The first step in the specific uptake of iron via siderophores in Gram-negative bacteria is the recognition and binding of a ferric siderophore by its cognate receptor. We investigated the molecular basis of this event through structural and biochemical approaches. FpvA, the pyoverdine–Fe transporter from Pseudomonas aeruginosa ATCC 15692 (PAO1 strain), is able to transport ferric–pyoverdines originating from other species, whereas most fluorescent pseudomonads are only able to use the one they produce among the more than 100 known different pyoverdines. We solved the structure of FpvA bound to non-cognate pyoverdines of high- or low-affinity and found a close correlation between receptor–ligand structure and the measured affinities. The structure of the first amino acid residues of the pyoverdine chain distinguished the high- and low-affinity binders while the C-terminal portion of the pyoverdines, often cyclic, does not appear to contribute extensively to the interaction between the siderophore and its transporter. The specificity of the ferric–pyoverdine binding site of FpvA is conferred by the structural elements common to all ferric–pyoverdines, i.e. the chromophore, iron, and its chelating groups.
Pseudomonas aeruginosa is an opportunistic pathogen, causing chronic lung disease in cystic fibrosis patients. This bacterium belongs to the large family of so-called fluorescent pseudomonads, because of their production of green fluorescent molecules named pyoverdines (PVDs). PVDs are siderophores, secreted by the bacteria in order to get access to iron. These molecules have in common a fluorescent chromophore derived from 2,3-diamino-6,7-dihydroxyquinoline attached via its carboxyl group to a peptide chain of 6–14 l- and d-amino acids (Meyer et al., 2008). The peptidic portion contains modified amino acids, particularly for the hydroxamate iron-chelating groups which are usually derived from a hydroxylated and formylated or acetylated lysine or ornithine residue. PVDs act as virulence factors during infection and they compete for iron with iron-containing proteins in the host. They are also involved in the regulation of other virulence factors such as exotoxin A and PrpL (Lamont et al., 2002; Buckling et al., 2007; Visca et al., 2007). They play a role in biofilm formation because the ferric uptake regulator (Fur) blocks biofilm formation until the bacteria acquire sufficient iron (Banin et al., 2005). Since the resistance of P. aeruginosa to antibiotics is enhanced by the formation of biofilms that protect the colonies from their environment, it is important to be able to interfere with the bacterial iron acquisition in the treatment of chronic infections.
During infection, PVD is synthesized and released in the organism where it competes for iron bound to transferrin (Xiao and Kisaalita, 1997). PVD chelates iron with a Ka of 1032 M−1, achieving this high affinity by using three bi-dentate groups: the catecholate of the chromophore and two hydroxamates from two modified amino acids on the peptide chain (Albrecht-Gary et al., 1994). Once loaded with iron, the ferric form of PVD is transported back into P. aeruginosa by a specific outer membrane receptor called FpvA. This transport is energy dependent and relies on direct interactions between FpvA and the inner membrane multisubunit TonB complex, which couples the proton motive force of the inner membrane to iron transport. Once in the periplasm, iron is released from the siderophore by a mechanism involving no degradation of the siderophore but probably a reduction of iron (Greenwald et al., 2007). Afterwards, PVD is recycled in the extracellular medium and iron is transported into the cytoplasm by an as yet unknown ABC transporter (Schalk et al., 2002).
The primary structures of nearly 70 different PVDs are presently known (Meyer et al., 2008). Within the P. aeruginosa species, strains can be subdivided into three types (siderovars) based on the primary structure of the PVD they produce. Following this classification, all PVDs can be placed into one of three structural groups: group I is characterized by a peptide with the four last residues forming an internal cycle; group II has a linear peptide with the last residue being an N-hydroxy(cyclo)Orn at the C-terminus; and group III is linear with an unmodified N-hydroxyOrn at the C-terminus (Fig. S1) (Budzikiewicz, 2001). Based on this diversity of structures and the usually high specificity between a given strain and its PVD, a method called siderotyping has been developed to differentiate the fluorescent Pseudomonas according to the type of PVD they produce (Fuchs et al., 2001; Meyer et al., 2002).
PAO1 (ATCC 15692) is the most extensively studied strain among P. aeruginosa and its genome has been fully sequenced (Stover et al., 2000). While a majority of fluorescent pseudomonads exclusively use their own PVD for iron uptake, PAO1 is not so restrictive and is also able to use other structurally different PVDs produced by Pseudomonas fluorescens ATCC 13525 or 18.1 (Meyer et al., 1999; Fuchs et al., 2001).
We have previously determined the crystallographic structures of FpvA from PAO1 bound or not to its natural ligand PVDI. Like other TonB-dependent receptors, FpvA consists of a 22-stranded beta barrel which is occluded by a plug domain (Cobessi et al., 2005). FpvA also has an N-terminal domain that is involved in PVDI signalling and which is very mobile with respect to the barrel domain (Brillet et al., 2007; Wirth et al., 2007). This structural information identified the residues of the binding site and also the conformational changes associated with the binding of PVDI. To further understand the molecular determinants of PVDs that govern their specificity and affinity for FpvA, we have investigated the uptake of iron in PAO1 using six PVDs, representative of the three structural groups, and solved the structures of FpvA loaded with the ferric forms of these PVDs. Our data show that the chromophore-iron-hydroxamate part of the different PVDs shares the same position in the FpvA binding site, while the amino acids at the C-terminus of the peptidic part of PVDs do not seem to be involved extensively in the recognition by FpvA. The difference of affinities with FpvA seems to be influenced by the structure of the first three to four amino acids of the PVDs.
In vivo binding to FpvA and iron release of the non-cognate ferric–PVDs
Six PVDs from different species (Table 1, lines 2–7) were compared with PVDI for their ability to displace iron-free PVDI bound to FpvA in vivo (Schalk et al., 2002; Greenwald et al., 2008). These same non-cognate ferric–PVDs were also assayed for iron release in vivo by fluorescence. Of these, the ferric complexes of PVDATCC27853 (group II) and PVDPa6 (group III) have been previously shown to be not used for iron uptake by the PAO1 strain, whereas of the remaining group I PVDs only PVDPfl18.1 and PVDATCC13525 have been shown to be utilized by FpvA (Meyer et al., 1999; Fuchs et al., 2001; Schons et al., 2005). To verify that FpvA, and not FpvB, was the receptor involved in the uptake of these heterologous PVDs, incorporation of 59Fe was quantified using the K691 strain, a PAO1 fpvA minus strain, and compared with the PAO1 strain. No significant incorporation of iron was detected with K691, for PVDI and the six non-cognate PVDs tested (data not shown).
The PVDs used for structural study or ligand binding are marked in bold characters (lines 1–7). The others (lines 8–19) are a collection of group I PVDs selected for their sequence similarity with the high-affinity PVDs as determined in this study.
d-amino acids are underlined. Cyclic peptides are shown with brackets [ ].
Determined by 59Fe incorporation in PAO1, expressed as percentages, incorporation of PVDI–59Fe by PAO1 representing 100%.
Determined by PVD–55Fe binding assay on PAD06 cells.
Reference of publication that first describes the corresponding PVD.
We have taken advantage of the intrinsic fluorescence of tryptophan and PVD to observe in vivo FRET, and direct fluorescence as a measure of ferric–PVD binding to FpvA, and subsequent dissociation of the iron–PVD complex. FRET measurements were performed as previously described, using K691(pPVR2), a PAO1 strain that overexpresses FpvA (Schalk et al., 2001). When grown in an iron-poor medium, these cells reach a state in which most of FpvA in the outer membrane is bound to iron-free PVDI. Originally reported to be apo-PVD, we have recently shown that the receptor purified from these cells is primarily bound to the fluorescent complex of aluminium–PVD. Thus, in vivo FRET is observed with an emission of PVDI-Al fluorescence at 447 nm while exciting the tryptophans of FpvA at 290 nm. Addition of PVDI–Fe diminishes the FRET signal by displacing the fluorescent PVDI-Al with the non-fluorescent PVDI–Fe. The signal is then recovered after PVDI–Fe transport, iron release and rebinding of PVDI-Al to FpvA (Schalk et al., 2001; Greenwald et al., 2007; 2008). Figure 1 shows the evolution of the FRET signal upon addition of the non-cognate ferric–PVDs to K691(pPVR2) cells. Addition of PVDG173–Fe, PVDATCC27853–Fe or PVDPa6–Fe does not diminish the FRET signal, whereas addition of PVDDSM50106–Fe, PVDPfl18.1–Fe and PVDATCC13525–Fe induces a drop in the intensity of the FRET signal. Although the signal from the PVDATCC13525–Fe complex is similar to what is observed with PVDI–Fe, this cannot be considered quantitative because the in vivo FRET intensity is affected by several factors that vary from sample to sample.
The in vivo release of iron from the different ferric–PVDs was monitored by observing the direct excitation fluorescence of the PVDs in a PVDI-deficient background. The lower panel of Fig. 1 is a time-course of fluorescence intensity after the addition of different ferric–PVDs to a PVDI-deficient strain that overexpresses FpvA [CDC5(pPVR2)]. An increase of fluorescence indicates that the PVD–iron complex has been transported, and then dissociated to yield the fluorescent iron-free PVD. As shown, the fluorescence intensity after the addition of PVDI–Fe, PVDDSM50106–Fe, PVDPfl18.1–Fe and PVDATCC13525–Fe increases and reaches a plateau after about 30 min. Consistent with the FRET data, PVDG173–Fe and PVDATCC27853–Fe are not transported. PVDPa6–Fe is slowly dissociated indicating that this ferric–PVD is transported but with much slower kinetics than for PVDI–Fe. This small degree of transport can only be observed with a sensitive technique like fluorescence and could be lost in background of a radioactive iron transport assay.
Some non-cognate PVDs have PVDI-like affinities
The affinities of the non-cognate ferric–PVDs for FpvA were measured by incubating 55Fe complexes with PAD06 cells at several concentrations and recording the bound radioactivity. The PAD06 strain is derived from PAO1 but is unable to synthesize PVDI (Takase et al., 2000). The results presented in Table 1 show that the ferric complexes of PVDDSM50106, PVDPfl18.1 and PVDATCC13525 have a Kd within one order of magnitude of that of PVDI–Fe for FpvA, whereas the others do not show significant binding. For the three other PVDs, no binding was observed with concentrations of PVD–55Fe of 80 μM, indicating that if binding occurs, it is with a Kd higher than 100 μM. These results are consistent with the fluorescence data and suggest that PVDG173–Fe, PVD27853–Fe and PVDPa6–Fe either have a very low affinity for FpvA or eventually do not bind, while PVD50106–Fe, PVDPfl18.1–Fe and PVD13525–Fe bind to FpvA, are transported across the outer membrane and dissociated in the periplasm like PVDI–Fe. These in vivo results mirror what we see in vitro for mixtures of pure FpvA with the different PVDs. In these experiments, the binding of PVD to FpvA is clearly observed as a ∼50% decrease in tryptophan fluorescence (data not shown). Thus, if FpvB is interfering, it is not having a significant effect on our measurements.
High- and low-affinity ferric–PVDs bind similarly to FpvA
The three-dimensional structures of FpvA bound to these ferric–PVDs were pursued in order to study the basis of the affinity and specificity of their interaction. We used the crystal form from the recently reported apo-FpvA structure to soak in the various ferric–PVDs (Brillet et al., 2007). For the ferric–PVDs that appeared not to bind in the previous assays we used a higher concentration in order to overcome their low affinity. Diffraction and eventually structures were obtained for all except PVDATCC27853–Fe whose soaking into the crystals damaged them beyond use for diffraction analysis.
The structures of the ferric–PVD complexes with FpvA are generally very similar to the apo-FpvA structure derived from the same crystal conditions, except for the residues of the PVD binding site, some of which undergo large conformation changes. As previously reported, there are two FpvA molecules in the asymmetric unit (chains A and B). In our soaking experiments, only one of the two (chain B) is bound to ferric–PVD (Fig. 2). We believe that this is due to constraints imposed by the crystal packing arrangement in which the extracellular loop comprising residues 650–660 of one FpvA molecule inserts into the binding pocket of the other and vice versa in a pseudo twofold symmetric manner. The slight difference in the two FpvA orientations provides sufficient space for only one PVD. As depicted in Fig. 2, a hypothetical PVD peptide chain transposed to the upper FpvA molecule (chain A) would clash with the FpvA loop from the lower molecule (chain B). Otherwise, the environments, and hence the structures, of the two FpvA molecules in the asymmetric unit are similar with the exception of the relative orientation of the N-terminal signal domain. In chain A, the signal domain is closer to the plug and the TonB box is sequestered into a beta strand with the plug domain. In chain B, the signal domain is much further from the plug and the connecting loop which includes the TonB box is disordered (not visible in the electron density). Based on the several structures of FpvA, the movement of the signal domain has been hypothesized to be critical for both its activity in upregulating FpvA and PVD expression as well as its acting as a lock on the TonB box, preventing TonB/FpvA interactions in the absence of ferric–PVD (Brillet et al., 2007). Therefore, another explanation for why only one of two FpvA molecules in the asymmetric unit can bind ferric–PVD is that the orientation of their signal domains dictates whether or not they are binding competent. However, this logic reverses the hypothesized order of events (PVD binding induces signal domain movement) and the simplest explanation is a matter of the steric constraints depicted in Fig. 2.
Interestingly, the high- and low-affinity PVDs were both able to bind to FpvA in the crystal, all of them leading to nearly identical conformational changes in FpvA. These observed conformational changes are likely to be the real effects of PVD–Fe binding because most artefacts can be excluded: we verified that the apo-FpvA structure is not changed by the soaking conditions and furthermore, the FpvA molecule of the same crystal whose binding site is sterically hindered undergoes no large structural changes upon PVD binding to its neighbour in the asymmetric unit. However, we cannot exclude that the crystal packing, including the dimer in the asymmetric unit, has an impact on how the ferric–PVDs bind to the FpvA chain B.
The structures of PVDI–Fe and the other ferric–PVDs bound to FpvA reveal a common mode of interaction with the receptor. The chromophore, the iron atom, and the hydroxamate groups that chelate iron are all buried into the binding pocket, while the C-terminal residues of the PVDs are more solvent exposed on the extracellular surface of the receptor. Although the structures are of medium to low resolution (Table 2) the binding of ferric–PVD in the crystal was easily located as the strongest peak in the Fo − Fc and 2Fo − Fc electron density (Fig. 3). Continuous density for each ferric–PVD was visible, although only at lower contour levels (0.5–1.0 σ). We did not obtain structural data on PVDATCC27853 due to poorly diffracting crystals. However it was likely bound to FpvA based on the typical dark brown colour change of crystals and the fact that, like the other PVDs, soaking with it severely reduced the diffraction quality. The fact that the low-affinity PVDs are binding in the same manner and affecting the same conformational changes in FpvA suggests that the specificity of the binding site is encoded in the common features of the PVDs, namely the chromophore, the iron and the chelating hydroxamates.
Table 2. Crystallographic data and refinement statistics.
Model (PDB ID)
Space group C121
aUnit cell a = 194 Åb = 130 Åc = 141 Å = 90° = 130° = 90°
2 FpvA molecules per asymmetric unit
The unit cell dimensions of apo-FpvA. The lengths of soaked crystals varied by up to 1.5% with a beta angle variation of less than 1 degree.
Rsym = ΣhΣi|Ii(h) − 〈I(h)〉|/ΣhΣiIi(h), where Ii(h) and 〈I(h)〉 are the ith and mean intensity over all symmetry-equivalent reflections h.
Rmeas = Σh(nh/nh − 1)1/2Σi|Ii(h) − 〈I(h)〉|/ΣhΣiIi(h), where Ii(h) and 〈I(h)〉 are the ith and mean intensity, and nh is the multiplicity over all symmetry-equivalent reflections h (Diederichs and Karplus, 1997)
R = Σ||FC| − |FO||/Σ|FO|, where |FC| is the calculated structure factor amplitude of the model, and |FO| is the observed structure factor amplitude; the Free R-factor was calculated against a random 5% test set of reflections that was not used during refinement. The same test set was used for all seven data sets.
RMSD, root-mean-square deviation from the parameter set for ideal stereochemistry (Engh and Huber, 1991).
Ferric–PVDs induce common conformational changes in FpvA
A comparison of the known FpvA structures reveals several conformational differences between the apo and PVD-bound states of the protein (Brillet et al., 2007; Wirth et al., 2007). However with these structures, the effects of the different crystal packing environments are difficult to estimate. Therefore only large conformational changes near the PVD binding site could be considered biologically relevant. In contrast to these previous structures, the apo and PVD-bound crystals in this study share the same crystal packing, having originated in identical conditions. The only difference among the crystals used in this study is the presence or absence of a ferric–PVD. The FpvA molecule that cannot bind to PVD in the crystal (chain A) acts as an internal control upon the introduction of ferric–PVDs. The magnitude of the differences among the chain A structures is the background, above which, differences can be attributed to the binding of the ferric–PVD. To better highlight these small changes, we mapped the vector sum of the main-chain displacements of the six PVD–Fe bound structures onto the C-alpha trace of the two unique FpvA chains (Fig. 4). The movements greater than the background level determined from the overlay of the chain A structures (∼0.3 Å) are colour coded from blue to red. In chain A, the barrel and plug have few significant differences among the structures and it is only the extracellular loops (which are directly affected by the binding of PVD in the pocket of chain B) and the signalling domain that move between the apo and ferric–PVD crystals. However an overlay of the chain A structures shows that the signalling domain movement is due to small non-isomorphisms in the crystals. Since only the barrel domains were used in the alignments for these calculations, small crystal to crystal variations in domain packing will show up as movements in the signal domain. In contrast to chain A, the chain B backbone atoms of both the barrel and plug domains undergo many more significant movements, highlighting a network of interacting residues that may energetically connect the PVD binding residues to those in the TonB box and signalling domain.
The largest conformational changes in FpvA occur for all of the PVDs tested in this study and are similar to what has been seen in the previous structures (Brillet et al., 2007; Wirth et al., 2007). These large changes occur mainly in the plug domain where the loop Thr-225–Gly-234 undergoes a large rearrangement, bringing the side-chains of Val-229 and Tyr-231 in contact with the chromophore and the first hydroxamate side-chain of the PVD respectively. The other large change is the Arg-204 side-chain which moves more than 8 Å, placing its guanido group near the bound iron (4 Å) and coming within hydrogen bonding distance of the catechol of the chromophore (2.5 Å). This arginine is at the centre of an extensive hydrogen-bonding network that is present only in the ferric–PVD-bound FpvA. This network includes the carbonyl from Val-229, stabilizing the new conformation of loop 225–234, as well as the Asp-201 side-chain. Asp-201 in turn forms a hydrogen bond to the amide of Ser-363 whose peptide bond with Trp-362 is flipped in all of the ferric–PVD structures. These large conformational changes near the binding site are somehow propagated towards the periplasmic side of the receptor, so as to signal the loaded status of the receptor to the TonB complex and/or the FpvR protein. However no significant conformational changes are detected in the periplasmic region upon binding of any of the ferric–PVDs.
Numerous small movements are consistently seen in all of the ferric-PVD structures, and there are likely even more that are too subtle to detect at this resolution or by crystallography in general. These movements are part of a conserved network of energetically coupled residues connecting the binding site to the persiplasmic side of the receptor. However, we observed two small structural differences between the high- and low-affinity PVDs that might account for their relative binding constants. In the L7 extracellular loop, Trp-599 and Tyr-600 move slightly upon binding of PVDI–Fe and the strong binders, while they remain relatively unchanged in the structure of FpvA bound to the weak binders PVDG173 and PVDPa6 (Fig. 5). The movement appears to be necessary to make space for the conformation of the first three residues of the PVD peptide chain. PVDG173 and PVDPa6 do not interact as extensively with Trp-599 and Tyr-600 and do not induce a movement in L7. A second subtle difference is that only the strong binders have an interaction via the carbonyl of their second residue with the side-chain of Asn-228. The distance is too long (3.6 Å) to be a strong hydrogen bond; however, the dipole–dipole interaction might contribute to the affinity difference. Aside from the first three residues of the high-affinity PVDs, there is little structural similarity between any of the PVD peptides. Furthermore, with the exception of the hydroxamate side-chains that are co-ordinating iron, the C-terminal portion of their peptide chains are more solvent exposed making fewer specific contacts with the receptor.
Structure of the bound PVDs correlates with FpvA affinity
A comparison of the structures of the different PVDs reveals a common scaffold around the iron. The iron constrains the geometry of the PVDs so that the chromophore and two hydroxamic acid moieties of the ferric–PVDs adopt the same conformation as can be seen in alignment of the six different FpvA-bound structures (Fig. 5). The sequences of the high-affinity binders (PVDI, PVDDSM50106, PVDPfl18.1 and PVDATCC13525) are similar for the first four amino acids after the chromophore with the first hydroxamate occurring in the fourth position. In addition to the iron core scaffold, only these four residues from the high-affinity PVDs have a common conformation (Fig. 5, box B). On the C-terminal side of the first formyl-hydroxy-ornithine (FoOHOrn) the backbone orientation is not conserved (Fig. 5, box C). For the low-affinity binders (PVDG173 and PVDPa6) the d-serine following the chromophore is the only conserved residue among the PVDs whose structures we solved and the conformations of their first few residues diverge as much as the rest of the peptide chain.
According to the observed structures, the FpvA-bound ferric–PVDs can be divided in three different regions: the iron, chromophore and hydroxamate groups that co-ordinate iron (this comprises the core of the PVD that is specifically recognized by the protein); the first three amino acids following the chromophore, the structure of which in the high-affinity PVDs allows unique contacts with FpvA (Trp-599 and Tyr-600 on loop L7 and Asn-228); the rest of the peptide chain, including the cyclic part of the group I PVDs, which is more exposed to the solvent.
It is thus tempting to establish rules that would govern the binding of a given PVD to FpvA according to the peptidic sequence. The structures suggest that a PVD with a peptidic sequence starting with D-Ser followed by Arg or Lys in the second position, a small residue in position 3, followed by the first hydroxamate in the fourth position would be a likely candidate for uptake via FpvA. To test this observation, a dozen more group I PVDs were tested in a cross-uptake assay with 59Fe-labelled ferric–PVDs and the relative uptake efficiency quantified for those that were positive in the cross-uptake test (Table 1). Except for PVDPfl12, the results are relatively in good agreement with our model: D-Ser in position 1 seems essential. The residue in position 2 is a basic amino acid (Arg, Lys or Orn), except for PVD96-312 (Ser). The presence of a small residue (Gly or Ser) on position 3 appears important but not essential, since PAO1 is able to use PVDLille25, PVD96-188, PVDD47 and PVD96-318, although with much lower efficiency. In general the shortening of the sequence between the chromophore and the first amino acid in the sequence involved in iron chelation has a strong negative effect.
The inability of PVDPfl12 to interact with FpvA is intriguing since it is only different in three positions compared with PVDPfl18.1 (insertion of Glu in position 10 and inversion of chirality at positions 5 and 6). Since these differences occur after the fourth position, there are clearly other critical factors than the identity of the first four residues. The Glu in PVDPfl12 suggests that this discrepancy could be the result of a charge repulsion effect. However, a more interesting possibility is that the sequence of the PVD peptide can determine its affinity for the receptor not simply by direct contacts with FpvA but indirectly by how it influences the handedness and thus the structure of its own ferric complex. It is clear from the structures of the six PVDs reported here that FpvA only recognizes one of the possible stereoisomers of the ferric–PVD complexes. A PVD that strongly favours the incorrect handedness for a receptor would have a lower affinity for that receptor.
Circular dichroism spectra of ferric PVDs reveal a variety of stereochemistry
We measured the circular dichroism (CD) spectra of the apo and ferric forms of the first seven PVDs from Table 1 as well as that of PVDPfl12. The apo-PVDs all had a positive cotton effect at 400 nm suggesting that the absolute configuration of C1 of the chromophore is (S) in all of the PVDs (Wendenbaum et al., 1983). The spectra were very different in the region from 240 to 300 nm suggesting that the apo-PVD peptides have distinct conformations near the chromophore (data not shown). In the case of the ferric–PVDs, the charge transfer band around 450 nm overlaps with the chromophore CD spectra in this region; however, the negative cotton effect around 470 nm present in all of the PVDs tested is indicative of an overall Δ configuration around the iron (Piper, 1961). Notably, it is only the Λ conformation that is bound to FpvA in the crystal structures. The solution structure of the PVDI–gallium complex has been shown to be a mixture of Λ and Δ conformers (Wasielewski et al., 2008). This suggests that even PVDI–Fe must isomerize before it can be fully bound to its own receptor. Thus the energy difference between the multiple stereoisomers existing in solution for a given ferric–PVD will have an effect on their Kd. According to our measurements, a subtle variation in sequence like that between PVDDSM50106 and PVDPfl18.1 (Orn versus Lys) can lead to striking differences in the CD spectra and hence the relative stability of the conformers. Thus, while we can observe the conformation of the FpvA-bound ferric–PVDs, we cannot predict from their sequences those PVDs that can achieve this conformation with a low enough energy in order to bind with high affinity. Interestingly, the CD spectra for PVDDSM50106 PVDATCC13525 and PVDPfl12 are similar when compared with the others that we measured. Here the relative stability of the 16 possible conformers (8 Λ and 8 Δ) may be the main factor determining which bind to FpvA with high affinity. Despite their similarities, PVDPfl12 still stands out from the others as having a much more negative cotton effect at 220 nm (black arrow, Fig. S2B), which is a feature common to the other three low-affinity PVDs (Fig. S2A). There are other more subtle differences in the chromophore and charge transfer band region of the CD spectra that are highlighted in the figure. The large differences in the CD spectra from 200 to 230 nm suggest that the peptide backbones of the ferric–PVDs adopt distinct conformations in solution.
The binding of ferric–PVD to FpvA is a crucial step for both iron uptake and the signals mediating the transcription of virulence factors. Among pseudomonads, there is a varying degree of cross-uptake of ferric–PVDs from other species. The fact that PAO1 can use PVDs synthesized by a different bacteria gives it a competitive advantage (Buckling et al., 2007). Likewise, it is disadvantageous for a bacterium to synthesize a PVD that is transported by a competing species. To gain some insight into how PAO1 is able to achieve its broad specificity for different PVDs, we elucidated the structure of FpvA bound to PVDs both that are used by PAO1 for iron uptake (PVDATCC13525, PVDPfl18.1 and PVDDSM50106) but also for some that are not used (PVDG173 and PVDPa6).
Iron uptake experiments assessed by FRET or using 55Fe on whole cells allowed us to identify quickly and with a high sensitivity which PVDs are transported by FpvA in PAO1 cells. For those PVDs that were transported in cross-uptake assays, the measured Kd was very close to the one determined for PVDI–Fe, while for the others no binding could be detected at up to 80 μM. Despite the large affinity differences, there was clearly binding in the soaked crystals of apo-FpvA when PVD was present as 2 mM ferric–PVD, indicating that the Kd of the non-binding siderophores should lie between 100 μM and 2 mM.
Thus it seems that the non-cognate PVDs that do not get transported by FpvA in vivo still retain enough PVDI-like features that they can bind with a Kd < 2 mM and with the same site specificity as the natural ligand. The fact that all of the ferric–PVDs induced nearly the same conformational changes in FpvA suggests that they all share some common features that are recognized for binding in this binding site. The overlay of the structures shows that this common region is likely to be the chromophore, the iron and its chelating groups (Fig. 5). Notably, the chirality of the octahedral iron co-ordination is the same for all of the ferric–PVDs. The remaining parts of the ferric–PVDs, consisting mostly of the peptide chain, diverge greatly in structure and share no common PVD–FpvA contacts.
The affinity determinants in the PVDs must then lie outside of the regions common to all PVDs. There are two main differences that are seen consistently for the high- and low-affinity PVD-bound structures. The first is the movement of the loop L7 that brings residues Trp-599 and Tyr-600 closer to the high-affinity PVDs but not with the low-affinity PVDs in which L7 moves little compared with apo-FpvA. This loop appears to be sensitive to the backbone conformation of the first three residues of the PVD chain and might play a role in PVD affinity. Asp-597 (also in L7) is near Arg-2 in PVDI or Lys-2 in PVDDSM50106, PVDPfl18.1 and PVDATCC13525, and although not within hydrogen bonding distance, there is a charge attraction. PVDG173 and PVDPa6 both lack a basic amino acid adjacent to the first Ser and therefore do not have this charge–charge interaction. Steric hindrance of Lys-2 in PVDDSM50106 has been shown to induce a 55-fold reduction of affinity to FpvA, supporting the significance of the Asp-597–Lys-2 interaction (Schons et al., 2005).
The other specific interaction for the high-affinity PVDs is a weak hydrogen bond or dipole–dipole interaction (3.6 Å) between the Asn-228 side-chain and the carbonyl of the second residue of the PVD peptide chain. This interaction is stabilized by a second hydrogen bond present only in the high-affinity structures between the Asn-228 side-chain and the backbone amide of Gly-230. For both of these potential affinity determinants, the conformation of the first three residues of the PVD chain is the critical factor. There are dozens of other interactions between the PVDs and FpvA so the overall binding affinity will have many other influences. Of these, the several aromatic residues in the beta barrel of FpvA have been shown to be important for ferric–PVD binding and transport (Shen et al., 2005).
The C-terminal regions of the ferric–PVD peptides are structurally very divergent. Apart from the side-chains of the residues involved in iron chelation that are constrained with a common orientation, the structural diversity suggests that this portion of PVD does not contribute extensively to the binding. It is worth noting that within the four high-affinity PVDs, the C-terminal peptide cycles of PVDI and PVDATCC13525 have a completely different orientation from those of PVDDSM50106 and PVDPfl18.1.
The L7 loop of FpvA in which Trp-599 and Tyr-600 move upon binding of the high-affinity PVDs has a counterpart in FecA, the Escherichia coli iron citrate transporter. FecA shows dramatic conformational changes of the loop L7 upon binding of iron citrate, and acts as a lock that closes the siderophore binding site, isolating it from the extracellular medium (Ferguson et al., 2002). While the conformational changes in L7 are subtler in FpvA, they also participate in the stabilization of the siderophore in its binding site. It is interesting to note that both FpvA and FecA have a homologous N-terminal domain involved in signal transduction. Upon binding of the ferric siderophore to its transporter, this N-terminal domain eventually interacts with an inner membrane protein, which in turn activates a cytoplasmic sigma factor (Beare et al., 2003; Braun et al., 2003). Furthermore, it has been shown that modifications in the L7 loop affect the signal transduction both for FpvA and for FecA, and eventually increase iron uptake (mutation S602C in FpvA) (Ferguson et al., 2007; Nader et al., 2007). The conformational changes observed for Trp-599 and Tyr-600 upon binding of PVDs on FpvA might then be involved in the signal transduction pathway, and it would be interesting to check whether or not the low-affinity PVDs can stimulate signalling when applied to cells at a high concentration.
While it is known that ferric siderophores can exist in solution with many conformations, our data indicate that FpvA and probably other receptors are specific for only one conformation. We have highlighted the chromophore, iron and hydroxamates as the common core of all ferric–PVDs; however, these structural elements that are common to all ferric–PVDs do not necessarily share a common lowest energy conformation. This lowest energy conformation is very sensitive to the peptide sequence as shown by the CD spectra of ferric–PVDs with similar sequences. It appears to be a variation in the conformation of these common elements that leads to a variation in affinities for the receptors. In addition to the role of the preferred conformations of the ferric–PVD, the chemical nature of the peptide sequence will influence the affinity through direct contacts with the receptor. Our data suggest that these direct contacts are more likely to be important for the residues near the chromophore where the structural variation among the bound high-affinity PVDs is the least.
Strains and plasmids
In addition to the wild-type PAO1 strain, we used two PVD-deficient strains of P. aeruginosa: CDC5 (Ankenbauer et al., 1986) and PAD06 (Takase et al., 2000) for the in vivo PVD dissociation and PVD–Fe ligand binding assays respectively. For the fluorescent in vivo displacement assays, we used the FpvA-deficient strain K691. To ensure high levels of FpvA expression for the fluorescence-based assays, we used cells transformed with the pPVR2 plasmid, a derivative of the E. coli–P. aeruginosa shuttle cloning vector pAK1900 carrying the fpvA gene on a 4.6 kb SphI insert (Poole et al., 1993).
In vivo fluorescence measurements
The fluorescence experiments were performed on K691(pPVR2) or CDC5(pPVR2) cells with a PTI (Photon Technology International TimeMaster; Bioritech) spectrofluorometer. For all experiments, the sample was stirred at 30°C in a 1 ml cuvette, the excitation wavelength (λex) was set at 290 nm (for the FRET experiments) or 400 nm (for direct excitation), and the emission of fluorescence (λem) was measured at 447 nm. The cells were grown to an OD600 of 1 in succinate medium and were washed once in 50 mM Tris-HCl (pH 8.0) and re-suspended in the same buffer to a final OD600 of 2. The fluorescence at 447 nm was measured every second for the duration of the experiment. Ferric–PVDs were added at a final concentration of 150 nM after 3 min for the FRET experiments and after 5 min in the direct excitation measurements. As a control, the same experiments were repeated in the absence of the siderophore (not shown for FRET measurement).
To obtain purified compounds required for analytical purposes, a previously described method (Meyer and Abdallah, 1978) was used, consisting of a chloroform-phenol extraction of the iron complexes of PVDs from the culture supernatants, followed by decomplexation and CM-Sephadex chromatography which allowed the separation of the different PVD forms as free ligands. Alternatively, pH 6-acidified culture supernatants were filtered on a XAD-4 Amberlite column. The PVDs were eluted with 50% methanol and further purified by chromatography on CM-Sephadex and Biogel P-2 (Budzikiewicz, 1993).
PVD-mediated 59Fe cross-uptake assay
Incorporation of 59Fe by PAO1 mediated by the different purified PVDs was performed as previously described (Meyer et al., 2002). Briefly, PAO1 cells from a 40 h culture in succinate medium were washed and re-suspended in fresh medium at an OD600 of 0.33. PVD–59Fe complexes were mixed with bacterial suspension, and incubated at 25°C. After 20 min incubation, the suspension was filtered and the radioactivity retained on the filter was counted. Results in Table 1 are expressed as percentage values where 100% corresponds to the incorporation of 59Fe mediated by PVDI.
For the in vivo determination of the apparent dissociation constants for the different PVD–Fe complexes binding to FpvA, we used the filtration assay as described previously (Schalk et al., 2001). After an overnight growth in succinate medium, PAD06 cells were prepared in 50 mM Tris-HCl (pH 8.0) buffer. The cells at an OD600 of 0.1 were incubated at 0°C in a final volume of 500 μl, in the presence of 200 μM of the protonophore carbonyl cyanide m-chlorophenylhydrazone (CCCP) and varying concentrations of siderophore–55Fe (0.1–80 μM; 2.1 Ci mmol−1), for 1 h. To evaluate the non-specific binding of siderophore–55Fe, the binding experiment was repeated, in parallel, for each concentration of siderophore with the same strain grown in LB medium, in which the expression of FpvA receptor is negligible.
FpvA was purified from CDC5(pPVR2) cells grown in succinate medium as previously described (Greenwald et al., 2008). Apo-FpvA at 6 mg ml−1 was crystallized as previously described in 1.3 M NaPO4 and MES, pH 6.5, using 2 μl each of protein and precipitant in a sitting drop (Brillet et al., 2007). After several weeks of crystal growth, 20 μl of ferric–PVD in mother liquor with 20% ethylene glycol was added directly to the crystals in the drop. The ferric–PVD concentration was 100 μM for the high-affinity PVDs and 2 mM for those PVDs that did not show significant affinity in the 55Fe ligand-binding experiment. The well solution was replaced with mother liquor in 20% ethylene glycol. After 6–24 h, the crystals were flash frozen in liquid nitrogen. Data collection was carried out at 100K on the EMBL beamlines BW7B and X11 at the DORIS storage ring, DESY, Hamburg. The structure of apo-FpvA (PDB entry: 205P) was used as a starting model for model building and refinement in COOT and REFMAC5. Libraries describing the geometric restraints of the PVD chromophore and the modified amino acids of the PVD were made using LIBCHECK. TLS refinement of signal domain and the plug/barrel domain of the two molecules was used and led to a lowering of R-free of 0.5–1.0%. The identity of the variable moiety on the PVD chromophore is unknown for the five heterologous PVDs and thus the occupancy of these atoms was set to zero for refinement. Co-ordinates and structure factors for the seven structures have been deposited in the Protein Data Bank (Table 2).
The measurements were performed on a Jasco J-815 spectrometer at 20°C in 0.5 and 0.1 cm path length cuvettes. The PVDs were dissolved at a concentration of 0.2 mM in 20 mM NaPO4 pH 8.0. The spectra were recorded from 600 to 240 nm in the 0.5 cm cuvette and from 260 to 200 in the 0.1 cm cuvette with buffer as a baseline. The ferric–PVDs were made by the addition of ferric citrate to a final concentration of 1.5 mM citrate and 0.3 mM Fe. The measurements were performed as with the apo-PVDs excepting that the baseline included the ferric citrate.
We thank SmithKline Beecham for generously providing carbenicillin. This work was partly funded by the Centre National de la Recherche Scientifique and the Association Vaincre la Mucoviscidose (French Association against Cystic Fibrosis). J.G. was supported by an EMBO postdoctoral fellowship.