Bacteria in an intense competition for iron: Keycomponent of the Campylobacter jejuni iron uptakesystem scavenges enterobactin hydrolysis productDaniel J. Rainesa, Olga V. Morozb, Elena V. Blagovab, Johan P. Turkenburgb, Keith S. Wilsonb,1, and Anne-K. Duhme-Klaira,1

aDepartment of Chemistry, University of York, Heslington, York YO10 5DD, United Kingdom; and bStructural Biology Laboratory, Department of Chemistry,University of York, Heslington, York YO10 5DD, United Kingdom

Edited by Kenneth N. Raymond, University of California, Berkeley, CA, and approved April 7, 2016 (received for review October 22, 2015)

To acquire essential Fe(III), bacteria produce and secrete siderophoreswith high affinity and selectivity for Fe(III) to mediate its uptake intothe cell. Here, we show that the periplasmic binding protein CeuE ofCampylobacter jejuni, which was previously thought to bind theFe(III) complex of the hexadentate siderophore enterobactin (Kd ∼ 0.4±0.1 μM), preferentially binds the Fe(III) complex of the tetradentateenterobactin hydrolysis product bis(2,3-dihydroxybenzoyl-L-Ser)(H5-bisDHBS) (Kd = 10.1 ± 3.8 nM). The protein selects Λ-configured[Fe(bisDHBS)]2− from a pool of diastereomeric Fe(III)-bisDHBS speciesthat includes complexes with metal-to-ligand ratios of 1:1 and 2:3.Cocrystal structures show that, in addition to electrostatic interactionsand hydrogen bonding, [Fe(bisDHBS)]2− binds through coordinationof His227 and Tyr288 to the iron center. Similar binding is observedfor the Fe(III) complex of the bidentate hydrolysis product 2,3-dihydroxybenzoyl-L-Ser, [Fe(monoDHBS)2]

3−. The mutation ofHis227 and Tyr288 to noncoordinating residues (H227L/Y288F) resultedin a substantial loss of affinity for [Fe(bisDHBS)]2− (Kd ∼ 0.5± 0.2 μM).These results suggest a previously unidentified role for CeuEwithin theFe(III) uptake system of C. jejuni, provide a molecular-level understand-ing of the underlying binding pocket adaptations, and rationalize re-ports on the use of enterobactin hydrolysis products by C. jejuni, Vibriocholerae, and other bacteria with homologous periplasmic bindingproteins.

siderophore | Fe(III) uptake | periplasmic binding protein | enterobactin |Campylobacter jejuni

With the rapid rise in bacterial resistance to antibiotics, abetter understanding of cooperative behavior in microbial

communities is urgently needed for the development of novel ap-proaches to controlling infections caused by resistant bacteria (1, 2).As an essential nutrient, iron is often a growth-limiting factor forbeneficial, commensal, and pathogenic bacteria alike, not onlydue to its low solubility in water under aerobic conditions at andaround neutral pH, but also because the host organism and com-peting microbes actively limit its availability (3, 4). Microorganismsevolved efficient Fe(III) uptake mechanisms to overcome thischallenge, a common strategy being the production of siderophores,small Fe(III)-chelating molecules with high affinity and selectivityfor Fe(III), with over 500 examples known to date (5). The sharingof siderophores is a recognized example for positive cooperativitythat has been linked to bacterial virulence (6). The best-character-ized siderophores are hexadentate ligands that form coordinativelysaturated, octahedral 1:1 complexes with Fe(III) (3, 5, 7), the moststudied being the triscatecholate enterobactin (H6-ENT) pro-duced by many enteric bacteria (8).In Escherichia coli, enterobactin is synthesized within the cell

and secreted through the cell membranes to capture Fe(III)from the environment. The resulting Fe(III)-enterobactin complex[Fe(ENT)]3− is recognized by the outer membrane receptor FepAand actively transported into the periplasm. In the periplasm,[Fe(ENT)]3− is sequestered by the periplasmic binding protein(PBP) FepB, which transfers it to an inner membrane trans-porter for further transport into the cytoplasm (9). Once there,

[Fe(ENT)]3− is hydrolyzed by an intracellular esterase to releaseFe(III) for use in the cell (8).Along with the development of structurally diverse siderophores,

microorganisms adapted their associated receptor and transportproteins for the uptake of the appropriate Fe(III) complexes (9). Togain a competitive advantage, many bacteria have evolved to poachsiderophores produced by other bacteria. Campylobacter jejuni, forexample, does not itself produce siderophores yet possesses anuptake system that is able to use siderophores from competingspecies (10). Initially, it was proposed that in C. jejuni [Fe(ENT)]3−

is transported across the outer membrane by the receptors CfrAand CfrB (11). Once in the periplasm, [Fe(ENT)]3− was proposedto bind to the PBP CeuE, the resulting complex enabling thetransport of the ferric siderophore into the cytoplasm (12, 13).In addition, increasing numbers of lower-denticity siderophores

are being isolated from bacterial cultures and found to coordinateFe(III) and mediate its uptake (14–18). For example, the trilactonebackbone of enterobactin makes it prone to hydrolysis, and al-though this lability is necessary to allow the intracellular release ofFe(III) from the siderophore, it in addition leads to its slow deg-radation in aqueous media (7, 19–21). Three hydrolysis productsare formed: tris(2,3-dihydroxybenzoyl-L-Ser) (H7-trisDHBS),bis(2,3-dihydroxybenzoyl-L-Ser) (H5-bisDHBS), and 2,3-dihydroxy-benzoyl-L-Ser (H3-monoDHBS), with all three found in the growthmedium of E. coli (Fig. 1). Although enterobactin, once secreted, isalso available to other cells (producers or nonproducers), it can

Significance

Almost all bacteria require Fe(III) for survival and growth. Tocompete successfully for this essential nutrient, bacteria developedvery efficient Fe(III) uptake mechanisms based on high-affinityFe(III) chelators, so-called siderophores. To gain a competitive ad-vantage, many bacteria have evolved to scavenge and effectivelypoach siderophores from other species. Enterobactin, one of thestrongest Fe(III) chelators known, is produced and secreted bymany enteric bacteria. We show that a key protein involved inFe(III) uptake in the foodborne pathogen Campylobacter jejuni isadapted to scavenge enterobactin hydrolysis products, a strategythat may enable the pathogen to more efficiently exploit side-rophores produced by other bacteria and hence their resources.

Author contributions: K.S.W. and A.-K.D.-K. designed research; D.J.R., O.V.M., E.V.B., andJ.P.T. performed research; D.J.R., O.V.M., J.P.T., K.S.W., and A.-K.D.-K. analyzed data; andD.J.R., K.S.W., and A.-K.D.-K. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

Data deposition: The atomic coordinates and structure factors have been deposited in theProtein Data Bank, www.pdb.org (PDB ID codes 5ADW and 5ADV). Experimental datasetsassociated with this paper have been deposited with the University of York library (www.york.ac.uk/library/info-for/researchers/datasets/).1To whom correspondence may be addressed. Email: anne.duhme-klair@york.ac.uk orKeith.Wilson@york.ac.uk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1520829113/-/DCSupplemental.

5850–5855 | PNAS | May 24, 2016 | vol. 113 | no. 21 www.pnas.org/cgi/doi/10.1073/pnas.1520829113

Dow

nloa

ded

by g

uest

on

Janu

ary

24, 2

021

only be used once because Fe(III) release requires its hydrolysis.The enterobactin hydrolysis products, however, could be used againas secondary, lower-denticity siderophores.It has been demonstrated that the human pathogens C. jejuni

(10, 22, 23) and Vibrio cholerae (24), the causes of food poisoningand cholera, respectively, can use enterobactin hydrolysis productsfor the uptake of Fe(III) from their environment. Both are knownnot to produce enterobactin.An alternative Campylobacter Fe(III) acquisition model that relies

on these linear hydrolysis products was recently proposed based onthe finding that Cee, the sole trilactone esterase of C. jejuni andCampylobacter coli, is located in the periplasm, i.e., these bacteria areunable to degrade enterobactin within the cytoplasm (25). The modelsuggests that, once the Fe(III) complex of enterobactin enters theperiplasm, its ester backbone is cleaved by Cee, which is highly ef-ficient in hydrolyzing both the Fe(III) complex and apo-enterobactin.The resulting hydrolysis products, mainly the tetradentate ligandbisDHBS5− and the bidentate ligand monoDHBS3−, are then used tomediate the subsequent transport of Fe(III) into the cytoplasm.The identification of the esterase Cee and in particular the ob-

servation that bisDHBS5− can be used independently of enter-obactin (25) raise important questions about the siderophorepreference of the PBP CeuE and its role in the iron uptake inC. jejuni. By using siderophore mimics, we previously demonstratedthat CeuE can bind the Fe(III) complexes of both hexadentate andtetradentate catecholate ligands (26, 27). Our cocrystal structuresrevealed that CeuE interacts with the coordinatively saturatedFe(III) complex of the hexadentate mimic MECAM6− throughelectrostatic interactions and hydrogen bonding, whereas it bindsthe coordinatively unsaturated complex of the tetradentate mimic4-LICAM4− by recruiting the side chains of two amino acid resi-dues (His227 and Tyr288) to complete the coordination sphere ofthe Fe(III) center. We established that His227 and Tyr288 areconserved among a subset of related PBPs, including VctP fromV. cholerae, and suggested that this subset of PBPs undergo similarstructural changes to adapt to the binding of lower-denticity sidero-phores (27). The recent report that V. choleraemost efficiently usestrisDHBS7− and bisDHBS5− for the acquisition of Fe(III) provideda partial confirmation of this prediction (24).Here, we report that CeuE binds [Fe(bisDHBS)]2− with much

higher affinity than the Fe(III) complex of enterobactin, reveal thestructural basis for this difference in binding strength, and examinekey aspects of the relevant Fe(III) coordination chemistry in solution.

ResultsCrystal Structures of CeuE-[Fe(bisDHBS)]2− and CeuE-[Fe(monoDHBS)2]

3−.A previously reported total synthesis of enterobactin was adapted to

synthesize H5-bisDHBS, with modifications aimed at simplifying theprocedures (28). The Fe(III) complex of bisDHBS5− was soakedinto apo-CeuE crystals, grown as reported previously (27). Experi-mental details for the preparation of the Fe(III) complex ofbisDHBS5−, crystal growth conditions, data collection, and re-finement are provided in Supporting Information. Crystallographicstatistics and a discussion of unit cell dimensions and average Bvalues can be found in Supporting Information (Table S1).Three crystals were selected; crystal I was soaked for 90 min,

crystal II for 24 h (PDB ID code 5ADW), and crystal III for 11 d(PDB ID code 5ADV). As expected, all three crystals were in spacegroup P1 with three protein monomers per asymmetric unit, con-sistent with the apo-CeuE structure (PDB ID code 3ZKW) (27).Crystal I showed no additional electron density in the CeuE bindingpockets, indicating that the 90-min soaking time was too short.Crystal II has additional electron density in the binding pockets

of each of the three independent CeuE monomers, which wasmodeled as [Fe(bisDHBS)]2− in chains A and B (Fig. 2 and Fig.S1). CeuE binds a single [Fe(bisDHBS)]2− species per monomer inan Fe(III):siderophore:protein ratio of 1:1:1. BisDHBS5− chelatesthe Fe(III) center in a tetradentate fashion with the four cat-echolate oxygen donors coordinated. The remaining two co-ordination sites of the distorted octahedral Fe(III) center areoccupied by donor atoms provided by the side chains of His227 andTyr288. In addition, electrostatic interactions and hydrogen bondsinvolving residues Arg118, Arg205, Arg249, and Lys121 contributeto the binding (Fig. 3). The diserine linker in bisDHBS5− forces theoxygen donors in the ortho and meta position to the amide groupson the catechol units to occupy adjacent coordination sites on theFe(III) center. Consequently, the remaining two coordination sitesare not equivalent. The site opposite the weaker catecholate donor(ortho position) is occupied by the oxygen donor of Tyr288,whereas that opposite the stronger oxygen donor (meta position) isoccupied by the nitrogen donor of His227, consistent with theirrespective trans-influence (29). Although the observed Fe(III)binding mode is similar to that which we reported previously forthe C2-symmetric siderophore mimic 4-LICAM4− (PDB ID code5A1J) (27), the diserine backbone of bisDHBS5− plays a key role indetermining the orientation of the unsymmetrical natural sidero-phore. In both chains A and B, the carboxylate group in thebackbone of bisDHBS5− provides an additional negative chargeand accepts a hydrogen bond from Arg249 (Fig. 3). This highdegree of complementarity indicates that the binding pocket ofCeuE is perfectly suited to the binding of the tetradentate ligand.In the binding pocket of chain C, there was much less density and

it was modeled as a half-occupied Fe(III) center with no co-ordinating ligand. The anomalous difference Fourier maps confirma peak of about 50% of that seen for the Fe(III) in chains A and B.

Fig. 1. Molecular structure of enterobactin, its hy-drolysis products, the siderophore mimic H6-MECAM,and a selection of tetradentate siderophores.

Raines et al. PNAS | May 24, 2016 | vol. 113 | no. 21 | 5851

CHEM

ISTR

YBIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Janu

ary

24, 2

021

There is some residual density around the Fe(III), but it was notpossible to satisfactorily model this as bisDHBS5− or monoDHBS3−.Although the binding pockets of chains A and B are both open tosolvent, access to the pocket of chain C is restricted by the N ter-minus of chain B. The resulting steric constraints explain the onlypartial occupation of this pocket (Fig. 2A).Crystal III has similar cell dimensions and packing to crystal II

(Table S2). There is electron density in each of the three bindingpockets (Figs. S1 and S2), the density is weak in the region corre-sponding to the backbone of bisDHBS5− (below the 1σ level), sug-gesting that the di-serine ester-linkage of bisDHBS5− was highlymobile or had hydrolyzed or a combination of both. The best fitto the density was attained when chain B was modeled with[Fe(bisDHBS)]2− bound while chains A and C were modeled withtwo monoDHBS3− ligands coordinating the Fe(III) center (Fig. 2).Remarkably, the orientation of the monoDHBS3− ligands in chain Cdiffered from that in chain A. In A, the oxygen donors in both theortho and meta position of the catecholamide units occupy adjacentcoordination sites on the Fe(III) center, as seen in [Fe(bisDHBS)]2−.In contrast, in chain C, one of the monoDHBS3− ligands was in aflipped position, with the attached serine of the backbone pointinginto the binding pocket. Here, the donors in the ortho position of thetwo catecholamide units coordinate trans to one another, whereas thedonors in themeta position remain coordinated cis to one another. Inthis isomer, His227 and Tyr288 occupy coordination sites trans to thecatecholate oxygen donor atoms in the meta position (Fig. 2). Thischange in orientation of the ligands is likely to be due to stericconstraints imposed by the N terminus of chain B (Fig. S3). CrystalIII thus reveals two alternative binding modes for the Fe(III) complexof the monomeric enterobactin hydrolysis product monoDHBS3−.In crystals II and III, the Fe(III) complexes are bound with the

metal center in the Λ-configuration, as previously observed in thecocrystal structures of (CeuE)2-[{Fe(MECAM)}2]

6− (26), CeuE-[Fe(4-LICAM)]− (27), FeuA-[Fe(bacillibactin)]3− (30), FeuA-[Fe(MECAM)]3−, and FeuA-[Fe(ENT)]3− (31).

Relevant Fe(III) Coordination Chemistry in Solution. The identificationof the Λ-configured 1:1 complex [Fe(bisDHBS)]2− in the cocrystal

structures prompted us to investigate whether this complex is also thepredominant species formed in solution. Due to the mismatch be-tween the four donor atoms provided by the two chelating catecholunits of bisDHBS5− and the preferred sixfold coordination environ-ment of Fe(III), the composition and structure of the complex(es)formed have long been a matter of debate. Based on a UV-visibleabsorbance spectroscopic study carried out in aqueous solution atpH 9, it was suggested that bisDHBS5− forms a 1:1 complex withFe(III) under these conditions and that the remaining two co-ordination sites are occupied by solvent molecules (15). In contrast,NMR spectroscopic investigations carried out in d6-DMSO withGa(III) as a diamagnetic substitute for Fe(III) indicated a metal-to-ligand ratio of 2:3. Consequently, the formation of a dinuclearcomplex of composition {Fe2(bisDHBS)3} was proposed (32).Stoichiometric control allowed the isolation of two different

Fe(III) complexes of a tetradentate siderophore mimic that con-sists of two ethyl-bridged catecholamides. A metal-to-ligand ratioof 2:3 resulted in the crystallization of a dinuclear triple-strandedhelicate of composition [Fe2(L)3]

6−, whereas a basic solution withmetal-to-ligand ratio of 1:1 produced crystals of hydroxo-bridgeddinuclear complexes of composition [{Fe(L)(OH)}2]

4− (29).We carried out experiments to rationalize these observations

and related the results to biologically relevant conditions.To probe the metal-to-ligand ratio, we used Job’s method of

continuous variation (33, 34), monitoring complex formation byboth electronic absorbance and CD spectroscopies. The spectro-scopic measurements were carried out in aqueous solution [5%(vol/vol) DMSO] at pH 7.5 using Fe(III) nitrate as the Fe(III)source. Nitrilotriacetic acid (NTA) was added to prevent the pre-cipitation of Fe(III) hydroxides. Control experiments ascertainedthat the presence of NTA does not interfere with the reaction ofFe(III) with bisDHBS5− under these conditions.Electronic absorbance spectra were recorded at selected metal-

to-ligand ratios in the region of the ligand-to-metal charge transfer(LMCT) band of [Fe(bisDHBS)]2− (Fig. 4A). The maximum of theLMCT band shifts significantly upon variation of the metal-to-ligand ratio. A red complex forms under conditions of excess li-gand. The λmax at ∼510 nm is consistent with the formation of atris(catecholate) species (35, 36). When the metal-to-ligand ratioapproaches 1:1, the color of the solution changes from red to purple.The observed λmax of around 565 nm is consistent with the for-mation of a bis(catecholate) species (36, 37). The Job plot confirmsthat a λmax around 510 nm is consistent with a metal-to-ligand ratioof 2:3, whereas a λmax around 565 nm is consistent with a metal-to-ligand ratio of 1:1 (Fig. 4B).In summary, the electronic absorbance spectra demonstrate that

in aqueous solution near neutral pH, both 1:1 and 2:3 complexesare formed upon reaction of Fe(III) with H5-bisDHBS (Fig. S4),with the species distribution being controlled by stoichiometry.Similarly, the visible region CD spectra vary with the metal-to-

ligand ratio (Fig. 5A). At metal-to-ligand ratios of 3:2 and 1:1, the

Fig. 2. (A) Ribbon representation of the three independent CeuE chains in theunit cell of crystal II (PDB ID code 5ADW) with a black dashed line representingsteric clash of chain B N terminuswith the binding pocket of chain C. (B–D) Cylinderrepresentation of the different ligand binding orientations found in crystals II andIII with electron density shown as 2Fobs-Fcalc, contoured at 1σ: (B) crystal II chainB; (C) crystal III chain A; and (D) crystal III chain C. Ligands are shown as cylinders:gray, carbon; blue, nitrogen; red, oxygen; and iron, coral; and Arg, Tyr, and His sidechains are shown as cylinders with carbon atoms colored as their ribbons.

Fig. 3. (A) Overlay of the three independent [Fe(bisDHBS)]2−units and key residuesin the respective substrate binding pocket of CeuE shown as cylinders with thecarbon atoms colored according to chain: crystal II chains A (blue) and B (gold), crystalIII chain B (green), dashed H bonds exemplified for crystal II chain B. (B) Schematicdiagram of key hydrogen-bonding interactions (charges omitted for clarity).

5852 | www.pnas.org/cgi/doi/10.1073/pnas.1520829113 Raines et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

24, 2

021

CD spectra show an intense positive band around 320 nm, a weakerpositive band around 410 nm, and a weak bisignate band centeredat 595 nm, which is near the LMCT absorption maximum of the 1:1species. Scarrow et al. (15) reported a similar CD spectrum for[Fe(bisDHBS)]2− isolated from growth cultures and assigned theintense near-UV band to ligand-based transitions and the weakervisible region features to LMCT transitions. Because the latter areindicative of the chirality at the Fe(III) center, the low intensity ofthe signal indicates that the chiral induction by the L-serine residuesin the backbone is only weak.Upon changing the metal-to-ligand ratio to 2:3, the maximum of

the near-UV band shifts from 320 to 340 nm and the visible regionnow shows a bisignate band with a positive signal at 415 nm and anegative signal at 550 nm, consistent with a negative Cotton effect.The position of the zero CD crossover at 503 nm is close to theLMCT absorption maximum (∼510 nm) of the Fe(III)-bisDHBS5−2:3 species. The resulting CD spectrum is characteristic of a solu-tion of Fe(III) tris(catecholamide) complexes in which the Δ-con-figuration dominates (38). This characteristic change in the CDprofile coincides with the color change from purple to red reflectedin the electronic absorbance spectra (Fig. 4A). The concomitantincrease in amplitude is consistent with the increase in the numberof coordinated catechol units, not only because each is expected tocontribute to the chiral induction effect, but also due to the pairwiseadditivity of exciton-coupled chromophores (39). The effect is alsoconsistent with the diastereoselective formation of a dinuclear triplehelicate, in which the enantiomerically pure ligand predeterminesthe helical direction and hence the chirality of both metal centers,as seen in similar supramolecular assemblies (40, 41).

Similar observations were reported for the amonabactins, a familyof bis(catecholate) siderophores in which the catecholate units are at-tached to the Ne atom of lysine residues present in their peptide-basedbackbones (16, 42, 43). However, the increase in Cotton effect upondecrease of the metal-to-ligand ratio from 1:1 to 2:3 is more pro-nounced with H5-bisDHBS than for the amonabactins. This is probablydue to a stronger chiral induction from the Cα of the serine backbonein the linear dimer, which is closer to the Fe(III) center than the Cα

of the lysine residues in the backbone of the amonabactins (Fig. 1).

Interactions Between CeuE and the Fe(III) Complexes of Siderophoresin Solution.CD. To probe the stereochemistry of the Fe(III) center in thepresence of CeuE in solution, CD spectra were recorded in the

Fig. 4. (A) Selected electronic absorbance spectra [colors from black to lightblue ordered by metal-to-ligand ratio as indicated; [M] + [L] = 0.4 mM; 0.1 MTris·HCl with 5% (vol/vol) DMSO; pH 7.5]. (B) Job plot for the reaction ofH5-bisDHBS with Fe(III), obtained by following the absorbance at 512 nm (redtriangles) and 563 nm (blue circles). The absorbance values are averages of twoexperiments (error bars indicate the difference between the runs).

Fig. 5. (A) CD spectra of solutions of H5-bisDHBS or Fe(III) plus H5-bisDHBS atkey metal-to-ligand ratios [0.1 M Tris·HCl buffer with 5% (vol/vol) DMSO, pH 7.5,150mMNaCl, [M] + [L] = 0.4 mM]. (B and C) CD spectra of solutions of Fe(III) plusH5-bisDHBS (B) or Fe(III) plus enterobactin (C) recorded after addition of increasingincrements of CeuE as indicated (0.1 M Tris·HCl buffer, pH 7.5, 150 mM NaCl).

Raines et al. PNAS | May 24, 2016 | vol. 113 | no. 21 | 5853

CHEM

ISTR

YBIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Janu

ary

24, 2

021

wavelength range of the LMCT band. Series of spectra wereobtained during a titration of a solution of CeuE into solutions ofFe(III) plus H5-bisDHBS or Fe(III) plus enterobactin in buffer (pH7.5). The titrations allowed the changes in spectral features uponbinding of the respective Fe(III) siderophore complex to CeuE tobe observed (Fig. 5 B and C).The first CD spectrum was obtained after addition of Fe(III) and

H5-bisDHBS to the buffer, followed by equilibration. It confirmsthat, in the absence of CeuE, the L-serine backbone in bisDHBS5−

has only a weak influence on the stereochemistry at the Fe(III)center, with Δ-configured complexes in slight excess.The spectra recorded upon addition of CeuE indicate inversion of

configuration at the metal center with isodichroic points at 305 and544 nm. The positive band at 320 nm changes into a negative band at330 nm. The weak positive signal across the LMCT range developsinto a broad negative band with a minimum at 395 nm and a positiveband with a maximum at 600 nm, consistent with a Λ-configuredFe(III) center (38). The series of spectra confirms that CeuE selectsfor Λ–configured [Fe(bisDHBS)]2− and shows that the diastereo-isomers that remain in solution are able to reequilibrate quickly.For comparison, an analogous titration was carried out with a

solution containing Fe(III) and enterobactin (Fig. 5C). As expectedfrom previous reports, the initial CD spectrum of [Fe(ENT)]3−

is characteristic of Fe(III) tris(catecholamide) complexes withΔ-configuration (38). Upon addition of CeuE, the bands decreasein intensity, indicating a shift in the equilibrium between Δ- andΛ-configured complexes is caused by the Fe(III) complex ofenterobactin binding to CeuE. However, even after addition of 1.6equivalents of CeuE, Δ-configured [Fe(ENT)]3− still predominates.The observed selection of Λ-configured complexes by CeuE indi-

cates that the chiral L-serine backbone, which has a Δ-configuration–inducing effect, slightly hinders rather than facilitates protein binding.The resulting moderation of binding affinity is consistent with theintermediary role of the periplasmic binding protein within the irontransport pathway. A precise tuning of affinity is essential to allow thetransfer of the cargo to the inner membrane transporter in the finalstep of the Fe(III) transport chain.Fluorescence spectroscopy. To allow a quantitative comparison of thebinding strength, the affinity of CeuE for [Fe(bisDHBS)]2− and[Fe(ENT)]3− was determined using intrinsic fluorescence quenchingexperiments. Aliquots of the respective Fe(III)-siderophore solution(1:1 ratio, 12 μM in aqueous 40 mM Tris·HCl, 150 mM NaCl) wereadded successively to a solution of CeuE (240 nM in aqueous 40 mMTris·HCl, 150 mM NaCl, pH 7.5) and the fluorescence intensity wasmeasured after equilibration (λexc = 280, λem = 310–410 nm) (Fig.S5). The dissociation constants (Kd) were obtained from three in-dependent measurements, as described in Supporting Information.Surprisingly, CeuE was found to bind [Fe(bisDHBS)]2−

with much higher affinity (Kd = 10.1 ± 3.8 nM) than [Fe(ENT)]3−

(Kd ∼ 0.4 ± 0.1 μM). This observation shows that the direct co-ordination of the His227 and Tyr288 side chains to the Fe(III)center more than compensates for the loss of electrostatic attractioncaused by the decrease in negative charge upon moving from ahexadentate to a tetradentate catecholate siderophore.To confirm the importance of His227 and Tyr288 for the binding

affinity of CeuE to [Fe(bisDHBS)]2−, site-directed mutagenesis wasused to replace the Fe(III)-coordinating amino acids with non-coordinating, yet sterically similar residues. As anticipated, mutationof His227 to leucine (H227L) and Tyr288 to phenylalanine (Y288F)resulted in a clear drop in binding affinity. The dissociation constantobtained with the double mutant (Kd ∼ 0.5 ± 0.2 μM) indicatesweak binding, comparable to that observed for CeuE-[Fe(ENT)]3−

(Kd ∼ 0.4 ± 0.1 μM). Both dissociation constants are about twoorders of magnitude higher than that for CeuE-[Fe(bisDHBS)]2−

(Kd = 10.1 ± 3.8 nM). In contrast, the dissociation constant obtainedfor CeuE-[Fe(bisDHBS)]2− is similar to those previously publishedfor the binding of hexadentate tris(catecholate) siderophore com-plexes to PBPs, such as FeuA and FepB (10–50 nM) (30, 31, 44–46).

DiscussionEnterobactin is one of the most efficient mediators of bacterialFe(III) uptake, and increasing evidence indicates that its hydro-lysis products, in particular bisDHBS5−, play an important role inthe competition for essential Fe(III) because they can be exploitedby siderophore poachers, such as C. jejuni.Our studies demonstrate that CeuE, the periplasmic binding

protein of C. jejuni, selects Λ-[Fe(bisDHBS)]2− from a mixture ofspecies in solution. In the absence of the protein, bisDHBS5− wasfound to coordinate to Fe(III) in metal-to-ligand ratios of both 1:1and 2:3, depending on the relative concentrations of Fe(III) andsiderophore present in solution. In both complexes, chiral in-duction from the Cα of the L-serine backbone causes the Fe(III)center(s) to adopt preferentially the Δ-configuration in aqueousbuffer at pH 7.4, with weak (1:1 complex) to moderate (2:3 com-plex) diastereoselectivity. The ability of CeuE to select a definedFe(III) complex from a mixture is consistent with its role in theFe(III) transport process, where it functions as a periplasmicgatekeeper and is required to deliver the correct Fe(III) complex tothe inner membrane transporter. CD spectra recorded during thetitration of CeuE into a solution of [Fe(bisDBHS)]2− indicated aninversion of configuration from Δ to Λ upon addition of the pro-tein. The effect is caused by CeuE driving the equilibrium byselecting for Λ-configured complexes; the two diastereomers thatremain in solution reequilibrate quickly.Of particular significance is the revelation that CeuE binds

[Fe(bisDBHS)]2− ∼100 times more tightly than [Fe(ENT)]3−, withKd = 10.1± 3.8 nM and Kd ∼ 0.4± 0.1 μM, respectively. This markeddifference strongly suggests that [Fe(bisDHBS)]2− is the preferredligand for the protein rather than [Fe(ENT)]3− as previously thought.Our cocrystal structures provide molecular-level insights into the

binding pocket adaptations that give rise to this pronounced dif-ference in affinity and are the first (to our knowledge) to reveal theinteractions of a natural tetradentate siderophore with its cognatePBP. In addition to electrostatic and hydrogen-bonding interac-tions between the binding pocket and [Fe(bisDHBS)]2−, the crystalstructure showed the direct coordination of two amino acid sidechains (His227 and Tyr288) to the Fe(III) center. In contrast, thebinding of the Fe(III) complex of MECAM, the hexadentateenterobactin mimic, was found earlier to rely solely on electrostaticinteractions and hydrogen bonding (26). The oxygen donor ofTyr288, which is directly coordinated in CeuE-[Fe(bisDHBS)]2−,is positioned 3.5 Å away from the metal center in (CeuE)2-[{Fe(MECAM)}2]

6−, where it participates in hydrogen bonding.His227 is disordered forming part of a flexible loop in both theapo-CeuE and (CeuE)2-[{Fe(MECAM)}2]

6− structures. Hence, wepropose that His227 and Tyr288 enable CeuE to preferentially bindthe tetradentate enterobactin hydrolysis product rather than intacthexadentate enterobactin. This proposal is supported by the observa-tion that the replacement of His227 and Tyr288 with noncoordinatingamino acids leads to a drastic drop in affinity for [Fe(bisDHBS)]2−.There is an equivalent direct coordination of His227 and Tyr288

in the crystal structure of CeuE-[Fe(monoDHBS)2]3−, in which two

monoDHBS3− ligands, the hydrolysis products of bisDHBS5−, co-ordinate to the Fe(III) center. In both crystals, the Fe(III) com-plexes are bound with the metal center in Λ-configuration.A sequence homology search of reported 3D structures (Fig.

S6) suggests that a subset of related Fe(III)-siderophore bindingproteins shares a preference for the hydrolysis products of enter-obactin over intact enterobactin. The PBP of the Vct uptake systemin V. cholerae has a high similarity to CeuE and contains the co-ordinating histidine and tyrosine residues (27). Hence, our resultsrationalize the recent discovery that V. cholerae is able to use thelinear derivatives of enterobactin, but not enterobactin itself (24).We therefore propose that CeuE preferentially selects for

[Fe(bisDHBS)]2− in the periplasm and mediates its uptake into thecell. As a result, the Fe(III) complex of enterobactin is retarded inthe periplasm where it is exposed to esterase-catalyzed hydrolysis toits linear components. Because it has been reported that C. jejunican use bisDHBS5− as a siderophore independently of enterobactin(10) and the discovery of the trilactone esterase Cee in the periplasm

5854 | www.pnas.org/cgi/doi/10.1073/pnas.1520829113 Raines et al.

Dow

nloa

ded

by g

uest

on

Janu

ary

24, 2

021

of C. jejuni (25), it appears that CeuE, previously considered to be a[Fe(ENT)]3− binding protein, is instead adapted to preferentiallybind [Fe(bisDHBS)]2−. This ability to use the enterobactin hydrolysisproduct for Fe(III) uptake and hence to avoid the metabolic costs ofsiderophore production provides C. jejuni with a competitive ad-vantage when it encounters enterobactin-secreting bacteria eitherwithin the host or contaminated water.In contrast, in E. coli, the PBP FepB does not possess equivalent

tyrosine and hisitidine residues and the esterase that cleaves theFe(III) complex of enterobactin is located in the cytoplasm. Con-sequently, the Fe(III)-siderophore uptake systems of the side-rophore scavenger C. jejuni and the siderophore producer E. colishow significant differences.Taken together, these insights strongly suggest an important

in vivo role for the tetradentate siderophore and its cognate PBP in

the uptake of Fe(III) by Campylobacter and a number of otherpathogenic bacteria.

MethodsExperimental details including generalmethods, compound synthesis (Fig. S7) andcharacterization, site-directed mutagenesis, protein crystallization, X-ray datacollection, structure solution and refinement (Tables S1 and S2), and fluorescencequenching data (Figs. S5 and S8) can be found in Supporting Information.

ACKNOWLEDGMENTS. We thank Mr. S. Hart for assistance with datacollection, Dr. A. Leech for help with CD spectroscopy, and the DiamondLight Source for access to beamlines I02 and I04 (proposal; mx-7864). Wethank the UK Biotechnology and Biological Sciences Research Council andthe Engineering and Physical Sciences Research Council (Grant EP/L024829/1)for financial support.

1. Jiricny N, et al. (2010) Fitness correlates with the extent of cheating in a bacterium.J Evol Biol 23(4):738–747.

2. Griffin AS, West SA, Buckling A (2004) Cooperation and competition in pathogenicbacteria. Nature 430(7003):1024–1027.

3. Schalk IJ, Guillon L (2013) Fate of ferrisiderophores after import across bacterial outermembranes: Different iron release strategies are observed in the cytoplasm or peri-plasm depending on the siderophore pathways. Amino Acids 44(5):1267–1277.

4. Raymond KN, Allred BE, Sia AK (2015) Coordination chemistry of microbial irontransport. Acc Chem Res 48(9):2496–2505.

5. Hider RC, Kong X (2010) Chemistry and biology of siderophores. Nat Prod Rep 27(5):637–657.6. West SA, Buckling A (2003) Cooperation, virulence and siderophore production in

bacterial parasites. Proc Biol Sci 270(1510):37–44.7. Miethke M, Marahiel MA (2007) Siderophore-based iron acquisition and pathogen

control. Microbiol Mol Biol Rev 71(3):413–451.8. Raymond KN, Dertz EA, Kim SS (2003) Enterobactin: An archetype for microbial iron

transport. Proc Natl Acad Sci USA 100(7):3584–3588.9. Krewulak KD, Vogel HJ (2008) Structural biology of bacterial iron uptake. Biochim

Biophys Acta 1778(9):1781–1804.10. Naikare H, et al. (2013) Campylobacter jejuni ferric-enterobactin receptor CfrA is

TonB3 dependent and mediates iron acquisition from structurally different catecholsiderophores. Metallomics 5(8):988–996.

11. Xu F, Zeng X, Haigh RD, Ketley JM, Lin J (2010) Identification and characterization of anew ferric enterobactin receptor, CfrB, in Campylobacter. J Bacteriol 192(17):4425–4435.

12. Miller CE, Williams PH, Ketley JM (2009) Pumping iron: Mechanisms for iron uptake byCampylobacter. Microbiology 155(Pt 10):3157–3165.

13. Stahl M, Butcher J, Stintzi A (2012) Nutrient acquisition and metabolism by Cam-pylobacter jejuni. Front Cell Infect Microbiol 2:5.

14. Knosp O, von Tigerstrom M, Page WJ (1984) Siderophore-mediated uptake of iron inAzotobacter vinelandii. J Bacteriol 159(1):341–347.

15. Scarrow RC, Ecker DJ, Ng C, Liu S, Raymond KN (1991) Iron(III) coordination chemistry oflinear dihydroxyserine compounds derived from enterobactin. Inorg Chem 30(5):900–906.

16. Telford JR, Raymond KN (1997) Amonabactin: A family of novel siderophores from apathogenic bacterium. J Biol Inorg Chem 2(6):750–761.

17. Brandel J, et al. (2012) Pyochelin, a siderophore of Pseudomonas aeruginosa: Physicochemicalcharacterization of the iron(III), copper(II) and zinc(II) complexes.Dalton Trans 41(9):2820–2834.

18. Miethke M (2013) Molecular strategies of microbial iron assimilation: From high-affinity complexes to cofactor assembly systems. Metallomics 5(1):15–28.

19. Brickman TJ, McIntosh MA (1992) Overexpression and purification of ferric enter-obactin esterase from Escherichia coli. Demonstration of enzymatic hydrolysis ofenterobactin and its iron complex. J Biol Chem 267(17):12350–12355.

20. Cohen SM, Meyer M, Raymond KN (1998) Enterobactin protonation and iron release:Hexadentate tris-salicylate ligands as models for triprotonated ferric enterobactin.J Am Chem Soc 120(25):6277–6286.

21. Lin H, Fischbach MA, Liu DR, Walsh CT (2005) In vitro characterization of salmochelin andenterobactin trilactone hydrolases IroD, IroE, and Fes. J Am Chem Soc 127(31):11075–11084.

22. Hancock RE, Hantke K, Braun V (1977) Iron transport in Escherichia coli K-12.2,3-Dihydroxybenzoate-promoted iron uptake. Arch Microbiol 114(3):231–239.

23. Hantke K (1990) Dihydroxybenzoylserine—a siderophore for E. coli. FEMS MicrobiolLett 55(1-2):5–8.

24. Wyckoff EE, Allred BE, Raymond KN, Payne SM (2015) Catechol siderophore transportby Vibrio cholerae. J Bacteriol 197(17):2840–2849.

25. Zeng X, Mo Y, Xu F, Lin J (2013) Identification and characterization of a periplasmictrilactone esterase, Cee, revealed unique features of ferric enterobactin acquisition inCampylobacter. Mol Microbiol 87(3):594–608.

26. Müller A, Wilkinson AJ, Wilson KS, Duhme-Klair A-K (2006) An [Fe(mecam)2]6− bridge

in the crystal structure of a ferric enterobactin binding protein. Angew Chem Int EdEngl 45(31):5132–5136.

27. Raines DJ,MorozOV,Wilson KS, Duhme-Klair A-K (2013) Interactions of a periplasmic bindingprotein with a tetradentate siderophore mimic. Angew Chem Int Ed Engl 52(17):4595–4598.

28. Rastetter WH, Erickson TJ, Venuti MC (1981) Synthesis of iron chelators. Enterobactin,enantioenterobactin, and a chiral analog. J Org Chem 46(18):3579–3590.

29. Enemark EJ, Stack TDP (1996) Spectral and structural characterization of two ferriccoordination modes of a simple bis(catecholamide) ligand: Metal-assisted self-assemblyin a siderophore analog. Inorg Chem 35(10):2719–2720.

30. Peuckert F, Miethke M, Albrecht AG, Essen LO, Marahiel MA (2009) Structural basisand stereochemistry of triscatecholate siderophore binding by FeuA. Angew Chem IntEd Engl 48(42):7924–7927.

31. Peuckert F, et al. (2011) The siderophore binding protein FeuA shows limited pro-miscuity toward exogenous triscatecholates. Chem Biol 18(7):907–919.

32. Bergstrom CP, Lu MC, Bell CL (1991) NMR studies of dimeric 2,3-dihydroxy-N-benzoylserine. J Nat Prod 54(4):1003–1008.

33. Job P (1928) Formation and stability of inorganic complexes in solution.Ann Chim 9:113–203.34. Renny JS, Tomasevich LL, Tallmadge EH, Collum DB (2013) Method of continuous

variations: Applications of job plots to the study of molecular associations in organ-ometallic chemistry. Angew Chem Int Ed Engl 52(46):11998–12013.

35. Harris WR, et al. (1979) Coordination chemistry of microbial iron transport com-pounds. 19. Stability constants and electrochemical behavior of ferric enterobactinand model complexes. J Am Chem Soc 101(20):6097–6104.

36. Harris WR, Raymond KN, Weitl FL (1981) Ferric ion sequestering agents. 6. Thespectrophotometric and potentiometric evaluation of sulfonated tricatecholate li-gands. J Am Chem Soc 103(10):2667–2675.

37. Murakami Y, Nakamura K (1963) The chelating behavior of catechol-4-sulfonate withiron (III) ion. Bull Chem Soc Jpn 36(11):1408–1411.

38. Karpishin TB, Stack TDP, Raymond KN (1993) Stereoselectivity in chiral iron(III) andgallium(III) tris(catecholate) complexes effected by nonbonded weakly polar inter-actions. J Am Chem Soc 115(14):6115–6125.

39. Liu M, Zhang L, Wang T (2015) Supramolecular chirality in self-assembled systems.Chem Rev 115(15):7304–7397.

40. Enemark EJ, Stack TDP (1995) Synthesis and structural characterization of a stereo-specific dinuclear gallium triple helix: Use of the trans-influence in metal-assisted self-assembly. Angew Chem Int Ed Engl 34(9):996–998.

41. Albrecht M, et al. (2011) “Induced fit” in chiral recognition: Epimerization upon di-merization in the hierarchical self-assembly of helicate-type titanium(IV) complexes.Angew Chem Int Ed Engl 50(12):2850–2853.

42. Telford JR, Leary JA, Tunstad LMG, Byers BR, Raymond KN (1994) Amonabactin: Characteriza-tion of a series of siderophores fromAeromonas hydrophila. J AmChemSoc 116(10):4499–4500.

43. Telford JR, Raymond KN (1998) Coordination chemistry of the amonabactins, bis(cat-echolate) siderophores from Aeromonas hydrophila. Inorg Chem 37(18):4578–4583.

44. Abergel RJ, Zawadzka AM, Hoette TM, Raymond KN (2009) Enzymatic hydrolysis oftrilactone siderophores: Where chiral recognition occurs in enterobactin and bacilli-bactin iron transport. J Am Chem Soc 131(35):12682–12692.

45. Sprencel C, et al. (2000) Binding of ferric enterobactin by the Escherichia coli peri-plasmic protein FepB. J Bacteriol 182(19):5359–5364.

46. Miethke M, Skerra A (2010) Neutrophil gelatinase-associated lipocalin expresses an-timicrobial activity by interfering with L-norepinephrine-mediated bacterial iron ac-quisition. Antimicrob Agents Chemother 54(4):1580–1589.

47. Sowinski JA, Toogood PL (1996) Synthesis of an enantiomerically pure serine-derivedthiazole. J Org Chem 61(22):7671–7676.

48. Yu X, Dai Y, Yang T, Gagné MR, Gong H (2011) Facile synthesis of salmochelin S1, S2,MGE, DGE, and TGE. Tetrahedron 67(1):144–151.

49. Schobert R, Stangl A, Hannemann K (2006) Conjugates of methyl 6-amino-penicillanate with biscatechol-hydroxamate chelators: Synthesis and siderophoricactivity. Tetrahedron 62(33):7799–7808.

50. Igarashi Y, et al. (2012) Catechoserine, a new catecholate-type inhibitor of tumor cellinvasion from Streptomyces sp. J Antibiot (Tokyo) 65(4):207–209.

51. Winn MD, et al. (2011) Overview of the CCP4 suite and current developments. ActaCrystallogr D Biol Crystallogr 67(Pt 4):235–242.

52. Winter G (2010) Xia2: An expert system for macromolecular crystallography datareduction. J Appl Cryst 43(1):186–190.

53. Kabsch W (2010) XDS. Acta Crystallogr D Biol Crystallogr 66(Pt 2):125–132.54. Evans P (2006) Scaling and assessment of data quality. Acta Crystallogr D Biol

Crystallogr 62(Pt 1):72–82.55. Murshudov GN, et al. (2011) REFMAC5 for the refinement of macromolecular crystal

structures. Acta Crystallogr D Biol Crystallogr 67(Pt 4):355–367.56. Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Features and development of Coot.

Acta Crystallogr D Biol Crystallogr 66(Pt 4):486–501.57. Kuzmic P (1996) Program DYNAFIT for the analysis of enzyme kinetic data: Applica-

tion to HIV proteinase. Anal Biochem 237(2):260–273.

Raines et al. PNAS | May 24, 2016 | vol. 113 | no. 21 | 5855

CHEM

ISTR

YBIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Janu

ary

24, 2

021