Enantioselective Epoxidation by Flavoprotein Monooxygenases Supported by Organic Solvents

Styrene and indole monooxygenases (SMO and IMO) are two-component flavoprotein monooxygenases composed of a nicotinamide adenine dinucleotide (NADH)-dependent flavin adenine dinucleotide (FAD)-reductase (StyB or IndB) and a monooxygenase (StyA or IndA). The latter uses reduced FAD to activate oxygen and to oxygenate the substrate while releasing water. We circumvented the need for the reductase by direct FAD reduction in solution using the NAD(P)H-mimic 1-benzyl-1,4-dihydronicotinamide (BNAH) to fuel monooxygenases without NADH requirement. Herein, we report on the hitherto unknown solvent tolerance for the indole monooxygenase from Gemmobacter nectariphilus DSM15620 (GnIndA) and the styrene monooxygenase from Gordonia rubripertincta CWB2 (GrStyA). These enzymes were shown to convert bulky and rather hydrophobic styrene derivatives in the presence of organic cosolvents. Subsequently, BNAH-driven biotransformation was furthermore optimized with regard to the applied cosolvent and its concentration as well as FAD and BNAH concentration. We herein demonstrate that GnIndA and GrStyA enable selective epoxidations of allylic double bonds (up to 217 mU mg−1) in the presence of organic solvents such as tetrahydrofuran, acetonitrile, or several alcohols. Notably, GnIndA was found to resist methanol concentrations up to 25 vol.%. Furthermore, a diverse substrate preference was determined for both enzymes, making their distinct use very interesting. In general, our results seem representative for many IMOs as was corroborated by in silico mutagenetic studies.


Introduction
Styrene and indole monooxygenases (StyA or IndA) are the initial enzymes in degradation pathways of human-made aromatic pollutants, such as styrenes, within several bacteria [1][2][3]. Similar to the general detoxification pathways in higher organisms, hydrophobic compounds are typically oxidized in bacteria to increase their water solubility and create a functional moiety, which can be easily metabolized by downstream enzymes [4,5]. For example, styrene monooxygenases perform the oxidation of double bonds with molecular oxygen and an additional reduction equivalent to form an epoxide along with water as side product [6,7].
Due to their role as mainly detoxifying enzymes, SMOs and IMOs accept a large variety of substrates. Although the highest activities were described for conjugated double bonds in direct proximity of an Nevertheless, there are some significant drawbacks that have to be overcome before SMOs or IMOs can be used on an even larger industrial scale. First, SMOs and IMOs depend on a suitable electron donor to reduce their flavin adenine dinucleotide (FAD) cofactor after each reaction cycle. In nature, nicotinamide adenine dinucleotide (NADH) is consumed stoichiometrically and regenerated in vivo, but it is too expensive to use with isolated enzymes on larger scales.
Among the enzyme family, SMOs and IMOs can be differentiated into mono-and twocomponent systems [7,21]. The first contain a NADH binding domain which allows them to reduce their FAD cofactor autonomously but limits them exclusively to their natural cosubstrate. In contrast, two-component oxygenases require an additional reductase (StyB or IndB) which supplies the oxygenase with reduced FAD in nature. Furthermore, the self-sufficient mono-component enzymes of these monooxygenases show a lower oxygenase activity than the others [12,22,23].
Notably, in artificial systems, two-component enzymes can also use FADH2 from the solution [24]. Thus, cofactor reduction by chemical methods without the need for NADH as an expensive cosubstrate is possible. Commonly, two-component SMOs and IMOs are cost-efficiently fueled by (a) the reduction of FAD by chemical agents (e.g., sodium thiosulfate or ruthenium complexes [8,25]), (b) the use of NAD(P)H-mimics [26] or (c) the electrochemical reduction of the cofactor [27].
In this study, the NAD(P)H-mimic 1-benzyl-1,4-dihydronicotinamide (BNAH) was preferred as cheap and easy to synthesize electron donor for the FAD cofactor in solution [11].
In addition to the typical need to apply expensive reductants, the limited solubility of many organic substrates in aqueous solutions (e.g., 2.9 mM for styrene) is a key problem and restricts substrate conversion to the lower millimolar range using isolated biocatalysts [28]. For higher amounts, whole cells have to be used in two-phase systems or with cosolvents [29]. In particular, styrene derivatives with hydrophobic side chains have a comparably low water solubility [30]. In Nevertheless, there are some significant drawbacks that have to be overcome before SMOs or IMOs can be used on an even larger industrial scale. First, SMOs and IMOs depend on a suitable electron donor to reduce their flavin adenine dinucleotide (FAD) cofactor after each reaction cycle. In nature, nicotinamide adenine dinucleotide (NADH) is consumed stoichiometrically and regenerated in vivo, but it is too expensive to use with isolated enzymes on larger scales.
Among the enzyme family, SMOs and IMOs can be differentiated into mono-and two-component systems [7,21]. The first contain a NADH binding domain which allows them to reduce their FAD cofactor autonomously but limits them exclusively to their natural cosubstrate. In contrast, two-component oxygenases require an additional reductase (StyB or IndB) which supplies the oxygenase with reduced FAD in nature. Furthermore, the self-sufficient mono-component enzymes of these monooxygenases show a lower oxygenase activity than the others [12,22,23].
Notably, in artificial systems, two-component enzymes can also use FADH 2 from the solution [24]. Thus, cofactor reduction by chemical methods without the need for NADH as an expensive cosubstrate is possible. Commonly, two-component SMOs and IMOs are cost-efficiently fueled by (a) the reduction of FAD by chemical agents (e.g., sodium thiosulfate or ruthenium complexes [8,25]), (b) the use of NAD(P)H-mimics [26] or (c) the electrochemical reduction of the cofactor [27].
In this study, the NAD(P)H-mimic 1-benzyl-1,4-dihydronicotinamide (BNAH) was preferred as cheap and easy to synthesize electron donor for the FAD cofactor in solution [11].
In addition to the typical need to apply expensive reductants, the limited solubility of many organic substrates in aqueous solutions (e.g., 2.9 mM for styrene) is a key problem and restricts substrate conversion to the lower millimolar range using isolated biocatalysts [28]. For higher amounts, whole cells have to be used in two-phase systems or with cosolvents [29]. In particular, styrene derivatives with hydrophobic side chains have a comparably low water solubility [30]. In addition, only a few enzymes with uncommonly capacious substrate binding sites are known to accept those large styrene derivatives [11]. Nevertheless, chiral styrene oxide derivatives with hydrophobic side chains are important industrial precursors for several drugs (Figure 1). In this study, we maximized the production of these compounds in an overall process optimization by applying various water-miscible solvents to increase the amount of substrate in solution and make it accessible for the enzymes. Our results show remarkably high product formation in reactions catalyzed by the IMO designated as GnIndA from Gemmobacter nectariphilus DSM15620 in presence of organic solvents.
To the best of our knowledge, solvent-tolerant SMOs or IMOs have not been reported before. Herein, we describe the systematical reaction optimization toward several water-miscible cosolvents for the probable solvent-tolerant GnIndA and GrStyA from Gordonia rubripertincta CWB2. In addition, our work highlights criteria which will help to identify solvent-tolerant enzymes based on their sequence and allows creating more robust mutants.

Reaction Condition Optimization
GrStyA and GnIndA were successfully produced by recombinant expression as described previously [11]. Both enzymes were obtained in pure form after His-tag purification using an N-terminal His 10 -tag and yielded 15 mg L −1 for GrStyA and 50 mg L −1 for GnIndA ( Figure S1). Since both enzymes were described to accept styrene derivatives with large side chains as substrate [11], we aimed to maximize the product amount in a BNAH-driven process.
Because GnIndA and GrStyA belong to the subgroup of two-component systems, they were expected to accept NAD(P)H-mimics and their activity was established with BNAH as reduction equivalent (standard conditions: 20 mM Tris-HCl pH 7.5, 10 mM BNAH, 50 µM FAD, 1 mM dithiothreitol, 5 vol.% glycerol, 20 U/mL catalase and 10 vol.% MeOH for GnIndA or 5 vol.% DMSO for GrStyA). Dithiothreitol was applied to support enzyme stability as reported earlier [22]. For the sake of comparability of in-house performed assays, the protocol was kept unchanged, even though adding dithiothreitol was not mandatory for herein tested enzymes (data not shown). For both enzymes, the optimal BNAH concentration was determined to be 15 mM while for GrStyA a comparable high product amount was also observed at a concentration of 10 mM BNAH. These concentrations are equal to a 200-to 300-fold BNAH excess toward FAD (Figure 2a). For GnIndA, lower product amounts with 10 mM BNAH were observed in contrast to otherwise similar values in comparison with GrStyA.
Catalysts 2020, 10, x FOR PEER REVIEW 3 of 14 addition, only a few enzymes with uncommonly capacious substrate binding sites are known to accept those large styrene derivatives [11]. Nevertheless, chiral styrene oxide derivatives with hydrophobic side chains are important industrial precursors for several drugs (Figure 1). In this study, we maximized the production of these compounds in an overall process optimization by applying various water-miscible solvents to increase the amount of substrate in solution and make it accessible for the enzymes. Our results show remarkably high product formation in reactions catalyzed by the IMO designated as GnIndA from Gemmobacter nectariphilus DSM15620 in presence of organic solvents.
To the best of our knowledge, solvent-tolerant SMOs or IMOs have not been reported before. Herein, we describe the systematical reaction optimization toward several water-miscible cosolvents for the probable solvent-tolerant GnIndA and GrStyA from Gordonia rubripertincta CWB2. In addition, our work highlights criteria which will help to identify solvent-tolerant enzymes based on their sequence and allows creating more robust mutants.

Reaction Condition Optimization
GrStyA and GnIndA were successfully produced by recombinant expression as described previously [11]. Both enzymes were obtained in pure form after His-tag purification using an Nterminal His10-tag and yielded 15 mg L −1 for GrStyA and 50 mg L −1 for GnIndA ( Figure S1). Since both enzymes were described to accept styrene derivatives with large side chains as substrate [11], we aimed to maximize the product amount in a BNAH-driven process.
Because GnIndA and GrStyA belong to the subgroup of two-component systems, they were expected to accept NAD(P)H-mimics and their activity was established with BNAH as reduction equivalent (standard conditions: 20 mM Tris-HCl pH 7.5, 10 mM BNAH, 50 µM FAD, 1 mM dithiothreitol, 5 vol.% glycerol, 20 U/mL catalase and 10 vol.% MeOH for GnIndA or 5 vol.% DMSO for GrStyA). Dithiothreitol was applied to support enzyme stability as reported earlier [22]. For the sake of comparability of in-house performed assays, the protocol was kept unchanged, even though adding dithiothreitol was not mandatory for herein tested enzymes (data not shown). For both enzymes, the optimal BNAH concentration was determined to be 15 mM while for GrStyA a comparable high product amount was also observed at a concentration of 10 mM BNAH. These concentrations are equal to a 200-to 300-fold BNAH excess toward FAD (Figure 2a). For GnIndA, lower product amounts with 10 mM BNAH were observed in contrast to otherwise similar values in comparison with GrStyA.  In addition, the FAD saturation concentration was found to be higher than for GrStyA (Figure 2b). GrStyA reached >80% of the maximal produced styrene oxide supplied with 25 µM FAD and no further increase of product amount was observed for cofactor concentrations above 40 µM (Figure 2b). In contrast, GnIndA required more than 60 µM FAD to reach >90% of the maximal substrate production and reached the maximal product formation at a FAD concentration of 100 µM.
These findings indicate a lower affinity of GnIndA toward FADH 2 . The BNAH-driven FAD reduction in solution is enzyme-independent, but the uptake of reduced FADH 2 correlates with the affinity of the enzyme toward the cofactor. Since the effect of a lower affinity is only detectable under limiting conditions, different product formation rates can only be observed in low FAD concentration ranges as it was demonstrated for GnIndA ( Figure 2).
After the optimization of cofactor and cosubstrate concentrations, we tested the influence of organic solvents on the enzymatic reaction. The cosolvent is required to increase the solubility of the anticipated substrates in this study and to make those styrene derivatives with hydrophobic side chains accessible for the biocatalyst. In addition, BNAH is also hardly soluble in aqueous solutions and likewise requires cosolvents. In total, nine water-miscible solvents with logP O/W values ranging from −1.35 (dimethyl sulfoxide) to +0.46 (tetrahydrofuran) were applied to the enzymatic reaction in concentrations up to 25 vol.% ( Figure 3, Table S1). In this concentration range, the dissolved amount of the most hydrophobic substrate 1-phenyl-1-cyclohexene could be increased between a factor of three (methanol) and 100 times (1-propanol). This increase marks the upper limit for effects achieved by the cosolvent regarding the substrate solubility ( Figure S2). Since styrene is about 1.6 times better soluble in water by the means of the logP O/W value (Table 1), a less pronounced effect in solubility increase can be expected by solvent addition for the more polar styrene. Nevertheless, the unsubstituted styrene was chosen as standard substrate to quantify the effects of each solvent and its concentration on the overall enzyme reaction. As measure of the effect, the produced amount of styrene oxide after 2 h of reaction time was used. In addition, the FAD saturation concentration was found to be higher than for GrStyA ( Figure  2b). GrStyA reached >80% of the maximal produced styrene oxide supplied with 25 µM FAD and no further increase of product amount was observed for cofactor concentrations above 40 µM ( Figure  2b). In contrast, GnIndA required more than 60 µM FAD to reach >90% of the maximal substrate production and reached the maximal product formation at a FAD concentration of 100 µM.
These findings indicate a lower affinity of GnIndA toward FADH2. The BNAH-driven FAD reduction in solution is enzyme-independent, but the uptake of reduced FADH2 correlates with the affinity of the enzyme toward the cofactor. Since the effect of a lower affinity is only detectable under limiting conditions, different product formation rates can only be observed in low FAD concentration ranges as it was demonstrated for GnIndA ( Figure 2).
After the optimization of cofactor and cosubstrate concentrations, we tested the influence of organic solvents on the enzymatic reaction. The cosolvent is required to increase the solubility of the anticipated substrates in this study and to make those styrene derivatives with hydrophobic side chains accessible for the biocatalyst. In addition, BNAH is also hardly soluble in aqueous solutions and likewise requires cosolvents. In total, nine water-miscible solvents with logPO/W values ranging from -1.35 (dimethyl sulfoxide) to +0.46 (tetrahydrofuran) were applied to the enzymatic reaction in concentrations up to 25 vol.% ( Figure 3, Table S1). In this concentration range, the dissolved amount of the most hydrophobic substrate 1-phenyl-1-cyclohexene could be increased between a factor of three (methanol) and 100 times (1-propanol). This increase marks the upper limit for effects achieved by the cosolvent regarding the substrate solubility ( Figure S2). Since styrene is about 1.6 times better soluble in water by the means of the logPO/W value (Table 1), a less pronounced effect in solubility increase can be expected by solvent addition for the more polar styrene. Nevertheless, the unsubstituted styrene was chosen as standard substrate to quantify the effects of each solvent and its concentration on the overall enzyme reaction. As measure of the effect, the produced amount of styrene oxide after 2 h of reaction time was used. Relative styrene oxide amount formed by GnIndA (left) and GrStyA (right) in presence of nine organic solvents: In general, for GnIndA a remarkably high product formation up to 15 vol.% cosolvent was observed while MeOH and THF were outstanding due to the stable product formation at higher concentrations and the increased product formation in lower concentration ranges, respectively. Except for DMSO and EtOH, a significantly reduced product amount was detected for GrStyA already at 10 vol.% cosolvent. The relative styrene oxide amount is normalized to 5 vol.% MeOH (GnIndA, 0.70 mM styrene oxide) resp. 4 vol.% MeOH (GrStyA, 0.47 mM styrene oxide). * not measured. Relative styrene oxide amount formed by GnIndA (left) and GrStyA (right) in presence of nine organic solvents: In general, for GnIndA a remarkably high product formation up to 15 vol.% cosolvent was observed while MeOH and THF were outstanding due to the stable product formation at higher concentrations and the increased product formation in lower concentration ranges, respectively. Except for DMSO and EtOH, a significantly reduced product amount was detected for GrStyA already at 10 vol.% cosolvent. The relative styrene oxide amount is normalized to 5 vol.% MeOH (GnIndA, 0.70 mM styrene oxide) resp. 4 vol.% MeOH (GrStyA, 0.47 mM styrene oxide). * not measured.

GrStyA GnIndA
Catalysts 2020, 10, x FOR PEER REVIEW 6 of 14 in the relative activity toward E-and Z-β-methylstyrene, being more active in the conversion of the Z-substrate.  For GrStyA, Z-β-methylstyrene, β,β-dimethylstyrene and 1,2-dihydronaphtalene were found to show the same or higher enzyme activity than toward styrene. Interestingly, the same compounds that contain the previously mentioned residue in Z-position cause a decreased epoxidation activity in GnIndA. In addition, for GrStyA a reduced activity for E-β-methylstyrene was observed which indicates an overall Z-selectivity. Furthermore, GrStyA converts 1,2-dihydronaphtalene at a comparable rate than styrene while 1-phenyl-1-cyclohexene is converted significantly slower (7 ± 1%). Thus, GrStyA was found to be Z-selective.
In general, GrStyA catalyzed the epoxidation of all tested substrates and overall, a lesser reduction in activity was observed for most of the substrates. Nevertheless, the actual activity of GnIndA is still higher than of GrStyA for most tested substrates.
For substrates with ether residues at the aromatic ring, a substitution-specific decrease in activity was observed for both enzymes while the relative product formation was slightly higher for GrStyA. Interestingly, the position of the ether moiety correlates similarly with the activity decrease: A methoxy group in meta-position was accepted best by both enzymes while the same substituent in para-position caused the lowest conversion rate, if converted at all, for both enzymes. These positionspecific correlations in enzymatic activity indicate a similar substrate binding at the aromatic ring Catalysts 2020, 10, x FOR PEER REVIEW 6 of 14 in the relative activity toward E-and Z-β-methylstyrene, being more active in the conversion of the Z-substrate.  For GrStyA, Z-β-methylstyrene, β,β-dimethylstyrene and 1,2-dihydronaphtalene were found to show the same or higher enzyme activity than toward styrene. Interestingly, the same compounds that contain the previously mentioned residue in Z-position cause a decreased epoxidation activity in GnIndA. In addition, for GrStyA a reduced activity for E-β-methylstyrene was observed which indicates an overall Z-selectivity. Furthermore, GrStyA converts 1,2-dihydronaphtalene at a comparable rate than styrene while 1-phenyl-1-cyclohexene is converted significantly slower (7 ± 1%). Thus, GrStyA was found to be Z-selective.
In general, GrStyA catalyzed the epoxidation of all tested substrates and overall, a lesser reduction in activity was observed for most of the substrates. Nevertheless, the actual activity of GnIndA is still higher than of GrStyA for most tested substrates.
For substrates with ether residues at the aromatic ring, a substitution-specific decrease in activity was observed for both enzymes while the relative product formation was slightly higher for GrStyA. Interestingly, the position of the ether moiety correlates similarly with the activity decrease: A methoxy group in meta-position was accepted best by both enzymes while the same substituent in para-position caused the lowest conversion rate, if converted at all, for both enzymes. These positionspecific correlations in enzymatic activity indicate a similar substrate binding at the aromatic ring Catalysts 2020, 10, x FOR PEER REVIEW 6 of 14 in the relative activity toward E-and Z-β-methylstyrene, being more active in the conversion of the Z-substrate.  For GrStyA, Z-β-methylstyrene, β,β-dimethylstyrene and 1,2-dihydronaphtalene were found to show the same or higher enzyme activity than toward styrene. Interestingly, the same compounds that contain the previously mentioned residue in Z-position cause a decreased epoxidation activity in GnIndA. In addition, for GrStyA a reduced activity for E-β-methylstyrene was observed which indicates an overall Z-selectivity. Furthermore, GrStyA converts 1,2-dihydronaphtalene at a comparable rate than styrene while 1-phenyl-1-cyclohexene is converted significantly slower (7 ± 1%). Thus, GrStyA was found to be Z-selective.
In general, GrStyA catalyzed the epoxidation of all tested substrates and overall, a lesser reduction in activity was observed for most of the substrates. Nevertheless, the actual activity of GnIndA is still higher than of GrStyA for most tested substrates.
For substrates with ether residues at the aromatic ring, a substitution-specific decrease in activity was observed for both enzymes while the relative product formation was slightly higher for GrStyA. Interestingly, the position of the ether moiety correlates similarly with the activity decrease: A methoxy group in meta-position was accepted best by both enzymes while the same substituent in para-position caused the lowest conversion rate, if converted at all, for both enzymes. These positionspecific correlations in enzymatic activity indicate a similar substrate binding at the aromatic ring Catalysts 2020, 10, x FOR PEER REVIEW 6 of 14 in the relative activity toward E-and Z-β-methylstyrene, being more active in the conversion of the Z-substrate.  For GrStyA, Z-β-methylstyrene, β,β-dimethylstyrene and 1,2-dihydronaphtalene were found to show the same or higher enzyme activity than toward styrene. Interestingly, the same compounds that contain the previously mentioned residue in Z-position cause a decreased epoxidation activity in GnIndA. In addition, for GrStyA a reduced activity for E-β-methylstyrene was observed which indicates an overall Z-selectivity. Furthermore, GrStyA converts 1,2-dihydronaphtalene at a comparable rate than styrene while 1-phenyl-1-cyclohexene is converted significantly slower (7 ± 1%). Thus, GrStyA was found to be Z-selective.
In general, GrStyA catalyzed the epoxidation of all tested substrates and overall, a lesser reduction in activity was observed for most of the substrates. Nevertheless, the actual activity of GnIndA is still higher than of GrStyA for most tested substrates.
For substrates with ether residues at the aromatic ring, a substitution-specific decrease in activity was observed for both enzymes while the relative product formation was slightly higher for GrStyA. Interestingly, the position of the ether moiety correlates similarly with the activity decrease: A methoxy group in meta-position was accepted best by both enzymes while the same substituent in para-position caused the lowest conversion rate, if converted at all, for both enzymes. These positionspecific correlations in enzymatic activity indicate a similar substrate binding at the aromatic ring Catalysts 2020, 10, x FOR PEER REVIEW 6 of 14 in the relative activity toward E-and Z-β-methylstyrene, being more active in the conversion of the Z-substrate.  For GrStyA, Z-β-methylstyrene, β,β-dimethylstyrene and 1,2-dihydronaphtalene were found to show the same or higher enzyme activity than toward styrene. Interestingly, the same compounds that contain the previously mentioned residue in Z-position cause a decreased epoxidation activity in GnIndA. In addition, for GrStyA a reduced activity for E-β-methylstyrene was observed which indicates an overall Z-selectivity. Furthermore, GrStyA converts 1,2-dihydronaphtalene at a comparable rate than styrene while 1-phenyl-1-cyclohexene is converted significantly slower (7 ± 1%). Thus, GrStyA was found to be Z-selective.
In general, GrStyA catalyzed the epoxidation of all tested substrates and overall, a lesser reduction in activity was observed for most of the substrates. Nevertheless, the actual activity of GnIndA is still higher than of GrStyA for most tested substrates.
For substrates with ether residues at the aromatic ring, a substitution-specific decrease in activity was observed for both enzymes while the relative product formation was slightly higher for GrStyA. Interestingly, the position of the ether moiety correlates similarly with the activity decrease: A methoxy group in meta-position was accepted best by both enzymes while the same substituent in para-position caused the lowest conversion rate, if converted at all, for both enzymes. These positionspecific correlations in enzymatic activity indicate a similar substrate binding at the aromatic ring Catalysts 2020, 10, x FOR PEER REVIEW 6 of 14 in the relative activity toward E-and Z-β-methylstyrene, being more active in the conversion of the Z-substrate.  For GrStyA, Z-β-methylstyrene, β,β-dimethylstyrene and 1,2-dihydronaphtalene were found to show the same or higher enzyme activity than toward styrene. Interestingly, the same compounds that contain the previously mentioned residue in Z-position cause a decreased epoxidation activity in GnIndA. In addition, for GrStyA a reduced activity for E-β-methylstyrene was observed which indicates an overall Z-selectivity. Furthermore, GrStyA converts 1,2-dihydronaphtalene at a comparable rate than styrene while 1-phenyl-1-cyclohexene is converted significantly slower (7 ± 1%). Thus, GrStyA was found to be Z-selective.
In general, GrStyA catalyzed the epoxidation of all tested substrates and overall, a lesser reduction in activity was observed for most of the substrates. Nevertheless, the actual activity of GnIndA is still higher than of GrStyA for most tested substrates.
For substrates with ether residues at the aromatic ring, a substitution-specific decrease in activity was observed for both enzymes while the relative product formation was slightly higher for GrStyA. Interestingly, the position of the ether moiety correlates similarly with the activity decrease: A methoxy group in meta-position was accepted best by both enzymes while the same substituent in para-position caused the lowest conversion rate, if converted at all, for both enzymes. These positionspecific correlations in enzymatic activity indicate a similar substrate binding at the aromatic ring Catalysts 2020, 10, x FOR PEER REVIEW 6 of 14 in the relative activity toward E-and Z-β-methylstyrene, being more active in the conversion of the Z-substrate.  For GrStyA, Z-β-methylstyrene, β,β-dimethylstyrene and 1,2-dihydronaphtalene were found to show the same or higher enzyme activity than toward styrene. Interestingly, the same compounds that contain the previously mentioned residue in Z-position cause a decreased epoxidation activity in GnIndA. In addition, for GrStyA a reduced activity for E-β-methylstyrene was observed which indicates an overall Z-selectivity. Furthermore, GrStyA converts 1,2-dihydronaphtalene at a comparable rate than styrene while 1-phenyl-1-cyclohexene is converted significantly slower (7 ± 1%). Thus, GrStyA was found to be Z-selective.
In general, GrStyA catalyzed the epoxidation of all tested substrates and overall, a lesser reduction in activity was observed for most of the substrates. Nevertheless, the actual activity of GnIndA is still higher than of GrStyA for most tested substrates.
For substrates with ether residues at the aromatic ring, a substitution-specific decrease in activity was observed for both enzymes while the relative product formation was slightly higher for GrStyA. Interestingly, the position of the ether moiety correlates similarly with the activity decrease: A methoxy group in meta-position was accepted best by both enzymes while the same substituent in para-position caused the lowest conversion rate, if converted at all, for both enzymes. These positionspecific correlations in enzymatic activity indicate a similar substrate binding at the aromatic ring Catalysts 2020, 10, x FOR PEER REVIEW 6 of 14 in the relative activity toward E-and Z-β-methylstyrene, being more active in the conversion of the Z-substrate.  For GrStyA, Z-β-methylstyrene, β,β-dimethylstyrene and 1,2-dihydronaphtalene were found to show the same or higher enzyme activity than toward styrene. Interestingly, the same compounds that contain the previously mentioned residue in Z-position cause a decreased epoxidation activity in GnIndA. In addition, for GrStyA a reduced activity for E-β-methylstyrene was observed which indicates an overall Z-selectivity. Furthermore, GrStyA converts 1,2-dihydronaphtalene at a comparable rate than styrene while 1-phenyl-1-cyclohexene is converted significantly slower (7 ± 1%). Thus, GrStyA was found to be Z-selective.
In general, GrStyA catalyzed the epoxidation of all tested substrates and overall, a lesser reduction in activity was observed for most of the substrates. Nevertheless, the actual activity of GnIndA is still higher than of GrStyA for most tested substrates.
For substrates with ether residues at the aromatic ring, a substitution-specific decrease in activity was observed for both enzymes while the relative product formation was slightly higher for GrStyA. Interestingly, the position of the ether moiety correlates similarly with the activity decrease: A methoxy group in meta-position was accepted best by both enzymes while the same substituent in para-position caused the lowest conversion rate, if converted at all, for both enzymes. These positionspecific correlations in enzymatic activity indicate a similar substrate binding at the aromatic ring Catalysts 2020, 10, x FOR PEER REVIEW 6 of 14 in the relative activity toward E-and Z-β-methylstyrene, being more active in the conversion of the Z-substrate.  For GrStyA, Z-β-methylstyrene, β,β-dimethylstyrene and 1,2-dihydronaphtalene were found to show the same or higher enzyme activity than toward styrene. Interestingly, the same compounds that contain the previously mentioned residue in Z-position cause a decreased epoxidation activity in GnIndA. In addition, for GrStyA a reduced activity for E-β-methylstyrene was observed which indicates an overall Z-selectivity. Furthermore, GrStyA converts 1,2-dihydronaphtalene at a comparable rate than styrene while 1-phenyl-1-cyclohexene is converted significantly slower (7 ± 1%). Thus, GrStyA was found to be Z-selective.
In general, GrStyA catalyzed the epoxidation of all tested substrates and overall, a lesser reduction in activity was observed for most of the substrates. Nevertheless, the actual activity of GnIndA is still higher than of GrStyA for most tested substrates.
For substrates with ether residues at the aromatic ring, a substitution-specific decrease in activity was observed for both enzymes while the relative product formation was slightly higher for GrStyA. Interestingly, the position of the ether moiety correlates similarly with the activity decrease: A methoxy group in meta-position was accepted best by both enzymes while the same substituent in para-position caused the lowest conversion rate, if converted at all, for both enzymes. These positionspecific correlations in enzymatic activity indicate a similar substrate binding at the aromatic ring 1,2-Dihydronaphtalene -100 ± 1 n.d. This simplification allows no differentiation between the contributions of either the enzyme stability, increased activity, or the substrate accessibility in solution; it rather shows the sum of all effects and is sufficient in order to develop an efficient enzymatic process in the first place.
The obtained styrene oxide amount was normalized to the amount produced in presence of 5 vol.% (GnIndA) or 4 vol.% methanol (GrStyA) (Figure 3). For the enzymatic reaction catalyzed by GnIndA, a remarkably high product formation after addition of various solvents in concentrations up to 15 vol.% was found, while methanol was outstanding with still 43% product production at 25 vol.% organic solvent in solution. In presence of up to 15 vol.% acetonitrile, an increased product amount was observed in comparison to 5 vol.% MeOH and other solvents at the respective concentrations. The highest increase in product formation of +72% relative to the result in 5 vol.% MeOH was found for 5 vol.% THF, which is the solvent with the most positive logP O/W value (+0.46). The detected styrene oxide amount for solvents with positive logP O/W values (THF, 1-and 2-propanol) is comparably high at 5 vol.% in general, but a significant decrease in product formation can be observed for higher concentrations. It has to be highlighted that the enantiomeric excess (ee) was investigated for selected solvents and no variation was observed under any investigated condition (GnIndA: >80% (S)-styrene oxide; Table S2).
In contrast to GnIndA, a decreased product formation in presence of most of the tested solvents was observed for the same reaction catalyzed by GrStyA. Only for DMSO and ethanol a product amount above 80% relative to standard conditions was detected at 10 vol.% cosolvent (Figure 3).

Substrate Spectrum
After optimization of the reaction conditions, we tested the enzymatic activity of both enzymes on eight substrates which contain aliphatic residues at the vinyl chain or methoxy modifications at the aromatic ring (Table 1). For all substrates, the water solubility is equal to or up to 1.6-times lower than for the unsubstituted styrene by means of the logP O/W value. In addition, all substrates exhibit a larger sterical size than styrene due to their substitution patterns. The substitutions were chosen to cover diverse positions within the molecule in order to compare the influence of variations at these positions on the enzymatic activity. Ether moieties were selected in paraand meta-position relative to and between the reactive double bond and the phenyl unit to investigate their influence on substrate recognition by the enzyme. The Eand Z-selectivity was investigated using a methyl group at the β-vinyl carbon in Eand Z-β-methylstyrene. In comparison to this small residue, a six-membered ring in 1-phenyl-1-cyclohexene and 1,2-dihydronaphtalene was selected to determine the E-and Z-selectivity as example for larger substituents.
Based on the findings described in the previous section, 10 vol.% methanol was applied as cosolvent for GnIndA in these experiments because it allowed the best combination of produced styrene oxide and substrate concentration in solution (Figure 3 and Figure S2). For GrStyA, a cosolvent concentration of 5 vol.% DMSO was chosen due to the high product formation in presence of DMSO. Furthermore, DMSO dissolved twice the substrate amount than ethanol at 5 vol.% respectively, for which also a stable product formation up to 10 vol.% was detected ( Figure S2). The remaining reaction conditions were applied as described in the previous section. For the determination of product formation rates, the respective activity on styrene under optimal conditions was used as 100% value.
Our results show a four-fold higher activity of GnIndA on the standard substrate styrene than observed for GrStyA. In return, for the less active GrStyA, an enantiomeric excess of >98% (S)-styrene oxide was detected while for GnIndA an ee-value >80% (S)-styrene oxide was observed. Furthermore, many substrate-dependent differences in the product formation rates were found. For GnIndA with the higher activity for styrene oxidation, the acceptance of larger substrates is overall reduced. For most substrates, a more than 35% reduced activity was observed while no product could be detected for 4-methoxystyrene and 1,2-dihydronaphtalene. The latter is representative for all tested substrates containing a residue in Z-position to the aromatic ring: For all of these substrates, a reduced enzyme activity was observed. In contrast, for E-β-methylstyrene (observed activity~358 mU mg −1 ), a 69 ± 3% increased activity relative to styrene (observed activity~212 mU mg −1 ) was determined indicating a clear E-selectivity of the enzyme. It is especially noteworthy that GrStyA shows an opposite behavior in the relative activity toward Eand Z-β-methylstyrene, being more active in the conversion of the Z-substrate.
For GrStyA, Z-β-methylstyrene, β,β-dimethylstyrene and 1,2-dihydronaphtalene were found to show the same or higher enzyme activity than toward styrene. Interestingly, the same compounds that contain the previously mentioned residue in Z-position cause a decreased epoxidation activity in GnIndA. In addition, for GrStyA a reduced activity for E-β-methylstyrene was observed which indicates an overall Z-selectivity. Furthermore, GrStyA converts 1,2-dihydronaphtalene at a comparable rate than styrene while 1-phenyl-1-cyclohexene is converted significantly slower (7 ± 1%). Thus, GrStyA was found to be Z-selective.
In general, GrStyA catalyzed the epoxidation of all tested substrates and overall, a lesser reduction in activity was observed for most of the substrates. Nevertheless, the actual activity of GnIndA is still higher than of GrStyA for most tested substrates.
For substrates with ether residues at the aromatic ring, a substitution-specific decrease in activity was observed for both enzymes while the relative product formation was slightly higher for GrStyA. Interestingly, the position of the ether moiety correlates similarly with the activity decrease: A methoxy group in meta-position was accepted best by both enzymes while the same substituent in para-position caused the lowest conversion rate, if converted at all, for both enzymes. These position-specific correlations in enzymatic activity indicate a similar substrate binding at the aromatic ring while the recognition at the reactive double bond was found to be different due to clear isomer preferences for the enzymes (GrStyA: Z-selective, GrIndA E-selective).
In addition to the substrate spectrum, the conversion of the standard substrate styrene to styrene oxide by both enzymes was investigated in a higher resolved time scale (Figure 4). GnIndA reached a plateau in product amount of 0.70 ± 0.03 mM styrene oxide (35.0% conversion) after 30 min while GrStyA produced a maximal amount of 0.47 ± 0.04 mM styrene oxide (23.5% conversion) after 80 min. Further investigations showed that GrStyA is stable under the reaction conditions and the cosubstrate limits the reaction, although a five-fold excess of BNAH was supplied ( Figure S3). The enzymes used most of the supplied BNAH to produce hydrogen peroxide instead of performing the epoxidation of the substrate. It can be estimated that GnIndA has an uncoupling rate of 86% while GrStyA shows 91% uncoupling. This uncoupling effect was also described for other enzymes, is typically influenced by the type of electron donor, and was reported for NAD(P)H-mimics before [31][32][33]. Nevertheless, this obvious drawback of BNAH can be overcompensated by continuous addition of the cosubstrate ( Figure S3) which is justified by the low price and easy provision of the compound.
Catalysts 2020, 10, x FOR PEER REVIEW 7 of 14 while the recognition at the reactive double bond was found to be different due to clear isomer preferences for the enzymes (GrStyA: Z-selective, GrIndA E-selective).
In addition to the substrate spectrum, the conversion of the standard substrate styrene to styrene oxide by both enzymes was investigated in a higher resolved time scale (Figure 4). GnIndA reached a plateau in product amount of 0.70 ± 0.03 mM styrene oxide (35.0% conversion) after 30 min while GrStyA produced a maximal amount of 0.47 ± 0.04 mM styrene oxide (23.5% conversion) after 80 min. Further investigations showed that GrStyA is stable under the reaction conditions and the cosubstrate limits the reaction, although a five-fold excess of BNAH was supplied ( Figure S3). The enzymes used most of the supplied BNAH to produce hydrogen peroxide instead of performing the epoxidation of the substrate. It can be estimated that GnIndA has an uncoupling rate of 86% while GrStyA shows 91% uncoupling. This uncoupling effect was also described for other enzymes, is typically influenced by the type of electron donor, and was reported for NAD(P)H-mimics before [31][32][33]. Nevertheless, this obvious drawback of BNAH can be overcompensated by continuous addition of the cosubstrate ( Figure S3) which is justified by the low price and easy provision of the compound.

In Silico Analysis
Because we found the IMO GnIndA from Gemmobacter nectariphilus DSM15620 to be active even in the presence of substantial amounts of organic solvents while the SMO GrStyA from Gordonia rubripertincta CWB2 was not, we focused on sequential and structural differences of the enzymes to explain the observed behavior. For this, the structure of the SMO PpStyA from Pseudomonas putida S12 (PDB accession: 3IHM) was used as template since it is the only known SMO structure so far [34]. Sharing 60% sequence identity with GrStyA and 29% with GnIndA, PpStyA was used to create valid models using the software Modeler 9.23 without the need for further optimization ( Figure S4) [35,36].
The PpStyA structure was subjected to the FireProt and HotSpot Wizard webservers which highlight so-called hot spot amino acids based on energy minimization and evolutionary mutations [37,38]. Compared to the structural model, these amino acids were found to be located mainly at junctions and interaction surfaces of secondary structure elements which are important for the overall protein stability ( Figure 5). In addition, FireProt suggested mutations for these locations to increase the thermostability of the structure (Figure 5a, Table S3). The mutations are further differentiated in energy mutants which result from the artificial energy optimization with random mutagenesis of the structure and in evolutionary mutants which highlight natural occurring mutations in determined hot spots.

In Silico Analysis
Because we found the IMO GnIndA from Gemmobacter nectariphilus DSM15620 to be active even in the presence of substantial amounts of organic solvents while the SMO GrStyA from Gordonia rubripertincta CWB2 was not, we focused on sequential and structural differences of the enzymes to explain the observed behavior. For this, the structure of the SMO PpStyA from Pseudomonas putida S12 (PDB accession: 3IHM) was used as template since it is the only known SMO structure so far [34]. Sharing 60% sequence identity with GrStyA and 29% with GnIndA, PpStyA was used to create valid models using the software Modeler 9.23 without the need for further optimization ( Figure S4) [35,36].
The PpStyA structure was subjected to the FireProt and HotSpot Wizard webservers which highlight so-called hot spot amino acids based on energy minimization and evolutionary mutations [37,38]. Compared to the structural model, these amino acids were found to be located mainly at junctions and interaction surfaces of secondary structure elements which are important for the overall protein stability ( Figure 5). In addition, FireProt suggested mutations for these locations to increase the thermostability of the structure (Figure 5a, Table S3). The mutations are further differentiated in energy mutants which result from the artificial energy optimization with random mutagenesis of the structure and in evolutionary mutants which highlight natural occurring mutations in determined hot spots.  does not cause hydrophobic interaction with the β-sheet amino acids. The S249F mutation increases the interaction between both secondary structure elements. In the GnIndA model, W251 causes hydrophobic interactions as well but additionally rigidifies the structure by a hydrogen bond to N192 of the β-sheet below. B: In the PpStyA structure, two helices are connected by hydrogen bonds between N51 (red) and N309 while A49 contributes little to the hydrophobic interactions in this central interaction point of three α-helices. By means of FireProt, stronger interactions are achieved by the double mutation A49F/N51F (∆∆G = −4.4 kJ mol −1 ). The GnIndA model already contains F50 while C48 contributes more to the hydrophobic interactions than A49 in the PpStyA structure. C: The mutation Q155D leads to an alternative hydrogen bond network forcing a more rigid α-helical structure while the tightness of the hydration shell is strengthened due to a charge increase by the K154R mutation.
Energy mutation simulations caused in most cases the introduction of large hydrophobic amino acids while the evolutionary mutations are more diverse, albeit expressed in less extreme structural or functional amino acid exchanges (Table S3). Thus, evolutionary mutations strengthened mainly existing interactions by different mechanisms: through an increase of hydrophobic interactions (e.g., S249F or A49F and N51F, Figure 5bA,B), an introduction of electrostatic interactions (e.g., Q155D, Figure 5bC), an increase of rigidity (e.g., N285P, not shown) or through the variation of charges to increase the tightness of the hydration shell (e.g., K154R, Figure 5bC).
As mentioned before, most of the suggested mutations for PpStyA were found at interaction surfaces of secondary structure elements. As an example, S249 is located in the PpStyA structure at the tip of the interaction site of the upper helix with the β-sheet below and does not contribute to any hydrophobic interaction with the surrounding amino acids (Figure 5bA). To increase the interaction between both secondary structure elements and prevent unfolding at this position, the replacement of serine by a phenylalanine (S249F) was suggested by FireProt. In the GnIndA model, tryptophan (W251) was found in this position which allowed not only increased hydrophobic interactions but additionally rigidifies the structure by a hydrogen bond to N192 of the β-sheet. This and further examples (Figure 5bB,C) highlight that mutations suggested by the FireProt and HotSpot Wizard webservers which are found in the native GnIndA sequence fulfill the same purpose with similar contributions to the overall protein stability. Based on these data, we state that the presence of mutations in the PpStyA structure suggested by FireProt which are found in the protein sequence of interest correlate with an increased thermo-and solvent stability of the protein as well.
For quantification of the effects, a multiple sequence alignment of PpStyA with GnIndA together with eight closest related IMOs and GrStyA together with 12 related SMOs revealed that half of the suggested mutations for the PpStyA structure were naturally present in the GnIndA sequence (18 out of 36, Figure S5). Except for one single mutation (Y328F), all evolutionary mutations are present in the GnIndA sequence (17 of 18, Table S3). In the PpStyA structure, theses mutations cause a predicted difference in thermostability of ∆∆G = −21.5 kJ mol −1 . From the suggested energy mutations, only a single mutation was found in the GnIndA sequence (A251Y) which contributes to structural stability of ∆∆G = −12.1 kJ mol −1 in the PpStyA structure optimized by FireProt. However, in addition, three correlating residues were found in the GnIndA sequence (C49, V226 and W249) which have the same functionalities than the suggested mutations, so that they probably contribute similarly to improved structural stability.
In contrast, native GrStyA contains only 19% of the mutations suggested by FireProt (7 out of 36). These variations should still result in an increased stability compared to native PpStyA (∆∆G = −10.2 kJ mol −1 ). Still, GnIndA can be expected to have a higher stability regarding elevated temperatures which will in last consequence correlate with the enhanced tolerance toward organic solvents.

Discussion
Two flavin containing monooxygenases, the SMO GrStyA and the IMO GnIndA, were probed for their applicability in the stereoselective epoxidation of styrene and eight differently substituted derivatives. Their relative activity and their tolerance toward the varied substitution patterns differed significantly, equally their tolerance toward organic solvents in reaction scenarios relevant for a potential industrial application of these enzymes. The herein presented data give a highly differential picture of these interesting biocatalysts worthy of more in-depth investigation.
While we observed a difference in oxygenation activity for flavin-based two-component monooxygenases in previous studies [11], an in-depth screening of substrate-, cosubstrate-and cosolvent-dependent reactivity has not been reported so far. The conversion of styrene derivatives with a hydrophobic substitution pattern was confirmed for both enzymes under investigation. While GrStyA shows an overall broader substrate spectrum, a generally higher activity was observed for GnIndA in most cases.
The major difference between both enzymes is their altered product formation rate in the presence of organic solvents. For GnIndA, a constant product amount was detected in presence of most tested solvents up to a concentration of 15 vol.%. Furthermore, in presence of 25 vol.% methanol, 43% residual styrene oxide production was observed and the addition of 5 vol.% THF resulted in the highest increase in product amount for all solvents and concentrations tested. In contrast, addition of 10 vol.% cosolvent resulted in significantly decreased product formation when GrStyA was used as biocatalyst for all solvents except for DMSO and ethanol. For those, a product amount above 80% relative to unsubstituted styrene was still detected.
To explain the observed differences with respect to supplied solvents, structural calculations were performed. The styrene monooxygenase PpStyA from Pseudomonas putida S12 (PDB accession: 3IHM) was used in our in silico approach since it is the only reported SMO structure so far ( Figure S5). By means of FireProt one can predict potential mutations to reach a more (thermo)stable enzyme. This tool can be supported by the HotSpot Wizard server, predicting potential sites to improve overall protein stability. Solvent-and thermostability often go hand in hand and thus we employed these tools together to rationalize our findings. The results revealed that GnIndA contains half of the suggested mutations (18 out of 36, FireProt) and shows functionally altered amino acids in highlighted stability hot spots in 86% of all cases (36 out of 42, HotSpot Wizard). By analysis of these amino acids in context of the protein structure, we showed a functional conservation in the model for GnIndA for these highlighted amino acids and demonstrated a structure-sequence dependency. Based on these data, we deduce higher stabilities for proteins that intrinsically contain the in silico suggested mutations in their native sequence. GrStyA naturally contains only 19% of the FireProt (7 out of 36) and 52% HotSpot Wizard sites (22 out of 42). These results indicate a higher thermostability of GnIndA which correlates with the observed increased tolerance toward organic solvents.
The IMO GnIndA produces a remarkably high styrene oxide amount in presence of several water-miscible solvents. In addition, the enzyme contains a unique set of mutations among related IMO which correlate with structural changes suggested by the FireProt webserver in order to generate a more stable protein. Combining both facts, we consider GnIndA to be the first solvent-tolerant IMO.
Interestingly, seven of the 18 crucial amino acids of GnIndA that are altered in comparison to PpStyA are conserved in the other nine aligned IMOs while no such conservation was found for the aligned SMO sequences (Table S3). In general, the chances to find solvent tolerance can be estimated to be higher for IMOs than SMOs. Further investigations on sequence level will show if other enzymes with higher stability than GnIndA exist in the subfamily. sulfoxide for GrStyA. An appropriate amount of enzyme (3.0 µM GrStyA, 2.7 µM GnIndA) was added and the solution was preheated to 30 • C for ten minutes at 750 rpm shaking before the reaction was started by addition of 2 mM substrate. Samples were taken immediately after substrate addition and after 2 h reaction time. If not indicated differently, styrene was used as standard substrate for all comparative studies.
Cofactor, cosubstrate and cosolvent optimization. The enzymatic reaction was performed as described above, but the concentration of the investigated compound was chanced according to the following concentrations. The minimal required amount of dissolved FAD cofactor was identified by applying a concentration range from 0 to 100 µM FAD. For determination of the optimal cosubstrate concentration a range between 5 and 25 mM BNAH was applied in the standard conditions. To determine the optimal cosolvent and its concentration for GnIndA, 5 to 25 vol.% of a variety of different water-miscible solvents were tested in 5 vol.% increments. For GrStyA, 10 vol.% concentrations were tested. All conditions were compared by the amount of produced styrene oxide after 2 h reaction time normalized to the produced amount under the above-described standard conditions. Initial enzyme activity. The enzymatic reaction for all tested substrates was performed in triplicates under the standard conditions described above. The volume of the reaction increased to 250 µL and 30 µL samples were taken after 5, 10, 20 and 30 min of reaction time and analyzed by HPLC. The initial activity of the respective enzymes toward a certain substrate was determined relative to styrene by comparing the slope of the linear part in product formation to the slope observed for styrene as substrate.
Time-dependent product formation. The enzymatic reaction was performed under the standard conditions described above in a twelve-fold multiplicate. Within the first hour, samples were taken every ten minutes and then every 20 minutes, from three random vials to a maximum of three samples per vial to avoid effects caused by the loss of volume during the experiment.
Synthesis of product standards. For determination of the HPLC retention times for the enzymatic products, the following three reactions were compared: First, the purchased substrate standard was measured. Through addition of excess meta-chloroperoxybenzoic acid to the substrate containing measurement vial in solid form, the respective epoxide was synthesized chemically in situ, which resulted in the observation of a second peak corresponding to a more polar substance. The further addition of water containing 0.1 vol.% trifluoroacetic acid led to disappearance of the epoxide peak and to detection of an even more polar compound, which is assigned to the respective diol. The procedure was validated for styrene, for which commercial standards of styrene oxide and 1-phenylethane-1,2-diol were available.
HPLC analysis. The produced epoxide amount was quantified by HPLC measurements (Thermo Scientific, Oberhausen, Germany) using an isocratic method at 60 vol.% acetonitrile in water on a Knauer Eurospher 100-5 C18 Vertex plus column (size: 125 × 4 mm, pore size: 100 Å, particle size: 5 µm; Knauer, Berlin, Germany) as stationary phase. Compounds were detected by UV/vis absorbance at 214 nm. The system was calibrated using commercially available styrene and racemic styrene oxide.
GC-FID analysis. Chiral GC-FID analysis was used to determine the enantiomeric excess of the product. For this, the enzymatically produced epoxidation product was extracted from the reaction solution with one volume ethyl acetate. The organic phase was dried over anhydrous magnesium sulfate and analyzed using a Macherey-Nagel Hydrodex β-6TBDM column (length: 50 m, inner diameter: 0.25 mm; Macherey-Nagel, Düren, Germany) under isothermal conditions at 100 • C by means of GC-FID (Shimadzu, Duisburg, Germany). The calibration was performed using commercially available styrene and (S)-styrene oxide.
In silico calculation of hot spot amino acids. The only known structure of the styrene monooxygenase PpStyA from Pseudomonas putida S12 (PDB accession number: 3IHM) was used to calculate crucial amino acids for protein stability using the HotSpot Wizard [38] and FireProt [37] webservers. The results were compared by a multiple sequence alignment of 15 SMOs closely related to GrStyA and 10 IMOs related to GnIndA. The alignment was calculated using Mega X (10.1.5) [41]. From the absence or presence of suggested mutations in each sequence in comparison to PpStyA, conclusions about the overall stability were drawn.
Homology modeling. The structural model for GnIndA was constructed using Modeler 9.23 [35]. The structure of PpStyA from Pseudomonas putida S12 (PDB accession number: 3IHM) served as template together with the structure files obtained from the FireProt results (see above) as template.

Conclusions
To sum up, we successfully optimized the reaction conditions for two flavin containing two-component monooxygenases regarding the conversion of large hydrophobic substrates. The intriguingly different preference for Erespective Z-enantiomeric substrates by GrIndA resp. GrStyA bears a high potential for industrial applications, e.g., when it comes to enantiomer separation in complex mixtures.
Furthermore, GnIndA was found to be remarkably tolerant toward nine tested organic solvents. This can be explained by beneficial mutations in crucial residues matching the predicted increased stability based on in silico calculations. In addition, these suggested mutations can be used to construct the most stable protein sequence for a targeted search for even more stable IMOs, but at the same time give valuable insight into potential stabilization strategies also for SMOs. Respectively, we could simplify and conclude that IMOs are more stable biocatalysts.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4344/10/5/568/s1, Figure S1. SDS gel of the protein production, Table S1. Organic solvents used to optimize the enzymatic epoxidation of styrene, Figure S2. Concentration of 1-phenyl-1-cyclohexane in aqueous solution in presence of different organic cosolvents, Table S2. Enantiomeric excess of (S)-styrene oxide produced by GnIndA in presence of selected organic solvents, Figure S3. Continuously fed biotransformation of styrene to styrene oxide by GrStyA, Figure S4. Results of the quality check of the structural GnIndA model, Figure S5. Multiple sequence alignment of GnIndA, GrStyA and PpStyA with 22 related SMOs and IMOs, Table S3. FireProt mutations in comparison with their abundance among IMOs and SMOs.