A photo-enzymatic cascade to transform racemic alcohols into enantiomerically pure amines

: The consecutive photooxidation and reductive amination of various alcohols in a cascade reaction were realized by the combination of a photocatalyst and several enzymes. Whereas the photocatalyst (sodium anthraquinone-2-sulfonate) mediated the light-driven, aerobic oxidation of primary and secondary alcohols, the enzymes (various ω -transaminases) catalyzed the enantio-speciﬁc reductive amination of the intermediate aldehydes and ketones. The system worked in a one-pot one-step fashion, whereas the productivity was signiﬁcantly improved by switching to a one-pot two-step procedure. A wide range of aliphatic and aromatic compounds was transformed into the enantiomerically pure corresponding amines via the photo-enzymatic cascade.


Introduction
In recent years, significant attention has been devoted to the synthesis of amines [1]. Especially, enantiomerically pure amines are of interest in natural product synthesis, as intermediates in pharmaceutical synthesis, and for a variety of other chemical products [2][3][4]. Direct amination of alcohols has been studied extensively [1,5]. The main advantage of prompt conversion of alcohols to amines is that both the substrate and the product are in the same oxidation state, and, thus, theoretically the additional use of any redox equivalents is not required [6,7]. Furthermore, many of the required alcohols are already available on an industrial scale, which facilitates the mass application of the technology [8]. On the other hand, most of the currently applied reactions either possess poor chemo-selectivity and require harsh reaction conditions (e.g., the reaction of alcohol with ammonia over various heterogeneous catalysts, such as tungsten, cobalt, nickel, chromium) [5] or have low efficiency and produce toxic intermediates [9]. It is also worth mentioning that the majority of the current methodologies either only accept symmetric substrates or produce racemic products. To synthesize chiral amine moieties, ketones are one of the most commonly used precursors [1].
Various approaches have been reported to selectively oxidize alcohols to the corresponding carbonyl products [10][11][12][13]. Compared to transition metal-based approaches [14], photooxidation provides an atom-economic and environmentally benign alternative [10]. Commercially available TiO 2 is used most commonly as a photocatalyst, but the applicability is restricted by its UV absorption [15]. However, efficient oxidation of 1-pentanol has been reported by irradiation up to 480 nm upon doping TiO 2 with Nb 2 O 5 [16,17]. To further utilize visible light, numerous other heterogeneous catalysts have been tested [18]. By using graphitic carbon nitride (g-C 3 N 4 ) [19,20] and vanadium-oxide grafted to numerous carriers [21,22], efficient alcohol oxidation mediated by visible light has been reported. Besides heterogeneous systems, water-soluble sodium anthraquinone-2-sulfonate (SAS) is also known as an effective and inexpensive homogeneous photocatalyst [23][24][25][26][27]. Very recently we found that SAS is also very active for the oxidation of C-H bonds [27] and, therefore, can be implemented in photo-enzymatic cascades for the conversion of a very large substrate scope [28].
Even though photocatalysis provides an attractive approach to produce the carbonyl intermediate, typically, tedious protection/de-protection steps are required for the synthesis of chiral amines. These additional stages prolong the synthesis time and significantly decrease the atom efficiency of the reaction. To aid these shortcomings, biocatalytic approaches have been extensively studied [29][30][31][32]. Pyridoxal-5-phosphate (PLP)-dependent ω-transaminases (ω-TAs) provide an elegant way to convert a carbonyl group to an amine moiety at the expense of a sacrificial amine donor [31,33]. ω-TAs have been increasingly applied in industrial chemical synthesis, in particular for the manufacturing of active pharmaceutical ingredients (APIs) [29,34,35]. Other main applications deal with the production of surfactants [36], amino acids [37], and plastic fiber monomers [36,38,39]. To reach an acceptable conversion when L-or D-alanine is used as an amine donor, the reaction equilibrium of the transamination reaction must be shifted towards the product. Thus, one of the most common means to enhance the amine formation is to remove the pyruvate by-product through the addition of another enzyme, such as lactate dehydrogenase (LDH), alanine dehydrogenase (AlaDH), pyruvate decarboxylase (PDC), or an acetolactate synthase (ALS) [32]. On the other hand, the use of alternative amine donors provides a more favorable thermodynamic equilibrium. However, these molecules are either prohibitively expensive [39] or generate a co-product during the transamination that tends to polymerize, thus complicating product isolation and lowering product yield [40,41]. Therefore, it is not surprising that isopropylamine (IPA) has been used as the preferred amine donor, especially on an industrial scale. IPA is inexpensive and provides a much more favorable thermodynamic equilibrium compared to alanine [42], which can be further shifted via simple and selective evaporation of the low boiling acetone co-product [35]. Consequently, IPA was selected as the amine donor in this study.
Recently, numerous reports have been published about the effective conversion of a plethora of alcohols to (chiral) amines in multi-enzyme cascades [43,44]. Inspired by their results and the current advances in the selective photochemical oxidation of alcohols [18], herein we present a photo-enzymatic tandem reaction for the direct conversion of alcohols to (chiral) amines. Employing the consecutive oxidation and reductive amination steps in one pot, a further reduction of the environmental impact of the synthesis is possible (Scheme 1). As time-consuming intermediate isolation and purification steps are omitted, the required amount of organic solvents is minimized, which results in a lower E-factor (Environmental factor, defined as the mass of waste per mass of product) [45,46].

Scheme 1.
Schematic representation of the photo-enzymatic cascade reaction and the examined substrates.

Results and Discussion
In the first set of experiments, we evaluated a range of homogeneous and heterogeneous photocatalysts available in our laboratory.
More specifically, water-soluble sodium anthraquinone-2-sulfonate (SAS) and heterogeneous graphitic carbon nitride (g-C 3 N 4 ), niobium(V)-oxide-doped titanium-dioxide (TiO 2 -Nb 2 O 5 ), and vanadium oxide (VO) deposited on various carriers, such as alumina (VO-Al 2 O 3 ), zirconium-dioxide (VO-ZrO 2 ), and graphitic carbon nitride (VO-g-C 3 N 4 ), were evaluated. As a model reaction, we chose the oxidation of rac-1-phenylethanol (5a) to acetophenone (5b) (Figure 1). Starting from 25 mM of the starting material, SAS mediated the full conversion of 5a (TN = 30), while the best heterogeneous catalyst, g-C 3 N 4 , produced only 6.2 mM of product with 20% conversion. The discrepancy between conversion and product concentration in the case of SAS is probably due to the oxidative decomposition of the product. Since the other photocatalysts fell back significantly in their performance (Figure 1), we focused our attention on SAS and g-C 3 N 4 . We performed the cascade reaction shown in Scheme 1 using SAS and g-C 3 N 4 in a one-pot one-step fashion, i.e., adding the starting material and all catalysts at the beginning ( Figure 2). For the reductive amination of the intermediate ketone, two ω-transaminases were applied: the (R)-selective ω-TA from Aspergillus terreus (ATωTA) and the (S)-selective ω-TA from Bacillus megaterium (BMωTA). In the case of g-C 3 N 4 , virtually no alcohol conversion was observed. The trace amounts of rac-5c originated from small acetophenone impurities within the commercial substrate. We attribute this lack of catalytic activity to the absorption of the biocatalysts to the g-C 3 N 4 surface, thereby passivating it for the desired oxidation reaction.
The cascade reaction using SAS, however, produced 2.5 mM (S)-and 1.5 mM (R)-1-phenylethyl amine in 48 h. Even though enantiomerically pure amines were obtained, the performance of this system fell back behind our expectations. Therefore, we set out to investigate the limitations of the one-pot one-step procedure. Further characterization of the reaction conditions revealed some limitations of the one-pot one-step procedure to be as follows: (i) the catalytic activity decreased steadily in the presence of light and SAS. This may be attributed to an oxidative inactivation and degradation of the biocatalysts by photoexcited SAS and the used light ( Figure S9); (ii) we found that the rate of the first step (SAS-mediated oxidation of phenylethanol) was significantly slower in the case of the one-pot one-step procedure. This may be attributed to SAS's activity on the stoichiometric amine donor and the product of the reductive amination step (Figures S11-S13). Therefore, we turned our attention to a one-pot two-step procedure, wherein we first performed the photochemical alcohol oxidation followed by supplementation of the reaction mixture containing the reaction components needed for the reductive amination in a second step. Compared to the one-pot one-step case, the rate of the oxidation increased meaningfully, and the amount of product 5c doubled with excellent enantiomeric excess (>99%) (Figure 3). To gain more insight into the performance of the ω-transaminases, we varied the enzyme, amine donor, and cofactor concentrations. An increase in the enzyme concentration resulted in a higher initial reaction rate. However, approx. 2.9 mM of product was formed for all cases after 28 h using 1 M of IPA at pH 9 ( Figure S23). Using the same concentration of IPA but increasing the pH to 10.2 led to a progressive increase in the yield above 6 mM ( Figure S26). These data are in agreement with previous publications, which indicate that the deprotonated form of IPA in aqueous solution might act as the actual amine donor [35,47,48]. Testing different concentrations of IPA at pH 9 (0.5, 1, 2 M) resulted in both higher initial rate and final product concentration ( Figure S24), as also the amount of free IPA amino donor in solution increased accordingly. We concluded that the final obtained conversion was determined by the actual concentration of free IPA in solution and the thermodynamic equilibrium constant of the reaction [42], and not by the stability of the biocatalyst. Therefore, the productivity of our system can be increased in the future by process engineering aimed at the selective removal of the acetone co-product. Finally, varying the concentration of the cofactor PLP did not show any noticeable influence on the product formation ( Figure S25), which is also in agreement with previous findings. Conversely, an excessively high concentration of PLP is not only inconvenient economically but can also inhibit the ω-TA [49]. Based on the above findings, we chose the following parameters for our next investigation: [with ω-TA enzyme] = 10 mg/mL, [IPA] = 1 M, and [PLP] = 1 mM. The reactions were performed at 30 • C, an optimal temperature for both the photooxidation and reductive amination steps.
To further explore the scope of the one-pot two-step procedure, we investigated a broader scope of starting materials including aliphatic, aromatic, chiral, and non-chiral substrates (Scheme 1 and Table 1).  Aromatic substrates were oxidized with good (>80%) to excellent (>95%) conversion. However, by increasing the length of the aliphatic side chain from 1-phenylethanol (5a) to 1-phenylbutanol (7a), the yield decreased significantly. This is likely the result of oxidative decomposition, which is enhanced by the inductive effect of the aliphatic side chain [47]. This is also supported by the results for benzyl alcohol (9a) and 3-chlorobenzyl alcohol (10a). With the presence of an electron withdrawing group on the aromatic ring, which slightly reduces the activity of the hydroxyl group, the selectivity increased. It is also worth mentioning that besides oxidation, 1,2,3,4-tetrahydro-naphtol (8a) reacted with SAS, which was indicated by the absence of the characteristic yellow color of the photocatalyst after irradiation. In contrast to the aromatic substrates, aliphatic compounds demonstrated moderate conversion (>30%). Interestingly, C5 compounds (1a, 2a) showed better yield than C6 compounds (3a, 4a).
Amongst the prochiral compounds (Table 2), the entire range of substrates was transformed with excellent stereoselectivity (>99% ee). Acetophenone (5b) was accepted by all enzymes. In the case of aromatic compounds (5b-7b), the conversion decreased progressively with the increasing length of the alkyl chain attached to the aromatic ring. However, VFωTA was an exception, as it similarly mediated the reductive amination of both propiophenone (6b) and butyrophenone (7b) with 35% conversion. Reductive amination of 8b was also feasible using BMωTA with excellent stereospecificity, which has been challenging for other systems [47,50]. Amongst the aliphatic substrates, 2-pentanone (2b) and 2-hexanal (4b) were converted to detectable levels of product only by ATωTA and VFωTA, respectively. However, the modest performances of other enzymes are likely to be the result of the mediocre initial concentration of the substrate, as both compounds have been reported to be efficiently converted to the corresponding chiral amine [47]. In the case of aldehydes, excellent conversions were achieved (Table 3). Besides VFωTA, all the enzymes completely converted hexanal (3b) to hexylamine (3c). However, it is also important to mention that similar to the other aliphatic compounds, the initial substrate concentration was low. Both benzaldehyde (9b) and 3-chlorobenzaldehyde (10b) were fully converted to the corresponding amine.

Materials and Methods
All chemicals were purchased Sigma-Aldrich (Zwijndrecht, The Netherlands), Fluka (Buchs, Switzerland), Acros (Geel, Belgium) or Alfa-Aesar (Karlsruhe, Germany) with the highest purity available and used without further treatment. The ω-transaminases were prepared via recombinant expression in Escherichia coli as described in detail in the Supplementary Materials.

Reaction Conditions for the One-Pot Two-Step Cascade
If not stated otherwise, in a 4 mL vial, 2 mL MilliQ water containing approx. 10 mM substrate and 0.75 mM SAS were irradiated with visible light (Osram Halolux CERAM 205W light bulb; λ > 400 nm) at 30 • C. The reaction mixture was gently stirred with a magnetic stirrer. At intervals, aliquots were taken, extracted with ethyl acetate, and analyzed by gas chromatography without derivatization (see Supplementary Materials). After the irradiation, 250 µL reaction mixture was diluted to 500 µL by using NaPi (100 mM) as such that the final samples contained NaPI (50 mM), IPA (1M), PLP (1 mM), and 10 mg/mL lyophilized whole cells overexpressed with (R)-or (S)-selective ω-TA. The pH of the system was adjusted to pH 9. Samples were incubated on a shaking plate at 30 • C with 600 rpm shaking speed. At intervals, aliquots were taken, extracted with ethyl acetate, derivatized, and analyzed by gas chromatography (see Supplementary Materials).

Conclusions
In summary, a reaction sequence of consecutive photooxidation and reductive amination of various alcohols in one pot is reported. The tandem reaction with sodium anthraquinone-2-sulfonate (SAS) as a photocatalyst was also feasible in a one-pot one-step approach; however, the yield was significantly inferior to the one-pot two-step case. The system was tested on a plethora of aromatic and aliphatic substrates. Even though all of the compounds were converted, the photooxidation of aromatic substrates proceeded much faster and with a higher yield. In the second step of the cascade reaction, recombinant transaminases originating from Aspergillus terreus (ATωTA), Bacillus megaterium (BMωTA), Chromobacterium violaceum (CVωTA), Pseudomonas fluorescens (PFωTA), and Vibrio fluvialis (VFωTA) were used for the reductive amination. All of the non-chiral substrates were converted with high conversion and the prochiral substrates with excellent enantiomeric excess.
One issue en route to the preparative application is the inactivation of the biocatalysts by the photoexcited photocatalysts. This may be overcome by special separation of the photocatalytic oxidation step from the reductive amination step. For this, for example, immobilized catalysts in a flow chemistry setup may be a doable approach.