A redox-mediated Kemp eliminase

The acid/base-catalysed Kemp elimination of 5-nitro-benzisoxazole forming 2-cyano-4-nitrophenol has long served as a design platform of enzymes with non-natural reactions, providing new mechanistic insights in protein science. Here we describe an alternative concept based on redox catalysis by P450-BM3, leading to the same Kemp product via a fundamentally different mechanism. QM/MM computations show that it involves coordination of the substrate's N-atom to haem-Fe(II) with electron transfer and concomitant N–O heterolysis liberating an intermediate having a nitrogen radical moiety Fe(III)–N· and a phenoxyl anion. Product formation occurs by bond rotation and H-transfer. Two rationally chosen point mutations cause a notable increase in activity. The results shed light on the prevailing mechanistic uncertainties in human P450-catalysed metabolism of the immunomodulatory drug leflunomide, which likewise undergoes redox-mediated Kemp elimination by P450-BM3. Other isoxazole-based pharmaceuticals are probably also metabolized by a redox mechanism. Our work provides a basis for designing future artificial enzymes.


System preparation and setup for computational analysis
The initial structures of P450-BM3 were taken from PDB code of 1JPZ 1 . The substrate 5nitrobenzisoxazole was docked into the active site of P450-BM3 using AutoDock Vina tool 2 in Chimera 3 . Missing hydrogen atoms were added by module leap of Amber 14 4 . The force field for the heme moiety in the resting state (Fe(III)) was taken from the literature 5 , while the force field for the one-electron reduced state (Fe(II)) was parameterized using "MCPB.py" 6 . The general AMBER force field (GAFF) 7 was used for the substrate 5-nitrobenzisoxazole, while the partial atomic charges and missing parameters were obtained from the RESP method 8,9 , using HF/6-31G* level of theory. 15 Na + ions were added into the protein surface to neutralize the total charges of the systems. Finally, the resulting system was solvated in a rectangular box of TIP3P 10 waters extending up to minimum cutoff of 10 Å from the protein boundary. The Amber ff14SB force field 11 was employed for the protein in all of the Molecular Dynamics (MD) simulations

Computational details for MD simulations
After proper parameterizations and setup, the resulting system's geometries were minimized (5,000 steps for steepest conjugate and 10,000 steps for conjugate gradient) to remove poor contacts and relax the system. The systems were then annealed from 10 to 300 K under the constant amount of substance (N), volume (V) and temperature (T) (NVT ensemble) for 50 ps with a weak restraint of 5 kcal mol -1 (Å 2 ) -1 . Subsequently, the systems were maintained for 1 ns of density equilibration under constant amount of substance (N), pressure (P) and temperature (T), i.e. isothermal-isobaric (NPT ensemble) at target temperature of 300 K and the target pressure of 1.0 atm using Langevin-thermostat 12 and Barendsen barostat 13 with collision frequency of 2 ps and pressure relaxation time of 1 ps, with a weak restraint of 1 kcal mol -1 (Å 2 ) -1 . This 1 ns of density equilibration is not identical with conformational equilibration, but rather a weakly restrained MD in which we slowly relax the system to achieve a uniform density after heating dynamics under periodic boundary conditions. Thereafter, we removed all restraints applied during heating and density dynamics and further equilibrated the systems for ∼3 ns to get well settled pressure and temperature for conformational and chemical analyses. This was followed by a productive MD run, for each system, for 100 ns. During all MD simulations, the covalent bonds containing hydrogen were constrained using SHAKE 14 , and particle mesh Ewald (PME) 15 was used to treat long-range electrostatic interactions. All MD simulations were performed with GPU version 16 of Amber 14 package.  25 , labeled as B2. All the QM/MM transition states (TSs) were located by relaxed potential energy surface (PES) scans followed by full TS optimizations using the P-RFO optimizer implemented in the HDLC code 26 . The empirical dispersion energy correction was calculated forall species by using the DFT-D3 program 27 .

Chemicals
All chemicals were purchased from Sigma-Aldrich, Tokyo Chemical Industry (TCI) or Alfa Aesar and used without further purification. NMR spectra were recorded on a Bruker Avance 300 ( 1 H: 300 MHz, 13 C: 75 MHz) spectrometer using TMS as internal standard (d=0). Compound 2 was prepared according to a published protocol 28

GC-MS analysis
The reactions were initiated by addition of 1 mM NADPH in 500 µL sodium phosphate buffer containing P450-BM3 WT (1 µM) and 2 mM substrate 1 at 25°C, 800 rpm for 10 min. After reaction, HCl was added to adjust the pH value to 1-2 until the solution became colorless. The product was extracted with dichloromethane (0.5 mL X 2), the organic phase was separated by centrifugation, and dried over Na 2 SO 4 . The solvent was completely evaporated, followed by addition of 100 µL acetonitrile and then subjected to GC-MS analysis for product identification.
The analysis was conducted using an achiral column (Rtx-1, 29.5 m× 0.25 mm ×0.25 mm), with injector and detector temperatures at 230°C and 350°C, respectively. Temperature program: 60°C to 340°C at 5°C min -1 , then hold at 340°C for 10 min. The retention time for 5-nitrobenzisoxazole and its Kemp product are 19 and 23 min, respectively.