How Many O-Donor Groups in Enterobactin Does It Take to Bind a Metal Cation?

Article abstract of DOI:10.1002/chem.201900453

The E. coli siderophore enterobactin, the strongest Fe^III chelator known to date, forms hexacoordinate complexes with Si^IV, Ge^IV, and Ti^IV. Synthetic protocols have been developed to prepare non-symmetric entero

Full text of DOI:10.1002/chem.201900453

A Journal of
Accepted Article
Title: How many O-donors in enterobactin does it take to bind a metal?
Authors: Todor Baramov, Bianca Schmid, Ho Ryu, Jinhoon Jeong,
Leonard von Eckardstein, Mu-Hyun Baik, and Roderich
Süssmuth
This manuscript has been accepted after peer review and appears as an
Accepted Article online prior to editing, proofing, and formal publication
of the final Version of Record (VoR). This work is currently citable by
using the Digital Object Identifier (DOI) given below. The VoR will be
published online in Early View as soon as possible and may be different
to this Accepted Article as a result of editing. Readers should obtain
the VoR from the journal website shown below when it is published
to ensure accuracy of information. The authors are responsible for the
content of this Accepted Article.
To be cited as: Chem. Eur. J. 10.1002/chem.201900453
Link to VoR: http://dx.doi.org/10.1002/chem.201900453
Supported by

10.1002/chem.201900453
Chemistry - A European Journal
How many O-donors in enterobactin does it take to bind a metal?
Todor Baramov,[a] Bianca Schmid,[a] Ho Ryu,[b] Jinhoon Jeong,[b] Karlijn Keijzer, Leonard von Eckardstein,[a]
Mu-Hyun Baik,*[b] and Roderich D. Süssmuth*[a]
Abstract
The E. coli siderophore enterobactin, the strongest FeIII–chelator known to date, was found to form
hexacoordinate complexes with SiIV, GeIV, and TiIV. Synthetic protocols were developed to prepare non-symmetric
enterobactin analogs with varying denticities. Various benzoic acid residues were coupled to the macrocyclic lactone
to afford a diverse library of ligands. These enterobactin analogs were bound to SiIV, GeIV, and TiIV, and investigated
using experimental and computational techniques. The binding behavior of the synthesized chelators enabled the
assessment of the contribution of each of the phenolic hydroxy groups in enterobactin to metal complexation. It was
found that at least four O-donors are needed for enterobactin derivatives to act as metal binders. DFT-calculations
indicate that the strong binding behavior of enterobactin is the results of diminished translational entropy penalty, a
common feature of the chelate effect, and the structural arrangement of the three catechols that allows the triseryl
base to be installed without distorting the preferred local metal binding geometry of the catecholates.
[a] Dr. T. Baramov, B. Schmid, Dr. L. von Eckardstein, Prof. Dr. R. D. Süssmuth
Institut für Chemie
Technische Universität Berlin, Fakultät II, Strasse des 17. Juni 124, 10623 Berlin, (Germany)
E-mail: suessmuth@chem.tu-berlin.de
[b] H. Ryu, J. Jeong, Prof. Dr. M.-H. Baik
Department of Chemistry
Korea Advanced Institute of Science and Technology (KAIST), Daejeon 34141, (Republic of Korea)
Center for Catalytic Hydrocarbon Functionalizations, Institute for Basic Science (IBS), Daejeon 34141, (Republic
of Korea)
E-mail: mbaik2805@kaist.ac.kr
1
This article is protected by copyright. All rights reserved.

10.1002/chem.201900453
Chemistry - A European Journal
Introduction
Siderophores are small biomolecules used by microorganisms to extract iron from their environment.[1]
A
prominent example is enterobactin (Ent), which is produced by gram-negative enteric bacteria such as E. coli or
Salmonella. Siderophores are the strongest FeIII–chelating molecules known to date with FeIII–Ent, for example,
having an estimated binding constant of 1049 M–1.[2] Under standard conditions, this binding constant translates to a
free energy of approximately –67 kcal/mol. Motivated by the curiosity about these extremely high binding constant,[3]
the utility in metal chelation therapy,[4] and to better understand the biological processes related to bacterial iron
uptake,[5] considerable efforts were made to understand the biosynthesis, transport, and recognition processes
mediated by these molecules.[6] To derive an unified concept for how siderophores bind metals in general, a
fundamental understanding of both, the energetics and kinetics of the iron binding mechanism is needed.
Unfortunately, the extreme binding constant makes experiments that rely on the assessment of equilibrium states
impossible and monitoring the evolution of intermediates during iron uptake proved difficult. In addition, the FeIII-d5
center can adopt many different electronic states that are notoriously difficult to understand.
One possible way of studying the metal binding behavior is to offer metals other than iron to Ent and to
study its interaction with these mismatched cases. A desirable study that has never been pursued systematically is
to vary the molecular composition of Ent, for example by sequentially decreasing the ligand denticity and studying
the differential binding behavior. These studies may allow for establishing the general metal binding characteristics
of Ent to provide a conceptual foundation with which to examine iron binding in more detail. We prepared a diverse
library of Ent analogs that serve as models for the ferric complexation and the mechanism of iron release in the
microbial cell. Two different binding modes known as salicylate and catecholate type chelation have been suggested
for the FeIII–Ent complex, where the degree of protonation determines the binding mode that is adopted by
enterobactin (Scheme 1).[7] The chemical behavior of these newly prepared enterobactin derivatives towards forming
complexes with SiIV, GeIV, and TiIV was investigated by HR-ESI-MS and NMR spectroscopy. Unlike the natural target
FeIII, these ions do not give rise to complicated electronic states upon binding to Ent.
Recently, we discovered that Ent can bind to silicon at physiological pH to form a hexacoordinate tris-
catecholato complex of silicon 2a.[8] In subsequent studies we prepared the germanium and titanium analogs 2b
and 2c (Scheme 1) and characterized the complexes by single crystal X-ray diffraction analysis.[9] Neutral and
negatively charged hexacoordinate SiIV–complexes have been previously studied and catecholato complexes of SiIV
in penta- and tetracoordinate geometries are known.[10] Analogous examples of penta- and hexaoxocoordinate
GeIV–complexes were reported.[11] Interestingly, TiIV also forms hexacoordinate and pentacoordinate catecholato
complexes.[12]
Scheme 1. Structures of (a) enterobactin 1 and its hexacoordinate SiIV–, GeIV–, and TiIV–complexes in (b)
catecholate and (c) salicylate binding motif.
2
This article is protected by copyright. All rights reserved.

10.1002/chem.201900453
Chemistry - A European Journal
Results and discussion
The precise reason for what makes enterobactin bind iron so tightly is not known, but clearly the pseudo
octahedral environment that the three catecholate arms offer and the six O-donors that work in concert are key
elements. To assess how critical the presence of the six oxygen atoms are for metal binding, we prepared a series
of Ent-derivatives that contain different numbers of the hydroxy groups, thus affording a series of ligands that offer
5-, 4-, 3- and 2 coordination sites to the formation of a complex, which are expected to be less effective in binding
metals compared to Ent. Our long-term vision is that these Ent-mimics will allow for probing for partial aspects of
metal binding to Ent. In some of these Ent-derivatives, several regioisomers are possible, as illustrated in Scheme
2b. To precisely describe the binding motif, we represent the number and position of the phenolic hydroxy groups
in the label of the ligand, as denoted in Table 1. In compound 10, for example, there is one hydroxy group on each
of the three aromatic arms at the proximal position to the triseryl base, which is denoted by the label [10/10/10]. The
regioisomer 13, in which all three hydroxy groups are placed at the distal position, is labeled [01/01/01].
Scheme 2. Structures of (a) analogous complexes of 2a, 2b and 2c (n: number of O atoms) and (b) enterobactin
analogs missing hydroxy groups at different positions. For binding labels see also Table 1.
Enterobactin analogs 8–15 were prepared by coupling the corresponding benzoic acids 16–19 (Scheme 3)
with the triseryl macrolactone salt 20.[13] The symmetric enterobactin analog 10 required a reaction of acid 17 with
lactone 20 in presence of DIC/HOBt, followed by hydrogenolysis of the fully protected derivative 21, as shown in
Scheme 3a. The same synthetic approach did not work equally well for the enterobactin analog 13 [01/01/01]. After
some experimentation, we found that acid 18 requires much stronger activation and P(OMe)3/I2[14] was identified as
the best coupling reagent giving target compound 22 in 22% yield. Hydrogenolysis afforded the desired deoxy–
enterobactin 13 in high yield.
3
This article is protected by copyright. All rights reserved.

10.1002/chem.201900453
Chemistry - A European Journal
Table 1. Labels of the ligands with varying number of O-donors. Abbreviations: B: benzoic acid; 2-HB: 2-
hydroxybenzoic acid; 3-HB: 3-hydroxybenzoic acid; DHB: dihydroxybenzoic acid.
Compound
B
0[a]
0
0
0
0
0
0
1
2-HB
3-HB
DHB
Label
1
8
9
10
11
12
13
14
15
Ent
0
1
2
3
0
0
0
0
0
0
0
0
0
1
2
3
0
0
3
2
1
0
2
1
0
2
1
[10/11/11]
[10/10/11]
[10/10/10]
[01/11/11]
[01/01/11]
[01/01/01]
[00/11/11]
[00/00/11]
2
[a] Number of the corresponding benzoic acid moieties in the ligand.
Scheme 3. Synthesis of (a) symmetric and (b) non-symmetric enterobactin analogs: (i) for 21: 17, DIC, HOBt, Et3N,
CH2Cl2, RT; for 22: 18, P(OMe)3, I2, Et3N, CH2Cl2, RT; (ii) 10 % Pd/C, H2, EtOAc, EtOH, RT; (iii) P(OMe)3, I2, Et3N,
CH2Cl2, , RT; DIC = N,N’-diisopropylcarbodiimide, HOBt = 1-hydroxybenzotriazole. The synthetic yields of 23–28
ranges from 24 to 49 %. See more details in the Supporting information.
The synthesis of enterobactin analogs carrying a combination of catecholic, phenolic and phenylic arms to
present a non-symmetric binding site for the metals, as described above, is exceedingly difficult. It is for that reason
that only two such enterobactin analogs were reported to date and they were prepared using the hydrolytically stable
TREN platform.[7b] Because the serine macrolactone base of enterobactin is hydrolytically sensitive, our target
molecules pose additional challenges. The standard strategy of coupling 20 with benzoic acids and subsequently
protecting the remaining amine groups failed to produce the desired product and resulted in decomposition of the
4
This article is protected by copyright. All rights reserved.

10.1002/chem.201900453
Chemistry - A European Journal
macrolactone platform.[15] Similarly, the use of trityl protecting groups on a precursor of 20, envisioned to undergo
the coupling reaction with benzoic acid, was unsuccessful. A solution to this intrinsic synthetic challenge was found
by initially reacting the lactone salt 20 in P(OMe)3/I2-mediated coupling with a 1:1 mixture of the benzoic acids 16
and 17 or 18 to give a mixture of all four possible products. Benzylated enterobactin analogs were successfully
separated by reversed phase chromatography and deprotected by catalytic hydrogenation providing the
corresponding deoxy–enterobactins in high yields. This strategy was used for the synthesis of all desired non-
symmetric enterobactin analogs 23–28. Finally, the complex formation was carried out in two different solvent
systems. First, an aqueous solution (H2O/MeOH/acetone 9:0.5:0.5) containing ligands 8–15 in 0.15 mM
concentration was treated with 15 mM methanolic solutions of Si(OMe)4, Ge(OMe)4, or Ti(Oi-Pr)4, respectively, to
form the SiIV–, GeIV–, and TiIV–complexes 3–7. Second, the complexation was repeated in an organic solvent system
consisting of a 1:1 mixture of MeOH and acetone, to which the reagents were added neatly.
To develop a deeper understanding of the binding properties of the non-symmetric enterobactin analogs,
the free energy of formation for different SiIV–, GeIV–, and TiIV–complexes in DMSO were calculated using density
functional theory (DFT). Table 2 summarizes the results of the metal binding experiments and the DFT-computed
binding energies. Initially, we anticipated that the metal binding behavior will depend on the solvent and prepared
an aqueous and a non-aqueous environment for our experiments. Interestingly, we found that both solvents give
nearly identical results, i.e. systems that bind the metal ions in aqueous environment also show such behavior in
the organic solvent environment. The only exception is seen for ligand 14 [00/11/11], which binds TiIV in the organic
medium, but does not in the aqueous solution (vide infra). In general, the expected and observed trend is that the
overall binding strength is directly proportional to the number of O-binding sites that the ligand provides. As we
delete the O-binding site sequentially, the binding energy should decrease and after a critical number of O-binding
sites is reached, no binding should take place. And this of course the observable behavior overall.
Table 2. Binding properties of enterobactin 1 and its analogs 8–15 (a) detected by HR-ESI mass spectrometry
(negative and positive mode; experiments were carried out in MeOH/acetone 1:1). (b) DFT-calculated energies for
the chelation of SiIV, GeIV, and TiIV by different non-symmetric derivatives employing a dielectric constant, ε = 46.68
for DMSO. All energies are in kcal mol–1. The details about how binding energies were calculated are in the
Supporting information.
SiIV
(b) ΔG(sol)
GeIV
(b) ΔG(sol)
TiIV
(b) ΔG(sol)
Denticity
of ligand
Binding
mode[a]
Metal
coord. #
Entry
Ligand
Label
(a) Exp.
(a) Exp.
(a) Exp.
1
2
cat
sal
–31.98
–47.46
–45.07
1
6
Ent
6
✓
✓
✓
33.14
21.31
6.64
3
4
cat
salb
5
6
–18.94
–11.65
–29.24
–25.55
–26.65
–27.15
8
5
5
[10/11/11]
[01/11/11]
✓
✓
✗
✗
✗
✓
✓
✓
✗
✗
✓
✓
✓
✓
✗
5
11
cat
5
–18.05
–29.70
–28.86
6
7
cat
sal
4
6
2.17
4.93
–2.10
–8.41
4.50
9
4
4
3
[10/10/11]
[01/01/11]
[10/10/10]
–16.81
8
12
10
cat
4
8.21
0.61
4.58
9[c]
10
cat
sal
4
6
8.22
4.47
5.36
–4.20
–6.54
16.33
11[c]
12[d]
13[d]
13
14
15
3
4
2
[01/01/01]
[00/11/11]
[00/00/11]
cat
cat
cat
4
6
4
11.35
0.50
6.95
–12.13
16.11
4.64
–10.50
11.40
✗
✗
✗
✗
✓
✗
✗
✓
✗
21.09
[a] cat = catecholate, sal = salicylate binding mode. [b] One 2-hydroxybenzoic acid adopts salicylate binding mode. Additionally coordinated by [c] one methoxide (SiIV
and GeIV), and one i-propyloxide (TiIV) and [d] two methoxides (SiIV and GeIV), and two i-propyloxides (TiIV). DFT-calculations at B3LYP-D3 / cc-pVTZ(-f) // B3LYP /
LACVP** level of theory were performed.
5
This article is protected by copyright. All rights reserved.

10.1002/chem.201900453
Chemistry - A European Journal
Table 2 summarizes our experimental and computational findings on how the three metal ions bind to the
enterobactin derivatives. Beginning with the unmodified enterobactin system 1, we observed that all metals form
tightly bound, hexacoordinate complexes and our DFT-calculations estimate the binding free energies to be –32.0,
–47.5, and –45.1 kcal/mol for the SiIV–, GeIV–, and TiIV–Ent complexes, respectively (entry 1 in Table 2). Since the
carbonyl amide oxygen of the 2-hydroxybenzoic acid moiety gives access to the salicylate binding mode, both the
catecholate and salicylate binding modes were considered. Our computational models suggest that the catecholate
binding where the catechol-oxygens are engaged by the metal is thermodynamically most preferred for the
unmodified enterobactin ligand 1 with the binding energies in salicylate binding mode being 50-60 kcal/mol higher
in energy, as enumerated in entry 2 of Table 2. These energies strongly suggest that salicylate binding is not possible
for these d0-ions and is a binding mode that is special for iron and more complex transition metal ions. Given that
the estimate for the iron binding free energy is in the range of –67 kcal/mol,[2d] these computed energies in
catecholate binding mode are reasonable and highlight that enterobactin is a capable ligand for many metal cations,
but iron is clearly special among them. Removing one of the oxygen binding sites from enterobactin has little effect
on the general metal binding ability and we detected the metal-bound complexes for both pentadentate ligands 8
[10/11/11] and 11 [01/11/11] experimentally. Interestingly, our DFT-calculations show that there is a significant loss
in binding energy for all three metal ions with the predicted free energy of binding being –18.9, –29.2, and –26.7
kcal/mol for SiIV–, GeIV–, and TiIV–[10/11/11] complexes, respectively (entry 3 in Table 2). The losses in free energy
of binding of 13.1, 18.3, and 18.4 are much larger than anticipated, since the average M–O bond strengths corrected
for solvation effects, are between 5–8 kcal/mol, and they emphasize that the binding energies of the six M–O bonds
are not linearly additive. The salicylate binding mode allows for ligand 8 to adopt a 6-coordinate binding by engaging
the carbonyl amide oxygen of 2-hydroxybenzoic acid moiety. Despite having six M–O bonds vs the five M–O bonds
in the catecholate binding mode, the catecholate binding is preferred for SiIV and GeIV and both modes are
isoenergetic for TiIV (entry 4, Table 2). Thus, we can conclude again that even with the advantage of an additional
M–O bond, the salicylate binding is not a meaningful alternative for the catecholate binding. The [10/11/11] and
[01/11/11] binding modes (entries 3 and 5) do not give rise to any meaningful difference in binding energies.
Figure 1. DFT-calculated structures of the 6- and 5-coordinate complexes. (a) M–Ent (b) M–[10/11/11] (c) M–
[01/11/11] complexes. All hydrogens attached to carbons are omitted for clarity. The bond lengths and bond angles
denoted in color are average values. The unit of length is angstrom (Å).
Figure 1 compares the optimized structures of the 6- and 5-coordinate complexes. All three metal ions adopt
pseudo octahedral and pseudo square-pyramidal geometries with the SiIV–O bonds being shortest in the range of
1.84–1.69 Å, followed by GeIV–O bonds in the range of 1.94–1.79 Å and TiIV–O bonds being around 2.02–1.85 Å.
In the pseudo square-pyramidal structures, the axial M–O bonds are consistently shorter by ~0.05 Å for all metal
6
This article is protected by copyright. All rights reserved.

10.1002/chem.201900453
Chemistry - A European Journal
ions, as there is no ligand in the trans disposition to exert a trans influence. Since the M–O bond lengths increase
in the order of SiIV < GeIV < TiIV, the O–M–O angles denoted α in Figure 1a naturally decrease in the order of SiIV >
GeIV > TiIV ranging from 89 to 79°. It is interesting that the [10/11/11] and [01/11/11] binding motifs can be
accommodated easily both energetically and also structurally, as the cyclic lactone base can change its tilt-angle to
extend the reach of one phenolic arm. Clearly, such a structural distortion of the base will become more challenging
as more than one of the three catecholate arms are changed.
The ligands 9 and 12 are designed to provide a tetradentate binding environment by combining one
catecholate and two phenolate groups and these ligands gave more dramatic differences in metal binding ability.
Experimentally, we were unable to detect any binding of SiIV to 9. GeIV and TiIV, on the other hand, show metal
complexes with 9, indicated by ion signals in the mass spectrometer formally corresponding to complexes of the
type 4b and 4c, as presented in Scheme 2a. In good agreement with this experimental finding, our DFT calculations
indicate that binding SiIV is energetically unfavorable by ~2–8 kcal/mol, whereas negative and barely positive binding
energies of –8.4 and 0.6 kcal/mol are obtained for GeIV binding to 9 and 12, respectively. TiIV binding energies are
negative at –16.8 kcal/mol for 9, but slightly positive at 4.6 kcal/mol for 12, thus, predicting that only 9 should give a
metal ion bound complex, while no such complex should be seen for ligand 12. Computationally capturing these
absolute energies is difficult and whereas the binding energy of 4.6 kcal/mol for the putative TiIV binding to 12 is not
in agreement with the experimental observation, it is clear that the four-coordinate binding mode constitutes the
breaking point, where the metal ion binding energies are close to zero.
Figure 2 illustrates the DFT-predicted structures of the minimum energy tetradentate complexes of all three
metal ions. All three metal ions have tetrahedral geometries with the SiIV–O bonds being near 1.68–1.64 Å, the
GeIV–O bonds near 1.78–1.74 Å, and TiIV–O bonds in the range of 1.85–1.81 Å. In the tetrahedral structures, two of
four O-donors are involved in the independent phenolic arms and the others in the catecholic arms. Whereas O–M–
O angles, denoted β, where the two O-donors are involved in the catecholic arms are little different between the
[10/10/11] and [01/01/11] binding motifs due to structural rigidity, there are ~10 degree difference of O–M–O angles,
marked α, in which the two O-donors are belong to the respective phenolic arms in the case of SiIV– and GeIV–
complexes. In the structure of TiIV–[01/01/11], the catecholic arms bind at the metal in a manner that does not fully
maximize the M–O binding interaction. The angle between O–M–O plane and catecholic arm, which is denoted γ,
in the case of TiIV–complex indicates this ineffective binding of catecholic arms to the metal.
Figure 2. DFT-calculated structures of the tetradentate complexes. All hydrogens attached to carbons are omitted
for clarity. The bond lengths and bond angles denoted in color are average values. The unit of length is angstrom
(Å).
7
This article is protected by copyright. All rights reserved.

10.1002/chem.201900453
Chemistry - A European Journal
Ligand 12 [01/01/11] does not bind SiIV or GeIV, but time-dependent 1H NMR measurements and ion signals
in the mass spectrometer confirmed that TiIV–[01/01/11] was formed. These findings underline the importance of
the hydroxy groups in ortho position of dihydroxybenzoic acid (DHB) in Ent for complex formation. Thus, four O-
donors are sufficient for metal binding, but they have to be arranged properly. The ortho hydroxy groups of DHB is
important, because they allow the metal to adopt the salicylate and catecholate binding motif. Thus, in principle, this
ligand could provide hexacoordination to the cation where the two 2-hydroxybenzoic acids participate in salicylate
binding and the 2,3-dihydroxybenzoic acid adopts the catecholate binding. Due to steric reasons, such mixed
salicylate/catecholate binding is not possible in the case of 12 [01/01/11], however. Our calculations show in good
agreement with these observations that the free energy difference for metal binding with 9 in the catecholate binding
mode is much higher than that in the salicylate binding mode as shown in Table 2. The silicon complex of 9 displays
an unfavorable binding energy in both the salicylate and catecholate binding modes and experiments agree, as no
complex was detected.
Neither of the tridentate ligands, 10 [10/10/10] nor 13 [01/01/01], showed any metal binding abilities.
Similarly, the bidentate ligand 15 [00/00/11] was also unable to engage any of the three cations SiIV, GeIV, and TiIV.
Note that ligand 10 [10/10/10] can bind the metal ions as a salicylate type complex at best. Computationally, this
binding motif was predicted to be much less favorable than the catecholate binding at the given conditions, again
demonstrating a good correlation between experiment and computations, as summarized in Table 2.
Figure 3. DFT-calculated structures of the tridentate complexes. All hydrogens attached to carbons are omitted for
clarity. The bond lengths and bond angles denoted in color are average values. The unit of length is angstrom (Å).
Figure 3 describes the optimized structures of the tridentate complexes. As is similar with the tetradentate
complexes, all three ions adopt pseudo tetrahedral geometries. The SiIV–O bonds are in the range of 1.66–1.64 Å,
GeIV–O bonds near 1.77–1.75 Å, and TiIV–O bonds around 1.85–1.77 Å. Since these tridentate ligands provide three
O-donors, the metal center is in insufficient coordination for satisfying minimum requirements to adopt a tetrahedral
geometry. That leads that one alkoxide ion remains associated with the metal center upon metal complexation.
While the alkoxide anion placed in the external side of the Ent-analogs in the case of M–[10/10/10] complexes, that
of M–[01/01/01] complexes occupies the inner space of the Ent-analogs due to sufficient flexibility of the ligand.
Having [01/01/01] binding motifs requires more distortion of the structure resulting in a loss of binding energies.
It is clear that the structural changes upon metal binding and the ability of the ligands to accommodate them
play an important role. Because the three arms carrying the O-donors are connected through the triseryl base, the
structures are rigidified and certain arrangements are stabilized, while others become difficult, if not impossible. To
better understand how the triseryl base impacts metal binding, we carried out calculations where the base was
removed and the O-donors containing fragments were simplified to catechols and phenols. By comparing structures
8
This article is protected by copyright. All rights reserved.

10.1002/chem.201900453
Chemistry - A European Journal
and energies of the Ent-derivatives to the fragment ligand systems, it is possible to quantify the structural distortion
both in structure and energy that the triseryl base imposes, as conceptualized graphically in Figure 4a.
Table 3. DFT-calculated energies for the chelation of SiIV, GeIV, and TiIV by different fragmented ligands employing
a dielectric constant, ε = 46.68 for DMSO. All energies are in kcal mol–1.
ΔG[Frag]
GeIV
ΔΔG([Ent]–[Frag])
Entry
Frag
Frag1
Frag2
Ent
SiIV
TiIV
SiIV
GeIV
TiIV
1
2
3
–9.79
–26.61
–22.48
Ent
–22.19
–10.74
–11.63
–20.85
–11.12
–10.66
–22.59
–14.50
–12.29
[10/11/11]
[01/11/11]
–7.31
–18.58
–14.36
4
5
[10/10/11]
[01/01/11]
0.46
1.24
–8.33
–3.98
–10.02
Frag3
6.97
4.59
3.95
0.63
The free energies of metal binding by these fragments ΔG[Frag] are enumerated in Table 3 and they show
substantially weaker binding compared to the enterobactin analogues. For example, whereas the metal binding
energies ΔG[Ent] of the unmodified enterobactin were –32.0, –47.5, and –45.1 for SiIV, GeIV, and TiIV, respectively,
the removal of the triseryl base changes the free energy of binding to –9.8, –26.6, and –22.5 kcal/mol, respectively.
For understanding this dramatic energy difference of 22.2, 20.9, and 22.6 kcal/mol, respectively, two features are
important. First, since the tryserine base turns the three catechols formally into a single ligand, we expect to see a
chelate effect that prefers the Ent ligand. The chelate effect relies on reducing the entropic penalty of metal-ligand
binding by tethering multiple ligands together. Specifically, the translational entropy of the reactant state is reduced
in this manner, leading to a more favorable free energy of binding. Secondly, the structural distortion that is imposed
due to the presence of the triseryl ring is expected to play a role. In our fragment model, this energy can be easily
calculated by moving the untethered ligands to the positions found in the tethered analogues.
Figure 4. (a) Conceptual energy diagram. The circle shape indicates the O-binding aryl moieties. (b) Structural
distortion trends and (c) Chelate effect. R1 = OH or H, R2 = Me (SiIV and GeIV) or i-Pr (TiIV).
9
This article is protected by copyright. All rights reserved.

10.1002/chem.201900453
Chemistry - A European Journal
The translational entropy contributions by far dominate the free energy change, as a detailed energy
decomposition analysis given in the supporting information indicates. Figure 4c enumerates the translational entropy
components of each reactant and product species for the unmodified enterobactin and the analogous fragment
model. In accordance with the experiment, the metal ion is modeled as a tetra-alkoxide ion and the catecholes are
protonated. Thus, upon metal binding, four equivalents of the methanol and two protons are released. Thus, in this
reaction seven species are released from initially two molecular entities and the translational entropy change for this
process ranges from –46 to –49 kcal/mol at room temperature. If the triseryl base is removed, the same number of
species is generated from four participating molecules and, thus, the translational entropy change is diminished to
a range of –23 to –26 kcal/mol. In other words, the tethering of the three catechol arms increases the translational
entropy component of the free energy by 22.9, 22.8, and 22.8 kcal/mol for SiIV, GeIV, and TiIV, respectively, which
accounts practically for all of the free energy changes between the two models. Interestingly, the structural energy
penalty is found to be negligible ranging from 0.2 to 2 kcal/mol. Thus, the triseryl base does not at all distort the
three catechols from their most favorable binding geometry.
A similar analysis of the modified enterobactin derivatives with reduced number of O-donors paints a very
consistent picture regarding the entropy penalties and shows that the triseryl tethering lowers the free energy of
binding by ~20 kcal/mol. This is easy to understand, given that the contribution of the translational entropy to the
free energy at room temperature is approximately 10 kcal/mol for any molecule[16] and the tethering reduces the
number of independent particle on the reactant side by two, forming a single chelating ligand out of three ligands.
Interestingly, the structural distortion energies increase significantly for the ligands with reduced number of O-donors,
as illustrated in Figure 4b. The 5-coordinate ligands display structural distortion energies in the range of 8–12
kcal/mol and 12–24 kcal/mol is seen for the 4-coordinate cases, thus, cancelling out the chelate effect to a
substantial amount. This finding highlights that enterobactin is special, because it is perfectly constructed to benefit
from the much reduced entropy penalty of the chelate effect without incurring any incompatibility in structure of the
catechol ligands.
The 1H NMR spectra of the tridentate ligands 10 and 13 show no evidence of metal complexes being formed.
Because these ligands offer only three O-donors, and all three metal ions require four ligands, the fourth ligand is
assumed to be the alkoxide anion from the reactant state in our DFT-calculations. Consequently, the entropy gain
associated with the liberation of the alkoxide ions upon binding with the enterobactin derivatives is reduced by
roughly 10 kcal/mol, which makes the binding free energy to become positive for silicon and germanium (entry 9
and 10), whereas –4.2 and –6.5 kcal/mol are obtained for the catecholate and salicylate binding motifs of TiIV.
Therefore, although DFT-calculations and experiments are in slight disagreement quantitatively, it is clear that
having four O-donors is a critical requirement for metal binding.
10
This article is protected by copyright. All rights reserved.

10.1002/chem.201900453
Chemistry - A European Journal
Figure 5. DFT-calculated structures of the M–[00/11/11] complexes. All hydrogens attached to carbons are omitted
for clarity. The bond lengths denoted in color are average values. The unit of length is angstrom (Å).
Lastly, the 14 [00/11/11] ligand allows for accessing a different binding motif with four O-donors, namely,
where only two of the three aryl groups carry O-donor groups. We found that GeIV and TiIV readily bind to this ligand,
while no complex was detected with SiIV. DFT-calculations gave binding free energies of –12.1 and –10.5 kcal/mol
for the first two cations, and +0.5 for Si, respectively (Table 2). The calculated structures show that the two
catecholate ligands occupy the equatorial positions and the alkoxides become the axial ligands to complete the
octahedral coordination geometry as shown in Figure 5. If tetrahedral conformation is enforced in a computer
simulation, significant distortion in one phenolic arm is found, giving rise to a much weaker binding than what is seen
for octahedral geometry.
Figure 6. Comparison between the 1H NMR spectra (400 MHz) of (a) [00/11/11] and its complexes 5b and 5c; (b)
[10/11/11] and the potassium salts of its complexes 3a, 3b, and 3c; (c) [01/11/11] and the potassium salts of its
complexes 6a, 6b, and 6c in [D6]DMSO.
The reactions of the new enterobactin analogs with Si(OMe)4, Ge(OMe)4, and Ti(Oi-Pr)4 were studied in
greater detail using time-dependent NMR experiments in [D6]DMSO. The GeIV and TiIV-complexes 3b–c and 6b–c
11
This article is protected by copyright. All rights reserved.

10.1002/chem.201900453
Chemistry - A European Journal
were formed within 1 h without having to add any base.[9] On the contrary, the corresponding SiIV complexes 3a and
6a were only obtained after adding 3 equivalents of Et3N, presumably to assist the metal binding kinetics by
deprotonating the catechols. Ligands 9 and 12 did not show any reaction with Si(OMe)4, but when treated with
Ge(OMe)4 and Ti(Oi-Pr)4 minor changes in the 1H NMR spectra were observed. It was not possible to assign these
signals to the expected complexes 4b–c and 7b–c unequivocally, although the ESI-MS of samples taken from the
solutions showed trace amounts of the complexes as displayed in Table 2. Compound 14 was converted into the
corresponding GeIV and TiIV complexes 5b–c within 1 h as confirmed by ESI-MS and corresponding signals in the
1H NMR (Figure 6).
To learn more about the structure of the newly obtained complexes compounds 3a–c and 6a–c were
prepared as potassium salts and 5b–c as neutral complexes. Figure 6 shows the different 1H NMR spectra of the
complexes and their corresponding ligands. Interestingly, the spectra of the GeIV and TiIV–complexes 5b–c and 3b–
c show distinct signals, whereas the spectra of 3a and 6a–c seem to represent a mixture of several compounds.
Because ESI-MS analysis had already confirmed the stability of the compounds in solution, it was assumed that the
complexes can undergo conformational changes in solution.
Our observations are interesting with respect to the ability of enterobactin to form catecholate and salicylate
type complexes reported for FeIII and GaIII.[5,6a,6b] SiIV, GeIV, and TiIV–chelation by the synthetic enterobactin analogs
is favored if the ligand is at least tetradentate. Moreover, the presence of two catechol moieties in the tetracoordinate
ligands favors significantly M–chelation (M = SiIV, GeIV, and TiIV). FeIII is able to form salicylate type complexes as
demonstrated before by Raymond et al.,7 but our computations suggest that only catecholate type complexes are
relevant with SiIV, GeIV, and TiIV.
Conclusions
We prepared a comprehensive library of non-symmetric enterobactin analogs that allows for the first time
to study how enterobactin analogs with reduced numbers of O-donors behave in comparison to unmodified
enterobactin. Studies on how these enterobactin derivatives react with SiIV, GeIV, and TiIV ions in solution revealed
that at least four hydroxy groups must be present to obtain stable metal complexes. In contrast to what was
previously observed for FeIII, the three cations examined were incapable of adopting the salicylate binding mode.
Comparison of the complex formation rates of the SiIV–, GeIV–, and TiIV–complexes reveals that GeIV and TiIV–
chelation by the enterobactin analogs is faster than the SiIV–chelation. The new library of systematically derivatized
enterobactin analogs constitutes a versatile test platform for ion binding studies. The three ions were chosen for our
initial study, because they are easy to handle and they lack valence electrons in d-orbitals. Having established a
foundation for how to utilize the diverse library of enterobactin analogs, future work will aim at studying how more
complex transition metal ions, such as the biological target iron, but also other metal ions interact with the
enterobactin analogs.
Supporting Information Available
For full experimental details, computational details, computed energy components, Cartesian coordinates, and
vibrational frequencies of all of the DFT-optimized structures, see Supporting Information.
ORCID
Ho Ryu: 0000-0001-8450-991X
Mu-Hyun Baik: 0000-0002-8832-8187
Roderich D. Süssmuth: 0000-0001-7027-2069
12
This article is protected by copyright. All rights reserved.

10.1002/chem.201900453
Chemistry - A European Journal
Conflicts of interest
There are no conflicts to declare.
Acknowledgements
RDS thanks the Deutsche Forschungsgemeinschaft (DFG SU 239/18) and MHB acknowledges financial support
from the Institute for Basic Science (No. IBS-R10-A1) in Korea for financial support.
Keywords: Siderophore, Enterobacin, Chelate effect, Silicon, Germanium, Titanium
Notes and references
[1] R. C. Hider, X. Kong, Nat. Prod. Rep. 2010, 27, 635–657.
[2] (a) J. R. Pollack, J. B. Neilands, Biochem. Biophys. Res. Commun. 1970, 38, 989–992. (b) I. G. O’Brien, F.
Gibson, Biochim. Biophys. Acta 1970, 215, 393–402. (c) K. N. Raymond, E. A. Dertz, S. S. Kim, Proc. Natl.
Acad. Sci. U. S. A. 2003, 100, 3584–3588. (d) T. Zheng, E. M. Nolan, Metallomics 2012, 4, 866–880.
[3] (a) B. J. Hernlem, L. M. Vane, G. D. Sayles, Wat. Res. 1999, 33, 951-960. (b) A. Nair, A. A. Juwarkar, S. K.
Singh, Water, Air, and Soil Pollution 2007, 180, 199–212.
[4] (a) J. R. Boelaert, M. D. Locht, J. V. Cutsem, V. Kerrels, B. Cantinieaux, A. Verdonck, H. W. Landuyt, Y. J.
Schneider, J. Clin. Invest. 1993, 91, 1979–1986. (b) J. B. Neilands, J. Biol. Chem. 1995, 270, 26723–26726.
[5] K. D. Krewulak, H. J. Vogel, Biochimica et Biophysica Acta 2008, 1778, 1781–1804.
[6] (a) M. A. Fischbach, H. Lin, D. R. Liu, C. T. Walsh, Nat. Chem. Biol. 2006, 2, 132–138. (b) M. Sandy, A. Butler,
Chem. Rev. 2009, 109, 4580–4595.
[7] (a) S. M. Cohen, M. Meyer, K. N. Raymond, J. Am. Chem. Soc. 1998, 120, 6277–6286. (b) S. M. Cohen, K. N.
Raymond, Inorg. Chem. 2000, 39, 3624–3631. (c) R. J. Abergel, J. A. Warner, D. K. Shuh, K. N. Raymond, J.
Am. Chem. Soc. 2006, 128, 8920–8931.
[8] T. Schmiederer, S. Rausch, M. Valdebenito, Y. Mantri, E. Mꢀsker, T. Baramov, K. Stelmaszyk, P. Schmieder,
D. Butz, S. I. Mꢁller, K. Schneider, M.-H. Baik, K. Hantke, R. D. Sꢁssmuth, Angew. Chem., Int. Ed. 2011, 50,
4230–4233.
[9] T. Baramov, K. Keijzer, E. Irran, E. Mösker, M.-H. Baik, R. Süssmuth, Chem. Eur. J. 2013, 19, 10536–10542.
[10] (a) C. Chuit, R. J. P. Corriu, C. Reye, J. C. Young, Chem. Rev. 1993, 93, 1371–1448. (b) R. R. Holmes, Chem.
Rev. 1996, 96, 927–950. (c) O. Seiler, C. Burschka, T. Fenske, D. Troegel, R. Tacke, Inorg. Chem. 2007, 46,
5419–5424. (d) B. Theis, J. Weiß, W. P. Lippert, R. Bertermann, C. Burschka, R. Tacke, Chem. Eur. J. 2012,
18, 2202–2206.
[11] (a) A. Boudin, G. Cerveau, C. Chuit, R. J. P. Corriu, C. Reye, Organometallics 1988, 7, 1165–1171. (b) G.
Cerveau, C. Chuit, R. J. P. Corriu, C. Reye, Organometallics 1991, 10, 1510–1515. (c) A. C. Sau, R. R. Holmes,
Inorg. Chem. 1981, 20, 4129–4135. (d) Á. Pintér, G. Haberhauer, Chem. Eur. J. 2008, 14, 11061–11068.
[12] (a) B. A. Borgias, S. R. Cooper, Y. B. Koh, K. N. Raymond, Inorg. Chem. 1984, 23, 1009–1016. (b) M. Albrecht,
Chem. Eur. J. 1997, 3, 1466–1471. (c) S. Priya, M. S. Balakrishna, J. T. Mague, Chem. Lett. 2004, 33, 308–
309. (d) S. Burk, M. Albrecht, K. Hiratani, J. Incl. Phenom. Macrocycl. Chem. 2008, 61, 353–359. (e) L. T.
Dulatas, S. N. Brown, E. Ojomo, B. C. Noll, M. J. Cavo, P. B. Holt, M. M. Wopperer, Inorg. Chem. 2009, 48,
10789–10799. (f) R. Uppal, H. P. Israel, C. D. Incarvito, A. M. Valentine, Inorg. Chem. 2009, 48, 10769–10779.
13
This article is protected by copyright. All rights reserved.

10.1002/chem.201900453
Chemistry - A European Journal
[13] M. Meyer, J. R. Telford, S. M. Cohen, D. J. White, J. Xu, K. N. Raymond, J. Am. Chem. Soc. 1997, 119, 10093–
10103.
[14] Q.-L. Luo, L. Lv, Y. Li, J.-P. Tan, W. Nan, Q. Hui, Eur. J. Org. Chem. 2011, 2011, 6916–6922.
[15] X. Yu, Y. Dai, T. Yang, M. R. Gagné, H. Gong, Tetrahedron 2011, 67, 144–151.
[16] H. Ryu, J. Park, H. K. Kim, J. Y. Park, S.-T. Kim, M.-H. Baik, Organometallics 2018, 37, 3228–3239.
14
This article is protected by copyright. All rights reserved.

10.1002/chem.201900453
Chemistry - A European Journal
TOC graphic
15
This article is protected by copyright. All rights reserved.