Biocatalytic hydrogen-borrowing cascades

Abstract

The exquisite chemoselectivity and the intrinsic compatibility of enzymes have been widely exploited during the past decade for the development of multi-step biocatalytic reactions in one-pot. In this context, hydrogen-borrowing cascades permit to maximise the atom-efficiency through the internal recycling of redox equivalents, which avoids the use of additional oxidants or reductants. Here, we describe the state-of-the-art in the field of biocatalytic hydrogen-borrowing cascades and provide a future perspective for a wider implementation in organic synthesis.

Introduction

Multi-step chemical reactions in one-pot offer economic as well as environmental benefits because they avoid the need for isolation and purification of intermediates along a synthetic route (1–3). For instance, that results in a reduced consumption of solvents for isolation procedures and energy for evaporation and mass transfer. In this context, biocatalytic hydrogen-borrowing cascades possess the most elevated atom-economy, since the electrons liberated in the oxidative step are quantitatively consumed in the concurrent reductive step (4). A concurrent oxidation-reduction process in a single vessel without compartmentalisation is possible as a consequence of the exquisite chemoselectivity and intrinsic compatibility of enzymes. Such systems are inspired from Nature’s complex redox networks. This contribution aims at providing an overview in this field and a guideline for future developments.

Discussion

Racemisation, isomerisation and disproportionation

The simplest biocatalytic hydrogen-borrowing reaction is the racemisation of optically active alcohols (5). Although destroying central chirality can be perceived as an undesired reaction, racemisation is needed for either the recycling of undesired enantiomers after kinetic resolution, or in dynamic kinetic resolution processes. In general, this process requires two stereocomplementary alcohol dehydrogenases (ADHs), since non-stereoselective ADHs are rare in nature (Scheme 1a). However, non-selective ADHs suitable for racemisation of alcohols were created by rational enzyme engineering (6). The racemisation system relies on the reversibility of the dehydrogenase reaction so that both alcohol enantiomers are in equilibrium via the ketone intermediate. The driving force for the racemization is the increase of entropy.

Scheme 1.

Hydrogen-borrowing processes for the racemisation of alcohols (a), the isomerisation with generation of a stereogenic centre (b, c) and asymmetric disproportionation (d).

Properly balanced catalytic amounts of the oxidised and reduced form of nicotinamide adenine dinucleotide coenzyme (NAD(P)) are sufficient to enable the dynamic process. A variation of this method is the racemisation of a-hydroxy carboxylic acids, albeit anoxic conditions were required to avoid accumulation of the a-keto intermediate (7).

Biocatalytic hydrogen-borrowing isomerisation was first shown in the interconversion of morphine to hydromorphinone via the combination of a NAD-dependent ADH in the oxidative step and a NADP-dependent ene-reductase (ERed) in the reductive step (8). Due to the divergent cofactor specificity of the two enzymes (NAD⁺ vs NADP⁺), a third enzyme (i.e. transhydrogenase) was required for shuttling the hydride from NADH to NADP⁺. In a similar concept, allylic alcohols can be isomerised into ketones (Scheme 1b) (9). The concept was exemplified for the isomerization of cyclohex-2-en-1-ol into cyclohexanone. Nevertheless, conversion and chemoselectivity were limited to 60% and 90%, respectively, due to the occurrence of an over-reduction to cyclohexanol.

The combination of EReds with aldehyde dehydrogenases (AldDHs) allowed for the hydrogen-borrowing conversion of a-substituted α,β-unsaturated aldehydes into optically active a-substituted carboxylic acids (Scheme 1c) (10). The set-up of the cascade was particularly challenging as: i) the AldDH has to be chemoselective for the oxidation of the aldehyde moiety of the intermediate (i.e. leaving the substrate untouched); ii) the intermediate is an a-substituted chiral aldehyde and therefore it is prone to unwanted racemisation that can reduce the optical purity of the final product. Hence, in this study, a systematic methodology for creating hydrogen-borrowing biocatalytic cascades was presented; the method allowed for obtaining the final products with up to >99% conversion, 95% chemoselectivity and 99% enantiomeric excess. Another independent work previously reported the same cascade as a proof of concept (11).

The asymmetric biocatalytic Canizzaro-type reaction catalysed the disproportion of enolizable aldehydes to afford, in theory, equimolar amounts of alcohols and carboxylic acids. In practise, the disproportion of 10 mM substrate in presence of equimolar NAD⁺ and NADH (0.5 mM each) provided imbalanced mixtures of products (± 10%). An elegant application was the disproportionation of a racemic α-substituted aldehyde to give enantioenriched mixtures of products, as a consequence of the in-situ spontaneous racemisation of the starting material as well as the enantioselectivity of the ADHs (Scheme 1d) (12).

Synthesis of lactones

The first example of in vitro biocatalytic hydrogen-borrowing reaction dates back 1991, reporting the combination of an ADH with a Baeyer-Villiger monooxygenase (BVMO) for the redox self-sufficient conversion of cyclic alcohols to lactones (Scheme 2a) (13). The method was exemplified for the conversion of endo-bicyclo alcohols with up to 95% conversion, elevated chemoselectivity (>95%) but imperfect e.e. (11% - 86%). Afterwards, again Willetts and coworkers identified new BVMOs, which permitted to improve the stereoselectivity of the cascade (e.e. >95%) (14). The same methodology was exploited two decades later for the conversion of cyclohexanol to ε-caprolactone in different independent works. In a first report, cyclohexanol (10 mM, 30 mL scale reaction) was converted into ε-caprolactone with 55% isolated yield and perfect purity (15). The main issue was the deactivation of the cyclohexanone monooxygenase from Acinetobacter calcoaceticus (CHMO) above 10 mM (free enzyme) and 30 mM (immobilised enzyme) concentration of cyclohexanol. In contrast, a second concomitant work reported the highest productivity for ε-caprolactone (6.45 g L⁻¹, 97% conversion) at 60 mM substrate concentration, although the same CHMO was apparently applied.(16) Bornscheuer and coworkers extended the cascade via including a third enzyme, namely Candida antarctica lipase A, for performing the concurrent oligomerisation of ε-caprolactone (17). The system was run in a truly hydrogen-borrowing fashion up to 60 mM reaction scale (99% conversion). The reported reaction at 200 mM scale required stoichiometric amounts of acetone and glucose with compartmentalization of ADH and CHMO in different E. coli resting cells.

Scheme 2.

Hydrogen-borrowing cascades for the conversion of alcohols into lactones.

Kroutil and coworkers demonstrated the hydrogen-borrowing conversion of cyclohexanol to ε-caprolactone at 200 mM scale, affording 99% conversion (i.e. 96% lactone and 3% cyclohexanone) (18). Liese and coworkers have recently reported a kinetic analysis of the process (19).

An alternative strategy is the convergent conversion of a mixture of cyclohexanone and 1,6 hexanediol (2:1 equivalent ratio) to give, in theory, 3 equivalents of ε-caprolactone (Scheme 2b) (20). On a 50 mL scale, Kara and coworkers converted cyclohexanone (98 mg, 2 eq.) and 1,6 hexanediol (59 mg, 1 eq.) into 146 mg of a mixture of: ε-caprolactone, poly-ε-caprolactone and 1,6 hexanediol (1:1.5:0.6). In a follow-up study, they improved remarkably the overall efficiency of the cascade with respect to conversion into ε-caprolactone (70%) and turnover numbers (4 x 10⁵ for ADH, 5.8 x 10³ for CHMO and 10³ for NADPH) (21). Very recently, the substrate scope was further increased to four and five membered ring ketones as well as bicyclic ketones (22).

Synthesis of amines

The direct conversion of alcohols into amines is an important chemical transformation. The first biocatalytic, hydrogen-borrowing amination of alcohols made use of three enzymes in a concurrent two-step cascade (Scheme 3a) (23). In the first step, an ADH oxidises the primary alcohol to the aldehyde intermediate at the expense of NAD⁺. In the second step, a ω-transaminase (ωTA) transfers the amino group from the donor (l-alanine) to the acceptor (the aldehyde intermediate) generating the amine product and pyruvate. Finally, a l-alanine dehydrogenase (AlaDH) connects the two cycles through the reductive amination of pyruvate to alanine. In the last reaction, ammonia is consumed and NADH is oxidised back to NAD⁺. The applicability of the method was demonstrated for the diamination of α,ω-diols that proceeded with up quantitative conversion at 50 mM scale. The requirement for 5 equivalents of l-alanine (250 mM), besides the excess of NH₄⁺ (275 mM), erodes the atom-efficiency of the method. The extension of the method to the asymmetric amination of secondary alcohols led to a maximum of 54% conversion along with a considerable accumulation of the ketone intermediate (32%) (24). The same method was applied in a subsequent work for the amination of ether alcohols (10 mM) requiring sub-stoichiometric l-alanine (0.5 mM), ammonium species (80 mM) and NAD⁺ (1 mM); the final concentration of the amine product never exceeded 40% conversion (25).

Scheme 3.

Hydrogen-borrowing cascades for the conversion of alcohols into amines.

Interestingly, it was calculated that the reductive amination of pyruvate to give l-alanine is the only exergonic step, which therefore drives the overall equilibrium. In the same way, the mono-amination of isosorbide (300 mM) was demonstrated (7% conversion) (26). Finally, Wendisch and coworkers extended the concept from in vitro to in vivo. They created an artificial operon encoding ADH, ωTA and AlaDH and under optimised conditions, resting E. coli cells supplemented with l-alanine (100 mM) and NH₄⁺ (100 mM) aminated quantitatively 1,10-decanediol (10 mM) with a productivity of 1.16 mmol g_cells^-1 h^-1 (27,28).

The break-through in the field came in 2015 when Mutti, Turner and coworkers presented the dual-enzyme, asymmetric, hydrogen-borrowing amination of alcohols.(29,30). The method profited of the implementation of amine dehydrogenases (AmDHs): a new class of enzymes capable of performing the reductive amination of carbonyl compounds (31–34). The cascade depicted in Scheme 3b possesses the highest possible atom efficiency as an alcohol is converted into an enantiopure amine, consuming ammonium and generating water as the sole coproduct. A broad panel of structurally diverse aromatic and aliphatic secondary alcohols (20 mM) was aminated with up to 96% conversion and 99% e.e., whereas primary alcohols were aminated quantitatively. 2 M of NH₃/NH₄⁺ provided the thermodynamic pull for the reaction (under the reaction conditions ΔG° ≈ 0). The amination of racemic alcohols required the simultaneous use of two stereocomplementary ADHs in combination with the AmDH. Another independent work showed the same concept (35). In related publications, the deracemisation of mandelic acid (30 mM) was coupled with a dual-enzyme hydrogen-borrowing amination to afford l-phenylglycine (94% conversion, e.e. >97%) at the expense of ammonium chloride (200 mM) (36,37).

Other cascades

Dual-enzyme hydrogen-borrowing cascades combining a P450 monooxygenase with an ADH enabled to convert cyclic alkanes to ketones with turnover frequencies exceeding 10³ and product titre up to 6 mM (Scheme 4a) (38,39).

Scheme 4.

Other hydrogen-borrowing cascades: functionalization of cycloalkanes (a), synthesis of L-lactate (b) and parallel kinetic asymmetric transformation (c).

The conversion of ethanol to l-lactate (Scheme 4b) represents an example of a hydrogen-borrowing cascade with an additional intermediate reaction (40). An ADH oxidises ethanol (100 μM, continuous feed) to acetaldehyde at the expense of NAD⁺. Then, a pyruvate (de)carboxylase (PDC) catalyses the carboxylation of acetaldehyde to pyruvate. Finally, a lactate dehydrogenase performs the reduction of pyruvate, which also regenerates NAD⁺. As a proof of principle, 87 μM l-lactate were obtained within 95 h. An inherent requirement for a hydrogen-borrowing process is the perfect atom-economy. Biocatalytic redox reactions involving systems for cofactor recycling (e.g. glucose/GDH, formate/FDH, 2-propanol) transfer a hydride from a sacrificial cosubstrate to the substrate or viceversa, and generate a coproduct. As the coproduct is actually waste, these cascades are not (strictly speaking) hydrogen-borrowing. A notable exception is the parallel interconnected asymmetric transformation (PIKAT). In this case, an oxidation and a reduction are running in tandem consuming different substrates and producing different products (i.e. both valuable) with internal recycling of NAD(P). Examples are the resolution of racemic alcohols combined with: i) the asymmetric synthesis of β-haloalcohols catalysed by ADH (Scheme 4c);(41) ii) the asymmetric biocatalytic sulfoxidation or the Baeyer-Villiger oxidation catalysed by a BVMO; (42,43); iii) the asymmetric reduction of α-diketones (44).

Conclusion

Future research in this field foresees the integration of hydrogen-borrowing cascades into extended artificial biocatalytic pathways both in vitro and in vivo. An early outstanding example is the “modular” combination of the hydrogen-borrowing cascades depicted in Scheme 2a and 3a with an intermediate hydrolytic step for the redox self-sufficient conversion of cyclohexanol to nylon-6 monomer at 50 mM scale (18). Another example is the synthesis of substituted caprolactones from variously substituted cyclohex-2-en-1-ols (45,46). Extensive research in protein engineering of oxidoreductases will be fundamental for a successful implementation of biocatalytic hydrogen-borrowing cascades in organic synthesis at laboratory as well as industrial scale (47–49). In particular, we anticipate the need for a tool-box of oxidoreductases with a broad substrate scope and increased stability at elevated substrate concentrations and in presence of cosolvents. Enzyme immobilisation will also contribute to achieve higher turnovers via recyclability of the biocatalysts (50,51). Furthermore, existing methodologies for enzyme compartmentalisation will be valuable for circumventing possible enzymatic cross-activity in complex artificial biocatalytic pathways (52). Another viable option in this field is the implementation of flow chemistry, which will permit to physically separate different hydrogen-borrowing modules (53). Finally, the possibility to replace efficiently NAD(P) coenzymes with man-made biomimetics might contribute to the overall economic profitability of the hydrogen-borrowing cascades (54).