated by chromatofocusing (Kils et al., 1999; Fuchs and Budzikiewicz, 2001). These isoforms have the same peptide moiety, but differ in the modification of the l-Glu attached to the initial d-Tyr, precursor of the chromophore (Fig. 1) (Koedam et al., 1994; Meyer et al., 1997). This l-Glu can be transformed to α-ketoglutaric acid, succinamide or malamide, or be hydroxylated to free acid (Schafer et al., 1991). These modifications follow the citric acid cycle. The discovery of different isoforms of ferribactin produced by Pseudomonas fluorescens ATCC 17400 suggests that this modification process may even occur before chromophore cyclization (Schäfer et al., 2006). However, it is unclear if this occurs in the extracellular medium or during the storage in the periplasm (Baysse et al., 2002), and whether it involves enzymes.

Pyoverdine biosynthesis in Pseudomonads

Analyses of Pseudomonas genomes suggest that there are analogous biosynthetic pathways for other pyoverdines in other strains and species (Ravel and Cornelis, 2003; Visca et al., 2007). PvdL, the first NRPS involved in the assembly of the PVDI backbone is highly conserved among all fluorescent pseudomonads suggesting that the first steps of siderophore assembly – attachment of a fatty acid to a Glu, followed by the anchoring of l-Tyr and l-Dab forming the chromophore – are common to all pyoverdines and even all species. This conservation of PvdL among fluorescent Pseudomonads is consistent with the presence of the dihydroquinoline-type chromophore in all pyoverdines. The NRPS genes pvdI, pvdJ and pvdD and the export ABC transporter pvdE are, however, highly divergent explaining the diversity of the peptide moiety sequence among pyoverdines (Smith et al., 2005).

The genes involved in PVD biosynthesis are often separated into different clusters in fluorescent Pseudomonads, except in Pseudomonas syringae, where they form one large cluster (Ravel and Cornelis, 2003). The distribution of biosynthetic machinery genes to multiple loci could confer an evolutionary advantage. The conserved chromophore biosynthetic gene (PvdL) is in each case located at a separate locus, and consequently able to interact with any of the more variable NRPS responsible for the biosynthesis of the peptide chain (Ravel and Cornelis, 2003). The term ‘genomic cassette’ has been proposed for these variable genomic regions (Spencer et al., 2003) and could explain the genetic diversity of the Pseudomonad pyoverdines. In addition, genes coding for the outer membrane ferri-pyoverdine transporter FpvA and for the pyoverdine export ABC transporter PvdE (Fig. 3) map to the same locus as the peptide chain biosynthetic genes (Ravel and Cornelis, 2003; Spencer et al., 2003). The transporters FpvA and PvdE import and export, respectively, pyoverdines with a different peptide moieties in different fluorescent Pseudomonad strains. The crystallographic structure of FpvA clearly indicates a specific interaction between the transporter and the peptide moiety of PVDI (Greenwald et al., 2009).

Figure 3.

Secretion of PVDI by PvdRT-OpmQ. The ATP-dependent efflux pump PvdRT-OpmQ is involved in the secretion of newly synthesized PVDI. Biosynthesis of this siderophore involves assembly of the PVDI backbone in the bacterial cytoplasm and maturation in the periplasm. PVDI that has already transported and released iron ions into the bacterial periplasm is also recycled into the extracellular medium by PvdRT-OpmQ. The chelator picking up iron in the periplasm and transporting it across the inner membrane into the cytoplasm has not been identified yet. Any unwanted PVDI–metal complexes imported into the periplasm by the outer membrane transporter FpvA are expelled as a PVDI–metal complex to the extracellular medium by PvdRT-OpmQ without any metal release from the siderophore. For all these different PVDI forms, PvdRT-OpmQ collects the siderophore from the periplasm and transports it across the outer membrane to the extracellular medium. OpmQ is predicted to be a porin-like β-barrel located in the outer membrane with a large periplasmic extension, PvdT an ATP-dependent inner membrane protein, and PvdR a periplasmic adaptor protein.

Pyoverdine diversity is a defence against ferrisiderophore stealing. Siderophore secretion is necessary for the acquisition of extracellular iron, but extracellular diffusion makes the siderophore available to any cell with an appropriate receptor, and invites cheating (Griffin et al., 2004). In iron-restricted environments, it is clearly beneficial for a strain to make a siderophore with a structure differing from what other strains can import. This evolutionary dynamic must lead to the continual generation of new siderophores allowing the bacterium to compete effectively with other organisms for iron.

Storage of PVDI in P. aeruginosa periplasm

Observations of P. aeruginosa cells by fluorescence microscopy and cell fractionation studies indicate that the periplasm abounds with fluorescent PVDI, suggesting that newly synthesized siderophore is stored in this cell compartment (Yeterian et al., 2010b). The fluorescent properties of PVDI are conferred by the chromophore confirming that newly synthesized PVDI with a formed chromophore and ready to be secreted is stored, and uniformly distributed, in the periplasm (Fig. 2C). PVDI biosynthesis is a complex process, so such storage facilitates rapid and large-scale secretion of the siderophore.

PVDI secretion

The pyoverdine locus on the P. aeruginosa PAO1 chromosome contains three genes pvdRT-opmQ (PA2389–91) coding for an efflux pump, the object of little attention until recently, when it was shown to be involved in PVDI secretion. These genes are adjacent to genes required for PVDI synthesis and transport (http://www.pseudomonas.com). This efflux pump is predicted to have the same organization as other bacterial efflux pumps (Li and Nikaido, 2009; Nikaido and Takatsuka, 2009) with PvdT being an inner membrane protein, PvdR a periplasmic adaptor protein and OpmQ an outer membrane protein with a β-barrel domain inserted in the outer membrane and a large periplasmic extension. PvdT (PA2390) shares 43% sequence identity (63% similarity) with MacB and, like MacB, is predicted to have four membrane-spanning helices (Lewenza et al., 2005). In MacB, the first transmembrane helix separates a predicted ATP-binding cytoplasmic domain from a large (∼ 200 residues) periplasmic domain (Kobayashi et al., 2003; Xu et al., 2009). The cytoplasmic ATP-binding domain contains the Walker A (GxxGxGKST, residues 48–56) and B motifs (IILADE, residues 172–177); it also contains the linker peptide (LSGGQQQRVS, residues 152–162) and the D (GALD, residues 180–183) and Q loops (FIFQ, residues 97–100), which are part the nucleotide-binding domain (NBD) characteristic of ATP-binding cassette transporters (Davidson and Chen, 2004). PvdR (PA2389) is predicted to be periplasmic, like other adaptor (MFP) proteins in efflux systems, and shares 34% sequence identity with MacA. OpmQ is predicted to comprise a porin-like β-barrel located in the outer membrane with a large periplasmic extension and has 49% sequence similarity to OprM, a well-characterized efflux system outer membrane protein (Akama et al., 2004).

Secretion (recycling) of PVDI that has already transported and released iron into the bacteria

Surprisingly, PVDI secretion was first investigated using P. aeruginosa cells unable to produce PVDI. Following incubation in the presence of PVDI–Fe, the fluorescence in the periplasm of these cells increased (Greenwald et al., 2007), and subsequently fluorescent apo PVDI was excreted (Schalk et al., 2002; Greenwald et al., 2007). This efflux corresponds to recycling into the extracellular medium of apo siderophore that has released the iron ions inside the bacteria (Schalk et al., 2002; Greenwald et al., 2007). In a pvdRT-OpmQ mutant, this secretion of apo PVDI is severely impaired (Imperi et al., 2009; Yeterian et al., 2010a), suggesting that the route used by the apo siderophore to cross the outer membrane to the extracellular medium relies on the efflux system PvdRT-OpmQ (Fig. 3). This inhibition of fluorescent apo PVDI secretion is associated with periplasmic accumulation of apo PVDI as observed by fluorescent microscopy and cell fractionation (Imperi et al., 2009; Yeterian et al., 2010a).

Actually in P. aeruginosa cells, extracellular PVDI–Fe is transported across the outer membrane by two specific outer membrane transporters FpvA [the major transporter (Poole et al., 1993b; Schalk et al., 1999; Wirth et al., 2007)] and FpvB [a second transporter (Ghysels et al., 2004)]. This uptake is dependent from the proton motive force of the inner membrane, which is used to activate FpvA and FpvB via an inner membrane complex called the TonB machinery (Postle and Larsen, 2007). This step of PVDI–Fe translocation across the outer membrane has been the subject of extensive investigations the last 10 years as summarized in these reviews (Schalk, 2008; Schalk et al., 2009; 2012). Iron is released from the siderophore PVDI in the periplasm (Schalk et al., 2002; Greenwald et al., 2007) and not in the cytoplasm as for the ferrichrome and enterobactin pathways in Escherichia coli (Braun, 2003). The mechanism involves no degradation or chemical modification of PVDI, but apparently does involve iron reduction as shown with uptake assays using PVDI–Ga complexes (Greenwald et al., 2007). PVDI–Ga binds to the PVDI–Fe outer membrane transporter FpvA with an affinity close to that of PVDI–Fe (Ki of 16 nM; the Ki for Fe3+ is 6 nM under the same conditions) and is transported into P. aeruginosa with the same efficiency as its corresponding ferric complex (Folschweiller et al., 2002). However, uptake assays with PVDI–Ga demonstrate an accumulation of this complex in the periplasm of wild-type P. aeruginosa (Greenwald et al., 2007; Yeterian et al., 2010a) [PVD-Ga being fluorescent (Folschweiller et al., 2002)]. The PVDI–Ga complex does not dissociate, most likely because gallium only exists in its Ga3+ oxidation state and cannot be reduced. Therefore PVDI–Ga can be considered as an inhibitor of any PVDI–Fe dissociation mechanism involving iron reduction. This periplasmic PVDI–Ga accumulation in PVDI-deficient P. aeruginosa cells strongly suggests that iron disociates from PVDI in the periplasm. This is consistent with the observed periplasmic accumulation of apo PVDI in pvdRT-opmQ mutants incubated in the presence of PVDI–Fe (Imperi et al., 2009; Yeterian et al., 2010a). There is thus diverse evidence that iron is released from PVDI in the periplasm by a reduction mechanism, and that the resulting apo siderophore is recycled to the extracellular medium by PvdRT-OpmQ; thus, PVDI starts a new iron uptake cycle (Schalk et al., 2002).

Secretion of newly synthesized PVDI

Elucidating the secretion of newly synthesized PVDI was more complex than investigating PVDI recycling. The first studies carried out on wild-type cells and pvdRT-opmQ mutants led to the conclusion that PvdRT-OpmQ was not involved in secretion of newly synthesized PVDI (Imperi et al., 2009; Yeterian et al., 2010a). Further experiments with strains that are unable to incorporate PVDI–Fe complexes (FpvA and FpvB mutants), led to the opposite conclusion: PvdRT-OpmQ is the efflux system involved in secretion of newly synthesized PVDI (Hannauer et al., 2010) (Fig. 3). PVDI–Fe is not incorporated into these fpvA and fpvB mutants, such that no PVDI can be recycled. Consequently, any apo PVDI present in these cells is newly synthesized PVDI. pvdRTopmQfpvAfpvB mutant produced 30% less PVDI in the extracellular medium than the same cells expressing PvdRT-OpmQ (Hannauer et al., 2010). In parallel, cellular fractionation and epifluorescence microscopy analyses showed that fluorescence, corresponding to newly synthesized PVDI, accumulated in the periplasm of cells lacking PvdRT-OpmQ (Hannauer et al., 2010). The ability of bacteria lacking PvdRT-OpmQ to secrete PVDI, though less efficiently than wild-type cells, indicates the presence of at least one other unidentified system that can also secrete PVDI. Pseudomonas aeruginosa contains a large number of efflux systems (Stover et al., 2000) and probably another one takes over PVDI secretion in the absence of PvdRT-OpmQ. It has been suggested that MexAB-OprM also contributes to PVDI secretion (Poole et al., 1993a) although recent research indicates that this is unlikely to be the case (Imperi et al., 2009).

Secretion (efflux) of PVDI–metal complexes

Pseudomonas aeruginosa PvdRT-OpmQ is also involved in efflux of PVDI–metal complexes, other than PVDI–Fe, present in the bacterial periplasm (Hannauer et al., 2012a) (Fig. 3). How do these PVDI–metal complexes appear in the periplasm? Siderophores efficiently chelate many metals other than iron (Schalk et al., 2011) and this was mostly demonstrated for PVDI, which chelates Ag+, Al3+, Cd2+, Co2+, Cu2+, Fe3+, Ga3+, Hg2+, Mn2+, Ni2+ and Zn2+ (Braud et al., 2009). Since all pyoverdines produced by Pseudomonads have similar structures, this property is very probably valid for all chelators of this family. However, its strongest affinity is for iron ions, and this has been demonstrated by competition experiments between iron ions and other metal ions for binding to PVDI (Braud et al., 2009). Competition experiments with PVD-55Fe showed that the PVD–metal complexes displace PVD-Fe from FpvA, the outer membrane transporter responsible for the import of extracellular PVDI–Fe complexes, with inhibition constants of between 2.9 nM for PVD-Fe and 13 μM for PVD-Al (Hannauer et al., 2012a). The accumulation of both metals and fluorescence corresponding to PVD in cells expressing FpvA but not in an FpvA-deleted mutant confirmed that this transporter is able to transport PVD–metal complexes other than PVD-Fe into P. aeruginosa (Hannauer et al., 2012a). Thus, FpvA displays little metal specificity for binding and similarly for metal uptake. These observations explain how PVDI–metal complexes can be imported into the bacterial periplasm. However, discordant findings concerning the ability of FpvA to transport PVDI–metal complexes other than PVDI–Fe have been published (Braud et al., 2009; Hannauer et al., 2012a), but it is important to note that the amounts of metals found to be accumulated via the FpvA/PVD pathway depend strongly on the way the uptake assays are carried out. Current knowledge of the molecular mechanisms involved in iron acquisition by the FpvA/PVD pathways indicates that the uptake of PVD–metal complexes depends on (i) the affinities of PVD for the metal, (ii) the affinities of FpvA for the PVD–metal complexes, and (iii) the efficacy with which FpvA transports these PVD–metal complexes after their binding to the transporter binding site (Braud et al., 2009; Hannauer et al., 2012a). Consequently the presence, or absence, of trace metals in the incubation buffer used for the uptake assay and the way in which PVDI is prepared can substantially affect the results obtained. When cells were incubated with the PVD–metal complexes in succinate medium, almost no metal accumulation was observed except for Mg2+ and Al3+ (Braud et al., 2009). Succinate medium contains 127 μM Mg2+ and 1.96 μM Al3+, which presumably compete with the other metals tested for the formation of PVD–metal complexes. Tris-HCl buffer contains less trace elements and therefore its use provides more reliable data about the ability of the FpvA/PVDI pathway to transport metals other than iron ion (Hannauer et al., 2012a). Moreover, it is important to use purified PVDI to form the PVDI–metal complexes and not to use growth supernatant of P. aeruginosa cultures without further purification (Braud et al., 2009).

Deletion of genes encoding the PvdRT-OpmQ efflux pump clearly increased metal ion and PVDI accumulation in P. aeruginosa, implicating this pump in the expulsion of PVDI–metal complexes into the extracellular medium (Hannauer et al., 2012a). PVDI fluorescence only accumulated in the periplasm and not in the cytoplasm (Hannauer et al., 2012a), indicating that the PVD–metal complexes were probably not transported further into the bacterial cells. Instead, they are directly exported back into the extracellular medium, from the periplasm, by PvdRT-OpmQ. Indeed, no data are currently available showing a dissociation of metals other than iron from PVDI in P. aeruginosa periplasm.

Implications for metal homeostasis of the cellular organization of PVDI biosynthesis

The presence of a fatty acid chain on cytoplasmic PVDI precursors results in this strong iron chelator being concentrated at the inner membrane and prevents its diffusion all throughout the cytoplasm. PvdA, one of the first enzymes involved in PVDI biosynthesis, is also located at the inner membrane and where it concentrates at the old cell pole, suggesting that PVDI assembly must occur in this area of the bacterium. This cellular organization avoids diffusion of siderophores through the bacterial cytoplasm and can be assumed to favour efficient biosynthesis. This organization leads to the question of the localization of the other cytoplasmic enzymes and whether there are siderosomes at the old cell pole of P. aeruginosa cells.

It is surprising that P. aeruginosa cells store PVDI in the periplasm. This siderophore is able to chelate a number of metal ions in addition to iron (Baysse et al., 2000; Braud et al., 2009; Schalk et al., 2011), and therefore potentially chelate any free metal present in this cell compartment, and possibly also metals bound to metalloproteins. The formation constants of PVDI complexes with Zn2+, Cu2+ and Mn2+ are between 1017 and 1022 M−1 (Chen et al., 1994), whereas that with Fe3+ is 1032 M−1 (Albrecht-Gary et al., 1994). It is therefore plausible that all these PVDI molecules stored in the periplasm are in some way inactivated, possibly by an as yet unidentified chaperone protein. Another possibility proposed by Waldron and Robinson is that: ‘metals are not in competition for a limited pool of chelators (proteins and siderophores), but rather the chelators compete for a limited pool of metals’ (Robinson, 2007; Waldron and Robinson, 2009). Thus, metal occupancy of metalloproteins would be a function of relative rather than absolute affinities of the different proteins for metals. In this model, any metal ion in the periplasm would be chelated by the PVDI present if no specific periplasmic protein, with higher affinity, is able to sequester it. Indeed, storage of newly synthesized PVDI all throughout the bacterial periplasm (Fig. 2) and the ability of PvdRT-OpmQ to excrete PVDI–metal complexes (Fig. 3) provide the bacteria with an efficient periplasmic barrier to stop the entry of unwanted metal ions into the bacterial cytoplasm. If any toxic metal diffuses via porins into the bacterial periplasm, the PVDI, which is an efficient metal chelator and present at high concentration, will sequester the metals and the complexes formed will be exported out of the cell by PvdRT-OpmQ. Consequently, metals other than iron ion have little chance to penetrate into P. aeruginosa cells further than the periplasm unless it has a high affinity for a specific periplasmic binding protein. Summing up, this high periplasmic concentration of stored PVDI may (i) avoid unwanted metal diffusion into the cytoplasm, (ii) regulate the intracellular concentration of all metals and (iii) avoid non-specific protein-metal interactions. Investigations on the ability of the periplasmic PVDI to chelate metals in the periplasm would be interesting and help elucidate the role of PVDI in Pseudomonad iron homeostasis, but also that of other metals.

In conclusion, the cellular organization of PVDI biosynthesis is such as to protect against disruption of iron or other metal homeostasis in P. aeruginosa. In other words, PVDI storage in the bacterial periplasm suggests that siderophores may even have a wider function than just feeding bacteria with iron: they may have as well a key role in iron and other metal homeostasis and bacterial metal tolerance.

Acknowledgements

This work was partly funded by the Centre National de la Recherche Scientifique, by grants from the Centre International de Recherche au Frontière de la Chimie (FRC), from the ANR (Agence Nationale de Recherche, ANR-08-BLAN-0309-02) and from the Association Vaincre la Mucoviscidose. L. Guillon had a fellowship from the Centre International de Recherche au Frontière de la Chimie (FRC).

Ancillary