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.

Abstract: Three enzymes of the Mentha essential oil biosynthetic pathway are highly homologous,n amely the ketoreductases (À)-menthone:(À)-menthol reductase and (À)menthone:(+ +)-neomenthol reductase,and the "ene" reductase isopiperitenone reductase.Weidentified arare 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 acomplete loss of native activity and aswitch between ene reduction and ketoreduction. This suggests the importance of ac atalytic glutamate vs.t yrosine residue in determining the outcome of the reduction of a,b-unsaturated alkenes,due to the substrate occupying different binding conformations,a nd possibly also to the relative acidities of the two residues.T his simple switchinmechanism by asingle amino acid substitution could potentially generate al arge number of de novo ene reductases.
Asequence alignment of the three ketoreductases MMR, MNMR, and salutaridine reductase (SalR;4 5-49 %h omology to Mentha enzymes) from Papaver somniferum Lw ith IPR showed each enzyme contained typical SDR-like motifs, such as those involved in central b-sheet stabilization, and aT GxxxGhG motif ( Figure S5). [4b] Thel atter motif in the Mentha enzymes contains the motif TGxxKGIG,predictive of apreference for NADP(H) over NAD(H). [7] Akey difference in the sequences between the ketoreductases and IPR was asubstitution of the highly conserved catalytic Ty rresidue for Glu (238 in IPR). An further sequence alignment of over 500 SDRs revealed only four additional enzymes had substitutions of the active-site Ty rresidue (results not shown). One of these enzymes was IPR from arelated Mentha sp., which also contained an active-site Glu. Interestingly,t he aldo-keto reductase superfamily contains both ketoreductases (e.g. aldose reductase) and double bond reductases (e.g. D 4-3ketosteroid 5b-reductase) with high sequence homologies. [8] In this case,asubstitution of an active-site His for aG lu residue discriminated between ketoreduction and double bond reduction. [9] Therefore,w ei nvestigated the role of the different catalytic acid residues in IPR (Glu 238) and MNMR (Tyr 244) in the reaction mechanism.
We determined crystal structures of both MNMR and IPR (Figure 1), the latter in combination with NADP + ,alkene 3a, and b-cyclocitral (non-substrate). Crystallographic methodology,d ata refinement statistics,a nd detailed structural descriptions can be found in the Supporting Information (Table S2 and associated discussion). Thecrystal structures of apo-IPR and the 3a-and b-cyclocitral-bound complexes were solved by molecular replacement using the known SalR crystal structure (PDB 3O26;1 .2 and 1.7 r esolution, respectively;T able S2). [10] Thep resence of clear density in the F 0 ÀF c map for NADP + ( Figure 1B)s uggested IPR had   [11] scavenged it from host cells during protein expression. Substrate 3a was bound to the active with the C = Cb ond close to,a nd parallel with, the nicotinamide ring of NADP + , and close (3.19 ) to the site of hydride transfer ( Figure 1B right and Figure S6a). Thec arbonyl oxygen atom of 3a hydrogen bonds with Glu 238 and the highly conserved Ser 182 and sits at an equal distance (3.15 ) from both residues.Awater molecule hydrogen bonds with Glu 238 and the ribose ring of NADP + ,s uggesting am echanistic role for this water molecule.Conserved residue Asn 154 is hydrogenbonded to Glu 238, while Lys242 forms hydrogen bonds to the ribose of NADP + and awater molecule,indicating its role in stabilizing NADP + and contributing to ap roton relay. [3] The IPR-b-cyclocitralc o-crystal structure shows the ligand binds in an on-active conformation compared to 3a binding (Figure S6b). No other major changes in residue positions were observed in the co-crystal structures. TheM NMR structure was solved by molecular replacement using IPR as the search model (resolution 2.3-2.7 ; Table S2), and was found to be structurally similar (rmsd 0.97 ; Figure 1Bleft). Acoenzyme-bound MNMR structure was obtained by soaking crystals with NADP + ,h owever no structures were obtained with 1a,b within the active site. Major structural differences were not observed between apoprotein and NADP + -bound forms.A dditional discussion on the crystal structures of the Mentha enzymes and related proteins is found in the Supporting Information (Figures S7-S9).
As expected, the conserved Glu 238 of IPR occupied the position of Ty r244 in MNMR, with adistance between Cb of 3a and the NADPH hydride of 3.18 i nt he co-crystal structure.T herefore the bulkier MNMR Ty r244 likely positions substrates in adifferent conformation compared to that observed for 3a in IPR ( Figure 1B right) because of the larger side-chain bulk of tyrosine.T his is consistent with the helix of the flap domain (MNMR) being shifted compared to that in IPR ( Figure 1B right), to accommodate binding of 1a,b.T his structural comparison suggests that this rare residue substitution might be responsible for the switch in activity seen for IPR to NADPH-dependent 1,4 conjugate reduction of the a,b-unsaturated carbonyl compound 3a to 4a.
Based on prior mechanistic studies and our structural studies,w ep ropose mechanisms of action for both ketoreduction (MNMR) and double bond reduction (IPR) in SDRs. [3,4,7,12] Ketoreduction follows typically an ordered "bibi" mechanism, where the coenzyme binds first and leaves last. MNMR appears to follow this classical SDR ketoreduction mechanism for 1a to 2b and 1b to 2d (Scheme 2A). [1a] Thea lcohol product is formed by the transfer of ah ydride from NADPH to the carbonyl carbon atom of the substrate with facial selectivity.I nt he case of SDRs,t he 4pro-S hydride is transferred, in contrast to MDRs that catalyze 4-pro-R hydride transfer. [4a] Concurrent with hydride attack, the carbonyl oxygen atom takes ap roton from the conserved Ty r244 residue acting as acatalytic acid. This starts ac ascade of proton transfers through the NADP + coenzyme and Lys248, terminating with removal of ap roton from aw ater molecule.T he conserved Ser 188 residue likely functions to stabilize the substrate,w hile Lys248 hydrogen bonds with the nicotinamide ribose moiety,lowering the pK a of the Ty r244-OH to promote proton transfer. [3] Residue Asn 160 in SDRs interacts with the conserved Lys248 and bulk solvent via water molecule(s), forming aprotein relay or hydrogen-bonded solvent network (Scheme 2A). This likely helps to stabilize the position of Lys248, thereby assisting the overall ketoreduction mechanism. [3] Thes tructure of the IPR-3a co-crystal reveals that Glu 238 positions the substrate to allow hydride addition at the C = Cbond of 3a,rather than the carbonyl carbon atom. In the proposed IPR double bond reduction mechanism, hydride transfer from NADPH to the 4-position of the a,b-unsaturated carbonyl system of 3a results in formation of the respective enolate ion (Scheme 2B), which then accepts ap roton from the conserved residue Glu 238 to generate the more stable enol. Residue Glu 238 abstracts aproton from an earby water molecule that may initiate as imilar proton transfer cascade to that seen in MNMR. Formation of cisisopulegone 4a then proceeds by Glu 238 abstracting the proton, previously donated to the substrate,r esulting in reformation of the carbonyl group.A lternatively an onenzymatic water-mediated step may occur. Concomitantly,t he enolate double bond accepts aproton from water, giving the 1,4 conjugate reduction product (Scheme 2B). This mechanism is possible in IPR as the side chain of Glu 238, unlike the Ty rs ide chain, readily dissociates to its conjugate base in water. To test this hypothesis further,w eg enerated the variants IPR E238Y and MNMR Y244E and performed biotransformation reactions to detect ketoreduction and/or double bond reduction (Table 2). We tested IPR E238Y at pH 6.0, consistent with the preference for lower pH values of the wild-type enzymes,i na ddition to reactions at pH 7.0 for comparison with the MNMR Y244E variant. IPR E238Y showed no double bond reduction with any substrate tested (3a,b and 5a-d), however it performed minor ketoreduction with substrate 3a to form the equivalent alcohol products 8a ( Table 2, entries 1and 2). Additionally it showed MNMR-like activity towards Mentha compounds 1a,b,f orming primarily 2b and 2d,r espectively ( Table 2, entries 3-6), although the product yields and enantiopurity were lower than with wildtype MNMR. Interestingly,r eactions with 1b at pH 7.0 generated as lightly higher yield of products,b ut they were obtained in near racemic form ( Table 2, entry 6. Therefore, replacing of active-site Glu by Tyrhas converted the enzyme from an ene reductase into aketoreductase,albeit with lower catalytic efficiencya nd enantiospecificity. In the case of MNMR variant Y244E, ketoreduction was not seen with any substrate tested (1a,b, 3a,b,a nd 5a-d).
Minor double bond reduction was detected with substrate 5c to form 6c(  Figure S4), suggesting further mutations are required to form amore active ene reductase.
Interestingly,s tudies with mechanistically different enzymes of the class Ia ldolase family (transaldolase and fructose-6-phosphate aldolase) have shown that the change of the nature of the catalytic acid/base can have as ignificant effect on the reaction mechanism. [14,15] However,the effect of active-site spacial changes by residue substitution needs to be considered. Fore xample the lack of ketoreduction of wildtype IPR with 3a and 3b may be due to ap reference for binding in ac onformation consistent with double bond reduction, while the steric bulk of Ty ri nI PR variant E238Y may orient the substrate in ap osition suitable for ketoreduction. Further studies will be needed to determine the relative contribution of catalytic residue type vs.s teric constraints in determining the overall mechanism of the catalysis.
We have pinpointed asimple mechanistic switch between ene-reductase and ketoreduction activity in the SDR superfamily.T his simple mechanistic switch, in addition to other residue substitutions to improve catalytic efficiency, could potentially transform SDR ketoreductases into novel ene reductases and provide attractive routes to novel enereductase catalysts.T his would reduce the dependence on traditional FMN-containing OYEs for the biocatalytic reduction of a,b-unsaturated alkenes and complications (reaction rates,y ields,a nd product enantiopurity) that arise when OYEs are affected by molecular oxygen. [13] Access to an ew class of ene reductases would open up the possibility of developing new catalytic specificities typical of the SDR superfamily for the reduction of a,b-unsaturated alkenes.   1a,b, 3a,b,and 5a-d;5mm), enzyme (5 mm or 10 mmfor IPR and MNMR, respectively), NADP + (10 mm), glucose (15 mm), GDH (10 U), and enzyme (2 mm). The reaction solutions were agitated at 25 8 8C for 24 ha t130 rpm. Product identification was performed by both comparingretention times with authentics tandards and identification by GCMS on aDB-WAX column (only GCMS identification for product 8a). Figure S10 gives the GCMS spectra traces of the additionalproducts and their respective substrates.
[b] Product yield and enantiomeric excess were determined by GC analysis using DB-WAX and Chirasil-DEX-CB columns, respectively. nd = not determined due to low product yield.