Mechanistic Studies of Fatty Acid Activation by CYP152 Peroxygenases Reveal Unexpected Desaturase Activity

The majority of cytochrome P450 enzymes (CYPs) predominantly operate as monooxygenases, but recently a class of P450 enzymes was discovered, that can act as peroxygenases (CYP152). These enzymes convert fatty acids through oxidative decarboxylation, yielding terminal alkenes, and through α- and β-hydroxylation to yield hydroxy-fatty acids. Bioderived olefins may serve as biofuels, and hence understanding the mechanism and substrate scope of this class of enzymes is important. In this work, we report on the substrate scope and catalytic promiscuity of CYP OleTJE and two of its orthologues from the CYP152 family, utilizing α-monosubstituted branched carboxylic acids. We identify α,β-desaturation as an unexpected dominant pathway for CYP OleTJE with 2-methylbutyric acid. To rationalize product distributions arising from α/β-hydroxylation, oxidative decarboxylation, and desaturation depending on the substrate’s structure and binding pattern, a computational study was performed based on an active site complex of CYP OleTJE containing the heme cofactor in the substrate binding pocket and 2-methylbutyric acid as substrate. It is shown that substrate positioning determines the accessibility of the oxidizing species (Compound I) to the substrate and hence the regio- and chemoselectivity of the reaction. Furthermore, the results show that, for 2-methylbutyric acid, α,β-desaturation is favorable because of a rate-determining α-hydrogen atom abstraction, which cannot proceed to decarboxylation. Moreover, substrate hydroxylation is energetically impeded due to the tight shape and size of the substrate binding pocket.


CO-titration for determination of the concentration of active cytochrome P450s
For determination of the concentration of purified cytochrome P450s (OleT, CYPBSß and CYPCla) the COtitration assay was used as described in the original protocol by Omura and Sato [3] .
For determination of the concentration of active cytochrome P450 a few milligrams of sodium dithionite were dissolved in KPi-buffer (950 µL; pH 7.5, 100 mM) and filled into a cuvette. Afterwards the purified enzyme (50 µL) was added to the KPi-buffer containing the sodium dithionite and absorbance was measured on a photometer (Agilent Technologies Cary 60 UV-VIS; software: CARYWinUV Scan). The reduced enzyme was used as blank over the range from 400 -520 nm absorbance. Followed by this the reduced enzyme solution was gassed with CO for 30 s. Absorption was measured again in the range from 400 -520 nm to obtain a differential spectrum of purified cytochrome P450s after CO-titration. The concentration of active cytochrome P450 could be determined through absorption at 450 nm, while the absorbance maximum of inactive cytochrome P450 is usually at 420 nm.
Based on the law of Lambert-Beer, the concentration of active cytochrome P450 was calculated in the following way: The absorbance at 500 nm (baseline) was subtracted from that at 450 nm (active cytochrome P450). c active enzyme [µM] = abs * 1000 * 20 91 * d

Model set-up
Our calculations start from the substrate bound P450 OleTJE coordinates (4L40 pdb file) reported by Leys et al. [14] The active site (heme)iron(III)-water group was replaced Compound I model by manually adjusting the Fe-O distance to 1.63Å and by removing the protons from water. Subsequently, we used standard protonation and solvation protocols as reported by Thiel et al [15] previously at pH = 7. The system was neutralized with Mg 2+ and Clions and iteratively solvated in a sphere with radius of 40Å. Finally, a molecular dynamics simulation was run for 3ns without constraints ( Figure S1). Based on the displacement of the substrate bound residues and the overlay of several snapshot structures of the MD simulation ( Figure S2), we set up a cluster model that captured the full active site region and the substrate binding positioning. The MD simulation shows that the substrate binding pocket is very rigid and as such a cluster model was selected for initial studies.

Methods
We utilized density functional theory (DFT) with the unrestricted B3LYP [16,17] hybrid density functional theory for geometry optimizations and frequencies in the gas phase to perform mechanistic investigations of the oxidation reactions of 2-methyl-butyric acid by cytochrome P450 Compound I (Cpd I). The first set of calculations, "Part A", were run on a simplified iron(IV)-oxo porphyrin radical cation model for Cpd I, whereby the axial cysteinate residue was abbreviated with thiolate and all porphyrin substituents replaced by hydrogen atoms. We calculated the general reactivity landscape and electronic structure of intermediates and transition states with the S-Enantiomer of the substrate, leading to either hydroxylation, desaturation or decarboxylation products.
In all calculations the Model A optimizations were done with a 6-31G* [18,19] basis set employed on all atoms, except on iron where LANL2DZ with ECP was used. [20] To improve the energies a single point with a larger basis set, i.e. 6-311+G* on all atoms, was performed. All calculations were run in Gaussian-09, [21] and were done on the lowest energy doublet and quartet spin states.
In a second set of calculations, "Part B", a comprehensive cluster model was selected, whereby the small model was expanded with residues representing the substrate binding pocket. These studies use the B3LYP-GD3 [16,17,22] density functional theory method in combination with the def2-SVP [23] basis set on all atoms (BS1, Model B). The model is based on several docked conformations of both S-and Renantiomers in the crystal structure coordinates of P450 OleTJE [24] with the aims of reproducing the experimentally observed enantioselectivity of the substrates and the barriers of hydrogen atom transfers to Cpd I. An improved basis set (BS2), incorporating def2-TZVP on all atoms and an implicit PCM(Water) solvent model, was employed to correct the energies for all results for this model.
The S and R enantiomers were docked into the 4L40 protein databank (pdb) file with the SwissDock [25] web service. Of the obtained results, three viable low-energy R conformations (1, 4 and 7) and five viable S conformations were retrieved (1, 2, 3, 5 and 6). All other iterations had the substrate located outside the active site. Based on these structures a model was created that contained the iron(IV)-oxo porphyrin cation radical with thiolate axial ligand and substrate (as above) and in addition included the side chains of residues located close to the substrate in the active site, which resulted in a model with a total of 163 atoms. All cluster models were based on the crystal structure coordinates and constraints were placed on several atoms to keep the features of the protein, while only changing the substrate enantiomer and conformations to make energy comparisons plausible. To prevent the structure from changing dramatically from the original crystal structure orientation, we put constraints on several atoms. In particular, restrictions were placed on the heme to avoid rotation and translation, as well as on the backbone carbon atoms of all amino acid residues. A trimmed proline was protonated to account for the lost backbone carbon bond. Figure S1: Molecular dynamics simulation (without constraints) using the CHARMMM forcefield on the fully solvated enzyme structure of substrate bound in the vicinity of Compound I. Figure S2: Overlay of the active site regions of the snapshots for structures obtained from the MD simulation of Figure S1 after 0ns (brown), 1ns (blue), 2ns (purple) and 3ns (green). The heme and axial cysteinate groups are fixed, while all other atoms were unconstrained. As observed from the landscape of the simplified CpdI model, the alpha hydrogen atom abstraction is preferred over a beta hydrogen atom abstraction, with a barrier of 13.5 kcal/mol on the low-spin surface and 16.0 kcal/mol in the high-spin state. After the alpha hydrogen atom abstraction, an electron is immediately transferred to form a stable neutral substrate intermediate instead of a radical intermediate, and reduced Cpd II is formed, i.e. an iron(III)-hydroxo porphyrin complex. Then, a second barrier may be competitive for desaturation or Rebound on the LS state, leading to the alcohol or desaturation products respectively, in good agreement with experiment.
An alternative pathway is the beta hydrogen atom abstraction, which is slightly higher in energy than alpha hydrogen atom abstraction, but appears to lead to dominant decarboxylation products. In particular, the low spin pathway proceeds with a single barrier of 16.6 kcal/mol, whereas on the high spin surface the process undergoes a barrier of 21.8 kcal/mol, stabilized as an intermediate and after a very small barrier of less than 2 kcal/mol, forms the decarboxylation product.       However, a stronger substrate binding does not necessarily imply faster reactivity as will be seen next.

Control Reactions to Exclude Spontaneous (Non-Enzymatic) Desaturation and Secondaray Enzymatic Activities
To confirm the enzyme-driven desaturation reaction on substrate 5a, control reactions were performed without enzyme and with α-and β-hydroxy compounds 5c and 5d under the standard reaction conditions.
The desaturation product was detected as the corresponding cyclopropane (cyclopropanation due to excess of the derivatization agent; m/z = 176). This allowed a differentiation between enzymatically produced desaturation product and desaturation product spontaneously formed from via S33 decomposition of β-hydroxy compound 5d in the GC inlet (m/z = 162). Derivatization of reference 5e confirmed this.
Overall no non-enzymatic formation of the desaturation product was found by spontaneous decomposition of 5a, 5c or 5d by the reaction conditions and derivatization allowed to differentiate between enzymatic desaturation and thermal decomposition of 5d. 8 Analytical methods

Quantification of substrates and products and derivatization
After thawing the samples, the reactions were quenched by adding hydrogen chloride (aq., 5 N, 100 µL) followed by extraction with ethyl acetate (500 µL containing 0.1% 1-decanol as internal standard). After separation of the phases via centrifugation, the organic phase was dried over Na2SO4. An aliquot (150 µL) was mixed with methanol (60 µL) in a 1 mL GC-vial and esterification was done by adding TMSCHN2 (20 µL, 2M in hexane). Derivatized samples were analyzed with GC-FID and GC-MS. The underivatized samples were used for chiral GC-FID and HPLC analysis.
For determination of the conversion, calibration of 1a -5a was done in a range from 10 mM to 3 mM. For that stock-solutions of the substrate standards (200 mM -60 mM) were prepared.
In a 4 mL glass vial KPi-buffer (950 µL; pH 7; 100 mM) was mixed with the stock solutions containing the product/substrate standard (50 µL). The samples were treated exactly as described for the biotransformations (incubation, freezing and workup). In order to achieve an accurate quantification, the biotransformations as well as the calibrations for substrate and products were done at the same time.

GC-FID and GC-MS analysis
For GC-FID analysis an Agilent HP-5 column (30 m x 320 μm, 0.25 μm film) was used for separation of the analytes. GC-MS analysis was done with an Agilent HP-5 MS column (30 m x 320 μm, 0.25 μm film). The same column was used for Headspace GC-MS. Chiral GC-FID analysis was done by a Varian Dex-CB column or a MN hydrodex-β-6TBDM column.
HPLC analysis of chiral substrates was achieved by a DAICEL CHIRALCEL OJ or OD-H or CHIRALPAK AD-H column.
Quantification of substrates and products was done on GC-FID (He) and further GC-MS of each biotransformation was done for product identification. For that product standards were used as references. Table S30 shows the GC methods used for achiral analysis. All samples were measured as the corresponding methyl ester.  Chiral analysis of substrates and products was achieved on GC-FID (H2) and HPLC (organic). For the substrates and products different methods and columns were used. Biotransformations were either measured as methyl ester or underivatized. Racemic product and standards were used as reference compounds. Table S32 shows the methods used for chiral analysis on GC-FID.    [26] 11.67 11.45 2a 3.51 3.45 2c [27] 12.9 13.4 3a [28] 52.6 52.7 3c [29] 9.6 12.6 4a 9.0 8.9 4c [30] 4.4 4.0 5a 8.5 8.4 5c [31] 7.6 6.8 5d [32] 15.2 14.4