Better than Nature: Nicotinamide Biomimetics That Outperform Natural Coenzymes

The search for affordable, green biocatalytic processes is a challenge for chemicals manufacture. Redox biotransformations are potentially attractive, but they rely on unstable and expensive nicotinamide coenzymes that have prevented their widespread exploitation. Stoichiometric use of natural coenzymes is not viable economically, and the instability of these molecules hinders catalytic processes that employ coenzyme recycling. Here, we investigate the efficiency of man-made synthetic biomimetics of the natural coenzymes NAD(P)H in redox biocatalysis. Extensive studies with a range of oxidoreductases belonging to the “ene” reductase family show that these biomimetics are excellent analogues of the natural coenzymes, revealed also in crystal structures of the ene reductase XenA with selected biomimetics. In selected cases, these biomimetics outperform the natural coenzymes. “Better-than-Nature” biomimetics should find widespread application in fine and specialty chemicals production by harnessing the power of high stereo-, regio-, and chemoselective redox biocatalysts and enabling reactions under mild conditions at low cost.

mNADHs (mNADH4 (1-5)b) were synthesised to bind in the active site of the ERs generating a "dead-end" complex which should be suitable for crystallisation and structural biochemical characterisation. The general structure of the tetrahydro-biomimetics is as follows: Pd catalyst on carbon (25 mg; extent of labelling 10% w w -1 loading matrix activated carbon support, Sigma Cat. 20, 569.9) was added into a Schlenk tube containing dried ethanol (6 mL). Then, biomimetic 1 (100 mg, 0.467 mmol) was added under inert atmosphere (no stirring). The solution became yellow.

S3
Finally, the Schlenk tube was evacuated and filled with hydrogen at atmospheric pressure. The reaction was run for approximately 20 h. A small aliquot of the reaction mixture was analysed by GC-MS as well as TLC. After removal of the Pd/C, the solution was colourless, indicating that all the starting material was reduced. The work-up was performed by filtering over celite and extensive washing with methanol. The clear solution was evaporated to afford a colourless oil (80 mg of crude product).
Judging from GC-MS analysis, the conversion into the desired reduced biomimetic 1b was 50% whereas the rest of the material were side-products generated by further reduction and/or cleavage of 1b.
Therefore, the reaction was repeated as previously described using biomimetic 1 (100 mg) and Pd/C (25 mg) in dried ethanol (6 mL). The reaction time was shortened to 2h and 30 min to determine whether a better chemoselectivity could be obtained. Reducing the reaction time led to the formation of higher amounts of the desired product 1b. The composition of the product mixture (determined by GC-MS) was: 82% of 1b, 13% of the fully reduced by-product (i.e. 1-benzylpiperidine-3-carboxamide) and 4% of cleaved derivatives.
The aliquots of the crude mixture from the first and the second reaction were combined and purified together by column chromatography on silica.
Conditions for column chromatography: GC-MS analysis showed that the "mono"-reduced desired product 2b was formed with 31% conversion.
The work-up was performed by filtering over celite and extensive washing with methanol. The clear solution was evaporated to afford a colourless crude oil that was purified by column chromatography.
Conditions for column chromatography: The crude mixture was purified by column chromatography on silica. GC-MS showed that the desired product 3b was formed with quantitative conversion (> 99%). Therefore, in this case, the hydrogenation was completely chemo-and regioselective, even when the reaction time was prolonged to 16 h.

S10
The work-up was performed by filtering over celite and extensive washing with methanol.

Expression and purification of the ERs
All the enzymes used during this study were expressed and purified as described previously. [2][3][4] TsOYE was additionally purified by anion exchange chromatography using a HiPrepQ HP 16/10 column (GE Healthcare). The heat-shock purified protein was dissolved in start buffer (20 mM Tris/HCl, pH 8.0 buffer) and loaded onto the column. The elution of the protein was performed with a gradient between start buffer and elution buffer (20 mM Tris/HCl, 1M NaCl, pH 8.0 buffer).
All proteins were stored as concentrated stocks in 50 mM MOPS-NaOH, pH 7.0 buffer, supplemented with 5 mM CaCl2 at -20 °C.

Determination of the molar extinction coefficient for mNADHs (1-5)
24-32 mg of mNADHs (1)(2)(3)(4)(5) were dissolved in a certain amount of DMSO (600 µL for 1 and 2, 2.5 mL for 3 and 4 and 50 mL for 5) and filled up with buffer (50 mM MOPS-NaOH, pH 7.0, supplemented with 5 mM CaCl2) to 500 mL using a volumetric flask. Then, 4 dilutions were prepared by pipetting 5, 10, 20 and 40 mL in 100 mL volumetric flasks. UV-vis absorbance spectra were recorded and the absorbance of selected wavelengths was plotted against the concentration of the biomimetic-solution. The slope of the linear fit represents the molar extinction coefficient.

Figure S1
. UV-vis absorbance spectra of mNADHs (1-5) at 4 different concentrations. The insets represent the absorbance at the peak-maximum plotted against the biomimetic concentration for obtaining the molar extinction coefficient. S18

Reductive half-reaction
The reductive half-reactions between ERs, NAD(P)H and mNADHs were investigated using a High-Tech stopped-flow spectrophotometer (TgK Scientific Limited). The experiments were prepared and run under anaerobic conditions in N2-environment as previously described. All measurements were acquired in MOPS-NaOH buffer (50 mM, pH 7) supplemented with CaCl2 (5 mM The reductive half-reaction was modelled as shown in (1) (1) where ER(ox) is the enzyme bound flavin in its oxidised state, [ER(ox)-NAD(P)H] is the enzyme-coenzyme complex and ER(red) is the enzyme bound flavin in its reduced state upon immediate NAD(P) + release after the reduction process. The dissociation constant KD is therefore determined according to equation (2): (2) Figure S2 to Figure S5 show the hyperbolic fits for the reductive half-reaction of ERs with NAD(P)H and mNADHs.        The following table shows the enantiomeric excess for the conversion of ketoisophorone 6 to levodione (R)-6a using 12 different ERs. The table with conversions can be found in the main paper ( Table 3).

Discussion of the results from biotransformations
The applicability of the mNAD biomimetics in organic synthesis as inexpensive alternatives to natural nicotinamide coenzymes was evaluated for the asymmetric reduction of an activated alkene as the test substrate. An extended panel of 12 ERs from the OYE family (PETNR, TOYE, OYE2, OYE3, XenA, XenB, LeOPR1, NerA, MR, TsOYE, DrOYE and RmOYE) was employed for the conversion of ketoisophorone 6 to levodione (R)-6a (see Table 3, main paper).
The results of the screening showed that the biomimetics were overall well accepted, the only exception being biomimetic 5. This finding is in line with the kinetic parameters previously described in and MR, that generally afforded poor conversions (<42%) with all the mNADHs.
As expected, the source of hydride did not have a significant influence on the enantiomeric excess (

mNAD + s recycling with the rhodium complex
So far, biocatalytic reactions with the biomimetics were carried out using stoichiometric amounts of coenzyme with respect to the substrate. Since mNAD + 1a is known to be reduced by the rhodium complex [Cp*Rh(bpy)(H2O)] 2+ , a recycling system was implemented in the reaction (Scheme S1).
Scheme S1. The rhodium complex [Cp*Rh(bpy)(H2O)] 2+ was synthesized as previously described. 6 Starting from previously optimized reaction conditions for a different system with the natural nicotinamide coenzyme, 7  The full data set is reported in Table S4. Applying the same reaction conditions, a conversion of more than 30% was obtained using NAD + and the biomimetic 1a (Table S4, entry 2 and 3). The optical purity was the same (ee 85%). The reason of the imperfect optical purity might stem from a non-stereoselective chemical side reduction catalysed by the Rh complex (Table S4, entry 6). However, a further optimisation of the reaction conditions led to 66% conversion with an enantiomeric excess of 85%.   All crystals were flash frozen by plunge freezing in liquid nitrogen (no additional cryo protection was required) prior to data collection at Diamond Light Source Ltd.

Data Collection and Refinement
Individual data sets were collected from single cryo frozen crystals of XenA at Diamond Light Source. All data were scaled and processed using Xia2 8 and subsequently rebuilt and refined with iterative cycles of rebuilding and refinement carried out in COOT 9 and Phenix 10 . Validation with both MOLPROBITY 11 and PDB_REDO 12 was integrated into the iterative rebuild process. Complete data collection and refinement statistics can be found in Table S5.

Kinetic parameter for the conversion of cinnamaldehyde by NtDBR using NAD(P)H and mimics 1 and 3
Steady-state kinetic data for the conversion of cinnamaldehyde by NtDBR using NADPH as coenzyme were  Table S7 Figure S12. Reaction rates for NtDBR with a fixed concentration of coenzyme (6 mM) and varied concentrations of cinnamaldehyde. YcnD was expressed and purified as described previously. 14 The activity for YcnD with NAD(P)H and biomimetics (1-5) was tested spectrophotometrically. First, the UV-vis absorbance spectrum of the coenzyme was recorded. Then, enzyme (final concentration 400 nM) was added and after an incubation time of 5 min the spectrum was recorded again. Immediately after the addition of the coenzyme, the solution containing the flavin-dependent enzyme became colourless, indicating the reduction of the flavin. Then, the flavin underwent slowly reoxidation by molecular oxygen. A bleach of the UV-vis absorbance spectra of the coenzyme showed activity for YcnD towards NAD(P)H and biomimetics 1, 2, 3 and 5 which can be seen in Figure S13.