The (+)-pulegone reductase gene from Mentha piperita and (−)-pulegone reductase gene from Nepeta tenuifolia were cloned into the pET28a vector under control of the bacteriophage T7 gene promoters using NheI and HindIII, respectively. The (−)-pulegone reductase gene from Agastache rugose was cloned into the pET28a vector under control of the bacteriophage T7 gene promoters using BamHI and SacI. Each resulting plasmid was transformed into E. coli strain BL21(DE3) (Invitrogen). A single colony of the resulting transformants was used to inoculate 50 mL of LB broth containing 50 mg/mL kanamycin, and the culture was incubated at 37°C for 16 h with shaking. An aliquot (10 mL) was used to inoculate 1 L of LB broth containing 50 mg/mL kanamycin, followed by the incubation with agitation at 37°C, 180 rpm until the OD₆₀₀ reached 0.8–1.0. Subsequently, the culture was supplemented with isopropyl-b-D-thiogalactoside to the final concentration of 0.5 mM for protein expression, and then the culture was incubated for 16 h at 20°C. Cells were harvested by centrifugation (5,000 × g; 15 min at 4°C), re-suspended in buffer A (10 mM Tris-HCl, pH 8.0, 200 mM NaCl, and 5% glycerol), and lysed using an EmulsiFlex-C5 cell disruptor (Avestin). The lysate was centrifuged (20,000 × g; 30 min at 4°C) and the cell debris was removed. The supernatant was loaded onto a 5 mL column of Ni²⁺-NTA-agarose (Qiagen) pre-equilibrated with buffer A. The column was washed with 10 × 5 mL buffer A containing 25 mM imidazole followed by elution with 30 mL buffer A containing 200 mM imidazole. The eluate was concentrated to ∼10 mg/mL and further purified by gel filtration chromatography on a HiLoad 16/60 Superdex 200 prep grade column (GE Healthcare) in 10 mM Tris-HCl, pH 8.0, 50 mM NaCl, 5 mM MgCl₂, and 1 mM DTT. The peak fractions were collected and concentrated to the final concentration of 20 mg/mL in the same buffer using 30 kDa MWCO Amicon Ultra-15 centrifugal ultrafilters (EMD Millipore), and stored in aliquots at –80°C. The yields of the proteins were ∼20 mg/mL, and purities were ∼95%. Site-directed mutations were prepared using one step PCR method, and the mutated proteins were expressed and purified by following the same protocol as the wild-type proteins. High and low molecular weight (mass) calibration kits (GE Health-care) were used to calibrate the molecular mass of wild type MpPR.

The reduction reaction (0.4 mL) catalyzed by pulegone reductase was performed in buffer B (50 mM KH₂PO₄, 10% sorbitol, 1 mM DTT, pH 7.5), containing 20 μM substrate [(+)-pulegone (CAS No: 89-82-7) or (−)-pulegone (CAS No: 3391-90-0), Sigma], 10 mM NADPH tetrasodium salt hydrate (CAS No:2646-71-1, Sigma), 6 mM glucose-6-phosphate, 20 U glucose-6-phosphate dehydrogenase (Solarbio), and 30 μM MpPR or 36 μM NtPR. 0.2 mL of n-hexane was added on the top of the reaction solution. Reaction was carried out at 31°C for 1 h (MpPR) and 16 h (NtPR) with slowly stirring. The reaction was terminated by placing the reaction vial at –20°C for 2 h. The upper organic phase was transferred into a new 2 mL glass vial containing a conical glass insert and immediately analyzed by GC-MS and chiral GC. The negative control containing inactive enzyme was generated by heating the reconstitution mix for 15 min at 95°C.

GC-MS analysis was performed by Agilent 7890B/7000C, equipped with a HP 5MS capillary column (30 m × 0.25 mm; film thickness 0.25 μm). The programmed temperatures of the column were set as follows: 85°C for 4 min, 85–130°C at 5°C/min, 130°C for 2 min, 130–140°C at 5°C/min, 140°C for 3 min. Ion source temperature was set at 230°C. Electron ionization (EI) mass spectra were acquired over the mass range 50–500 Da at the energy of 70 eV. Chiral GC analysis was carried out using Agilent 8860, equipped with a CYCLODEX-B capillary column (30 m × 0.32 mm; film thickness 0.25 μM). The programmed temperatures of the column were set as follows: 80–95°C at 2°C/min, 95–110°C at 0.5°C/min. Injector temperature and carrier gas (Nitrogen) detector temperature were set at 250°C.Nitrogen was used as the carrier gas at a flow rate of 1 mL/min with an injection volume of 1 μL, no split.

The reduction reaction was initiated in a 400 μL solution containing 50 mM KH₂PO₄, 10% sorbitol, 1 mM DTT, 10 mM NADPH tetrasodium salt hydrate, 6 mM glucose-6-phosphate, 20 U glucose-6-phosphate dehydrogenase (Solarbio), together with the wild-type and mutated MpPR, ArPR, and NtPR (Supplementary Table 1). Other conditions are the same as the in vitro enzyme catalysis assay. Quantitative analysis was performed by the comparison of the peak areas of products to the standards of known concentrations obtained from chiral GC analysis (Agilent 8860 equipped with a HP-5 capillary column; the column condition is the same as described above). All biotransformation reactions were performed in at least duplicates, and the results are averages of the data. The yields of menthone and iso-menthone at each concentration were calculated by the comparison of the peak areas of products to the standards of known concentrations. The kinetic parameters K_m and k_cat were calculated using Equation 1.

where V₀ is the initial velocity, [E] is the enzyme concentration, [S] is the substrate concentration, V_max is the maximum velocity; K_m is the Michaelis constant and k_cat is calculated from V_max/[E].

Robotic crystallization trials were performed for MpPR and NtPR as well as co-crystallization with NADPH tetrasodium salt hydrate and (+)- or (−)-pulegone using a Griffin liquid handling system (Art Robbins Instruments), commercial screening solutions (Emerald Biosystems, Hampton Research, and Qiagen), and the sitting-drop vapor-diffusion technique (drop: 0.2 μL protein plus 0.2 μL screening solution; reservoir: 60 μL screening solution; 20°C). 1,200 conditions were screened. Under several conditions, MpPR crystals appeared within 2 weeks. Conditions were optimized using the hanging-drop vapor-diffusion technique at 20°C. The optimized crystallization condition for MpPR was as follows: 1 M ammonium sulfate, 0.1 M Bis-Tris (pH 5.5), 1% W/V PEG 3,350 at 20°C. Crystals were transferred to reservoir solution containing 20% (v/v) glycerol and flash-cooled with liquid nitrogen. Diffraction data were collected from cryo-cooled crystals at SSRF BL17U. Data were processed using HKL2000 (Otwinowski and Minor, 1997) and CCP4i (Winn et al., 2011). The resolution cut-off criteria were: (i) I/σ > = 2.0, (ii) CC_1/2 (highest resolution shell) > 0.5.

The structure of MpPR was solved by molecular replacement with MOLREP (Adams et al., 2010; Vagin and Teplyakov, 2010) using the structure of native raspberry ketone synthase from Rubus idaeus (PDB ID 6EOW) as a starting model. The molecular replacement solution was good, and an automatic model building was performed with Phenix (Adams et al., 2010). Additional model building was done manually with Coot (Emsley and Cowtan, 2004) and refined with Phenix. The final model of MpPR was refined to 2.7 Å resolution. The final model for MpPR was refined to R_work and R_free of 0.27/0.29 (Table 1).

All molecular docking studies were performed using Autodock4.2 package (Morris et al., 2009). Briefly, crystal structure of MpPR enzyme was docked with (+)-pulegone. The molecule was added with non-polar hydrogens and assigned partial atomic charges using AutoDockTools (ADT) (Morris et al., 2009). The coordinates of NADP(H) and (+)-pulegone in MpPR structure was generated based on the coordinates of hydroxybenzalacetone from the crystal structure of RiDBR (PDB ID: 6EOW) and the coordinates of p-coumaryl aldehyde of the crystal of AtDBR (PDB ID: 2J3J) in combination with CORINA Classic online service. A grid box with 40 × 40 × 40 grid points and 0.2 Å grid spacing centered roughly at the pulegone binding position was used as the searching space. 100 runs of Larmarckian Genetic Algorithm were performed to search the protein-ligand interactions. The results were clustered and ranked. Result analyses and figure rendering were performed using PyMOL. The structure of (−)-pulegone reductase from N. tenuifolia was modeled by the online artificial intelligence tfold2 program.¹

The crystal structure of MpPR was deposited into Protein Data Bank under accession number 7EQL. UniProt IDs: Q6WAU0 for MpPR; Q39172 for AtDBR; Q9SLN8 for NtDBR; G1FCG0 for RiDBR; A0A5N5GUE7 for MdDBR. National Center Bioinfomratic Center (NCBI) accession number: MZ504956 for NtPR; MZ504957 for ArPR.

Pulegone reductase from Mentha piperita (MpPR) that catalyzes the reduction of the C2–C8 double bond of (+)-pulegone to (−)-menthone using NADP(H) as a co-factor has been established for decades (Lange, 2015). Besides M. piperita, many other medicinal herbs, such as Nepeta tenuifolia and Agastache rugosa are also producers of essential oils, in which menthone and menthol are the major components (Chen et al., 2019). Intriguingly, it has been reported that N. tenuifolia and A. rugosa produce (+)-menthone and (+)-menthol as the major metabolites, in contrast to M. piperita that predominantly biosynthesizes (−)-menthone and (−)-menthol (Chen et al., 2019).

To investigate the underlying reasons for the substrate selectivity between M. piperita and N. tenuifolia or A. rugosa, we first compared the evolutionary relatedness of these three plants by performing molecular phylogenetic analysis based on a cascade of 28S-18S-5.8S rDNA sequences from their genomes. The results clearly showed that M. piperita, N. tenuifolia, A. rugose, Sesamum indicum, and Amborella trichopoda were clustered into a big group. More importantly, M. piperita, N. tenuifolia, and A. rugose were further classified into an individual subcluster, indicating that the three plants may be evolutionarily related (Supplementary Figure 2). Next, we analyzed menthone biosynthesis pathway in the genus M. piperita in which was well studied before. Limonene synthases, isopiperitenone reductase, and pulegone reductase have been proven to be able to catalyze the committed step in the biosynthesis of menthone until now (Supplementary Figure 1B; Ringer et al., 2003, 2005). Moreover, the structure-function relationships underlying the formation of limonene enantiomers in limonene synthases and candidate active site residues with critical roles in catalyzing reactions that involve accommodating reaction intermediates of opposite enantiomeric series have been identified (Hyatt et al., 2007; Srividya et al., 2020). However, there is no information concerning the biosynthesis mechanism of limonene downstream product enantiomers and relevant biosynthesis enzymes like isopiperitenone reductase or pulegone reductase so far. Therefore, we performed RNA-seq analysis on N. tenuifolia that produces (−)-pulegone and (+)-menthone in order to locate the candidate genes of (−)-pulegone reductase (Liu et al., 2021). Based on our N. tenuifolia RNA-seq data, three candidate genes of pulegone reductase (cluster-16657.51589, cluster-16657.19187, cluster-16657.38628) sharing high identities (∼65, ∼75, and ∼64%, respectively) with MpPR were proposed to be NtPR (Supplementary Figure 3). We then selected cluster-16657.51589 as our target gene in view of its highest expression level in leaves and the leaves contain the highest ratio of monoterpenes (e.g., pulegone, menthone) among all the tissues of N. tenuifolia (Supplementary Figure 4). Subsequently, we performed sequence alignment analysis of NtPR (cluster-16657.51589) and MpPR together with another four DBRs (AtDBR, NtDBR, MdDBR, and RiDBR) in the MDR superfamily. The alignment revealed that most residues in the proposed substrate binding pocket across all aligned MDR superfamily proteins are conserved (Supplementary Figure 5, labeled with green triangles). Moreover, NtPR and MpPR showed high sequence similarity with the major exception that six amino acid residues (Leu56, Tyr78, Phe281, Val282, Val284, and Tyr287 numbered as in MpPR, Ser59, Asp81, Tyr283, Leu284, Tyr286, and Arg289 numbered as in NtPR) in the proposed pulegone binding pocket are different (Supplementary Figure 5), suggesting these amino acids might contribute to substrate or product specificity. Thus, we cloned NtPR and MpPR genes and expressed the respective pulegone reductase in Escherichia coli BL21(DE3) strain for biochemical and structural studies (Supplementary Figure 6).

To experimentally validate the substrate specificity of NtPR and MpPR, we firstly carried out in vitro enzyme activity assays, where the pulegone reductase was fed with (+)- and (−)-pulegone, respectively. The products were analyzed and quantified using gas chromatography-mass spectrometry (GC-MS). The results clearly showed that NtPR has a higher preference for (−)-pulegone to generate (−)-isomenthone (57%) and (+)-menthone (43%), confirmed by chiral GC and the fragments generated from electron impact mass spectrometry (EIMS) by comparison with standards (Figures 1A–D). In contrast, MpPR is more inclined to convert (+)-pulegone to the major (−)-menthone (69%) and the minor (+)-isomenthone (31%) as reported previously (Figures 1A–D; Lange, 2015). Next, we plotted the Michaelis–Menten curve and calculated the enzyme kinetic parameters for NtPR after feeding with (+)- or (−)-pulegones. Apparently, NtPR has lower catalytic activities on (+)- or (−)-pulegone [K_cat 0.39 10^–4s^–1 for (+)-pulegone or 0.53 10^–4s^–1 for (−)-pulegone] compared to that of MpPR [K_cat 863.66 10^–4s^–1for (+)-pulegone and 1595.06 10^–4s^–1 for (−)-pulegone]. NtPR displayed higher binding affinity toward (−)-pulegone (K_m 57.18 μM) than that for (+)-pulegone (K_m = 163.30 μM), whereas the V_max for both (+)- and (−)-pulegones appears similar (Figure 1B), confirming that (−)-pulegone is the more favorable substrate that NtPR can bind rather than (+)-pulegone. In comparison with NtPR, MpPR displayed higher binding affinity toward (+)-pulegone (K_m 3.00 μM) than that for (−)-pulegone (K_m 8.63 μM) (Figure 1B). To further investigate the substrate selectivity of MpPR and NtPR, we fed the pulegone reductases with other alkene double bond-containing substrates, including (+)-menthone, (−)-menthone, (+)-limonene, (−)-limonene, (+)-menthofuran, (−)-(1R,4S)-p-mentha-2,8-dien-1-ol, carveol, (−)-perilllc alcohol, and (−)-carvone. The results revealed that MpPR and NtPR can adopt (+)- and (−)-pulegones as substrate exclusively, and they do not show any activity toward other substrates (Supplementary Figure 7), suggesting MpPR and NtPR are the DBRs with high substrate specificity.

Using NtPR sequence as a query, we further investigated the evolutionary relationship of NtPR, MpPR, and other DBRs in the MDR superfamily. 67 DBRs belonging to the MDR superfamily from plants that have complete genome data in KEGG database were chosen for the maximum-likelihood phylogenetic analysis. As expected, the phylogenetic tree revealed that NtPR forms a unique clade, which suggests NtPR may be evolutionarily distinct from other known MDR reductases (Figure 2 and Supplementary Figure 8). In contrast, the clade where MpPR locates has more homologous members. The results inspired us to perform in-depth structural analysis of NtPR and MpPR toward the better understanding of substrate and product stereoselectivity.

To elucidate the molecular basis underlying the differences in substrate and product specificities, we attempted to obtain co-crystals of MpPR and NtPR with NADP(H) and (±)-pulegone. Finally, the crystal structure of MpPR was obtained at 2.7 Å resolution (Figure 3A), although the attempts to acquire crystal of NtPR and co-crystals of the enzyme with either NADP⁺, (+)-pulegone or both NADP⁺ and (+)-pulegone failed, due to the hydrophobic and volatile nature of pulegone. Statistics of data collection and model refinement for MpPR crystal are summarized in Table 1. Alternatively, the structure of NtPR was in silico modeled using the tfold2 program (Figure 3B), an artificial intelligence-based online structural modeling tool. By structure superimposition analysis, the overall structures of both MpPR and NtPR are similar to the known MDR superfamily enzymes RiDBR, NtDBR, AtDBR, and MdDBR (the RMSDs are 0.99, 0.85, 1.07, and 1.04 Å, Figures 3B–F), which form a different clade in the phylogenetic tree (in green, Figure 2).

The crystal structure of MpPR revealed an asymmetric unit containing only one monomer of apo-MpPR. The PDBePISA server calculations and gel filtration of the protein suggested that the biologically relevant form of the enzyme is a homodimer as reported previously for other MDR superfamily members (Supplementary Figure 9). The structure consists of two typical conserved N-terminal and C-terminal domains that are connected by a short loop. The N-terminal catalytic domain (residues 1–134 and 310–349) includes three a-helices and nine b-sheets forming a twisted partial b-barrel-like structure, while the C-terminal nucleotide coenzyme-binding Rossmann-fold domain (residues 135–309) features seven a-helices and six b-sheets, forming a typical six-stranded, parallel b-sheet sandwiched by three helices on each side (Figure 3A). The two domains are separated by a cleft containing a deep pocket that accommodates the cofactor and forms the active site (Figure 3A).

We did not observe the unambiguous electronic density for NADP(H) in the map of co-crystal of MpPR in complex with NADP(H), but two extra density peaks accounting for the phosphate group of NADP(H) were present, probably due to the crystal packing. Sequence alignment of MpPR with other DBRs (e.g., AtDBR and NtDBR) in the MDR superfamily showed that most residues predicted to interact with NADP(H) are conserved and located at the nucleotide coenzyme-binding domain of MpPR (Supplementary Figure 5, shown as purple rectangle). In order to locate the relative position of NADP(H) in MpPR, we superimposed the phosphate group of NADP(H) in the crystal structure of MpPR with that in AtDBR (PDB ID 2J3J and 2J3K). The superimposed model exhibited that the residues Asn51, Ser164, Lys189, Tyr205, and Asn331 in MpPR may form hydrogen bonds with the phosphate groups and the ribose rings of NADP(H) (Figure 4A). A number of hydrophobic residues consisting of Pro52, Tyr53, Met135, Ala162, Val165, Ala184, Cys251, Met253, Val254, Phe281, Val282, and Val283 were found surrounding the NADP(H) backbone to further stabilize the NADP(H) molecule (Figure 4A).

Due to the lack of co-crystal structure of MpPR with the substrate, we docked (+)-pulegone into the MpPR structure based on the ternary complex structure of its homolog AtDBR (PDB ID 2J3J and 2J3K). In contrast to the predicted NADP(H) binding pocket, residues Tyr53, Leu65, Tyr78, Met135, Tyr257, Phe281, Val282, Val283, Val284, and Tyr287 form a hydrophobic network stabilizing (+)-pulegone mainly via hydrophobic and van der Waals interactions (Figure 4B). In addition, residues Ser77 and Ser100 also contribute to the binding of (+)-pulegone by van der Waals interaction. Residues Tyr53 and Arg57 may interact with the ketone moiety of (+)-pulegone through polar interactions to further stabilize the binding of (+)-pulegone (Figure 4B). All these residues form a feasible (+)-pulegone binding pocket in the structure of MpPR.

To validate the predicted residues in pulegone binding pocket, we generated 25 single mutants by mutating 13 most relevant residues in the binding pockets of MpPR and assessed the effect of point mutation on enzyme activity and the yield of final product (Supplementary Figure 6). In vitro enzyme catalysis assays showed that the wild-type MpPR can convert (+)-pulegone to the products (−)-menthone and (+)-isomenthone with the ratio of 2:1 (Figure 5A and Supplementary Figure 10). The yield of the products decreased when the MpPR mutants L56A, R57A, M135A, Y257A, F281A, V282A, V283A, V284A, and S77G were used in the reactions. Based on the docking model of MpPR with NADP(H) and (+)-pulegone, the mutation R57A may disrupt the potential weak hydrogen bonding interaction with the ketone group of (+)-pulegone, thus reducing the enzyme activity slightly. The key residues Leu56, Val282, and Val284 in the binding pocket was proposed to interact with (+)-pulegone through hydrophobic effect. The hypothesis was evidenced by the mutation of Leu56, Val282, and Val284 to the non-polar residues with larger hydrophobic side chain (Leu56 to Ile and Val, Val282 to Leu, and Val284 to Phe, Leu, and Tyr) led to a remarkable increase in the yield of products when compared with the wild-type enzyme (Figure 5). Site-directed mutagenesis analysis validated the amino acid residues Leu56, Arg57, Ser77, Tyr257, Phe281, Val282, Val283, and Val284 in MpPR binding pocket are critical to substrate binding and catalysis, consistent with the interactions revealed from the docking model of MpPR in complex with NADP(H) and (+)-pulegone.

Once the putative residues (i.e., Leu56, Arg57, Ser77, Tyr257, Phe281, Val282, Val283, and Val284) in the (+)-pulegone binding pocket of MpPR were validated, we set out to explore the underlying mechanism on why MpPR has preference to (+)-pulegone rather than (−)-pulegone as observed in our GC-MS analysis and enzyme kinetic study (Figure 1). Firstly, (+)- and (−)-pulegones were docked into the crystal structure of MpPR with NADP(H), respectively, and we observed that the angle of C-5 methyl group (5-Me) of pulegone in the binding pocket likely determines the substrate selectivity of MpPR. Specifically, 5-Me of (+)-pulegone perfectly interacts with the hydrophobic network constituted by Tyr53, Leu56, Tyr78, M135, Val282, Val283, Val284, and Tyr287, stabilizing the binding between MpPR and (+)-pulegone. In contrast, 5-Me of (−)-pulegone does not totally fit into the hydrophobic network and may destabilize the binding with MpPR (Figures 6A,B). To validate the hypothesis, we calculated the forces and energies generated during the binding between pulegone and MpPR using molecular dynamic (MD) simulations. The free energy of binding (ΔG_bind^e) calculated for (+)-pulegone to MpPR (–98.95 kcal/mol) is significantly lower than that of (−)-pulegone (–89.17 kcal/mol), indicating a tendency to adopt (+)-pulegone as the substrate by MpPR (Figure 6C). Furthermore, the non-polar interaction (ΔG_GA^d) might contribute to the free energy of binding to the most extent with the values of –29.97 kcal/mol for (+)-pulegone and –43.12 kcal/mol for (−)-pulegone (Figure 6C). The MD simulations also predicted and compared the contribution of each amino acid residue involved in the non-polar interaction with (+)- and (−)-pulegones. MD simulations results showed that the residues Tyr78, Met135, Val282, Val283, Val284, Tyr287, and particularly Tyr53 constitute much stronger hydrophobic interactions with (+)-pulegone compared with those for (−)-pulegone, suggesting these residues are highly likely involved in stereoselectivity of the substrate (Figure 6D). The calculated results are highly consistent with our hypothesis that the interaction between these residues and 5-Me of (+)-pulegone stabilize their binding and thus led to the substrate specificity.

To further experimentally investigate the importance of these predicted residues involved in substrate stereoselectivity, we mutated the hydrophobic residues Leu56, Val282, and Val284 in the binding pocket of (+)-pulegone in MpPR to the corresponding residues Ser59, Leu285, and Tyr287 in NtPR. The enzyme kinetic measurements of the wild-type MpPR revealed that its binding affinity to (+)-pulegone (K_m ∼3.00 μM) is about threefold higher than that to (−)-pulegone (K_m ∼8.63 μM). The mutations L56S, V282L, and V284Y led to 8–100-fold decrease in the binding affinity to both (+)- and (−)-pulegone (Figures 7A–E). Importantly, the mutation V282L and V284Y almost diminished the substrate stereoselectivity of MpPR as evidenced by the very similar K_m values for both (+)- and (−)-pulegone, possibly attributed to the larger side chains of the mutants that strengthen the hydrophobic interaction with (−)-pulegone (Figures 7A–E). In contrast, the mutation L56S did not display a significant effect on the substrate selectivity, suggesting Leu56 is not involved in the substrate specificity.

In vitro enzyme catalysis assays coupled with GC-MS analysis showed that the wild-type MpPR can convert (+)-pulegone to the products (−)-menthone and (+)-isomenthone with the ratio of 2:1. This observation encouraged us to explore the potential mechanism on the product stereoselectivity of MpPR. Given that the only difference between the structures of (−)-menthone and (+)-isomenthone lies in the angle of the 2-isopropyl group (2-iPr), we therefore proposed that the product stereoselectivity may be attributed to the stereochemistry of 2-iPr, which could result in different interaction strengths and binding affinities to MpPR. To test this hypothesis, we first docked (−)-menthone and (+)-isomenthone into the crystal structure of MpPR, respectively. The docking model showed that the 2-iPr of (−)-menthone is well positioned in the binding pocket of MpPR and stabilized by Tyr53, Leu56, Val282, Val283, and Val284, while the 2-iPr of (+)-isomenthone crashes the main chain of residue Tyr53 in the pocket, thus requiring higher energies to generate and stabilize with MpPR (Figures 8A,B). Subsequently, the binding affinities of (−)-menthone and (+)-isomenthone to MpPR were calculated using MD simulations. The enzyme kinetic parameters supported our deduction from the structural analysis as the free binding energy (ΔG_bind^e) for (−)-menthone (–119.43 kcal/mol) is higher than that for (+)-isomenthone (–112.75 kcal/mol) (Figure 8C). Moreover, the MD simulations also predicted the potential contributions of the surrounding residues to the non-polar hydrophobic interactions with (−)-menthone or (+)-isomenthone. The residues Tyr53, Leu56, Tyr78, Met135, Val282, Val283, and Val284 have major contributions to the binding of the products (Figure 8D). Particularly, Tyr53, M135, and V284 are inclined to stabilize (−)-menthone rather than (+)-isomenthone as revealed by the lower ΔG_GA^d values, whereas Tyr78, Val282, and Val283 have larger contribution to the stabilization of (+)-isomenthone through hydrophobic interactions (Figure 8D).

Next, we performed an in-depth analysis of the site-directed mutagenesis data to evaluate the impact of these residues on the stereoselectivity of products (−)-menthone and (+)-isomenthone. The ratio of the products (−)-menthone and (+)-isomenthone was calculated for each mutant and normalized based on those for the wild-type MpPR. The results revealed that the mutants L56S, L56I, L56V, S100I, S100A, V282T, V284F, V284L, V284Y, and V284A significantly reduced the ratio of (−)-menthone to (+)-isomenthone, but have little impact on the total yield of the products (Figures 5A,B), suggesting that the residues Leu56, Ser100, Val282, and Val284 may play critical roles in the stereoselectivity of the products (−)-menthone and (+)-isomenthone, which was consistent with our docking and MD simulation results described above.

In the enzyme catalysis assay and GC-MS analysis, we found that MpPR can also adopt (−)-pulegone as its substrate but with low affinity to generate (+)-menthone (∼46%) and the (−)-isomenthone (∼54%) (Figure 1C, Supplementary Figure 10, and Supplementary Table 1). To investigate the stereoselectivity of MpPR for products (+)-menthone and (−)-isomenthone, we performed docking study and MD simulations. The docking analysis show that the product either (+)-menthone or (−)-isomenthone can well fit the proposed product binding pocket and does not have big crash with the surrounding residues (Supplementary Figures 11A,B), indicating (+)-menthone or (−)-isomenthone may be all favored by MpPR. The MD simulations results supported the hypothesis because the free binding energy of product (+)-menthone (–141.26 kcal/mol) and (−)-isomenthone (–116.74 kcal/mol) is both relatively low (Supplementary Figure 11C), which is consistent with our GC-MS data. Furthermore, the residues Y53, L56, V282, and V284 in the binding pocket of MpPR are predicted to mainly contribute to the non-polar hydrophobic interactions with (+)-menthone and (−)-isomenthone in the MD simulations (Supplementary Figure 11D). To experimentally validate these residues responsible for product stabilization, we performed mutagenesis analysis and assayed the abilities of mutants to convert (−)-pulegone to (+)-menthone and (−)-isomenthone. Compared with the wild-type enzyme, the mutation L56S, Y78D, and V282L decreased the ratio of products (+)-menthone and (−)-isomenthone, like the results obtained using the native substrate (+)-pulegone (Supplementary Figure 12). In contrast, the mutation L56I, L56V, R57A, Y78A, M135A, V282A, and V284A significantly improved the percentage yield of (+)-menthone, suggesting the residues Leu56, Arg57, Tyr78, Met135, Val282, and V284 might be critical to the stereoselectivity of products (+)-menthone and (−)-isomenthone (Supplementary Figure 12B).