Rationalizing the Unprecedented Stereochemistry of an Enzymatic Nitrile Synthesis through a Combined Computational and Experimental Approach

Abstract In this contribution, the unique and unprecedented stereochemical phenomenon of an aldoxime dehydratase‐catalyzed enantioselective dehydration of racemic E‐ and Z‐aldoximes with selective formation of both enantiomeric forms of a chiral nitrile is rationalized by means of molecular modelling, comprising in silico mutations and docking studies. This theoretical investigation gave detailed insight into why with the same enzyme the use of racemic E‐ and Z‐aldoximes leads to opposite forms of the chiral nitrile. The calculated mutants with a larger or smaller cavity in the active site were then prepared and used in biotransformations, showing the theoretically predicted decrease and increase of the enantioselectivities in these nitrile syntheses. This validated model also enabled the rational design of mutants with a smaller cavity, which gave superior enantioselectivities compared to the known wild‐type enzyme, with excellent E‐values of up to E>200 when the mutant OxdRE‐Leu145Phe was utilized.


Site directed mutagenesis via Quick-Change PCR
To construct OxdRE-mutants with increased or decreased cavity size, the cavity forming amino acids were changed by means of site directed mutagenesis. As method the Quick-Change PCR was used. In Table 1 the components for the Quick-Change PCR are listed. The designed primers are listed in Table 2. As example Met29Ala_fw and Met29Ala_rw are the forward and reverse primers for the mutation of the Methionine29 towards Alanine29. The same goes for the Leu147Ala_fw, Leu147Ala_rw, Ala145Gly_fw and Ala145Gly_rw. Also, the mutated sequences are displayed in red. The Quick-Change PCR was done with the following program in Table 3.

DPN1 -Treatment, Ligation and Transformation
The PCR-product was treated with DPN1 for 1 hour at 37 degrees. After the DPN1 treatment the PCR-product was treated with T4ligase, which is not a necessary step, but recommended one to ligate the PCR-product to a closed vector. This closed vector was then transformed to competent Bl21 DE3 (Star) cells.
For the transformation the competent cells, which are stored at -80°C, are placed in a vessel with ice. Then the ligation solution or the plasmid of choice is added to the competent cells and incubated for 30 minutes. Afterwards the cells are heat shock treated for 90 seconds at 42°C. Then 700 µL LB-Medium is added and the cell solution is incubated for 3 hours at 37°C and 1200 rpm. The suspension was then centrifuged at 11.000 g and 600 µL of LB-Medium was discarded. The cell pellet was then resuspended again in the left 100 µL medium. The cell suspension of 100 µL was then poured over an LB-agar-plate and incubated over night at 37°C

Agar plate preparation
The agar plates were prepared by dissolving 10 g of LB-medium, 6 g of agar in 400 mL of water. The solution is then sterilized using an autoclave. The thick solution of agar is cooled down until ~60°C, before adding the respective antibiotics (100 µg/ml endconentration). Afterwards the solution is poured into sterile plates.

Preculture
The preliminary cultures were prepared in test tubes. 5 ml of LB-medium and the desired antibiotic with a concentration of 50 µg/ ml were transferred to the sterile test tubes. The colonies were removed from the plate with a toothpick and then used to inoculate the precultures.

Main culture
For the main culture, sterilized 500 mL-Flask with LB-Medium (400 mL) and Kanamycin (50 µg·mL -1 ) were used. The cultivating started by adding 5 ml of the preculture. The main culture was incubated at 20°C, 155 rpm for 72 hours. The so-called leaky expression was the most suited one for the expression of aldoxime dehydratase.  Table 4 and loaded with 10 μL of 1 mg/mL protein solutions. The protein concentrations were measured via standard Bradford assay, using Bradford reagent and protocol from Sigma-Aldrich. The samples were mixed with 6 μL of loading buffer and filled the rest with water to a total volume of 30 μL and an end concentration of 1 mg/ml. A voltage of 80 V was set for the collecting gel. When the separation gel was reached, the voltage was increased to 120 V, the maximum current reached was 15 mA. The stained gel shown in figure 1.  The SDS-PAGE shows that the mutants are as well expressed as the wild type OxdRE, but they are not as stable as the wild type, with much smaller bands in the crude extract as in the lysate. For example, the OxdRE-3G-mutant is not stable at all with no protein band on the SDS-PAGE for the crude extract row.

Aldoxime synthesis 3.1 General procedure 1 (GP 1): nitroaldol condensation of fluoro-benzaldehyde (1) with nitromethane
The synthesis was carried out in analogy to Betke et al. 1 Fluoro-benzaldehyde (1.0 eq.), ammonium acetate (0.1 eq.) and nitromethane (6.5 eq.) were dissolved in acetic acid. The solution was heated to reflux for 24 hours. The complete conversion was determined by means of TLC (cyclohexane/ethyl acetate, 6:1, v/v). Water (1:1, v/v) was then added and the phases separated. The solid was filtered off and recrystallized from ethanol.

General procedure 2 (GP 2): Michael-addition of methyl-magnesium-bromide to fluoro-nitrovinyl benzene
In analogy to Betke et al 1 , the synthesis was carried out in a heat dried flask under argon atmosphere. Dry toluene was placed in the flask and cooled to 0 ° C. After addition of the fluoro-nitrovinyl benzene (1 eq), a 3 M solution of methyl magnesium bromide in diethyl ether (1.5 eq) was added dropwise under argon reflux and further stirring. After three hours of reaction time at 0 ° C., a conversion control was conducted by means of TLC (cyclohexane/ethyl acetate, 6:1, v/v). The reaction mixture was mixed with a saturated ammonium chloride solution (1:1, v/v). The phases were separated, and the aqueous phase was extracted three times with ethyl acetate (1:1, v/v). The combined organic phase was washed with saturated sodium chloride solution (1:3, v/v). The organic phase was dried over magnesium sulfate and the solvent was then evaporated in vacuo. The crude product was then purified by automated column chromatography (cyclohexane/ethyl acetate 6:1, v/v) to give the racemic nitroalkane as oil.
Yield: 4.32 g, 53%. The 1 H-NMR-spectroscopy were compared with an analog compound from literature 1

General procedure 3 (GP 3): Synthesis of rac-(E/Z)-fluoro-phenylpropanal oxime using benzyl bromide
The synthesis was carried out in a heat dried flask under argon atmosphere. Dry THF was KOH (85 wt.% pellets, 1.05 eq.) was dissolved in dry THF, 4 Å molecular sieve were added, and the suspension was stirred for half an hour. Benzyl bromide (1.1 eq.) and tetrabutylammonium iodide (TBAI, 0.05 eq.) were added. The respective nitroalkane was then slowly added, dropwise. The suspension was stirred at room temperature for three hours. The conversion was controlled by means of TLC. Water (1:1 v/v) was added and the phases were separated. The aqueous phase was extracted three times with ethyl acetate. The combined organic phases were washed with brine and dried over magnesium sulfate. The solvent was removed under reduced pressure and the crude product obtained as an oil. Rac-(E/Z)-fluoro-phenylpropanal oximes were purified via automated column chromatography using the biotage system (cyclohexane/ethyl acetate), the mixture of E-and Z-isomers were obtained as colorless solid or as an oil. The aggregate state was dependent on the E/Z ratios.

General procedure 4 (GP 4): Copper catalyzed nitrile synthesis
Copper(II)acetate (0.1 eq.) was dissolved in acetonitrile at RT. Rac-(E/Z)-fluoro-phenylpropanal oximes (1.0 eq.) were added and heated to reflux for 90 minutes. Conversion was controlled by means of TLC (cyclohexane/ethyl acetate, 6: 1 v/v). The solvent was removed under reduced pressure to give the crude product. The crude product was dissolved in cyclohexane/ethyl acetate (8:1, v/v) and purified by means of a glass frit silica gel (silica gel, 3 cm filling height). The pure product was obtained as an oil. 1

General procedure for docking with MOE
The docking was performed with MOE (Molecular Operating Environment). 2 First the protein was prepared, therefore the co-crystal of OxdRE (PDB 3A16) was used. The preparation included protonation, adding missing atoms, preparing the heme structure and minimizing the protein energy in the corresponding force-field (MMFF94). The ligands were prepared using MOE to design the ligands and minimize within the force-field of interest. Over many optimization steps regarding the placement and scoring, including also reaction mechanism and experimental data, a final docking method was developed to find realistic protein ligand structures. These protein ligand structures (pose) were then used for general rationalizing of the enzyme's selectivity. In the following part procedure for a successful docking is tailored.

Initial docking:
The PDB data 3A16 contained the crystal structure of OxdRE as co-crystal with propanal oxime as ligand was loaded into the software MOE (molecular operating environment). Using the co-crystal and the postulated mechanism for aldoxime dehydratase we could tailor the binding motif of the ligand. According to the mechanism the aldoxime has to bind the heme iron(II) and need to be coordinated via serine and histidine. The serine is here the hydrogen acceptor and the histidine the hydrogen donor. The co-crystal does only show the serine and the heme interaction towards the aldoxime as ligand (Scheme 1). Even though the hydrogen donation towards the aldoxime is not recognized, the oxygen πelectrons show towards the histidine, which indicates the right ligand binding. Furthermore, the binding motif itself was then quantified in binding distances, angles and dihedrals. Afterwards the same ligand (propanal oxime) was redocked into the active site. The redocked structure gave us the variance of these quantified values, which we then could use as "cut-off" values for other ligands (Scheme 2).

Scheme 2:
Schematic representation of the bond angles, distances and dihedrals as 2D structure.

Docking with ligands of interest phenyl propanal oxime derivatives
The enzyme can convert PPOX derivatives into opposite enantiomers depending on the isomer used for the reaction. First of all, for each substrate with two enantiomers and two isomers we generated in total 4 ligands for the docking study. The ligands were then docked inside the active site using GDP 1 (general docking procedure 1). The results gave us a hint how the ligand could fit inside the pocket (active site) of the enzyme. The final poses for each ligand did not show any interaction patterns in terms of hydrogen bonds or metal ligation. Also the amount of possible poses are way too many and different, which does not allow any comparison between each ligand.

Placement:
Triangle matcher: Poses are generated by aligning ligand triplets of atoms on triplets of alpha spheres in a more systematic way than in the Alpha Triangle method. Pharmacophore: Information from the pharmacophore is directly used to place the ligand in the binding site. This is the preferred method when a pharmacophore is present, but will not work if there is no pharmacophore.

Pharmacophore generation
The pharmacophore docking describes a placement method where the ligand is forced into specified areas for indicated interaction patterns, such as hydrogen bonds metal ligation and hydrophobic or aromatic interaction. The pharmacophore docking is more powerful when a lead structure and a mechanism is already known. In this case, the co-crystal as the lead structure, as well as the mechanism is already known and given 3 . In figure 2 the generated pharmacophore is shown. The orange sphere describes an aromatic interaction and the green sphere a hydrophobic interaction. Both of these pharmacophores are not essential just giving a possible position for either an aromatic or a hydrophobic part of the ligand. The nitrogen is a hydrogen acceptor (blue sphere) with the right angle the electron pair need to show in the direction of the iron(II), which is why the bigger blue sphere (hydrogen acceptor projection) is around the iron(II). The hydroxyl group is hydrogen acceptor as well as a donator. Therefore, the sphere for the hydroxyl group has also two different projection spheres one for hydrogen acceptor projection (towards histidine) and one for hydrogen donator projection (towards serine). This projection spheres are created according to the co crystal and the mechanism behind the aldoxime dehydration. The position and the projections are determined with the measured angles, dihedrals and distances. The pharmacophore as such is the result of the correct binding distances, angles and dihedrals.
The pharmacophore can be used to define the binding motif such as the position of the functional group and its orientation. For example, a good pose is not just be described by the presence of the ligand inside the pocket but by the orientation of the aldoxime ligand, in which the aldoxime function shows a specific position with specified angles (His-O-Ser ~115°), dihedrals (C-N-O-O, ~85°) and distances (Fe-N, ~2.5 Å; O-Ser, ~2.8 Å; O-His, 2.7 Å) towards the heme group and the catalytic triad. Only when these prerequisites are fulfilled, the subsequent dehydration step can proceed. With these defined parameters in hand, a pharmacophore query was created which then allowed us to find ligand poses with this specified binding motif. Even though metalloenzymes are not well parameterized 4.5 a combination of MMF94 force field and the scoring functions London dG and Affinity dG proved to be suited for this issue. The protein model was prepared using the preparation kit given by MOE. As a representative example, the 2D-figure in Figure 2 shows the interaction pattern of the enantio-and diastereomerically pure ligand (Z,R)-3FPPOX with the active site of the enzyme OxdRE-WT. The ligand shows a hydrogen donation to the Ser219 and the His320 donates a hydrogen towards the oxygen of the ligand, while the nitrogen is coordinated by the Fe II -metal center of the heme. The phenyl group of the ligand interacts with the porphyrin ring via π-πstacking and the aldoxime function for each isomer is always in the same position.

Set up and parameters
Forcefield and parameter set up. MMFF94 is well parametrized for nitrogen interactions. Docking was performed with the MMF94 force field. In the following the potential set up with MOE is shown (left side). Enabled are all potential energy terms for bonded, van der Waals and electrostatic interaction and restraints. The cut off for non-bonded interaction with 8 to 10 was not changed. For solvation we used the reaction field model. The threads value just indicates the number of processors which were used for the calculation. In addition, in the following the docking panel in MOE is also shown (right side).

Affinity dG Scoring:
This function estimates the enthalpic contribution to the free energy of binding using a linear function: where the f terms fractionally count atomic contacts of specific types and the C's are coefficients that weight the term contributions to the affinity estimate. The individual terms are 2 :

London dG Scoring (default):
The London dG scoring function estimates the free energy of binding of the ligand from a given pose. The functional form is a sum of terms: where c represents the average gain/loss of rotational and translational entropy; E flex is the energy due to the loss of flexibility of the ligand (calculated from ligand topology only); f HB measures geometric imperfections of hydrogen bonds and takes a value in [0,1]; c HB is the energy of an ideal hydrogen bond; f M measures geometric imperfections of metal ligations and takes a value in [0,1]; c M is the energy of an ideal metal ligation; and Di is the desolvation energy of atom i. The difference in desolvation energies is calculated according to the formula where A and B are the protein and/or ligand volumes with atom i belonging to volume B; R i is the solvation radius of atom i (taken as the OPLS-AA van der Waals sigma parameter plus 0.5 Angstrom); and c i is the desolvation coefficient of atom i. The coefficients {c, c HB , c M , c i } were fitted from ~400 x-ray crystal structures of protein-ligand complexes with available experimental pK i data. Atoms are categorized into ~12 atom types for the assignment of the c i coefficients. The triple integrals are approximated using Generalized Born integral formulas. 2

Energy adjustment
After finding the right placement method, the scoring of these poses had to be adjusted, since only the structures, which had the essential interactions should be in the comparison data file and the energy values should represent the experimental data. The weighting of the parameters for hydrogen bond, metal ligation, hydrophobic contact and hydrophobic-polar interaction were increased. The adjustment in parameters led to improved protein-ligand-structure and also the associated energy values. These values and structures would then describe the best approximation of reality. General docking procedure 1 Receptor is set to 3A16 including the protein, the solvent and the heme as receptor. Docking site was the complete active site including all amino acids next to the heme group. No walls of constraints were used. Pharmacophore was set to none. No electron density file. The ligand was set to MOE selected atoms, which includes all selected ligands for the calculation. As placement the triangle matcher and as refinement the rigid receptor option was used. The scoring functions for initial score was London dG and for final score affinity dG. Number of placement poses were set to 200 and the number of shown poses to 20.

General docking procedure 2
Receptor is set to 3A16 including the protein the solvent and the heme as receptor. Docking site was the complete active site including all amino acids next to the heme group. No walls of constraints were used. The designed pharmacophore file was chosen for the pharmacophore set up. No electron density file. The ligand was set to MOE selected atoms, which includes all selected ligands for the calculation. As placement the pharmacophore search were chosen and as refinement the rigid receptor option was used. The scoring functions for initial score was London dG and for final score affinity dG. Number of placement poses were set to 200 and the number of shown poses to 5.

Docking results with wild type OxdRE
Docking of PPOX derivatives for OxdRE-WT was performed according to GDP2. Also, the partial energy contribution of the aldoxime function was determined.  The energy contribution of the aldoxime function is random ( Figure 3) there is no dependency between the enantioselectivity and the interaction pattern of the functional group. Figure 3 shows the sum of energy contributions of the functional group in dependency of the formed enantiomer. For example, when taking the value obtained from the ΔG-value for the S-enantiomer of a specific E-or Zaldoxime minus the ΔG-value for the R-enantiomer of this E-or Z-aldoxime, 2FE has a ΔΔG-value of 1.87 kcal mol -1 , which is positive and predicts the S-enantiomer (2FER-2FES, -7.36 kcal mol -1 -(-9.31 kcal · mol -1 )), therefore the Δ-sum-energy has to be also positive, but in fact it is not. A dependency with the enantioselectivity would be indicated with an alternating graph.

Docking results with OxdRE-L145F
Docking of PPOX derivatives for the L145F mutant of OxdRE was performed according to GDP2. The partial energy contribution was not determined. Red numbers indicate low docking quality. Docking results with OxdRE-M29A-L145A-A147G Docking of PPOX derivatives for the triple point mutant of OxdRE was performed according to GDP2. The partial energy contribution was not determined. Red number indicates low docking quality.

Biocatalytic chiral nitrile synthesis
The biotransformation was performed with whole cells of Oxd -variants. First the buffer (PPB 50 mM, pH 7) was placed into a 2 mL Eppendorf vessel. Then the substrate solution, containing a FPPOX derivate in DMSO (100 µL, 100 mM), was added to the buffer to an end concentration of 10 mM FPPOX and 10% DMSO. The temperature was arranged to 20°C. Afterwards the enzyme as whole cell solution was added to the reaction mixture (0.3 U). The reaction solution was stirred for 1h min at 1000 rpm. The reaction was quenched and extracted with 800 µL of cyclohexane. The organic phase was analyzed via HPLC.

SFC-HPLC
The conversion was determined using a calibration curve for the SFC HPLC and using 1 H-NMR. The enantiomeric excess was determined by means of chiral, supercritical fluid chromatography (SFC-HPLC). The OB-H column from Chiracel TM served as stationary phase. The mobile phase consisted of CO 2 /isopropanol mixture (98:2, v/v). The method runs for 30 mins at 1.0 ml/min, the injection volume is 20 μL and the measurement runs at 20 ° C. The absorbance was measured at 210 nm.

MS and HRMS
The routine MS-spectra were measured via Esquire 3000. The Agilent Techn. 6220 TOF LCMS with ESI as ionization method was used for the accurate mass measurement.