Pinpointing a Mechanistic Switch Between Ketoreduction and “Ene” Reduction in Short‐Chain Dehydrogenases/Reductases

Abstract Three enzymes of the Mentha essential oil biosynthetic pathway are highly homologous, namely the ketoreductases (−)‐menthone:(−)‐menthol reductase and (−)‐menthone:(+)‐neomenthol reductase, and the “ene” reductase isopiperitenone reductase. We identified a rare catalytic residue substitution in the last two, and performed comparative crystal structure analyses and residue‐swapping mutagenesis to investigate whether this determines the reaction outcome. The result was a complete loss of native activity and a switch between ene reduction and ketoreduction. This suggests the importance of a catalytic glutamate vs. tyrosine residue in determining the outcome of the reduction of α,β‐unsaturated alkenes, due to the substrate occupying different binding conformations, and possibly also to the relative acidities of the two residues. This simple switch in mechanism by a single amino acid substitution could potentially generate a large number of de novo ene reductases.


General Reagents and Analytical Techniques
All chemicals and solvents were purchased from commercial suppliers, except where specified, and were of analytical grade or better. The synthesis of monoterpenoids 1a-b, 2a-d, 3a-b and 4a-b were performed as described previously. [1] Product standard (R/S)-levodione 6c was synthesised as described previously. [2] All monoterpenoid compounds were dissolved as stock solutions in absolute ethanol, with a final concentration of 2 % (v/v) in the reactions. The concentration of nicotinamide coenzymes (Melford) was determined by the extinction coefficient method (340 = 6220 M -1 cm -1 ). Steady-state kinetic analyses were performed on a Cary UV-50 Bio UV/Vis scanning spectrophotometer using a quartz cuvette (1 mL; Hellma) with a 1 cm path length.
7890A GC system with FID detector. Analyses were performed on a DB-WAX column (30 m; 0.32 mm; 0.25 μm film thickness; JW Scientific) using the monoterpenoid separation method described previously. [3] Unknown products were analysed on an Agilent Technologies 7890B GC system with a 5977A MSD extractor EI source detector using the same DB-WAX column. In this method the injector temperat helium with a flow rate of 3 mLmin -1 and a pressure of 8.3 psi. The program began at 40°C with a hold for 2 min followed by an increase of temperature to 210°C at a rate of 15°C/minute, with a hold at 210 °C for 3 min. The mass spectra fragmentation patterns were entered into the NIST/EPA/NIH 11 (mass spectral library for identification of a potential match. Product enantiomeric excess was determined by analysing reactions by GC using a Chirasil-DEX-CB column (Varian; 25 m, 0.32 mm, 0.25 m). In this method the injector temperature was at 180°C with a split-less injection. The carrier gas was helium with a flow rate of 1 mLmin -1 and a pressure of 5.8 psi. The program began at 70°C with an increase of temperature to 150°C at a rate of 20°C/minute and a hold for 3 min. This was followed by an increase of temperature to 190°C at a rate of 2°C/minute and a hold for 3 min.

Enzyme production and purification
The protein sequences for the following M. piperita enzymes were obtained from UniProt (http://www.uniprot. org): i) MMR (Q5CAF4), ii) MNMR (Q06ZW2) and iii) IPR (Q6WAU1). The respective gene sequences were designed and synthesised by GenScript (USA), incorporating codon optimisation techniques of rare codon removal for optimal expression in E. coli. The genes for MMR and IPR [1] were sub cloned individually into pET21b (Novagen) via NdeI/XhoI restriction sites, without a stop codon, to incorporate a C-terminal His6-tag. In the case of MNMR, a highly expressing construct was produced by sub cloning the gene into pET28b, via NdeI/XhoI restriction sites, to generate a N-and C-terminally His6-tagged protein for characterisation studies. Each construct was transformed into the E. coli strain BL21(DE3)pLysS (Stratagene) for soluble protein over-expression according to the manufacturer's protocol. A general His6-tagged protocol for the production and purification of each individual His6-tagged protein was performed as described previously. [1,3] In the case of IPR, an additional purification step was performed using a Q Sepharose column (GE Healthcare) pre-equilibrated in buffer A (50 mM Tris pH 8.0 containing 1 mM -mercaptoethanol and 10% glycerol). Purified IPR was eluted in a gradient to buffer B (50 mM Tris pH 8.0 containing 1 M NaCl, 1 mM -mercaptoethanol and 10% glycerol). All purified enzymes were dialysed into cryobuffer (10 mM Tris pH 7.0 containing 10 % glycerol), and flash frozen in liquid nitrogen for storage at -80°C. Protein concentration was determined using the Bradford and extinction coefficient methods. [5] Purity was assessed by SDS-PAGE, using 10-12% Mini-PROTEAN® TGX Stain-Free™ gels and Precision Plus protein unstained markers (BioRad) according to the manufacturers instructions.
Additional IPR and MNMR constructs were generated in pETM11 containing only a cleavable Nterminally His6-tagged for protein crystallisation studies. The genes encoding IPR and MNMR were amplified from the pET21b constructs, using primer sets TTTCAGGGCGCCATGGC GGAAGTCCAACGCTATGCTC/ GGTGGTGGTGCTCGATTAATACAGAGCCAGTGCTTTGT CACG and TTTCAGGGCGCCATGGGTGACGAAGTGGTTGTGGATC/GGTGGTGGTGC TCGATCAATACAGACAAAACGCTTCATCG respectively. A modified pET24b vector (pET-M11) was digested with restriction enzymes NcoI and XhoI. The amplified genes were cloned into pET-M11 vector using the InFusion cloning (Clontech) technique, according to the manufacturers protocol. The final constructs contain a N-His6 tag followed by a TEV protease cleavage site.
The constructs were transformed into BL21 (DE3) cells according to the manufacturers instructions. Starter cultures were composed of terrific broth (50 mL) containing kanamycin (15 gmL -1 ) and thr E. coli constructs, and incubated at 37°C for 4 hours. This was diluted into further terrific broth medium (2 L) containing kanamycin (15 gmL -1 ) and was grown at 37°C until the OD600 reached 0.5-0.7. The temperature was reduced to 18°C and recombinant protein expression was induced by the addition of IPTG (0.1 mM). After 15 hours, cells were harvested by centrifuging at 6000g for 10 minutes. The cells were resuspended in lysis buffer (25 mM Tris-HCl pH 8.0) containing 150 mM NaCl, 1X protease inhibitor cocktail (Roche) and DNase (10 μg/ml). The cells were lysed using a sonicator (Bandelin) with a probe set at 40% amplitude with cycles of 10 seconds ON and 20 seconds OFF for 10 minutes. The lysed cells were centrifuged at 40000g for 30 minutes and the supernatant was passed through a 0.2 micron filter and injected onto a 5ml HisTrap column (GE HealthCare). The column was washed with lysis buffer containing up to 40mM imidazole. The enzymes were eluted with a gradient from 40 to 200 mm imidazole.
The proteins were passed through a gravity-flow desalting column (CentriPure P100) equilibrated with lysis buffer and treated with TEV protease at 4°C for 16 hours. To remove the His-tagged TEV protease, samples were loaded onto a 5ml HisTrap column. The flow through from the HisTrap column containing IPR or MNMR without the His tag was concentrated using the VivaSpin centrifugal concentrator (10kDa cutoff) and injected onto a Hiload Superdex 75 26/60 column. The pure fractions from the gel filtration chromatogram peak were concentrated to 15 mg/ml.

Site directed mutagenesis of MNMR and IPR. Variants of MNMR (Y244E) and IPR (E238Y)
were generated by site-directed mutagenesis using the Stratagene QuickChange whole plasmid synthesis protocol. The PCR reaction with IPR was performed using the C-terminally His6-tagged gene in pET21b and the following oligonucleotides: CCGCATTTCGCAGCTTACCG TGTGTCAAAGGCG and CGCCTTTGACACACGGTAAGCTGCGAAATGCGG. For MNMR mutagenesis, both the constructs in pET28b and pETM11 were used with the following oligonucleotides: GCCGCATTTCAGTGCTGAAAAAGTCTCCAAGGCGG and CCGCCTTGG AGACTTTTTCAGCACTGAAATGCGGC. Base substitutions in the oligos are shown in italics. Following template removal by selective restriction digest (DpnI), PCR products (~50 ng) were transformed into the E. coli strain BL21(DE3)pLysS according to the manufacturer's protocol. Each mutant was grown on LB agar containing ampicillin (100 µgmL -1 ; IPR) or kanamycin (34 µgmL -1 ; MNMR) for 24 h at 37°C. Colonies (3) of each mutant were grown and fully sequenced to confirm the presence of the required mutation. Variant enzyme production and purification was performed as for the wild-type enzymes, except no Q-Sepharose step was performed for MNMR. Enzyme kinetics. Standard ketoreductase reactions (1 mL) were performed in KR buffer (50 mM Tris pH 7.0) containing dithiothreitol (1 mM; DTT), NADPH (50 M), monoterpenoid (1 mM) and enzyme (30 nM to 2 M). Reactions were followed by continuously monitoring NADPH oxidation at 340 nm for 1 min at 25C. Standard double bond reductase reactions (1 mL) were performed in DB buffer (50 mM KH2PO4, 12.5 mM K2HPO4 pH 5.5-5.8 for IPR and 50 mM Tris pH 7.0 for MMR/MNMR) containing dithiothreitol (1 mM; DTT), NADPH (100 M), monoterpenoid (1 mM) and enzyme (30 nM to 2 M). Reactions were monitored continuously at 340 nm as described above. To determine the optimal pH for each enzyme in both steady state and biotransformations, reactions were performed in a buffer composed of three buffer salts (12.5 mM tri-sodium citrate, 12.5 mM KH2PO4, 12.5 mM K2HPO4 and 12.5 mM CHES) instead of using Tris buffer. All steady state reactions were performed in at least duplicate.
Biotransformations. Ketoreductase reactions (1.0 mL) with purified enzymes were performed in BT buffer (50 mM Tris pH 7.0) containing the monoterpenoid (5 mM), NADP + (10 μM), glucose (15 mM), glucose dehydrogenase (GDH from Pseudomonas sp.; 10 U; Sigma Aldrich) and enzyme (2-5 M). Reactions were shaken at 30C for 24 h at 130 rpm, and terminated by extraction with ethyl acetate (0.9 mL) containing an internal standard (0.05 % (S)-limonene or 0.1 % secbutylbenzene). All reaction extracts were dried using anhydrous magnesium sulphate, and analysed by GC. Double bond reductase reactions were performed in a similar manner, except the buffer was composed of phosphate buffer (50 mM KH2PO4/K2HPO4 pH 5.5) for IPR and BT buffer (50 mM Tris pH 7.0) for MMR/MNMR. Quantitative analysis was carried out by a comparison of product peak areas to standards of known concentrations. Products were identified by comparison with authentic standards. All biotransformation reactions were performed in at least duplicates, and the results are averages of the data.

Protein crystallography
Crystallisation trials were setup using a Mosquito robot (TTP Labtech) with a drop ratio of 1:1 in MRC 3-drop plates and incubated at 4°C. Eight commercially available screens (Molecular Dimensions and Microlytic) were used for initial screening. IPR crystals were obtained in several conditions in the Morpheus screen. [6] The X-ray data for IPR were collected from crystals grown in well positions C1, D1, E1, F1, and G1. The MNMR crystals were obtained when 200 nL of protein was mixed with 200 nL of mother liquor containing 0.1M imidazole pH 8 and 1M sodium citrate. Before flash cooling crystals with liquid nitrogen, IPR crystals were cryo-protected by washing with mother liquor, while MNMR crystals were cryo-protected by washing with mother liquor supplemented with 20% glycerol. For obtaining protein-ligand complexes, IPR crystals were incubated for 1 hour in mother liquor supplemented with 10mM NADPH and either 25 mM 3a or 25 mM β-cyclocitral. Similarly, MNMR crystals were incubated in mother liquor supplemented with 10 mM NADPH and 20% glycerol.

Structure solution for IPR and MNMR
The X-ray data were collected at Diamond Light Source beamlines I02, I03 and I04. MNMR datasets and IPR bound to β-cyclocitral were processed by automated pipeline implemented in xia2, [7] using xds [8] and xscale. Other IPR datasets were processed manually using DIALS and aimless [9] in the CCP4 suite. [10] The IPR structure bound to NADP + was solved by molecular replacement using SalR structure (PDB 3O26) as the search model in Phaser. [11] The structures of IPR bound to 3a or β-cyclocitral were subsequently solved using difference Fourier methods. The MNMR structures were solved by molecular replacement using the IPR structure as the search model. Automated model building by Autobuild, [12] as implemented in Phenix, [13] built the majority of the residues for both IPR and MNMR. The structures were completed by repeated rounds of manual model building in Coot [14] and refinement using phenix_refine. [15] The structures were validated using Molprobity [16] and PDB_REDO. [17] The crystallograpic data summary and refinement parameters are listed in Table S2. The atomic coordinates and structure factors (codes 5LCX: IPR/NADP; 5LDG: IPR/NADP/3a; 5L4S: IPR/NADP/β-cyclocitral; 5L51: apo-MNMR and 5L53: MNMR/NADP) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (http://www.rcsb.org/).

Supplementary enzyme crystal structure discussion
The unique difference between the structures exists in the "flap-like" domain: SalR's flap domain is composed of 2 short β-strands connected by 3 α-helices whereas the IPR's flap domain is composed of 2 β-strands connected by a single elongated α-helix and a long loop (Figure 1a). Importantly, the IPR's flap domain is twisted inwards, compared to the SalR's flap domain, which aligns the α-helix almost parallel to the β-sheet of the core domain. Surface representation of SalR structure shows it is exposed to the solvent from the nicotinamide end of NADP + , but the rotation of the flap domain in IPR closes this solvent channel ( Figure S7). IPR also shares structural similarities with carbonyl reductases (CBRs) with the core structures being homologous but differ in the region that caps the substrate. In human CBR [18] , there is no flap domain and instead a short loop connects α4 and β4 and leaves the substrate-binding region open at the top ( Figure S8a). In chicken liver CBR [19] a different loop, which connects α5 and β5, extends long and acts as a cap for the substrate ( Figure  S8b). Analysis of IPR homologs (396 sequences) using the ConSurf server [20] revealed the presence of highly conserved residues around the NADP + binding site and most of the residues in the flap domain are highly variable ( Figure S9). Based on the above factors, it is clear that the flap domain in IPR/SalR is species specific. Moreover, it is unlikely to interact with the substrate directly (see below) and might be involved in interaction with partner protein(s) in their respective pathways.  Figure S5. Primary sequence alignment of four SDR enzymes. The three ketoreductases are salutaridine reductase (SALR; UniProt ID: Q071N0) from Papaver somniferum L, [21] and two M. piperita enzymes menthone:(+)-neomenthol reductase (MNMR; UniProt ID: Q06ZW2) and (−)menthone:(−)menthol reductase (MMR; UniProt ID: Q5CAF4). [22] The double bond reductase IPR (isopiperitenone reductase from M. piperita; UniProt ID: Q6WAU1) [23] is separated from the ketoreductases by a dotted line. Conserved residue positions highlighted in blue and green are involved in the binding of the nicotinamide coenzyme and stabilisation of the central -sheet, respectively. Active site residues are shown in red. The sequence alignment was performed by Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo/). The fo-fc densities (blue mesh), contoured at 3σ, are shown for the substrates. This figure was prepared using CCP4mg [24] .

Supplementary Figures
A B Figure S7. Surface representation of SalR and IPR structures. Side views of (A) SalR (PDB 3O26), (B) IPR bound to NADP + and (C) IPR bound to NADP + and 3a. Solvent exposed nicotinamide ring of NADP + in SalR and 3a-bound IPR structures are indicated by arrows. This figure was prepared using CCP4mg [24] . . The long loop, which connects α5 and β5 and caps the active site in chicken carbonyl reductase is shown in red. This figure was prepared using CCP4mg [24] . B A Figure S9. Conserved surface analysis of IPR. ConSurf server [20] was used to map the conserved residues based on 369 IPR homologous sequences. NADP + and 3a are displayed as ball and stick and are coloured green and orange respectively. The variable flap domain is indicated by dotted lines. This figure was prepared using Chimera [25] .   An overlay of the IPR structure with SalR gave an rmsd of 1.04 Å (over 264 residues), indicating high structural similarity. IPR has a core SDR-like structure, with a Rossmann fold domain composed of seven parallel twisted β-strands flanked by four α-helices on the front and three αhelices on the back side (Figure 1a). A SalR-like "flap-like" domain (residues 100-141), mostly predicted in plant reductases, connects α4 with β4 and caps the substrate and cofactor-binding region along with a large loop (residues 265-282) that originates from the core domain. The helix from the flap domain, which caps the active site, has no direct interaction with 3a, as the side chains are at a distance of >3.8 Å from the substrate.
Compound β-cyclocitral binds to the active site in a different orientation compared to 3a, tilted outwards with respect to the nicotinamide ring ( Figure S6b), with a change in the Glu238 side chain conformation. This positions the C=C bond of β-cyclocitral in an orientation inconsistent with hydride transfer (5.42 Å).