Mechanism for the reactivation of the peroxidase activity of human cyclooxygenases: investigation using phenol as a reducing cosubstrate

It has been known for many years that the peroxidase activity of cyclooxygenase 1 and 2 (COX-1 and COX-2) can be reactivated in vitro by the presence of phenol, which serves as a reducing compound, but the underlying mechanism is still poorly understood. In the present study, we use phenol as a model compound to investigate the mechanism by which the peroxidase activity of human COXs is reactivated after each catalytic cycle. Molecular docking and quantum mechanics calculations are carried out to probe the interaction of phenol with the peroxidase site of COXs and the reactivation mechanism. It is found that the oxygen atom associated with the Fe ion in the heme group (i.e., the complex of Fe ion and porphyrin) of COXs can be removed by addition of two protons. Following its removal, phenol can readily bind inside the peroxidase active sites of the COX enzymes, and directly interact with Fe in heme to facilitate electron transfer from phenol to heme. This investigation provides theoretical evidence for several intermediates formed in the COX peroxidase reactivation cycle, thereby unveiling mechanistic details that would aid in future rational design of drugs that target the peroxidase site.

Protein structure refinement. The X-ray structure of human COX-1 (PDB code: 6y3c 14 ) and human COX-2 (PDB code: 5kir 15 ) are used as templates for computational docking analyses. For comparison, the X-ray structure of sheep COX-1 protein (PDB code: 1q4g 16 ) and mouse COX-2 protein (PDB code: 3nt1 17 ) in complex Scheme 1. A proposed mechanism for the reactivation of the catalytic cycle of the COX peroxidase by phenol. PPIX is shown as an abbreviated parallelogram, and the imidazole ring of His388 as Im. Also, PGG 2 is for prostaglandin G 2 , and PGH 2 for prostaglandin H 2 . The proposed mechanisms for reaction 1 and 2 are depicted beneath the main scheme. In these two reactions, two protons attack the oxygen atom of Fe=O, resulting in the formation of H 2 O molecule. Detailed molecular orbital analysis of reaction 1 is shown in Scheme 2.
Scientific RepoRtS | (2020) 10:15187 | https://doi.org/10.1038/s41598-020-71237-x www.nature.com/scientificreports/ with protoporphyrin IX (PPIX) containing an Fe III ion (abbreviated as PPIXFe III ) are also included in the computational docking analysis. All small molecules that are non-covalently attached to the COX proteins are removed, and then the amino acid residues in the protein structures are renumbered according to the known sequences. The PPIXFe III is individually add to the peroxidase sites of human COX-1 and COX-2. The Clean Protein module in Discovery Studio is used to complete the side chains for amino acid residues, to correct bonding and bond orders, and to add hydrogens back. Notably, the PPIXFe III component in both COX-1 and COX-2 are modified into PPIX •+ Fe IV =O, PPIX 2+ Fe III and PPIX + Fe III for further docking analysis. The reason for using PPIX •+ Fe IV = O, PPIX 2+ Fe III and PPIX •+ Fe III in place of PPIXFe III is described later in the Results section. The Prepare Protein module in Discovery Studio is used together with the CHARMm force field for protein preparation 18 . flexible docking. For flexible docking, we use the Find Sites from Receptor Cavities module to identify the binding site in the prepared COX-1 and COX-2 structures. According to our earlier study, the target site is the peroxidase active site in these two COX proteins 12 . We select all amino acid residues within a 5 Å reach of the target site and allow them to have flexible side chains. The SBD Site Sphere is centered at the target site and then expanded to the radius size around 13 Å. Under the Flexible Docking mode with the maximum number of residues for creating side chain conformations set to 10, the Simulated Annealing docking method is then applied to dock phenol into the target site of COX-1 and COX-2. Docking analyses are separately carried out for COX-1 and COX-2, with phenol in both ionized and nonionized states. The whole structure of each COX protein is further minimized with the CHARMm force field 18 . This docking method is validated by performing a "self-docking" exercise using the crystal structure of mouse COX-2 in complex with celecoxib (PDB id: 3ln1) 19 . Specifically, we remove celecoxib from the experimentally-determined structure of the complex, and then use the computational docking program to dock it back inside the COX-2 protein. The root mean square of deviation (RMSD) between the docked ligand pose and its experimentally-determined pose in the crystal structure is computed, which is found to be 0.63 Å, suggesting that the docking method employed in this study has a very good ability to predict the correct binding pose(s) of a small ligand inside the specified COX binding pocket ( Figure S1).
calculation of binding energy. After completing flexible docking, the Calculate Binding Energies module is used to find complexes with the lowest binding energy value. The free energy of binding for a receptor-ligand complex is computed from the free energies of the complex, the target protein, and the ligand. According to Discovery Studio, the free energy values are separately computed using the CHARMm force field and the Poisson-Boltzmann equation with the non-polar surface area (PBSA) method 18,20 . In this approach, the Poisson-Boltzmann equation is solved numerically on a three-dimensional (3D) grip, and the computed electrostatic potential is used to estimate the electrostatic solvation free energy. The ligand conformational entropy is also taken into consideration during the free binding energy calculation. The following equation is used to calculate the binding energy (ΔG binding ) between phenol and the COX-1 or COX-2 protein: where G complex is the absolute free energy of the complex, G COX is the absolute free energy of the COX protein, and G ligand is the absolute free energy of the ligand 21,22 . The ΔG binding value is used to reflect the relative interaction affinity between the COX enzyme and the phenol molecule.
Quantum chemistry calculation of geometry and Gibbs free energy. The molecular geometry optimization and vibrational frequency calculation are performed on a Dell PowerEdge R730 Server with Gaussian 09 W calculation software (Revision D.01, Gaussian, Wallingford, CT) 23 . Geometrical structure and electronic information of molecules are investigated in the vapor phase using the B3LYP-D3 method (namely, the Becke's three-parameter hybrid functional and the Lee-Yang-Parr correlation functional method) [24][25][26][27][28] . where is a diagonal matrix, whose diagonal values correspond to the overlap integrals between the newly transformed alpha and beta orbitals. Normally, such new overlap integrals should approximate 1.0. The parameter U www.nature.com/scientificreports/ (V) is the transformation matrix between the original orbitals and the new orbitals of α (β) spin 37 . The coefficient matrix of the biorthogonalized orbitals can be calculated as: The energy of biorthogonalized orbitals is evaluated as the expectation value of Fock operator: where F AO is the Fock matrix in original basis function, and C(µ, i) corresponds to the coefficient of basis function µ in biorthogonalized orbital i . After energy calculation, biorthogonalized orbitals are ranked according to their energies (average of energy of α orbital and its β counterpart) 37 .

Results and discussion
First, the Flexible Docking function of Discovery Studio is used to dock the phenol molecule in both non-ionized and ionized states into the peroxidase active site with an oxygen atom covalently attached to the Fe IV ion (abbreviated as Fe IV =O) of COX-1 and COX-2 to predict the potential binding mode and orientation of phenol's O atom toward the Fe atom in PPIX •+ Fe IV =O. Results from our recent studies 38,39 suggest that the binding interaction of a reducing substrate under ionization is dramatically enhanced in comparison with the non-ionizing state. Therefore, in the present study, we also choose to perform docking analysis using both non-ionized and ionized phenol molecules. Under these conditions, we find that either the binding energies between phenol molecule (non-ionized or ionized) and PPIX •+ Fe IV =O are positive, or the distances between phenol's oxygen and the Fe IV ion are longer than 5 Å, which indicate that the presence of the Fe=O group in PPIX •+ Fe IV =O would prevent the binding of phenol or phenol ion inside the peroxidase active site close to the heme group ( Fig. 1, Table 1).
Next we further test the possibility of whether phenol in its non-ionized or ionized state can be docked inside the peroxidase active sites of COX-1 and COX-2 when the O atom is absent in the Fe=O group (namely, in the form of PPIX 2+ Fe III ). We find that both ionized and non-ionized phenol molecules can be docked inside the active site of COX-1 and COX-2 ( Fig. 1, Table 1). As expected, ionized phenol has a much lower binding energy level and is much closer to Fe than non-ionized phenol, indicating that the former can bind more favorably inside the peroxidase sites of COX-1 and COX-2 than the latter. The binding energies of phenol ion are -157.46 kcal/ mol for COX-1 and -193.48 kcal/mol for COX-2, respectively. As is shown in Fig. 1, the oxygen atom of phenol ion is very close to the heme Fe when O is removed from the Fe=O group. The Fe-O distances of the docked phenol ion are 2.10 Å for COX-1 and 2.35 Å for COX-2, which is sufficiently close for electron transfer to take place. Similar results are also obtained using sheep COX-1 and mouse COX-2 as target proteins (data are summarized in Figure S2). The high degree of similarity in their docking results is not surprising, as these proteins share a high degree of sequence homology (89% for human and sheep COX-1 proteins and 85% for human and mouse COX-2 proteins).
This result leads to the suggestion that the binding of phenol inside the peroxidase active sites of COX-1 and COX-2 likely occurs only after the O atom is removed from Fe IV =O. Regarding the possible mechanism of oxygen removal from Fe IV =O, we hypothesize that two protons might bind to the oxygen atom in PPIX •+ Fe IV =O and then the oxygen atom is removed by forming H 2 O (depicted in Scheme 1). To test this hypothesis, we conducted a theoretical analysis of the binding process of two protons to the O atom in PPIX •+ Fe IV =O. Since these calculations are extremely time-consuming, here we adopt a slightly simplified approach by using Por •+ Fe IV =O (Por •+ refers to porphyrin π radical) as a model compound for Gaussian calculation. In this model, the His388, which binds to the heme iron in COX enzymes, is simplified into an imidazole ring (abbreviated as Im). Notably, this simplified model has also been used in earlier studies on cytochrome P450 enzymes [40][41][42] .
Before examining our proposed cycle, we first calculate the electronic structures including spin density ( Table 2, Table S1 and S2) and molecular orbitals of all intermediates ( 2 Por •+ Fe IV =O and 4 Por •+ Fe IV =O are shown as examples in Scheme S1 and Figure S3). Molecular orbital analyses indicate that for either doublet or quartet state, there are 3 unpaired electrons located in π * (d yz -p y ), π * (d xz -p x ), and a 2u orbitals. This result is consistent with the spin density result ( Table 2). The sum of the spin densities of iron and oxygen atoms is approximately 2.0, which corresponds to the two singly-occupied orbitals π * (d yz -p y ) and π * (d xz -p x ). The spin density of porphyrin is approximately 1.0 (or -1.0), which corresponds to the singly-occupied a 2u orbital of the porphyrin aromatic ring (Scheme S1 and 2). Such electronic structures (doublet and quartet states) match well with the results from previous studies of cytochrome P450, a protein with a similar heme structure in the active site [40][41][42] .
Optimized geometries of the complexes Por •+ Fe IV =O, Por 2+ Fe III and PorFe III in the lowest doublet, quartet and sextet spin states are schematically depicted in Fig. 2. All three electronic states in Fig. 2 correspond to the minimum energy structures that have zero imaginary frequencies. In terms of geometry, the doublet and quartet states of Por •+ Fe IV =O closely resemble each other. In addition, the quartet and sextet states of Por 2+ Fe III are also very similar. In the case of the PorFe III structures, the epoxide moiety is in a plane almost orthogonal to the porphyrin ring bisecting the plane between the four nitrogen atoms. In particular, the quartet state has a substantially shorter Fe-imidazole bond length of 2.10 Å in 4 Por •+ Fe IV =O compared to 2.06 Å in 2 Por 2+ Fe III .
The reaction of the first proton binding to Por •+ Fe IV =O is depicted in the lower panel of Scheme 1 as reaction 1. This mechanism is supported by our quantum mechanics calculations as follows. As the first proton binds to the O atom in 2 Por •+ Fe IV =O (Fig. 3), the bond length of Fe IV -O is slightly increased from 1.64 to 1.83 Å (  (Table 3), which also reflects the weakening of the Fe-O bond following the bonding of the first proton. Similarly, 4 Por •+ Fe IV =O and 6 Por •+ Fe IV =O show the same trend in these states. Moreover, the spin density analysis suggests that the binding of the first proton changes the electronic structure of iron and porphyrin (Table 2). From 2 Por •+ Fe IV =O to 2 Por 2+ Fe III -OH, the spin density of the porphyrin changes from − 1.0 to 0.0, and its NBO charge increases from 0.7 to 1.6 ( Table 2). This indicates that the original single electron on the a 2u orbital of the porphyrin is transferred to another orbital. Furthermore, the  www.nature.com/scientificreports/ sum of the spin densities of iron and oxygen atoms is reduced from 2.0 (1.1 + 0.9) to 1.0 (0.9 + 0.1). To explain this phenomenon, we perform molecular orbital analysis and find that after the binding of the first proton, the original singly-occupied π * (d xz -p x ) becomes a fully-occupied d xz orbital with lower energy, and the original singly-occupied a 2u orbital becomes empty (Scheme 2). The singly-occupied π * (d yz -p y ) persists during the protonation. As such, the binding of the first proton breaks one Fe-O bond, and the corresponding Fe-O antibonding orbital π * (d xz -p x ) receives one electron from the a 2u orbital of porphyrin and changes to d xz orbital (Scheme 2). Hence, the protonation product of 2 Por •+ Fe IV =O is 2 Por 2+ Fe III -OH. In addition, energy analysis also shows that this reaction is favorable because the free energy change is − 29.70, − 24.28 and − 29.80 kcal/mol for doublet, quartet, and sextet states, respectively (Table 2). Based on the structure of Por 2+ Fe III -OH which already has the first proton bond to the O atom in 2 Por 2+ Fe III =O (Fig. 3), next we further calculate whether the second proton would bind to the O atom in 2 Por 2+ Fe III -OH. We find the distance between the O atom and Fe atom in 2 Por 2+ Fe III -OH increases from 1.83 to 2.30 Å (Table 3),  www.nature.com/scientificreports/ from phenol ion to the porphyrin ring of 2 Por 2+ Fe III . Combining the above conclusion together with the analysis of the molecular orbitals of two intermediates, we propose that an electron of the phenol ion enters into the a 2u orbital of porphyrin. Similar electron transfers are also observed in other states of multiplicity, though the molecular orbitals involved may differ. Ultimately, the changes of Gibbs free energy demonstrate that such reduction processes are all favorable for doublet, quartet, and sextet states of Por 2+ Fe III ( Table 2, Scheme 4). After the first electron is transferred from one phenol molecule to PPIX 2+ Fe III , Por •+ Fe III is formed in COX-1 and COX-2. Then we employ Flexible Docking approach again to dock the second phenol molecule in both nonionized and ionized states into the peroxidase active site of COX-1 and COX-2 with PPIX + Fe III (Figs. 5 and 6, Figure S4). As predicted, the results are very similar to those with PPIX 2+ Fe III . In the case of COX-1, the binding energy value is − 15.84 kcal/mol for non-ionized phenol and − 100.71 kcal/mol for ionized phenol, and in the case of COX-2, the binding energy value is − 7.13 kcal/mol for non-ionized phenol and − 139.32 kcal/mol for ionized phenol ( Table 1). The distances between the O atom of non-ionized phenol and the Fe atom of PPIX + Fe III are 12.06 and 11.94 Å, respectively, for COX-1 and COX-2 (Table 1). However, when phenol is ionized, the distances between its O atom and the Fe of PPIX + Fe III are much closer, at 2.11 Å for both COX-1 and COX-2 (Table 1). It is predicted that these distances are sufficiently short for effective electron transfer to take place. Similar results are also obtained using sheep COX-1 and mouse COX-2 as target proteins (summarized in Figure S5).
When the protein surface for peroxidase active sites of COX-1 and COX-2 is modeled, it is noted that most of the amino acid residues inside the active site where phenol ion binds are neutral or basic ( Figure S6). Therefore, it is speculated that the protein microenvironment in or near the active site likely would be conducive to the binding of an ionized phenol molecule. In addition, it is of note that an earlier study showed that the distal histidine (His207) may facilitate the ionization of hydroperoxide inside the peroxidase active site 43 . Similarly, it is also observed in this study that the phenol ion is very close to His207 (Figs. 4 and 6, Figure S4), making it likely that His207 may aid in stabilizing ionized phenol inside the peroxidase site. In addition, His207 may facilitate the ionization of phenol through formation of a salt bridge with phenol ion's Oatom (Figs. 4 and 6).
The results from docking analysis suggest that ionized phenol can bind far more favorably inside the peroxidase sites of COX-1 and COX-2 than non-ionized phenol for electron donation; this result is expected as similar results were also obtained in our recent studies 38,39 . Based on known pKa value of phenol (9.98) 44 , the extent of phenol ionization around physiologically-relevant pH range (pH 7.0 -7.4) would be very small, around 1-3‰. Despite the small percentage of ionized phenol formed in aqueous condition, its existence would make it possible to serve as a reducing cosubstrate for the COX enzymes, partly also owing to its much higher affinity for the peroxidase active sites.
Following the above docking analysis, next we further analyze the process of Por •+ Fe III (Por 2+ Fe II for singlet) reduction by the second electron via mapping spin densities and NBO charges. According to our analyses, 1 Por 2+ Fe II and 3 Por •+ Fe III are reduced to 2 PorFe III ; 5 Por •+ Fe III is reduced to 4 PorFe III ; 7 Por •+ Fe III is reduced to  (Table 2). Therefore, we propose that the electron donated by the second phenol ion is transferred to the a 2u orbital of porphyrin (Scheme 5). Similar reduction processes can occur for other states of multiplicity although different molecular orbitals would participate in the electron transfer. In addition, we also calculate the change of Gibbs free energy in such processes. The second reduction is also found to be spontaneous, with free energy decrease by 29.05, 34.60 and 32.25 kcal/mol for the formation of doublet, quartet, and sextet PorFe III , respectively (Table 2 and Scheme 4). Details of the potential energy profiles for the regeneration of the peroxidase activity of COX-1 and COX-2 are assembled in Scheme 4. In general, three states of multiplicity of Por •+ Fe IV can all spontaneously undergo the catalytic cycle accelerated by phenol. While the quartet and doublet states at the beginning of the reaction . Phenol ion is shown as sticks, with different colors representing different atomic elements. Fe III is shown as sphere and colored in pink. All the nearby residues are shown in line, with PPIX •+ ring in orange, H388 in marine, Y385 in dark green, H207 in raspberry, V291 in yellow, and L294 in light blue. All intermolecular interactions that facilitate the binding of phenol ion are shown in dash line, with metal-acceptor in gray, salt bridge in cyan, Pi-alkyl in pink, and Pi-Pi in magenta.

Scheme 5.
Electron transfer from phenol ion to 3  www.nature.com/scientificreports/ cycle have relatively low energy levels compared to the energy level of the sextet state, the quartet state has the lowest energy level in the last two reactions (Scheme 4). In comparison, the doublet state has the highest energy levels in the last two reactions. This information suggests that the quartet state likely is both more stable and thermodynamically more favored to proceed through the whole reaction cycle. Scheme 4 also depicts several multiplicity shifts during the whole cycle, which suggests that recombination of electrons within the orbitals may be frequent in the cycle. In our proposed cycle (Scheme 1), we hypothesize that during the reduction processes, electrons are not directly transferred from phenol ion to porphyrin. Electrons are firstly transferred to Fe (Fe III in Por 2+ Fe III is reduced to Fe II in Por 2+ Fe II ) and then to porphyrin (Por 2+ Fe II is converted to Por •+ Fe III ). Therefore, it is postulated that Fe acts as a bridge of electron transfer from phenol ion to porphyrin, which is highly probable given the strong electrophilicity and polarizability of Fe and the close distance between the docked phenol ion and Fe (Table 1). In partial support of this hypothesis, Mulliken charges of atoms show that Fe ion indeed has a stronger electrophilicity compared to the porphyrin ring (Fig. 7). In addition, as shown in Scheme 4, 1 Por 2+ Fe II , which is formed by electron transfer directly to Fe, displays a considerably higher energy level (about 22.46 kcal/ mol higher) compared to 3 Por •+ Fe III , 5 Por •+ Fe III and 7 Por •+ Fe III (Table 2). This relatively large energy difference between states of multiplicity is not observed in other groups of intermediates. Therefore, it is suspected that 1 Por 2+ Fe II may serve as a reactive intermediate during the reduction and then undergo electron recombination to form 3 Por •+ Fe III , 5 Por •+ Fe III or 7 Por •+ Fe III (Scheme 3). Further experimental or theoretical studies are needed to test this hypothesis. It is of note that in Scheme 1, only Por •+ Fe IV =O has been identified in an earlier experimental study 45 , while other theoretically-viable intermediates or transition states have not yet been identified in experimental settings thus far.
It is of note that our recent studies have shown that some of the naturally-occurring phenolic compounds such as quercetin and myricetin are high-affinity reducing substrates for the peroxidase catalytic site of COX-1 and COX-2 11,12 . Since flavonoids have similar functional groups like phenol, the results of our present study may also shed light on the mechanism of COX activation by flavonoids. The knowledge gained from this as well as earlier studies 11,12,38,39 might aid in the design of structural analogs which can bind inside the peroxidase sites of COX-1 and COX-2 but are incapable of donating electrons for catalytic reactivation. These compounds would function as a novel type of COX inhibitors that selectively target the peroxidase sites of the enzymes and are different from clinically-used NSAIDs, which selectively target the cyclooxygenase sites. In the present study, by employing molecular docking and quantum chemistry calculation, we investigate the mechanisms by which phenol molecule activates the peroxidase catalytic cycle of COX-1 and COX-2. Computational quantum chemistry analysis shows that two protons bind sequentially to the O atom in PPIX •+ Fe IV =O, resulting in oxygen removal and formation of H 2 O and PPIX 2+ Fe III . Following this initial reaction, two phenol ions bind in sequence inside the peroxidase sites of the enzymes, resulting in the reduction of PPIX 2+ Fe III to its resting form PPIXFe III , with each phenol ion donating an electron. The results of our present study indicate that during the reactivation process by phenol, Fe only acts as a bridge for electron transfer, which enables the passage of two electrons from two phenol ions to Fe as a transient intermediate, and then quickly to porphyrin.