*2.5. Evaluation of Double Substitution at Sites P76 and R330 of OYE2y*

To investigate the effect of the double substitutions on P76 and R330, we firstly conducted site-directed mutagenesis of P76 to C and R330 to H, D, W, and C, resulting in the four variants (Table 4). The *e.e.* values and yields of the resulting variants were higher than those of the variant P76C, but similar to those of the corresponding R330H, R330D, R330W, and R330C. On the other hand, the other set of double substitutions was created by site-directed mutagenesis of R330 to H and P76 to M, G, and S (Table 4). Among them, the (*E*)-citral-derived *e.e.* values were significantly increased up to >99% (*R*) with relatively higher yields (64.09%~73.88%). In the (*E*/*Z*)-citral reduction, the variants P76M/R330H, P76M/R330H, and P76M/R330H also exhibited full (*R*)-enantioselectivity despite lower product yields (9.12–15.83%).


**Table 4.** The catalytic performance of double substitution variants of OYE2y at sites P76 and R330 a.

<sup>a</sup> Data present mean values ± SD from three independent experiments. (*E*)-citral contained 98.38% geranial and 1.62% neral, (*Z*)-citral contained 96.84% neral and 3.16% geranial, and (*E*/*Z*)-citral contained 58.45% geranial and 41.55% neral.

#### **3. Discussion**

For the synthesis of (*R*)-citronellal from OYE-mediated citral reduction, low-cost (*E*/*Z*)-citral is the industrial desire in contrast to geranial and neral [9]. However, (*E*/*Z*)-citral reduction remains challenging due to limited chemoselectivity and enantioselectivity. On the one hand, the presence of multiple C=C and C=O double bonds of citral makes the whole-cell biocatalyst impossible to avoid side reactions [33]. Thus, the purified enzyme is commonly required as biocatalyst. On the other hand, the hydrogenated products from the geometric isomers geranial and neral are usually enantiocomplementary when the wild type OYE was used as biocatalyst, reducing the reaction's enantioselectivity. The features of (*E*/*Z*)-citral reduction make it difficult to implement a high-throughput screening (HTS) method for determining the enantioselectivity of the large variant libraries [34–36]. Typically, the samples must be examined individually by chiral GC, which requires at least thirty minutes per sample. To keep the variant library as minimum as possible, the best strategy for enantioselectivity alteration turns to be site-saturation mutagenesis of individual key residue(s) rather than the HTS-based directed evolution [18,37,38]. Based on the strategy of site-saturated mutagenesis, several groups have successfully increased (*R*)-enantioselectivity in the (*E*/*Z*)-citral reduction, and our study further demonstrated that it was feasible to achieve the full (*R*)-enantioselectivity in the OYE-mediated (*E*/*Z*)-citral reduction through protein engineering.

The R330X and P76X variant libraries yielded 17 and five variants with improved (*R*)-enantioselectivity, respectively, indicating that the subtle change of structure would significantly affect the enantioselectivity. In the variants R330H and P76C, the amino acid pair R and H possessed electrically charged side group, while the side groups of the amino acid pair P and C was polar and uncharged. It was suggested that the substitution could give priority to amino acid residue(s) with similar side group in terms size, polarity and charge, if no clear structure–function relationship was available. Furthermore, the (*Z*)-citral-derived *e.e.* value of R330H (71.92%, *R*) was much higher than that of OYE2p (26.5%, *R*) [7], suggesting that the role of S13, S59, and V289 could not be ignored for the (*R*)-enantioselectivity. For the single substitutions, the (*R*)-enantioselectivity improvement in the (*E*/*Z*)-citral reduction was mainly attributed to the enantioselectivity inversion in the (*Z*)-citral reduction, meanwhile, the same variant reduced (*E*)-citral to (*R*)-citronellal with similar *e.e.* values. Different from the single substitutions, the double substitutions led to further improved (*R*)-stereoselectivity in the (*E*)-citral reduction rather than (*Z*)-reduction, which was rarely observed in the enantioselectivity alternation of OYEs.

To get detailed insights into the molecular mechanism, the models of OYE2y were created from the OYE1 structure (PDB code: 1OYB) using SWISS-MODEL and molecular docking was conducted using the program AutoDock Vina. The substrate and FMN were docked in silico into the models of OYE2y and its variants. In the model of wild type OYE2y, a conserved H192/N195 pair formed hydrogen bonds with the carbonyl oxygen of α,β-unsaturated carbonyl compounds; a hydride was enantioselectively transferred to the substrate C<sup>β</sup> atom from FMNH2; and the Y197 residue provided a proton to the substrate Cα atom as an electron acceptor [17]. From the model of wild type OYE2y, the distances from Cβ of neral to the side groups of P76 and R330 were calculated to be 14.92 and 20.84 Å, respectively (Figure 3). Similarly, the distances from C<sup>β</sup> of geranial to the side groups of P76 and R330 were calculated to be 14.49 and 21.16 Å, respectively. It was assumed that the residues affecting the enantioselectivity of OYEs might directly interact with the substrate [25]. Our results indicated that the residue distant from active sites, e.g., R330 in OYE2y, could also be pivotal for determining the enantioselectivity of OYEs. Furthermore, substrate modeling into the wild type enzyme revealed two different binding modes for the two isomers geranial and neral, leading to products with different enantioselectivity. The docking analyses of variants P76C, R330H, and P76M/R330 suggested that the reversed enantioselectivity in the neral reduction was due to the flipped binding orientation that placed the opposite face of the alkene above the *si* face of the FMN cofactor [24]. Meanwhile, the same variant reduced (*E*)-citral with preserved (*R*)-enantioselectivity derived from the same binding orientation as wild type OYE2y (Figure 4). Although it was acceptable that enantioselectivity was controlled by tuning the orientation of substrate in the binding sites, how subtle changes control the orientation of substrate binding remains to be unraveled, and X-ray crystallography of the variants in the future study would benefit to clarify the subtle structural differences at the substrate-binding site.

**Figure 3.** The residues targeted for site-saturation mutagenesis in the homology model of OYE2y. The model structure of OYE2y was constructed with the crystal structure of OYE1 (PDB code: 1OYB) as template. Distances between the Cβ atom of geranial (**A**) and neral (**B**) and side chains of P76 and R330 were determined. Green, carbon atom; blue, nitrogen atom; tangerine, oxygen atom; white, hydrogen atom; orange, phosphorus atom.

**Figure 4.** The binding modes of citral isomers in OYE2y and its variants R330H (**A**), P76C (**B**), and P76M/R330H (**C**) leading to either (*R*)- or (*S*)-citronellal. The catalytic residues H192, N195, Y197, and the prosthetic group FMN were depicted. Green, carbon atom; blue, nitrogen atom; tangerine, oxygen atom; white, hydrogen atom; orange, phosphorus atom.
