Expanding the substrate scope of a bacterial monoterpene synthase for the production of sesquiterpenoid and diterpenoid products

TL;DR

Researchers modified a bacterial enzyme to produce larger terpenoids by changing its active site, enabling it to work with longer substrates.

Contribution

The first successful engineering of a monoterpene synthase to accept diterpene substrates through rational mutations.

Findings

Combinatorial mutations expanded the enzyme's activity to sesquiterpenoids and diterpenoids.

The triple mutant W58A-F74A-F179A showed highest activity with diterpene substrates.

The modified enzyme retained its core 1,6 cyclisation cascade across all substrates.

Abstract

1,8‐cineole synthase from Streptomyces clavuligerus (bCinS) is the only known bacterial terpene synthase that shows exclusive activity towards the monoterpene substrate geranyl diphosphate (GPP; C10). Unlike most plant terpene synthases, bCinS is a high‐fidelity enzyme producing 1,8‐cineole as the predominant product (> 95%). A large number of bulky aromatic residues in the active site steer the carbocationic intermediates down a single path and restrict the conversion of larger prenyl‐diphosphate substrates. Previously, we have shown that a single Phe‐to‐Ala mutation (F74A or F179A) allows bCinS to convert farnesyl diphosphate (FPP; C15) into sesquiterpenoid products, including sesquicineole and germacrene A. Here, we made combinatorial mutations of aromatic active site residues to further expand the substrate scope of bCinS. The F74A‐F179A double variant was not only more active than the wild type but showed increased activity towards FPP over GPP, with sesquicineole and cineole as the main products from these substrates, respectively. Computational active site volume analysis identified an additional residue, W58A, that unlocked activity towards the diterpene substrate geranylgeranyl diphosphate (GGPP; C20), with the W58A‐F74A‐F179A triple variant showing the highest activity on this substrate. Remarkably, these key variants all appear to use the same 1,6 cyclisation cascade to form their main products from GPP, FPP, and GGPP. These results show that even high‐fidelity terpene synthases such as bCinS can be engineered to accept different prenyl‐pyrophosphate substrates without affecting the fundamental reaction cascade.

We have converted the only known true bacterial monoterpene synthase, cineole synthase from Streptomyces clavuligerus (bCinS, C10 substrate), to a highly competent sesquiterpene synthase (C15) with a minimum number of rational mutations. By comparison with diterpene synthases (C20), we were then able to bestow diterpene synthase activity on bCinS. This is the first time the doubling of the substrate length has been reported for a monoterpene synthase.

Figures

In vivo terpenoid production in engineered E. coli and proposed cyclisation pathways for the main products produced by bCinS and previously produced variants. The C5 isoprenoid precursors isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP) are supplied by a heterologous mevalonate (MVA) pathway and converted to geranyl diphosphate (GPP) and farnesyl diphosphate (FPP), the substrates for mono‐ and sesquiterpenoids, by a heterologous GPPS and endogenous FPPS (IspA). The platform can be expanded to diterpenoid production by replacing GPPS with IspAM22 (D2G, C155G), which is capable of producing the C20 substrate geranylgeranyl diphosphate (GGPP).

Expanding the mono‐ and sesquiterpenoid repertoire for bCinS. (A) In vivo product profiles obtained for wt‐bCinS and F74A/F179A single/double variants obtained in the presence of geranyl diphosphate synthase (GPPS) or the geranylgeranyl diphosphate (GGPP)‐producing IspAM22. (B) In vivo product profiles obtained for F74A and N305A single and double variants in the presence of GPPS. Averages of three biological replicates are shown. The main products obtained for each substrate are shown above the bars. Acyclic, cyclic and hydroxylated cyclic monoterpenoids are shaded in green from light to dark, and acyclic, cyclic and hydroxylated cyclic sesquiterpenoids are shaded in orange from light to dark. The data for wt‐bCinS and bCinS‐N305A were taken from Leferink et al. [13] and bCinS‐F74A and bCinS‐F179A from Leferink et al. [14]. Full product profiles are shown in the Tables S3 and S4.

Active site volume analysis of selected bCinS variants. (A) Crystal structure of bCinS (PDB: 5NX7 [10]). The fluoro‐geranyl diphosphate (f‐GPP)/fluoro‐neryl diphosphate (f‐NPP) ligand is shown in yellow sticks, the Mg2+ ions in purple spheres, and the 13 residues forming the active site are highlighted [24]. Structure visualised by pymol [38]. (B) Average active site volumes of various terpene synthases (blue) and the substrate volumes of their main substrate (grey). Error bars show the standard deviation. Averages of 2 mTSs, 58 sqTSs, 35 dTSs and 13 stTSs, taken from ref. [3]. (C) Predicted active site pocket volumes of bCinS 5NX6 and in silico‐generated variants, including the single variants W58A, F74A and F179A, the double variants F74A‐F179A, W58A‐F74A and W58A‐F179A, the triple variants F74A‐F78A‐F179A and W58A‐W74A‐F179A, and the quadruple variant W58A‐F74A‐F78A‐F179A using CastP [25]. All active site pocket volumes detected in our variants of 5NX6 and 5NX7 are summarised in Table S8 in the Supporting Information.

W58A unlocks diterpenoid production in bCinS. Estimated mono‐, sesqui‐ and diterpenoid product titres obtained for the different bCinS W58A‐containing variants in the presence of geranyl diphosphate synthase (GPPS) or IspAM22. Averages of three biological replicates are shown. Acyclic, cyclic and hydroxylated cyclic monoterpenoids are shaded in green from light to dark. Acyclic, cyclic and hydroxylated cyclic sesquiterpenoids are shaded in orange from light to dark. Suspected acyclic, cyclic and hydroxylated cyclic diterpenoids are shaded in blue from light to dark. Full product profiles and standard deviations are shown in Table S5.

Substrate‐specific activities for wt‐bCinS and key variants. (A) k cat values for conversion of geranyl diphosphate (GPP, green) or farnesyl diphosphate (FPP, orange) to all detected mono‐ or sesquiterpenoid products, respectively. wt‐bCinS acts on GPP only. Sesquiterpenoid products were not quantifiable for W58A‐F74A‐F179A. k cat values are given in units of min−1. (B) k cat values for specific conversion of GPP to cineole (blue) or FPP to sesquicineole (pink). W58A‐F74A‐F179A makes linalool (yellow) from GPP, and trace amounts of (E)‐β‐farnesene from FPP (not shown) as the expanded active site cannot force the cyclic conformation necessary for cineole or sesquicineole biosynthesis. Averages are calculated from duplicate experiments and the error bars show the standard deviation.

Molecular docking of wt‐bCinS and in silico variants. (A) Geranyl diphosphate (GPP, green) adopts a precyclic conformation suitable for 1,6 cyclisation (after isomerisation about C2 = C3) when docked into wt‐bCinS (PDB: 5NX6, teal). No suitable conformations are achieved with farnesyl diphosphate (FPP). (B) When docked into bCinS‐F74A, GPP and FPP (orange) both adopt a precyclic conformation (with respect to 1,6 cyclisation) suitable for cineole and sesquicineole synthesis, respectively. The additional isoprene unit of FPP occupies the cavity created by mutation of F74. The mutated residue is highlighted in pink, and the new cavity is illustrated by dots. (C) GPP and FPP again adopt a 1,6 precyclic conformation when docked into bCinS‐F74A‐F179A. (D) In bCinS‐W58A‐F74A‐F179A, the additional cavity created by mutation of W58 allows docking of GGPP in a 1,6 precylic conformation suitable for geranyl‐cineole synthesis. The two Mg2+ ions from 5NX6 are shown as green spheres, and the conformers shown are the top scoring docking poses with the correct pyrophosphate (PPi) orientation (Tables S11–S16). Structures visualised using pymol [38].

Full text

Introduction

Terpenoids are a large group of structurally complex and diverse natural products of considerable industrial importance, for example as pharmaceuticals, cosmetics or biofuels. Terpenoids are synthesised from the universal C_5_ isoprenoid building blocks dimethyl allyl diphosphate (DMAPP) and isopentenyl diphosphate (IPP). DMAPP and IPP are combined in head‐to‐tail condensation reactions by prenyl diphosphate (prenyl‐PP) synthases to generate the acyclic terpenoid precursors geranyl diphosphate (GPP, C_10_), farnesyl diphosphate (FPP, C_15_), geranylgeranyl diphosphate (GGPP, C_20_) and geranyl‐farnesyl diphosphate (GFPP, C_25_), the precursors for mono‐, sesqui‐, di‐ and sesterterpenoids, respectively. The vast structural diversity observed amongst terpenoids is the result of the action of terpene synthases (TSs), which convert the acyclic substrates into structurally complex products via high‐energy carbocation cyclisation reactions [1]. The TS reaction is initiated by ionisation of the substrate, either via metal‐dependent diphosphate abstraction (class I TSs) or protonation (class II TSs). Many TSs are active on a single substrate only. However, increasingly, TSs that act on multiple prenyl‐PP substrates are being recognised. A few notable examples are spata‐13,17‐diene synthase from Streptomyces xinghaiensis, which has sesqui‐, di‐, and sesterterpene synthase activity [2], and VenA from S. venezuelae, a diterpene synthase with promiscuous mono‐ and sesquiterpene synthase activity, which was further engineered to a sesterterpene synthase [3]. Interestingly, many of these multi‐substrate TSs are from bacterial sources [4].

Most bacterial TSs are sesquiterpene synthases (sqTSs) or diterpene synthases (dTSs) [5]. Unlike in plants, where monoterpene synthases (mTSs) are common, to date only a few bacterial mTSs have been identified, with 1,8‐cineole synthase from S. clavuligerus (bCinS) the only known bacterial mTS that shows activity towards GPP only [6]. All other known bacterial TSs with mTS activity show promiscuous activity towards GPP, often resulting in acyclic monoterpenoid by‐products such as β‐myrcene and linalool in addition to their more complex, cyclic sesquiterpenoid major products from FPP [7]. The only other known microbial mTSs with GPP as their main or only substrate are a fungal cineole synthase (HypA from Hypoxylon sp.) [8] and a fungal linalool synthase (Ap.LS from Agrocybe pediades) [9].

Previously, we solved the crystal structure for bCinS, a class I TS [10], and proposed the cyclisation cascade for 1,8‐cineole formation from GPP (Fig. 1). The initial step in the cascade, the metal‐dependent ionisation of GPP resulting in the geranyl cation, is strictly conserved in all class I mTSs. In order to form the first cyclic intermediate, the α‐terpinyl cation, the geranyl cation is required to isomerise to the linalyl cation, which occurs via linalyl‐diphosphate, permitting the subsequent 1,6‐cyclisation. After the formation of the α‐terpinyl cation, nucleophilic attack by water generates an alkyloxonium species, which can undergo a second cyclisation followed by deprotonation to form the final product, 1,8‐cineole (cineole). Cineole is one of the most complex structures that can be formed from GPP, and the shape of the bCinS active site contour and bCinS' overall protein dynamics determine the nature of the product [11]. Many mTSs that catalyse the formation of cyclic products yield mixtures due to a lack of control over the reaction cascade resulting in premature quenching. These enzymes also often display a high degree of functional plasticity, where relatively few amino acid substitutions can sufficiently reshape the active site contour to change the product outcome; this is particularly true for plant mTSs [12]. Unlike most plant mTSs, bCinS is a high‐fidelity enzyme resulting in a product mixture that contains > 95% cineole from GPP. As a result, bCinS shows relatively low levels of functional plasticity, and only a few mutations have been shown to alter the product profile for bCinS. The N305A mutation disrupts the water attack step that incorporates the oxygen atom into the final product, resulting in non‐hydroxylated products redirected from the α‐terpinyl cation [13]. In addition, a single Phe mutation (F74A or F179A) can unlock activity towards FPP in bCinS [14]. Interestingly, in bCinS‐F74A, FPP follows the same cyclisation path as GPP, yielding sesquicineole as the main product from FPP following 1,6‐cyclisation. In bCinS‐F179A, however, FPP follows a distinct cyclisation pathway to GPP, resulting in 1,10‐cyclised sesquiterpenoids like germacrene A from FPP, while still producing cineole from GPP (Fig. 1).

In vivo terpenoid production in engineered E. coli and proposed cyclisation pathways for the main products produced by bCinS and previously produced variants. The C5 isoprenoid precursors isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP) are supplied by a heterologous mevalonate (MVA) pathway and converted to geranyl diphosphate (GPP) and farnesyl diphosphate (FPP), the substrates for mono‐ and sesquiterpenoids, by a heterologous GPPS and endogenous FPPS (IspA). The platform can be expanded to diterpenoid production by replacing GPPS with IspAM22 (D2G, C155G), which is capable of producing the C20 substrate geranylgeranyl diphosphate (GGPP).

Related studies on other TSs have also demonstrated that a single or relatively few mutations allow the conversion of larger prenyl‐PP substrates. For example, mutation of N338 in a cineole synthase from sage (SfCinS) yielded mostly non‐hydroxylated 1,6‐cyclised sesquiterpenoid products. Similar to N305 in bCinS, N338 in SfCinS is implicated in the nucleophilic water attack step [15]. More recently, the dTS for venezuelaene A (VenA) was rationally engineered into a sesterterpene synthase (stTS) following the incorporation of six mutations [3]. Similarly, a systematic mutational analysis of the active site of the sqTS 7‐epi‐α‐eudesmol synthase (SvES) revealed two key residues which, when mutated, can turn SvES into a dTS or even an stTS, an increase of two isoprene units from C_15_ to C_25_ [16]. In light of these recent engineering studies, it appears that the functional plasticity observed for many TSs regarding product outcome also extends to substrate acceptance. In fact, the identification of a functional switch between sqTS and dTS activity identified by using ancestral TS reconstruction [17] suggests that this ‘easy‐switching’ between different prenyl‐PP substrates is a common feature of TS evolution and may be exploited to rationally engineer these enzymes to equip them with new functionalities with relatively few mutations.

In this study, we tested this hypothesis on bCinS, an enzyme that is active on GPP only and shows high fidelity towards 1,8‐cineole, which constitutes > 95% of the product mixture of wild‐type bCinS (wt‐bCinS). Our initial aim was to employ a rational design approach to increase the sesquiterpenoid repertoire of bCinS via introduction of combinations of F74, F179, and N305 mutations. Indeed, combining N305A with F74A yielded non‐hydroxylated sesquiterpenoid products, albeit at low efficiency. A combination of F74A and F179A mutations, however, resulted in a particularly efficient sqTS, producing much higher sesquicineole yields from FPP than previously mutated synthases [18], despite wt‐bCinS not accepting FPP, a gain of function not previously reported for a mTS. Further active site analysis and molecular modelling identified an additional mutation, W58A, that unlocked activity towards GGPP and expanded the product scope of bCinS to include what appear to be new‐to‐nature 1,6‐cyclised diterpenoid products. The most efficient bCinS dTS variant has three mutations and represents a doubling in substrate size acceptance from C_10_ to C_20_, a gain of function not previously achieved for a mTS. The substrate preference of these variants was then tested in vitro using purified enzymes, and the products were rationalised using molecular docking. These results give a striking insight into the malleability of TSs in general and bCinS in particular, a strict mTS once thought to show low levels of functional plasticity. Furthermore, engineering substrate acceptance while exploiting existing reaction chemistries represents a new possible route to high‐value sesqui‐ and diterpenoids, potentially without any significant loss of native catalytic activity.

Results and discussion

Expanding the sesquiterpenoid product scope of bCinS

Previously, we established that bCinS‐F74A produces cineole from GPP and sesquicineole from FPP, utilising the same reaction cascade for both substrates following initial 1,6‐cyclisation. However, where bCinS‐F179A also produces cineole from GPP, it follows a different cyclisation cascade with FPP, resulting predominantly in 1,10‐cyclised sesquiterpenoid products such as germacrene A [14]. In addition, we have shown that N305 is essential for the hydroxylation step in the bCinS reaction cascade: mutation of N305 yields non‐hydroxylated monoterpenoid products redirected from the α‐terpinyl cation [13] (Fig. 1). We hypothesised that a combination of N305A and F74A mutations would result in alternative unhydroxylated sesquiterpenoid products not previously produced by bCinS. Similarly, combining F74A and F179A in a single variant might further expand the sesquiterpenoid product scope of bCinS.

All product profiles presented here were determined using our ‘Plug and Play’ E. coli terpenoid production platform [19], unless stated otherwise (Tables S1 and S2). The platform consists of a refactored heterologous mevalonate (MVA) pathway for the production of the isoprenoid precursors IPP and DMAPP, as well as a truncated GPP synthase (GPPS) from Abies grandis for the synthesis of the monoterpenoid precursor GPP. In addition, the presence of the native E. coli FPP synthase (IspA) results in overproduction of FPP in the presence of the heterologous MVA pathway (Fig. 1). This platform allows for rapid mono‐ and sesquiterpenoid product profile determination via simple analysis of organic culture overlays without the need for protein purification and negates the need for expensive substrates as they are delivered via heterologous and endogenous isoprenoid precursor supply pathways. In the absence of an efficient TS, the endogenous alkaline phosphatase PhoA can convert excess GPP and FPP to geraniol and farnesol, respectively [20, 21, 22]; as such, these products were omitted from all product profile analyses. Details of all products obtained for each variant in this study are shown in Tables S3–S6 and Figs S1–S7 in the Supporting Information available online.

The F74A‐N305A double mutation resulted in a relatively poor catalyst, with total terpenoid titres not exceeding 5 mg per litre organic overlay (mg·L_org_ ^−1^), a 100‐fold reduction compared to wt‐bCinS [13]. However, the variant is capable of producing non‐hydroxylated 1,6‐cyclised sesquiterpenoids, as hypothesised, with β‐sesquiphellandrene the main product from FPP, equivalent to the monoterpenoid β‐phellandrene produced by N305A (Figs 2A and S1). The reduced activity is most likely due to the N305A mutation, which on its own reduces the product titres 12‐fold compared to wt‐bCinS. The F74A‐F179A double variant, on the other hand, resulted in total product titres exceeding that of wt‐bCinS, with cineole as the main product from GPP (43%, 159 mg·L_org_ ^−1^) and sesquicineole the main product from FPP (81%, 227 mg·L_org_ ^−1^) (Figs 2B and S5). Notably, the contribution of the F179A mutation on the product profile is seemingly masked, as the product mixture detected is near identical to the F74A single variant—that is to say, the sesquiterpenoids detected were 1,6 cyclised and not 1,10 cyclised like for the F179A single variant. As both F74A and F179A single variants accept FPP, and the F74A‐F179A double variant is a catalytically competent sqTS, we were intrigued to see if the active site of this double variant was large enough to accommodate the C_20_ diterpenoid substrate GGPP.

Expanding the mono‐ and sesquiterpenoid repertoire for bCinS. (A) In vivo product profiles obtained for wt‐bCinS and F74A/F179A single/double variants obtained in the presence of geranyl diphosphate synthase (GPPS) or the geranylgeranyl diphosphate (GGPP)‐producing IspAM22. (B) In vivo product profiles obtained for F74A and N305A single and double variants in the presence of GPPS. Averages of three biological replicates are shown. The main products obtained for each substrate are shown above the bars. Acyclic, cyclic and hydroxylated cyclic monoterpenoids are shaded in green from light to dark, and acyclic, cyclic and hydroxylated cyclic sesquiterpenoids are shaded in orange from light to dark. The data for wt‐bCinS and bCinS‐N305A were taken from Leferink et al. [13] and bCinS‐F74A and bCinS‐F179A from Leferink et al. [14]. Full product profiles are shown in the Tables S3 and S4.

Adjusting the E. coli terpenoid production platform for diterpenoid production

To allow our terpenoid production platform to produce diterpenoids, the GPPS was replaced with an IspA variant that has GGPP synthase (GGPPS) activity (D2G‐C155G, ‘IspAM22’) [23]. To establish the functionality of the newly introduced GGPPS, diterpenoid production was tested with three previously identified bacterial dTSs from Roseiflexus castenholzii DSM 13941 (RcDTPS), Chryseobacterium sp. CF314 (CsDTPS), and C. luteum (ClDTPS) [7]. In all three cases, multiple terpenoid products with molecular ions (MI) of m/z 272 were detected by GC–MS analysis, suggesting the formation of diterpene hydrocarbon products (molecular formula C_20_H_32_), with estimated titres ranging from 30 to 160 mg L_org_ ^−1^ (Figs S2–S4). In addition, for ClDTPS at least two distinct terpenoid products with an MI of m/z 290 were also detected, suggesting the formation of hydroxylated diterpenoids (molecular formula C_20_H_34_O, see Table S6 in the Supporting Information for full product profiles and titres). No mono‐ or sesquiterpenoid products were detected for these enzymes. The nature of the diterpenoid products was not previously established and could not be determined here from the MS spectra alone. However, confident that the system could produce diterpenoids at sufficient levels, we proceeded with the expression of bCinS‐F74A‐F179A in our GGPP‐producing strain. The products obtained in the presence of IspAM22 were identical to those obtained in the presence of GPPS, albeit at different ratios, but no diterpenoid products were detected (Figs 2 and S5). As this variant still yields high mono‐ and sesquiterpenoid titres in the presence of IspAM22, we hypothesised that the lack of diterpenoid products was most likely caused by an active site that is still too restrictive for GGPP and not a lack of precursor supply. Therefore, we next performed an active site analysis of wt‐bCinS and putative variants to determine if additional mutations could unlock dTS activity.

Active site volume analysis predicts additional mutations for dTS activity in bCinS

Previous studies on TSs have shown that there is a clear correlation between the nature of the active site residues, the size of the active site cavity and the length of the main prenyl‐PP substrate converted. An inverted correlation was found between the van der Waals volumes of a defined set of 13 active site residues and the volume of the main prenyl‐PP substrate converted for a large number of TSs, with bCinS having the bulkiest active site residues of all TSs analysed (924 Å^3^) [24]. In another study, the average active site volume of available TS structures was found to be correlated with the length of the main prenyl‐PP substrate converted, with bCinS displaying the smallest active site volume of all TSs analysed (373 Å^3^) [3]. This is all in accordance with the fact that bCinS is the only known bacterial TS that is strictly active towards GPP. The 13 residues forming the active site in bCinS are highlighted in Fig. 3A. The sum of the van der Waals volumes of these residues in bCinS‐F74A‐F179A is 778 Å^3^, which is closer to the average value for dTSs (776 Å^3^) than sqTSs (855 Å^3^), suggesting that this double variant could accommodate the GGPP substrate (Table S7). However, the active site cavity volume (469 Å^3^), calculated using CastP protein cavity analysis [25], is closer to the average volume of sqTSs (438 Å^3^) than dTSs (526 Å^3^) (Fig. 3B, Table S8). This latter analysis therefore appears to be a better predictor for substrate acceptance in bCinS than the sum of the van der Waals volume of its active site residues. However, combining the information from both methods allowed the prediction of additional bCinS mutations which could potentially unlock diterpenoid synthesis.

Active site volume analysis of selected bCinS variants. (A) Crystal structure of bCinS (PDB: 5NX7 [10]). The fluoro‐geranyl diphosphate (f‐GPP)/fluoro‐neryl diphosphate (f‐NPP) ligand is shown in yellow sticks, the Mg2+ ions in purple spheres, and the 13 residues forming the active site are highlighted [24]. Structure visualised by pymol [38]. (B) Average active site volumes of various terpene synthases (blue) and the substrate volumes of their main substrate (grey). Error bars show the standard deviation. Averages of 2 mTSs, 58 sqTSs, 35 dTSs and 13 stTSs, taken from ref. [3]. (C) Predicted active site pocket volumes of bCinS 5NX6 and in silico‐generated variants, including the single variants W58A, F74A and F179A, the double variants F74A‐F179A, W58A‐F74A and W58A‐F179A, the triple variants F74A‐F78A‐F179A and W58A‐W74A‐F179A, and the quadruple variant W58A‐F74A‐F78A‐F179A using CastP [25]. All active site pocket volumes detected in our variants of 5NX6 and 5NX7 are summarised in Table S8 in the Supporting Information.

From the 13 residues involved in shaping the bCinS active site identified by Hou et al. [24], which include F74 and F179, two additional candidates for mutation in bCinS were identified in W58 and F78. All other residues were disregarded for different reasons: F77 is involved in carbocation stabilisation early on in the reaction cascade [14] and R174 is involved in substrate binding as part of the effector triad [26, 27]. A number of residues are small, and mutation would yield little to no volume increase (V54, G70, G178, and A301). Other residues have no effect on the product profile, as mutation to Ala resulted in variants active on GPP only with cineole as the main product (M180, I216, and N217, see Table S3). From our previous work on bCinS, we know that F78A is inactive, but F78L produces cineole, so F78 likely has a role in active site contouring rather than direct cation‐π interactions with the carbocationic intermediates. W58 was not previously mutated, but mutation of the nearby A301 causes a steric clash with W58, disrupting the active site contour and preventing the formation of the cyclic α‐terpinyl cation (the only product observed was acyclic linalool, in small quantities) [14]. The active site volumes of dTS crystal structures range from 458 to 630 Å^3^ [3]. Active site volume analysis of various bCinS variants suggested that triple (497–510 Å^3^) and quadruple (539 Å^3^) bCinS variants containing a combination of F74A, F78A, F179A, and W58A mutations might have active sites big enough to convert GGPP. The double variants also have active site volumes that fall within the range observed for dTS crystal structures, but where the F74A‐F179A and W58A‐F179A double variants were at the lower end of this range (460–468 Å^3^), the W58A‐F74A double variant has a distinctly larger volume (490 Å^3^, Fig. 3C). Following this analysis, the W58A and F78A mutations were introduced into bCinS as a series of double, triple and quadruple variants with the existing F74A and F179A mutations.

W58A mutation unlocks diterpenoid synthesis in bCinS

As expected, based on our previously reported results on A301F, the W58A mutation on its own resulted in a variant that was relatively poorly active when expressed in the presence of GPPS, with α‐terpineol as the main monoterpenoid product at very low titres (53%, 3.3 mg·L_org_ ^−1^) [13]. Interestingly, unlike the other aromatic substitutions F74A and F179A, no sesquiterpenoid products were detected for W58A in the presence of GPPS, despite a larger predicted active site volume for the latter variant (Figs 4 and S6). The poor activity of W58A and alternative main product confirm the important role of W58 in efficient cineole formation [14]. However, unlike A301F, W58A has retained some cyclisation activity (α‐terpineol), suggesting that the additional space created in W58A is less detrimental to cyclisation than the steric clash in A301F. Addition of a single Phe to Ala mutation yielded the W58A‐F74A and W58A‐F179A double variants, both of which show improved total product titres over W58A alone (7–8 mg·L_org_ ^−1^), which improved even further when expressed in the presence of IspAM22 (18–35 mg·L_org_ ^−1^). The increase in product titres in the presence of IspAM22 is mostly due to increased sesquiterpenoid titres, suggesting that these variants prefer FPP over GPP. Both double variants produce linalool as the main product from GPP along with very small amounts of cyclic monoterpenoids and several cyclic and acyclic sesquiterpenoid products from FPP, including α‐bisabolol, E‐β‐farnesene and *trans‐*nerolidol. Unexpectedly, we observed very small amounts of diterpenoids in these W58A‐containing double variants when expressed in the presence of IspAM22. At least three distinct diterpenoids were produced by both variants: two products containing an MI with m/z 272 (molecular formula C_20_H_32_) and one hydroxylated product containing an MI with m/z 290 (molecular formula C_20_H_34_O), with estimated product titres ranging from 0.5 to 2 mg·L_org_ ^−1^. None of these products are observed in the presence of GPPS, providing additional evidence that these products are indeed diterpenoids (Figs 4 and S6). This was a surprising observation for the W58A‐F179A variant, as the F74A‐F179A variant, which does not show activity towards GGPP, has a slightly larger predicted active site volume. These results suggest that active site shape is as important to product outcome as active site volume in bCinS. Further mutagenesis yielded the W58A‐F74A‐F179A triple variant, which produced the highest diterpenoid titres for a bCinS variant observed so far. The same three distinct diterpenoid products as in the two double variants were observed, with the hydroxylated product as the main product at 3 mg·L_org_ ^−1^ (40% of total). This variant is no longer capable of producing cyclic compounds from GPP, with linalool the only observed monoterpenoid product, along with the same primarily acyclic sesquiterpenoids observed for both W58A‐containing double variants. Similar to the W58A‐containing double variants, the triple variant also results in higher overall product titres in the presence of IspAM22 due to increased sesquiterpenoid production (Figs 4 and S6). The quadruple variant W58A‐F74A‐F78A‐F179A had poor activity, resulting in barely detectable (E)‐β‐farnesene and a single diterpenoid hydrocarbon, at < 1 mg·L_org_ ^−1^ each. Full product profiles and estimated product titres are shown in Tables S3–S5 in the Supporting Information.

W58A unlocks diterpenoid production in bCinS. Estimated mono‐, sesqui‐ and diterpenoid product titres obtained for the different bCinS W58A‐containing variants in the presence of geranyl diphosphate synthase (GPPS) or IspAM22. Averages of three biological replicates are shown. Acyclic, cyclic and hydroxylated cyclic monoterpenoids are shaded in green from light to dark. Acyclic, cyclic and hydroxylated cyclic sesquiterpenoids are shaded in orange from light to dark. Suspected acyclic, cyclic and hydroxylated cyclic diterpenoids are shaded in blue from light to dark. Full product profiles and standard deviations are shown in Table S5.

All F78A‐containing variants were poorly active (Fig. S7). The F74A‐F78A variant did not produce any detectable terpenoid products, similar to F78A, which produces very small amounts of linalool only (0.3 mg·L_org_ ^−1^) [14]. The W58A‐F74A‐F78A variant yields very small amounts of acyclic sesquiterpenoids only (0.3–1 mg·L_org_ ^−1^ each). The F74A‐F78A‐F179A variant also produces sesquiterpenoids only with IspAM22, but this time the main products include the cyclic products sesquicineole (10 mg·L_org_ ^−1^) and α‐bisabolol (4 mg·L_org_ ^−1^), meaning this variant is also capable of producing 1,6‐cyclised products. Both the W58A and F78A single mutations severely disrupt the activity of bCinS, but additional mutation of aromatic residues partially restores activity. However, the presence of the W58A mutation appears essential to obtain bCinS variants capable of producing diterpenoid products from GGPP.

The bCinS 1,6 cyclisation machinery remains intact in the W58A‐containing variants

Product identification for the diterpenoids using GCMS alone is not straightforward. The number of possible structures that can be produced from GGPP is very large, and 1,6‐cyclised diterpenoids are not common in nature, increasing the likelihood that compounds detected are not represented in the NIST database. As such, for none of the three observed diterpenoids produced by our bCinS variants was a definitive match found in the NIST database. However, apart from the observed MI peaks at m/z 272 and 290, there are other ions that hint towards the nature of the products observed.

For example, the dominant ion in the MS spectrum of the hydroxylated diterpenoid product (MI = m/z 290) has an m/z of 139.1, which is otherwise exclusively observed in cineole and sesquicineole, making it likely that this product is geranyl‐cineole, the diterpenoid equivalent of those compounds (Figs S8 and S9). Furthermore, prediction of the fragmentation patterns of cineole, sesquicineole and geranyl‐cineole using the competitive fragmentation modelling for electron ionisation (CFM‐ID) web tool [28] reveals that this ion with m/z 139.1 represents the bicyclic oxygen‐containing ring system characteristic of cineole, so this product is indeed likely to be geranyl‐cineole. The MS spectrum of the diterpene hydrocarbon (MI = m/z 272) produced by all the variants active on GGPP (Table S5) shows similarities to the MS spectra of β‐myrcene and (E)‐β‐farnesene and is most likely β‐springene, the diterpenoid equivalent of those compounds (Figs S8 and S9). Finally, the third diterpene product observed in several of our W58A‐containing variants has the closest match with geranyl‐linalool from the NIST database. However, since this compound has an apparent MI with m/z 272, and geranyl‐linalool would be expected to have an MI with m/z 290, its exact identity is more difficult to determine than the other two major products. The MS spectra obtained using GCMS for the three diterpenoid products compared to their closest match in the NIST database are shown in Fig. S9 in the Supporting Information. Interestingly, not only does the triple variant W58A‐F74A‐F179A result in higher diterpenoid titres, but the proportion of geranyl‐cineole produced from GGPP increases to 40%, up from 13 to 27% produced by the double variants, suggesting the triple variant has a more intact 1,6 cyclisation machinery than the double variants.

To gain more confidence about the identity of these diterpenoid products, the product mixture produced by bCinS‐W58A‐F74A‐F179A (highest dTS activity) was analysed using a more sensitive quadrupole time‐of‐flight (qToF) GC method. From the qToF analysis, a number of additional diterpenoid products were detected with MI of m/z 272 or m/z 290 (Fig. S10), further illustrating the ability of this variant to convert GGPP to diterpenoids. The three peaks with the highest peak intensities are likely the same products that were detected using GCMS. The MS spectra show similarities despite the different ionisation methods, and for the product that likely represents geranyl‐cineole, peaks with m/z 139.1 and m/z 290 are again observed (Figs S9 and S10). The additional products all have a closest match with geranyl‐α‐terpinene (1,6 cyclised) from the NIST database, albeit with low probability scores (~10%), suggesting they represent alternative 1,6 cyclised diterpenoids whose precise identity cannot be confirmed. The diterpenoid titres were too low to isolate them for NMR analysis and, unfortunately, when we scaled up our cultures in an attempt to overcome this, the relative yields were even lower. Positive product identification is also hampered by the fact that acyclic and 1,6 cyclised diterpenoid products are rare in nature. The acyclic diterpene β‐springene is an insect repellent found in the essential oils of various plants [29], and geranyl‐linalool is produced by a geranyl‐linalool synthase from the plant Tripterygium wilfordii [30]. Diterpenoids resulting from 1,6 cyclisation, including geranyl‐α‐terpinene and geranyl‐p‐cymene, have been identified in the essential oils of various plant species [31, 32]. Geranyl‐cineole, however, is to the best of our knowledge a new‐to‐nature diterpenoid product.

Strikingly, if the major diterpenoid compound is geranyl‐cineole, this suggests that the bCinS 1,6 cyclisation machinery remains intact in the diterpenoid‐producing variants as it did in the sesquiterpenoid‐producing double variants, all of which produce a form of cineole from their preferred substrate. This retention of reaction path on larger substrates is unusual for engineered TSs. In most cases, the cyclisation reaction on the larger substrates yields products that show less complexity. For example, wild‐type VenA produces a complex tetracyclic diterpenoid product, whereas the variant with stTS activity yields a monocyclic product [3]. Similarly, SvES yields a bicyclic sesquiterpenoid product, with the variants F77I and F77A showing dTS and stTS activity, but both yielding monocyclic products [16]. A notable exception is Ap.LS, a fungal linalool mTS, where mutation of Y299 yielded variants that not only convert the larger substrate FPP, but are also capable of producing cyclic products from this substrate, whereas the wild‐type enzyme produces acyclic linalool from GPP. Molecular modelling based on the structure of wt‐Ap.LS found that the additional space created by Y299A allowed better binding of FPP in the active site with less torsion strain compared to wt [33]. There is still a change in the carbon scaffold of the major product; however, unlike our variants presented here. These results, and those presented here, suggest that rational engineering of substrate scope in TSs may currently be more predictable than altering product outcome from the same substrate [34] and could present an opportunity to ‘reverse’ the evolution of efficient microbial monoterpene synthases to reobtain sesquiterpene synthase function without affecting the specific reaction pathway.

In vitro kinetic assays reveal substrate‐specific activities for bCinS and variants

All results shown so far were obtained using our in vivo E. coli terpenoid production platform, either in the presence of GPPS or IspAM22, in addition to native IspA. As a result, the exact supply of GPP, FPP and GGPP in each strain was not known. To determine the substrate‐specific activities of the most active variants, these enzymes were purified and incubated with each substrate separately. These variants were: wt‐bCinS, F74A, F74A‐F179A and W58‐F74A‐F179A. All four purified enzymes yielded the same product mixtures as we observed in vivo (Figs S11–S17 and Tables S9 and S10). We then determined the apparent k _ cat _ (henceforth ‘k _ cat _’) of each variant in our given experimental conditions, as described in the methods section. Because the promiscuity of our bCinS variants increases with more mutations, we calculated the k cat values for both main product formation and total substrate turnover; otherwise, the activity of a variant may seem lower if its main product represents a smaller proportion of the total product mixture.

We determined the k cat of wt‐bCinS to be 4.14 ± 0.68 min^−1^ with respect to cineole formation. For the conversion of GPP to all detected products, the k cat was 4.38 ± 0.70 min^−1^, a slightly higher value that includes the small amount of linalool and α‐terpineol also produced (Fig. 5). No in vitro activity was detected with FPP or GGPP for wt‐bCinS.

Substrate‐specific activities for wt‐bCinS and key variants. (A) k cat values for conversion of geranyl diphosphate (GPP, green) or farnesyl diphosphate (FPP, orange) to all detected mono‐ or sesquiterpenoid products, respectively. wt‐bCinS acts on GPP only. Sesquiterpenoid products were not quantifiable for W58A‐F74A‐F179A. k cat values are given in units of min−1. (B) k cat values for specific conversion of GPP to cineole (blue) or FPP to sesquicineole (pink). W58A‐F74A‐F179A makes linalool (yellow) from GPP, and trace amounts of (E)‐β‐farnesene from FPP (not shown) as the expanded active site cannot force the cyclic conformation necessary for cineole or sesquicineole biosynthesis. Averages are calculated from duplicate experiments and the error bars show the standard deviation.

For F74A, we determined the k cat for cineole formation to be 1.24 ± 0.21 min^−1^. For conversion of GPP to all detected products, the k cat was 1.94 ± 0.30 min^−1^. The larger difference between these two values is a reflection of the increased promiscuity of F74A compared to wt‐bCinS: in F74A, cineole represents ~65% of the product mixture compared to 95% in wt‐bCinS (the other two products are linalool and α‐terpineol, like in wt‐bCinS). TSs handle highly reactive intermediates and are known to privilege control of the reaction over catalytic activity [1]. This is typically observed in terms of product distribution but appears also to be the case with substrate acceptance: the k cat for the conversion of FPP to sesquicineole by F74A was 2.68 ± 0.37 min^−1^, and 3.79 ± 0.63 min^−1^ for all detected sesquiterpenoid products. This means F74A has activity on FPP comparable to wt‐bCinS's activity on GPP, a remarkable result considering wt‐bCinS is one of the most efficient bacterial TSs and is entirely incapable of converting FPP in its native form. This suggests bCinS may have sacrificed considerable sesquiterpene synthase activity at some point during its evolutionary history in order to become a high‐fidelity monoterpene synthase. Sesquicineole represents ~73% of the product mixture from FPP for F74A, the other ~27% being α‐bisabolol, the sesquiterpenoid equivalent of α‐terpineol.

F74A‐F179A was an even more efficient synthase: the k cat for cineole formation was 3.03 ± 0.48 min^−1^, and for conversion of GPP to all detected monoterpenoid products was 5.07 ± 0.86 min^−1^, where cineole was ~60% of the product mixture, and the two other products were again linalool and α‐terpineol. Like F74A, this variant is more promiscuous than wt‐bCinS, but its k cat with respect to total turnover of GPP is actually higher than wt‐bCinS. The k cat for conversion of FPP to all sesquiterpenoid products for F74A‐F179A was 5.49 ± 0.10 min^−1^, the highest we observed for any variant on any substrate. Not only that, but F74A‐F179A was a high‐fidelity synthase, producing sesquicineole at ~92% of the product mixture. As such, the k cat for sesquicineole formation was only slightly lower at 5.04 ± 0.07 min^−1^. The slight difference in main product proportions compared to the in vivo assays is due to the extended post‐induction period in our Plug and Play method, which allows the accumulation of small‐quantity by‐products (Table S3) that do not reach detectable levels in these short‐lived kinetics experiments.

We attempted to detect the diterpenoid products that we observed in vivo for W58‐F74A‐F179A, but this was not successful. The low in vivo titres (0.5–3.0 mg·mL^−1^) suggest this variant has only low activity on C_20_ GGPP, so it is perhaps no surprise that we failed to detect these products under kinetic conditions. Furthermore, the additional C_5_ isoprene unit in GGPP caused problems with substrate solubility in our assay set‐up, which likely also contributed to low turnover. With GPP, we detected small quantities of linalool (k cat = 0.24 ± 0.04 min^−1^), and from FPP trace amounts of (E)‐β‐farnesene, whose kinetic value we could not determine. This is consistent with what we observed in vivo, where the enhanced size of the active site in this triple variant hinders the formation of cyclic mono‐ and sesquiterpenoid products, likely because it cannot force these shorter substrates into the necessary precyclic conformation. This becomes apparent during molecular docking discussed later.

These data confirm the results of our in vivo assays and show that F74A and F74A‐F179A are both highly efficient bifunctional mono/sesquiterpene synthases with higher activity on FPP than GPP. Compared to wt‐bCinS, which is incapable of converting FPP, F74A possesses activity on FPP comparable to that of wt‐bCinS on GPP, with only a moderate loss of activity on the latter. F74A‐F179A actually possesses greater activity on GPP than wt‐bCinS and even higher activity on FPP. Most surprisingly, although F74A and F74A‐F179A are more promiscuous than wt‐bCinS, the major products (ranging from 60 to 92% of total) are still cineole (GPP) or sesquicineole (FPP), demonstrating that bCinS's 1,6 cyclisation machinery remains intact in these variants and that these changes to the active site architecture are almost exclusively confined to substrate acceptance. This is striking given that (1) most modifications to terpene synthases result in enzymes with lower overall activity [1] and/or control of reaction path; and (2) wt‐bCinS is such a high‐fidelity enzyme with respect to both substrate acceptance and product outcome. It therefore seems likely that wt‐bCinS evolved from an efficient sesquiterpene/bifunctional synthase into a strict monoterpene synthase by restriction of active site volume, as we have previously speculated [13, 14]. As discussed, genuine bacterial monoterpene synthases are rare, and they bear greater resemblance to bacterial sesquiterpene synthases than they do to monoterpene synthases from plants [1], of which there are many examples [12]. Understanding the evolutionary relationship between TS sequence and substrate acceptance is essential to the engineering of TSs with desired product outcomes. This is especially true in the context of engineered microbes where substrate availability cannot always be quantitatively controlled, and the production of FPP cannot simply be downregulated due to its importance in many cellular processes [35].

Molecular docking of GPP, FPP and GGPP into bCinS and variants

To rationalise the products of our bCinS variants with GPP, FPP and GGPP, and the preferences observed in our in vitro assays, each of the three substrates was docked into wt‐bCinS and in silico‐generated variants (Fig. 6). For the 20 top conformers, we recorded (1) the predicted binding affinity (kcal·mol^−1^); (2) whether the orientation of the PPi moiety was appropriate with respect to the Mg^2+^ cluster and the ‘sensor’ residue [26] R174; and (3) the number of clashes between the ligand and the enzyme, defined as a van der Waals overlap of ≥ 0.4 Å (Tables S11–S16).

Molecular docking of wt‐bCinS and in silico variants. (A) Geranyl diphosphate (GPP, green) adopts a precyclic conformation suitable for 1,6 cyclisation (after isomerisation about C2 = C3) when docked into wt‐bCinS (PDB: 5NX6, teal). No suitable conformations are achieved with farnesyl diphosphate (FPP). (B) When docked into bCinS‐F74A, GPP and FPP (orange) both adopt a precyclic conformation (with respect to 1,6 cyclisation) suitable for cineole and sesquicineole synthesis, respectively. The additional isoprene unit of FPP occupies the cavity created by mutation of F74. The mutated residue is highlighted in pink, and the new cavity is illustrated by dots. (C) GPP and FPP again adopt a 1,6 precyclic conformation when docked into bCinS‐F74A‐F179A. (D) In bCinS‐W58A‐F74A‐F179A, the additional cavity created by mutation of W58 allows docking of GGPP in a 1,6 precylic conformation suitable for geranyl‐cineole synthesis. The two Mg2+ ions from 5NX6 are shown as green spheres, and the conformers shown are the top scoring docking poses with the correct pyrophosphate (PPi) orientation (Tables S11–S16). Structures visualised using pymol [38].

The bCinS crystal structure (PDB: 5NX6) contains the inhibitor fluoro‐neryl‐diphosphate (f‐NPP), a fluorinated, isomerised analogue of GPP, in a precyclic conformation suitable for 1,6 cyclisation [10]. To achieve this true ‘1,6 precyclic’ conformation, GPP must first undergo isomerisation about its C2 = C3 bond, which occurs after PPi abstraction and brings C1 adjacent to C6 (Fig. 1). A ‘1,6 precyclic’ conformation in our docking experiment is therefore defined as one in which the substrate resembles f‐NPP in 5NX6, except for this difference in stereochemistry.

GPP docked into 5NX6 (Fig. 6A) yields a 1,6 precyclic configuration with the correct PPi orientation and zero clashes (Table S11), consistent with wt‐bCinS′ major products derived from the 1,6‐cyclised α‐terpinyl cation. This conformation is essentially identical to that of the f‐NPP ligand from 5NX6, which helps validate the docking procedure (Fig. S18A). When FPP is docked into the same structure, some conformers possess the correct PPi orientation but with a high number of clashes, resulting in a significantly more restricted conformation. This suggests that FPP is a poor substrate, as motion required for chemistry will be impacted, which is in agreement with the fact that wt‐bCinS cannot convert FPP into sesquiterpenoids. Similarly, docking of GGPP into 5NX6 forces the substrate to coil up in the active site, resulting in even more clashes and lower binding affinities (Table S11).

In the in silico‐generated F74A variant, GPP adopts a very similar pose to in wt‐bCinS, with a precyclic conformation suitable for 1,6 cyclisation (Fig. 6B), consistent with the fact that cineole is retained as the major product. However, the space created by the F74A mutation now allows successful binding of FPP. Notably, the geometry of FPP is very similar to GPP from C1 to C10, with the additional C_5_ isoprene unit occupying the cavity created by the removal of the aromatic ring of F74 (Fig. 6B). This is consistent with the formation of sesquicineole, which proceeds by the same 1,6 cyclisation as cineole and may be considered ‘cineole with a prenyl tail’. These conformers have the correct PPi orientation and one (GPP) or zero (FPP) clashes with active site residues (Table S12). Interestingly, the binding energy of FPP was ~1.0 kcal·mol^−1^ more negative than GPP, consistent with the preference for FPP we observed for this variant in our in vitro experiments, although these binding energies serve only as a proxy for substrate affinity. GGPP can dock with the correct PPi orientation but with more clashes, perhaps explaining this variant's inability to process this substrate.

The docking results for F74A‐F179A were very similar to F74A. Both GPP and FPP adopt a precyclic conformation with respect to 1,6 cyclisation (Fig. 6C), with correct PPi orientation and zero clashes (Table S13). The additional space created by the F179A mutation allows FPP to loop back on itself slightly more, bringing C1 even closer to C6 than in F74A, and presumably providing more space for substrate rearrangement. This small change may explain the improved activity of F74A‐F179A compared to F74A, although the cavity created by the F74A mutation is clearly more important to sesquicineole synthesis, as shown by the fact that the F179A single variant makes 1,10 cyclised products such as germacrene A [14] Once again, GGPP can dock with correct PPi orientation but only with a significant number of clashes, consistent with this variant's inability to produce diterpenoids.

W58A‐F74A‐F179A is the only variant in which GGPP can dock with the correct PPi orientation and no clashes with active site residues (Table S14). The cavity created by mutation of W58 accommodates the additional isoprene unit, allowing the earlier part of the hydrocarbon chain to adopt a precyclic conformation with respect to 1,6 cyclisation (Fig. 6D). As such, the geometry from C1 to C10 resembles that of GPP when docked into the other variants, consistent with the formation of 1,6 cyclised diterpenoids like geranyl‐cineole. As hypothesised earlier, the increased size of the active site results in linear conformations of the shorter substrates. FPP can adopt the correct PPi orientation but is no longer precyclic, with its hydrocarbon chain burrowing into the deepened active site (Fig. S18B). With GPP, none of the conformers adopted an orientation similar to NPP in 5NX6: the most feasible conformer is ‘flipped’ with respect to the coordination of its PPi moiety to the metal cluster, and also adopts a linear conformation (Fig. S18B).

Finally, for further validation of the docking protocol, we performed the same analysis on the F179A single variant, which makes 1,10‐cyclised products from FPP when assayed experimentally (Table S15). Gratifyingly, the best docking FPP conformers in this variant adopted a 1,10‐precyclic conformation strongly resembling the germacryl cation instead of the 1,6‐bisabolyl conformations observed for the other variants discussed (Fig. S19). Moreover, the best conformers for GPP were still 1,6‐precyclic (Fig. S20), consistent with the fact that F179A makes cineole (1,6) from GPP, despite making germacrene A (1,10) from FPP. This highlights the ability of the docking protocol not only to capture differences in active site shape between variants but also between substrates for each variant.

Conclusion

bCinS is the only known bacterial TS that is strictly active on the monoterpenoid substrate GPP (C_10_). A large number of bulky aromatic residues constrain the active site, allowing the enzyme to catalyse cineole formation with remarkable precision. Even though this highly constrained active site is essential for the efficient conversion of GPP to cineole, we have shown that mutation of select active site aromatic residues results in bCinS variants that can convert increasingly larger substrates. A single F74A mutation converts bCinS from a strict mTS to a competent sqTS, producing sesquicineole from FPP at high titres, which can be increased by the addition of F179A. As such, F74A and F74A‐F179A exhibited k cat values for FPP turnover similar to or higher than that for GPP turnover in wt‐bCinS. Both enzymes show a marginally reduced ability to convert GPP to cineole, but this is primarily due to their increased promiscuity on this substrate, with F74A‐F179A actually demonstrating improved total turnover of GPP compared to wt‐bCinS. Further addition of the W58A mutation allows the enzyme to convert GGPP (C_20_) to several diterpenoids, one of which is very likely to be geranyl‐cineole, a new‐to‐nature product. A substrate size increase of two isoprenoid units (from C_10_ to C_20_), doubling the prenyl chain length, is a remarkable result for an enzyme that shows no substrate promiscuity in its native form, and is reported here for a monoterpene synthase for the first time. Interestingly, bCinS′ 1,6 cyclisation machinery remains intact on the larger substrates, with F74A and F74A‐F179A predominantly producing cineole from GPP and sesquicineole from FPP, while W58A‐F74A‐F179A appears to produce geranyl‐cineole as its main product from GGPP. This highlights the difference between active site residues involved in controlling reaction paths vs. substrate acceptance, a key distinction that will allow researchers to rationally engineer TSs for altered substrate selection while retaining specific reaction pathways. As activity on the larger substrates increases with additional mutations, promiscuity with GPP and FPP increases, until in the presence of the W58A mutation GPP can no longer be cyclised and FPP is converted predominantly to acyclic products—the increased space in the active site allows these substrates to bind in an extended conformation not suitable for cyclisation. This ‘easy‐switching’ between prenyl‐PP substrates of varying lengths appears to be a common feature in TS evolution, especially for bacterial TSs, and can be exploited to rationally engineer new functionalities with relatively few mutations, even for highly selective, high‐fidelity TSs such as bCinS.

Materials and methods

Chemicals

All terpenoid standards (including β‐myrcene, limonene, 1,8‐cineole, linalool, α‐terpineol, geraniol, cis/trans‐nerolidol and farnesol) were purchased from Sigma‐Aldrich. GPP and FPP were synthesised from geraniol and farnesol, respectively, as described previously [7] for use in initial testing. For kinetic assays, GPP, FPP and GGPP were purchased from Sigma‐Aldrich.

Bacterial strains and media

All E. coli strains were routinely grown in Lysogeny Broth (LB, Formedium) or on LB agar plates including antibiotic supplements as appropriate (ampicillin, 100 μg·mL^−1^; kanamycin, 50 μg·mL^−1^). For site‐directed mutagenesis, cloning, plasmid propagation and terpenoid production, E. coli DH5α cells were used (NEB 5α New England Biolabs, Hitchin, Hertfordshire, UK). The latter was performed in phosphate buffered terrific broth (PBTB, Formedium). Protein overexpression was performed using BL21 (DE3) cells (NEB) in PBTB.

Cloning and site‐directed mutagenesis

The native IspA gene was amplified from the genome of E. coli DH5α (NEB 5α) using primer pair IspA‐D2G‐IF_Fw and IspA‐IF_Rv, and the resulting IspA‐D2G variant was cloned into the pBb_bCinS plasmid backbone linearised with pBbB2a‐IF_Fw and pBbB2a‐IF_Rv using the InFusion protocol (Takara Bio, London, UK) according to the manufacturer's instructions. Introduction of the C155G mutation using IspA‐C155G_Fw and Rv was performed using the QuikChange site‐directed mutagenesis method (Stratagene, Cheadle, Cheshire, UK) according to the manufacturer's instructions and resulted in the final pBb‐IspAM22_bCinS plasmid. Three previously identified bacterial dTSs from Chryseobacterium sp. CF314 (EJL71407), Roseiflexus castenholzii DSM 13941 (ABU58787) and Chryseobacterium luteum [7] (KFE96946) were amplified from their corresponding pET vectors using pET‐IF_Fw/Rv and cloned into the pBb‐IspAM22 backbone linearised with primer pair vector‐IF_Fw/Rv using the InFusion protocol. For protein overexpression, selected bCinS variants were amplified from the pBb plasmid backbone using primer pair pET‐IF_Fw/Rv and cloned into the pETM‐11 backbone linearised with primer pair vector‐IF_Fw/Rv using template pET‐bCinS. Correct insertion of all fragments was confirmed by standard Sanger Sequencing (Eurofins, Ebersberg, Germany). Mutations were introduced into bCinS (WP_003952918) using the QuikChange site‐directed mutagenesis method, using plasmids pBb‐GPPS‐bCinS or pBb‐IspAM22‐bCinS encoding N‐terminally His‐tagged bCinS as templates and a mutagenic primer pair [10]. Correct introduction of all mutations was confirmed by Sanger Sequencing. All oligonucleotides used in this study are shown in Table S1 and all plasmids used in this study are shown in Table S2 of the Supporting Information.

Terpenoid production in E. coli

For mono‐ and sesquiterpenoid production, a pBb‐GPPS_TS plasmid was co‐transformed with plasmid pMVA, and for diterpenoid production, a pBb‐IspAM22_TS plasmid was co‐transformed with pMVA and grown as described [14, 19]. Briefly, expression strains were inoculated in PBTB supplemented with 0.4% glucose in glass screw‐capped vials and induced for 48 h at 30 °C with 50 μm Isopropyl β‐D‐1‐thiogalactopyranoside (IPTG) and 25 nm anhydro‐tetracycline (aTet). A 20% (v/v) n‐nonane layer was added to capture the volatile terpenoid products. After induction, the organic layer was collected, dried over anhydrous MgSO_4_, and mixed at a 1 : 1 ratio with ethyl acetate containing 0.01% (v/v) sec‐butylbenzene as an internal standard. The samples were analysed by GC–MS.

Expression and purification of selected bCinS variants

E. coli BL21(DE3) cells were freshly transformed with a pET‐bCinS (or variant) plasmid (Table S2), and induced with 0.1 mm IPTG for 24 h at 16 °C. The recombinant proteins were purified as described previously [10, 35]. Briefly, the cells were harvested and resuspended in buffer A (25 mm Tris pH 8.0, 150 mm NaCl, 1 mm DTT, 4 mm MgCl_2_ and 5% (v/v) glycerol). The cells were lysed by sonication, and the debris was removed by centrifugation. The supernatant was loaded onto a His‐Trap column (Cytiva, Marlborough, MA, USA) pre‐equilibrated with buffer A containing 25 mm imidazole. The column was washed with buffer A containing 25 mm imidazole, and the proteins were eluted by increasing the imidazole concentration to 200 mm. Fractions containing the purified proteins were pooled, concentrated and desalted using a PD10 column (BioRad, Watford, Hertfordshire, UK) prior to storage at −80 °C.

In vitro enzyme assays

All in vitro reactions with purified enzyme were performed in duplicate in buffer A. The reaction mixtures (0.3 mL) contained 0.25–1 mm GPP, FPP or GGPP and 1.0–1.175 μm purified enzyme, and a 100% (v/v) n‐nonane layer was added to capture the volatile terpenoid products. The vials were incubated at 37 °C with gentle shaking for 5–120 min. 120 min reactions were performed with in‐house substrate to confirm the in vivo product profiles and determine the linear kinetic phase of the reaction. 5 min reactions were then performed with commercial substrate to determine apparent k cat according to k cat = V max/[E]. The reaction was stopped by placing the vials on ice, followed by immediate vortexing for 30 s. The n‐nonane overlay was collected, dried over anhydrous MgSO_4_, and mixed at a 1 : 1 ratio with ethyl acetate containing 0.001% (v/v) sec‐butylbenzene as an internal standard. The samples were analysed by GC–MS.

GC–MS analysis

GC–MS analysis was performed as previously described [7]. Terpenoid‐containing samples were injected onto an Agilent Technologies (Cheadle, Cheshire, UK) 7890B Gas Chromatograph system equipped with an Agilent Technologies 5977A MSD. The terpenoid products were separated on a DB‐WAX column (30 m × 0.32 mm i.d., 0.25 μm film thickness, Agilent Technologies). The injector temperature was set at 240 °C with a split ratio of 20 : 1 (1 μL injection). The carrier gas was helium with a flow rate of 1 mL·min^−1^ and a pressure of 5.1 psi. The oven program used was as follows: 50 °C (1 min hold), ramp to 68 °C at 5 °C min^−1^ (2 min hold), and ramp to 230 °C at 25 °C min^−1^ (2 min hold). The ion source temperature of the mass spectrometer (MS) was set to 230 °C, and spectra were recorded from m/z 50 to m/z 250 for mono‐ and sesquiterpenoids or m/z 50 to m/z 300 for diterpenoids. Compound identification was carried out using authentic standards where available or comparison to reference spectra in the NIST library of MS spectra and fragmentation patterns as described previously [19]. To attempt product identification of unknown diterpenoids, MS fragmentation patterns of suspected compounds were predicted using the online CFM‐ID tool [28] in the absence of a close match in the NIST library. Product titres were estimated using average values of experimentally determined relative response factors of all available terpenoid standards in relation to the internal standard. For more accurate diterpenoid product identification, diterpenoid‐containing samples were also injected onto an Agilent Technologies 7200 quadrupole time‐of‐flight (qToF) mass spectrometer equipped with an Agilent Technologies 5977A MSD. The products were separated on a VF‐5MS capillary column (30 m × 250 μm i.d., 0.25 μm film thickness, Agilent Technologies). The oven program used was as follows: of 50–300 °C with a temperature gradient of 20 °C·min^−1^ and hold for 5 min at 300 °C was used. The injector temperature was set at 250 °C with a split ratio of 10 : 1 (1 μL injection). The carrier gas was helium with a flow rate of 1 mL·min^−1^ and a pressure of 5.1 psi. The ion source temperature of the MS was set to 250 °C, and spectra were recorded from m/z 50 to m/z 450 with electron impact mode (70 eV).

Active site volume prediction

Active site volume prediction was performed on PDB: 5NX6 and 5NX7 using CASTp [25] with a radius probe of 1.4 Å. In silico mutations were introduced using the pymol software package [36, 37, 38]. Active site ligands and waters were included in the analysis.

Molecular docking

Docking of GPP, FPP and GGPP was performed using Autodock Vina version 1.2.3 [39], with input PDBQT files prepared using AutodockTools 1.5.7 [40]. Chain A of the 5NX6 crystal structure, with all non‐protein atoms except the active site Mg^2+^ ions removed, was used for wt‐bCinS, and in silico mutations were introduced using UCSF ChimeraX [41] for the variants. All protein residues were kept rigid, all ligand single bonds were rotatable, and an exhaustiveness of 50 was used to search for 20 conformations. For wt‐bCinS, a search volume of 12 Å per side centred on C8 of inhibitor fluoro‐neryl‐diphosphate (f‐NPP) in the crystal structure was used. For the variants, a slightly larger search volume of 16 Å per side centred on the midpoint between the GPP C8 and W58 CZ2 was used to account for the increased active site volume.

Conflict of interest

The authors declare no conflict of interest.

Author contributions

NGHL conceptualised the study, generated the protein variants and associated in vivo product profiles, and interpreted the data. JNW purified the protein variants, performed the in vitro enzyme assays, and interpreted the kinetics and docking data. LOJ generated the in silico variants, performed the molecular docking, and interpreted the docking data. NGHL and JNW wrote the paper. NSS supervised the project. All authors have given approval to the final version of the manuscript.

Supporting information

Table S1. Oligonucleotides used in this study. Table S2. Plasmids used in this study. Table S3. bCinS mono‐ and sesquiterpenoid product titres obtained in the presence of GPPS. Table S4. bCinS mono‐ and sesquiterpenoid product titres obtained in the presence of IspAM22. Table S5. bCinS diterpenoid product titres obtained in the presence of IspAM22. Table S6. Diterpenoid product titres for the three bacterial dTSs obtained in the presence of IspAM22. Table S7. Van der Waals volume analysis of 13 TS active site residues. Table S8. Active site pocket volumes of wt‐bCinS and in silico‐generated variants. Table S9. In vitro monoterpenoid product profiles for wt‐bCinS and variants incubated with excess GPP for 5 min at 37 °C. Table S10. In vitro sesquiterpenoid product profiles for wt‐bCinS and variants incubated with excess FPP for 5 min at 37 °C. Table S11. Wt‐bCinS docking. Table S12. F74A docking. Table S13. F74A‐F179A docking. Table S14. W58A‐F74A‐F179A docking. Table S15. F179A docking. Table S16. W58A docking. Table S17. W58A‐F74A docking. Fig. S1. GCMS analysis of the terpenoid production strain containing bCinS‐F74A‐N305A. Fig. S2. GCMS analysis of terpenoid production strain containing RsDTPS. Fig. S3. GCMS analysis of terpenoid production strain containing CsDTPS. Fig. S4. GCMS analysis of terpenoid production strain containing ClDTPS. Fig. S5. GCMS analysis of terpenoid production strains containing bCinS‐F74A‐F179A. Fig. S6. GCMS analysis of terpenoid production strains containing bCinS‐W58A‐variants. Fig. S7. GCMS analysis of terpenoid production strains containing bCinS‐F78A‐variants. Fig. S8. Identification of relevant sesquiterpenoids in the absence of authentic standards. Fig. S9. Identification of diterpenoid products in the absence of authentic standards. Fig. S10. GC‐qToF analysis of bCinS‐W58A‐F74A‐F179A produced diterpenoids. Fig. S11. GCMS total ion chromatogram (TIC) for in vitro turnover of GPP by wt‐bCinS. Fig. S12. GCMS total ion chromatogram (TIC) for in vitro turnover of GPP by F74A. Fig. S13. GCMS total ion chromatogram (TIC) for in vitro turnover of FPP by F74A. Fig. S14. GCMS total ion chromatogram (TIC) for in vitro turnover of GPP by F74A‐F179A. Fig. S15. GCMS total ion chromatogram (TIC) for in vitro turnover of FPP by F74A‐F179A. Fig. S16. GCMS total ion chromatogram (TIC) for in vitro turnover of GPP by W58A‐F74A‐F179A. Fig. S17. GCMS total ion chromatogram (TIC) for in vitro turnover of FPP by W58A‐F74A‐F179A. Fig. S18. Molecular docking of wt‐bCinS and in silico variants. Fig. S19. Molecular docking of FPP into F179A. Fig. S20. Molecular docking of GPP and GGPP into F179A.