The Role of Electrostatics in Siderophore Recognition by the Immunoprotein Siderocalin

Abstract

Iron is required for virulence of most bacterial pathogens, many of which rely on siderophores, small-molecule chelators, to scavenge iron in mammalian hosts. As an immune response, the human protein Siderocalin binds both apo and ferric siderophores in order to intercept delivery of iron to the bacterium, impeding virulence. The introduction of steric clashes into the siderophore structure is an important mechanism of evading sequestration. However, in the absence of steric incompatibilities, electrostatic interactions determine siderophore strength of binding by Siderocalin. By using a series of isosteric enterobactin analogs, the contribution of electrostatic interactions, including both charge-charge and cation-π, to the recognition of 2,3-catecholate siderophores has been deconvoluted. The analogs used in the study incorporate a systematic combination of 2,3-catecholamide (CAM) and N-hydroxypyridinonate (1,2-HOPO) binding units on a tris(2-aminoethyl)amine backbone, [tren(CAM)_m(1,2-HOPO)_n where m = 0, 1, 2, or 3 and n = 3 – m]. The shape complementarity of the synthetic analog series was determined through small molecule crystallography and the binding interactions investigated through a fluorescence-based binding assay. These results were modeled and correlated through ab initio calculations of the electrostatic properties of the binding units. Although all the analogs are accommodated in the binding pocket of Siderocalin, the ferric complexes incorporating decreasing numbers of CAM units are bound with decreasing affinities (K_d = >600, 43, 0.8 and 0.3 nM for m = 0-3). These results elucidate the role of electrostatics in the mechanism of siderophore recognition by Siderocalin.

Introduction

Iron has a range of biologically relevant oxidation states including Fe(II), Fe(III) and Fe(IV), and is therefore a platform for electron transfer, redox chemistry and oxygen transport in living systems.² However, the availability of ferric ion is limited by its insolubility in aqueous, pH-neutral environments³, and the presence of ferrous ion poses the risk of the production of destructive oxygen radicals.⁴ In response to these environmental pressures for tight regulation, both mammals and microorganisms have developed efficient and competitive systems to acquire, transport and store iron.^3,5,6

Of the 3-4 grams of iron in the average human body, approximately 2.5 grams participate in oxygen transport in hemoglobin, another 0.4 grams is engaged in other cellular functions, and the small remaining portion of iron is bound by transport and storage proteins such as extracellular transferrins and intracellular ferritin.^7,8 With nearly all the iron confined to such proteins, the concentration of “free” extracellular iron is estimated at 10⁻²⁴ M.⁹ This iron containment strategy serves multiple physiological functions; first, to solubilize ferric ions which would otherwise be unavailable in a biological environment, second, to prevent ferrous ions from participating in unregulated production of reactive oxygen species (ROS) via the Fenton reaction, and finally to act as a first line of defense against microbial infection by withholding iron.

Like mammals, microorganisms have developed effective systems to secure their iron supply. Many bacteria synthesize and secrete low-molecular weight, highly iron-selective chelators called siderophores to collect iron from the extracellular environment.⁸ There are hundreds of known siderophore structures with a variety of denticities, chelating units and backbone scaffolds (Figure 1). Usually bacteria synthesize more than one type of siderophore (often a catecholate and a hydroxamate or citrate) by taking advantage of this modular ligand assembly. In addition, most microorganisms are able to utilize xenosiderophores, expressing receptors for siderophores that the microbe does not itself produce.^10,11 The use of divergent, multi-siderophore systems by successful pathogens has proven to be an advantage against a number of environmental pressures including the bacteriostatic action of the immune system in a mammalian host.

Figure 1.

Selected siderophore structures, illustrating a variety of backbones and chelating moieties. Iron binding atoms are shown in red. High-affinity, archetypal siderophores enterobactin and bacillibactin, from Gram-negative and Gram-positive bacteria respectively, are sequestered by the immune system protein Siderocalin (Scn) while aerobactin (Escherichia coli), salmochelin-S4 (Salmonella enterica), petrobactin (Bacillus anthracis) and alcaligin (Bordetella pertussis) are not bound by Scn and confer virulence in the mammalian host.

While containment of iron in a mammalian host is important as a preliminary protection against infection, many pathogens can overcome this limitation by producing siderophores, such as enterobactin and bacillibactin, that have a strong kinetic and thermodynamic advantage for removal of iron from transferrin.¹² In response to this action, the mammalian innate immune system produces Siderocalin (Scn; see Figure 2, a.k.a. lipocalin 2, NGAL, uterocalin and 24p3) a 20 kDa lipocalin that binds both ferric and apo siderophores thereby preventing iron delivery.^13-16 The calyx (or binding site) of Scn is shallow and lined with positive residues (K125, K134 and R81) that promote the binding of a number of different siderophores, including enterobactin and bacillibactin, which have the highest known affinities for iron. The recognition and the sub-nanomolar dissociation constant for the Scn/ferric enterobactin interaction is mediated by electrostatic bonds (including both coulombic and cation-π) in the absence of good shape complementarity, atypical for tight protein/small-molecule binding. Nonetheless, recent work has shown that the most successful (and deadly) pathogens evade the host immune system by utilizing siderophores that are not recognized by Scn.

Figure 2.

Enterobactin bound within the β-barrel of Scn (left). A space-filling, close-up view of the calyx showing positively charged residues in blue that shape the binding pocket (center). Cationic residues K125, K134, and R81 create a cyclically permuted cation-π interaction with enterobactin (right).

There are several examples in the literature that describe the substrate shape limitations of the Scn calyx. Successful pathogens such as Bacillus anthracis and Salmonella enterica use these shape constraints to their advantage by producing “stealth siderophores”, salmochelin^17,18 and petrobactin¹⁵ correspondingly (Figure 1), which are not recognized because of gross steric incompatibilities with the rigid calyx. Hydroxamate-based siderophores, which form neutral complexes, are another class of siderophores not bound by Scn.¹⁴ Pathogens such as Escherichia coli and Bordetella pertussis, both of which can utilize enterobactin, instead rely on hydroxamate siderophores, aerobactin¹⁹ and alcaligin²⁰, respectively (Figure 1), for virulence in a mammalian host.²¹ While these systems similarly demonstrate that siderophore-mediated iron acquisition in pathogenic bacteria is closely associated with successful evasion of Scn, it raises the question: do both steric and electronic features of siderophores contribute to Siderocalin recognition?

The work presented here aims to determine the relative contributions of the electrostatic components, both coulombic and cation-π, to siderophore recognition through the use of an isosteric series of tren-based enterobactin analogs (tren(CAM)_m(1,2-HOPO)_n 1-4) (Figure 3). The analogs incorporate a systematic combination of catecholamide (CAM) and 1,2-hydroxypyridinone (1,2-HOPO) metal binding units. The isosteric nature of the ferric complexes was verified though small molecule X-ray crystallography.²² The sequestration by Scn of the apo, Fe^III and V^IV complex analogs was evaluated using fluorescence spectroscopy. A series of ab initio calculations were performed to study cation binding to the π face of the aromatic units. The Scn interactions with a larger analog series (5-7) incorporating N-methyl-3,2-hydroxypyridinone binding units (Me-3,2-HOPO, Figure 3) were inspected as a parallel control for the isosteric 1,2-HOPO series. These studies demonstrate the importance of cation-π interactions to recognition, which, in concert with steric constraints imposed by the relatively rigid calyx, enables Scn to target unmodified catecholate siderophores but restricts binding of modified, virulence-associated “stealth” siderophores.

Figure 3.

Isosteric enterobactin analogs (1-4) and larger Me-3,2-HOPO analogs (5-7) used to study the influence of the electrostatic recognition of siderophores by Scn. The bidentate CAM and 1,2-HOPO groups have the same steric profile but pose different electrostatic interactions, whereas the Me-3,2-HOPO unit is similar electronically to the 1,2-HOPO unit but incorporates added steric bulk. Iron binding atoms shown in red.

Results

Ligand and Metal Complex Synthesis

We have previously shown that ferric tren(CAM)₃, a well studied synthetic analog of enterobactin, binds to Scn with an affinity comparable to the ferric complex of the natural siderophore.¹⁸ This and other studies indicated that the backbone is not an important part of recognition by Scn.¹⁸ Syntheses of tren(CAM)₃ ²³ , tren(1,2-HOPO)₃ ²⁴ and tren(Me-3,2-HOPO)₃²⁵ have been reported previously. The synthesis of the mixed binding unit analogs (Scheme 1) is based on methods previously described in the literature for multiple substitution of tren.^18,26 Slow addition of a stoichiometric (2:1) amount of thiazolide-activated benzyl-protected CAM, 1,2-HOPO or Me-3,2-HOPO to the commercially available tren scaffold afforded the bis-substituted tren intermediate (8, 9 or 10) and a third binding unit was subsequently coupled, yielding the benzyl-protected ligands (11, 12, 13 or 14). The ligands were deprotected using acidic conditions (1:1 HCl:HOAc) and recrystallized from methanol and diethyl ether to afford the HCl salt of the ligands (2, 3 and 5, 6) in good yields. The iron complexes of all ligands were either prepared in situ (for fluorescence studies) or isolated (for crystallization). One equivalent of Fe(acac)₃ was added to each of the ligands in methanol followed by a stoichiometric amount of KOH to neutralize the HCl salt and deprotonate the ligand. In order to attain the complexes as solids, the metal complexes were recrystallized from diethyl ether. Potassium counter ions were associated with the negatively charged complexes.

Scheme 1.

Synthesis of tren-based enterobactin analogs with mixed binding units. Reagents and conditions: (i) 0.5 equivalents tris-(2-aminoethyl)amine, CH₂Cl₂; (ii) 1 equivalent of Bn-1,2-HOPO-thiaz, Bn-Me-3,2-HOPO-thiaz, or Bn₂CAM-thiaz, CH₂Cl₂; (iii) 1:1 HCl:AcOH; (iv) Fe(acac)₃, KOH, MeOH, under N₂.

Solid-state structures of [Fe^III(tren(CAM)_m(1,2-HOPO)_n)]^m− and Structural Overlay of Isosteric Series

In order to demonstrate that the ligand series proposed would only test the electronic recognition of the ligands and their complexes and not their steric interaction, the X-ray structures of the ferric complexes were compared. While the structure of the trianionic ferric tren(CAM)₃ complex has been previously described²⁷, the solid state characterization of the dianion, monoanion and neutral ferric complexes are described here (Figure 4).

Figure 4.

Structural diagrams (ORTEP), top and side views, of [Fe^IIItren(CAM)₂(1,2-HOPO)]²⁻, A; [Fe^IIItren(CAM)(1,2-HOPO)₂]¹⁻ , B; and [Fe^IIItren(1,2-HOPO)₃]⁰, C. Solvent molecules, counter ions and hydrogens are omitted for clarity (50% probability elipsoides). Top and side views of MERCURY overlaid structures of [Fe^IIItren(CAM)₃]³⁻ and [Fe^IIItren(1,2-HOPO)₃]⁰, D with an RMS of 0.476 Å for the 34 atom pairs.

All three complexes, 2K⁺[Fe^III(2)]²⁻, K⁺[Fe^III(3)]¹⁻ and [Fe^III(4)]⁰ crystallize in the monoclinic space group P2₁/c (Table S1) while the published structure of 3K⁺[Fe^III(1)]³⁻ crystallizes in the cubic space group P2₁3. As described for [Fe^III(1)]³⁻ and [V^IV(enterobactin)]²⁻, the amide protons in these structures are directed inward and participate in a hydrogen bonding network with the o-hydroxyls typical of these types of complexes.^27,28 The potassium counter ions were either coordinated to an amide carbonyl or situated under the iron center interacting with the catecholate m-hydroxyls. There was no evidence of K⁺-π interactions in these small molecule structures.

The tren-based ligands are achiral and therefore afford a racemic mixture of complexes ( Δ and Λ). Two molecules of the same metal center chirality were used for the structural overlay. Since the backbone is not important to Scn recognition, only the binding units including the amides were used in the shape comparison. The ferric catecholamide units and the ferric hydroxypyridinone amide units of [Fe-tren(CAM)₃]³⁻ and [Fe-tren(1,2-HOPO)₃]⁰ were overlaid and distance minimized²², yielding an RMS deviation for the 34 atom pairs of 0.476 Å (Figure 4). This shows that the spatial arrangement of the complexes is virtually identical and Scn recognition should therefore only differ due to the electrostatic nature of the complexes.

Fluorescence Binding Assay

The affinities of Scn for the analogs were determined by a fluorescence-based binding assay as previously described.¹⁴ A 100 nM solution of Scn was titrated with between two and ten equivalents of substrate, depending on the affinity. The binding curves were fit by nonlinear least squares refinement and the resulting dissociation constants calculated.²⁹ Four sets of substrates were used in this study. The Fe^III complexes and the V^IV complexes of the tren(CAM)_m(1,2-HOPO)_n series were used as a direct comparison of complex charge. The apo ligands were the third set of substrates that was used to explore the recognition of the aromatic units. The fourth set of substrates was the ferric complexes of a series of ligands incorporating the structurally similar Me-3,2-HOPO unit (Figure 3) rather than the 1,2-HOPO unit. This alternative series combines both steric and electronic perturbations as a parallel control for the isosteric 1,2-HOPO series. Binding results are shown in Figure 5 and Table 1. Crystallographic analyses of complexes between Scn and the isosteric 1,2-HOPO series show that, in general, these compounds bind in the calyx in the same orientation as enterobactin (Clifton et al., manuscript in preparation), making them ideal compounds for isolating the relative effects of cation-π versus coulombic bonds.

Figure 5.

Fluorescence quenching titrations measuring the affinity of Scn for isosteric analogs (1-4) in the Fe^III (top left), V^IV (top right) and apo (bottom left) forms, as well binding curves with the Fe^III complexes of mixed CAM ligands incorporating a bulkier Me-3,2-HOPO (bottom right). Experiments were performed with 100 nM Scn samples buffered at 7.4 with TBS. An excitation wavelength of 281 nm was used and emission monitored at 340 nm. The symbols represent the experimental fluorescence data (tren(CAM)_m(HOPO)_n : m=3, n=0; : m=2, n=1; : m=1, n=2; : m=0, n=3) and the calculated fits are shown as lines.

Table 1.

Scn affinities for tren-based analogs and their metal complexes.^a, ^b

		Kd
	FeIII	VIV	no metal
trenCAM3	0.3±0.1 nM18	1.3±0.3 nM	1.2±0.2 nM
trenCAM2(1,2-HOPO)	0.8±0.3 nM	40±9 nM	1.4±0.3 nM
trenCAM2(Me-3,2-HOPO)	0.12±0.01 μM	n.d.	n.d.
trenCAM(1,2-HOPO)2	43±2 nM	0.22±0.07 μM	0.14±0.01 μM
trenCAM(Me-3,2-HOPO)2	0.15±0.02 μM	n.d.	n.d.
tren(1,2-HOPO)3	> 0.6 μM	0.30±0.03 μM	0.25±0.03 μM
tren(Me-3,2-HOPO)3	> 0.6 μM	n.d.	n.d.

^aDissociation constants are reported in units of either nM or μM depending on the magnitude in order to maintain consistent significant figures.

^bSee supplementary information for a bar graph representation of Scn dissociation constants (Figure S1).

A decrease of affinity was observed within the ligand series for both the ferric and vanadium complexes with decreasing charge. In a comparison of a ferric complex to the vanadium complex with the same ligand, there is a decrease in affinity for every pair of substrates. In addition the charge-matched complexes are bound with comparable K_d’s for the pairs of complexes. The more highly charged complexes (3− and 2−) are recognized with fairly strong affinities with only slight charge effect. The complexes of lesser charge (1−, neutral, and 1+) experience a more dramatic decrease in recognition compared to larger anions, such that binding was barely detectable under the experimental conditions.

The apo ligands were also tested for Scn binding. The charges on the ligands were estimated from the previously determined protonation constants for tren(CAM)₃ and tren(1,2-HOPO)₃. Since the 1,2-HOPO unit is a relatively acidic binding unit (in tren(1,2-HOPO)₃ pK_aHOPO = 3.60, 4.32, 5.62³⁰) the hydroxyl will be deprotonated in our experimental conditions (pH = 7.4). For the more basic CAM unit (in tren(CAM)₃ pK_aCAM = 6.71, 8.61, 8.75, 11.3, 12.1, 12.9²³) only one of the meta oxygens is deprotonated at 7.4, therefore the overall ligand charge at physiological pH is 1−. The fluorescence binding data indicated that Scn affinity decreased with increasing 1,2-HOPO content, in that tren(CAM)₃ is recognized with sub-nanomolar affinity while tren(1,2-HOPO)₃ is bound in a micromolar affinity range.

Relative to the 1,2-HOPO series, the Me-3,2-HOPO analog series provides a platform for comparison of the influence of shape versus charge. An additional methyl group functionalizes the N group on the binding unit, placing more steric demand on the bottom side of the binding unit. The addition of both one and two Me-3,2-HOPO units decreases the affinity three orders of magnitude compared to the parent tris-catecholate. The complex with three Me-3,2-HOPO units is not bound.

Ab initio Calculations of Quadrupole Moments and Cation-π interaction energies

Two sets of ab initio calculations were performed to examine the interactions of the π surfaces of a siderophore with cationic residues in the calyx of Scn. The methyl amide bidentate units of CAM and 1,2-HOPO were used as a simplified model (Figure 6). In order to mimic coordination to a metal center the 1,2-HOPO and CAM units were coordinated to the alkaline metal ions, sodium and calcium respectively, to form the neutral metal complexes. The free ligand forms of the bidentate units were also studied. These four molecules were geometry optimized and the quadrupole tensor components were determined at the RHF/6-311G** level of theory.³¹

Figure 6.

Schematic of the quadrupole moment of benzene with negative density above and below the ring represented in red (A). Representative input structure for calculations of cation-π interactions energies with the model aromatic units, as shown here the cation (in pink) was fixed at 2.47 Å above the centroid of the M(1,2-HOPO) aromatic (B). Computational studies describe the electrostatic interactions of metal binding units of siderophores with positive residues of the calyx. The bottom panel includes quadrupole moments (calculated at the level RHF/6-311G** of theory), cation-π interaction energies (calculated at MP2/6-311++G** the level of theory), molecular structures and electrostatic potentials (with a range of −0.18 kcal/mol (red) to +0.18 kcal/mol (blue)) for the four model aromatic units used in this study.

Next the cation interaction energies with the π surfaces of the four aromatic units were determined at the MP2/6-311++G** level of theory and the basis set superposition error was corrected for with the counterpoise method.³¹ A cation was placed above the centroid of the aromatic at 2.47 Å as described in the literature³² (Figure 6B). In these calculations a sodium ion was used as a non-polarizable model cation, as is common practice in these types of calculations.^32,33 Since these molecules have alternate binding sites for the cation (for example the amide carbonyl) the geometry was fixed with the cation above the centroid of the previously optimized π system. In these calculations a purely electrostatic model was employed and therefore, did not account for the influence of polarizability in the cation-π system. Nonetheless this model has proven suitable for comparison of cation-π interactions within an aromatic series.³²

The anisotropy of the tensor components of the four molecules used in these calculations was calculated as Θ_zz = Q_zz – 0.5(Q_xx + Q_yy). The quadrupole moments calculated for the neutral molecules were within the range of values reported for similarly functionalized aromatics.³³ Both CAM molecules were found to have more negative Θ_zz values than their 1,2-HOPO counterparts. In comparing the protonated molecules to the metal complexes, both CAM and 1,2-HOPO experienced a more negative Θ_zz upon metal binding (Figure 6). Calculated cation interaction energies with these metal binding units confirmed that as Θ_zz becomes more negative, the binding energy increases.

Discussion

Recognition by Coulombic Interactions

A comparison of the dissociation constants of the Fe^III complexes and the V^IV complexes provides direct evidence of the effect of substrate charge on recognition. Previous work has shown that the Scn recognition of [V^IV(enterobactin)]²⁻ is slightly weaker than [Fe^III(enterobactin)]³⁻.¹⁸ The present study expands on this result, demonstrating the generality of the charge effect on Scn recognition. Since there is a significant difference in binding between the neutral and monoanionic complexes compared to the diaionic and trianionic complexes, it appears that Scn favors at least dianionic substrates for strong recognition. In addition, Fe^III and V^IV complexes of the same charge (different ligands), for example [Fe^III(tren(CAM)(1,2-HOPO)₂)]¹⁻ and [V^IV(tren(CAM)₂(1,2-HOPO))]¹⁻, have the same binding affinity within experimental error.

The catecholate binding unit, when compared to the hydroxamate, has enhanced stability with iron that increases from the bidentate to tetradentate to hexadentate complexes. For example, while the tris-hydroxamate desferrioxamine B has a pM value of 26.6 the tris-catecholate enterobactin has a pM value of 35.3 (where pM = -log[Fe^III] at [L] = 10 μmol L⁻¹ and [Fe^III] = 1 μmol L⁻¹)³⁴ making it a far superior ligand for iron binding.³ Catecholate complexes of ferric ion are highly anionic due to the dianionic bidentate unit.³⁵ Therefore, Scn impedes iron scavenging via catecholate siderophores by recognizing these anionic substrates through coulombic electrostatic binding interactions.

Recognition by Cation-π Interactions

A cation-π interaction has a distance dependence of 1/rⁿ where n<2, similar to the 1/r dependence in a coulombic interaction.³⁶ However, while a coulombic attraction provides a strong interaction between two charged groups, the cation-π interaction creates a specific interaction dependent upon geometric constraints. The calculated quadrupole moments and cation-π interaction energies for the 2,3-catecholamide structures support the experimental observation that in general the CAM containing ligands have stronger binding affinities with Scn than the 1,2-HOPO ligands. This becomes most apparent when comparing the binding trend for the apo ligands. While the tris-substituted 1,2-HOPO ligand, with an overall 3− charge, binds weakly to Scn, the tris-substituted CAM ligand has an overall 1− charge and is bound much more tightly. Thus it appears that the cation-π interaction of the 2,3-catecholamide is more important to binding than the coulombic attraction between the anionic substrate and the cationic calyx. The recognition of apo catecholate siderophores is biologically relevant, since it has previously been demonstrated that both apo bacillibactin and apo enterobactin are bound with high affinity by Scn.^15,18

Overall Recognition Mechanism

This study has focused on the importance of the electronic character of a siderophore to Scn recognition, demonstrating that the electrostatic nature, including the quadrupole of aromatics and the overall charge of the substrate, is important to selectivity. Whereas previous studies have shown that large adducts also affect selectivity by introducing intolerable steric clashes, here we show that even the N-methyl substituent on the Me-3,2-HOPO unit is not accommodated. It has been shown previously that neither the backbone nor the metal, nor the chirality at the metal center is important for recognition by Scn.^15,18 The features of the metal-binding units seem to single-handedly influence whether Scn binds a substrate.

One of the most remarkable features of Scn is that it can bind a number of substrates of differing structures while maintaining an affinity comparable to that of cognate siderophore receptors, in order to act effectively as a bacteriostatic agent. The recognition mechanism relies on interactions between the protein and key sub-structures of the substrate rather than global recognition of the entire ligand. This recognition degeneracy may also enable the protein to participate in moonlighting functions. Scn has been implicated in several other physiological events unrelated to bacteraemia such as iron trafficking, tissue development and apoptosis.^37-39 Recognition mechanisms that broaden the specificity of Scn-ligand interactions may, therefore, extend to as-yet uncharacterized endogenous ligands that mediate pleiotropic functions of Scn.

Conclusion

Scn has an affinity for ferric enterobactin similar to the siderophore’s bacterial membrane receptor FepA, making it an effective competitor as part of an immune response against bacterial iron theft. Scn also binds ferric bacillibactin with equal affinity. Bacillibactin and enterobactin are the archetypal siderophores of Gram-positive and Gram-negative bacteria, respectively. Both are tripodal 2,3-catecholates based on trilactone backbones, which provides these compounds unrivaled affinities for ferric ion. But, because of Scn, neither siderophore contributes to virulence.

Based on this study, we conclude that hydroxamates do not bind to Scn, in part, due to the formation of neutral ferric complexes and the lack of aromatic binding units, the hallmark structural unit recognized by Scn. Hydroxamates, while having a lower affinity for iron compared to catecholate siderophores, may be the workhorse siderophores for many organisms because they are not impeded by Scn. Many microorganisms rely on a dual siderophore system for iron acquisition, often including a catecholate and a hydroxamate. While this strategy offers several advantages, including the ability to adapt to varying degrees of environmental stress, one reason to employ a hydroxamate siderophore in a mammalian host is to ensure virulence in the presence of the bacteriostatic action of the immune system.

Experimental

General

Unless otherwise noted, starting materials were obtained from commercial suppliers and were used without further purification. Flash silica gel chromatography was performed using Merck silica gel (40-7 mesh). Microanalyses were performed by the Microanalytical Services Laboratory, College of Chemistry, University of California at Berkeley, Berkeley, CA 94720. Mass spectra were recorded at the Mass Spectrometry Laboratory, College of Chemistry, University of California at Berkeley, Berkeley, CA 94720. Melting points were taken on a Büchi melting apparatus and are uncorrected. All ¹H NMR and ¹³C NMR spectra were recorded on AV-300 or AVB-400 Bruker FT spectrometers unless otherwise noted. Tren(CAM)₃ (1)²³, tren(1,2-HOPO)₃ (4)²⁴, tren(Me-3,2-HOPO)₃ (7)²⁵, Bn₂-2,3-CAM-thiaz²⁶, Bn-1,2-HOPO-thiaz²⁴, Bn-Me-3,2-HOPO-thiaz⁴⁰, and [Fe^IIItren(CAM)₃]^{3− 27}, were prepared according to published procedures.

Synthesis

The synthesis of all of the mixed ligands followed the same process, as illustrated in Scheme 1. The synthetic procedure of tren(CAM)₂(1,2-HOPO) is fully described, as well as the synthesis of the iron complex [Fe^IIItren(CAM)₂(1,2-HOPO)]²⁻, as representative examples for the syntheses all of the other intermediates, ligands and iron complexes. Complete characterizations for all new compounds are included.

Bn-tren(CAM)₂(1,2-HOPO) (11)

2,3-Benzyloxybenzoylmercaptothiazoline (Bn-2,3-CAM-thiaz) (1.334 g, 3.062 mmol, 2 equiv) was dissolved in chloroform (150 mL) and added dropwise via capillary over four days to tris(2-aminoethyl)amine (0.229 mL, 1.531 mmol, 1 equiv) in dichloromethane (50 mL). The resulting mixture was evaporated to dryness and the presence of the bis-substituted intermediate (8) was confirmed by proton NMR at which point the crude material was carried into the next step without further purification. ¹H NMR (300 MHz, CDCl₃): δ 7.92 (t-br, 2H), 7.59-7.09 (m, 22H), 7.07 (d, J = 8.1 Hz, 4H) 5.10 (s, 4H, CH₂), 5.02 (s, 4H, CH₂), 3.18 (q, J = 6.3 Hz, 4H, CH₂), 2.45 (t, J = 6 Hz 2H, CH₂), 2.35 (t, J = 6.6 Hz 6H, CH₂), 1.86 (s, 2H, NH₂) ppm.

The bis-substituted intermediate (8) (0.990 g, 1.271 mmol) was dissolved in CH₂Cl₂, and added to N-Benzyloxypyridin-2-one-6-mercaptothiazoline (Bn-1,2-HOPO-thiaz) (0.440 g, 1.271 mmol, 1 equiv) in CH₂Cl₂. The reaction mixture was stirred until the starting material was consumed (the solution went from yellow to clear). The reaction mixture was extracted with NaOH (1M) 3x, dried over Na₂SO₄ and condensed. The resulting residue was applied to a silica column and product eluted with 2:98 CH₃OH:CH₂Cl₂. The fractions containing the protected ligand (11) were combined and the solvent evaporated to obtain the product as clear oil. Yield: 62%. ¹H NMR (300 MHz, CDCl₃): δ 7.82 (t-br, 2H), 7.65 (t-br, 1H), 7.32-6.98 (m, 32H), 6.53 (d, J = 7.8 Hz, 1H), 6.11 (d, J = 5.1 Hz, 1H), 5.33 (s, 2H), 5.09 (s, 4H), 4.99 (s, 4H), 3.12 (m, 6H), 2.48 (t-br, 2H), 2.27 (t, J = 6.3 Hz, 4H) ppm; ¹³C NMR (75 MHz, CDCl₃): δ 165.92, 161.26, 158.99, 152.04, 146.94, 143.92, 138.29, 136.93, 136.80, 134.42, 130.56, 129.43, 129.12, 129.08, 129.06, 128.83, 128.67, 128.14, 127.70, 124.79, 123.72, 123.17, 117.11, 105.29, 53.99, 53.69, 52.19, 38.15, 37.62 ppm; (+)FABMS (MH⁺) calc. 1006, obs. 1006.

tren(CAM)₂(1,2-HOPO) (2)

Bn-tren(CAM)₂(1,2-HOPO) (11) (2.580 g, 2.565 mmol) was dissolved in 1:1 HCl:CH₃COOH. The mixture was stirred overnight and the solution evaporated to dryness. The product was recrystallized by dissolving the crude residue in minimum amount of CH₃OH, and dripped slowly into 50mL of stirring ether in a polypropylene tube. The slurry was then centrifuged and the ether decanted to obtain the deprotected ligand (2) as a white solid. Yield: 68%. Mp: 152-155 °C. ¹H NMR (300 MHz, DMSO): δ 9.12 (m, 3H), 7.39 (dd, J = 6.9, 2.1 Hz, 1H), 7.33 (d, J = 7.2 Hz, 2H), 6.94 (d, J = 6.9 Hz, 2H), 6.70 (t, J = 7.8 Hz, 2H), 6.59 (dd, J = 9, 1.5 Hz, 1H), 6.40 (dd, J = 6.9, 1.5 Hz, 1H), 3.73 (m, 6H), 3.37 (m, 6H), ppm; ¹³C NMR (75 MHz, DMSO-d₆): δ 170.13, 157.48, 149.31, 146.20, 137.18, 119.09, 118.16, 117.71, 115.07, 104.38, 51.51, 33.85 ppm; (+)FABMS (MH⁺) calc. 556, obs. 556; Anal. Calcd. (Found) for C₂₆H₂₉N₅O₉·HCl·H₂O: C, 49.64 (49.86); H, 5.13 (5.44); N, 11.13 (10.84).

[Fe^IIItren(CAM)₂(1,2-HOPO)]²⁻ ([Fe^III(2)]²⁻)

Tren(CAM)₂(1,2-HOPO) (2) (24.3 mg, 0.041 mmol) was dissolved in methanol (10 mL) and degassed. Fe(acac)₃ (14.5 mg, 1 equiv.) was dissolved in methanol (5 mL) and added to the ligand solution followed by 3 equivalents of KOH (0.095 M in MeOH). The solution was degassed and stirred under N₂ overnight. The iron complex solution was condensed (not to dryness) and precipitated with ether. The purple solid was filtered. Yield 20.1 mg (81 %). Mp > 300 °C. (−)ESMS (M-H⁻) calc. 607, obs. 607; Anal. Calcd. (Found) for [C₂₆H₂₄N₅O₉Fe]⁻² 2K⁺·3H₂O: C, 42.29 (42.08); H, 4.10 (4.19); N, 9.48 (9.12).

tren(CAM)(1,2-HOPO)₂ (3)

The bis-substituted intermediate (9) confirmed by ¹H NMR: (300 MHz, CDCl₃): δ 7.42 (t-br, 2H), 7.24 (m-br, 2H), 7.12 (t, J = 6.9 Hz, 2H), 6.42 (d, J = 9 Hz, 2H), 6.10 (d, J = 6.6 Hz 2H, ArH), 5.21 (s, 4H, CH₂), 3.15 (m, 6H), 2.29 (m, 4H), 2.04 (m, 2H) ppm.

Yield of 12 (clear oil): 59%; ¹H NMR (300 MHz, CDCl₃): δ 8.08 (t, J = 6.3 Hz, 1H), 7.47-6.93 (m, 24H), 6.85 (t, J = 8.1 Hz, 2H), 6.58 (d, J = 9.3 Hz, 2H), 6.08 (d, J = 6.6 Hz, 2H), 5.30 (s, 4H), 5.13 (s, 2H), 4.90 (s, 2H), 3.14 (m, 4H), 3.05 (m, 2H), 2.42 (m, 4H), 2.15 (m, 2H) ppm; ¹³C NMR (75 MHz, DMSO-d₆): δ 170.10, 160.69, 157.49, 146.22, 141.46, 137.22, 130.56, 129.43, 129.12, 129.08, 129.06, 128.83, 128.67, 128.14, 127.70, 120.05, 118.15, 117.70, 115.03, 104.38, 51.07, 33.89 ppm; (+)FABMS (MLi⁺) calc. 923, obs. 923.

Yield of 3 (white powder): 67%. Mp: 161-163 °C. ¹H NMR (300 MHz, DMSO): δ 9.15 (m, 3H), 7.40 (m, 3H), 6.95 (d, J = 6.9 Hz, 1H), 6.70 (t, J = 7.8 Hz, 1H), 6.60 (dd, J = 7.5, 1.8 Hz, 2H), 6.43 (dd, J = 5.4, 1.5 Hz, 2H), 3.70 (m, 6H), 3.45 (m, 6H) ppm; ¹³C NMR (75 MHz, DMSO-d₆): δ 161.40, 158.19, 150.04, 146.92, 142.16, 137.92, 105.08, 51.70, 34.61 ppm; (+)FABMS (MH⁺) calc. 557, obs. 557; Anal. Calcd. (Found) for C₂₅H₂₈N₆O₉·HCl·H₂O: C, 47.65 (47.83); H, 4.96 (5.16); N, 13.30 (12.85).

Yield of [Fe^III(3)]¹⁻ (bluish purple solid): 70 %. Mp > 300 °C. (−)ESMS (M-H) calc. 608, obs. 608. Anal. Calcd. (Found) for [C₂₅H₂₄N₆O₉Fe]⁻¹ K⁺·2H₂O·CH₃OH: C, 43.65 (43.49); H, 4.50 (4.19); N, 11.75 (11.35).

tren(CAM)₂(Me-3,2-HOPO) (5)

Yield of 13 (clear oil): 77%; ¹H NMR (300MHz, CDCl₃): δ 7.84 (t-br, 2H), 7.79 (t-br, 1H), 7.64 (t, J = 4.8Hz, 2H), 7.08-7.50 (m, 29H), 6.99 (d, J = 7.2 Hz, 1H), 6.64 (d, J = 7.2 Hz, 1H), 5.29 (s, 2H), 5.14 (s, 4H), 5.04 (s, 4H), 3.52 (s, 3H), 3.16 (q-br, J = 6 Hz), 3.10 (t, J = 4.0 Hz, 2H), 2.31 (s-br, 6H) ppm; ¹³C NMR (300MHz, CDCl₃):δ 164.8, 162.9, 159.0, 151.3, 146.2, 145.7, 136.2, 136.0, 131.8, 130.5, 128.4, 128.2, 127.8, 127.3, 127.2, 123.9, 122.5, 116.3, 104.3, 75.8, 74.2, 70.7, 52.1, 51.7, 45.7, 37.2, 36.8 ppm; (+)-FABMS: m/z: 1020.5 (MH+).

Yield of 5 (white powder): 73%. Mp: 115-118 °C. ¹H NMR (400MHz, DMSO-d₆): δ 12.15 (s-br, 2H), 11.35 (s-br, 1H), 10.06 (s-br, 1H), 9.26 (s-br, 1H), 9.01 (t-br, 2H), 8.68 (t-br, 1H), 7.27 (d, J = 7.6Hz, 2H), 7.14 (d, J = 7.2Hz, 1H), 6.92 (d, J = 7.6 Hz, 2H), 6.68 (t, J = 8.0 Hz, 2H), 6.45 (d, J = 7.2 Hz, 1H), , 3.72 (m-br, 6H), 3.46 (m 9H) ppm; ¹³C NMR (400MHz, DMSO-d₆):δ 172.0, 170.1, 166.1, 158.1, 149.4, 147.5, 146.2, 127.7, 119.1, 118.1, 117.7, 117.0, 115.0, 102,9, 51.6, 45.4, 36.9, 33.8, 21.1 ppm; (+)-FABMS: m/z: 570.5 (MH+); Anal. Calcd (Found) for C₂₇H₃₁N₅O₉·2H₂O (FW 605.59): C, 53.55 (53.28); H, 5.83 (5.95); N, 11.56 (11.21).

Yield of [Fe^III(5)]²⁻ (reddish purple solid): 80%. Mp > 300 °C. (−)FABMS: m/z: 621 [(HM)⁻]; Anal. Calcd (Found) for K₂[FeC₂₇H₂₆N₅O₉]·3H₂O (FW 752.61): C, 43.09 (42.78); H, 4.29 (4.12); N, 9.31 (9.01).

tren(CAM)(Me-3,2-HOPO)₂ (6)

Yield of 14 (clear oil): 65%. ¹H NMR (400 MHz, CDCl₃): δ 1.64 (s, 6H), 3.08 (s, br, 4H), 3.10 (s, br, 4H), 3.58 (s, br, 2H), 3.61 (s, br, 2H) 5.16 (s, 4H), 5.61 (s, 4H), 6.54 (d, J = 7.2 Hz, 2H), 7.06 (d, J = 7.2 Hz, 2H), 7.29-7.43 (m, 20H), 7.61 (d, J = 8.4 Hz, 1H), 7.82 (d, J = 8.4 Hz, 1H), 8.24 (m, br, 3H) ppm. (+)-FABMS: m/z 946.0 (MH⁺). Anal. Calcd (Found) for C₅₅H₅₆N₆O₉: C, 69.90 (70.17); H, 5.97 (6.22); N, 8.89 (8.75).

Yield of 6 (off white solid): 88%. Mp: 103-105 °C. ¹H NMR (500 MHz, DMSO-d₆): δ 3.46 (br, 12H), 3.70 (s, br, 6H), 6.44 (d, J = 7.0 Hz, 2H), 6.68 (t, J = 8.0 Hz, 1H), 6.93 (d, J = 7.5 Hz, 1H), 7.14 (d, J = 7 Hz, 2H), 7.26 (d, J = 7.5 Hz, 1H), 8.67 (br, 2H), 8.98 (br, 1H), 9.91 (br, 1H), 12.13 (br, 1H) ppm. ¹³C NMR (500 MHz, DMSO-d₆): δ 34.03, 36.90, 51.51, 102.81, 115.01, 116.88, 117.70, 118.09, 127.69, 146.18, 147.54, 149.34, 158.05, 166.08, 170.17 ppm. (+)-FABMS: m/z 585 (MH⁺). Anal. Calcd (Found) for C₂₇H₃₂N₆O₉·0.5HCl·1MeOH·3H₂O: C, 47.56 (47.89); H, 6.06 (6.02); N, 11.88 (11.81).

Yield of [Fe^III(6)]¹⁻ (red solid): 85%. Mp > 300 °C. (−)-ESMS: m/z 636.1 ([M+2H]⁻). Anal. Calcd (Found) for C₂₇H₂₈FeKN₆O₉·2HCl·0.5H₂O·MeOH: C, 39.00 (39.01); H, 4.09 (3.99); N, 9.75 (9.59). UV-VIS: λ = 526 nm (ε = 3500 M⁻¹.cm⁻¹), 413 nm (ε = 5000 M^{− 1}.cm⁻¹). IR: ν 1637 m, 1551 s, 1529 s, 1457 m, 1294 m, 1240 s, 1225 s, 1159 w, 743 m, 717 m, cm⁻¹.

Fluorescence quenching binding assay

Fluorescence quenching of recombinant Scn was measured on either a Jobin Yvon fluoroLOG-3 fluorimeter or a Cary Eclipse fluorescence spectrophotometer. A 5 nm slit band pass for excitation and a 10 nm slit band pass for emission was used with a high voltage detector. An excitation wavelength of λ_exc = 281 nm was used and emission was collected at λ_em = 340 nm. Measurements were made at a protein concentration of 100 nM in buffered aqueous solutions, plus 32 μg/mL ubiquitin (Sigma) and 5% DMSO. Fluorescence values were corrected for dilution upon addition of substrate. Fluorescence data were analyzed by nonlinear regression analysis of fluorescence response versus substrate concentration using a one-site binding model as implemented in DYNAFIT.²⁹ All standard binding experiments were done at pH 7.4 using TBS aqueous buffer. Control experiments were performed to ensure the stability of the protein at experimental conditions, including the dilution and the addition of DMSO and Ubiquitin.

Substrate solutions preparation

Metal complex solutions were freshly prepared in situ. An aliquot of a DMSO stock solution of the free ligand (4 mM, 25 μL) and metal salt (1 eq.) were combined, vigorously shaken and diluted with TBS buffer to form the metal complex at a concentration of 0.1 mM. The solutions were equilibrated for 1h and diluted to a final concentration of 6 μM in 5% DMSO, aqueous buffer (pH 7.4). Apo ligand solutions were prepared analogously without the addition of a metal salt.

Structure solution and refinement

X-ray data sets were collected either at the Advanced Light Source ([Fe^IIItren(CAM)₂(1,2-HOPO)]²⁻ and [Fe^IIItren(CAM)(1,2-HOPO)₂]¹⁻) or on a SMART diffractometer ([Fe^IIItren(1,2-HOPO)₃]⁰). The data sets were collected, solved and refined as previously described.^41-45 Structural overlays were calculated in MERCURY.²²

[Fe^IIItren(CAM)₂(1,2-HOPO)]²⁻

Crystals of ferric tren(CAM)₂(1,2-HOPO) were grown from a solution of complex in DMF diffused with a mixture of diethyl ether and DMF. The complex crystallized as pink plates. Two potassium counterions, a DMF molecule, and four waters (three of which were disordered) were found in the asymmetric unit.

[Fe^IIItren(CAM)(1,2-HOPO)₂]¹⁻

Crystals of ferric tren(CAM)(1,2-HOPO)₂ were grown from a solution of complex in DMF diffused with a mixture of diethyl ether and DMF. The complex crystallized as blue rhombohedral plates. A potassium counterion and two disordered DMF molecules were found in the asymmetric unit.

[Fe^IIItren(1,2-HOPO)₃]⁰

Crystals of ferric tren(1,2-HOPO)₃ were grow from a saturated solution of complex in dichloromethane by evaporation. The complex crystallized as bluish plates. Two disordered dichloromethane molecules were found in the asymmetric unit.

Computational Methods

Computational studies were conducted at the Molecular Graphics and Computation Facility, College of Chemistry, University of California, Berkeley. To determine the quadrupole moments, Θ_zz, the aromatic structures were geometry optimized and characterized via frequency calculations at the RHF/6-311G** level of theory in the Gaussian 03³¹ package. To determine the aromatic-cation interaction energies the aromatic structures, the sodium cation, and the aromatic-cation complexes were characterized via a frequency calculation at the MP2/6-311++G** level of theory and the aromatic-cation interaction energies corrected for basis set superposition error (BSSE) with the counterpoise method in the Gaussian 03³¹ package. In the aromatic-cation calculations the sodium ion was fixed at a distance of 2.47 Å above the centroid of the aromatic unit.