Discovery of Novel Haloalkane Dehalogenase Inhibitors

Abstract

Haloalkane dehalogenases (HLDs) have recently been discovered in a number of bacteria, including symbionts and pathogens of both plants and humans. However, the biological roles of HLDs in these organisms are unclear. The development of efficient HLD inhibitors serving as molecular probes to explore their function would represent an important step toward a better understanding of these interesting enzymes. Here we report the identification of inhibitors for this enzyme family using two different approaches. The first builds on the structures of the enzymes' known substrates and led to the discovery of less potent nonspecific HLD inhibitors. The second approach involved the virtual screening of 150,000 potential inhibitors against the crystal structure of an HLD from the human pathogen Mycobacterium tuberculosis H37Rv. The best inhibitor exhibited high specificity for the target structure, with an inhibition constant of 3 μM and a molecular architecture that clearly differs from those of all known HLD substrates. The new inhibitors will be used to study the natural functions of HLDs in bacteria, to probe their mechanisms, and to achieve their stabilization.

INTRODUCTION

Haloalkane dehalogenases (HLDs; EC 3.5.1.8) are enzymes that catalyze the hydrolytic cleavage of carbon-halogen bonds in alkyl halides (Fig. 1) with a wide range of potential applications in biocatalysis, biodegradation, biosensing, decontamination, and cell imaging (1). In structural terms, they belong to the α/β-hydrolase superfamily (2–4). The HLDs have broad substrate specificity, enabling them to catalyze conversion of diverse chlorinated, brominated and iodinated alkanes, alkenes, cycloalkanes, alcohols, epoxides, carboxylic acids, esters, ethers, amides, and nitriles (5, 6).

FIG 1.

General scheme of the reaction mechanism of HLDs. Enz, enzyme.

The first known members of this family were isolated from the bacteria Xanthobacter autotrophicus GJ10, Sphinghomonas paucimobilis UT26, and Rhodococcus rhodochrous NCIMB13064 (3, 7, 8), which colonize environments that have been heavily contaminated with halogenated pollutants, such as 1,2-dichloroethane, 1,2,3,4,5,6-hexachlorocyclohexane, and 1-chlorobutane. In these microorganisms, the HLDs were found to be components of metabolic pathways that enable the microbes to utilize otherwise toxic haloalkanes as their sole source of carbon and energy. The HLDs have been recently discovered in a wider range of organisms, including symbiotic bacteria such as Bradyrhizobium japonicum USDA110 and Mesorhizobium loti MAF303099 (9, 10), pathogenic bacteria such as Agrobacterium tumefaciens C58 and Mycobacterium spp. (11, 12), and the eukaryotic organism Strongylocentrotus purpuratus (13). Even though HLDs have been studied intensively over the last 25 years and isolated from many different environments and species, most of their biological functions remain elusive. For instance, it is undoubtedly interesting that the plant pathogen Agrobacterium tumefaciens C58 carries the HLD-encoding gene datA on its tumor-inducing plasmid while there is no clear link between HLD activity and tumorigenesis (14). Similarly, the presence of three different HLD genes in the genome of the human pathogen Mycobacterium tuberculosis H37Rv (15) suggests that HLDs are important for its survival, but it is not immediately obvious why an organism that colonizes human tissues would require enzymes that cleave carbon-halogen bonds. One way to study the natural function of an enzyme is to use specific inhibitors. However, no attempts to systematically identify HLD inhibitors have yet been reported. Such molecules would be extremely useful in studies on the natural functions of HLDs in bacteria, but they may also find use in detailed kinetic and mechanistic studies and as enzyme stabilizers. Inhibitors may also facilitate enzyme crystallization by increasing internal structural stability, and noncovalent inhibitors are useful during long-term storage of proteins.

Here we present a systematic search for HLD inhibitors. Two different and complementary search techniques were adopted: a ligand-based approach and a structure-based approach. The ligand-based approach involved the rational design of inhibitors based on the structures of the enzymes' known substrates, while the structure-based approach relied on a virtual screen of candidate inhibitors against an experimentally determined HLD structure, more specifically targeted to the molecules noncovalently bound to the enzyme active site. Both approaches were expected to provide noncovalent competitive inhibitors. The set of molecules selected by these theoretical approaches was tested on inhibitory effects using conventional activity and kinetic assays. The discovered inhibitors show a wide range of binding affinities with interesting selectivity for individual HLDs.

MATERIALS AND METHODS

Preparation of ligands for molecular docking.

The clean drug-like subset of the ZINC database (16) was searched for molecules satisfying the following selection criteria: xlogP value of ≤5, molecular weight between ≥150 and ≤500, number of H bond donors of ≤5, and number of H bond acceptors of ≤10. This yielded 149,662 hit molecules, whose three-dimensional structures were downloaded and filtered to remove those with Tanimoto similarity coefficients in excess of 0.8. The resulting molecular library was enriched with six ligand-based inhibitors that were chosen to act as positive controls. The structures of these six inhibitors were built in Avogadro (17). Input files in Sybyl mol2 format were converted into an AutoDock-compatible format using MGLTools (18).

Preparation of the receptor structure for molecular docking.

The crystal structure of HLD DmbA (PDB ID 2QVB) was selected for use as a receptor in virtual screening (19). Gasteiger charges and AutoDock atom types were assigned using MGLTools, and hydrogen atoms were added using the H++ server at pH 7.5 (20). The hydrogen atom bound to the NE2 atom of His 280 was deleted to reflect the catalytic mechanism of the HLDs. Before performing binding energy calculations, the structure was subjected to energy minimization using the Sander module of AMBER 11 (21) with the ff03.r1 force field (22). The geometry optimization protocol involved performing 250 steepest descent steps followed by 750 conjugate gradient energy minimization steps. The convergence criterion for the energy gradient was set to 0.1 kcal · mol⁻¹ · Å⁻¹. The nonbonded cutoff and dielectric multiplicative constant for electrostatic interactions were set to 50 Å and 4 r_ij, respectively.

Molecular docking, rescoring, and clustering.

The active site of DmbA was selected as the target for molecular docking, which was performed using AutoDock Vina (23). The region of the active site selected for molecular docking was set to 20 × 19 × 20 Å, centered at (20.45; 14.59; 12.77) Å. The center of this region was located between the positions of the nucleophile and the catalytic histidine. The docked conformations were rescored using NNScore 2.0 (24), which evaluates the conformation of a molecule with 20 distinct neural-network scoring functions. The final score was obtained by averaging the scores given by these 20 functions. The docked conformations were clustered according to the common features of their binding modes. The clustering analysis was performed using AuposSOM (25) with the default settings for all parameters other than map_size, which was changed to 6 × 5 in order to increase the maximal number of clusters. A tree representation of the clustering results was generated using Dendroscope (26).

Calculation of binding energies.

The free-energy differences between the bound and free states of the receptor and various ligands were calculated by the molecular-mechanics/generalized-Born surface area (MM/GBSA) method. Free-energy differences were calculated by combining gas-phase energy contributions with solvation free-energy components calculated using an implicit solvent model for each species. Force field parameters for the docked conformations of ligands were prepared using the Antechamber and Prmchk modules of AmberTools 1.5. AM1-BCC charges (27) were assigned to individual atoms of ligands with the antechamber module of AmberTools 1.5. Input topologies for the receptor, ligands, and receptor-ligand complexes were prepared with the Leap module of AMBER 11, using the ff03.r1 force field for proteins and the general amber force field (28) for ligands. The PBradii were set to mbondi2 (29). The structures of receptor-ligand complexes were subjected to two rounds of energy minimization using the Sander module of AMBER 11. The first round involved a short minimization with 250 steepest descent steps followed by 750 conjugate gradient minimization steps. The convergence criterion for the energy gradient was set to 0.1 kcal · mol⁻¹ · Å⁻¹, while the nonbonded cutoff and dielectric multiplicative constant for the electrostatic interactions were set to 50 Å and 4 r_ij, respectively. The second round of minimization was done in an implicit solvent with the following parameters: 100 steps of steepest descent followed by 400 steps of conjugate gradient energy minimization, a nonbonded cutoff of 16 Å, and an interior dielectric constant of 2 (30) and with the generalized-Born model parameter set to 2 (29, 31). The convergence criterion for the energy gradient was set to 0.1 kcal · mol⁻¹ · Å⁻¹. The nonpolar contribution to the solvation energy was computed with the LCPO (linear combination of pairwise overlaps) model (32). Finally, an MM/GBSA refinement of the binding energy was performed on the minimized structure using the Python script MMPBSA.py (33) from AmberTools 1.5, with the settings from the second round of minimization. Finally, a consensus score for each conformation of the docked ligands was calculated by averaging the ranks obtained using MM/GBSA and NNScore 2.0.

Enzyme expression and purification.

Four optimized recombinant genes, linB-His6, dhaA-His6, dbjA-His6, and dmbA-His6, were subcloned into the expression vector pET21b. Escherichia coli BL21(DE3) cells were grown at 37°C in LB medium with ampicillin as a selection marker (final concentration, 100 μg · ml⁻¹) and induced by adding isopropyl-β-d-thiogalactopyranoside (IPTG) to a final concentration of 1 mM once the culture reached an optical density at 600 nm (OD₆₀₀) of 0.4. Overexpression was carried out at 20°C for 8 h. Harvested cells were disrupted by sonication using a Hielscher UP200S sonicator (Teltow, Germany) set at 0.3-s pulses and 85% amplitude. Crude extracts were purified by metallo-affinity chromatography on a 5-ml nickel-nitrilotriacetice acid (Ni-NTA) Superflow column (Qiagen, Germany) as reported elsewhere (34). Eluted fractions containing the desired proteins were dialyzed against 50 mM phosphate buffer. The homogeneity, purity, and expression level of each protein were evaluated by SDS-PAGE; all of the proteins used in subsequent experiments had purities above 95%.

Enzyme activity assay for ligand-based inhibitors.

All reactions were performed at 25°C in 25-ml Reacti flasks sealed with Mininert valves. The reaction mixtures consisted of 10 ml of 100 mM (pH 8.6) glycine buffer and 10 μl of 1,2-dibromoethane (DBE) as the substrate, together with 10 μl of the inhibitor being tested, and were initiated by injection of 200 μl of 0.2-mg · ml⁻¹ enzyme. The progress of the enzymatic reaction was monitored by withdrawing 1-ml samples of the reaction mixture 0, 7, 14, 21, and 28 min after the start of the experiment and promptly quenching the samples by mixing them with 100 μl of 35% nitric acid. The released halide ions in the acid-quenched samples were then complexed with mercuric thiocyanate and ferric ammonium sulfate and determined on a Sunrise microplate reader (TECAN, Austria) at 460 nm (35). The enzymes' dehalogenation activity was quantified as the rate of product formation over time after correction for the rate of abiotic hydrolysis.

Enzyme activity assay for structure-based inhibitors.

Compounds identified by virtual screening as potential DmbA inhibitors and selected for experimental evaluation were purchased in powder form at crystalline purity from MolPort (Molport, Latvia). Library stock solutions were prepared by dissolving 1 mg of the relevant compound to a concentration of 5 mg · ml⁻¹ in 200 μl of 99% dimethyl sulfoxide (DMSO). Serial dilutions of the stock solutions were then prepared using pure DMSO to reach a final inhibitor concentration of 0.039 mg · ml⁻¹. The library of structure-based inhibitors identified by virtual screening was evaluated using a microtiter plate screening assay. Reaction progress was monitored using a modification of Holloway's assay (36). The pH indicator phenol red was added to 1.0 mM HEPES buffer (pH 8.0) to a final concentration of 60 μM and used to prepare reaction mixtures in transparent 96-well microtiter plates. The mixtures comprised 120 μl of buffer with indicator and DBE (9.3 mM final concentration), 15 μl of inhibitor in DMSO, and 15 μl of enzyme. The total volume of each reaction mixture was thus 150 μl. The enzyme-catalyzed haloalkane hydrolysis reaction caused a gradual decrease in the pH of the reaction mixture, which was detected by monitoring the resulting change in absorbance at 550 nm using a FLUOstar Optima spectrometer (BMG Labtech, USA). The level of inhibition achieved with each compound was calculated with reference to control reactions performed in the absence of inhibitor. The average reaction rates observed in three independent experiments were fitted to a logistic dose-response model (equation 1) using Origin 9.1 (OriginLab, USA), and the IC₅₀ for each inhibitor was determined from the inflection point of the corresponding fitted curve.

$$v=\frac{A_1\hspace{0.17em}-\hspace{0.17em}{A}_2}{1\hspace{0.17em}+{\hspace{0.17em}\left(\frac{\left[I\right]}{{\text{IC}}_{50}}\right)}^p}\hspace{0.17em}+\hspace{0.17em}{A}_2$$ (1)

In equation 1, v is the observed reaction velocity, A₂ and A₁ are lower and upper limit, respectively, p is slope, [I] is the concentration of the inhibitor, and IC₅₀ is the inhibitor concentration causing 50% inhibition of enzymatic activity.

Protein stability by differential scanning calorimetry.

Thermal unfolding of 1 mg · ml⁻¹ DmbA was conducted in the absence and presence of 93 μM inhibitor 22 in 1 mM HEPES buffer (pH 8.0). The system's heat capacity was monitored using a VP-capillary differential scanning calorimetry system (MicroCal, USA). Experiments were performed at temperatures of 20 to 80°C, ramping at 1°C · min⁻¹. Baseline subtraction and peak maximum determination were performed using Origin 7.0 (OriginLab, USA) with the DSC plugin provided by MicroCal.

Analysis of binding by isothermal titration calorimetry.

All calorimetry reactions were performed at 25°C in 1 mM HEPES buffer (pH 8.0) containing 10% DMSO, with final enzyme and inhibitor 22 concentrations of 43 μM and 300 μM, respectively. All measurements were performed using a MicroCal VP-isothermal titration calorimetry system (GE Healthcare Life Sciences, Sweden). An inhibitor solution prepared by dissolving 1 mg of compound 22 in 200 μl of 99% DMSO was injected stepwise (28 injections of 10 μl each) into a solution of the enzyme in the HEPES buffer inside the reaction chamber. The injection duration was 20 s, with 150-s intervals between injections. The stirring speed was set to 502 rpm, and the feedback mode was set to high. The calorimetric data were evaluated using a single-site binding model in Origin 7.0 (OriginLab, USA).

Analysis of inhibition mechanism by isothermal titration calorimetry.

The dependence of the rate of DmbA reaction on the concentration of DBE was measured in the presence of 0, 2.5, 5, and 10 μM compound 22. All reactions were performed at 25°C in 1 mM HEPES buffer (pH 8.0) containing 10% DMSO. The substrate solution was prepared by adding 10 μl of DBE to 4 ml of reaction buffer containing an appropriate inhibitor concentration and incubated in a water bath at 37°C for 30 min. The final concentration of substrate for each measurement was determined by gas chromatography equipped with flame ionization detection (GC-FID) and a Cyclosil-B LTM-II capillary column (Agilent, USA). DBE was extracted from substrate solution into methanol containing 1,2-dichloroethane as an internal standard. Injection of 1-μl samples on the column was done at 250°C; the oven temperature was increased from 40 to 220°C at a 30°C · min⁻¹ gradient followed by 6 min at a constant temperature of 220°C. The flow of carrier gas was constant at 2.0 ml · min⁻¹. The kinetic measurements were performed using a MicroCal isothermal titration calorimetry (ITC) system (GE Healthcare Life Sciences, Sweden). The substrate solution was titrated into the measurement cell containing enzyme solution of identical composition to avoid generation of a dilution heat. Each injection increased amounts of a substrate in the reaction cell while pseudo-first-order conditions were maintained. A total of 28 injections of 10 μl, each lasting 20 s, were carried out during titration. Instrumental feedback mode was set to none. The protein concentration used was between 0.9 and 5.0 μM, depending on the amount of inhibitor present during the reaction. The reaction rates reached after every injection (in units of thermal power) were converted to enzyme turnover by using apparent molar enthalpy (ΔH_app) (47):

$$\text{rate}=\frac{d\left[P\right]}{dt}=\frac{1}{\text{V}\hspace{0.17em}\times \hspace{0.17em}\mathrm{\Delta }{H}_{\text{app}}}\times \frac{dQ}{dt}$$

where [P] is the molar concentration of product generated, V is the volume of the solution in the reaction cell, and Q is enzyme-generated thermal power. Apparent molar enthalpy of the DBE conversion by DmbA (ΔH_app = 0.094 kcal · mol⁻¹) was determined in a separate experiment that allowed the reaction to proceed to completion. In the experiment, 6.86 nmol of DBE was fully converted by DmbA, and the total heat of conversion was obtained by integration of the ITC signal (47):

$${\mathrm{\Delta }H}_{\text{app}}=\frac{1}{{\left[S\right]}_{\text{total}}\times V}\times {\int }_{t\hspace{0.17em}=\hspace{0.17em}0}^{t\hspace{0.17em}=\hspace{0.17em}\infty }\frac{dQ\left(t\right)}{dt}\hspace{0.17em}\times \hspace{0.17em}dt$$

where [S] is the molar concentration of substrate converted. The average from the three experiments was used.

Observed reaction rates were globally fitted using Origin 9.0 software (OriginLab, USA) according to equation 2:

$$v=\frac{V_{\max \hspace{0.17em}}\times {\left[S\right]}^n}{K_S^n\times \frac{(1+\left[I\right]/K_i)}{(1+\beta \times \left[I\right]/\alpha \times K_i)}+{\left[S\right]}^n\times \frac{(1+\left[I\right]/\alpha \times K_i+{\left[S\right]}^m/K_{si}^m)}{(1+\beta \times \left[I\right]/\alpha \times K_i)}}$$ (2)

where v is reaction velocity, V_max is limiting maximal velocity, K_s is substrate concentration producing half occupation of the binding site, [S] is substrate concentration, [I] is inhibitor concentration, K_si is dissociation constant of SES complex (substrate inhibition constant), K_i is dissociation constant of enzyme inhibitor (EI) complex, n is Hill coefficient describing cooperative binding of substrate, m is Hill coefficient describing cooperative mode of substrate inhibition, α is the factor by which K_s changes when I occupies the enzyme (interaction factor) or K_i changes when enzyme is occupied by S, and β is the factor by which the productivity (rate of breakdown of ESI complex to EI + P) is affected.

RESULTS

Ligand-based inhibitors.

The first set of inhibitors was rationally designed based on structural similarity to the best known HLD substrates. Due to their similarity with substrates, these molecules are expected to bind to the enzyme active site and competitively affect the enzymatic reaction. Since HLDs cannot cleave carbon-fluorine bonds (37), we selected fluorinated analogues of common HLD substrates as potential noncovalent inhibitors. The set of ligand-based fluorinated analogues consisted of 1-fluoropropane (FPE), 1-fluorohexane (FHE), 1-fluorocyclohexane (FCH), and 1,3-difluoropropane (DFP). This set was augmented with two other halogenated compounds, bromocyclopropane (BCP) and 1-iodo-2,2-dimethylpropane (NPI), which are also very similar to known HLD substrates but do not readily undergo bimolecular nucleophilic substitution in the active sites of these enzymes.

We assayed the inhibitory properties of these six ligand-based inhibitors (Fig. 2A) against four widely studied HLDs—DbjA from Bradyrhizobium japonicum USDA110 (10), DhaA from Rhodococcus rhodochrous NCIMB13064 (38), DmbA from Mycobacterium tuberculosis H37Rv (15), and LinB from Sphingobium japonicum UT26 (34)—using DBE as the test substrate. Testing concentrations were close to solubility limits (5 to 15 mM), and all of the ligand-based inhibitors exhibited significant effects on most of selected HLDs. Interesting variability in the specificity of the tested dehalogenases to ligand-based inhibitors was indicated. The widest specificity was observed for LinB, which was significantly inhibited by all of the tested inhibitors. On the other hand, narrow specificity was observed for DmbA, whose activity was significantly reduced by only one of the tested inhibitors (Fig. 3; see also Table S1 in the supplemental material). FCH was the inhibitor with the broadest impact, reducing the activities of all of the tested enzymes between 40% and 90%. DFP was the weakest inhibitor, reducing activity by no more than 20% for any of the tested enzymes. The strongest and the most specific inhibition was observed for FHE, which caused complete inhibition of DbjA when present at its saturated concentration. For this best case, the subsequent experiments using descending concentrations of FHE were performed to provide an exact IC₅₀ of 2.7 mM (see Fig. S1). A trend relating specificity of the inhibitors and substrates can be deduced from the correlation matrix comparing the enzymes and theirs inhibition levels (see Table S2). While the more inhibited LinB, DhaA, and DbjA belong to substrate specificity group 1, the least inhibited DmbA belongs to the substrate specificity group 2 and the same trend can be observed from the correlation table, where correlation coefficients for LinB, DhaA, and DbjA are significantly higher than for DmbA.

FIG 2.

Chemical structures of ligand-based (A) and structure-based (B) HLD inhibitors. While the ligand-based inhibitors are structurally similar to known substrates, the structure-based inhibitors differ strongly from any substrate of the target enzymes. FPE, 1-fluoropentane; FHE, 1-fluorohexane; FCH, 1-fluorocyclohexane; DFP, 1,3-difluoropropane; BCP, bromocyclopropane; NPI, 1-iodo-2,2-dimethylpropane. A complete list of structure-based inhibitors and corresponding PubChem codes can be found in Table S3 in the supplemental material.

FIG 3.

Effects of ligand-based inhibitors on the activity of DmbA (gray), DbjA (white), LinB (black), and DhaA (cross hatched). Each value represents the average from at least three independent experiments; error bars indicate standard deviations. In each case, 100% activity corresponds to the enzyme's activity toward DBE in the absence of inhibitors. FPE, 1-fluoropropane (8.6 mM); FHE, 1-fluorohexane (7.5 mM); FCH, 1-fluorocyclohexane (8.8 mM); DFP, 1,3-difluoropropane (11.3 mM); BCP, bromocyclopropane (14.9 mM); NPI, 1-iodo-2,2-dimethylpropane (7.7 mM).

Structure-based inhibitors.

Since the ligand-based approach provided only weak inhibitors with a millimolar range of effective concentration, we decided to apply a structure-based approach using virtual screening. DmbA was selected as the target enzyme because (i) its crystal structure has been solved at a high resolution, (ii) it originates from the pathogenic bacterium Mycobacterium tuberculosis H37Rv, which primarily colonizes human tissues and whose role in catalysis of dehalogenation reactions is not obvious (15, 39, 40), (iii) genetic engineering of mycobacteria is complicated due to the extremely slow growth and high pathogenicity of the organisms, and (iv) it was the enzyme least affected by the ligand-based inhibitors.

In total, we docked 142,662 structurally diverse molecules into the active site of DmbA. The predicted binding energies ranged from −10.7 to 43.8 kcal · mol⁻¹. The 10,000 molecules with the lowest binding energies, ranging from −10.7 to −7.8 kcal · mol⁻¹, were selected for further investigation. Improved binding-energy estimates for this set of molecules were obtained by the molecular-mechanics/generalized-Born surface area (MM/GBSA) method, and the corresponding enzyme-ligand complexes were rescored using NNScore 2.0. The ligands were divided into 30 separate clusters based on their interactions with individual active-site residues, with each cluster containing between 117 and 691 molecules. One hundred molecules were selected from the best-ranked hits, with the number of molecules taken from individual clusters proportional to the cluster size. Each cluster was represented by at least one molecule to ensure that the selected set reflected the diversity of identified interaction patterns. The 100 selected molecules were then assessed to determine their availability for experimental testing; the DmbA complexes of those found to be available were inspected visually using PyMol (41) to determine whether the ligands effectively blocked the active site of DmbA (see Fig. S2 in the supplemental material). This visual analysis revealed that the selected molecules have structures very different from those of previously known HLD ligands (Fig. 2B). Additionally, all six ligand-based inhibitors were added to the virtual screening library as positive hits. However, none of the molecules ranked in the final top 100 list; the best, 1-chloro-2,2-dimethylpropane, ranked as the 5,392th molecule. Lower ranking may have following origins: first, the dissociation constants of substrates are millimolar, and second, the ligand-based molecules are much smaller, resulting in formation of possibly lower numbers of interactions than the structure-based molecules.

Twenty-five of the compounds identified by virtual screening, one from each cluster whenever possible, were tested experimentally. Five clusters were not represented because the corresponding compounds were not commercially available. IC₅₀s were determined for each of the 25 compounds by measuring their effects on the rate of DBE hydrolysis catalyzed by DmbA at various inhibitor concentrations. Any molecule with an IC₅₀ below 1,000 μM was defined as an inhibitor (Fig. 4). Using this threshold, 17 of the 25 tested compounds were found to be inhibitors, giving a hit rate of 68%. All of the compounds identified by virtual screening had structures that differed strongly from the enzymes' known substrates, and their binding affinities greatly exceeded those of both the enzymes' native substrates and the best ligand-based inhibitors. Kinetic analysis revealed that all of these compounds were partial inhibitors (see Fig. S3). By partial inhibition we understand the cases when enzyme inhibitor complexes maintain a certain level of catalytic activity. The decreased activity consolidated at an equilibrium where any further increase in inhibitor concentration did not significantly affect the reaction rate.

FIG 4.

Observed IC₅₀s for the structure-based inhibitors. All compounds with IC₅₀s below 1,000 μM were assigned as inhibitors.

Significant correlation between the predicted docking scores, MM/GBSA results, and inhibition constants were not observed. The examined inhibitors did not vary significantly in the predicted energies, while all molecules ranked in the top 100. The process of molecule identification used in this study cannot predict whether the discovered inhibitor will be capable of complete or partial inhibition. More rigorous approaches involving molecular dynamics would be necessary for predictions of such quantities with a certain degree of confidence. Implementing additional computational steps into the workflow would be possible at the expense of ease and throughput of the method.

The binding of the best structure-based inhibitor was studied in detail by molecular docking and isothermal titration calorimetry (ITC). Molecular docking revealed a binding mode where the triazole part of the molecule is localized in between the halide-stabilizing residues (Asn39 and Trp110) at the distances enabling formation of a hydrogen bond (Fig. 5A; see also Fig. S4 in the supplemental material). A similar binding motif has been observed in the crystal structure of the HLD-based HaloTag with its stabilizer, where the molecule has a tetrazole part located at the similar position (42). Analysis of the enzyme-inhibitor complex by ITC indicates the single binding site model with an n of 0.91 ± 0.01, K_d of 3.37 ± 0.12 μM, ΔH of −5.24 ± 0.12 kcal · mol⁻¹, and ΔS of 8.14 ± 0.12 cal · mol⁻¹ · °C⁻¹, where n is number of binding sites, K_d is dissociation constant, ΔH is a change in enthalpy, and ΔS is a change in entropy (Fig. 5B).

FIG 5.

Molecular docking, calorimetry, and kinetic analysis of binding of the best structure-based inhibitor 22 to DmbA. (A) The most probable binding mode of the best structure-based inhibitor (in gray sticks) in the DmbA active site (gray surface) according to the molecular docking. The residues of catalytic pentade are shown in sticks. (B) Integrated heat change of ITC data for the interaction between inhibitor 22 and DmbA. The solid line represents the best fit using a single site binding model. (C) DmbA inhibition kinetics in the presence of 0 μM (squares), 2.5 μM (triangles), 5.0 μM (diamonds), and 10.0 μM (circles) concentrations of the best structure-based inhibitor. Solid lines represent a global fit to the data according to equation 1. (D) Differential scanning calorimetry data, showing the difference in the melting temperature of DmbA in the absence (dashed line) and in the presence (solid line) of inhibitor 22.

Next, we performed series of kinetic measurements to unveil the inhibition mechanism of the best inhibitor. The steady-state kinetic constants were determined at four different concentrations of the inhibitor 22 (Fig. 5C). Kinetic data were subjected to a global fit analysis testing three standard inhibition models: (i) competitive, (ii) noncompetitive, and (iii) mixed inhibition. The best fit was obtained by using the partial mixed-type inhibition model (equation 2). The mixed-type mechanism supports formation of enzyme-inhibitor complex (EI) but also simultaneous binding of both substrate and inhibitor, resulting in an enzyme-substrate-inhibitor (ESI) complex with reduced production efficiency (β = 0.066 ± 0.018) compared to that of ES complex. The interaction factor value of 0.864 ± 0.073 indicates that a substrate and an inhibitor bind in a cooperative manner. The resulting equilibrium dissociation constant for enzyme-inhibitor complex, 3.58 ± 0.29 μM, matching the value obtained from ITC binding experiment, is 3 orders of magnitude lower than that of the best ligand-based inhibitor.

We also employed the differential scanning calorimetry to investigate the effect of inhibitor 22 on protein stability. The presence of inhibitor 22 at the concentration of 93 μM raised the melting temperature of DmbA from 52°C to 58°C, demonstrating that the inhibitor's binding stabilized the protein significantly (Fig. 5D). The same experimental design with the other five best inhibitors (no. 1, 18, 9, 15, and 8) yielded an increase of melting temperature between 1.2°C and 2.1°C.

We examined also the specificity of the three best inhibitors (no. 22, 1, and 18) by investigating their effects on three extensively studied HLDs: DbjA, DhaA, and LinB. None of the three inhibitors had any discernible effect on the catalytic activities of these enzymes, indicating that the structure-based inhibitors do not bind to other dehalogenases despite of their high sequence identity with DmbA: 68% for LinB, 44% for DhaA, and 41% for DbjA. Molecular docking revealed that compounds 22, 1, and 18 all bound unfavorably to these enzymes, without forming hydrogen bonds to halide-stabilizing residues as in the case of DmbA (see Fig. S5 in the supplemental material). We therefore conclude that the structure-based inhibitors identified by virtual screening against the structure of DmbA are highly specific for this enzyme.

DISCUSSION

Here we have conducted a systematic search for competitive noncovalent inhibitors of HLD enzymes. Two very different approaches were used, the ligand-based approach and the structure-based approach.

The ligand-based approach is focused on candidate inhibitors whose structures resembled those of the enzymes' known substrates. The resulting inhibitors are expected to bind into the active site in a way similar to that of the substrates, but they cannot undergo enzyme-catalyzed dehalogenation because they either contain strong carbon-fluorine bonds or are not amenable to nucleophilic displacement fore tested enzymes, but they exhibited significant inhibition effect only at millimolar concentrations. Such a finding is in a good agreement with mechanistic analysis of the HLD reaction mechanism, showing binding constants in the millimolar range (43). The most potent of the six tested molecules was 1-fluorohexane, which had an IC₅₀ of 5.4 mM toward DbjA. Interestingly, the substrate specificity mirrored the specificity for the ligand-based inhibitors. Enzymes belonging to the same specificity group were also similarly affected by ligand-based inhibitors. The major advantage of the ligand-based inhibitors is their nonspecificity, which allows them to modulate the reactivity of many related HLDs. Nonspecificity is, however, reflected also by their low affinity. Nevertheless, the best ligand-based inhibitor outperformed a recently reported additive that binds to the active site and facilitates crystallization of DatA (K_d > 20 mM) (44). We note that inhibitors with millimolar-range effectiveness are still unsuitable for most of the practical applications, which usually require inhibition constants in the micro- to nanomolar range (45, 46).

The second approach exploited the crystal structure of DmbA from Mycobacterium tuberculosis H37Rv, which had previously been determined to atomic resolution (19). By virtually screening a large library of inhibitors against this structure, we obtained 17 novel inhibitors from 25 tested molecules, representing a 68% hit rate. The determined inhibition constants are up to 3 orders of magnitude lower than that of the best ligand-based inhibitor. All of the compounds identified in this way had structures that differed strongly from the enzymes' known substrates, and their binding affinities greatly exceeded those of both the enzymes' native substrates and the best ligand-based inhibitors. Based on the isothermal titration calorimetry binding analysis, we concluded that the best structure-based inhibitor occupies a single binding site with a dissociation constant of 3.37 ± 0.12 μM. The mechanism of inhibition was studied in more detail by a steady-state inhibition kinetics. The kinetic data indicated that the best structure-based inhibitor follows a hyperbolic mixed-type inhibition pattern with a dissociation constant of 3.58 ± 0.29 μM. The mixed-type inhibition suggests a possibility of simultaneous binding of substrate and inhibitor to the enzyme active site and provides an explanation for the partial inhibition observed with several tested inhibitors. The overall kinetic mechanism for the best inhibitor includes substrate inhibition and enzyme cooperativity. We are aware of the fact that the determined inhibition mechanism is valid for inhibitor 22 only and cannot be extrapolated to other molecules described in this study.

Interestingly, all structure-based inhibitors were strictly specific for DmbA, having no effect on other tested HLDs. This suggests that the structure-based approach may be a general way of obtaining inhibitors that are unique to the targeted protein. Highly specific molecules can be used as molecular probes in chemical biology studies. It could be, for example, useful for a study of the biological role of DmbA, one of three different HLDs in Mycobacterium tuberculosis H37Rv strain K, and the most abundantly expressed proteins after phagocytosis by the human monocytic cell line U-937 (40). Additionally, the best structure-based inhibitor showed strong stabilization effects, raising the melting temperature of DmbA by 6°C. Similar effects have been observed with HaloTag stabilizers (42) and with DatA enzyme during crystallization experiments (44). Increased thermostability of the enzyme, while retaining its activity, represents an interesting strategy for extension of its half-life during industrial biocatalysis.

In summary, 6 ligand-based and 17 structure-based HLD inhibitors show an interesting range of binding affinities and selectivities for individual target proteins. The best inhibitors are expected to find use in the analysis of enzymes' biological role, reaction mechanism, protein stabilization supporting crystallization, long-term storage, and protection during immobilization or industrial process.

Funding Statement

This work was supported by the Grant Agency of the Czech Republic (P503/12/0572), the Ministry of Education of the Czech Republic (LO1214 and LH14027), and the European Regional Development Fund (ICRC CZ.1.05/1.1.00/02.0123 and CZ.1.07/2.3.00/20.0239). MetaCentrum and CERIT-SC are acknowledged for providing access to computing and storage facilities (LM2010005 and CZ.1.05/3.2.00/08.0144). J.B. was supported by the Employment of Best Young Scientists for International Cooperation Empowerment (CZ.1.07/2.3.00/30.0037) project cofinanced by the European Social Fund and the state budget of the Czech Republic.