Comprehensive analysis of the mouse cytochrome P450 family responsible for omega-3 epoxidation of eicosapentaenoic acid

Metabolites generated via oxygenation of the omega-3 double bond (omega-3 oxygenation) in eicosapentaenoic acid (EPA) have recently been identified as novel anti-inflammatory lipid mediators. Therefore, oxygenase(s) responsible for this metabolic pathway are of particular interest. We performed genome-wide screening of mouse cytochrome P450 (CYP) isoforms to explore enzymes involved in omega-3 oxygenation of EPA. As a result, 5 CYP isoforms (mouse Cyp1a2, 2c50, 4a12a, 4a12b, and 4f18) were selected and identified to confer omega-3 epoxidation of EPA to yield 17,18-epoxyeicosatetraenoic acid (17,18-EpETE). Stereoselective production of 17,18-EpETE by each CYP isoform was confirmed, and molecular modeling indicated that chiral differences stem from different EPA binding conformations in the catalytic domains of respective CYP enzymes.

Prepared EPA was docked into the BM-3 and CYP1A2 receptors individually using Glide version 7.2 8,9,10 (Schrödinger Release 2016-3: Glide, Schrödinger, LLC, New York, NY, 2016) and Glide induced-fit docking, respectively 11,12,13 (Schrödinger Release 2016-3: Induced Fit Docking protocol; Glide version 7.2. and Prime version 4.5, Schrödinger, LLC, New York, NY, 2016). Because the BM-3 crystal structure binds N-palmitoylglycine (long-chain fatty acid) 2 , which is structurally similar to EPA, Glide could predict the EPA binding pose in the BM-3 pocket. By contrast, CYP1A2 is co-crystallized with α-naphthoflavone 3 , which has a much different structure from EPA than N-palmitoylglycine (the BM-3 ligand). As a result, several amino acids in the binding cavity sterically hindered EPA binding and no docking poses were generated in Glide. In induced-fit docking of CYP1A2, the Phe226, Leu382 and Leu497 side chains were trimmed. Docking simulations in both BM-3 and CYP1A2 were performed using the OPLS-3 force field 7 . The grid box was placed on the center of the originally bound ligand.
To obtain a docking pose consistent with the experimental data, a specific range (between 0.3 and 0.8) of scaling factors for van der Waals radius were used for the receptor and ligand.
Using the docking procedures described, only a few poses that were positionally and chirally appropriate for EPA metabolism were obtained for both BM-3 and CYP1A2 structures.
This low success rate can be attributed to the pocket conformation of the docking receptors, which is basically the same as in the crystal structures and therefore not suitable for EPA binding and metabolism. To refine the obtained EPA-CYP complexes in Glide docking, the structures were subsequently subjected to MD simulations, which can optimize side chain and backbone atoms of proteins such that the receptor pocket is rearranged to fit the docked EPA.
Calculations were conducted for 120 ns using DESMOND version 4.7 14,15 with an OPLS-3 force field 7 . In the MD simulations, the simple point charge (SPC) model 16 for water was used in a cubic box with periodic boundary conditions containing a 10 Å buffer distance to each dimension and the system was electrically neutralized at 0.15 M NaCl. Before performing the productive runs for 120 ns, the systems were relaxed by the default protocol for the NPT ensemble, which consists of five steps: (i) 100 ps NVT ensemble with Brownian dynamics at 10 K with solute restrained; (ii) 12 ps simulation in the NVT ensemble using a 10 K Berendsen thermostat with solute heavy atoms restrained; (iii) 12 ps simulation in the NPT ensemble using a 10 K Berendsen thermostat and a 1 atm Berendsen barostat with solute heavy atoms restrained; (iv) 12 ps NPT ensemble using a 300 K Berendsen thermostat and a 1 atm Berendsen barostat with solute heavy atoms restrained; and (v) 24 ps NPT ensemble using a 300 K Berendsen thermostat and a 1 atm Berendsen barostat with no restraints. In the 120 ns productive runs, 300 K temperature and 1 atm pressure were maintained using the Nosé-Hoover thermostat 17 and Martyna-Tobias-Klein barostat algorithms 18,19 . The short-range electrostatic interactions were analyzed using a cut-off value of 9.0 Å. To treat long-range electrostatic interactions, the particle-mesh Ewald method was used in which a tolerance value of 1e-9 was set 20 . A multiple time step approach, RESPA, was employed where default values of 2.0, 2.0 and 6.0 fs were used for bonded, short-range nonbonded, and long-range nonbonded electrostatic interactions, respectively 21 . To enable the time step, the SHAKE algorithm was used 22 .
In MD simulations, distance restraints were applied (k=3.0) between the metabolized position of EPA and the Fe 2+ ion of the heme (C17-Fe 2+ and C18-Fe 2+ ) because the pocket in the initial structure had enough space. Without the restraints, the metabolized position of EPA moved slightly away from the Fe 2+ ion (in this case, the pocket failed to be optimized for EPA).    Figure S1. continued    Figure S1. continued    1a1  1a2  1b1  2a4  mock  2a5   2a12  2a22  2b9  2b10  2b13  2b19   2b23  2ab1  2c29  2c37  2c38  2c39   2c40  2c44  2c50  2c54  2c55  2c65   2c66  2c67  2c68  2c69  2c70  2d9   2d10  2d11  2d12  2d22  2d26  2d34   2d40  2e1  2f2  2g1 2j5 2j6    Figure S2. continued   Figure S5. Bound EPA in the last snapshot of MD simulation of the BM-3 structure is shown in green. Docked EPA in the same structure using Glide is shown in magenta. Heme is indicated in red, and the three most interacted residues with EPA (Arg47, Tyr51, and Gln73) are shown in blue. The binding pocket calculated by SiteMap (Schrodinger, LLC) is showed in mesh. Figure S6. Bound EPA in the last snapshot of MD simulation of human CYP1A2 is shown in green. Docked EPA in the same structure using Glide is shown in magenta. Heme is indicated in red, and the three most interacted residues with EPA (Thr118, Ser122, and Asn312) are shown in blue. Located behind the binding cavity, Ser122 is hidden by the surface pocket. The binding pocket calculated by SiteMap (Schrodinger, LLC) is shown in mesh. Figure S7. RMSD value for the Cα atoms of BM-3 (blue) and for the EPA (red) during the MD simulation. The reference is the EPA and BM-3 complex predicted by Glide. Figure S8. MD simulation of BM-3. (A) The interaction summary between residues and EPA in MD simulation of BM-3. The three most interacted residues with EPA are shown with interaction percentages. Interaction criterion is the distance less than 2.5 Å between the residue and EPA (if the residues have a hydrogen bond via a water atom, the criteria is less than 2.    N312A