Structural and biophysical characterization of the Burkholderia pseudomallei IspF inhibitor L-tryptophan hydroxamate

Abstract

The enzyme 2-methylerythritol 2,4-cyclodiphosphate synthase, IspF, is essential for the biosynthesis of isoprenoids in most bacteria, some eukaryotic parasites, and the plastids of plant cells. The development of inhibitors that target IspF may lead to novel classes of anti-infective agents or herbicides. Enantiomers of tryptophan hydroxamate were synthesized and evaluated for binding to Burkholderia pseudomallei (Bp) IspF. The L-isomer possessed the highest potency, binding BpIspF with a K_D of 36 µM and inhibited BpIspF activity 55% at 120 µM. The high-resolution crystal structure of the L-tryptophan hydroxamate (3)/BpIspF complex revealed a non-traditional mode of hydroxamate binding where the ligand interacts with the active site zinc ion through the primary amine. In addition, two hydrogen bonds are formed with active site groups, and the indole group is buried within the hydrophobic pocket composed of side chains from the 60 s/70 s loop. Along with the co-crystal structure, STD NMR studies suggest the methylene group and indole ring are potential positions for optimization to enhance binding potency.

The Gram-negative bacterium Burkholderia pseudomallei (Bp) is the causative agent of melioidosis, also known as Whitmore’s disease.^1–6 This disease is predominately found in southeast Asia and northern Australia, but has also been observed in the Western Hemisphere, in locations such as Puerto Rico, an unincorporated territory of the United States.⁶ The bacteria spreads when humans, animals, and plants come into contact with contaminated water and soil.⁶ The Center for Disease Control and Prevention considers B. pseudomallei a Tier 1 pathogen and potential bioterrorism agent that could lead to a large-scale public health crisis.⁴ Current treatments of melioidosis requires 3–6 months of intravenous administration of ceftazidime every 6–8 h or meropenem every 8 h, followed by oral antimicrobial therapy with trimethoprim-sulfamethoxazole or doxycycline.^2,3 Ceftazidime resistance has been observed in patients following chemotherapy or in cases of relapse.¹ B. pseudomallei is intrinsically resistant to a multitude of antibiotic classes, such as β-lactams, fluoroquinolones, and clavulanic acid, due to target mutation.⁵

The methylerythritol phosphate (MEP) pathway, a seven step biosynthetic pathway contributing to the synthesis of isoprenoids, was discovered in 1993 and is a promising target for novel anti-infective agents as well as herbicides (Figure 1).^7–9 The seven step pathway synthesizes the two five carbon isoprenoid precursors, dimethylallyl pyrophosphate (DMAPP) and isopentyl pyrophosphate (IPP).⁹ The MEP pathway is essential in several pathogenic organisms such as Plasmodium falciparum, Mycobacterium tuberculosis, Escherichia coli, Burkholderia pseudomallei, and others.^10–14 Since the MEP pathway is absent in mammals, which use the mevalonate (MVA) pathway to synthesize isoprenoids, ¹⁵ inhibitors targeting the MEP pathway may cause fewer issues, such as host toxicity. One example of such an inhibitor is fosmidomycin, which targets MEP enzyme IspC and has demonstrated its potential as a therapeutic agent against the pathogen P. falciparum.^16,17

Fig. 1.

Methylerythritol phosphate (MEP) pathway.^7–9.

The fifth enzyme in the pathway, 2-C-methyl-D-erythritol 2,4-cyclodiphosphate synthase (IspF), catalyzes the cyclization of 4-diphosphocytidyl-2-C-methyl-D-erythritol 2-phosphate (CDP-MEP) to 2-C-methyl-D-erythritol 2,4-cyclodiphosphate (MEcPP).¹⁹ This reaction involves the nucleophilic attack of the β-phosphate of the CDP moiety by the phosphate of the MEP moiety, producing the products, MEcPP and CMP.^14,20 IspF has an overall sequence identity of about 40% across species, while residues involved in substrate binding, catalysis, and enzymatic function are highly conserved. Due to the highly conserved active site, the development of inhibitors targeting BpIspF may be used as starting points for the development of inhibitors targeting other organisms that use the MEP pathway. Structural studies have revealed the IspF enzyme exists as a homotrimer with active sites at the protomer-protomer interfaces (Figure 2A).^{10,14, 21} Each active site can be divided into three sub-pockets (Figure 2B), including MECP/Phosphate, ribose, and cytosine binding sites.²⁰ The MECP/Phosphate pocket contains a divalent zinc ion, Zn²+, which is necessary for catalysis and interacts with the β-phosphate of the CDP moiety.^14,21 The MECP site is capped by a lipophilic loop, which, based on crystal structure data, is often unstructured in the absence of a bound ligand.¹⁴ Reported examples of IspF inhibitors include thiozolopyrimidine and arylbisulfonamide compounds, which were reported to inhibit A. thaliana and P. falciparum IspF activity at low micromolar concentrations.^22–24 The bis-sulfonamides are known metal ion binders, which would facilitate targeting the IspF active site zinc ion.^23,24

Fig. 2.

Schematic diagrams based on crystal structures of IspF. A) Ribbon diagram of E. coli IspF (PDB: 1JY8¹⁴ ) highlighting products MEcPP and CMP (magenta-left and right, respectively) occupying the active site located at the protomer-protomer interface and the 60 s/70 s loop (black). Each monomer chain is displayed in a different color (green, white, and red). B) Stick representation of the three different sub-binding pockets (MECP/Phosphate, ribose, and cytosine) of the B. pseudomallei IspF (PDB: 3QHD¹⁰) active site. Structure includes compounds FOL955, which resides in the MECP/Phosphate sub-pocket, and cytidine, which resides in the ribose and cytosine sub-pockets (magenta- FOL955 and cytidine). IspF residues are colored by protomer domain (green or white) and the catalytic zinc ion is indicated by grey sphere. Images created using PyMol.^{18. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)}

Zinc-binding groups (ZBGs) are commonly used in inhibitor design targeting zinc-dependent matrix metalloproteinases (MPPs), enzymes that facilitate the breakdown of connective tissue.^25,26 Hydroxamic acid is an example of a commonly used ZBG.²⁷ Ramsden et al. reported that the hydroxamate group increased the potency for ligands targeting E. coli IspF.²⁸ Specifically, a hydroxamic acid of L-tryptophan was found to possess a K_D of ~ 2 µM, as measured by surface plasmon resonance (SPR). To support the development of new IspF inhibitors and evaluate cross-species inhibition activity, L-tryptophan hydroxamate was synthesized and subjected to biophysical and structural studies to assess its interactions with BpIspF.

In our hands, starting from L-tryptophan methyl ester, the reported synthesis of L-tryptophan hydroxamate required multiple recrystallization steps to obtain pure material.^29,30 Here, we used a concise, three-step synthesis starting with L-tryptophan (Scheme 1). The primary amine of L-tryptophan was protected with benzyl chloroformate, followed by carbodiimide coupling with O-benzylhydroxylamine to provide compound 2. The doubly protected precursor 2 allowed for purification of the product by flash-column chromatography on silica gel. Finally, both protecting groups were removed via hydrogenation of 2 in one step to afford L-tryptophan hydroxamate (3) with an overall yield of 49%. The same method was used to synthesize D-tryptophan hydroxamate.

Scheme 1.

Chemical synthesis of L-tryptophan hydroxamate. D-tryptophan hydroxamate was synthesized from D-tryptophan via this method. Abbreviations: CbzCl = Benzyl chloroformate, Bn = Benzyl.

The inhibition potency of compound 3 against BpIspF was evaluated using an HPLC activity assay.³² After initiating the reaction, enzyme activity was measured by quenching a small aliquot of the reaction every few minutes. Aliquots were subjected to the removal of proteins with a 3000 molecular weight cut-off filter, followed by chromatographic separation of CDP-ME2P and CMP using a C18 Rocket column. The CMP product concentration was calculated based on the observed peak size. In the presence of 120 µM compound 3, 55% inhibition was observed (Supplemental Figure 1).

Isothermal titration calorimetry was used to investigate the binding thermodynamics of BpIspF with L-tryptophan hydroxamate (3) and related analogs at 25 °C. This approach provided a full thermodynamic profile of the binding event (K_b, ∆G°, ∆H°, -T∆S°, and stoichiometry). All tryptophan derivatives were evaluated at physiological pH (7.4), as shown in Table 1. Almost identical binding affinities were observed for 3 with both B. pseudomallei and E. coli IspF, which were primarily enthalpically-driven events. The similar affinities appear in line with the high degree of active site residue identity between both species. Notably, the observed affinity of 3 toward E. coli IspF, as measured by ITC, was 10-fold weaker than that reported previously (K_D ~ 2 µM) using surface plasmon resonance.²⁸ Removing the hydroxamate group entirely, analogs L- and D- tryptophan, resulted in undetectable binding, suggesting the hydroxamate/active site zinc ion interaction is the major energetic driving force of IspF binding. Interestingly, changing the stereochemistry from L- to D-tryptophan hydroxamate resulted in less than a twofold drop in affinity and similar thermodynamic signatures when binding BpIspF. This suggests the stereochemistry is less relevant for specific interaction(s) and/or that new interactions are made that are roughly energetically equal.

Table 1.

Thermodynamic analysis of tryptophan and tryptophan hydroxamates binding to B. pseudomallei and E. coli IspF at 25 °C. Data were analyzed using the manufacturer’s ITC add-on in Origin with a simple 1:1 binding model, which allows the stoichiometry to float and assumes a Hill coefficient of 1.0. *stoichiometry was fixed during data analysis following method of Turnbull, et al.³¹

Compound	pH	KD,app (μM)	N	∆G° (kcal/mol)	∆H° (kcal/mol)	−T∆S° (kcal/mol)
L-tryptophan hydroxamate (3) (vs E. coli)	7.4	22 ± 9	1.2 ± 0.3	−6.4 ± 0.2	−4 ± 2	−3 ± 2
L-tryptophan hydroxamate (3) (vs B. pseudomallei)	6.0	848±2	1*	−4.19 ± 0.01	1.1 ± 0.2	−5.2 ± 0.2
	7.4	36±9	1.5 ± 0.2	−6.1 ± 0.2	−7.4 ± 0.9	1.9 ± 0.9
	8.0	24±3	1.4 ± 0.4	−6.29 ± 0.09	−8.3 ± 0.2	2 ± 1
D-tryptophan hydroxamate	7.4	64±7	0.7	−5.7 ± 0.1	−6.650 ± 0.001	0.9 ± 0.1
L-tryptophan	7.4	>1,000	1*	‒	‒	‒
D-tryptophan	7.4	>1,000	1*	‒	‒	‒

L- and D-tryptophan hydroxamate analogs, which possessed higher affinity against both species of IspF, enabled the determination of the binding stoichiometry. The observed stoichiometries for the closely related analogs ranged from 0.7 to 1.5 against the two IspF species. While detailed investigations into the thermodynamics of IspF/ligand interactions are limited, previously, Thelemann et al. observed a binding stoichiometry of 1.4 for bis-sulfonamide compounds binding Arabidopsis thaliana IspF,²⁴ which falls within the range observed here. Based on the trimeric structure of IspF, the ligand binding stoichiometries were anticipated to be either 1.0 or 3.0. The observed differences in stoichiometry may reflect issues in determining an accurate concentration of the ligands. Here, the extinction coefficient of tryptophan, which does not have the hydroxamate moiety, was used to determine the concentration. Overall, the values suggest a stoichiometry of 1.0, which would imply the presence of cooperativity between binding sites.

The binding thermodynamics of 3 with BpIspF was studied at pH values slightly above and below typical physiological conditions (pH 6.0 and 8.0) to evaluate potential changes in binding due to changes in protonation states of the protein or ligand. Existing crystal structures of BpIspF in complex with ligands have been evaluated under such pH conditions. As compared to binding at pH 7.4, while there was only a negligible change in the binding thermodynamics at pH 8.0, the binding of 3 at pH 6.0 was approximately 25-fold weaker and became entropically driven. The pH dependent change in the binding constant suggests a proton-linked binding event. The two most likely ionizable contributors include His36 and the hydroxamate moiety of the ligand (Figure 3), which have estimated pK_a values in water of 6.0 and 8.0, respectively. It is important to note that the linked protonation complicates the analysis of the underlying thermodynamics, as the enthalpy of the ionizable group, as well as the experimental buffer can contribute to the overall observed change in enthalpy due to changes in protonation upon ligand binding.

Fig. 3.

Stick model of L-tryptophan hydroxamate’s interactions within the BpIspF active site. L-tryptophan hydroxamate (green-carbon, red-oxygen, blue-nitrogen); IspF chain a and b residues (white-carbon, blue-nitrogen, red-oxygen), zinc (gray sphere). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

To gain structural insight into the interactions of 3 with BpIspF, a 1.47 Å resolution structure of the complex was obtained using X-ray crystallography (see supplemental Table 1 and supplemental methods). Similar to many published ligand/IspF co-crystal structures, only one of the three active sites showed full ligand occupancy based on the observed electron density. This observation is consistent with the observed binding stoichiometry of 1.0. It is perhaps unexpected that there are no obvious changes in structure (such as structural rearrangements that preclude binding at the other two sites), which could help explain the apparent allosteric binding mechanism. In fact, several active site residues in the remaining two sites, such as those from the 60 s/70 s loop and Gly35, His36 and Ser37 could not be fully modeled due to lack of electron density. Consequently, it is possible that binding to one site stabilizes the structure in a way that propagates to destabilize binding to the remaining sites. However, such as model may be difficult to determine on crystal structure data alone since IspF structures, particularly apo structure, often possess less well-defined active sites.

As expected, 3 was found located near the zinc ion in the active site; however, the interactions with the metal ion were unexpected, as the oxygen atoms of the hydroxamic acid group did not interact directly with the metal ion, as is observed in most hydroxamate/zinc interactions.²⁶ Instead, 3 interacts with the active site zinc ion using the primary amine, with a distance of 1.9 Å, comparable to the 2.0 Å distances between the zinc ion and IspF residues Asp10, His12, and His44 (Figure 3). Together these four groups adopt a tetrahedral geometry with the metal ion. Notably, the hydroxamate nitrogen of 3 is located 2.9 Å from the zinc in pseudo trigonal bipyramidal coordination. Hydrogen bonds appear between the hydroxamate alcohol group and the side chain of Ser37, as well as the indole nitrogen with the main chain oxygen of Phe63. Finally, the indole ring extends into the MECP/Phosphate pocket, specifically the 60 s/70 s loop, which is frequently unstructured in apo IspF structures. The hydrophobic makeup of this pocket (including Ile59, Phe63, Phe70) complements the hydrophobic ligand. Two water molecules are observed with partial occupancy, which appear to hydrogen bond with main chain amides of His36 and Ser37; however, due to the close overlap with 3 (including the alcohol group), it is not likely they are present at the same time as the ligand.

Saturation transfer difference (STD) NMR is a ligand-based NMR technique for the study of protein ligand interactions and has been applied to study of IspF-ligand interactions.³³ To further analyze the binding of L-tryptophan hydroxamic acid to BpIspF, STD-NMR was performed. Table 2 presents the results relative to the proton of highest saturation transfer. Protons on heteroatoms are not observed in the presence of D₂O due to exchange. In general, the results are consistent with the x-ray crystal structure (Figure 3). The proton α to the carbonyl and the amine (HG) received 50–55% saturation transfer, and likely only received transfer through the proximity to the hydroxamic acid group. The methylene protons (HF), which are directed towards the open ribose pocket, have the least transfer in the compound. The highest transfers were all located on the indole, which fits into the lipophilic pocket. All the protons on the indole (HA through HE) have moderate to high saturation transfer relative to the highest transfer, HA. Protons HC and HE had overlapping resonances and could not be distinguished at any saturation time. The HC and HE protons had an 80–85% relative saturation transfer average. Saturation was slightly lower for HD, which faces back toward the metal ion. The lowest relative saturation transfer on the indole moiety was HB. Interestingly, HB, despite its proximity to HA, is in the deepest part of the 60 s/70 s hydrophobic pocket, where it is not able to fully fill the pocket. Overall, potential modification to improve the affinity of 3 might focus the lower saturation transfer regions, including the methylene, stereocentric carbon, and the hydrophobic cavity filling HB position.

Table 2.

Relative saturation transfer results for L-tryptophan hydroxamate with BpIspF.

Proton (H)	Relative Percent Saturation Transfer
A	100%
B	60–65%
C*	80–85%
D	75–80%
E*	80–85%
F	25–30%
G	50–55%

^*HC and HE had overlapping resonances; values represent their grouped average.

In addition to revealing the atomic level details of the compound 3/BpIspF complex, the high-resolution structure revealed distinct differences in the central chamber of the IspF trimer. Structural studies on E. coli IspF revealed geranyl diphosphate (GPP) bound to IspF’s central cavity (Figure 4).^34,35 The GPP binding site in these E. coli IspF crystal structures is characterized by a salt-bridge between Arg142 and the terminal phosphate on GPP, as well as a deep hydrophobic central chamber that complements the lipophilic GPP tail. NMR and mass spectrometry analysis identified additional diphosphate species, such as isopentyl diphosphate and farnesyl diphosphate, that bind this cavity, suggesting these downstream MEP pathway metabolites may serve a role in feedback regulation.³⁵ However, enzymatic studies by Bitok et al. did not observe direct feedback inhibition of E. coli IspF in the presence of diphosphates, such as GPP or farnesyl diphosphate (FPP).³² Instead, 2C-methyl-D-erythritol 4-phosphate (MEP), the upstream product of IspC, along with downstream product FPP inhibited IspF activity. Figure 4 highlights the significant difference in residue composition in BpIspF. While the arginine residues at the chamber’s entrance are conserved, the central chamber is significantly polar, where almost each residue is more polar/charged, as compared to the E. coli counterpart. This observation suggests the BpIspF central chamber does not likely contribute to MEP/FPP regulation in the same way observed in E. coli. No electron density the presence of downstream metabolites was observed in the 3/BpIspF structure. Instead, a network of water molecules fills the cavity. Furthermore, STD NMR studies with geranyl diphosphate (GPP) did not show significant interaction with BpIspF (supplemental info). These differences in the central chamber of different IspF species and how they relate to regulation may play an important role in better understanding the observed stoichiometry differences with certain IspF inhibitors.

Fig. 4.

Stereoview of a superposition of the central core from E. coli and B. pseudomallei IspF. Two subunits of the trimer are displayed with the third subunit, located in the foreground, not displayed to aid visualization of the core. E. coli IspF (white) with GPP-magenta and manganese ion (blue sphere), BpIspF- blue; PDB IDs: E. coli (1H48)³⁴ and B. pseudomallei (5L03). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

In summary, we report a concise, three-step synthesis of IspF inhibitor L-tryptophan hydroxamate (3), followed by biophysical and structural studies describing the binding interaction with BpIspF. As expected, the hydroxamic acid group is critical to binding, as both L- and D-tryptophan did not show appreciable binding. While 3 possessed the highest binding affinity against BpIspF (K_D,app of 29 ± 5 µM), it was somewhat surprising that changing the stereochemistry to D-tryptophan hydroxamate did not drastically reduce the binding affinity. The x-ray crystal structure of the 3/BpIspF complex revealed ligand binding occurred through an unconventional mechanism where the primary amine of tryptophan, and not the hydroxamic acid oxygen atoms, was observed in coordination with the active site zinc ion. Based on the active site interactions, it is likely that D-tryptophan hydroxamate, which binds less than two-fold weaker than 3, uses a modified or distinct binding mechanism.

The co-crystal structure and STD-NMR analysis of 3 with BpIspF helps guide the development of more potent inhibitors. Modifications to the methylene and/or indole ring may be of particular interest. The latter can consist of modifications to more fully fill the hydrophobic pocket formed from the 60 s/70 s loop. Furthermore, compound 3 exhibited almost identical binding affinities between B. pseudomallei and E. coli IspF, suggesting the compound’s binding determinants rely on the high homology of the active site. This also suggests that development of analogs of 3 may possess activity against other IspF species, broadening its potential therapeutic use.