Genomics-driven discovery of chiral triscatechol siderophores with enantiomeric Fe(iii) coordination

Ferric complexes of triscatechol siderophores may assume one of two enantiomeric configurations at the iron site. Chirality is known to be important in the iron uptake process, however an understanding of the molecular features directing stereospecific coordination remains ambiguous. Synthesis of the full suite of (DHBL/DLysL/DSer)3 macrolactone diastereomers, which includes the siderophore cyclic trichrysobactin (CTC), enables the effects that the chirality of Lys and Ser residues exert on the configuration of the Fe(iii) complex to be defined. Computationally optimized geometries indicate that the Λ/Δ configurational preferences are set by steric interactions between the Lys sidechains and the peptide backbone. The ability of each (DHBL/DLysL/DSer)3 diastereomer to form a stable Fe(iii) complex prompted a genomic search for biosynthetic gene clusters (BGCs) encoding the synthesis of these diastereomers in microbes. The genome of the plant pathogen Dickeya chrysanthemi EC16 was sequenced and the genes responsible for the biosynthesis of CTC were identified. A related but distinct BGC was identified in the genome of the opportunistic pathogen Yersinia frederiksenii ATCC 33641; isolation of the siderophore from Y. frederiksenii ATCC 33641, named frederiksenibactin (FSB), revealed the triscatechol oligoester, linear-(DHBLLysLSer)3. Circular dichroism (CD) spectroscopy establishes that Fe(iii)–CTC and Fe(iii)–FSB are formed in opposite enantiomeric configuration, consistent with the results of the ferric complexes of the cyclic (DHBL/DLysL/DSer)3 diastereomers.


Introduction
Chirality is universally signicant in biological reactions, including the essential microbial process of iron acquisition mediated by small-molecule chelators known as siderophores. The specic three-dimensional structure of an Fe(III)-siderophore complex plays a role in the ability of a bacterium to recognize, acquire, and extract iron from it. 1,2 The triscatechol siderophores enterobactin (Ent) and bacillibactin (BB) each coordinate Fe(III) with three 2,3dihydroxybenzoyl (DHB) ligands framed on a macrolactone derived from three L Ser or L Thr residues, respectively (ESI Fig. S1 †). Unlike Ent, BB also contains a glycine residue inserted between the macrolactone core and DHB. In stark contrast to Fe(III)-Ent 3À , which forms exclusively in the D conguration, 3,4 Fe(III)-BB 3À adopts the opposing L conguration. 3 Several related triscatechol siderophores are further distinguished from Ent and BB by the presence of a chiral amino acid inserted between DHB and the oligoester backbone, including cyclic trichrysobactin [CTC; Dickeya chrysanthemi EC16] with D Lys, 5 as well as the linear tris-L Ser scaffolds of trivanchrobactin with D Arg, 6 and turnerbactin with L Orn. 7 Structurally, the inuence amino acids exert on the conguration at the Fe(III) site is incompletely understood; evolutionarily, these structural differences hint that chirality confers a competitive advantage in microbial iron uptake.
To understand the factors controlling stereospecic Fe(III) coordination of the expanded triscatechol-triserine siderophore CTC, (DHB D Lys L Ser) 3 , we synthesized the full suite of cyclic (DHB L/D Lys L/D Ser) 3 diastereomers of CTC (Fig. 1). We report that circular dichroism (CD) spectroscopic measurements of the Fe(III) complexes of these ligands allow the relationship between siderophore chirality and the conguration at the Fe(III) site to be dened. Computational modeling of the ferric complexes reveals steric interactions between the Lys sidechains and the peptide backbone dictate the congurational preference. Fe(III) complexation by each (DHB L/D Lys L/ a linear triserine backbone and chelating DHB units. The genomics data, in combination with our spectroscopic investigation of isolated FSB, establish that it is a natural diastereomer of linear trichrysobactin. Circular dichroism (CD) spectra reveal that FSB and CTC coordinate Fe(III) in opposing enantiomeric conguration, consistent with the setting of the handedness of Fe(III) coordination by the stereochemistry at Lys.

Results and discussion
Chirality of Fe(III)-[(DHB L/D Lys L/D Ser) 3

] diastereomers
Chiral triscatechol siderophores and synthetic analogs are capable of coordinating labile metal ions with a thermodynamic preference for a specic stereochemistry at the metal center. 8 The presence of chirality at both the metal center and in the ligand renders the D and L stereoisomers diastereomeric and energetically inequivalent. To establish the relationship between the chirality at the amino acid adjacent to the catecholamide, the stereochemistry of the triserine macrolactone, and the stereochemistry at Fe(III), we synthesized the four C 3 -symmetric cyclic diastereomers (DHB L Lys L Ser) 3 , (DHB D Lys L Ser) 3 , (DHB L Lys D Ser) 3 , and (DHB D Lys D Ser) 3 (Scheme S1 †) of which (DHB D Lys L Ser) 3 is structurally identical to CTC. Well-established methodology to construct the cyclic triserine macrolactone (1 in Scheme S1 †), 9 provided a convenient synthetic platform to access CTC and related diastereomers ( Fig. S2-S6, Tables S1 and S2 †). In the absence of crystallographic information, circular dichroism (CD) spectroscopy can provide information on the stereochemical conguration of optically-active metal complexes. 10 As expected, enantiomeric pairs of ligands, such as (DHB L Lys L Ser) 3 and (DHB D Lys D Ser) 3 , coordinate Fe(III) with opposite handedness, as indicated by the CD spectra (Fig. 2). Two prominent CD bands at 435 nm and 545 nm arise from characteristic LMCT transitions and are therefore sensitive to the chirality at the iron center. When comparing diastereomeric ligands, we observed that the CD spectra of Fe(III)-[(DHB L Lys L Ser) 3 Table 1). ‡ Comparison of the signs of the Cotton effects for the Fe(III)-[(DHB L/D Lys L/D Ser) 3 ] complexes with those of Fe(III)-Ent 3À and Fe(III)-BB 3À , for which the chirality at the metal center is known, allows the conguration of the new complexes to be determined. (DHB L Lys L Ser) 3 and (DHB L Lys D Ser) 3 both form D complexes, whereas (DHB D Lys D Ser) 3 and (DHB D Lys L Ser) 3 both form L complexes. These results suggest that the handedness of metal-ion chelation is set by the chirality of the Lys unit, and not the chirality of the triserine macrolactone. In comparison, the D conguration of Fe(III)-Ent 3À and L conguration of Fe(III)-enantioEnt 3À has been attributed to nonbonding interactions within the chiral triserine macrolactone. [11][12][13] Computational modeling To better understand the mechanism by which amino acid chirality dictates the congurational preferences of the Fe(III)-[(DHB L/D Lys L/D Ser) 3 ] complexes, the structures and energies of the four enantiomeric pairs of diastereomers were optimized computationally (PBE0/6-311++G(d,p)) ( Fig. 3).
Comparing the energies of the Fe(III) complexes of a given ligand with different handedness, we observe complete agreement with the CD spectroscopic results. For example, the energy of L-Fe(III)-[(DHB D Lys L Ser) 3 ] is lower than that of D-Fe(III)-[(DHB D Lys L Ser) 3 ], consistent with the formation of the L complex in aqueous solution (Fig. 2).
Insight into the origin of the differential stabilities of the Fe(III)-[(DHB L/D Lys L/D Ser) 3 ] complexes comes directly from the   optimized geometries (Fig. 4). In this molecular framework, the Lys sidechains are able to wrap around the complex so as to allow each terminal ammonium group to hydrogen-bond with the carbonyl of the DHB unit of an adjacent arm. We observe, however, that this interaction is present in all of the optimized geometries, preferred and non-preferred. Closer analysis revealed that the prime inuence of the Lys residue chirality is the impact that it has on j (N-C carbonyl -C a -N torsion angle). It is well established that certain values of j are unfavorable for polypeptides, contributing, for example, to the characteristic distribution of protein dihedral angles in Ramachandran plots. Specically, favorable j values are those that prevent the amino acid side chain from eclipsing the adjacent carbonyl. 14 In the Fe(III)-[(DHB L/D Lys L/D Ser) 3 ] complexes, combination of either D conguration at Fe(III) and D Lys, or L and L Lys, produce j angles near AE60 , which introduces a steric clash between the carbonyl O atom and the Lys side chain (Fig. 5). In contrast, the combination of L and D Lys, as occurs in CTC, produces no such clash.
It is noteworthy that, of all the diastereomeric combinations of metal chelation handedness and amino acid chirality, our calculations predict that the most stable structures are those assumed by D-Fe(III)-[(DHB L Lys D Ser) 3 ] and its enantiomer L-Fe(III)-[(DHB D Lys L Ser) 3 ]. Organisms have adopted this stability by using (DHB D Lys L Ser) 3 , which is the siderophore CTC, for iron acquisition. The fact that the other diastereomers that we investigated also form Fe(III) complexes gives rise to the question of whether they too might be used biologically.

Genomic screen for catechol-based siderophores
Inspired by the discovery of other naturally occurring siderophores with D-and L-amino acidsthat is, trivanchrobactin ( D Arg), and turnerbactin ( L Orn)we initiated a search for biosynthetic gene clusters (BGCs) encoding diastereomers of CTC. The biosynthesis of chrysobactin (i.e., DHB D Lys L Ser) in D. dadantii 3937 requires genes encoding 2,3-DHB synthesis, as well as the non-ribosomal peptide synthetase (NRPS) CbsF with an epimerization, E, domain to convert L Lys to D Lys. 15,16 In contrast to D. dadantii 3937, the plant pathogen D. chrysanthemi EC16 produces not only the monocatechol chrysobactin, but also the triscatechol macrolactone CTC. We found that the genome of the D. chrysanthemi EC16 contains a BGC homologous to the cbs locus of D. dadantii 3937 (genome sequence reported herein; Tables S3 and S4 †). Genome mining revealed similar but distinct BGCs in several Yersinia genomes, including the BGC freABCEF of opportunistic pathogen Yersinia frederiksenii ATCC 33641 (Tables S5 and S6 †). The fre locus contains genes encoding 2,3-DHB synthesis, as well as the NRPS FreF with adenylation domains selecting for L Lys and L Ser. However, FreF lacks an E domain, implicating biosynthesis of a siderophore comprised of DHB L Lys L Ser units (Fig. 6).

Frederiksenibactin and cyclic trichrysobactin siderophores
Siderophores from Y. frederiksenii ATCC 33641 were extracted and puried from a low-iron culture (  3 , respectively. 5 In contrast to D. chrysanthemi EC16, which produces trichrysobactin in both cyclic and linear forms, 5 we have only been able   in the culture supernatant of Y. frederiksenii ATCC 33641. We have named this new siderophore frederiksenibactin (FSB). We note that the related triscatechol siderophores trivanchrobactin and turnerbactin are also linear and that their cyclic forms have not been detected in biological systems. 6,7 Marfey's analysis 17 establishes the presence of L Lys and L Ser in FSB, consistent with the genomic prediction (Fig. S9 †). The proposed structure of FSB was conrmed by 1 H and 13 C NMR spectroscopic data, which were assigned through 1 H-1 H COSY, 1 H-13 C HSQC, and 1 H-13 C HMBC NMR data (Fig. S10-S14 †).
While the NMR spectral data of FSB (Table S7 †) are similar to those of CTC, several features conrm the mass spectrometric results indicating that FSB is a linear compound. Specically, the three Ser residues are inequivalent (Fig. S11 †). The Ser methylene protons involved in the backbone ester linkages, C16/C16 0 , at 4.25-4.46 ppm are shied signicantly downeld relative to the corresponding protons on C16 00 at 3.67 ppm and 3.78 ppm, which are adjacent to the unmodied Ser hydroxyl group. Additionally, the protons on the three methine carbons (C15, 4.59 ppm; C15 0 , 4.69 ppm; C15 00 , 4.41 ppm) are inequivalent, as are the protons on the chiral methine carbons derived from Lys (C9, C9 0 and C9 00 , 4.50-4.65 ppm). The 1 H NMR spectrum of FSB is consistent with related asymmetric linear triscatechol siderophores trivanchrobactin 6 and turnerbactin. 7 Thus, FSB is a novel siderophore and a natural diastereomer of linear trichrysobactin.  Table 1) also establishes that linearization of the trilactone does not signicantly affect the conguration of the ferric complex. Earlier work revealed that linearization of Ent also does not invert its overall congurational preference, however, a small fraction of the L enantiomer is formed. 11,13 Our earlier work with the cyclic Fe(III)-[(DHB L/ D Lys L/D Ser) 3 ] complexes suggests that the opposing chirality observed for ferric complexes of FSB and CTC is likely due to the stereochemistry of the Lys residue adjacent to the catecholamide and not due to the linear or cyclic nature of the triserine backbone.

Fe(III) exchange between FSB and CTC
Surprisingly little is known about the exchange of Fe(III) among triscatechol siderophores. CD spectroscopy is uniquely poised to monitor Fe(III) exchange between optically-active siderophores. The intensity of the L-Fe(III)-CTC CD bands decrease upon addition of equimolar FSB as a result of formation of nearly equimolar L-Fe(III)-CTC and D-Fe(III)-FSB (Fig. 8A). Moreover the equivalent equilibration approached from reaction of D-Fe(III)-FSB with CTC is also observed (Fig. 8B). Interestingly, a weak negative band at 435 nm and a weak positive   band at 550 nm are formed aer four hours of equilibration (Fig. 8B), suggestive of a slight preference favoring formation of Fe(III)-CTC over Fe(III)-FSB, consistent with the increased stability constant of macrocyclic ligands. 13 The intensity of this band diminishes upon further equilibration, potentially due to hydrolysis of the labile macrolactone. Under neutral pH conditions with 100 mM Fe(III)-CTC and 100 mM FSB, the magnitude of the CD signal decreases within hours of mixing, indicating that exchange occurs on a relatively short, biologically relevant time scale.

Conclusions
In sum, BGCs encoding synthesis of the triscatechol siderophores CTC and FSB were identied and the structure of The suite of cyclic and linear (DHB L/D Lys L/D Ser) 3 siderophores and analogs raises further signicant questions regarding the effect of a mismatched Dand L-Fe(III) conguration on microbial iron uptake and growth. For example, discrimination at the outer membrane receptor protein could prevent uptake of the wrong Fe(III)-enantiomer as has been observed with Fe(III) complexes of pyochelin and enantiopyochelin. [18][19][20] If iron uptake is insensitive to the Fe(III)-enantiomer chirality, discrimination could still occur at other points including the iron-release process, as is observed in Bacillus subtilis in the Fes-catalyzed hydrolysis of the macrolactone of Fe(enantioEnt) 3À which is required for release of iron. 1 Additionally, it may be possible for the relevant siderophoreinteracting proteins to invert the conguration of a mismatched Fe(III)-siderophore complex upon binding, as has been observed for the periplasmic binding protein CeuE of Campylobacter jejuni. [21][22][23] Siderophores are primarily extracellular metabolites and facile Fe(III) exchange observed between triscatechol siderophores is likely of biological consequence within complex microbial communities. Certainly, the rate of release of the newly synthesized apo siderophores during growth of Y. frederiksenii ATCC 33641 and D. chrysanthemi EC16, which is occurring over the time scale of hours to days, could be exchanging Fe(III) within hours with other triscatechol siderophores, as evinced by the CD results (Fig. 8). In fact Fe(III) exchange between the triscatecholate siderophores is orders of magnitude faster than Fe(III) exchange between hydroxamate siderophores or between hydroxamate and catecholate siderophores. 24 Experiments addressing the questions raised above are in progress, as well as the question of whether BGCs encoding the synthesis of the diastereomers of trivanchrobactin, (DHB D/ L Arg L/D Ser) 3 , and turnerbactin, (DHB L/D Orn L/D Ser) 3 , are present in microbial genomes. The discovery of frederiksenibactin and its relationship to CTC exemplies the structural variability of microbial siderophores and provides a natural system to determine the signicance of chirality within siderophoremediated microbial iron-uptake pathways.

Experimental
General experimental procedures UV-visible absorbance and circular dichroism spectroscopy were measured on an Agilent Cary 300 UV Vis spectrophotometer and a Jasco J-1500 CD spectrophotometer, respectively. 13 C NMR spectroscopy was performed on a Bruker Advanced Neo 500 MHz spectrometer equipped with a prodigy cryoprobe at RT. All 1 H, COSY, HMBC, HSQC NMR spectroscopy was performed on a Varian Unity 600 MHz spectrometer at RT. Chemical shis were referenced through residual solvent peaks [ 1 H (DMSO-d 6 ) 2.50 ppm, 13 C (DMSO-d 6 ) 39.51 ppm]. Mass spectrometry analysis of Y. frederiksenii ATCC 33641 supernatant extracts and puried FSB was carried out on a Waters Xevo G2-XS QToF with positive mode electrospray ionization coupled to an ACQUITY UPLC H-Class system with a waters BEH C18 column. Y. frederiksenii ATCC 33641 culture extracts were analyzed using a linear gradient of 0-30% CH 3 CN (+0.1% formic acid) in ddH 2 O (+0.1% formic acid) over 10 min. For MS/MS analysis, a collision energy of 15 eV was employed. HR-ESIMS analysis of synthetic compounds was carried out on a Waters LCT Premier ESI TOF introduced into the ESI by direct infusion via a syringe pump.

General synthetic procedures
All reactions performed under an argon atmosphere were carried out using a high-vacuum line, standard Schlenk techniques, and dry solvents. DMF, DCM, and DMSO-d 6 were stored over 3Å molecular sieves for at least 72 h prior to use. N,N 0 -Diisopropylethylamine (DIPEA) was puried by distillation over ninhydrin (Â3) and was subsequently stored over 3Å molecular sieves. N a -Boc-N 3 -Cbz-L-lysine and N a -Boc-N 3 -Cbz-D-lysine were acquired from Bachem. All other reagents (including those used for Marfey's analysis) were purchased from Sigma-Aldrich.

Synthesis of the cyclic (DHB L/D Lys L/D Ser) 3 diastereomers
Established peptide coupling methodology 25 was employed to construct the two key amide bonds in 4 (Scheme S1 †). Reaction of chiral triamine 1 with HATU (3 eq.), Boc-Lys(Z)-OH (3 eq.), and DIPEA (9 eq.) cleanly affords intermediate 2 ( Step a, Scheme S1 †). Removal of the N a -Boc protecting groups (step b, Scheme S1 †) and subsequent coupling to benzyl-protected 2,3-dihydroxybenzoic acid (step c, Scheme S1 †) yields 3 in an 81% yield over two steps. Global deprotection by hydrogenolysis over 10% Pd/C (step d, Scheme S1 †) yields (DHB L/D Lys L/D Ser) 3 , (4) as an enantiopure product. Initial synthetic efforts in which the direction of peptide coupling was reversed were highly susceptible to epimerization at the Lys stereocenter, consistent with the observed chiral instability of N a -acylated amino acids upon activation as a HOBT or HOAT ester. 26 N a -Boc-N 3 -Cbz-L-lysine was substituted for N a -Boc-N 3 -Cbz-D-lysine in the synthesis of (DHB D Lys L Ser) 3 and (DHB D Lys D Ser) 3 . N-Trityl-L-serine was substituted for N-trityl-D-serine in the synthesis of 1 to yield (DHB L Lys D Ser) 3 and (DHB D Lys D Ser) 3 .
Synthesis of N,N 0 ,N 00 -tris[N a -Boc-N 3 -Cbz-L-lysinyl]cyclotri-Lseryl trilactone, 2. N a -Boc-N 3 -Cbz-L-lysine (502 mg, 1.32 mmol) was dissolved in 10 mL of dry DMF under an argon atmosphere and cooled in an ice bath. HATU (502 mg, 1.32 mmol) and DIPEA (836 mL, 4.8 mmol) were added at 0 C and the ask was subsequently taken out of the ice bath and stirred for 3 min. Triserine trilactone hydrochloride (148.5 mg, 0.4 mmol), prepared according to literature procedure, 27 was added as a solid to the ask and the reaction was stirred overnight at RT. The solvent was removed in vacuo and the crude reaction mixture was brought up in DCM and rinsed quickly with 1 M HCl (30 mL, Â3) and brine (30 mL). The organic layer was concentrated and then loaded onto a silica column. Purication by ash chromatography using a gradient of 2-4% MeOH in DCM afforded 2 as a colorless solid. (76% yield). 1  Compound 2 (404.6 mg, 0.3 mmol) was added to a dry ask under argon and dissolved in 6 mL dry DCM. The ask was cooled in an ice bath and 4 mL of TFA were added. Aer stirring for 1.5 h at RT, full deprotection of the boc groups was observed by TLC. Volatiles were removed in vacuo and the pale yellow oil was brought up in 5 mL of dry DMF. In a separate ask, 2,3dibenzyloxybenzoic acid (341 mg, 0.99 mmol), HATU (376 mg, 0.99 mmol), and DIPEA (627 mL, 3.6 mmol) were added to 5 mL of dry DMF under an argon atmosphere and stirred for 3 min at RT. The contents of the rst ask were then transferred to the reaction mixture via syringe and the reaction was le to stir overnight at RT. The reaction mixture was concentrated, loaded onto a silica column, and then puried by ash chromatography using a gradient of 1-3% MeOH in DCM. Fractions were combined and concentrated to yield 3 as a white solid. (81% yield over 2 steps) 1 65.1, 70.3, 75.1, 116.5, 121.4, 124.2, 127.7, 127.9,  128.0, 128.1, 128.3, 128.4, 128.9, 136.6, 136.7, 137.2, 145.6,  151.6, 156.0, 162.3, 164.9, 169.3, 171.8  N,N 0 ,N 00 -Tris[2,3-dihydroxybenzoyl-L-lysinyl]cyclotri-L-seryl trilactone (DHB L Lys L Ser) 3 , 4. Compound 3 (399.5 mg, 0.2 mmol) was dissolved in 10 mL of 60% THF (aq.) + 0.5% acetic acid under an atmosphere of argon. 10% Pd/C (100 mg) was carefully added, and a balloon of hydrogen attached to a threeway ushing adapter was tted to the round bottom. The atmosphere was evacuated and back-lled with hydrogen four times and stirred under an atmosphere of hydrogen for 24 h at RT. The catalyst was then ltered off, rinsed with 25 mL of DMF, and concentrated to yield a dark-red oil. The crude reaction, deemed mostly pure by NMR, was further puried by semipreparative HPLC on a YMC-Actus 20 Â 250 mm C18 ODS-AQ column using a linear gradient of 15% MeOH in ddH 2 O (+0.1% triuoroacetic acid) to 40% MeOH in ddH 2 O (+0.1% triuoroacetic acid) over 25 min. HPLC fractions were concentrated and subsequently lyophilized to yield 4 as a white solid. (65% yield) 1   and NaHCO 3 (1 M, 20 mL) were added and the solution was briey vortexed and placed on a heating block (40 C) for 1 h. 10 mL of 2 M HCl was then added to quench the reaction and solutions were stored at À20 C in the dark prior to analysis. Amino acid standards were derivatized according to the same procedure. Derivatized hydrolysis products of FSB were separated by HPLC on a YMC 4.6 Â 250 mm C18-AQ column using a gradient from 10% CH 3 CN in ddH 2 O (0.05% triuoroacetic acid) to 40% CH 3 CN in ddH 2 O (0.05% triuoroacetic acid) over 60 min. Derivatized hydrolysis products of (DHB L/D Lys L/D Ser) 3 were separated by HPLC on a YMC 4.6 Â 250 mm C18-A column using a gradient from 10% CH 3 CN in TEAP buffer (50 mM, pH 3.00) to 40% CH 3 CN in TEAP buffer over 60 min. Derivatized hydrolysis products were co-injected with derivatized amino acid standards to determine the constituent amino acids of FSB and to determine the extent of epimerization during synthesis of synthetic (DHB L/D Lys L/D Ser) 3 . Three peaks corresponding to FDAA-derivatized lysine were observed, corresponding to products derivatized at either the a-amine, 3-amine, or both amines.
FDAA-derivatized D Ser co-eluted with L Lys and D Lys derivatized at the 3-amine under the conditions used for Marfey's analysis of FSB (YMC C18-AQ column).

Preparation of Fe(III)-complexes and circular dichroism spectroscopy
Fe(III)-complexes of the (DHB L/D Lys L/D Ser) 3 diastereomers and FSB for CD spectroscopy were prepared in citrate-phosphate buffer (50 mM, pH 7.40) by mixing a solution of FeCl 3 [2.45 mM, 0.1 M HCl (aq)] with 1.0 equivalent of the desired apo-ligand. Formation of the Fe(III)-complex was tracked by UV-visible spectroscopy by observing the absorbance at 498 nm. The resulting solution was equilibrated for 30 min in the dark prior to analysis by CD spectroscopy.
Full CD spectra were acquired using the following parameters: 4 s D.I.T., 1 nm bandwidth, 50 nm s À1 scanning speed, with 3 accumulations. Fe(III) exchange assays were performed by preparing pre-equilibrated Fe(III)-complexes of either FSB or CTC as described above. At time t ¼ 0, an equimolar amount of the opposing apo-ligand was added to the Fe(III)-complex and the resulting solution was gently vortexed. CD spectra were acquired as a single accumulation at 20 min intervals using the following parameters: 400-600 nm; 2 s D.I.T., 1 nm bandwidth, and 100 nm s À1 scan speed.

Computational modeling
Electronic structure calculations were performed using Gaussian 16. 28 The structures of the following four complexes  3 ]. Note that the structures of the corresponding L isomers were not optimized, because each is an enantiomer of one of the four D complexes listed above, and therefore energetically equivalent. The input geometries were generated manually. The Fe(III) centers were treated as high-spin (S ¼ 5/2) and the Lys residues were protonated to afford neutral complexes (z ¼ 0). Optimizations were performed at the PBE0/6-311++G(d,p) level of theory with Grimme's D3 empirical dispersion correction and tight convergence criteria. [29][30][31][32] Implicit aqueous solvation was included using a conductorlike polarizable continuum model (CPCM). The energy values presented in Fig. 3  Dickeya chrysanthemi EC16 (ATCC 11662) was obtained from the ATCC and maintained on Difco Luria-Bertani (LB) agar plates at 30 C. A liquid LB culture was inoculated from a single colony and incubated for 18 h at 30 C and 180 rpm. Genomic DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen) according to the manufacturer's instructions for Gram-negative bacteria. Extracted DNA was quantied by a Qubit 2.0 uorometer (Invitrogen). Library preparation and sequencing were performed by the Microbial Genome Sequencing Center (Pittsburgh, PA): paired-end libraries were prepared according to Baym et al. 34 and sequenced on the NextSeq 550 platform (Illumina), generating 4 232 664 pairs of 2 Â 150 bp reads.

Bacterial growth and siderophore isolation
Yersinia frederiksenii ATCC 33641, obtained from the American Type Culture Collection (ATCC), was cultured on Difco Luria Bertani (LB) Miller (BD biosciences) medium plates. A single colony of Y. frederiksenii ATCC 33641 was inoculated into 50 mL of Difco LB Miller (BD biosciences) media and grown overnight at 30 C, shaking at 180 rpm. A portion of the overnight culture (5 mL) was then inoculated into low-iron minimal media (2 L, pH 7.0) containing sodium succinate (4 g L À1 ), K 2 HPO 4 (6 g L À1 ), KH 2 PO 4 (3 g L À1 ), NH 4 Cl (1 g L À1 ), CaCl 2 $2H 2 O (20 mg L À1 ), and MgSO 4 $7H 2 O (200 mg L À1 ) in an acid-washed 4 L Erlenmeyer ask. The culture was shaken at RT, 180 rpm for 72 h. Cultures were harvested in the late log phase of growth by centrifugation (SLA-3000 rotor, ThermoScientic) at 6000 rpm for 30 min at 4 C. Culture supernatants were decanted into a clean, acid-washed Erlenmeyer ask containing 100 g of Amberlite XAD-4 polystyrene resin, which was shaken at 120 rpm for 4 h at 4 C. The resin was ltered from the supernatant, rinsed with 100 mL of 90/10% ddH 2 O/MeOH, and then eluted with 250 mL of 95 : 5% MeOH/ddH 2 O. The eluent was concentrated under reduced pressure to a volume of 30 mL and stored at 4 C prior to analysis. Frederiksenibactin and the related monocatechol and dicatechol compounds were puried by semi-preparative RP-HPLC on a YMC-Actus 20 Â 250 mm C18 ODS-AQ column using a linear gradient of 15% MeOH in ddH 2 O (+0.1% triuoroacetic acid) to 40% MeOH in ddH 2 O (+0.1% triuoroacetic acid) over 25 min.

Data availability
The dra genome sequence of Dickeya chrysanthemi EC16 was deposited at NCBI under the BioProject ID PRJNA690813.