QM/MM Study of the Fosfomycin Resistance Mechanism Involving FosB Enzyme.

Multidrug-resistant organisms contain antibiotic-modifying enzymes that facilitate resistance to a variety of antimicrobial compounds. Particularly, the fosfomycin (FOF) drug can be structurally modified by several FOF-modifying enzymes before it reaches the biological target. Among them, FosB is an enzyme that utilizes l-cysteine or bacillithiol in the presence of a divalent metal to open the epoxide ring of FOF and, consequently, inactivate the drug. Here, we have used hybrid quantum mechanics/molecular mechanics (QM/MM) and molecular dynamics (MD) simulations to explore the mechanism of the reaction involving FosB and FOF. The calculated free-energy profiles show that the cost to open the epoxide ring of FOF at the C2 atom is ∼3.0 kcal/mol higher than that at the C1 atom. Besides, our QM/MM MD results revealed the critical role of conformation change of Cys9 and Asn50 to release the drug from the active site. Overall, the present study provides insights into the mechanism of FOF-resistant proteins.

Introduction

Fosfomycin (FOF) is a broad-spectrum antibiotic that binds covalently to Cys115 of the UDP-N-acetylglucosamine-3-enolpyruvyltransferase (MurA) enzyme, an important target for antimicrobial agents through peptidoglycan biosynthesis.[1] However, several FOF-modifying enzymes that have been found inactivate the drug, modifying the structure of FOF before it reaches the target.[2−4] Under these conditions, the mechanism of resistance to FOF is related to the inactivation reaction promoted by three metalloenzymes (FosA, FosB, and FosX) and two kinases (FomA and FomB).[5] Particularly, the metalloenzymes are responsible to open the epoxide ring of FOF employing different substrates, as reviewed in previous literature.[4]

In the metalloenzyme group, the FosB enzyme has been extensively used as a model for a better understanding of the FOF resistance mechanism.[6−8] FosB is a divalent metal-dependent thiol-S-transferase implicated in the FOF resistance among many pathogenic Gram-positive bacteria.[8] Besides, it has been widely detected in the chromosomes and plasmids of many low G + C Gram-positive bacteria, including Bacillus subtilis, Bacillus anthracis, Bacillus cereus, Staphylococcus aureus, Staphylococcus epidermidis, and Enterococcus faecium.[8] The coordination of FOF by various divalent metals is similar, and the geometry of metal coordination with bound FOF can be described as highly distorted, five-coordinate trigonal bipyramidal.[9] Thus, the catalytic mechanism of FosB consists of a nucleophilic addition reaction between either l-cysteine (l-Cys) or bacillithiol (BSH) and FOF leading to a modified compound with no antibiotic properties (Figure ).

Figure 1

Reaction catalyzed by the FosB enzyme in the presence of l-Cys or BSH, utilizing a divalent cation M2+.

A previous theoretical study analyzed the catalytic mechanism of the FosA enzyme and described the factors which drive the regioselectivity and chemoselectivity of the reaction.[10] FosA inactivates FOF by the addition of glutathione (GSH) to the oxirane ring of the drug.[11,12] Thus, in contrast to FosB, FosA is a Mn2+-dependent GSH S-transferase and requires K+ for optimal activity. In other words, FosA shows monovalent metal dependence.[13]

More recently, the catalytic mechanism of FOF-resistant kinase A (FomA) has been investigated through a hybrid quantum mechanics/molecular mechanics (QM/MM) approach.[14] The QM/MM approach deals with complex systems by treating the parts of the system in different settings, where the quantum mechanics and the Newtonian counterpart are joined by a buffered region.[15] This approach has been successfully used by our group to investigate many enzymatic reactions[16,17] and, particularly, for biological systems related to antibiotic resistance.[18,19] Herein, we present an extensive semiempirical QM/MM umbrella sampling (US) simulation study to explore the reaction catalyzed by the FosB enzyme.

Results and Discussion

Regiochemistry of the Reaction Catalyzed by FosB

The reaction catalyzed by the enzyme FosB contributes to the emergence of bacterial resistance to the antibiotic FOF since it has its structure modified before reaching its target (the MurA enzyme).[1,20] As substrates, FosB utilizes l-Cys or BSH. FosB from B. cereus, zinc, and l-cysteine–FOF ternary complex is deposited in the protein data bank under the code 4JH8.[9] After replacing Zn2+ for Mg2+ and applying a protocol to obtain the energy minimized system, US simulations were performed to provide structural and energetic insights into the inactivation reaction of FOF by FosB–l-Cys complex, as described in Figure . A total of 60 ps of QM/MM MD simulations at the PM6/MM level was carried out for each US window for the two reaction coordinates (RCs) proposed. It provided a good overlap between adjacent windows (Figure S1, Supporting Information).

Figure 6

Windows in the US simulations were defined as a linear combination of distances d1 – d2 and d1′ – d2′. “d1” describes the nucleophilic attack of the thiol group of l-Cys to the C1 atom of FOF, while “d2” describes the bond breaking of the C1 atom and the oxirane oxygen of FOF; d1′ describes the nucleophilic attack of the thiol group of l-Cys to the C2 atom of FOF, and d2′ describes the opening ring of the epoxide group of FOF at C2.

The free-energy profiles corrected at the wB97XD/6-31G(d,p) level are presented in Figure . The corresponding potential of mean force (PMF) at the PM6/MM level with error bars is shown in Figure S2 of the Supporting Information file. The corrected PMF profile for path1 reaction reveals an exergonic pathway where the product state is 11.5 kcal/mol lower than the reactant state. The theoretical free-energy barrier obtained at the wB97XD/6-31G(d,p) level is 14.7 kcal/mol, which is in good agreement with the experimental kinetic data of FosB from B. cereus, S. aureus, and B. subtilis, reported in the range of 0.9–604 s–1.[8,21] By using the Eyring equation,[22] the activation energy values range from 13.7 to 17 kcal/mol. For path2, the reaction free energy is ∼3.0 kcal/mol higher than that obtained for path1 and less exergonic. This reactional behavior can be supported by experimental observations that show FosB opening the FOF oxirane ring at the C1 atom.[8,23]

Figure 2

PMF for opening FOF oxirane ring at the C1 atom (black line) and C2 atom (red line) at the DFT/MM level. The reaction utilizes l-Cys as a thiol substrate in the presence of Mg2+. The reaction was sampled at the PM6 level and corrected at the wB97XD/6-31G(d,p) level.

However, it is well established that under normal physiological conditions, the opening of the FOF oxirane ring can occur either at C1 or C2 atom.[5,24] It should be highlighted that FosB catalyzes the inactivation reaction of FOF by a nucleophilic attack of the thiol group (from l-Cys or cacillithiol) to the C1 atom of the FOF oxirane ring, while MurA utilizes the thiol group of catalytic Cys115 to attack the C2 atom of FOF and promote the opening of its oxirane ring. Interestingly, nucleophilic substitutions at the C1 atom of FOF are hard to be performed; however, in the presence of the FosB enzyme, the addition reactions at C1 are favored under experimental conditions.[8] Then, by comparing this experimental evidence with our theoretical findings, the regioselectivity of the opening of the FOF oxirane ring by the FosB–l-Cys complex can be explained by the increase of 3.0 kcal/mol in the activation free energy for the nucleophilic attack at the C2 atom of FOF (path2) when compared with the attack at the C1 atom (path1). Also, as reported above, the product state of path2 is less thermodynamically favorable than the product state of path1. The larger thermodynamic driving force for the attack at C1 than for the C2 atom of FOF reflects the larger stabilization of the l-Cys thiol adduct with FOF as a result of unfavorable steric and electrostatic interactions at the active site.

A previous theoretical study carried out by Liao and Thiel (2013) describes the regioselectivity of FosA by analyzing a molecular model for the uncatalyzed reaction of FOF, methanethiol, and two water molecules.[10] According to them, the nucleophilic attack at both C1 and C2 atoms occurs in a concerted step with a very similar activation barrier (∼30 kcal/mol), which agrees with the experimental evidence for the uncatalyzed reaction.[13] Then, our results for the enzymatic reactions reveal the strong influence of the enzymatic environment for this reaction. Based on that, we should expect that the catalytic power of FosB as well as its regioselectivity is related to the preorganization of the enzyme active site. Significant works have been discussed that in water, the free energy of moving from the ground state to the transition state (TS) is larger than in the enzyme. The catalytic groups are preorganized in the correct direction and thus do not have to pay a large preorganization energy.[25−28]

Structural Comparison with X-ray Structures

The RC used during US QM/MM calculations is a linear combination of two atomic distances d1 and d2, which describes the nucleophilic attack to the C1 atom and the opening of the oxirane ring of FOF, respectively. Then, changes of d1 and d2 can be observed by taking representative ensembles of the reactant, TS, and product states along the PMF profile, as can be seen in Figure . For d1, we can observe a decrease from 3.0 Å at the reactant state to 2.1 Å at the TS and 1.7 Å at the end of the reaction, which characterized the S–C bond of l-Cys and FOF in the product state. Simultaneously, for the d2 distance, we can observe an increase from 1.4 Å at the reactant state to 1.8 Å at the TS and 2.4 Å at the product state. Therefore, the inactivation reaction of FOF by the FosB enzyme occurs via a concerted mechanism where the formation of the S–C1 bond (from nucleophilic attack) occurs simultaneously with the breakdown of the C1–O bond (opening of the oxirane ring of FOF), similar to that observed in simulations made for the enzyme FosA.[10]

Figure 3

Representative three-dimensional structures for the reactant (A), TS (B), and product (C), obtained during the reaction simulation. Distances are given in angstroms, Å.

On the inactivation reaction of FOF by the FosB–l-Cys complex, the product state is characterized when the S–C bond is completely formed (1.7 Å). At this point, the thiol group of residue Cys9 interacts with the O atom of the FOF oxirane ring through a hydrogen-bond interaction (1.4 Å), which contributes to a better understanding of how the l-Cys–FOF product releases from the active site of the FosB enzyme. Besides, the proton atom of the thiol group of Cys9 can be transferred to oxirane oxygen of FOF, leading to a reduction of the interaction between the inactivated drug and the Mg2+ ion, improving the release of l-Cys–FOF product from the catalytic site of the FosB enzyme.

Based on the results mentioned above, we fitted the product state from US QM/MM simulations and the crystal structure of PDB 4JH7(9) (FosB from B. cereus with Mn2+ and l-Cys–FOF product). Thus, a root-mean-square deviation (rmsd) of only 0.93 Å was observed, showing that our results are in good accordance with experimental data (Figure S3, Supporting Information). Interestingly, the coordination of antibiotics with the metal center of the active site of metalloenzymes is very similar for a variety of divalent metals, including Ni2+, Mn2+, Co2+, or Zn2+.[6,23] Comparison of the FosB–Mn(II)– FOF and FosB–Mn(II)-l-Cys–FOF product structures revealed a unique trigonal bipyramidal metal–ligand coordination geometry.[6] According to our simulations, the geometry when Mn2+ is replaced with Mg2+ remains the same. Therefore, once the M2+ pocket does not change significantly, the choice of Mg2+ as a divalent ion was motivated by the computational limitations and costs of sampling transition metals at semiempirical or ab initio levels of QM methods.

Finally, crystal structures provided by Thompson and co-workers[9,23] show a highly conserved water molecule hydrogen-bonded to Glu115. Our results showed that water positioned 5.2 Å away from the FOF C1 atom. For the homologue enzyme, it was demonstrated that a water attack at C1 is not suitable.[10] Thus, the functional relevance of this conserved water in the active site of FosB has remained unclear.

Energetic Contributions of Particular Residues on the Catalysis

Experimental evidence suggests that FosB utilizes Tyr39 as a general catalytic base.[7,8] A proton transfer from the thiol group of l-Cys to the hydroxyl group of Tyr39 occurs, and it becomes an essential step to promote the deprotonated l-Cys as a good nucleophile to attack the C1 atom of FOF. Tyr39 and Asn50 likely work in a concerted manner to ionize and stabilize the approaching thiolate during the nucleophilic attack of the antibiotic.[9] It should be noted in our simulations that l-Cys is already considered deprotonated by Tyr39. According to a previous QM/MM study based on the mechanism of the FosA enzyme, the activation of thiol by Tyr39 is not the rate-limiting step.[10] Interestingly, US QM/MM simulations at the PM6/MM level reveals that the Tyr39 residue strongly contributes to stabilizing the TS, thus decreasing the activation energy for the reaction.

In an attempt to understand how the enzymatic environment stabilizes or destabilizes the TS of the reaction catalyzed by FosB, we have analyzed the energetic contribution of key residues around 10 Å of the QM part of the QM/MM model. The energetic contribution of an individual residue to the total energy of a particular structure was computed by using the difference of energies when this particular residue is present and when it is replaced by the Gly residue.[29−31] The average values for the interaction energy and stabilization effects were computed for 400 snapshots from QM/MM US simulations. Thus, it can be observed for the path1 mechanism proposal that the Tyr39 residue has a strong stabilization effect on the TS, while the Arg96 residue has the most destabilizing effect on the TS (Figure ).

Figure 4

Relative stabilization pattern that represents the influence of amino acid residues 10 Å around the QM region on the TS considering the reactant state as the reference. The average values with error bars were computed for 400 snapshots from QM/MM US simulations.

The total interaction energy calculated for the attack at the C1 atom (path1) is −11.91 kcal/mol, which means a stabilizing effect of TS by the enzymatic environment. On the other hand, the total interaction energy calculated for the attack at the C2 atom (path2) is 11.82 kcal/mol, meaning a strong destabilizing effect of the enzymatic environment on the TS when this pathway is described. Particularly, His7 residue, which also coordinates the Mg2+ ion, presented a strong destabilizing effect when the nucleophilic attack occurs at the C2 atom of FOF (path2). However, for path1, a suitable favorable interaction was found.

Finally, the residue Asn50, which appears with a weak stabilizing effect (−0.45 kcal/mol) at the TS, seems to acquire an important role when the approximation of Cys9 to the oxirane oxygen of FOF occurs. At the product state, it can be observed that the Asn50 residue rotates to stabilize the Cys9 residue, which may indicate that Cys9 can donate a proton to the oxirane oxygen of FOF at the end of the reaction. It should be highlighted, the acid character of the Cys9 residue, once that in the absence of FOF, cocrystallization of FosB with either l-Cys or BSH results in a disulfide bond between these substrates and Cys9 residue.[23]

Computational Methods

Molecular Model

The crystal structure of FosB from B. cereus with zinc and l-cysteine–FOF ternary complex (PDB code: 4JH8)[9] was used as the initial model system. The first step after acquiring the enzyme model was to replace the divalent cation Zn2+ for Mg2+, once the metal activity assays indicated that Mn2+ and Mg2+ are likely to be the relevant cofactors under physiological conditions.[8] The H++ server (http://biophysics.cs.vt.edu/) was used to add hydrogen atoms to the enzyme. The tLeap module of the AMBER16 package[32] was used to solvate the systems in a cubic box with explicit TIP3P[33] water molecules. The GAFF[34] and ff14SB[35] force fields were used to build MM parameters to the ligands and protein, respectively. The RESP charges for the FOF and l-Cys were calculated on the Gaussian 09 software[36] using the Hartree-Fock method with 6-31G(d) basis.[37] For QM/MM simulations, the PM6 semiempirical method[38] was used to describe the potential energy of the QM region. The charge and spin multiplicity for the QM region were defined as −2 and 1, respectively. The valences of the QM atoms at the QM/MM boundary were satisfied using the “link atom” method.[39] Finally, a total of 69 atoms are included in the QM region (Figure ) to mimic the reactive part of the inactivation reaction of FOF and l-Cys catalyzed by FosB in the presence of Mg2+.

Figure 5

Atoms in the QM region of the FosB ternary complex.

US Simulations and PMF Calculations

Initially, before US simulations, the molecular model was minimized by 20,000 cycles of steepest descent and conjugate gradient[40] algorithms with positional restraints of 50.0 kcal/mol Å2 for the QM atoms. Next, the whole system was minimized with the progressive relaxation of restraints. We used 20 ps to raise the temperature gradually from 10 to 300 K. Afterward, 50 ps of unrestrained QM/MM MD simulations were carried out to equilibrate the system. The simulations were performed using periodic boundary conditions with a cutoff distance of 12.0 Å and a time step of 1.0 fs. The particle mesh Ewald method was used to calculate electrostatic interactions with a cutoff distance of 10 Å. The temperature was controlled using a Langevin thermostat with a collision frequency of 2.0 ps–1,[41] and the pressure was controlled using the Beredensen barostat with a 2.0 ps relaxation time.[42] Thus, to calculate the free-energy barriers that lead to the inactivation of FOF by FosB, the system was divided into a series of windows located along the RC (Figure ). To evaluate the regioselectivity of the inactivation reaction, two RCs were simulated: (a) RC = d1 – d2 = path 1, where “d1” describes the nucleophilic attack of the thiol group of l-Cys to C1 atom of FOF, while “d2” describes the bond breaking of the C1 atom and the oxirane oxygen of FOF; (b) RC′ = d1′ – d2′ = path 2, which describes the nucleophilic attack of the thiol group of l-Cys to the C2 atom of FOF (d1′), followed by the opening ring of the epoxide group of FOF at C2 (d2′). Each RC was constrained using a harmonic force constant of 200 kcal/mol Å2 and scanned from −2.10 to 0.80 Å in steps of 0.10 Å.

The samples in each window must overlap with the adjacent windows so that the impartial PMF can be reproduced by removing the polarization potential. The external polarization potential u in the window i is a harmonic function u = k(r – r)2, where r is the reference position and k is the harmonic force constant. All US simulations were performed using the SANDER module of the AmberTools16 software package. The PMF profile was built by using WHAM program.[43]

To improve the energetic analysis of the PMF profile obtained by PM6/MM US simulations, high-level QM corrections at the DFT level were computed as reported in the previous literature.[44,45] This approach provides that the interaction energy of the QM/MM does not change significantly at high and low levels of theory, as can be shown in eq EDFT(model) – EPM6(model) corresponds to the difference in the energies of the QM layer calculated from the optimized stationary points at PM6 and wB97XD/6-31G(d,p) levels. After performing the corrections for the reactants and products, the rest of the potential energy surface was corrected interpolating by increments added to each point. This correction factor is defined as the energy difference between two stationary points divided by the number of points along with the RCs.[44,45]

Conclusions

Here, we have found that FosB regioselectivity is related to the thiol attack on the C1 atom of FOF with a barrier 3.0 kcal/mol lower than the thiol attack on the C2 atom. The energetic contribution of individual residues revealed that Tyr39 and His7 presented the most stabilizing effect aiding the nucleophilic attack on the C1 carbon. Besides that, opening the epoxide ring of FOF at C1 leads to conformational changes of Asn50 that rotates in order to stabilize Cys9, an acidic residue that appears to be related to the protonation of the epoxide oxygen and facilitates the drug to escape the active site. These results will be potentially useful in the development of new FosB inhibitors and should help to treat infections by FOF-resistant bacteria.