Sesquiterpene Synthase‐Catalysed Formation of a New Medium‐Sized Cyclic Terpenoid Ether from Farnesyl Diphosphate Analogues

Abstract Terpene synthases catalyse the first step in the conversion of prenyl diphosphates to terpenoids. They act as templates for their substrates to generate a reactive conformation, from which a Mg2+‐dependent reaction creates a carbocation–PPi ion pair that undergoes a series of rearrangements and (de)protonations to give the final terpene product. This tight conformational control was exploited for the (R)‐germacrene A synthase– and germacradien‐4‐ol synthase–catalysed formation of a medium‐sized cyclic terpenoid ether from substrates containing nucleophilic functional groups. Farnesyl diphosphate analogues with a 10,11‐epoxide or an allylic alcohol were efficiently converted to a 11‐membered cyclic terpenoid ether that was characterised by HRMS and NMR spectroscopic analyses. Further experiments showed that other sesquiterpene synthases, including aristolochene synthase, δ‐cadinene synthase and amorphadiene synthase, yielded this novel terpenoid from the same substrate analogues. This work illustrates the potential of terpene synthases for the efficient generation of structurally and functionally novel medium‐sized terpene ethers.

Te rpene synthases catalyse the first step in the conversion of prenyl diphosphatest ot erpenoids. They act as templates for their substrates to generateareactive conformation,f rom which aM g 2 + -dependent reaction createsacarbocation-PP i ion pairt hat undergoes as eries of rearrangements and (de)protonations to give the final terpene product.T his tight conformational control was exploited for the (R)-germacreneA synthase-and germacradien-4-ol synthase-catalysed formation of am edium-sized cyclic terpenoid ether from substrates containing nucleophilic functional groups. Farnesyl diphosphate analogues with a1 0,11-epoxide or an allylic alcohol were efficiently converted to a1 1-membered cyclic terpenoid ether that was characterised by HRMS and NMR spectroscopica nalyses. Furthere xperiments showed that other sesquiterpene synthases, including aristolochenes ynthase, d-cadinene synthase and amorphadienes ynthase,y ieldedt his novel terpenoid from the same substrate analogues.T hisw ork illustrates the potential of terpene synthases for the efficient generation of structurally andf unctionally novel medium-sized terpene ethers.
Te rpenoids are the most diversec lass of natural productsa nd possess astounding complexity in structure and biological activity.T hey all arise from as mall pool of isoprenoid diphosphate precursors. Amongst the tens of thousands of known terpenoids are primary metabolites such as carotenoids and ubiquinones as well as al arge array of secondary metabolites that act, for example, as semiochemicals,p heromones,a ntioxidants, phytoalexins and cytotoxins. [1] The structural complexity of thesec ompounds presentsaformidable challenge for organic chemistry,a nd the synthetic generation of terpenoid variants is often difficult. For example, most active analogues of the antimalarial and anticancer drugs artemisinin [2] and paclitaxel [3] only have modificationsa tt he lactone group or the amino acid side chain, as other changes are synthetically difficult to introduce. Moreover,t erpene hydrocarbons are often heat sensitive and unstable under acidic conditions, and their chemicals ynthesis is cumbersome. [1] In the past, great efforts have been made to overcome the challenging total syntheses of such complex ring systems, but the synthesis of terpenoids does not generally compete with direct extractionf rom natural sources. [4] An alternative approach to the synthesis of terpenederived products exploits ad etailed understandingo ft erpene chemistry [5] andu ses biotransformations to convert substrate analogues to functionalised terpenoids.
Te rpene synthases are divided into two classes depending on the pathway used to form the initial carbocation. Class I synthases use the Mg 2 + -binding motif to ionise the substrate and form an allyl cation. In contrast, class II synthases form the initial carbocation by protonation of the distal doubleb ond or an epoxided erivative thereof. [5a] Class Iterpene synthases comprise am ostly hydrophobic active site,s urrounded by an ahelical barrel with two Mg 2 + -binding motifs at the entrance. [5a] Al arge body of work on probingt he chemical steps by using substratea nalogues, [6] mutagenesis, [7] putative reaction intermediates [8] and X-ray crystallography [5a, 9] combined with computational approaches [5c, 10] has provided ad etailedp ictureo f the mechanisms, by which these enzymes catalyse their reactions. [5a, b] Co-crystal structures of aristolochene synthasew ith variouss ubstrate analogues, Mg 2 + and PP i , [9c] in conjunction with molecular modelling, [10a] revealed the detailed physical steps that lead to the generation of the reactive Michaelis complex. The synthase first binds one Mg 2 + ion andt hen farnesyl diphosphate (FDP, 1); two more Mg 2 + ions follow,a nd this closes the active site, forming the Michaelis complex. [10a] Coordination of the diphosphate by the Mg 2 + ions triggerst he generation of af arnesyl cation (2)w hich is chaperoned by the active-site contour through as eries of electrophilic ring closures and rearrangements. Quenching of the final carbocation either by proton loss or by nucleophilic capture generates the terpenoid product. [5a, b] The active-site template steers this reaction cascade, distinguishing it from many potentially competing, similar energy pathways with exquisite precision.After catalysis of the initial CÀOb ond breakage, the role of the enzyme in these chemical steps appears to be largely to act as a template that steers the substrate through as eries of reactive conformations within an optimised electrostatic environment. [5b, c, 10b] The substrate adopts as pecific conformationi n the active site that directso ri nhibits site-selective nucleophilic attack by solvent water.
Harnessing the remarkable catalytic power of terpenes ynthases to generate terpenoids and evolvingthem in apredictable and tunable manner can open the door to an expansion of the terpenome.E xploring how terpene synthases control or inhibit the nucleophilic captureo fc arbocation intermediates has recently become af ocus of our research. We have explored how native germacradien-4-ol synthase from Streptomycesc itricolor (GdolS) mediates the nucleophilic capture of the final carbocation by water and have converted a d-cadinene synthase (DCS) from Gossypium arboreum into aG dolS by targeteds itedirectedm utagenesis;h ere as eries of loop movements at the active site allowed water ingress to capture the carbocation rather than making use of ab ound water molecule in the closed form of the active site. [9b, 11] Moreover,w eh ave investigated how FDP analogues containing nucleophilic functional groups are converted by some sesquiterpene synthases;h ence 12-OH-FDP was converted directly to dihydroartemisinic aldehyde by the amorphadienes ynthase( ADS) from Artemisia annua leading to the most concisea rtemisinin synthesis known and demonstrating that hydroxylated FDPs can act as efficient substrates for terpene synthases. [12] In the past, we have used (S)-germacrene Ds ynthase (GDS) and (R)-germacrene As ynthases (GAS;b oth from Solidago canadensis)t og enerate ap ool of modified germacrenes [13] from ar ange of fluorinated andm ethylated FDP analogues. One of these products is ap otent attractant of grain aphids and ap otentialc rop protection agent. [14] The efficiency of these transformations could be improved dramatically by using segmented-flow methods. [13,15] Using the understandingo ft erpene synthase chemistry gained from this and others' previous work, we surmised that this could be done by introducing an ucleophile into a substrate analoguet oi nterceptaknown carbocationic configuration generated through the templating effect of the enzyme.Asimple example to illustrate the proof of concept would be to intercept the germacryl cation that is formed initially by 1,10-terpene cyclases (i.e.,t hose that catalyse an initial 1,10 ring closure of the FDP substrate, vide infra).
Herein, we describe the use of two 1,10-cyclases, namely GdolS and GAS to synthesise an unnatural cyclic ether terpenoid from two synthetic oxygenated FDP analogues. The former was chosen as an example of this chemoenzymatic intramolecular capturea si ti sk nownn aturally to use nucleophilic quenching to capturethe final carbocation duringits catalytic cycle. The latter is ah ydrocarbons ynthase to exemplify that this can also be done by both types of sesquiterpenesynthase. Both are mechanistically simple with no subsequent cyclisations of the 1,10-cyclisation product (Scheme 1) and hence keep other potential active-site variables to am inimum.W e also show that severalo ther sesquiterpene synthases are capable of converting these substrates to the same cyclic ether.
Allylic alcohol-and epoxide-containing FDP analogues 7 and 8 were chosen as suitable substrates for this investigation as it was hypothesised that they would efficiently be converted to cyclic ether 10 (Scheme 2). During GAS and GdolS catalysis (Scheme 1), the initially formed farnesyl cation 2 is attacked by the C10=C11d ouble bond to generate germacryl cation 3, [16] so that the alcohol oxygen atom in 7 or the epoxide in 8 (Scheme 2) is ideally placed to interceptt he carbocation in intermediate 9 or 11,r espectively.P roton loss from 9 would then furnish cyclic ether 10.S imilarly,c yclisation of epoxy cation 11 should lead to carbocation 12.D eprotonation from C12 of 12,a si ss een with GAS, [17] should yield 10.A lthough this was considered the most likely outcome, alternative products such as 13, 14,o r15 (Scheme 2) had to be ruled out experimentally.C leavageo ft he C10ÀOb ond following nucleophilic attack at C1 with final proton loss would give the 12membered cyclic ether 15;a lternatively, a[ 1,2]-methyl shift from C11f ollowed by protonloss might result in 13 or 14.
Synthetic cDNAs for GAS and GdolS were overexpressed in Escherichia coli. The resulting enzymes were purified as previously described [9b, 18] and incubatedo na na nalytical scale with synthetically produced 7 and 8 (for synthetic details see Section S2 in the Supporting Information). The pentane-extractable products were analysed by GC-MS (Figure1 and Support-Scheme1.Reactions catalysed by GAS and GdolS with the natural substrate 1. Scheme2.Proposed reaction mechanism for the GAS-and GdolS-catalysed conversions of 7 and 8 to the medium-sized terpene ether 10. ChemBioChem 2018ChemBioChem , 19,1834ChemBioChem -1838 www.chembiochem.org 2018 The Authors. Publishedb yWiley-VCH Verlag GmbH &Co. KGaA, Weinheim ing Information). From the allylic alcohol 7,G dolS generated a single product with am olecular ion of m/z 220 (Figures S1 and S2). GAS produced the same product as only 50 %o ft he pentane-extractable products with the remaining 50 %c omprising am ixture of unidentified products ( Figure S3). Incubationo f epoxide 8 with GdolS and GAS also gave the same major compound. However,aminor second product was formed by GdolS ( Figure 1). The major product for both enzymess howed the same fragmentation pattern (Figures S4 and S5) in the mass spectrum, and co-elution of all four product mixtures confirmed an identicalm ajor product for each reaction (Figure S3). The major product was characteriseda sc ompound 10 (vide infra). The minor product ( % 10 %) produced by GdolS from 8 was identified by GC-MS as the epoxide derivativeo f (E)-b-farnesene 16 after comparison with as ample generated by conversion of 8 with (E)-b-farnesene synthase from Mentha x piperita (EBFS;F igure 1). [19] As both analogues proved to be substrates, other sesquiterpenes ynthases were tested. Aristolochenes ynthase from Penicillium roqueforti produced compound 10 as well as an unknown side product from both 7 and 8.O nt he other hand, DCS was only able to turn over 8 to 10,w hereas ADS only accepted 7 to produce 10 ( Figure S8). To estimate the efficiency with which GdolS turned over 8,c ompetitivei ncubations of 8 and 1 with GdolS were performed and analysed by GC-MSt of ollow the relative production of 6 and 10.I ncubations of 3 mm enzymew ith0 .125 mm 8 and 25 mm FDP led to similart urnovers of the two substrates. This result suggestst hat turnover for the conversion of 8 by GdolS is about2 0% of that of the natural substrate 1.N oc onversion was detectedf or any substrate in negative controlsi nw hich enzymew as absent.M oreover,t od emonstratet hat this is a specific templating effect of terpene synthases, 8 was incubated with alkaline phosphatase from bovine intestinal mucosa (Sigma-Aldrich)a sapositive control experiment.N of ormation of any cyclised product was observedw ith only 10,11-epoxy farnesol, that is, simple hydrolysis product was detectedi nt he pentane-extractable products only (Figures S9-S10).
To confirmt he structure of the major product 10,ap reparative-scalei ncubation of 8 with GdolS was performed. Ac olourless oil was isolated from the incubation of 60 mg of 8 (ESI) in 41 %y ield, which is at ypical value for naturalt erpenoids produced from FDP in batch processes. [13] Unoptimised segmented-flow procedures, which improvet he extraction of the hydrophobic product from the aqueous phase, improvedt he yield to 75 %. [13,15] Both substrates (7 and 8)w ereprepared and used in racemic form, and the products were analysed by GC on ac hiral stationary-phase and found to be am ixture of enantiomers with ar atio of 48:52 ( Figures S6 and S7). The slight deviation from ar acemate in the batch-generated product arises from different conversion rates of the two enantiomers, as one might expect. This was reinforcedw hen performed in flow ( Figure S11). The reaction in flow was performed over 1h,a nd ac learly faster turnover of one enantiomer occurred, whereas the batch incubation took severald ays and led to loss of the initial enantioselectivity seen in flow. Initial 1 HNMR spectroscopica nalysis of the batch product at room temperature in CDCl 3 showed broad,p oorly defined signals in some areas of the spectrum;t his hindered af ull assignment of the spectrum (Figure 2). Slow exchange on the NMR timescale at room temperature has been observed previously for similarm edium-sized ring systems; [20] hence variable-tem-peratureN MR spectra were measured between À50 8Ca nd + 50 8C. The structure was successfully elucidated at + 50 8C. Full assignmentso ft he 1 Ha nd 13 CNMR spectra are given in the Ta ble S1, they confirmed 10 as the major product generated by GdolS from 8.
At À50 8C, two conformations were apparent:f or example, the signala td H = 5.32 ppm, corresponding to the proton on C2, split into two resonances at d H = 5. 19 and 5.48 ppm (Figure 2) in an approximately 2:1r atio. Resonances for the minor conformation of 10 also appeared at À50 8Ci na ll other  regions of the spectrum,p articularly clearly for the protons on C1 and C10 (d H = 3.2-4.0 ppm) and in the alkyl region. A NOESY spectruma tÀ50 8Cs howed several distinct NOEs ( Figure 2a nd the Supporting Information), thus allowing some conformational restraints to be applied. Close proximity was apparent between the protons on C2 and C6, C13 and C15, and C10 and C12. Hence, the major conformation of (R)-10 is the down-downf orm, in which C14 and C15 are on the same side of the ring system( Figure 3), with (S)-10 being the mirror image. NOEs were not detectable for the minor isomer;h owever,t he most significant changes in the 1 HNMR spectrum, as temperature decreased, corresponded to the protons on C1 and C2, thus suggesting that the minor conformer is the updown conformation ( Figure 3), in which the methyl groups are on opposite sides of the ten-membered ring. These two alkyl groups change position most during the conformational transition.
The application of biologically active molecules is often dependento nt he availability of an efficient and economic synthesis for their production.T erpene synthases can potentially play key roles in the synthesis of blockbusting terpenoids and are already employed in engineered fermentation platforms. [21] However,s uch systems are still largely limited to the production of natural terpenoids as they rely on the extended metabolic pathways for FDP synthesis from simple primary metabolites such as acetyl-CoA. [1a] The goal to develop biotechnological platforms to generate novel oxygenatedt erpenoids in a programmable andb espoke manner relies on ad emonstration that terpene synthases can generate novel products in vitro. Although there are two examples in the literature in which unnatural heterocycles have been generated by terpene synthases, [22] ando fc ourse many examples of abortive products generated in mechanistic investigations of substrate analogues, [5][6][7] our work shows for the first time that an ovel terpenoidh eterocycle can be generated by knowledge-based design and conversion of as ubstrate analogue,t hereby openingu pt he use of terpenes ynthases to the preparation of complexc hiral organic compounds with potentially novel activities.