Active Site Loop Engineering Abolishes Water Capture in Hydroxylating Sesquiterpene Synthases

Terpene synthases (TS) catalyze complex reactions to produce a diverse array of terpene skeletons from linear isoprenyl diphosphates. Patchoulol synthase (PTS) from Pogostemon cablin converts farnesyl diphosphate into patchoulol. Using simulation-guided engineering, we obtained PTS variants that eliminate water capture. Further, we demonstrate that modifying the structurally conserved Hα-1 loop also reduces hydroxylation in PTS, as well as in germacradiene-11-ol synthase (Gd11olS), leading to cyclic neutral intermediates as products, including α-bulnesene (PTS) and isolepidozene (Gd11olS). Hα-1 loop modification could be a general strategy for engineering sesquiterpene synthases to produce complex cyclic hydrocarbons without the need for structure determination or modeling.

−4 In the structurally conserved Class I TSs, the diphosphate (PPi) of the substrate is coordinated by three Mg 2+ ions, with the hydrocarbon chain residing in a hydrophobic active site cavity. 1,5−12 While much progress has been made to decipher the biochemical details of (sesqui)TSs, altering the water capture behavior with targeted engineering remains challenging. 13nderstanding carbocation management in TSs would enable targeted engineering to obtain specific (novel) terpene products.
Patchoulol synthase (PTS) is a Class I (sesqui)TS that produces patchoulol as the main product upon incubation with FDP. 14 PTS is the key enzyme in patchouli oil biosynthesis, 14,15 a widely appreciated natural fragrance commonly used in the cosmetics industry.The common non-hydroxylated PTS side-product α-bulnesene may have potential as an antiplatelet aggregation agent (antagonist of PAF, the plateletactivating factor). 16Here, we report a combined computational and experimental approach to gain insights into patchoulol biosynthesis.We further demonstrate a strategy to engineer hydroxylating (sesqui)TSs to form complex non-hydroxylated (sesqui)terpenes: Hα-1 loop modification.A four-residue replacement leads to the formation of primarily isolepidozene (24) in germacradiene-11-ol synthase (Gd11olS) and a mixture of α-bulnesene (4) and germacrene A (3) in PTS.
First, we characterized the product profile and kinetics of wild-type PTS (PTS WT ) from P. cablin (details in Supporting Information).GCMS analysis indicated formation of patchoulol as major product (2, 60%), alongside several sesquiterpene hydrocarbons from FDP (Table S2), similar to the profile reported earlier. 14,17Kinetic constants for PTS (Table 1) were slightly different than the reported values (K M 4.0 μM and k cat 4.0 × 10 −4 s −1 ). 14The majority of the products identified (98%) logically derive from the germacryl cation, 17 while trace compounds (2%) likely arise from the humulyl cation (Scheme 1), suggesting multiple mechanisms operating concurrently.
We created a homology model using 5-epi-aristolochene synthase as template (details in Supporting Information). 5We subsequently modeled in (E,E)-FDP in various possible orientations consistent with C1−C10 cyclization and performed multiple independent molecular dynamics (MD) simulations.By monitoring the C1−C10 distance and propensity to form an R-germacryl cation (Figure S2), consistent with patchoulol formation, we predicted the likely PTS:FDP complex (Figure 1A): W276 delineates the pocket at the face opposite to the diphosphate; Y525 and Y531 on the J-K loop (residues 525−537) point into the active site pocket and interact with FDP.Y525 likely coordinates the water molecule(s) involved in hydroxylation, whereas W276 and Y531 ensure a cyclization-competent farnesyl carbocation conformation.
MD simulations indicate that the hydroxyl of Y525 is located proximal to the C6 position in FDP and likely helps coordinate water molecule(s) involved in carbocation quenching to form patchoulol. Consistent with this, a Y525F mutation resulted in only trace amounts of 2 (<2%) as hydroxylated product, instead generating sesquiterpene hydrocarbons β-caryophyllene (12, 40%) and α-bulnesene (4, > 30%, Figure 1B).Formation of 12, requiring C1−C11 cyclization, was confirmed by coinjection studies (Figure S15).The balance between C1−C10 and C1−C11 cyclization is clearly subtle, consistent with the distances measured in MD simulations of the predicted PTS:FDP complex (Figure S2D).The Y525A variant resulted in only non-hydroxylated products derived from C1−C10 cyclization along with ∼9% of 12. Tailing in the chromatogram before a sharp peak at the retention time of 4 suggested the presence of germacrene A (3). 18 Separation of these two products proved problematic on an achiral stationary phase hence GC analysis on a chiral stationary phase was employed to separate and quantify 3 and 4, indicating that 3 is the major product (>51%), with ∼31% of 4 (Figure S21, figures are less precise than other measurements due to this tailing).Similar GC analysis of the other variants also detected ∼9% 3 from Y525F (Figure S22).Y525A and Y525F have an ∼7-fold increase in K M compared to PTS WT with similar k cat (Table 1).We hypothesize that this is related to less efficient formation of a productive PTS:FDP:(Mg 2+ ) 3 complex by abolishing the hydrogen bond between Y525 and N449, a key residue for Mg 2+ coordination.This is consistent with PTS Y525W being inactive (Figure S9).Notably, it is possible that Y525 could reprotonate α-bulnesene (before activating the water for hydroxylation), similar to the suggested role of Y520 in Tobacco 5-epi-aristolochene synthase in reprotonation of germacrene A. 5 Eliminating the Y531 hydroxy (Y531F) leads to a profile similar to that of PTS WT , with one significant difference: 4 is now the major product (51%), at the cost of 2 (Figure 1B, Figure S23).Directed mutagenesis of PTS successfully produced variants that were incapable of hydroxylation.However, such specific targeted mutagenesis for structurally uncharacterized TSs is time-consuming and may not be transferable to other TSs.We previously were successful in significantly reducing hydroxylation by targeting the structurally conserved "kink" region of the G-helix in Gd11olS, with the G188A variant producing 88% of 24. 19The equivalent mutation in PTS (C405A) indeed led to primarily non-hydroxylated products with 36% 4, without impacting turnover (Table 1).However, formation of 2 (>31%) indicates that this strategy may not generally lead to significant reduction in hydroxylation.We therefore considered modifying the Hα-1 loop that undergoes a significant conformational change in the open-to-closed transition in Class I TSs and was suggested as a target to increase TS product diversity. 20In the closed form, this loop shields the reacting carbocation species from bulk water (avoiding premature quenching). 1 We first attempted changing the Hα-1 loop of Gd11olS, based on the loop sequence in the well-characterized, nonhydroxylating selinadiene synthase (SdS). 21We replaced 8 or 4 residues in Gd11olS ( 238 VEDEGELS 245 or 238 VEDE 241 ) with the equivalent residues in SdS ( 233 RRGSGYYL 240 or 233 RRGS 236 ).The 8-residue replacement led to inactivity; however, 4-residue replacement variant Gd11olS Hα-1 resulted in significantly reduced hydroxylation (8.3% germacredien-11ol, 23).The major product was isolepidozene (24, 81%, Figure 2B), the final neutral intermediate in production of 23, 19 alongside smaller amounts of germacrene D (11, 8.4%) and 3 (2.4%).To understand this reduction in hydroxylation, we performed multiple independent MD simulations of Gd11olS WT and Gd11olS Hα-1 in complex with 24 (details in the Supporting Information).In simulations of the Gd11olS WT :24 complex, the closest water molecule to C11 is most commonly positioned within 4 Å (>40% of simulation time).Notably, this water molecule is coordinated by H320 such that it can assist water attack by (transiently) accepting a proton, confirming our previous suggestion (Figure 2C; Gd11olS H320F reduces hydroxylation by ∼50%, in favor of 24). 19In equivalent simulations of Gd11olS Hα-1 , the closest water molecule is typically much further away (only 8% of simulation time within 4 Å) and dominated by a position deeper in the pocket (Figure 2C).This difference is consistently observed in 5 independent simulations.The change in the Hα-1 loop does not cause a direct structural change in the active site template around H320, but it does lead to a significant reduction in water placed in line for hydroxylation at isolepidozene C11, explaining 24 as the major product.Hα-1 loop replacement thus limits the availability of water that can quench carbocations.When the Hα-1 loop from Gd11olS in SdS is used, production of the neutral intermediate germacrene B (25) is increased (66% for 233 VEDE 236 , 98% for 233 VEDEGELS 240 , at the cost of catalytic efficiency; Figures S19 and S20, Table S3), but no hydroxylation is introduced, presumably because a hydroxylating water is not available.
Having established that the four-residue Hα-1 loop variant reduces hydroxylation in Gd11olS, we produced and characterized the equivalent PTS Hα-1 variant ( 458 KKRE 461 to RRGS).This resulted in exclusive sesquiterpene hydrocarbon formation, abolishing the formation of 2 (Figure 2B).The major products were 4 (46%) and 3 (40%), both neutral intermediates in the proposed mechanism (Scheme 1).To confidently identify the major products from PTS Hα-1 , the enzymatic reaction was scaled up (details in the Supporting Information) and the presence of 3 18,22 and 4 23 in similar proportions was confirmed by NMR spectroscopy (Figures S58−S62).Several mechanistic proposals for the biosynthesis of patchoulol have been reported. 15,17,24A recent detailed labeling study suggested that cyclization of FDP to patchoulol goes through reprotonation of two neutral intermediates (3 and 4). 17α-Bulnesene (4) then undergoes further rearrangement and water capture to form 2. The product profiles determined here (particularly PTS WT , PTS Y531F, and PTS Hα-1 ) support this as the most likely mechanism by indicating that subtle disruption of the active site primarily leads to the formation of these neutral intermediates.
Kinetic characterization of these variants showed increases in K M compared to WT (2.5-fold for PTS Hα-1 and 15-fold for Gd11olS Hα-1 ), with similar k cat values (Table 1).This likely reflects a somewhat hindered open-to-closed conformational change upon formation of the reactive enzyme−FDP complex.Overall, these results suggest that manipulating the Hα-1 loop can avoid water capture in hydroxylating (sesqui)TSs, with little change in initial cyclization propensity and kinetic efficiency.
In conclusion, we have demonstrated a combined computational and experimental approach to manipulate and understand water capture in (sesqui)TSs.Despite a lack of structural information and with limited mutagenic effort, we have obtained PTS Hα-1 and J-K loop variants that produce αbulnesene, a complex sesquiterpene capable of interfering with platelet aggregation, as the major product.Further, we show that mutagenesis of four amino acids in the structurally conserved Hα-1 loop results in avoiding water capture in two hydroxylating TSs with significantly different loop sequences.Such loop modification may thus be a more general strategy to obtain TS biocatalysts for highly complex terpenes with possible biological applications.

Scheme 1 .
Scheme 1. Proposed Pathways for Formation of Sesquiterpenes by PTS Variants a

Figure 2 .
Figure 2. Hα-1 loop variants of Gd11olS and PTS.(A) Closed structures of Gd11olS (PDB 5DZ2, green), PTS (homology model, see the Supporting Information, light blue), and SdS (PDB 4OKZ, orange, with Mg 2+ as green spheres and FDP with gray carbons), with Hα-1 loop sequences shown.For Gd11olS and PTS, the 4 residues of the Hα-1 loop targeted are colored magenta (and sequence in bold), with the following 4 residues (targeted in the 8-residue variant) blue.(B) TICs of products of Gd11olS and PTS WT and Hα-1 loop variants.(C) Histograms of the distance of the closest water to C11 in isolepidozene (that do not coordinate Mg 2+ ), as observed in 5 independent 30 ns simulations of Gd11olS WT and Gd11olS Hα-1 .