The biocatalytic approach of the asymmetric reduction of activated alkenes has emerged as a highly valuable tool for the synthesis of optically active compounds. Enoate reductases, NAD(P)H-dependent flavoproteins from the old yellow enzyme family, are catalyzing the chemo- and stereoselective carbon-carbon double bond reduction in a trans-specific fashion of a broad range of structurally diverse olefins bearing an electron withdrawing group (aldehydes, ketones, carboxylic acids/esters, lactones, nitriles, and nitroalkenes) thereby generating up to two new stereogenic centers (Hall et al., 2007, 2008a,b; 2010; Kosjek et al., 2008; Toogood et al., 2010). The non-racemic products of the bioreduction represent important building blocks. Prominent examples like Citronellal (Müller et al., 2006), LilialTM and HelionalTM (Stueckler et al., 2010), or Levodione (Liese et al., 2006) are used in various applications.
Due to the complexity of cofactor recycling in the past, the majority of asymmetric CC bioreduction were performed using whole fermenting cells, mostly baker´s yeast as well as anaerobic bacteria (Fuganti, 1990; Günther and Simon, 1995). However, due to the presence of competing carbonyl reductases, the chemoselectivies (i.e., CC versus CO reduction) were often low. In order to circumvent this limitation, purified ene-reductases were employed thereby necessitating an accompanying regeneration system for the nicotinamide cofactor [NAD(P)H]. Much effort has been made over the last decades in order to develop powerful in situ cofactor regeneration systems (Wichmann and Vasic-Racki, 2005; Woodyer et al., 2006).
While chemical and electrochemical methods often suffer from drastic problems such as enzyme deactivation or low total turnover numbers (TTN's), enzymatic approaches have emerged as the method of choice, particularly with regard to the recent success towards genetically engineered and biotechnologically improved cofactor recycling enzymes (Tishkov et al., 1999; Turner, 2009). Up to now, cofactor recycling in bioreductions using enoate reductases has been carried out by either complex electro-microbial/enzymatic cofactor recycling systems based on toxic mediators, such as ParaquatTM or the traditional coupled-enzyme approach using formate/formate dehydrogenase (FDH), glucose/glucose dehydrogenase (GDH), or glucose-6-phosphate/glucose-6-phosphate dehydrogenase (G6PDH), respectively (Hall et al., 2008a,b, 2010; Toogood et al., 2010). Recently, a new coupled-enzyme approach for NAD(P)H regeneration was reported (Kosjek et al., 2008) employing a phosphite dehydrogenase catalyzing the oxidation of phosphite to phosphate with the concurrent reduction of NAD(P)+ to NAD(P)H. However, only few data are available in literature concerning the applicability of this system.
Due to both low cost of substrate and favorable equilibrium the FDH regeneration system fulfills the conditions necessary for industrial standards and thus has been applied broadly. However, the presence of competitive carbonyl reduction activity in the FDH preparation (presumably caused by primary alcohol dehydrogenases present as impurities in commercial crude enzyme preparations) and product racemization both limit the use of this system in combination with enoate reductases for the reduction of activated olefins (Hall et al., 2007). The use of the GDH and G6PDH regeneration system to recycle NADH and NADPH, respectively, applied to ene-reduction turned out to be an adequate system yielding excellent results. However, the high costs of enzymes (especially GDH) and cosubstrate glucose-6-phosphate in case of G6PDH as well as difficulties in product isolation due to the high polarity of coproducts (gluconic acid or its phosphorylated analog) set a narrow limit to the potential of both systems.
Herein we report on an NADH-recycling system for the bioreduction of olefins by applying a highly efficient, inexpensive, and robust system based on concurrent alcohol-oxidation coupled to ene-reduction. In this system, the oxidation of the sacrificial sec-alcohol 2-propanol to acetone by the solvent-stable alcohol dehydrogenase (ADH-A) from Rhodococcus ruber (Edegger et al., 2006; Kosjek et al., 2004) has been coupled to the bioreduction of activated alkenes using ene-reductases (Fig. 1).
For ketone reduction employing 2-propanol as hydrogen donor a stripping process has been reported to remove the formed most volatile compound (acetone) to overcome the thermodynamic limitation and thereby enhancing the conversion (Goldberg et al., 2006). In order to get a deeper insight into the position of the equilibrium of the herein reported system, the Gibbs free energy (ΔG) for the redox reaction methyl-vinyl ketone + 2-propanol → 2-butanone + acetone used as a model was determined by quantum chemical calculations. ΔG was calculated to be −16 kcal mol−1 in aqueous solution (−20.5 kcal mol−1 for the gas phase). Consequently, complete conversion (>99%, ΔG = −RT ln K) is expected for the hydrogenation of the double bond of methyl-vinyl ketone employing 2-propanol as hydrogen donor. Thus, theoretically only a stoichiometric amount of cosubstrate is required.
Previous studies have shown that the majority of enoate reductases accept both NADH and NADPH at similar rates (Hall et al., 2010). In order to examine the potential of the 2-propanol/ADH-A system for the recycling of NADH, YqjM from Bacillus subtilis (Kitzing et al., 2005) and NCR from Zymomonas mobilis (Müller et al., 2007) were applied for the bioreduction of various substrates. The chosen substrates covered a broad spectrum on structurally and electronically diverse carbonyl moieties (α,β-unsaturated diester, cyclic and open-chain ketones, diketone, and aldehydes). Since an alcohol dehydrogenase (ADH-A) was used as recycling enzyme, competing carbonyl reduction posed a challenge to this approach. On the one hand substrate-depletion forming allylic alcohols as dead-end side products which are non-substrates for ene-reductases and on the other hand product depletion furnishing saturated alcohols due to over-reduction needed to be taken into account.
The screening was started with α-methylmaleic acid dimethylester 1a and the sterically more demanding 4-ketoisophorone 2a which were reduced to (R)-α-methylsuccinate [(R)-1b, e.e. up to >99%] and (R)-levodione [(R)-2b, e.e. up to >99%] using the formate/FDH recycling system (Hall et al., 2008a,b). Applying the 2-propanol/ADH-A (2 eq./2 U) recycling system, 1a was quantitatively converted to (R)-1b in excellent optical purity (conv. and e.e. ≥99%; TTN = 100, not optimized; Table I, entry 1) employing Zymomonas mobilis NCR-reductase. In case of YqjM from Bacillus subtilis, 1a was stereoselectively reduced to (R)-1b with high conversion (conv. = 90%, e.e. ≥ 99%, Table I, entry 1). The sterically more demanding substrate 2a was converted by both enzymes with good to excellent conversions at moderate stereoselectivity (Table I, entry 3). Fortunately, the carbonyl moieties of the diester and the sterically hindered cyclic diketone were not reduced by the alcohol dehydrogenase (ADH-A). Based on these promising results, the range of substrates was extended in order to examine the scope and limitations of the new recycling system. In order to gain more insight into the optimal reaction conditions, various amounts of the recycling enzyme (ADH-A, 0.01–8 U) as well as various cosubstrate concentrations (2-propanol, 0.6–20 eq.) were used. During these studies it was shown that the use of 2 U of ADH-A and a modest excess of 2 equivalents of cosubstrate turned out to be best in order to obtain high conversion but still keep the carbonyl reduction at a minimum (data not shown). An additional consumption of NADH due to the presence of oxygen-dependent NADH oxidases in the enzyme preparations was ruled out by running the biotransformations under argon atmosphere and obtaining comparable results as under air (data not shown).
n.c., no conversion; n.d., not determined; Absolute configurations of side products 3c, 4c, 5c, 5d, and 7c were not determined due to low conversion (<7%).
aStandard conditions: 10 mM substrate, 2-propanol (2 eq.)/ADH-A (2 U), reaction time 24 h.
c10 mM substrate, 2-propanol (2 eq.)/ADH-A (0.1 U), reaction time 24 h.
dCalculated based on alcohol side product.
Both cyclic α-substituted α,β-unsaturated ketones 2-methylcyclohexenone 3a and 2-methylcyclopentenone 4a were converted with good to excellent conversions to (R)-3b and (S)-4b, respectively, by both enzymes in high optical purities (Table I, entries 5 and 8), except for a modest enantiomeric excess of (S)-4b using NCR (Table I, entry 8). To our delight, the competing carbonyl reduction to the corresponding saturated alcohol causing product depletion was only observed in a negligible amount: less than 5% in case of the six-membered cyclic enone 3a (Table I, entry 7) and even less than 1% for the five-membered ring analog 4a (Table I, entry 10). The degradation of substrate by formation of the corresponding allylic alcohol (substrate depletion) was not observed at all. Hence, for both substrates, the 2-propanol/ADH-A system proved to be highly suitable for cofactor-regeneration yielding equally good results with respect to conversion and enantiomeric excess compared to the formate/FDH for 3a and glucose/GDH-systems for 4a (Hall et al., 2007, 2008a,b). Surprisingly, also the open-chain α,β-unsaturated ketone 5a, which is supposed to be a good substrate for alcohol dehydrogenase (ADH-A), was successfully transformed with high rates and enantiopurity to (S)-5b employing YqjM (conv. = 95%, e.e. = 94%, Table I, entry 11). For NCR good conversion albeit with moderate stereoselectivity (conv. = 88%, e.e. = 36%, Table I, entry 11) was obtained. Side products originating from substrate and/or product depletion were only detected in marginal quantities (<1% and <5% for the corresponding unsaturated and saturated alcohols, respectively; Table I, entries 12 and 13). In order to push the system limits, two aldehydes, 2-methylpent-2-enal 6a and citral 7a, were investigated. For these substrates, carbonyl reduction is expected to dominate due to the high carbonyl activity of the aldehyde moiety as compared to the ketone/ester substrates. Employing standard conditions, substrate 6a was mainly converted to the undesired saturated alcohol 6d (82% and 91% conversion for YqjM and NCR, respectively, Table I, entry 17) going in hand with low yields of the desired product (S)-6b (13% and 5% conversion for YqjM and NCR, respectively, Table I, entry 14). However, the competitive carbonyl reduction could be depleted by reducing the amount of ADH-A. Best conversion with lowest side product formation of 6c and 6d (corresponding unsaturated and saturated alcohol, respectively) was obtained using 0.1 U ADH-A and 2 eq. of cosubstrate (81% and 25% conversion for YqjM and NCR, respectively, Table I, entry 15). For citral 7a no chemoselective conversion to the desired product 7b was observed even by using only 0.1 U of ADH-A and the undesired –CHO reduction could not be avoided, leading to a mixture of isomers of the corresponding α,β-unsaturated alcohols (Z)-7c (nerol) and (E)-7c (geraniol) and the saturated analog 7d (citronellal) (Table I, entries 18–23). As anticipated, carbonyl reduction by the ADH-A recycling enzyme prevailed which makes the 2-propanol/ADH system inapplicable to enal-type substrates thereby showing the limitations of this approach. Consequently, one has to switch to the glucose/GDH system (Hall et al., 2008a,b) as a more expensive alternative.
We have successfully applied an alcohol-oxidation system for cofactor recycling to the bioreduction of olefins in a coupled-enzyme approach. This highly efficient system is based on concurrent alcohol-oxidation versus ene-reduction, in which an inexpensive sacrificial sec-alcohol (2-propanol) is oxidized to acetone by a solvent stable alcohol dehydrogenase (ADH-A from R. ruber) thereby providing a hydride via NADH to recycle the flavin cofactor of the ene-reduction catalyzed by enoate reductases. Due to the highly favorable position of the equilibrium towards the product side, theoretically only a stoichiometric amount of cosubstrate is required in order to obtain full conversion (ΔG = −20.5 kcal mol−1, equivalent to >99% conversion). From the viewpoint of atom economy, this system has a high atom efficiency fulfilling an important requirement for 'green' biotechnological processes. In order to examine the potential of this system, a range of substrates carrying structurally and electronically diverse carbonyl moieties (α,β-unsaturated ester, diketone, cyclic and open-chain ketones, and aldehydes) were applied. For all substrates undesired carbonyl reduction was negligible, except for highly activated aldehydes for which the system failed due to predominant reduction of the carbonyl moiety. With this study, we successfully contributed to the rising demand for effective, ecologically friendly and cheap cofactor recycling systems in order to make more complex cofactor-dependent enzymatic reactions attractive for preparative-scale applications.