Engineering class I terpene synthases for skeletal diversity: strategies and applications

Contents

• 1. Introduction

• 2. Structure-guided rational design
  • 2.1. Targeting active site residues for carbocation control and pathway redirection
  • 2.2. Rational engineering of active site architecture for skeletal reprogramming
  • 2.3. Controlling cyclization by engineering water accessibility
  • 2.4. Altering active site pocket size and shape

• 3. Evolutionary and comparative approaches
  • 3.1. Phylogeny-guided mutation
  • 3.2. Mimicking natural variation and functional switches

• 4. Mechanism-focused engineering
  • 4.1. Carbocation intermediate control
  • 4.2. Reaction BP control

• 5. Engineering beyond the active site
  • 5.1. Engineering structural elements
  • 5.2. Second-shell engineering
  • 5.3. Combinatorial mapping of catalytic landscapes approach

• 6. Semi-rational and random approaches
  • 6.1. Alanine scanning
  • 6.2. Saturation mutagenesis and error-prone PCR

• 7. Protein segment engineering
  • 7.1. Protein segment swapping for functional conversion

• 8. Emerging and focused engineering strategies
  • 8.1. Advanced applications of computational modeling and simulation
  • 8.2. Directed evolution and high-throughput screening
  • 8.3. Data-driven and machine learning approaches

• 9. Conclusions and perspectives

• 10. Conflicts of interest

• 11. Data availability

• Acknowledgements

• References

---

Abstract

Covering: up to August 2025

Terpenoids constitute nature's largest and most structurally diverse class of natural products, with extensive applications in medicine, agriculture, and fragrance industries. Class I terpene synthases (TSs) create this remarkable diversity by converting linear isoprenoid diphosphates into complex, often polycyclic frameworks through intricate carbocation cascades. This review examines strategies for engineering TSs to generate diverse terpene skeletons—a key objective in synthetic biology. We summarize four core approaches: structure-guided design targeting active sites, water networks, and conserved motifs; evolutionary methods leveraging natural variation and phylogenetic insights; mechanism-focused engineering controlling specific carbocation intermediates; and techniques extending beyond the active site through second-shell modifications and contact mapping. These approaches are complemented by semi-rational and random methods including alanine scanning, saturation mutagenesis, and directed evolution, often enhanced by computational modeling and high-throughput screening. While the complexity of TS catalysis and often weak sequence-function correlations create significant engineering challenges, integration of structural biology, computational simulations, diverse engineering techniques, and advanced screening methods is steadily improving outcomes. Future advances in machine learning, mechanistic understanding, screening technologies, and metabolic engineering integration will further expand access to novel terpenoid chemical space for biotechnological exploration.

---

1. Introduction

Terpenoids, also known as isoprenoids, represent the largest and most structurally diverse class of natural products, comprising over 100 000 identified compounds.^1–4 These molecules, derived from the universal five-carbon precursors isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP),^5,6 exhibit a vast array of biological functions and are invaluable resources for human applications, including pharmaceuticals (e.g., paclitaxel,^7,8 artemisinin⁹), agrochemicals,^10,11 fragrances,¹² and biofuels.^13,14 The remarkable chemical diversity of terpenoids originates primarily from the activity of terpene synthases (TSs), enzymes that catalyze the transformation of linear isoprenoid diphosphate substrates (e.g., geranyl diphosphate (GPP, C10),¹⁵ farnesyl diphosphate (FPP, C15),^16–18 geranylgeranyl diphosphate (GGPP, C20),^19,20 geranylfarnesyl diphosphate (GFPP, C25))²¹ and C30 substrate squalene or epoxysqualene^22,23 into a multitude of often complex cyclic and polycyclic hydrocarbon or alcohol skeletons.

TSs are broadly classified into two main types based on their reaction mechanisms and structural folds.^2,24 Class II TSs typically possess a βγ-domain structure and initiate catalysis by protonating the substrate double bond or epoxide group, often cyclizing larger substrates like triterpenes (e.g., squalene).^23,25–27 In contrast, class I TSs, the focus of this review, generally feature an α-helical fold (α-domain) and catalyze the ionization of the allylic diphosphate substrate, generating an initial carbocation.^28,29 This highly reactive intermediate then undergoes a series of intricate and precisely controlled intramolecular cyclizations, rearrangements (including hydride and methyl shifts), and quenching steps (typically deprotonation or water capture), often referred to as a carbocation cascade.^30,31 This complex cascade, guided by the enzyme's active site architecture, is the source of the vast skeletal diversity observed in mono-, sesqui-, di-, and sesterterpenes (C10–C25).³²

The characteristic α-helical fold of class I TSs creates a hydrophobic active site cavity lined with specific residues that stabilize transient carbocation intermediates, guide the substrate through the reaction coordinate, and ultimately dictate the final product structure.^28,29,33 Engineering class I TSs offers multiple advantages that majorly include (1) access to novel, non-natural terpene skeletons with unique bioactivities; (2) generation of analogs to optimize existing therapeutic compounds; (3) enabling sustainable bioproduction of valuable terpenoids that are otherwise obtained through inefficient extraction from natural sources or complex chemical synthesis; and (4) addressing limitations of wild-type (WT) enzymes in heterologous expression systems.^34–36 Among these advantages, the generation of novel terpenoid skeletons through enzyme engineering has emerged as a particularly compelling objective, as new scaffolds can lead to compounds with unique biological activities or improved properties.^37–39 This skeletal diversification capability is especially valuable, as chemical synthesis of complex terpenes is often challenging,^40,41 whereas engineering class I TSs offers a powerful biosynthetic alternative for accessing both known and novel terpenoid structures.^2,42 The inherent catalytic promiscuity and evolutionary plasticity of these enzymes make them attractive targets for protein engineering aimed at altering their product specificity and generating skeletal diversity.

Despite these opportunities, engineering TSs presents significant challenges: product specificity control is difficult due to the intrinsic promiscuity of many TSs; weak correlation exists between enzyme sequence and functional output; the complex, dynamic reaction mechanism involves multiple transient carbocation intermediates; and there is limited availability of high-resolution structures capturing catalytically relevant conformations. These factors constitute a “predictability barrier” that necessitates integrating rational approaches with empirical methods. This review focuses on the strategies employed to engineer class I TSs for the generation of diverse and novel terpenoid skeletons. We explore several core approaches, including: (i) structure-guided rational design targeting specific active site residues, water networks, or conserved motifs; (ii) evolutionary and comparative methods that leverage natural variation, phylogenetic insights, and functional switches identified between related enzymes; (iii) mechanism-focused engineering aimed at controlling specific carbocation intermediates or reaction branch points (BPs); (iv) techniques extending beyond the immediate active site, such as second-shell engineering and contact mapping; (v) protein segment engineering including domain swapping and chimeric approaches for functional conversion; and (vi) semi-rational and random approaches including alanine scanning, saturation mutagenesis, and directed evolution (Fig. 1). We also discuss emerging computational and data-driven methods incorporating machine learning and high-throughput screening technologies that show promise for future applications. These diverse strategies, often used in combination, navigate the complex catalytic landscape of these fascinating enzymes. Understanding and applying these strategies is crucial for unlocking the full biosynthetic potential of class I TSs and expanding the accessible chemical space of terpenoids for biotechnological applications.

Fig. 1 Overview of major engineering strategies for class I TSs to achieve skeletal diversity.

2. Structure-guided rational design

2.1. Targeting active site residues for carbocation control and pathway redirection

2.1.1. Reshaping the active site contour to remodel carbocation fates. The hydrophobic active site of a class I TS functions as a precisely sculpted catalytic pocket, guiding the complex carbocation cascade through a network of non-covalent interactions.^43–45 Consequently, remodeling the steric and electronic landscape of this pocket is a foundational strategy for reprogramming the product outcome. The catalytic plasticity of epi-isozizaene synthase (EIZS) from Streptomyces coelicolor has been extensively explored, serving as a canonical model for this engineering approach.^44,45

Altering steric bulk within the active site can effectively interrupt or redirect the cyclization cascade.⁴³ EIZS WT converts FPP into the complex tricyclic product epi-isozizaene with high fidelity (79% of the product mixture) (Fig. 2).⁴³ Its active site is lined with aromatic residues that stabilize the series of carbocation intermediates required for the intricate cyclization. Crystal structures (e.g., EIZS WT, PDB ID: 4LTV) revealed that F96A causes minimal structural perturbations but significantly alters the active site contour (Fig. 2B). This modification removes a key steric barrier and cation-π stabilizing interaction, leading to the premature termination of the cascade. As a result, the product profile is substantially inverted, with the acyclic (E)-β-farnesene becoming the major product (70%), demonstrating a complete skeletal simplification from tricyclic to linear. Further mutations targeting other aromatic residues, such as F95H and F198L, similarly redirected the cascade to yield alternative monocyclic (β-curcumene) and tricyclic (β-cedrene) skeletons, respectively. Notably, the F198L mutant redirected the cascade to produce 61% β-cedrene, of which 80% was (−)-β-cedrene and 20% was (+)-β-cedrene, highlighting a significant loss of stereocontrol and the plasticity of the catalytic pocket (Fig. 2A).⁴³

Fig. 2 Engineering of EIZS for sesquiterpene skeletal diversity. (A) Reaction scheme illustrating the diversification of sesquiterpene skeletons from FPP. While EIZS WT produces the tricyclic epi-isozizaene, structure-guided mutations at key residues in active site redirect the cascade to yield alternative acyclic, monocyclic, bicyclic, and tricyclic products. (B) Representative crystal structures of engineered EIZS mutants. The top-left panel shows the EIZS WT (green; PDB ID: 7KJ9) in complex with the inhibitor risedronate (RIS). The top-right panel displays the F95H mutant (cyan; PDB ID: 4LZ3) binding the carbocation analog benzyltriethylammonium (BTM). The bottom-left panel features the F95S/F198S double mutant (purple; PDB ID: 8V3K) also complexed with BTM, highlighting a significantly enlarged active site cavity that facilitates hydroxylation. Finally, the bottom-right panel presents a molecular model of the F96S mutant (yellow; PDB ID: 7KJE), which functions as a nerolidol synthase, accommodating its major alcohol product, nerolidol (NRD), providing a rationale for the enzyme's reprogrammed function.

In parallel, modulating the pocket's polarity provides an orthogonal handle to steer carbocation intermediates toward alternative reaction trajectories. The introduction of polar residues, such as in the EIZS F96S or F96Q mutants, fundamentally alters the electronic environment of the active site.⁴⁶ Crystallographic analysis (the unliganded F96S mutant, PDB ID: 6AXO) confirmed that these substitutions preserve the enzyme's architecture while creating localized electronic effects. These changes critically alter the stabilization of the homobisabolyl⁺, redirecting its cyclization pathway from the native tricyclic scaffold to a bicyclic framework, sesquisabinene A. The F96Q variant performs this conversion with near-perfect selectivity (97%).⁴⁷ Building on these findings, a synergistic double mutant (F95S/F198S) was designed; the core hypothesis was that this mutation would synergistically expand the active site to accommodate both the bisabolyl⁺ and a water molecule—an effect neither single mutant could achieve alone. This non-additive expansion was confirmed by active site volume measurements (WT: 279 ųvs. double mutant: 391 ų). Structural analysis of this variant at 1.47 Å resolution (PDB ID: 8V3K) revealed that the elongated active site not only redirects the product skeleton from tricyclic to monocyclic but also positions a water molecule for nucleophilic attack, thereby converting the enzyme into a specific α-bisabolol synthase with high yield (74% at 4 °C). Crucially, the engineered active site shares key structural features with a natural α-bisabolol synthase (AaBOS from Artemisia annua),⁴⁸ providing evolutionary validation for the design strategy.⁴⁶ These studies collectively exemplify how precise, structure-guided sculpting of the active site can effectively intercept and reroute the carbocation flux to unlock novel terpenoid scaffolds.

2.1.2. Harnessing computational insights to engineer reaction BPs. The rational engineering of class I TSs has been revolutionized by the integration of computational approaches that provide unprecedented molecular-level insights into carbocation cascade mechanisms.^49–51 High-resolution crystal structures combined with quantum mechanical calculations and molecular dynamics (MD) simulations reveal how specific residues control reaction BPs through stage-specific stabilization of carbocation intermediates.⁵² This computational framework enables the identification of critical molecular interactions that govern skeletal outcomes, transforming enzyme engineering from empirical screening to predictive design.^53,54 By mapping the energetic landscape of carbocation cascades and quantifying residue-substrate interactions, rational targeting of specific reaction coordinates now becomes possible to achieve desired skeletal reprogramming from complex polycyclic frameworks to alternative scaffolds.^55–60

Taxadiene synthase (TXS) engineering exemplifies the integration of multiple computational approaches to achieve precision skeletal reprogramming in the biosynthesis of paclitaxel (Taxol®), one of the most important anticancer agents.^57,58 The initial three-dimensional crystal structures (PDB ID: 3P5P) of TXS revealed its active site cavity, identifying a group of polar residues (S587, Q609, Y688, C719, and C830) as potential candidates for catalysis. Subsequent engineering efforts altered the product profile by manipulating carbocation intermediates. One common outcome was the enhanced production of taxa-4(20),11(12)-diene, a more favorable isomer for paclitaxel biosynthesis. The Y688L⁵⁹ or V584I⁵⁶ reorients the final taxen-4-yl⁺, making its C20 methyl protons more accessible for deprotonation and favoring the exo-cyclic double bond formation.^58,61 More reprogramming involves intercepting the catalytic cascade to yield entirely different carbon skeletons. Cembrene A, a monocyclic diterpene, is formed by halting the reaction after the first cyclization. This was achieved after molecular mechanic (MM)-based modeling predicted that the W753H mutation introduces an active base that prematurely terminates the reaction.⁵⁵ Similarly, V610F sterically blocks the path to the second cyclization, also yielding cembrene A (Fig. 3).⁵⁶ The reaction can also be terminated at the bicyclic stage to produce verticillene-type compounds.⁵⁶ This strategy was guided by computational modeling, which indicated that altering the active site's shape could interrupt the process. The models predicted that V584M/L⁵⁶ or Q609A/G,^56,59 the bicyclic intermediate is physically pushed closer to the diphosphate moiety, triggering its premature deprotonation.

Fig. 3 Engineering taxadiene synthase (TXS) for altered diterpene production. The reaction scheme illustrates the skeletal reprogramming of TXS, a key enzyme in the biosynthesis of the anticancer agent paclitaxel (Taxol®). The embedded crystal structure highlights the modular architecture of TXS and provides a close-up of the active site, where key residues like W753, V610, V584, Q609, and Y688 are selected for mutagenesis. While the TXS WT converts GGPP to a native tricyclic taxadiene skeleton, structure-guided mutations can intercept or redirect the complex carbocation cascade to generate alternative scaffolds, such as monocyclic cembrene A or a bicyclic verticillatriene skeleton.

In summary, these residues function as sequential gatekeepers that can be engineered to intercept the native tricyclic cascade. Y688 exclusively directs the final deprotonation, controlling the double bond's regioselectivity on the intact skeleton. W753 and V610 control the first cyclization; their mutation introduces either a catalytic base (W753H) or a steric block (V610F), halting the reaction to release a monocyclic product. V584 and Q609 shape the active site for the second cyclization; mutating them alters this geometry, forcing the premature release of a bicyclic intermediate before the final ring can form.

The bacterial diterpene synthase (DTS) VenA demonstrates how crystal structure-guided manipulation of aromatic cage architecture can achieve skeletal rearrangements through systematic residue targeting.³⁴ Crystal structures (PDB IDs, 7Y9H and 7Y9G) revealed seven critical aromatic residues (Y88, W107, Y108, F185, F215, F219, and F338) positioned within 5 Å of the GGPP, each contributing stage-specific stabilization energies ranging from −6.8 to −9.6 kcal mol⁻¹.³⁴ Quantum Mechanics/Molecular Mechanics (QM/MM) simulations demonstrated that Y88 and F215 jointly stabilize intermediate C during early cyclization stages after intermediate B undergoes a 1,3-hydride shift, while F185 and W107 control later reaction trajectories, particularly stabilization of non-classical carbocation intermediate F, which arises from the rearrangement and subsequent cyclization of the four-membered ring in intermediate E. The cascade involves a critical BP at the initial carbocation intermediate A, which in certain mutants (e.g., Y88A) can undergo direct 1,14-cyclization to form 14-membered monocyclic products like cembrene A. In the WT, the pathway proceeds linearly through a separate 1,14-cyclization (C → D), followed by a 2,15-cyclization toward intermediate E and ultimately the tetracyclic venezuelaene scaffold (Fig. 4). Systematic alanine scanning exploited these mechanistic insights to redirect the native 5/5/6/7 tetracyclic venezuelaene A pathway toward alternative skeletal frameworks. The Y88A mutation, disrupting early-stage intermediate stabilization, transformed products to acyclic β-springene and monocyclic cembrene A while retaining only 21.7% of WT activity. The W107A variant, affecting late-stage carbocation control, generated 3,7,18-dolabellatriene (5/11 bicyclic system) with a titer of 275.5 ± 9.4 mg L⁻¹ through an alternative 1,11-cyclization rather than native 1,10-cyclization. The F215A mutation produced dictytriene C, representing a 5/7 bicyclic framework transformation. Additionally, a systematic combination of six mutations (Y88A/W107A/V111A/T112A/F215A/F219A) expanded the active site volume from 516.1 ų to 641.5 ų, enabling substrate scope expansion to accommodate GFPP and produce the 14-membered monocyclic sesterterpene (S)-cericerene at 92.6% selectivity.

Fig. 4 Reprogramming the catalytic landscape of VenA through aromatic cage repositioning. (A) Products generated by VenA WT and its mutants from GGPP substrate. The upper panel shows alternative products including acyclic, monocyclic, and bicyclic diterpenes. The lower panel shows the native carbocation cascade pathway (intermediates A–G) leading to 5/5/6/7 tetracyclic venezuelaene A and hydroxylated products venezuelaenols A and B. Inset: VenA crystal structure (PDB ID: 7Y9G) showing key aromatic cage residues. (B) Conversion of GFPP to monocyclic (S)-cericerene by the hexuple mutant VenA Y88A/W107A/V111A/T112A/F215A/F219A.

TXS and VenA engineering collectively demonstrate that aromatic residue networks function as programmable molecular switches for skeletal reprogramming.^34,55–59 The success of this approach relies on detailed mechanistic understanding of stage-specific carbocation stabilization, enabling rational design of mutations that predictably redirect cyclization cascades toward desired skeletal outcomes. The integration of structural biology, quantum mechanical modeling, and systematic mutagenesis provides a robust framework for expanding the accessible chemical space of TSs beyond their native product profiles.

2.2. Rational engineering of active site architecture for skeletal reprogramming

Active site engineering represents the most direct approach for redirecting terpene cyclization cascades, leveraging the principle that the three-dimensional architecture of the catalytic pocket fundamentally controls carbocation stabilization, intermediate conformations, and reaction termination.² The hydrophobic active site of class I TSs functions as a precisely sculpted template where deliberate alterations to steric bulk, electronic character, and solvent accessibility can predictably intercept reaction intermediates at different stages of the cyclization cascade.⁶² This approach enables systematic diversification of molecular scaffolds, as demonstrated through successful engineering efforts across monoterpene, sesquiterpene, and diterpene synthase families.

2.2.1. Reshaping active sites in polycyclic sesquiterpene synthases. Structure-guided protein engineering, when combined with QM/MM simulations, has proven effective for reprogramming sesquiterpene synthases (STSs) that produce complex polycyclic frameworks. A comprehensive study of three STSs—BcBOT2, DbPROS, and CLM1—revealed the molecular basis for controlled skeletal diversity. The high-resolution crystal structures of the enzymes with a substrate mimic (PDB IDs: 8H6U, 8H72, 8H4P) provided detailed visualization of the carbocation stabilization networks responsible for guiding the cyclization trajectories (Fig. 5A).³⁵

Fig. 5 Structure-guided engineering of STSs for extensive skeletal diversification. (A) Close-up views of the active sites of three fungal STSs with a bound substrate analog (benzyltriethylammonium cation (BTAC), magenta) and Mg²⁺ ions (green spheres): BcBOT2 (gray, left), DbPROS (light blue, center), and CLM1 (cyan, right). (B) Product diversification achieved through mutagenesis. Gray region: BcBOT2-derived tricyclic alcohols and skeletal variants; blue region: DbPROS-derived tricyclic protoilludene-type products; purple region: CLM1-derived tricyclic and bicyclic products. Mutant designations are shown for each group. Mutant designations are indicated for each product group.

The engineering of BcBOT2, which naturally produces the tricyclic alcohol presilphiperfolan-8β-ol, demonstrates how single mutations can redirect complex cyclization cascades (Fig. 5B). The W94A mutation eliminated both steric bulk and aromatic stabilization, causing premature termination at the humulyl⁺ intermediate to yield the 11-membered macrocyclic α-humulene and its hydroxylated bicyclic derivative. The F99A variant produced the bicyclic β-caryophyllene and a related bicyclic alcohol through early termination at different carbocation intermediates. Most remarkably, the N325A mutant generated presilphiperfolene, which represents a ring-cleaved rearrangement product involving C-12 methyl migration—a skeletal reorganization from the native tricyclic framework. Additionally, this mutant produced germacrene A through alternative 1,10-cyclization. Similarly, modifications to DbPROS, which yields the tricyclic Δ⁶-protoilludene, led to significant skeletal simplification. The I193A mutation altered the active site geometry, shortening the C1–C10 distance from 3.6 Å to 2.6 Å, which promoted alternative 1,10-cyclization to form the germacrene A instead of proceeding through the normal tricyclic pathway. The F93A mutation removed both aromatic stabilization and steric constraints, leading to water attack and formation of a hydroxylated bicyclic product. CLM1 engineering revealed the most extensive skeletal diversification. While the WT produces the tricyclic alcohol longiborneol, site-directed mutagenesis generated diverse alternative frameworks. The T111A variant produced a mixture of mono- and bicyclic sesquiterpenes by disrupting dipole–cation interactions with early intermediates. The V83A and Y87A mutations yielded another set of products from the same intermediate through different termination mechanisms: a tricyclic hydroxylated alcohol and its corresponding tricyclic deprotonated olefin were formed, while another product resulted from an additional ring closure to form a distinct new tricyclic framework (Fig. 5B). These findings on STSs reveal how minimal alterations of active site properties can effectively redirect cyclization cascades toward a wide array of diverse carbon frameworks.³⁵

2.2.2. Reprogramming eunicellane synthases for divergent scaffolds. Eunicellane diterpenoids are a class of bioactive 6,10-bicyclic natural products whose synthases are often reprogrammed through structure-guided mutagenesis to generate novel skeletal variants.^63–68 Recent advances in computational modeling and structure-guided mutagenesis have enabled systematic reprogramming of bacterial eunicellane synthases to generate novel skeletal variants, providing critical insights into molecular determinants governing stereochemical precision and conformational dynamics.^65,69–71

Careful manipulation of aromatic residues can intercept carbocation intermediates. The DTS Bnd4, from Streptomyces sp. (CL12-4), produces the cis-fused 6/10 bicyclic diterpene benditerpe-2,6,15-triene through a 1,10-cyclization followed by 1,14-cyclization of GGPP.^70,71 A computer model of Bnd4 identified an active site residue, W316. W316A/H altered the reaction pathway from the native 1,10-cyclization to a 1,14-cyclization, yielding cembrenes with a 14-membered monocyclic skeleton as the major products. A similar approach was used for AlbS from Streptomyces albireticuli, which synthesizes the trans-fused 6/10 bicyclic diterpene albireticulene via initial 1,10-cyclization, followed by a critical 1,3-hydride shift, and subsequent 1,14-cyclization (Fig. 6A).⁷² Density functional theory (DFT) calculations on the WT mechanism identified a key monocyclic carbocation intermediate formed after the initial 1,10-cyclization. The calculations showed this intermediate must overcome an 11.1 kcal mol⁻¹ energy barrier for the subsequent 1,3-hydride shift to occur. This theoretical model provides a direct explanation for the experimental outcome of the Y214F mutation, a mutational target selected due to its homology with the critical Y197 residue in Bnd4.⁷¹ By altering a key residue, the mutant is unable to stabilize the transition state and facilitate the reaction over this energy barrier, thus intercepting the reaction at the monocyclic intermediate. As a result, the mutant's activity to form the native bicyclic product was reduced to 4% of WT, with the primary product being prenylgermacrene A, a 10-membered monocyclic germacrane.

Fig. 6 Skeletal diversification of eunicellane-type diterpenes through active site engineering. (A) Product diversification of bacterial eunicellane synthases (AlbS, Bnd4, and AriE) and their mutant variants. (B) Product diversification of coral-derived PcTS1 through systematic mutagenesis, showing various scaffolds from different mutants.

Direct crystallographic visualization provides unparalleled precision for systematic scaffold diversification. The coral-derived synthase PcTS1, resolved at 3.5 Å resolution (PDB ID: 8ZWZ), exemplifies how structural data guides comprehensive active site engineering. The WT enzyme catalyzes the formation of elisabethatriene, a diterpenoid featuring a 6/6 bicyclic biflorane skeleton through multiple cyclizations and hydride shifts (Fig. 6B).^42,73 The crystal structure reveals a distinctive class I TS α-helical fold characterized by the absence of helix K, elongation of helices H and M, and the presence of an additional helix O.⁶⁹ Systematic mutation of crystallographically defined positions produced notable scaffold diversity. The I254A mutant generated three distinct products: a 10-membered monocyclic intermediate from the initial 1,10-cyclization, and two constitutional isomers of the native biflorane skeleton resulting from deprotonation at alternative sites. This substitution enlarges the active site cavity, which provides the final carbocation intermediate with greater conformational freedom, leading to these varied quenching patterns. Similarly, the S338I mutation yielded a novel cage-like tricyclic structure by altering the stability of an earlier bicyclic intermediate formed after the 2,6-cyclization step, a change proposed to promote an additional, non-native cyclization event. Other designed mutations successfully intercepted specific intermediates along the catalytic cascade to produce a cyclopropane-containing product (I59L), a constitutional isomer of the native scaffold (F93Y), and an alternative tricyclic scaffold (F221V).⁶⁹

The first high-resolution eunicellane cyclase structure, AriE at 1.87 Å (PDB ID: 9LPB), revealed precise aromatic residue contributions to product specificity.⁶⁵ The crystal structure provides clear mechanistic explanations through its detailed active site architecture (volume 588.80 ų, surface area 481.47 Ų, depth 16.18 Å), where W73 and neighboring residues form a critical hydrophobic pocket. Molecular docking identified W73 as a critical regulatory hotspot, and saturation mutagenesis at this position altered product distributions. Modifications to the W73 disrupt the stabilization of specific carbocation intermediates and the trajectory of subsequent hydride shifts, preventing the formation of the native cis-6/10-bicyclic benditerpetriene and instead yielding diverse scaffolds including monocyclic cembrenes and products with inverted stereochemistry.⁶⁵ Conversely, the L96V mutation converted AriE into a selective cis-6/10-bicyclic eunicellane synthase by subtly refining the active site contour, which enhances stabilization for the bicyclic pathway while restricting the formation of cembrene byproducts.⁶⁵ These studies collectively demonstrate how computational frameworks can systematically unlock vast chemical diversity through rational manipulation of aromatic residue networks, water positioning, and stereochemical control elements across diverse TS families.

2.2.3. Aromatic residues as molecular gatekeepers. Aromatic residues within the hydrophobic active sites of class I TSs function as critical molecular gatekeepers that orchestrate carbocation cyclization pathways through precisely positioned cation-π interactions and hydrogen bonding networks.^74,75 These residues, particularly phenylalanine and tyrosine, serve dual roles as both stabilizing elements for specific carbocation intermediates and architectural components that control water access and conformational dynamics.⁷⁶ Structure-guided engineering of these aromatic gatekeepers, informed by high-resolution crystallography and quantum mechanical calculations,^47,77 has emerged as a transformative strategy for achieving skeletal reprogramming—enabling the transformation of complex polycyclic scaffolds into simplified monocyclic or alternative bicyclic frameworks through selective disruption of native cyclization cascades.

The engineering of Sorangium cellulosum cubebol synthase (ScCubS) exemplifies how multiple aromatic residues function in concert to control tricyclic cubebane formation (Fig. 7).⁷⁸ High-resolution crystal structure determination (1.80 Å resolution, PDB ID: 7ZRN) revealed that F104 stabilizes the C7 position of cation C through cation-π interactions, while F211 serves as a barrier against premature water quenching of reaction intermediates, with S206 positioning F211 correctly through a hydrogen-bonding network.⁷⁸ Substitutions at these critical positions substantially redirected cyclization pathways from the native tricyclic 10-epi-cubebol to alternative scaffolds. The F104Y mutation disrupted cation-π stabilization at C7, favoring water quenching of an earlier intermediate and effectively transforming the enzyme into a germacradien-4-ol synthase with ∼80% selectivity. This mutation shifted the major product to germacradien-4-ol at 7.59 mg L⁻¹, while 10-epi-cubebol production decreased from ∼45 mg L⁻¹ in the WT to 1.23 mg L⁻¹. The F211A variant removed the steric and aromatic barrier preventing water access, allowing deprotonation at intermediate C to produce γ-cadinene or water addition to form 1,10-di-epi-cubenol (bicyclic cadalane scaffold), while S206G disrupted the hydrogen-bonding network positioning F211, preventing the third cyclization and resulting in predominantly cadalane-type compounds.⁷⁸

Fig. 7 Redirecting the cyclization cascade of ScCubS. Reaction scheme showing the skeletal reprogramming of Sorangium cellulosum cubebol synthase (ScCubS). The cascade begins with metal-dependent ionization of FPP, where Mg²⁺ ions (green spheres in crystal structure) coordinate the pyrophosphate moiety to facilitate C–O bond cleavage. Inset shows the crystal structure (PDB ID: 7ZRN) highlighting key active site residues that control product specificity. The WT produces tricyclic 10-epi-cubebol (red), while mutations lead to monocyclic germacradien-4-ol (blue) or bicyclic products including γ-cadinene and 1,10-di-epi-cubenol (green).

Conserved tyrosines function as conformational gatekeepers that dictate cyclization trajectories through well-defined hydrogen bonding networks.^79–81 The Y573 position in Mentha spicata S-limonene synthase (LS) represents a paradigmatic example, as crystal structure (PDB IDs: 2ONG and 2ONH) analysis identified this residue as uniquely conserved in cyclic MTSs but variable in acyclic ones.⁸² Native enzyme converts GPP primarily to limonene (monocyclic, 96.69%) with minor bicyclic products through initial ionization to geranyl⁺, isomerization to linalyl diphosphate, re-ionization to linalyl⁺, 1,6-cyclization to terpinyl⁺, and final deprotonation (Fig. 8). The Y573F substitution, preserving aromaticity while eliminating the hydroxyl group, altered this distribution despite reducing activity to 2.79%—significantly increasing bicyclic compounds including α-pinene (2.76%), β-pinene (2.15%), and sabinene (4.77%) by allowing the terpinyl⁺ to undergo 2,7-cyclization to pinyl⁺ or alternative rearrangements. MD simulations revealed a critical hydrogen bond between Y573 and D496, with closer proximity between Y573 and carbons C7–C9 of the linalyl diphosphate intermediate in the cyclization-prone helical conformation. Unlike residues that primarily stabilize carbocations, Y573 functions earlier in the cascade by directing intermediate conformation and initial cyclization events.⁷⁹

Fig. 8 Aromatic residue control of monoterpene cyclization cascades in bacterial TSs. LS WT produces limonene (96.69%), while Y573F enables additional 2,7-cyclization yielding bicyclic pinenes and sabinene. bCinS WT forms 1,8-cineole through water capture at terpinyl⁺, whereas F77A produces acyclic linalool and F77H yields monocyclic products. Insets: crystal structures with fluorinated substrate analogues—non-reactive mimics that reveal how Y573 and F77/A301 position intermediates for specific cyclization outcomes. F3P: fluorinated GPP analogue; 0FV: (2Z)-2-fluoro-3,7-dimethylocta-2,6-dien-1-yl diphosphate.

Aromatic residues, particularly phenylalanine, can serve as molecular switches controlling carbocation cyclization through a combination of steric effects and potential cation-π interactions, with the relative contribution depending on spatial positioning. Structure-guided mutagenesis of bacterial cineole synthase (bCinS) coupled with QM/MM calculations identified F77 as the critical determinant of cyclization specificity. The F77A substitution fundamentally redirected product formation from 96% 1,8-cineole (bicyclic) to primarily linalool (acyclic), with QM/MM calculations revealing that F77 stabilizes the α-terpinyl⁺ intermediate with exceptional efficiency (−5.7 kcal mol⁻¹ interaction energy) compared to other phenylalanine (−1.2 to −2.8 kcal mol⁻¹). This stabilization enables water capture by the terpinyl⁺ to form the cyclic ether of 1,8-cineole (Fig. 8). The aromatic nature proved essential—F77W restored cineole production, while F77H redirected cyclization toward monocyclic products like limonene and α-terpineol through alternative deprotonation or hydroxylation pathways. Complementary engineering at position A301 demonstrated architectural control mechanisms, where A301V/L/F mutations eliminated cineole production by creating steric clashes with W58 that disrupted active site geometry necessary for productive cyclization.⁸³

2.3. Controlling cyclization by engineering water accessibility

TS active sites are typically hydrophobic to exclude water and prevent the premature quenching of carbocation intermediates.² A compelling engineering strategy, therefore, involves intentionally creating openings for water, which can intercept the catalytic cascade at specific points.⁸⁴ This not only generates novel hydroxylated products but can also redirect the reaction pathway away from subsequent cyclizations, yielding entirely different carbon skeletons.

This principle is well-demonstrated by the engineered conversion of δ-cadinene synthase (DCS), which naturally produces a bicyclic hydrocarbon, into a germacradien-4-ol synthase (GdolS) that selectively produces a monocyclic alcohol.^85,86 This transformation was achieved through two distinct but complementary approaches (Fig. 9). A highly effective “single-residue gatekeeper” strategy involved mutating W279, a residue that forms a hydrophobic barrier at the active site entrance. The W279A opened a water channel and inverted the product profile, shifting from 90% bicyclic δ-cadinene to 89% monocyclic germacradien-4-ol by allowing water capture at the monocyclic intermediate before the second cyclization, while maintaining high catalytic efficiency.⁸⁷ Alternatively, restructuring the more distant G-helix (e.g., with an N403P/L405H double mutant) also successfully opened a path for water to quench the monocyclic intermediate, yielding 93% germacrene D-4-ol, although the catalytic activity of this mutant was severely compromised relative to the WT.^88–90

Fig. 9 Two distinct strategies for engineering water-mediated catalysis to control STS product formation. Blue pathway: DCS WT produces bicyclic (+)-δ-cadinene, while W279A and N403P/L405H mutations redirect to monocyclic (−)-germacradien-4-ol (red) through water capture. Green pathway: SdS WT converts germacrene B to bicyclic selina-4(15),7(11)-diene, but A301Y/G305E arrests at monocyclic germacrene B by blocking the second cyclization.

Beyond controlling general water access, more precise engineering strategies have emerged that target specific residues involved in water activation for nucleophilic attack on carbocation intermediates.⁹¹ The case of bCinS demonstrates how identifying water-activating residues precisely enables controlled redirection of product formation through minimal mutations. Structural analysis and MD simulations identified N305 as the key residue coordinating a water molecule for nucleophilic attack on the α-terpinyl⁺ intermediate in bCinS (Fig. 9). The WT enzyme exhibits exceptionally high fidelity, producing bicyclic 1,8-cineole with 95% specificity and minor amounts of monocyclic α-terpineol (2%). N305A completely eliminated 1,8-cineole production, redirecting catalysis toward non-hydroxylated terpenes including monocyclic β-phellandrene and bicyclic β-pinene and sabinene. Despite this shift in product profile, the variant maintained substantial activity, although its total terpene production was markedly reduced compared to the high levels of the WT. MD simulations revealed that these mutations fundamentally disrupted the water coordination network, with quantum mechanics calculations suggesting that N305 likely functions as a transient proton acceptor in the final deprotonation step of 1,8-cineole formation. This skeletal transformation occurs because disrupting water activation prevents hydroxylation at the α-terpinyl⁺ stage, allowing alternative cyclization pathways and direct deprotonation to predominate.⁹¹

In selinadiene synthase (SdS), engineering of the K-helix successfully converted the enzyme from a bicyclic selinadiene producer to a producer of the monocyclic intermediate, germacrene B.⁹² The native pathway proceeds through germacrene B formation followed by a second cyclization to yield the bicyclic selina-4(15),7(11)-diene. Notably, the double mutant A301Y/G305E resulted in the exclusive production of germacrene B, completely halting the reaction cascade before the second cyclization. This high degree of product specificity, however, came at the cost of a 162.8-fold reduction in catalytic efficiency.⁹²

2.4. Altering active site pocket size and shape

Beyond modifying residues that directly interact with carbocation intermediates or control water access, reshaping the active site cavity itself offers another valuable engineering approach.^93,94 This approach encompasses both broad cavity expansion to accommodate new substrate classes and precise architectural adjustments to fine-tune product formation.⁹⁵

2.4.1. Expanding substrate scope and switching cyclization modes. Reshaping active site pocket dimensions represents a versatile methodology for transforming enzyme capabilities beyond their native substrate preferences and reaction modes.^93,94 Careful reshaping of the active site cavity enables engineered variants to accept larger isoprenoid precursors while redirecting cyclization trajectories from simple to complex scaffolds, or from one structural class to entirely different architectures. This approach transcends traditional active site engineering by physically reconfiguring the catalytic environment to enable reactions that would be impossible in the WT.⁹⁵

The STS 7-epi-α-eudesmol synthase (SvES) from Streptomyces viridochromogenes exemplifies how systematic cavity expansion can transform both substrate scope and product architecture (Fig. 10).⁹⁶ Structure-guided comparison with SdS identified key hydrophobic residues defining active site boundaries, particularly positions A76 and F77 located directly in front of the Asp-rich motif. SvES WT normally converts FPP to bicyclic 7-epi-α-eudesmol (66%) and monocyclic products, displaying modest substrate scope limited to C15 precursors. The A76G variant showed slightly reduced sesquiterpene production while gaining the ability to convert the larger GGPP (C20) into macrocyclic diterpenes cembrene A and nephthenol (28 ± 2%). This transformation represents not merely substrate expansion but a redirection from bi/monocyclic to macrocyclic scaffolds. More substantial results emerged from F77 substitutions, where systematic mutations (F77G/A/V/L/I/M/S/T/C) all enabled GGPP acceptance with yields ranging from 6.8 ± 0.4% (F77C) to 85 ± 6% (F77I). The substrate range was expanded further by smaller substitutions (F77G/A/V/S/T), which enabled the enzyme to accept GFPP (C25), yielding macrocyclic sesterterpenes with yields up to 11 ± 2% (F77A). The mechanism involves creating additional space to accommodate larger substrates while maintaining critical catalytic interactions, with products showing substrate-dependent stereoselectivity (14R for GGPP products, 14S for GFPP products).⁹⁶

Fig. 10 Expanding substrate scope and altering product skeletons in SvES through active site cavity engineering. (A) SvES WT converts FPP to sesquiterpene products including hedycaryol and 7-epi-α-eudesmol. (B) SvES A76G and F77 variants accept GGPP to produce macrocyclic diterpene products cembrene A and nephthenol. (C) SvES F77 variants accept GFPP to produce macrocyclic cericerene derivatives.

Complementing this expansion strategy, the fungal linalool synthase (Ap.LS) from Agrocybe pediades demonstrates how active site reshaping can simultaneously transform substrate preference and reaction mechanism. Resolution of the first crystal structure of Ap.LS at 1.99 Å (PDB ID: 8GY0) revealed a key constriction in the substrate tunnel formed by the Y299-S184-M77 trio, explaining the enzyme's native specificity for GPP and exclusive production of linear linalool. Notably, this enzyme exhibits high efficiency—44-fold and 287-fold higher than bacterial and plant counterparts (Fig. 11).⁹⁷

Fig. 11 Transforming substrate specificity and cyclization mechanism in Ap.LS from Agrocybe pediades. (A) Ap.LS WT converts GPP to linear linalool through geranyl and linalyl⁺ intermediates. (B) Ap.LS Y299 variants accept FPP as substrate, producing diverse products including linear nerolidol, monocyclic germacrene A and β-elemene, and bicyclic α-selinene and β-selinene through different cyclization pathways.

Saturation mutagenesis at position Y299 produced mechanistic transformation. Several variants (Y299A/C/G/Q/S) gained the ability to accept FPP as substrate and altered cyclization pathways, producing bicyclic α-selinene, bicyclic β-selinene, and monocyclic β-elemene. Notably, β-elemene is formed via a heat-induced Cope rearrangement of the initial enzymatic product, germacrene A, during analysis. The Y299G variant additionally generated linear nerolidol.⁹⁷ This represents a conversion from a highly specific monoterpene synthase (MTS) producing exclusively linear products to a promiscuous STS generating diverse cyclic scaffolds. Computational analysis revealed that the Y299A mutation significantly reduced FPP torsion strain energy (30.42 vs. 45.36 kcal mol⁻¹ in WT), creating space for the longer C15 chain while altering carbocation stabilization patterns. The biological relevance was confirmed in homologous enzymes Ap.LNS and Hf.LS, where equivalent mutations produced similar transformations.⁹⁷

Unlike traditional engineering that modifies residues directly interacting with intermediates, this approach physically reconfigures the catalytic environment to enable entirely new substrate capabilities and reaction modes. The ability to transform cyclization patterns from bicyclic to macrocyclic scaffolds (e.g., SvES) or convert linear monoterpene synthesis into cyclizing sesquiterpene production (e.g., Ap.LS) establishes active site cavity engineering as a transformative method for expanding terpene scaffold diversity.⁹⁷

2.4.2. Fine-tuning cavity architecture for complex scaffolds. While broad cavity expansion enables substrate scope transformation, precision architectural control offers exceptional fine-tuning capabilities for scaffold diversification within established reaction frameworks.⁹⁸ This methodology employs volumetric adjustments and selective sculpting of stabilizing elements to redirect cyclization pathways at critical decision points, enabling predictable control over complex polycyclic architectures.⁹⁹ Rather than wholesale cavity reshaping, these methods focus on specific spatial relationships that govern carbocation intermediate stability and reaction trajectory selection.¹⁰⁰

Systematic volume calibration exemplifies this precision approach, as demonstrated in the polytrichastrene synthase (CpPS) from Chryseobacterium polytrichastri.¹⁰¹ Originally mischaracterized as a linalool synthase, this enzyme was revealed to be a DTS producing tetracyclic polytrichastrene A and tricyclic polytrichastrol A featuring an unusual ethyl group in terpenes, alongside minor products wanju-2,5-diene and thunbergol. Homology-based analysis using the SdS crystal structure (PDB ID: 4OKM) as template identified key active site positions that control product specificity (Fig. 12).⁷⁴ The I66F substitution, designed to introduce an aromatic residue common in other DTSs, dramatically transformed the product profile. This single mutation generated five new compounds, including three with entirely new skeletons—the tetracyclic polytrichastrol B and polytrichastrene B, and the tricyclic bonnadiene-type bonn-2-en-11α-ol—along with wanju scaffold variants wanju-2,6-diene and wanju-2-en-6α-ol. The mutation achieved 195 ± 43% of WT activity, demonstrating that product diversity can be enhanced without sacrificing catalytic efficiency. Other adjustments—A87T, A192V, and their combination—shifted production toward the tricyclic wanju-2,5-diene scaffold while maintaining high activity levels (108–176% of WT).¹⁰¹

Fig. 12 Scaffold diversity generation through systematic volume calibration of the CpPS active site. CpPS WT produces tetracyclic polytrichastrene A and tricyclic polytrichastrol A along with tricyclic wanju-2,5-diene and thunbergol. CpPS I66F mutation (blue) generates alternative tetracyclic (polytrichastrene B, polytrichastrol B) and tricyclic (bonn-2-en-11α-ol) scaffolds. CpPS A192V mutation (orange) shifts production toward wanju-type scaffolds.

Sesterterpene synthases (StTSs) exemplify how architectural sculpting enables the formation of complex polycyclic systems.¹⁰² Unlike broad pocket modifications, this approach targets the three-dimensional arrangement of stabilizing residues to direct intricate macrocyclic architectures unique to larger terpene frameworks. Crystallographic analysis of fungal StTSs, NfSS (PDB ID: 8YLA) and PbSS (PDB ID: 8YL9), identified key aromatic residues (F191, F196, Y160, W164 in NfSS) forming cation-π interactions and aliphatic residues (T88, I92) facilitating hydrophobic interactions with carbocation intermediates (Fig. 13).¹⁰² Site-directed mutagenesis at these positions transformed the original 5/8/6/5 tetracyclic system into diverse scaffolds: the NfSS T88A and I92A alternatively yielded 5/10/5 tricyclic and 5/15 bicyclic skeletons. In PbSS, residue F102 controlled initial cyclization mode, with PbSS F102A redirecting C1–C15 linkage to C1–C14 linkage, yielding 14-membered ring systems. Notably, mutations of residues not directly involved in catalysis (PbSS W169A) simultaneously produced multiple scaffolds while maintaining WT activity levels, demonstrating the high sensitivity of complex polycyclic formation to architectural perturbations.¹⁰²

Fig. 13 Sculpting active site architecture in fungal sesterterpene synthases (StTSs) for product diversification. (A) Product diversification from NfSS showing transformation of the native 5/8/6/5 tetracyclic system to alternative scaffolds through T88A (red arrows) and I92A (blue arrows) mutations. (B) NfSS active site structure highlighting key aromatic and aliphatic residues. (C) Product diversification from PbSS showing products from F102A (red arrows) and W169A (blue arrows) mutants. (D) PbSS active site structure showing arrangement of stabilizing residues around substrate analog BTAC.

Supporting these engineerxing advances, structural studies of bacterial MTSs reveal critical determinants that could guide future precision modifications.¹⁰³ Structure analysis of bCinS (PDB IDs: 5NX6, 5NX7) and bLinS (PDB IDs: 5NX4, 5NX5) demonstrates how specific residue pairs control substrate acceptance: aromatic residues F77/F179 in bCinS constrict the active site for GPP and bicyclic products, while non-aromatic residues T75/C177 in bLinS create larger cavities accommodating both GPP and FPP for acyclic products. These structure–function relationships establish clear predictions for rational engineering: modifying aromatic residues to smaller variants could expand substrate scope, while introducing bulky aromatics might constrain cavities to favor cyclization.¹⁰³

2.4.3. Backbone engineering for skeletal rearrangement. Complementing aromatic network engineering, backbone carbonyl modification harnesses the catalytic potential of the protein main chain itself.¹⁰⁴ Multiscale QM/MM simulations of DTS EfTPS14 from Euphorbia fischeriana uncovered an unexpected mechanism where the main-chain carbonyl of I644 plays a dual acid-base role in mediating crucial proton abstraction-reprotonation sequences. This high-fidelity enzyme produces 93.4% ent-neoabietadiene (tricyclic ent-abietane-type diterpene), but computational analysis revealed how targeting backbone functionality could redirect this pathway.¹⁰⁵

The I644A mutation, designed to alter backbone catalytic function, completely transformed product specificity from ent-abietane to ent-pimarane skeleton, yielding exclusively ent-sandaracopimaradiene while maintaining catalytic efficiency (Fig. 14). QM/MM calculations revealed that EfTPS14 WT facilitates a concerted but asynchronous process (8.4 kcal mol⁻¹ barrier) where the I644 carbonyl mediates deprotonation at C14 followed by reprotonation at C16 and methyl migration. The I644A variant retained only deprotonation function while losing proton transfer capability, effectively arresting the reaction at an intermediate stage to produce ent-pimarane rather than ent-abietane skeletons. Additional computational optimization through the L765F/L762S double mutant redirected deprotonation regioselectivity to exclusively produce ent-abietadiene, demonstrating how distal residue modifications can fine-tune reaction outcomes.¹⁰⁵

Fig. 14 Skeletal interconversion in EfTPS14 via backbone carbonyl engineering. EfTPS14 WT produces tricyclic ent-neoabietadiene through complete cascade including methyl migration. EfTPS14 I644A disrupts proton transfer, yielding ent-sandaracopimaradiene (ent-pimarane skeleton). EfTPS14 L765F/L762S redirects final deprotonation to produce ent-abietadiene (ent-abietane skeleton).

By targeting protein backbone elements and aromatic transport networks through computational prediction, these strategies access control mechanisms previously overlooked in traditional protein design, enabling skeletal interconversions that would be challenging to achieve through conventional active site engineering approaches.¹⁰⁵

3. Evolutionary and comparative approaches

3.1. Phylogeny-guided mutation

The evolutionary trajectory of class I TSs provides a rich blueprint for rational enzyme engineering, as sequence divergence patterns across phylogenetic lineages directly correlate with functional diversification in cyclization modes and product scaffolds.^106–108 By systematically analyzing conserved and variable residues between enzymes with distinct product specificities, critical “plasticity hotspots” can be identified that serve as molecular switches controlling carbocation cascade fate.¹⁰⁹ This phylogeny-guided approach enables the precise manipulation of key evolutionary residues to redirect cyclization pathways, achieving skeletal transformations through minimal, high-impact mutations that mirror natural evolutionary processes.

The successful application of this strategy is exemplified by the rational engineering of amorpha-4,11-diene synthase (AaADS) from Artemisia annua, where phylogenetic analysis of six related STSs with distinct cyclization patterns guided site-directed mutagenesis (Fig. 15).¹¹⁰ Sequence alignment between ADS (bicyclic amorpha-4,11-diene producer), BOS (monocyclic α-bisabolol producer), and GAS (monocyclic germacrene A producer) identified differential residues that control first and second cyclization events. The T296V mutation, designed based on the finding that valine occupies this position in acyclic (E)-β-farnesene synthase, substantially shifted production from the native bicyclic scaffold to 93.7% linear (E)-β-farnesene. Structural modeling revealed that T296's hydroxyl group stabilizes the farnesyl⁺ during ionization, while valine's hydrophobic character disrupts this critical interaction, causing premature cascade termination. The systematic L374Y/L404V/L405I/G439S tetra-substitution, based on residues found in GAS, blocked the second cyclization through steric hindrance, yielding 80% monocyclic bisabolyl-type products including 61.5% cis-γ-bisabolene. Significantly, five of the seven critical residues (T296, L374, G439, T399, and T447) are located adjacent to conserved metal-binding motifs, establishing these regions as crucial leverage points for scaffold engineering.¹¹⁰

Fig. 15 Engineering ADS based on phylogenetic insights. (A) Sequence alignment of three functionally distinct STSs: ADS, GAS, and BOS. Red stars indicate key plasticity hotspots. (B) Reaction scheme showing skeletal reprogramming of ADS. ADS WT produces bicyclic amorpha-4,11-diene. AaADS T296V yields linear (E)-β-farnesene. AaADS L374Y/L404V/L405I/G439S produces monocyclic products including cis-γ-bisabolene and α/β-bisabolol.

The evolutionary hypothesis is that modern 1,10-cyclases evolved from promiscuous 1,11-ancestors through gene duplication. This hypothesis was experimentally validated using germacrene A synthase (GAS) from Solidago canadensis.¹¹¹ This enzyme naturally produces >96% monocyclic germacrene A (1,10-cyclization) and only ∼2% α-humulene (1,11-cyclization) (Fig. 16). Systematic substitution of G402 with amino acids of increasing steric bulk progressively shifted product distribution toward 1,11-cyclization products, with the G402C variant producing 62.5% α-humulene while maintaining half of WT catalytic efficiency. Parallel mutations in the homologous germacrene D synthase (GDS G404V) generated ∼20% bicyclic bicyclogermacrene, a unique cyclopropane-containing scaffold that structurally links both cyclization pathways. Mechanistic studies using [1-³H₁]-10-fluoro-FPP and deuterium-labeled [12,13-²H₆]-FPP revealed a germacrene-humulene rearrangement connecting the 1,10- and 1,11-pathways, likely through a bridged 1,10,11-carbocation intermediate. Kinetic isotope effects (k_H/k_D = 4.98 for GAS WT, 4.29–5.09 for G402 variants) indicated that G402 functions as an evolutionary ‘plasticity residue’ that can redirect cyclization preference while maintaining substantial catalytic activity.¹¹¹

Fig. 16 Reverting GAS to ancestral 1,11-cyclization function. GAS WT produces germacrene A (>96%) via 1,10-cyclization and minor α-humulene (∼2%) via 1,11-cyclization. GAS G402C shifts to 62.5% α-humulene production by favoring the 1,11-cyclization pathway.

Comprehensive phylogenetic analysis across plant STSs revealed that a single preNSE/DTE residue—the amino acid positioned four residues upstream of the conserved NSE/DTE ion-binding motif—functions as a conserved molecular switch between 1,10- and 1,11-cyclization modes (Fig. 17).¹¹² Statistical analysis demonstrated that 1,10-cyclases predominantly contain hydrophilic residues (C, S, G, T) at C440, while 1,11-cyclases harbor hydrophobic residues (A, L, I, V). Single-point mutations effectively reversed cyclization specificity: the C438G in SmSTPS3 shifted production from bicyclic 5-epi-eremophilene to 54.7% monocyclic elemol, while S454G in AaGAS transformed the profile from β-elemene to 74.4% elemol. Additionally, S441G mutation in AtTPS21 converted the enzyme from producing 92.9% bicyclic β-caryophyllene (9/4 bicyclic ring system) to 69.2% monocyclic germacrene A-11-ol (10-membered ring). Crystal structure analysis from tobacco epi-aristolochene synthase TEAS (PDB ID: 5EAU)²⁸ revealed this residue's proximity to the FPP isopropyl tail, suggesting it controls substrate folding to determine whether C1 attacks in a Markovnikov (1,10-cyclization) or anti-Markovnikov (1,11-cyclization) manner.¹¹²

Fig. 17 Role of the PreNSE/DTE residue in determining cyclization modes of plant STSs. Single point mutations reverse cyclization preference: AtTPS21 WT produces bicyclic β-caryophyllene, while AtTPS21 S441 G produces monocyclic germacrene A-11-ol; SmSTPS3 WT produces bicyclic (−)-5-epi-eremophilene, while SmSTPS3 C438 G and AaGAS S454G produce monocyclic elemol.

The integration of phylogenetic analysis with high-resolution crystallography enabled precise engineering of 1,8-cineole synthase (Sf-CinS1) from Salvia fruticosa, whose 1.95 Å (PDB ID: 2J5C) crystal structure identified N338 as a critical plasticity hotspot.¹¹³ This residue activates water for nucleophilic attack on the α-terpinyl⁺ intermediate, leading to ether formation in 1,8-cineole. The N338I mutation eliminated water activation, transforming the enzyme from producing primarily cyclic ether (72.4% 1,8-cineole) to bicyclic sabinene (48.3%) and monocyclic limonene (37%) (Fig. 18A). Optimization via the N338I/A339T/G447S/I449P/P450T quintuple mutation yielded an enzyme producing 86.8% sabinene. Notably, the N338A mutation enlarged the active site cavity sufficiently to accommodate FPP, converting the MTS into a STS producing 49% bicyclic trans-α-bergamotene (Fig. 18B), demonstrating how key evolutionary residues can control not only product scaffold but also substrate specificity.¹¹³

Fig. 18 Control of product profile and substrate specificity by a single residue switch in Sf-Cins1. Product outcomes controlled by N338 mutations in the GPP cyclization cascade. Sf-Cins1 WT yields bicyclic 1,8-cineole via water capture at α-terpinyl⁺. Sf-Cins1 N338I variant shifts to sabinene/limonene production by preventing water activation. Quintuple mutation Sf-Cins1 N338I/A339T/G447S/I449P/P450T optimizes sabinene formation. (B) The N338A mutation enables FPP acceptance, converting the MTS into a STS producing trans-α-bergamotene.

3.2. Mimicking natural variation and functional switches

Natural allelic variation represents evolution's most elegant solution for generating functional diversity through minimal genetic modification.¹¹⁴ Closely related enzyme variants that differ by only a few amino acids yet produce different product profiles provide exceptional opportunities to identify critical molecular switches controlling terpene scaffold formation. These naturally occurring “minimal intervention” experiments provide direct insights into the precise residues responsible for cyclization pathway control, allowing evolutionary divergence to be recreated through rational mutagenesis. This approach leverages the accumulated wisdom of natural selection to achieve maximal functional transformation with minimal structural perturbation, revealing how subtle changes in active site architecture can fundamentally redirect carbocation cascade trajectories.

3.2.1. Evolution-guided single residue switches in diterpene synthases. Comparative analysis of functionally divergent orthologs and paralogs has been particularly fruitful in engineering plant class I diterpene synthases (DTSs). A foundational example involves the identification of a single residue switch that controls the cyclization depth in ent-kaurene synthases (KSs) and related KS-like (KSL) enzymes. By comparing two rice enzymes—OsKSL5i, which produces the tetracyclic ent-isokaurene, and OsKSL5j, which yields the tricyclic ent-pimaradiene—researchers noted that despite sharing 98% amino acid identity, they exhibited distinct product specificities.¹¹⁵ Through structural modeling and mutagenesis, a critical isoleucine/threonine (I/T) switch was identified. The substitution of the conserved isoleucine with threonine (e.g., I664T in OsKSL5i) functions as a molecular “short-circuit,” prematurely terminating the carbocation cascade at the tricyclic ent-pimarenyl⁺ to produce ent-pimaradiene (Fig. 19A).¹¹⁵ Subsequent research extended the evolutionary and mechanistic scope of this discovery, revealing that this critical isoleucine is conserved across all Embryophyta.¹¹⁶ Analogous substitutions successfully converted not only angiosperm KSs but also the monofunctional MpKS from liverwort and the bifunctional PpCPS/KS from moss into tricyclic DTSs. Mechanistically, substituting isoleucine with alanine also results in scaffold abbreviation, attributed to the creation of a cavity that accommodates a water molecule to electrostatically stabilize the cation intermediate.¹¹⁶

Fig. 19 Isoleucine/threonine (I/T) switch in diterpene synthases. (A) The I/T switch in ent-CPP-utilizing KS/KSL enzymes terminates the cascade at tricyclic pimaradienes; secondary L/V substitutions provide additional tuning. (B) In (+)-CPP-utilizing diterpene synthases, the switch controls hydroxylation (IrTPS2, nezukol) versus deprotonation (IrKSL3a, isopimaradiene).

Further refinement of this system identified a secondary residue (typically leucine or valine) adjacent to the primary I/T switch that interactively affects product outcome.¹¹⁷ This secondary residue exerts an epistatic or additive effect, essentially “tuning” the active site cavity. For instance, in OsKSL5, pairing the primary reversal T664I with the secondary substitution L661V was required to fully restore specific tetracyclic ent-isokaurene production. This additive interactivity was further demonstrated in castor bean (Ricinus communis) KS enzymes, where double mutations significantly abbreviated cyclization cascades. The L633S/I636T substitution in RcKSL2 converted the native pentacyclic ent-trachylobane to a mixture of simplified tricyclic and bicyclic products, while the L629S/V632T substitution in RcKSL4 similarly redirected catalysis from tetracyclic ent-beyerene to simplified scaffolds (Fig. 19A). Moreover, rational combination of the secondary switch mutation L635V with the primary site mutation I638S in AtKS successfully engineered the predominant production of a novel hydroxylated diterpene, 8α-hydroxy-ent-pimar-15-ene, by likely relaxing active site confinement to better accommodate the attacking water molecule.¹¹⁷

Extending the I/T paradigm established by these foundational studies, the evolutionary principles governing residue switches in DTSs were found to extend beyond cyclization depth control to encompass the critical decision between hydroxylation and deprotonation pathways. Two highly homologous enzymes (98% identity) from Isodon rubescens, IrTPS2 and IrKSL3a, produce different products from the same substrate (+)-copalyl diphosphate despite their near-identical sequences. Position 522/523 emerged as important, with IrKSL3a exclusively generating isopimaradiene while IrTPS2 produces the hydroxylated nezukol. This functional divergence mirrors the I/T switch paradigm discovered in KS enzymes, wherein residues containing β-methyl groups versus hydroxyl-bearing residues determine reaction trajectory. Systematic analysis revealed that residues containing β-methyl groups control product formation through selective occlusion of water binding sites. QM/MM simulations demonstrated that when A523 in IrTPS2 is mutated to isoleucine, the β-methyl physically blocks water binding to S499/T527, preventing hydroxylation and redirecting catalysis toward deprotonation pathways (Fig. 19B). Conversely, the reciprocal mutation in IrKSL3a (I522A) alone produced minimal nezukol, but combining it with A498S to restore the water-binding dyad (IrKSL3a I522A/A498S) resulted in a 7.9-fold increase in nezukol production, demonstrating that both the absence of steric hindrance and the presence of hydrogen-bonding residues are required for hydroxylation. Notably, T527 in IrTPS2 is highly conserved across Lamiaceae DTSs, paralleling the conservation of threonine residues in the classical I/T switch and suggesting its fundamental role in water-mediated hydroxylation. This steric effect alters the reaction trajectory without disrupting overall active site architecture. Computational analysis quantified these effects through QM/MM free energy calculations: water addition in IrTPS2 is favorable (exothermic by −11.7 kcal mol⁻¹) with virtually no energy barrier, while deprotonation in IrKSL3a requires overcoming a substantial 17.0 kcal mol⁻¹ energy barrier. These energetic differences explain how single amino acid variations translate to different product distributions, demonstrating that the β-methyl versus hydroxyl-bearing residue principle operates not only in controlling cyclization depth (as shown in KS/KSL systems) but also in governing terminal quenching chemistry through modulation of the local chemical environment. The requirement for multiple coordinated mutations (I522A/A498S) to efficiently convert IrKSL3a function underscores the extended evolutionary process underlying the specialized production of hydroxylated diterpenes.¹⁰⁹

3.2.2. Exploiting minimal allelic differences for maximal functional divergence. The efficiency of this minimal intervention strategy is exemplified by grapevine VvTPS24 allelic homologs (i.e. VvGuaS and VvPNSeInt), which share 99.5% sequence identity and differ in only 6 amino acids, produce completely distinct sesquiterpene scaffolds.¹¹⁸ VvGuaS generates primarily 5/7 bicyclic sesquiterpenes α-guaiene (44%) and α-bulnesene (35%), while VvPNSeInt produces predominantly 6/6 bicyclic selina-4,11-diene (34%) and intermedeol (30%) from identical FPP. Homology modeling identified just two polymorphisms (T414S and V530M) located within 4 Å of the substrate binding site as the critical determinants of this cyclization divergence. Systematic mutagenesis revealed their cooperative mechanism: T414S maintained predominant α-guaiene production (41%) with minor shifts toward selinene-type products, while V530M substantially reduced α-guaiene (24%) with concomitant increase in selina-4,11-diene (24%) (Fig. 20). This redirection was most pronounced in the T414S/V530M double mutant, which almost completely redirected cyclization toward the selinene scaffold, reducing α-guaiene to merely 5% while increasing selina-4,11-diene to 40%, effectively converting VvGuaS into VvPNSeInt. The mechanistic basis involves methionine's sulfur atom stabilizing different carbocation intermediates through lone pair electrons, while steric changes redirect the second ring closure from a 6,2-cyclization (forming the 5/7 bicyclic guaiene skeleton) to a 7,2-cyclization (yielding the 6/6 bicyclic selinene scaffold).¹¹⁸

Fig. 20 Key polymorphisms for product specificity revealed by allelic variants of grapevine TPS24. VvGuaS WT produces 5/7 bicyclic α-guaiene and α-bulnesene. VvPNSeInt WT generates 6/6 bicyclic selina-4(14),11-diene. Single mutations (T414S or V530M) partially shift product profiles, while the double mutant T414S/V530M achieves complete functional conversion between scaffolds.

A parallel example of minimal intervention engineering is found in maize TSs TPS4 and TPS5, which share 98% sequence identity yet produce distinct product distributions.¹¹⁹ TPS5-Del1 yields sesquithujene (28.2%, bicyclic), (S)-β-bisabolene (26.6%, monocyclic), and (E)-β-farnesene (13.2%, acyclic), while TPS4-B73 produces 7-epi-sesquithujene (24.4%, bicyclic), (S)-β-bisabolene (29.1%, monocyclic), and (E)-β-farnesene (9.2%, acyclic). Comparative analysis identified positions 407–411 as key determinants, with the single G409A mutation altering product distribution by favoring (S)-β-bisabolene (monocyclic) production while reducing bicyclic compounds to barely detectable levels (Fig. 21). Adding the A410T mutation partially restored bicyclic product formation, demonstrating the interplay between adjacent residues. Structural modeling located position 409 within a conserved helix G kink at the active site bottom, where the additional methyl group in alanine versus glycine alters the formation rates of (R)- and (S)-bisabolyl⁺ intermediates. This subtle change explains why TPS5 derives approximately 95% of products from the (S)-bisabolyl⁺ pathway, while TPS4 utilizes both pathways equally, creating their divergent product profiles despite minimal sequence differences.¹¹⁹

Fig. 21 Influence of active site kink residue on product ratios in maize allelic TPS4 and TPS5. TPS4 WT produces 7-epi-sesquithujene (bicyclic product) while TPS5 WT yields sesquithujene. Both enzymes produce (S)-β-bisabolene (monocyclic product). TPS5 G409A shifts product distribution toward (S)-β-bisabolene by altering bisabolyl⁺ intermediate formation rates.

3.2.3. Systematic active site remodeling through multi-residue networks. While minimal allelic differences can achieve functional transformations, more divergent enzyme pairs often require comprehensive active site remodeling involving coordinated changes across multiple residues. This systematic engineering approach recognizes that terpene cyclization specificity emerges from complex networks of residue interactions rather than single molecular switches. Identification and manipulation of these multi-residue networks—including both substrate-contacting positions and second-shell structural elements—enables functional conversions between evolutionarily distant paralogs. This strategy encompasses three distinct but interconnected approaches: comprehensive active site substitution, identification of conserved functional motifs, and mapping epistatic networks that control the emergence of new catalytic capabilities.

The complexity of multi-residue engineering is illustrated by maize TS paralogs TPS4 and TPS10, which exhibit substantial evolutionary divergence (amino acid similarity of 57%) compared to the minimal differences seen in allelic variants.¹²⁰ TPS4 naturally converts FPP into bicyclic 7-epi-sesquithujene and monocyclic β-bisabolene, whereas TPS10 produces bicyclic (E)-α-bergamotene and acyclic (E)-β-farnesene from the same substrate. Sequence comparison combined with homology modeling identified 17 amino acid differences within the active site region between these enzymes (Fig. 22). Systematic mutagenesis revealed varying impacts on product specificity, with the Y382S substitution notably redirecting catalysis toward (E)-β-farnesene production. While the comprehensive TPS4-c17 variant incorporating all 17 active site mutations achieved only partial functional conversion (61.5% (E)-β-farnesene and 14.4% (E)-α-bergamotene), complete functional mimicry required two additional second-shell substitutions, I411F and R442K. The resulting enzyme (TPS4-c17 R442K/I411F) produced 56.5% (E)-β-farnesene and 21.0% (E)-α-bergamotene, approaching TPS10's natural distribution of 50.4% and 35.9%, respectively. Structural examination from homology modeling revealed that despite pointing away from the substrate-binding cavity, I411F at the N-terminal end of helix G2 indirectly influences active site geometry, while R442K affects the positioning of the G1 helix, demonstrating how secondary structural elements modulate active site conformation and product outcomes.¹²⁰

Fig. 22 Functional conversion between TPS4 and TPS10. TPS4 WT yields bicyclic 7-epi-sesquithujene and monocyclic β-bisabolene, while TPS10 produces structurally distinct bicyclic (E)-α-bergamotene and acyclic (E)-β-farnesene. TPS4-c17 variant partially converts the original bicyclic sesquithujene scaffold toward acyclic farnesene. Additional second-shell mutations (TPS4-c17 R442K/I411F) complete the skeletal transformation, shifting from sesquithujene/bisabolene scaffolds to bergamotene/farnesene frameworks characteristic of TPS10.

A more refined approach to multi-residue engineering is exemplified by A. annua TSs AaBOS and AaADS. Despite sharing 92% nucleotide sequence identity, AaBOS yields monocyclic α-bisabolol (92.8%) while AaADS generates bicyclic amorpha-4,11-diene as its major product. Through systematic domain-swapping and mutagenesis, a distinctive “BOS motif” (named after BisabOlol Synthase) comprising four key residues (V373, I395, N398, L399) that collectively control cyclization pathways was identified. These spatially proximal residues—with V373 on helix F and I395/N398/L399 on helix G1—form a functional unit within the substrate-binding pocket that acts as a molecular switch for product determination.⁴⁸ Engineering efforts on AaBOS produced variant AaBOS-M2 (V373N/L381A/I395V/N398I/L399T) that generated 68.8% γ-humulene, a monocyclic product with alternative ring closure pattern, representing significant improvement over the tetra-substituted AaBOS-M1 (V373N/I395V/N398I/L399T), which produced 29.3% γ-humulene (Fig. 23). Crystal structure analysis (PDB IDs: 4FJQ for AaBOS, 4GAX for AaBOS-M2) revealed specific roles for each plasticity residue: the L399T substitution proved essential for γ-humulene formation, with its hydroxyl group promoting 1,11-ring closure through altered carbocation stabilization, while L381A doubled γ-humulene production by reducing steric hindrance at the pocket rim. Reciprocal engineering in AaADS demonstrated the broader applicability of this approach, where the T399S mutation increased amorpha-4,11-diene conversion by 70% and catalytic efficiency (k_cat/K_m) by 71.7% through optimized active site hydrophilicity—demonstrating the potential of plasticity residue substitution for improving enzymatic performance.⁴⁸

Fig. 23 Identifying plasticity residues in AaBOS and AaADS synthases. AaBOS WT produces monocyclic α-bisabolol via bisabolyl⁺ intermediate. AaBOS-M1 and AaBOS-M2 variants redirect to monocyclic γ-humulene through alternative 1,11-cyclization of cis-humulyl⁺. AaADS WT generates bicyclic amorpha-4,11-diene. Reaction proceeds through transoid/cisoid NPP intermediates and E,Z-farnesyl⁺ before carbocation cyclization.

The most comprehensive multi-residue engineering strategy involves mapping epistatic networks that control fundamental catalytic transformations, as demonstrated in the evolutionary transition from linear to cyclic terpene production in A. annua.¹²¹ Structure-based Combinatorial Protein Engineering (SCOPE) generated an extensive library of ∼27 000 mutants by breeding natural variations from AaADS into (E)-BFS, enabling comprehensive exploration of the functional landscape. From this collection emerged Y402L as a dominant mutation that fundamentally transformed BFS function, producing 15 distinct cyclic terpenes (∼75% of total products) instead of linear compounds while maintaining near WT catalytic efficiency (k_cat/K_m = 5.9 × 10⁻³vs. 10.5 × 10⁻³ µM⁻¹ s⁻¹). Mechanistic analysis revealed that Y402L delays proton elimination and releases the substrate from its unproductive all-trans conformation, enabling 2,3-σ-bond rotation in the neutral nerolidyl diphosphate (NPP) intermediate for subsequent cyclization. Complex epistatic interactions formed the foundation of this functional shift: V467G acts as a second-site suppressor, reverting cyclization to linear products despite being ∼10 Å from Y402L, while Y430 adds further complexity through Y430A reactivating cyclization in the Y402L/V467G background (55% cyclic products) and enhancing cyclization when paired with Y402L alone (87% cyclic products) (Fig. 24). Quantitative fitness landscape analysis identified positions 402-467-430 as a strongly epistatic residue network (roughness/slope ratio = 0.57) collectively controlling cyclization emergence, with the Y402L/Y430A pair representing a fitness maximum within this network.¹²¹

Fig. 24 Y402L-driven product diversification in engineered BFS. BFS WT produces primarily linear (E)-β-farnesene, while the Y402L mutation enables production of 15 distinct products including linear (nerolidol, farnesol, α-farnesene), monocyclic (β-bisabolene, γ-curcumene, zingiberene, α-bisabolol), and bicyclic (β-sesquiphellandrene, ar-curcumene, cis-α-bergamotene, α-exo-bergamotene) scaffolds.

3.2.4. Conservation of molecular switches across phylogenetically distant species. The functional principles governing terpene cyclization extend far beyond individual enzyme families, revealing conserved molecular mechanisms that operate across phylogenetically distant organisms.¹²² Despite the structural diversity of TSs from bacteria, fungi, and plants, certain key residue positions and molecular switches exhibit conservation in their functional roles.^106,123 Investigation of these conserved mechanisms across different biological systems identifies universal engineering targets and validates the broader applicability of specific catalytic strategies. This cross-species approach not only illuminates fundamental principles of terpene cyclization but also facilitates development of generalizable engineering frameworks that transcend taxonomic boundaries.

The concept of gatekeeper residues emerges as a universal mechanism for controlling carbon skeleton formation across diverse enzyme architectures, as demonstrated in fungal haloacid dehalogenase-like terpene cyclases (HAD-TCs) from Antrodia cinnamomea.^124,125 Despite representing a unique enzyme family with distinctive architecture compared to plant TSs, these fungal enzymes employ similar spatial control mechanisms. AncC and AncA share 94% sequence identity yet generate different carbon skeletons: AncC forms a bicyclic drimane-type scaffold ((+)-albicanol) while AncA produces a monocyclic cyclofarnesoid scaffold ((R)-trans-γ-monocyclofarnesol) from identical FPP. Molecular docking combined with AlphaFold2 structural analysis pinpointed just 18 residue differences between their TC domains, with two gatekeeper residues in AncA (Y283/F375) emerging as critical determinants. These residues create a narrower active site pocket that restricts conformational freedom of reaction intermediates, directing cyclization toward monocyclic products. Individual mutations (Y283F or F375L) produced minimal effects, demonstrating cooperative gatekeeper function, while the double mutant (Y283F/F375L) transformed AncA into an efficient (+)-albicanol producer, effectively recreating AncC-like activity (Fig. 25). Domain-swapping experiments confirmed that the TC domain exclusively determines scaffold identity, with kinetic analysis revealing that cyclization occurs more rapidly than subsequent dephosphorylation, establishing specific architectural constraints rather than altered reaction rates as the key determinant.¹²⁴

Fig. 25 Control of skeleton formation by gatekeeper residues in fungal HAD-like terpene cyclases (TCs) AncA and AncC. (A) AncC pathway: class II TC domain cyclizes FPP to albicanyl pyrophosphate, followed by HAD domain dephosphorylation to yield bicyclic drimane-type (+)-albicanol. (B) AncA pathway: class II TC domain forms monocyclofarnesyl pyrophosphate intermediate, with HAD domain dephosphorylation producing monocyclic (R)-trans-γ-monocyclofarnesol. AncA Y283F/F375L double mutant converts to AncC-like activity.

High-resolution structural analysis provides crucial insights into conserved catalytic mechanisms, as exemplified by crystallographic studies of Coniophora puteana δ-cadinol synthases Copu5 and Copu9.¹²⁶ The first crystal structure of Copu9 was determined at 1.83 Å resolution (PDB ID: 7OFL), capturing a catalytically relevant state with three Mg²⁺ ions and substrate analog bound in the active site. This structural snapshot illuminated universal elements of sesquiterpene cyclization: the “effector triad” comprising R190, D191, and S193 in Copu9 (R182, D183, and S185 in Copu5) sits at the C-terminal end of the G1 α-helix and plays a crucial role in substrate ionization, while residues G194 and C195 in Copu9 (G186 and C187 in Copu5) position near the substrate's phosphate moiety to control carbocation formation and stabilization. Structure-guided mutagenesis targeting these conserved elements produced variants—Copu5 C187V, N220A, and S306N, along with Copu9 N228A and S314N—that generated monocyclic germacrene D-4-ol as a notable side product while maintaining (+)-δ-cadinol as the main product (Fig. 26). These mutations affect the active site architecture at the interface between hydrophilic and hydrophobic regions, specifically influencing the reaction trajectory of the germacryl⁺ intermediate and demonstrating how conserved structural motifs control cyclization outcomes across fungal species.¹²⁶

Fig. 26 Structural insights and mutagenesis of fungal δ-cadinol synthases Copu5 and Copu9. (A) Product profiles of WT and mutant. Copu5 and Copu9 WT convert FPP primarily to the bicyclic (+)-δ-cadinol, with minor amounts of α-cadinol, cubebol, and either tau-muurolene or delta-cadinene. Site-directed mutations (Copu5 C187V, N220A, S306N; Copu9 N228A, S314N) introduce formation of the monocyclic germacrene D-4-ol as an additional side product. (B) Crystal structure of Copu9 (PDB ID: 7OFL) showing the characteristic α-helical fold of class I TSs. (C) Active site close-up showing the catalytic machinery. The effector triad (R190, D191, S193), NSE motif residues (N228), and other key residues (G194, C195, S314) coordinate substrate binding and carbocation formation. Green spheres represent the three catalytically essential Mg²⁺ ions, while the bound substrate analog AHD (alendronate, a non-hydrolyzable FPP mimic) is shown in stick representation with its phosphonate groups coordinating the metal cluster.

The conserved nature of aromatic stabilization mechanisms is further demonstrated in bacterial TSs from the thermotolerant organism Rubrobacter radiotolerans, where leucine/phenylalanine switches profoundly influence product specificity through modulation of cation-π interactions.¹²⁷ Homology modeling and molecular docking guided systematic mutagenesis of nerolidol synthase (NerS) and trans-α-bergamotene synthase (BerS), revealing that the BerS L86F substitution approximately doubled (E)-β-farnesene (acyclic) production compared to WT (26.2 vs. 12.5 mg L⁻¹) while enhancing overall terpene yields (Fig. 27). This mutation significantly altered the (E)-β-farnesene:trans-α-bergamotene ratio from ∼17 : 1 to ∼10 : 1 and enabled production of trace amounts of β-bisabolene (monocyclic), a product absent in WT. The introduced aromatic ring at position 86 provides critical cation-π stabilization of reaction intermediates, enhancing substrate isomerization required for product diversification. Conversely, the BerS S82L variant increased product titers but favored non-isomerized products, shifting the ratio from ∼17 : 1 to ∼23 : 1 through steric hindrance that blocks crucial isomerization steps. Complementary evidence from NerS engineering, where F79A mutation reduced nerolidol (acyclic) production approximately 15-fold, underscored the importance of the phenylalanine residue—through both its steric bulk and potential aromatic interactions—in stabilizing isomerized carbocation intermediates. The thermotolerant origin of these enzymes was reflected in their optimal activity at 50 °C (turnover number of 0.38 min⁻¹, more than three times faster than at 30 °C), demonstrating how conserved aromatic stabilization mechanisms operate across diverse environmental conditions.¹²⁷

Fig. 27 Modulation of product outcome in NerS and BerS by a leucine/phenylalanine switch. BerS WT produces bicyclic trans-α-bergamotene and acyclic (E)-β-farnesene. BerS L86F increases (E)-β-farnesene production; BerS S82L shifts ratio toward (E)-β-farnesene. NerS WT and NerS F79A produce acyclic nerolidol, with F79A showing reduced activity. The red dashed box indicates products affected by L/F switching at position 86.

4. Mechanism-focused engineering

4.1. Carbocation intermediate control

Manipulating carbocation intermediates represents a particularly elegant approach to redirecting terpene cyclization cascades.¹²⁸ Unlike strategies focused solely on substrate binding or product release, this method targets the heart of TS chemistry—the reactive carbocation species that define skeletal architecture.¹²⁹ Modification of residues that stabilize specific cationic intermediates enables interception of cyclization pathways at precise BPs, redirecting reaction trajectories toward alternative carbon skeletons.^130,131 This strategy offers exceptional potential for generating structural diversity from common precursors and provides fundamental insights into the catalytic mechanisms underlying terpene cyclization.

Fine-tuning carbocation stabilization patterns offers even more streamlined control over diterpene cyclization, as demonstrated in conifer DTSs where a single hydroxyl-bearing residue determines the fundamental BP between bicyclic and tricyclic skeletal formation.¹³² Sequence alignment of conifer DTSs revealed a crucial pattern: abietadiene-producing enzymes from grand fir (AgAS) and Norway spruce (PaAS) share a conserved alanine at a key position, while the pimaradiene-specific synthase (PaPS) from Norway spruce features serine at the equivalent location. Introduction of this single modification (A723S) into AgAS fundamentally redirected product outcome from >95% tricyclic abietadienes to >95% tricyclic pimaradienes—predominantly isopimara-7,15-diene (75%) and sandaracopimara-8(14),15-diene (21%) (Fig. 28). The high efficiency of this conversion defied conventional expectations, as the A723S mutant maintained comparable catalytic rates to AgAS WT while exhibiting slightly increased specific activity (∼2-fold). Mechanistic analysis revealed that the introduced hydroxyl group selectively stabilizes the pimar-15-en-8-yl⁺ intermediate, creating a kinetic trap that favors direct deprotonation rather than rearrangement to the secondary pimar-8(14)-en-15 yl⁺ species required for progression to abietadienes. This stabilization likely occurs by modifying the electrostatic environment that would otherwise favor continued cyclization driven by ion-pairing with the diphosphate anion released during initial ionization.¹³²

Fig. 28 Control of abietadiene versus pimaradiene formation by a single hydroxyl-bearing residue. The A723S mutation in AgAS redirects cyclization from a tricyclic abietane scaffold to tricyclic pimarane scaffolds by preventing the key rearrangement required to form the abietane skeleton.

Structure-guided engineering of white spruce (Picea glauca) ent-kaurene synthase (PgKS) further demonstrates how steric effects, rather than polar interactions, can critically influence cyclization outcomes.¹³³ Homology modeling based on the T. brevifolia TXS crystal structure combined with molecular docking of ent-CPP identified I619 as a pivotal residue for cyclization control. This residue occupies a pivotal position on helix G2 within 3 Å of C-8 of ent-CPP—precisely where the critical pimarenyl⁺ forms during cyclization. Comprehensive sequence analysis revealed this position as strictly conserved among tetracyclic kaurene-producing enzymes but variable in those generating tricyclic pimaradiene-type products. Introduction of the I619A mutation fundamentally altered product outcome, redirecting cyclization from predominantly ent-kaurene (tetracyclic scaffold) to ent-pimaradiene (tricyclic scaffold) with 85% selectivity for the new product, nearly eliminating native ent-kaurene production (reduced to 0.3%). Molecular docking revealed that the native I619 residue prevents premature quenching of the pimaren-8-yl⁺ intermediate, allowing progression to the beyeran-16 yl⁺ intermediate and ultimately to tetracyclic ent-kaurene (Fig. 29). When replaced with alanine, the smaller side chain creates an environment permitting premature deprotonation of the pimarenyl⁺, effectively trapping the reaction at the tricyclic stage.¹³³

Fig. 29 Steric control of the pimarenyl⁺ fate in PgKS. The PgKS I619A mutation creates space for premature deprotonation at the pimaren-8-yl⁺ intermediate, redirecting the cascade from tetracyclic ent-kaurene to tricyclic pimaradiene.

4.2. Reaction BP control

Reaction BP control offers a comprehensive approach to redirecting terpene cyclization by focusing on decision points where reaction trajectories diverge.^134,135 This strategy maps the complete energetic landscape of cyclization pathways, identifying specific junctures where selective mutations can selectively favor desired transformations. Manipulation of active site geometries and electrostatic environments at these BPs systematically converts a single enzyme into variants with diverse product profiles, often achieving higher product specificity than through traditional approaches targeting isolated residues.

4.2.1. Computational design of multi-branch reaction networks. Computational design approaches enable systematic engineering of TSs for controlling cyclization outcomes, with the capability to map complete reaction networks and manipulate multiple decision points concurrently.¹³⁶ Advanced computational methods such as QM/MM simulations and structure-guided quantum chemical calculations identify and target the full spectrum of BPs within complex cyclization cascades.^137,138 This comprehensive approach allows for precise control over reaction trajectories, achieving transformations in product specificity while maintaining catalytic efficiency.¹³⁹ This integration of predictive modeling and experimental feedback establishes a robust methodology for transforming a single synthase into a suite of specialist biocatalysts, each optimized for a specific product.

Computational reaction BP mapping enabled the systematic redesign of MTS HCinS, with QM/MM simulations identifying and manipulating three critical decision points within its catalytic cascade.¹⁴⁰ Detailed computational analysis revealed the complete reaction network: BP1 governing initial cyclization of linalyl⁺, BP2 determining whether (R)-α-terpinyl⁺ undergoes hydroxylation or deprotonation, and BP3 controlling cyclization of (R)-α-terpineol (Fig. 30). Native HCinS predominantly produces 1,8-cineole (bicyclic terpene) with 87.4% specificity via a specific trajectory through these BPs. Engineering efforts at each BP generated variants with altered product profiles through precise manipulation of active site geometry and energetics. At BP3, the F236M mutation expanded the spatial distance between the side chains of F236 and F268 from 8.9 Å to 9.8 Å, increasing the cyclization energy barrier from 12.3 to 14.2 kcal mol⁻¹ and redirecting product formation toward (R)-α-terpineol (monocyclic hydroxylated terpene) with 55.1% specificity. Intervention at BP2 employed the N135H substitution to replace a hydroxylation-promoting asparagine with histidine, facilitating deprotonation and generating (R)-limonene (monocyclic terpene) with 87.6% specificity by altering the competitive balance between hydroxylation and deprotonation pathways. The T111A mutation targeted BP1, widening the binding cavity from 9.9 Å to 11.9 Å and allowing the substrate to maintain a Z-shaped conformation rather than adopting the U-shape required for cyclization, effectively preventing ring formation and producing myrcene (acyclic terpene) with 70.8% specificity. These engineered variants retained 55–95% of native HCinS's catalytic efficiency while achieving product specificity switching. When integrated into strains with enhanced GPP supply and increased gene dosage, fed-batch fermentations yielded 6.0 g L⁻¹ for 1,8-cineole, 4.3 g L⁻¹ for myrcene, 4.2 g L⁻¹ for (R)-limonene, and 3.8 g L⁻¹ for (R)-α-terpineol in 5L bioreactors—the highest reported titers for these compounds. Furthermore, the principles of this strategy were applied to phylogenetically distant TSs, suggesting its potential for broad applicability.¹⁴⁰

Fig. 30 Computational redesign of HCinS at multiple reaction branch points (BPs). The computational redesign of the monoterpene synthase HCinS at three distinct BPs enables the divergent production of four different terpene skeletons from a single enzyme. Targeted mutations at BP1 (T111), BP2 (N135), and BP3 (F236) systematically reprogram the cyclization cascade to selectively yield acyclic (myrcene), monocyclic (limonene, α-terpineol), or bicyclic (1,8-cineole) products.

Structure-guided quantum chemical design enables even more refined control over reaction BPs, as demonstrated in the engineering of bisabolene synthase (SydA) where crystal structure analysis (PDB ID: 8YZU) combined with computational modeling revealed and manipulated novel mechanistic pathways.¹⁴¹ Comparative analyses of active site contours with seven other TSs guided the identification of D99, D222, and W297 as critical determinants of product specificity at distinct BPs within the cyclization cascade. The D99A variant displayed exceptional product diversity, generating six different sesquiterpenes with monocyclic, bicyclic, and tricyclic structures, while functioning as a gatekeeper residue modulating active site pocket size via Mg²⁺ binding (Fig. 31). The D222K mutation achieved precise control over a newly identified 1,7-hydride shift BP, predominantly producing bicyclic α-cuprenene (66.7%) with reduced formation of the original monocyclic product (22.8%) and minor amounts of bicyclic β-chamigrene (10.5%). This precise control was further refined in the D222A/T185D/F187V triple mutant (SydA-M3), which achieved 90.3% selectivity for bicyclic α-cuprenene with complete elimination of the original product, while the W297Y variant significantly increased the formation of tricyclic 7-epi-β-cedrene from 6.1% in D99A to 40.5% (a 6.64-fold increase) by affecting base positioning critical for deprotonation steps (Fig. 31). Quantum chemical calculations and molecular docking studies revealed that these mutations alter the enzyme's ability to stabilize different carbocation intermediates and guide reaction pathways, with the catalytic mechanism involving FPP cyclization to the bisabolyl⁺ followed by divergent reaction pathways including the unique 1,7-hydride shift confirmed through deuterium-labeled FPP experiments.¹⁴¹

Fig. 31 Engineering a novel hydride shift BP in bisabolene synthase. (A) Systematic mutagenesis of SydA redirects the monocyclic bisabolyl⁺ intermediate toward diverse scaffolds: D222K enables a 1,7-hydride shift producing bicyclic α-cuprenene and β-chamigrene, while D99A and W297Y generate tricyclic epi-β-cedrene, demonstrating controlled progression from mono-to bi-to tricyclic products. (B) Close-up view of the SydA active site showing key residues D99, D222, W297, T185, and F187 coordinating with the Mg²⁺ cluster (green spheres) that control product specificity.

Rational control of aromatic residue clusters (ARCs) exemplifies computational-guided engineering for carbocation transport control, representing an orthogonal approach to traditional BP targeting.¹⁴² The bifunctional Pestalotiopsis fici nigtetraene synthase (PfNS) demonstrates how MD simulations can guide systematic redesign of critical transport networks that control skeletal complexity through regulated carbocation migration. Detailed simulations identified a distinctive ARC arrangement [F89-Y113-W317] in PfNS that differs significantly from ophiobolin F synthase (AcOS) [F66-F90-W189-W305], despite both enzymes sharing common carbocation intermediates. These computational insights facilitated the design of specific modifications to redirect cyclization pathways through controlled carbocation transport mechanisms.¹⁴²

The engineered PfNS W193L/T194W double mutant triggered functional transformation, completely abolishing production of the native 5/11 bicyclic nigtetraene in favor of novel 14-membered monocyclic and 5/15 bicyclic sesterterpenes.¹⁴² Progressive engineering culminated in the quadruple mutant PfNS F89A/Y113F/W193L/T194W, which achieved complete functional conversion to produce 5/8/5 tricyclic ophiobolin F at approximately 10% of AcOS WT yield, demonstrating successful skeletal reprogramming from bicyclic to tricyclic frameworks (Fig. 32). Computational analysis revealed the molecular basis: W194 forms edge-to-face interactions with F89, stabilizing C14 of GGPP at 5.2 ± 0.8 Å, while F89 creates critical C–H⋯π interactions with carbocation intermediates at 3.9 ± 0.3 Å. Theozyme calculations confirmed these interactions significantly reduce energy barriers (<0.5 kcal) for critical hydride shifts, explaining the mechanistic basis for skeleton transformation.^143,144 Volumetric analysis provided quantitative parameters for directing cyclization (active site volumes ranging from 166.0 ± 17.7 ų to 227.0 ± 17.7 ų), enabling rational design of different carbon skeletons.¹⁴²

Fig. 32 Engineering StTS by manipulating aromatic residue clusters (ARCs) for carbocation transport control. The W193L/T194W double mutation in PfNS abolishes native 5/11 bicyclic nigtetraene formation, instead producing fictetraenes A–C and 14-membered monocyclic cericerenes through Type B catalysis. The quadruple mutant F89A/Y113F/W193L/T194W achieves complete Type A conversion, generating 5/8/5 tricyclic ophiobolin F, demonstrating ARC-mediated progression from bicyclic to monocyclic/tricyclic scaffolds.

4.2.2. Precision engineering of critical decision points. While computational approaches provide comprehensive control over multi-branch reaction networks, experimental identification and selective manipulation of critical decision points offers a more streamlined strategy for achieving functional transformations.¹⁴⁵ This methodology centers on identifying specific molecular interactions that serve as decisive BPs in cyclization cascades, facilitating skeletal diversification through minimal, carefully positioned mutations. Understanding the fundamental mechanisms governing reaction complexity—such as intermediate stabilization patterns and polar group positioning—enables implementation of precise interventions that redirect entire cyclization pathways without requiring extensive computational modeling or systematic mutagenesis campaigns.¹⁴⁶

The value of focusing on critical intermediate stabilization is demonstrated in the bacterial DTS CotB2, where experimental identification of key residues controlling carbocation fate enabled scaffold transformations through selective mutagenesis.¹⁴⁶ CotB2 naturally converts GGPP to cyclooctat-9-en-7-ol, a complex 5/8/5 fused tricyclic structure with six chiral centers. Site-directed mutagenesis targeting residues involved in stabilizing reaction intermediates produced two variants with fundamentally altered cyclization trajectories: the CotB2 F107A mutant generated (R)-cembrene A, while the CotB2 W288G variant produced (1R,3E,7E,11S,12S)-3,7,18-dolabellatriene (Fig. 33). These scaffold transformations resulted from altered stabilization of critical carbocation intermediates at specific BPs in the cyclization cascade. The F107 residue appears to interact with and stabilize bicyclic intermediates in the native reaction pathway, and its replacement with alanine eliminated this stabilization, causing premature termination of the cyclization cascade immediately after initial ring formation. Similarly, W288 influences substrate folding and intermediate stabilization, with its substitution by glycine redirecting the reaction trajectory toward an alternative pathway. This experimental approach efficiently identified minimal mutations that fundamentally redirect complex cyclization cascades by focusing on residues likely to participate in specific intermediate stabilization rather than mapping the entire energetic landscape.¹⁴⁶

Fig. 33 Controlled redirection of diterpene cyclization through critical intermediates in CotB2. Single mutations in CotB2 dramatically simplify product complexity: F107A disrupts early cyclization yielding 14-membered monocyclic cembrene A, while W288G redirects the cascade to form 5/11 bicyclic dolabellatriene, both preventing formation of the native 5/8/5 tricyclic cyclooctat-9-en-7-ol.

An even more focused approach involves exploiting polar groups' role in determining skeletal complexity, as demonstrated in rice DTS OsKSL4 where strategic hydroxyl group placement critically influences reaction termination points.¹⁴⁷ A single T696I substitution transformed the enzyme's catalytic output from tricyclic syn-pimara-7,15-diene to complex rearranged tetracyclic aphidicol-15-ene with approximately 80% selectivity when presented with syn-CPP (Fig. 34). This mutation was positioned spatially near the C8-yl⁺ in the initial cyclization intermediate, where the hydroxyl group of T696 facilitates direct deprotonation of the pimar-15-en-8-yl⁺ intermediate in the WT, promptly terminating the reaction. Substituting the polar threonine with non-polar isoleucine removed this premature termination pathway, enabling the intermediate to undergo a complex sequence of transformations: a 1,2-hydride shift from C9 to C8, followed by secondary cyclization and Wagner–Meerwein ring rearrangement before eventual deprotonation. Notably, this substantial increase in reaction complexity occurred with minimal impact on catalytic efficiency (less than 2-fold decrease in yield), indicating the mutation primarily redirects reaction trajectory rather than disrupting overall enzyme function. The more hydrophobic environment created by the isoleucine substitution appears to better stabilize carbocation intermediates in proximity to the diphosphate co-product, enabling the enzyme to pursue more intricate reaction pathways and access complex product architectures that would otherwise be terminated by premature deprotonation.¹⁴⁷

Fig. 34 Modulation of diterpene cyclization complexity by single hydroxyl group positioning in OsKSL4. The T696I mutation eliminates hydroxyl-mediated premature deprotonation, allowing progression from tricyclic syn-pimaradiene through hydride shift and Wagner–Meerwein rearrangement to form the tetracyclic aphidicolane scaffold with a characteristic 5-membered ring.

5. Engineering beyond the active site

5.1. Engineering structural elements

Recognition that structural elements remote from the immediate active site can profoundly influence catalytic outcomes has led to innovative engineering strategies targeting dynamic protein regions. These approaches focus on modifying conserved motifs, flexible loops, and mobile structural elements that undergo conformational changes during catalysis, thereby altering substrate accessibility, carbocation stabilization patterns, and product selectivity.² Successful targets include the Hα-1 loop, a dynamic element which undergoes significant conformational changes to shield the active site from bulk solvent, secondary structural elements like the G-helix, conserved metal-binding motifs, and the dynamic C-terminal region.^148–150

Modifying the Hα-1 loop has proven effective for reversing hydroxylation patterns in TSs. This structurally conserved element undergoes significant conformational changes during catalysis and serves as a critical determinant of water accessibility to carbocation intermediates. Patchoulol synthase (PTS) and germacradien-11-ol synthase (Gd11olS) served as model systems, where transplantation of specific loop sequences from non-hydroxylating SdS yielded product shifts. The conversion of ²³⁸VEDE²⁴¹ to ²³³RRGS²³⁶ in Gd11olS and ⁴⁵⁸KKRE⁴⁶¹ to RRGS in PTS yielded substantial changes. Native PTS primarily synthesizes tricyclic patchoulol (60%), but the engineered PTS Hα-1 variant completely abolished hydroxylation, yielding instead bicyclic α-bulnesene (46%) and monocyclic germacrene A (40%) (Fig. 35). Similarly, Gd11olS Hα-1 sharply decreased germacradien-11-ol production from >90% to merely 8.3%, redirecting catalysis toward bicyclic isolepidozene (81%).¹⁴⁹

Fig. 35 Abolishing hydroxylation in STSs by engineering the Hα-1 loop. Transplanting the Hα-1 loop sequence eliminates water capture, converting PTS from a tricyclic alcohol (patchoulol) producer to generating bicyclic α-bulnesene and monocyclic germacrene A, while transforming Gd11olS from producing monocyclic germacradien-11-ol to bicyclic isolepidozene formation.

Complementing loop engineering, modifications of secondary structural elements have demonstrated equal promise for altering cyclization outcomes.¹⁵⁰ The G-helix kink in Gd11olS exemplifies how subtle changes to structural features can simultaneously influence carbocation stabilization and substrate positioning. The “kink” refers to a functionally critical bend in the G-α-helix that is essential for active site closure and substrate ionization. MD simulations identified the RQH site (R228, Q313, H320) and the G-helix kink region (residues 187–190) as critical determinants. Specific substitutions produced diverse product profiles: W312A substantially altered the product distribution, yielding large amounts of the acyclic nerolidol alongside the hydroxylated germacradien-11-ol and trace quantities of other cyclic hydrocarbons, including isolepidozene, while W312F redirected catalysis toward monocyclic germacrene A (71.5%). Additionally, the G188A mutation produced 88.4% bicyclic isolepidozene with minimal hydroxylation, demonstrating complete skeletal rearrangement (Fig. 36).¹⁵⁰

Fig. 36 Product diversification in GdolS and Gd11olS through site-directed mutagenesis. Gd11olS WT produces germacradien-11-ol, while W312A yields a mixture of acyclic nerolidol, germacradien-11-ol, and trace amounts of isolepidozene, and G188A accumulates bicyclic isolepidozene. GdolS WT produces germacradien-4-ol, while N218Q generates a mixture of germacradien-4-ol and germacrene A. These mutations demonstrate control over cyclization extent (acyclic to bicyclic) and hydroxylation, redirecting catalysis from hydroxylated sesquiterpenes to diverse terpenoid scaffolds.

Engineering conserved metal-binding motifs has revealed another effective approach for product diversification.¹⁴⁸ The NSE/DTE motif and adjacent residues in GdolS from Streptomyces citricolor exemplify this strategy. High-resolution structural analysis (PDB ID: 5I1U, 1.50 Å) revealed how subtle modifications to the metal coordination environment can alter product specificity. The N218Q mutation in the NSE motif converted FPP into a nearly equal mixture of hydroxylated monocyclic germacradien-4-ol (47.6%) and non-hydroxylated monocyclic germacrene A (50.7%), compared to exclusively germacradien-4-ol in the WT (Fig. 36). This substantial shift results from altered metal coordination geometry, enabling the diphosphate group to act as a base for premature carbocation quenching. Additional mutations targeting the active site periphery, including Y303F and E307M, produced smaller but significant product shifts (>2% and >4% germacrene A, respectively).¹⁴⁸

The dynamic C-terminal region represents perhaps the most complex target for engineering novel terpene scaffolds, as demonstrated in the bacterial DTS CotB2.³³ This mobile element, featuring the conserved WXXXXXRY motif, orchestrates active site closure, substrate positioning, and carbocation stabilization. Crystal structures (PDB IDs: 4OMG and 6GGI) captured both open and closed conformational states, revealing how C-terminal folding creates extensive hydrogen-bond networks connecting key catalytic motifs. Site-directed mutagenesis of aromatic residues in this region led to scaffold rearrangements: W288G transformed the natural cyclooctat-9-en-7-ol (5/8/5 fused ring system) into 3,7,18-dolabellatriene, while W186H similarly generated dolabellatriene by altering reaction trajectories (Fig. 37). Notably, F107A prevented crucial C2–C6 bond formation, yielding cembrene A through premature cyclization termination. QM/MM simulations quantified interaction energies between active site residues and carbocation intermediates (−10 to −20 kcal mol⁻¹), demonstrating how precisely positioned amino acids guide complex reaction cascades.³³

Fig. 37 Engineering novel diterpene scaffolds by targeting C-terminal dynamics in the CotB2. Aromatic residue mutations in the C-terminal region redirect the native 5/8/5 tricyclic cyclooctat-9-en-7-ol formation. CotB2 F107A disrupts C2–C6 bond formation yielding 14-membered monocyclic cembrene A, while W288G redirects the cascade to 5/11 bicyclic dolabellatriene. Inset shows the crystal structure of F107 and W288 positions critical for carbocation stabilization.

5.2. Second-shell engineering

While previous strategies have focused on direct manipulation of active site residues or structural elements that undergo dynamic conformational changes during catalysis, second-shell engineering targets a fundamentally different class of positions—residues that do not directly participate in substrate binding or catalysis but instead influence active site architecture through indirect interactions. These second-shell residues often participate in hydrogen-bonding networks, salt bridges, or structural elements that position catalytic residues for optimal function.¹⁵¹ Modifications to these positions can propagate structural changes to the active site, subtly altering reaction pathways without disrupting the core catalytic machinery. This approach offers unique advantages by enabling functional transformations while preserving the fundamental structural integrity of the enzyme's catalytic framework.^145,151,152

The potential of second-shell engineering is demonstrated in cattleyene synthase (CyS) from Streptomyces cattleya, where high-resolution structural analysis revealed how subtle modifications to non-substrate-contacting residues can reshape terpene cyclization landscapes.¹⁵¹ Three critical structures—apo-CyS (PDB ID: 7Y50, 2.00 Å), CyS-GGPP-Mg²⁺ complex (PDB ID: 7Y88, 1.87 Å), and the CyS C59A variant (PDB ID: 7Y87, 2.30 Å)—provided unprecedented insight into substrate binding and catalysis, with the CyS-GGPP-Mg²⁺ structure representing the first TS captured with its native GGPP substrate. The structure revealed an active site architecture where GGPP is surrounded by five aromatic residues (F62, W81, F86, W160, W318), four aliphatic residues (A190, A191, A229, L311), and three polar residues (C59, C82, N315). Systematic site-directed mutagenesis confirmed that the C59A substitution was particularly consequential, identifying C59 as a key second-shell residue that influences a first-shell residue without directly contacting the substrate (Fig. 38). This variant catalyzed the conversion of GGPP into six diterpenes with distinct skeletons: the original tetracyclic cattleyene, an alternative tetracyclic compound, two different tricyclic structures, another tetracyclic framework (allokutznerene), and a novel tetracyclic system with an unprecedented skeleton. The C59A variant produced approximately 6-fold higher total yields compared to F86A and W160A variants that generated similar diversity but with lower efficiency. Structural comparison between WT and C59A revealed the mechanistic basis for this functional shift: the thiol group of C59 forms a close interaction (3.4 Å) with the phenyl ring of F86 in CyS WT, orienting this residue toward the active site. The C59A mutation disrupts this interaction, prompting F86 to rotate 24° clockwise and shift slightly away from the substrate binding pocket. This structural rearrangement expands the active site cavity and alters the cation-π interactions that normally channel the reaction toward a single product, thereby reducing selectivity. The formation of the novel tetracyclic compound involves a particularly complex pathway mapped using 20 isotopomers of (¹³C)GGPP, revealing branching from a common intermediate through multiple steps involving ring closures, 1,2-hydride shifts, and skeletal rearrangements—a complex transformation enabled by a single second-shell substitution.¹⁵¹

Fig. 38 Second-shell C59A mutation in CyS unlocking skeletal diversity. The C59A substitution disrupts second-shell interactions with F86, expanding the active site cavity and enabling divergent reaction pathways from a common intermediate. Path a (black) leads to a native tetracyclic cattleyene, while paths b (red) and c (blue) generate alternative tetracyclic products including novel rearranged frameworks. Similar pathway diversification occurs with F86A and W160A mutations. Inset shows the active site architecture with key residues controlling pathway selection.

The versatility of second-shell engineering is further exemplified by Fusarium oxysporum fusoxypene synthase (FoFS), where saturation mutagenesis at a single key non-active site position generated extensive product diversity from GFPP.¹⁵² Systematic variation at position L89 in FoFS created variants with diverse product profiles from the GFPP.¹⁵² The L89Q variant proved particularly versatile, producing not only the native pentacyclic fusoxypene A, but also three novel sesterterpenes: fusoxypene D with a 5/6/8/5 fused tetracyclic skeleton, fusoxypene E featuring a 5/6/7/3/5 fused pentacyclic arrangement, and fusoxypene F with a 5/12/5 tricyclic framework, as well as pentacyclic (−)-astellatene. Other variants showed more focused product shifts: the L89C/A/G/T/N substitutions exclusively produced bicyclic sesterterpenes including bipolapene A, bipolapene B, and preterpestacin I, completely redirecting cyclization away from the native pentacyclic pathway, while L89D/M variants retained the original catalytic specificity (Fig. 39). Computational analyses using DFT calculations and MD simulations revealed that L89 mutations reconfigure hydrogen-bonding networks with second-shell residues N70 and Y74, which indirectly reposition the first-shell aromatic residue W69 within the catalytic pocket, modifying its interactions with carbocation intermediates through cation-π, dipole-π, or C–H⋯π mechanisms. These subtle changes redirect cyclization at critical BPs, yielding diverse ring systems without directly modifying the catalytic residues themselves.¹⁵²

Fig. 39 Sesterterpene diversity from second-shell L89 saturation in FoFS. Single amino acid substitutions at L89 dramatically redirect GFPP cyclization from native pentacyclic fusoxypene A to diverse scaffolds. L89Q generates tetracyclic fusoxypene D (5/6/8/5), pentacyclic fusoxypene E (5/6/7/3/5), tricyclic fusoxypene F (5/12/5), and pentacyclic (−)-astellatene, while L89C/A/G/T/N variants exclusively produce bicyclic bipolapenes, demonstrating second-shell control over skeletal complexity.

Building upon the single-position modifications demonstrated in CyS and FoFS, systematic contact mapping reveals even greater engineering potential when applied to multiple second-shell positions, as illustrated by comparing TEAS and Hyoscyamus premnaspirodiene synthase (HPS), which share 72% sequence identity yet produce fundamentally different products: TEAS generates predominantly eudesmane-type bicyclic compounds while HPS forms vetispirane-type spirocyclic structures from identical FPP.¹⁴⁵ Systematic tier-based mapping revealed that first-shell residues directly contacting the substrate were identical between the two enzymes, while critical determinants of product specificity resided in second-shell residues that shape the active site environment without directly contacting reaction intermediates (Fig. 40). Mutations informed by this mapping yielded catalytic redirections. The T402S/V516I double mutation in TEAS created a variant producing 4-epi-eremophilene (70% yield)—a previously undocumented bicyclic eremophilene-type compound resulting from altered regiospecificity of deprotonation at C6 rather than C8 of the eremophilenyl⁺ intermediate. Further refinement through a nine-residue variant (T274A/V291A/V372I/T402S/Y406L/S436N/I438T/I439L/V516I) generated 75% premnaspirodiene, effectively converting TEAS to mimic HPS function while maintaining comparable catalytic parameters (K_m = 6.24 µM, k_cat = 0.130). The complementary approach proved equally effective, with eight reciprocal mutations transforming HPS to produce 90% 5-epi-aristolochene, demonstrating bidirectional applicability. Mechanistic analysis revealed that these second-shell residues collectively influence catalytic trajectories by shaping active site geometry and dynamics rather than directly participating in substrate binding, with distal residues like V372I (located 12.5 Å from the active site center) influencing critical alkyl migration steps while T402S/V516I mutations determine regiospecificity of deprotonation.¹⁴⁵

Fig. 40 Engineering STSs TEAS and HPS using contact mapping and combinatorial mapping of catalytic landscapes. Systematic mutagenesis converts TEAS from producing bicyclic 5-epi-aristolochene to either the intermediate 4-epi-eremophilene or the spirocyclic premnaspirodiene (HPS product), demonstrating complete functional interconversion between eudesmane-type and vetispirane-type scaffolds through manipulation of second-shell residue networks.

This systematic engineering approach demonstrates how subtle modifications to residues surrounding the active site can redirect complex terpene cyclization cascades, providing another framework for rational engineering of terpene structural diversity through comprehensive mapping of extended interaction networks.

5.3. Combinatorial mapping of catalytic landscapes approach

The Combinatorial Mapping of Catalytic Landscapes approach represents a systematic advancement in TS engineering by comprehensively exploring functional sequence space through complete mutational libraries. Unlike sequential mutagenesis strategies, this method examines all possible mutational combinations among selected positions (2ⁿ variants for n positions), capturing both individual and combinatorial residue effects while revealing complex epistatic interactions that facilitate skeletal reprogramming across evolutionary landscapes.¹⁵³

Application of this approach to TEAS and HPS demonstrated its transformative potential for achieving skeletal diversification.¹⁵³ Despite sharing high sequence identity (as noted above), these enzymes produce fundamentally different carbon frameworks—TEAS generates bicyclic 5-epi-aristolochene, while HPS produces spirocyclic premnaspirodiene. A comprehensive combinatorial library encompassing all 512 possible combinations (2⁹) of nine phylogenetically guided differentiating residues revealed multiple discrete functional transitions between these distinct skeletal types. The study identified 418 active variants from 432 unique combinations, with 51.2% showing promiscuous activities, 36.8% maintaining TEAS-like specificity, 9.6% producing premnaspirodiene-like products, and only 2.4% generating 4-epi-eremophilene as the major product. The library revealed that product specificity changes varied greatly with mutational context, with average interneighbor distances ranging from 21 to 49 chemical distance units across different sequence backgrounds. Here, chemical distance represents a quantitative measure of product spectrum differences between enzyme variants, calculated as the Euclidean distance between their product distributions in four-dimensional space (representing percentages of 5-epi-aristolochene, 4-epi-eremophilene, premnaspirodiene, and minor products). Notably, an eight-mutation variant M8 (TEAS T274A/V291A/V372I/T402S/Y406L/S436N/I439L/V516I) achieved 81% premnaspirodiene production, surpassing the complete nine-mutation M9 variant (TEAS T274A/V291A/V372I/T402S/Y406L/S436N/I438T/I439L/V516I) (72% premnaspirodiene) (Fig. 40), demonstrating the non-linear nature of evolutionary trajectories toward new skeletal specificities. Furthermore, the strict dependence of 4-epi-eremophilene synthase activity on the T402S active-site mutation highlighted the critical role of active-site residues in directing skeletal reprogramming, despite the importance of distal mutations for overall catalytic optimization. DFT calculations provided mechanistic insights into these skeletal transformations, revealing higher energy barriers for methyl migration (8.4 kcal mol⁻¹) compared to methylene migration (4.8 kcal mol⁻¹), suggesting that thermodynamic control influences native skeletal specificities. Quantitative assessment across the entire mutational landscape revealed that approximately 7% of sequence backgrounds functioned as “hot spots” where single mutations produced functional shifts (average interneighbor distances >50 on the chemical distance scale, where larger values indicate greater divergence in product profiles), highlighting how specific mutational contexts can facilitate rapid skeletal diversification. This systematic exploration demonstrates how mutations distributed throughout the protein architecture—from direct active site positions to distal supporting residues—collectively redirect complex carbocation cyclization through subtle alterations in reaction geometry and intermediate stabilization.¹⁵³

6. Semi-rational and random approaches

6.1. Alanine scanning

Alanine scanning is a foundational approach for identifying critical functional residues in biosynthetic enzymes.¹⁵⁴ By systematically replacing active site amino acids with alanine—which eliminates side chain functionality beyond the β-carbon—this method reveals residues whose specific chemical properties are essential for catalysis or product selectivity.¹⁵⁵ The non-bulky, chemically inert nature of alanine makes it ideal for neutralizing side chain contributions while minimizing structural disruption.¹⁵⁶ When applied to TSs, alanine scanning often identifies residues that shape the active site pocket, influence carbocation stabilization, or participate in hydrogen-bonding networks critical for cyclization specificity.^157,158 The resulting functional changes provide insights into which positions most significantly influence product outcomes, creating a roadmap for more focused engineering efforts.

The transformation of mint (M. spicata) LS into diverse monoterpene-producing enzymes exemplifies how alanine scanning can guide systematic active site remodeling (Fig. 41).¹⁵⁷ Initial screening identified three critical positions—N345, L423, and S454—where alanine substitution increased formation of bicyclic pinene products.¹⁵⁷ Following established principles that postulate additive effects of plasticity residues,¹⁵⁷ the triple mutant N345A/L423A/S454A converted a highly specific LS (96.69 ± 0.29% monocyclic limonene) into an effective pinene synthase producing primarily bicyclic products (21.68 ± 2.64% α-pinene and 47.07 ± 3.87% β-pinene, total 68.75%) with minimal residual limonene (9.19 ± 1.57%).

Fig. 41 Alanine scanning-guided conversion of monoterpene synthases to alternative product specificities. The α-terpinyl⁺ intermediate serves as a central BP for diverse monoterpene products. LS WT produces monocyclic limonene. The N345I mutant generates monocyclic α/β-phellandrenes, whereas the N345A/L423A/S454A triple mutant yields bicyclic α/β-pinenes. Mutants N345S and N345D retain limonene production, while N345L shifts toward alternative scaffolds. In α-pinene synthase, substitutions at F482 (A/I/L/V/T) shift product specificity from bicyclic α-pinene to bicyclic sabinene.

Focused exploration of the N345 position revealed how side chain polarity controls cyclization pathway selection. The N345I mutation created a phellandrene synthase generating predominantly monocyclic products (29.47 ± 1.18% α-phellandrene and 39.40 ± 1.61% β-phellandrene, total 68.87%). Systematic substitution at this position demonstrated a clear structure–activity relationship: polar residues (N345S: 69.84 ± 0.73% limonene; N345D: 89.97 ± 0.47% limonene) maintained native limonene-producing activity, while non-polar substitutions (N345L: 25.04 ± 1.76% limonene; N345I: 18.48 ± 0.81% limonene) shifted product profiles toward alternative skeletal frameworks.¹⁵⁷

Mechanistic investigation revealed these mutations influence the fate of the α-terpinyl⁺ intermediate formed after initial cyclization of GPP. Structural analysis identified a “polar pocket” comprising multiple residues (C321, W324, N345, T349, S454, M458, H579, and Y573) that collectively stabilize the terpinyl⁺ and prevent carbocation migration toward the diphosphate co-product. Computational modeling confirmed the electrostatic steering effect of the diphosphate, with measurements from crystal structures showing the α-phosphorus positioned 2.72 Å, 4.70 Å, and 6.43 Å from C3, C5, and C7, respectively, supporting directional migration during bicyclic pinene formation.¹⁵⁷

Complementary studies on Paeonia lactiflora α-pinene synthase demonstrated how alanine scanning can identify aromatic residues critical for carbocation stabilization.¹⁵⁸ Homology modeling and computational docking identified F482 as a key position within 5 Å of the reaction center, stabilizes the pinyl⁺ through steric constraints and potential cation-π interactions. Single-point alanine substitution at this position, along with other aliphatic replacements (F482I, F482L, F482V, and F482T), transformed the native enzyme from producing 97.86% α-pinene (bicyclic 2,7-ring closure product) into an efficient sabinene synthase generating >90% sabinene—a bicyclic product formed via 6,7-hydride shift followed by 2,6-ring closure (Fig. 41). The F482V and F482L variants proved most effective, yielding 94.51% and 94.26% sabinene while retaining 29.50% and 49.98% of WT catalytic efficiency, respectively. MD simulations revealed that these mutations altered positioning of reaction intermediates within the active site, preventing stabilization of the pinyl⁺ and favoring the alternative sabinene-forming pathway.¹⁵⁸

6.2. Saturation mutagenesis and error-prone PCR

Saturation mutagenesis systematically explores the functional impact of all possible amino acid substitutions at specific positions in TSs.^159,160 This approach builds upon initial identification of catalytically important residues through alanine scanning, sequence comparisons, or random mutagenesis, providing comprehensive insights into how specific positions control reaction outcomes.^161,162 Testing all 20 possible amino acids (or judiciously strategically selected subsets) at critical positions enables optimization of desired properties, discovery of novel activities, and establishment of detailed structure–function relationships.¹⁶³ The technique proves especially effective for TSs, where subtle changes in active site composition can alter cyclization trajectories and product distributions.

Successful applications highlight saturation mutagenesis's versatility across diverse TSs. Saturation mutagenesis of cotton (+)-DCS generated the N403P/L405H double mutant producing 93% germacrene D-4-ol,⁸⁸ while systematic exploration of F205 in 2-methylenebornane synthase revealed substrate-dependent product profiles ranging from alcohols to novel C11 compounds.¹⁶⁴ Studies on FoFS demonstrated how non-catalytic substitutions at position L89 could generate distinct sesterterpene architectures, from tetracyclic structures (L89Q) to exclusively bicyclic products (L89C/A/G/T/N).¹⁵² In TXS, the Y688L mutation enhanced selectivity for specific isomers while Q609G redirected cyclization toward novel skeletons.⁵⁹ Smart library design strategies employing degenerate codons significantly improved screening efficiency, leading to the identification of variants such as V610F and V584M with completely altered product profiles.⁵⁶

Additionally, error-prone PCR complements rational approaches by introducing random mutations throughout the gene sequence, enabling exploration of unexpected beneficial mutations that influence protein dynamics through allosteric effects. This approach proved effective in engineering cotton (+)-δ-cadinene synthase, where random mutagenesis identified the L405S mutation that redirected cyclization from bicyclic (+)-δ-cadinene toward monocyclic germacrene D-4-ol, with subsequent systematic optimization yielding the highly efficient N403P/L405H double mutant.⁸⁸

These directed evolution approaches provide comprehensive functional information about critical residues, establishing their roles in determining product outcomes and creating opportunities for accessing novel terpene skeletons through systematic engineering.

7. Protein segment engineering

7.1. Protein segment swapping for functional conversion

Protein segment engineering leverages the modular architecture of TSs to achieve substantial functional transformations that often exceed the capabilities of single-residue mutations.¹⁶⁵ This approach recognizes that enzyme specificity frequently emerges from co-evolved amino acid networks rather than isolated catalytic positions, making domain swapping particularly effective for transferring complex functional properties between related TSs. By exchanging discrete structural modules, this strategy captures the integrated contributions of multiple residues acting in concert, enabling efficient traversal of large functional distances between different skeletal frameworks and substrate preferences.¹⁶⁶

The versatility of this approach extends across diverse TS classes. In fungal systems, systematic segment exchange studies between related bifunctional TSs have identified a critical cyclization control region (residues 60–69).⁹⁷ Chimeric constructs between fungal HAD-TCs have demonstrated that skeleton-determining functions can be isolated to the TC domain.¹²⁴ Recent structural studies of the bifunctional fusicoccadiene synthase (PaFS) revealed that cyclase domains preferentially interact with prenyltransferase domains from adjacent chains rather than their covalently linked partners, demonstrating how non-covalent interactions can mediate substrate channeling between domains.¹⁶⁶

The functional conversion between Cinnamomum camphora monoterpene and STSs demonstrates the potential of systematic domain swapping.¹⁶⁷ Despite sharing 97% DNA sequence identity, CiCaMS and CiCaSSy exhibit fundamentally different catalytic activities—CiCaMS produces linear monoterpenes (myrcene and linalool) from GPP, while CiCaSSy synthesizes complex polycyclic sesquiterpenes (tricyclic α-santalene, bicyclic β-santalene, and bicyclic trans-α-bergamotene) from FPP (Fig. 42A). Through systematic domain-swapping experiments, specific protein regions were exchanged between the two enzymes. Region R4 (a protein segment spanning positions 267–308 in the C-terminal domain, containing 5 amino acid differences) and Region R5 (a 12-amino acid stretch from positions 401–419 near the substrate-binding pocket, also with 5 substitutions) were individually transplanted from CiCaSSy into the CiCaMS backbone. These chimeric enzymes—MS_R4 (carrying all five R4 substitutions as a block) and MS_R5 (carrying all five R5 substitutions together)—gained STS activity, producing trans-α-bergamotene at 20.5 ± 0.7 and 28.1 ± 1.4 mg L⁻¹, respectively. The MS_R4R5 double hybrid, combining both regions, exhibited enhanced functional characteristics by producing all three sesquiterpenes (α-santalene, β-santalene, and trans-α-bergamotene) with a combined titer of 35.6 ± 15.2 mg L⁻¹, which approaches the production level of the WT CiCaSSy (43.8 ± 6.7 mg L⁻¹). This wholesale region exchange proved far more effective than any individual amino acid substitution (best single mutant MS_F294M: only 8.7 ± 0.4 mg L⁻¹).

Fig. 42 Conversion of CiCaMS MTS to STS function via region exchange. (A) CiCaMS WT converts GPP into linear products myrcene and linalool. (B) CiCaSSy WT converts FPP into three bicyclic sesquiterpenes. Through systematic region exchange, CiCaMS S267N/F294M/V401A/L403F/L404V/N442D gains STS activity, transforming from producing linear monoterpenes to generating bicyclic α-santalene, β-santalene, and trans-α-bergamotene, effectively mimicking CiCaSSy function.

Domain swapping revealed the modular nature of enzyme function, demonstrating that regions governing substrate recognition (GPP vs. FPP) could be functionally separated from those controlling cyclization trajectories.¹⁶⁷ Integration with selected point mutations ultimately achieved complete skeletal reprogramming. The MS_6S variant, incorporating six critical substitutions (S267N/F294M/V401A/L403F/L404V/N442D), reproduced CiCaSSy's characteristic product profile with 60% α-santalene, 25% trans-α-bergamotene, and 15% β-santalene, effectively converting a linear monoterpene-producing enzyme into a polycyclic STS (Fig. 42B).¹⁶⁷

Complementary studies on Sitka spruce MTSs revealed how domain swapping can identify single residues functioning as master switches for cyclization pathway control. Systematic exchanges between (+)-3-carene synthase (PsTPS-3car) and (−)-sabinene synthase (PsTPS-sab) narrowed specificity determinants to position F596 within helix J. The F596L substitution in PsTPS-sab redirected product formation from 44.7% (−)-sabinene (monocyclic) to 28.6% (+)-3-carene (bicyclic), while the reciprocal L596F mutation in PsTPS-3car2 shifted distribution from 67.5% (+)-3-carene to 37.4% (−)-sabinene. Triple mutations (A595G/F596L/L599F in PsTPS-sab) achieved 42.3% (+)-3-carene production, demonstrating amplified pathway redirection through coordinated substitutions (Fig. 43). Homology modeling revealed that F596 resides approximately 4 Å from the C4 carbon of the α-terpinyl⁺ intermediate, where phenylalanine likely provides steric constraints or cation-π interactions promoting sabinene formation while leucine facilitates the alternative 5,8-cyclization required for 3-carene production.

Fig. 43 Position 596 as a master switch controlling monoterpene product specificity. From the common α-terpinyl⁺, residue 596 directs four distinct cyclization fates: (1) F596G prevents cyclization, yielding linear myrcene; (2) PsTPS-sab WT or L596F (PsTPS-3car mutant) promotes monocyclic sabinene formation via cation-π stabilization; (3) PsTPS-3car WT or F596L substitutions enable bicyclic (+)-3-carene formation through 5,8-cyclization, with the triple mutant A595G/F596L/L599F showing enhanced selectivity; (4) F596E directs to monocyclic limonene through alternative deprotonation. A single position thus controls product complexity from linear to mono- and bicyclic scaffolds.

Kinetic analysis revealed that PsTPS-sab WT (producing primarily sabinene) exhibited significantly higher turnover rates (k_cat = 32.68 min⁻¹) than PsTPS-3car2 WT (producing primarily 3-carene, k_cat = 5.86 min⁻¹), demonstrating how the F596L substitution affects both product selectivity and catalytic efficiency.¹⁶⁸

8. Emerging and focused engineering strategies

Building upon the diverse strategies detailed in previous sections, emerging and particularly noteworthy engineering approaches offer new perspectives and tools for overcoming challenges in class I TS engineering, especially for achieving skeletal rearrangements.

8.1. Advanced applications of computational modeling and simulation

Computational tools form the core of powerful standalone strategies for tackling TS complexity. Specialized computational workflows, such as TerDockin, have been developed specifically for predicting binding modes of substrates and key carbocation intermediates within active sites. By combining chemical constraints and structural constraints, TerDockin predicts catalytically relevant ion-pair conformations consistent with experimental data.^169–171 TerDockin has been successfully applied to predict the binding modes of intermediates in bornyl diphosphate synthase, yielding results consistent with experimental observations and more complex QM/MM MD simulations.¹⁶⁹

QM/MM simulations combine quantum mechanics for the reacting center and molecular mechanics for the enzyme environment to simulate entire reaction pathways and calculate energy barriers for different steps.^104,172,173 These methods help understand how enzymes stabilize transition states and intermediates through electrostatic or steric effects, and can reveal mechanisms for controlling product specificity by raising energy barriers of undesired pathways.¹⁷⁴ QM/MM MD simulations have demonstrated how mutations in LS helices G1-G2 region (S454G, C457V, M458I) increased the distance between the α-terpinyl⁺ and potential bases, thereby preventing deprotonation to limonene and promoting reaction flow towards bicyclic products.¹⁷⁴

These approaches provide atomic-level details of transient species inaccessible experimentally, guide rational design, explain complex experimental outcomes, and predict binding modes even without experimental structures.¹⁷⁵ Their accuracy depends heavily on starting structure quality, force field parameters, and quantum mechanical methods used.

8.2. Directed evolution and high-throughput screening

Directed evolution mimics natural selection to rapidly optimize enzyme performance through creating large libraries of variants followed by high-throughput screening to identify those with desired properties. While error-prone PCR is commonly used to generate diversity, the screening step represents the true bottleneck for TS engineering.⁸⁸ Developing effective high-throughput screening for TSs presents particular challenges due to product volatility, limited or absent UV absorption, structural diversity, and the limitations of traditional GC-MS analysis. Several indirect approaches have emerged to address these challenges.

Substrate consumption-based screening has been successfully implemented using engineered strains with competitive carotenoid synthesis pathways, where TS activity is inferred from colony color intensity.¹⁷⁶ Byproduct detection relies on colorimetric methods to measure diphosphate release, reflecting overall activity but not distinguishing product profiles.^177,178 Biosensor-based approaches employ genetically encoded sensors that convert product concentration into detectable signals, though developing skeleton-specific sensors remains challenging.¹⁷⁹

Automated GC-MS workflows with liquid handling and multi-well plate autosamplers increase throughput for medium-scale library screening.¹⁸⁰ While directed evolution can discover unexpected beneficial mutations without requiring detailed structural knowledge, the screening bottleneck remains the major obstacle, particularly for identifying specific skeletal rearrangements.

8.3. Data-driven and machine learning approaches

With increasing TS sequence, structure, and functional data, machine learning approaches are emerging as promising tools for engineering.^181–184 These methods can learn patterns from sequence-function datasets to predict specificity of unknown sequences, such as STS preferences for specific carbocation precursors.¹⁸⁵

Machine learning and co-evolutionary analysis identify key residues or regions likely to influence product outcomes, guiding mutagenesis or library design. Co-evolution analysis combined with structural simulation has been used to identify hotspot regions in a 5-epi-aristolochene synthase, successfully improving enzyme activity through site-specific mutation.¹⁸⁶

These approaches can uncover complex patterns in large datasets, integrate diverse data types, and potentially overcome the poor sequence-function correlation often observed in TSs.^187,188 Co-evolutionary analysis reveals functionally important residue networks, including those distant from the active site.¹⁸⁹ While promising, these methods require substantial high-quality training data and careful management of potential biases.¹⁹⁰

Achieving significant skeletal remodeling in TSs typically requires integrating multiple strategies. An integrated, multidisciplinary approach combining computational modeling, ancestral reconstruction, high-throughput screening, and data-driven methods offers the effective path toward predictive engineering of TSs for novel product outcomes, including skeletal rearrangements.⁹⁹

9. Conclusions and perspectives

The diverse biological functions and applications of terpenoids—from pharmaceuticals to biofuels and fragrances—stem from their remarkable structural diversity. Engineering TSs to harness and expand this diversity represents a central goal in synthetic biology and natural product chemistry. This review has examined strategies for engineering class I TSs, including structure-guided rational design, evolutionary approaches leveraging natural variation, mechanism-focused engineering targeting carbocation intermediates, and methods exploring regions beyond the active site. These core approaches are complemented by semi-rational and random techniques like alanine scanning, saturation mutagenesis, and directed evolution, alongside computational modeling and high-throughput screening.

Engineering TSs for predictable skeletal diversification remains challenging due to their complex carbocation cascade mechanisms, inherent promiscuity, and the often weak correlation between sequence and function. Notably, while random mutagenesis approaches such as error-prone PCR and directed evolution have proven highly effective for optimizing enzyme properties like activity or stability, they have yielded limited success in generating fundamentally new terpene skeletons. This highlights the difficulty of navigating the vast and complex fitness landscapes required for such profound catalytic reprogramming. Consequently, progress is accelerating through the integration of structural biology, computational simulations, and diverse engineering techniques that allow for more precise interventions. Strategies combining rational design with evolutionary principles or computational predictions with experimental validation show particular promise for achieving significant skeletal changes.

Future advances will likely emerge from several directions: machine learning applications for predicting sequence–function relationships and designing novel variants; deeper mechanistic understanding through advanced spectroscopy and computational modeling; innovations in high-throughput screening using biosensors and microfluidic platforms; and integration of engineered enzymes into optimized microbial production systems. These developments will expand access to novel terpenoid structures with applications across medicine, agriculture, and materials science, demonstrating enzyme engineering's potential to reshape molecular diversity.

10. Conflicts of interest

There are no conflicts of interest to declare.

11. Data availability

This review article is based on previously published studies, and no new datasets were generated or analyzed. All cited sources are listed in the references section.

12. Acknowledgements

This work is supported in part by the National Natural Science Foundation of China Grant 82473823, the Natural Science Foundation of Jiangsu Province for Distinguished Young Scholars (BK20240097), the Jiangsu Provincial Major Science and Technology Special Project (BG2024046), and the Project Program of State Key Laboratory of Natural Medicines at China Pharmaceutical University (SKLNMZZ2024JS44).

13. References

1. X. Pan, J. D. Rudolf and L.-B. Dong, Nat. Prod. Rep., 2024, 41, 402–433.
2. D. W. Christianson, Chem. Rev., 2017, 117, 11570–11648.
3. T. Zeng, Z. Liu, J. Zhuang, Y. Jiang, W. He, H. Diao, N. Lv, Y. Jian, D. Liang, Y. Qiu, R. Zhang, F. Zhang, X. Tang and R. Wu, J. Chem. Inf. Model., 2020, 60, 2082–2090.
4. Z. Huang, K. A. Taizoumbe, C. Liang, B. Goldfuss, J. Xu and J. S. Dickschat, Angew. Chem., Int. Ed., 2023, 62, e202315659.
5. J. D. Rudolf, T. A. Alsup, B. Xu and Z. Li, Nat. Prod. Rep., 2021, 38, 905–980.
6. Y. Hoshino and L. Villanueva, FEMS Microbiol. Rev., 2023, 47, fuad008.
7. E. K. Rowinsky and R. C. Donehower, N. Engl. J. Med., 1995, 332, 1004–1014.
8. L. Min, J.-C. Han, W. Zhang, C.-C. Gu, Y.-P. Zou and C.-C. Li, Chem. Rev., 2023, 123, 4934–4971.
9. Y. Tu, Angew. Chem., Int. Ed., 2016, 55, 10210–10226.
10. S.-Q. Guo, Y.-X. Chen, Y.-L. Ju, C.-Y. Pan, J.-X. Shan, W.-W. Ye, N.-Q. Dong, Y. Kan, Y.-B. Yang, H.-Y. Zhao, H.-X. Yu, Z.-Q. Lu, J.-J. Lei, B. Liao, X.-R. Mu, Y.-J. Cao, L. Guo, J. Gao, J.-F. Zhou, K.-Y. Yang, H.-X. Lin and Y. Lin, Nature, 2025, 639, 162–171.
11. J. Zhang, Y. Zhang, J. Chen, M. Xu, X. Guan, C. Wu, S. Zhang, H. Qu, J. Chu, Y. Xu, M. Gu, Y. Liu and G. Xu, Nat. Commun., 2024, 15, 9233.
12. N. Luo, M. Turberg, M. Leutzsch, B. Mitschke, S. Brunen, V. N. Wakchaure, N. Nöthling, M. Schelwies, R. Pelzer and B. List, Nature, 2024, 632, 795–801.
13. R. Mewalal, D. K. Rai, D. Kainer, F. Chen, C. Külheim, G. F. Peter and G. A. Tuskan, Trends Biotechnol., 2017, 35, 227–240.
14. T. Zeng, H. Du and R. Wu, Chin. J. Nat. Med., 2025, 23, 1–18.
15. M. Weitman and D. T. Major, J. Am. Chem. Soc., 2010, 132, 6349–6360.
16. D. J. Miller and R. K. Allemann, Nat. Prod. Rep., 2012, 29, 60–71.
17. Q. Yang, J. Tian, S. Chen, Z. Yang, Z. Wang, H.-M. Xu and L.-B. Dong, Bioorg. Chem., 2024, 146, 107308.
18. D. Zhang, W. Du, X. Pan, X. Lin, F.-R. Li, Q. Wang, Q. Yang, H.-M. Xu and L.-B. Dong, Beilstein J. Org. Chem., 2024, 20, 815–822.
19. J. S. Dickschat, Angew. Chem., Int. Ed., 2019, 58, 15964–15976.
20. J. S. Dickschat, Nat. Prod. Rep., 2016, 33, 87–110.
21. K. Li and K. R. Gustafson, Nat. Prod. Rep., 2021, 38, 1251–1281.
22. Y. Li, J. Wang, L. Li, W. Song, M. Li, X. Hua, Y. Wang, J. Yuan and Z. Xue, Nat. Prod. Rep., 2023, 40, 1303–1353.
23. K. U. Wendt, K. Poralla and G. E. Schulz, Science, 1997, 277, 1811–1815.
24. D. W. Christianson, Chem. Rev., 2006, 106, 3412–3442.
25. R. Thoma, T. Schulz-Gasch, B. D'Arcy, J. Benz, J. Aebi, H. Dehmlow, M. Hennig, M. Stihle and A. Ruf, Nature, 2004, 432, 118–122.
26. H. Diao, N. Chen, K. Wang, F. Zhang, Y.-H. Wang and R. Wu, ACS Catal., 2020, 10, 2157–2168.
27. Z. Wang, T. A. Alsup, X. Pan, L.-L. Li, J. Tian, Z. Yang, X. Lin, H.-M. Xu, J. D. Rudolf and L.-B. Dong, Chem. Sci., 2025, 16, 310–317.
28. C. M. Starks, K. Back, J. Chappell and J. P. Noel, Science, 1997, 277, 1815–1820.
29. C. A. Lesburg, G. Zhai, D. E. Cane and D. W. Christianson, Science, 1997, 277, 1820–1824.
30. J. L. Faylo, T. A. Ronnebaum and D. W. Christianson, Acc. Chem. Res., 2021, 54, 3780–3791.
31. C. M. Paschall, J. Hasserodt, T. Jones, R. A. Lerner, K. D. Janda and D. W. Christianson, Angew. Chem., Int. Ed., 1999, 38, 1743–1747.
32. J. Qi, J. Wu, S. Kang, J. Gao, K. Hirokazu, H. Liu and C. Liu, Chin. J. Nat. Med., 2024, 22, 676–698.
33. R. Driller, S. Janke, M. Fuchs, E. Warner, A. R. Mhashal, D. T. Major, M. Christmann, T. Brück and B. Loll, Nat. Commun., 2018, 9, 3971.
34. Z. Li, L. Zhang, K. Xu, Y. Jiang, J. Du, X. Zhang, L.-H. Meng, Q. Wu, L. Du, X. Li, Y. Hu, Z. Xie, X. Jiang, Y.-J. Tang, R. Wu, R.-T. Guo and S. Li, Nat. Commun., 2023, 14, 4001.
35. T. Lou, A. Li, H. Xu, J. Pan, B. Xing, R. Wu, J. S. Dickschat, D. Yang and M. Ma, J. Am. Chem. Soc., 2023, 145, 8474–8485.
36. Z. Quan, A. Hou, B. Goldfuss and J. S. Dickschat, Angew. Chem., Int. Ed., 2022, 61, e202117273.
37. E.-Q. Li, C. W. Lindsley, J. Chang and B. Yu, J. Med. Chem., 2024, 67, 13509–13511.
38. S. Bag, J. Liu, S. Patil, J. Bonowski, S. Koska, B. Schölermann, R. Zhang, L. Wang, A. Pahl, S. Sievers, L. Brieger, C. Strohmann, S. Ziegler, M. Grigalunas and H. Waldmann, Nat. Chem., 2024, 16, 945–958.
39. P. Kirkpatrick, Nat. Rev. Drug Discovery, 2003, 2, 948.
40. S. E. Reisman and T. J. Maimone, Acc. Chem. Res., 2021, 54, 1815–1816.
41. C. S. Harmange Magnani, D. Q. Thach, K. T. Haelsig and T. J. Maimone, Acc. Chem. Res., 2020, 53, 949–961.
42. I. Burkhardt, T. De Rond, P. Y.-T. Chen and B. S. Moore, Nat. Chem. Biol., 2022, 18, 664–669.
43. R. Li, W. K. W. Chou, J. A. Himmelberger, K. M. Litwin, G. G. Harris, D. E. Cane and D. W. Christianson, Biochemistry, 2014, 53, 1155–1168.
44. X. Lin and D. E. Cane, J. Am. Chem. Soc., 2009, 131, 6332–6333.
45. R. P. Pemberton, K. C. Ho and D. J. Tantillo, Chem. Sci., 2015, 6, 2347–2353.
46. S. A. Eaton and D. W. Christianson, Biochemistry, 2024, 63, 797–805.
47. P. N. Blank, G. H. Barrow, W. K. W. Chou, L. Duan, D. E. Cane and D. W. Christianson, Biochemistry, 2017, 56, 5798–5811.
48. J.-X. Li, X. Fang, Q. Zhao, J.-X. Ruan, C.-Q. Yang, L.-J. Wang, D. J. Miller, J. A. Faraldos, R. K. Allemann, X.-Y. Chen and P. Zhang, Biochem. J., 2013, 451, 417–426.
49. R. Otten, R. A. P. Pádua, H. A. Bunzel, V. Nguyen, W. Pitsawong, M. Patterson, S. Sui, S. L. Perry, A. E. Cohen, D. Hilvert and D. Kern, Science, 2020, 370, 1442–1446.
50. S. J. Benkovic and S. Hammes-Schiffer, Science, 2003, 301, 1196–1202.
51. J. Jumper, R. Evans, A. Pritzel, T. Green, M. Figurnov, O. Ronneberger, K. Tunyasuvunakool, R. Bates, A. Žídek, A. Potapenko, A. Bridgland, C. Meyer, S. A. A. Kohl, A. J. Ballard, A. Cowie, B. Romera-Paredes, S. Nikolov, R. Jain, J. Adler, T. Back, S. Petersen, D. Reiman, E. Clancy, M. Zielinski, M. Steinegger, M. Pacholska, T. Berghammer, S. Bodenstein, D. Silver, O. Vinyals, A. W. Senior, K. Kavukcuoglu, P. Kohli and D. Hassabis, Nature, 2021, 596, 583–589.
52. T. Wang, X. He, M. Li, Y. Li, R. Bi, Y. Wang, C. Cheng, X. Shen, J. Meng, H. Zhang, H. Liu, Z. Wang, S. Li, B. Shao and T.-Y. Liu, Nature, 2024, 635, 1019–1027.
53. H. Tao, L. Lauterbach, G. Bian, R. Chen, A. Hou, T. Mori, S. Cheng, B. Hu, L. Lu, X. Mu, M. Li, N. Adachi, M. Kawasaki, T. Moriya, T. Senda, X. Wang, Z. Deng, I. Abe, J. S. Dickschat and T. Liu, Nature, 2022, 606, 414–419.
54. T. Klucznik, L.-D. Syntrivanis, S. Baś, B. Mikulak-Klucznik, M. Moskal, S. Szymkuć, J. Mlynarski, L. Gadina, W. Beker, M. D. Burke, K. Tiefenbacher and B. A. Grzybowski, Nature, 2024, 625, 508–515.
55. P. Schrepfer, A. Buettner, C. Goerner, M. Hertel, J. Van Rijn, F. Wallrapp, W. Eisenreich, V. Sieber, R. Kourist and T. Brück, Proc. Natl. Acad. Sci. U.S.A., 2016, 113, E958–E967.
56. S. He, I. I. Abdallah, R. Van Merkerk and W. J. Quax, Planta, 2024, 259, 87.
57. Z. Xiao, Q. Yang, X. Lin, F.-R. Li, X. Zhang, H.-M. Xu, Z. Wang, J. Wang and L.-B. Dong, Org. Lett., 2024, 26, 1640–1644.
58. M. Köksal, Y. Jin, R. M. Coates, R. Croteau and D. W. Christianson, Nature, 2011, 469, 116–120.
59. S. Edgar, F.-S. Li, K. Qiao, J.-K. Weng and G. Stephanopoulos, ACS Synth. Biol., 2017, 6, 201–205.
60. E. N. Kissman, M. B. Sosa, D. C. Millar, E. J. Koleski, K. Thevasundaram and M. C. Y. Chang, Nature, 2024, 631, 37–48.
61. Q. Jin, D. C. Williams, M. Hezari, R. Croteau and R. M. Coates, J. Org. Chem., 2005, 70, 4667–4675.
62. D. T. Major, ACS Catal., 2017, 7, 5461–5465.
63. J. Li, B. Chen, Z. Fu, J. Mao, L. Liu, X. Chen, M. Zheng, C.-Y. Wang, C. Wang, Y.-W. Guo and B. Xu, Nat. Commun., 2024, 15, 5940.
64. Z. Wang, Q. Yang, J. He, H. Li, X. Pan, Z. Li, H. Xu, J. D. Rudolf, D. J. Tantillo and L.-B. Dong, Angew. Chem., Int. Ed., 2023, 62, e202312490.
65. F.-R. Li, Q. Yang, J. He, X. Sun, X. Pan, H.-M. Xu, J. D. Rudolf and L.-B. Dong, Chem.–Eur. J., 2025, 31, e202500012.
66. Z. Li, S. Jindani, V. Kojasoy, T. Ortega, E. M. Marshall, K. A. Abboud, S. Loesgen, D. J. Tantillo and J. D. Rudolf, Beilstein J. Org. Chem., 2024, 20, 1320–1326.
67. Z. Li, B. Xu, T. A. Alsup, X. Wei, W. Ning, D. G. Icenhour, M. A. Ehrenberger, I. Ghiviriga, B.-D. Giang and J. D. Rudolf, J. Am. Chem. Soc., 2023, 145, 22361–22365.
68. Z. Li and J. D. Rudolf, J. Ind. Microbiol. Biotechnol., 2023, 50, kuad027.
69. B. Chen, J. Mao, K. Xu, L. Liu, W. Lin, Y.-W. Guo, R. Wu, C. Wang and B. Xu, Sci. Adv., 2025, 11, eadv0805.
70. B. Xu, W. Ning, X. Wei and J. D. Rudolf, Org. Biomol. Chem., 2022, 20, 8833–8837.
71. C. Zhu, B. Xu, D. A. Adpressa, J. D. Rudolf and S. Loesgen, Angew. Chem., Int. Ed., 2021, 60, 14163–14170.
72. Z. Li, B. Xu, V. Kojasoy, T. Ortega, D. A. Adpressa, W. Ning, X. Wei, J. Liu, D. J. Tantillo, S. Loesgen and J. D. Rudolf, Chem, 2023, 9, 698–708.
73. P. D. Scesa, Z. Lin and E. W. Schmidt, Nat. Chem. Biol., 2022, 18, 659–663.
74. P. Baer, P. Rabe, K. Fischer, C. A. Citron, T. A. Klapschinski, M. Groll and J. S. Dickschat, Angew. Chem., Int. Ed., 2014, 53, 7652–7656.
75. J. N. Whitehead, N. G. H. Leferink, L. O. Johannissen, S. Hay and N. S. Scrutton, ACS Catal., 2023, 13, 12774–12802.
76. J. A. Aaron, X. Lin, D. E. Cane and D. W. Christianson, Biochemistry, 2010, 49, 1787–1797.
77. Y. J. Hong and D. J. Tantillo, Nat. Chem., 2014, 6, 104–111.
78. J. N. Whitehead, N. G. H. Leferink, G. Komati Reddy, C. W. Levy, S. Hay, E. Takano and N. S. Scrutton, ACS Catal., 2022, 12, 12123–12131.
79. J. Xu, J. Xu, Y. Ai, R. A. Farid, L. Tong and D. Yang, Arch. Biochem. Biophys., 2018, 638, 27–34.
80. M. J. Calvert, S. E. Taylor and R. K. Allemann, Chem. Commun., 2002, 2384–2385.
81. J. A. Aaron and D. W. Christianson, Pure Appl. Chem., 2010, 82, 1585–1597.
82. D. C. Hyatt, B. Youn, Y. Zhao, B. Santhamma, R. M. Coates, R. B. Croteau and C. Kang, Proc. Natl. Acad. Sci. U.S.A., 2007, 104, 5360–5365.
83. N. G. H. Leferink, A. M. Escorcia, B. R. Ouwersloot, L. O. Johanissen, S. Hay, M. W. Van Der Kamp and N. S. Scrutton, ChemBioChem, 2022, 23, e202100688.
84. Y. Yonetani, J. Chem. Phys., 2018, 149, 175102.
85. V. González Requena, P. L. Srivastava, D. J. Miller and R. K. Allemann, ChemBioChem, 2024, 25, e202400290.
86. J. A. Faraldos, D. J. Miller, V. González, Z. Yoosuf-Aly, O. Cascón, A. Li and R. K. Allemann, J. Am. Chem. Soc., 2012, 134, 5900–5908.
87. M. Loizzi, V. González, D. J. Miller and R. K. Allemann, ChemBioChem, 2018, 19, 100–105.
88. Y. Yoshikuni, V. J. J. Martin, T. E. Ferrin and J. D. Keasling, Chem. Biol., 2006, 13, 91–98.
89. S. Cesaro-Tadic, D. Lagos, A. Honegger, J. H. Rickard, L. J. Partridge, G. M. Blackburn and A. Plückthun, Nat. Biotechnol., 2003, 21, 679–685.
90. D. S. Wilson and A. D. Keefe, Curr. Protoc. Mol. Biol., 2001, 8.3.1–8.3.9 , Chapter 8, Unit 8.3.
91. N. G. H. Leferink, K. E. Ranaghan, J. Battye, L. O. Johannissen, S. Hay, M. W. Van Der Kamp, A. J. Mulholland and N. S. Scrutton, ChemBioChem, 2020, 21, 985–990.
92. P. L. Srivastava, S. T. Johns, A. Voice, K. Morley, A. M. Escorcia, D. J. Miller, R. K. Allemann and M. W. Van Der Kamp, ACS Catal., 2024, 14, 11034–11043.
93. L. A. Johnson and R. K. Allemann, Chem. Commun., 2025, 61, 2468–2483.
94. P. S. Karunanithi and P. Zerbe, Front. Plant Sci., 2019, 10, 1166.
95. B. Gu, B. Goldfuss and J. S. Dickschat, Angew. Chem., Int. Ed., 2023, 62, e202215688.
96. H. Xu and J. S. Dickschat, ACS Catal., 2023, 13, 12723–12729.
97. T. Rekha, D. Sharma, F. Lin, Y. K. Choong, C. Lim, C. Jobichen and C. Zhang, ACS Catal., 2023, 13, 4949–4959.
98. H. Kim, N. Srividya, I. Lange, E. W. Huchala, B. Ginovska, B. M. Lange and S. Raugei, ACS Catal., 2022, 12, 7453–7469.
99. N. G. H. Leferink and N. S. Scrutton, ChemBioChem, 2022, 23, e202100484.
100. D. Lei, Z. Qiu, J. Qiao and G.-R. Zhao, Biotechnol. Biofuels, 2021, 14, 147.
101. A. Hou, B. Goldfuss and J. S. Dickschat, Angew. Chem., Int. Ed., 2021, 60, 20781–20785.
102. M. Xu, H. Xu, Z. Lei, B. Xing, J. S. Dickschat, D. Yang and M. Ma, Angew. Chem., Int. Ed., 2024, 63, e202405140.
103. V. Karuppiah, K. E. Ranaghan, N. G. H. Leferink, L. O. Johannissen, M. Shanmugam, A. Ní Cheallaigh, N. J. Bennett, L. J. Kearsey, E. Takano, J. M. Gardiner, M. W. Van Der Kamp, S. Hay, A. J. Mulholland, D. Leys and N. S. Scrutton, ACS Catal., 2017, 7, 6268–6282.
104. Y.-H. Wang, H. Xu, J. Zou, X.-B. Chen, Y.-Q. Zhuang, W.-L. Liu, E. Celik, G.-D. Chen, D. Hu, H. Gao, R. Wu, P.-H. Sun and J. S. Dickschat, Nat. Catal., 2022, 5, 128–135.
105. K. Gao, K. Xu, P. Lin, J. Zhu, R. Wu and J. Zi, ACS Catal., 2024, 14, 14233–14241.
106. Q. Jia, R. Brown, T. G. Köllner, J. Fu, X. Chen, G. K.-S. Wong, J. Gershenzon, R. J. Peters and F. Chen, Proc. Natl. Acad. Sci. U.S.A., 2022, 119, e2100361119.
107. F. Chen, D. Tholl, J. Bohlmann and E. Pichersky, Plant J., 2011, 66, 212–229.
108. X. Chen, T. G. Köllner, Q. Jia, A. Norris, B. Santhanam, P. Rabe, J. S. Dickschat, G. Shaulsky, J. Gershenzon and F. Chen, Proc. Natl. Acad. Sci. U.S.A., 2016, 113, 12132–12137.
109. B. Jin, K. Xu, J. Guo, Y. Ma, J. Yang, N. Chen, T. Zeng, J. Wang, J. Liu, M. Tian, Q. Ma, H. Zhang, R. J. Peters, G. Cui, R. Wu and L. Huang, ACS Catal., 2024, 14, 2959–2970.
110. X. Fang, J.-X. Li, J.-Q. Huang, Y.-L. Xiao, P. Zhang and X.-Y. Chen, Biochem. J., 2017, 474, 2191–2202.
111. V. Gonzalez, S. Touchet, D. J. Grundy, J. A. Faraldos and R. K. Allemann, J. Am. Chem. Soc., 2014, 136, 14505–14512.
112. J.-Q. Huang, D.-M. Li, J.-X. Li, J.-L. Lin, X. Tian, L.-J. Wang, X.-Y. Chen and X. Fang, Org. Biomol. Chem., 2021, 19, 6650–6656.
113. S. C. Kampranis, D. Ioannidis, A. Purvis, W. Mahrez, E. Ninga, N. A. Katerelos, S. Anssour, J. M. Dunwell, J. Degenhardt, A. M. Makris, P. W. Goodenough and C. B. Johnson, Plant Cell, 2007, 19, 1994–2005.
114. H. Gee and R. Howlett, Nature, 2009, 457, 807.
115. M. Xu, P. R. Wilderman and R. J. Peters, Proc. Natl. Acad. Sci. U.S.A., 2007, 104, 7397–7401.
116. M. Jia and R. J. Peters, Front. Plant Sci., 2016, 7, 1765.
117. M. Jia, K. Zhou, S. Tufts, S. Schulte and R. J. Peters, ACS Chem. Biol., 2017, 12, 862–867.
118. D. P. Drew, T. B. Andersen, C. Sweetman, B. L. Møller, C. Ford and H. T. Simonsen, J. Exp. Bot., 2016, 67, 799–808.
119. T. G. Köllner, C. Schnee, J. Gershenzon and J. Degenhardt, Plant Cell, 2004, 16, 1115–1131.
120. T. G. Köllner, J. Degenhardt and J. Gershenzon, Plants, 2020, 9, 552.
121. M. Salmon, C. Laurendon, M. Vardakou, J. Cheema, M. Defernez, S. Green, J. A. Faraldos and P. E. O'Maille, Nat. Commun., 2015, 6, 6143.
122. G. Wei, F. Eberl, X. Chen, C. Zhang, S. B. Unsicker, T. G. Köllner, J. Gershenzon and F. Chen, Sci. Rep., 2020, 10, 14944.
123. J. Shang, D. Feng, H. Liu, L. Niu, R. Li, Y. Li, M. Chen, A. Li, Z. Liu, Y. He, X. Gao, H. Jian, C. Wang, K. Tang, M. Bao, J. Wang, S. Yang, H. Yan and G. Ning, Curr. Biol., 2024, 34, 3550–3563.
124. T. Chen, C. Chen, C. Lee, R. Huang, K. Chen, Y. Lu, S. Liang, M. Pham, Y. K. Rao, S. Wu, R. Chein and H. Lin, Angew. Chem., Int. Ed., 2023, 62, e202215566.
125. W. Du, Q. Yang, H. Xu and L.-B. Dong, Chin. J. Nat. Med., 2022, 20, 737–748.
126. M. Ringel, N. Dimos, S. Himpich, M. Haack, C. Huber, W. Eisenreich, G. Schenk, B. Loll and T. Brück, Microb. Cell Fact., 2022, 21, 64.
127. J. N. Whitehead, N. G. H. Leferink, S. Hay and N. S. Scrutton, ChemBioChem, 2025, 26, e202400672.
128. M. M. Ravn, R. J. Peters, R. M. Coates and R. Croteau, J. Am. Chem. Soc., 2002, 124, 6998–7006.
129. R. Schwartz, S. Zev and D. T. Major, Methods Enzymol., 2024, 699, 265–292.
130. X. Tang, F. Zhang, T. Zeng, W. Li, S. Yin and R. Wu, ACS Chem. Biol., 2020, 15, 2820–2832.
131. T. A. Pemberton and D. W. Christianson, J. Antibiot., 2016, 69, 486–493.
132. P. R. Wilderman and R. J. Peters, J. Am. Chem. Soc., 2007, 129, 15736–15737.
133. P. Zerbe, A. Chiang and J. Bohlmann, Phytochemistry, 2012, 74, 30–39.
134. S. Perez Rafael, A. Vallée-Bélisle, E. Fabregas, K. Plaxco, G. Palleschi and F. Ricci, Anal. Chem., 2012, 84, 1076–1082.
135. E. M. Paradise, J. Kirby, R. Chan and J. D. Keasling, Biotechnol. Bioeng., 2008, 100, 371–378.
136. D. J. Tantillo, Nat. Prod. Rep., 2011, 28, 1035–1053.
137. S. R. Hare and D. J. Tantillo, Beilstein J. Org. Chem., 2016, 12, 377–390.
138. D. Ray, S. Das and U. Raucci, J. Chem. Inf. Model., 2024, 64, 3953–3958.
139. K. Raz, S. Levi, P. K. Gupta and D. T. Major, Curr. Opin. Biotechnol., 2020, 65, 248–258.
140. L. Su, P. Liu, W. Liu, Q. Liu, J. Gao, Q. Zhao, K. Jia, X. Sheng, H. Ma, Q. Wang and Z. Dai, ACS Catal., 2024, 14, 17699–17715.
141. M. Wu, I. Torrence, Y. Liu, J. Wu, R. Ge, K. Ma, D. Liu, J. Ren, S. Fan, M. Ma, J. B. Siegel, D. J. Tantillo, W. Lin and A. Fan, J. Am. Chem. Soc., 2025, 147, 10413–10422.
142. W. Zhang, X. Wang, G. Zhu, B. Zhu, K. Peng, T. Hsiang, L. Zhang and X. Liu, Angew. Chem., Int. Ed., 2024, 63, e202406246.
143. H. Sato, K. Saito and M. Yamazaki, Front. Plant Sci., 2019, 10, 802.
144. X. Zhang, J. DeChancie, H. Gunaydin, A. B. Chowdry, F. R. Clemente, A. J. T. Smith, T. M. Handel and K. N. Houk, J. Org. Chem., 2008, 73, 889–899.
145. B. T. Greenhagen, P. E. O'Maille, J. P. Noel and J. Chappell, Proc. Natl. Acad. Sci. U.S.A., 2006, 103, 9826–9831.
146. A. Meguro, Y. Motoyoshi, K. Teramoto, S. Ueda, Y. Totsuka, Y. Ando, T. Tomita, S. Kim, T. Kimura, M. Igarashi, R. Sawa, T. Shinada, M. Nishiyama and T. Kuzuyama, Angew. Chem., Int. Ed., 2015, 54, 4353–4356.
147. D. Morrone, M. Xu, D. B. Fulton, M. K. Determan and R. J. Peters, J. Am. Chem. Soc., 2008, 130, 5400–5401.
148. D. J. Grundy, M. Chen, V. González, S. Leoni, D. J. Miller, D. W. Christianson and R. K. Allemann, Biochemistry, 2016, 55, 2112–2121.
149. P. L. Srivastava, S. T. Johns, R. Walters, D. J. Miller, M. W. Van Der Kamp and R. K. Allemann, ACS Catal., 2023, 13, 14199–14204.
150. P. L. Srivastava, A. M. Escorcia, F. Huynh, D. J. Miller, R. K. Allemann and M. W. Van Der Kamp, ACS Catal., 2021, 11, 1033–1041.
151. B. Xing, H. Xu, A. Li, T. Lou, M. Xu, K. Wang, Z. Xu, J. S. Dickschat, D. Yang and M. Ma, Angew. Chem., Int. Ed., 2022, 61, e202209785.
152. X. Wang, Y. Huang, W. Zhang, K. Lv, X. Li, Z. Wang, L. Zhang, T. Hsiang, L. Zhang, L. Ouyang and X. Liu, Synth. Syst. Biotechnol., 2024, 9, 380–387.
153. P. E. O'Maille, A. Malone, N. Dellas, B. Andes Hess, L. Smentek, I. Sheehan, B. T. Greenhagen, J. Chappell, G. Manning and J. P. Noel, Nat. Chem. Biol., 2008, 4, 617–623.
154. K. L. Morrison and G. A. Weiss, Curr. Opin. Chem. Biol., 2001, 5, 302–307.
155. C. W. Wood, A. A. Ibarra, G. J. Bartlett, A. J. Wilson, D. N. Woolfson and R. B. Sessions, Bioinformatics, 2020, 36, 2917–2919.
156. B. C. Cunningham and J. A. Wells, Science, 1989, 244, 1081–1085.
157. J. Xu, Y. Ai, J. Wang, J. Xu, Y. Zhang and D. Yang, Phytochemistry, 2017, 137, 34–41.
158. J. Xu, G. Peng, J. Xu, Y. Li, L. Tong and D. Yang, Phytochemistry, 2021, 181, 112573.
159. V. Sayous, P. Lubrano, Y. Li and C. G. Acevedo-Rocha, Biochim. Biophys. Acta, Proteins Proteomics, 2020, 1868, 140321.
160. X. F. Cadet, J. C. Gelly, A. Van Noord, F. Cadet and C. G. Acevedo-Rocha, Methods Mol. Biol., 2022, 2461, 225–275.
161. S. Kille, C. G. Acevedo-Rocha, L. P. Parra, Z.-G. Zhang, D. J. Opperman, M. T. Reetz and J. P. Acevedo, ACS Synth. Biol., 2013, 2, 83–92.
162. C. G. Acevedo-Rocha, M. T. Reetz and Y. Nov, Sci. Rep., 2015, 5, 10654.
163. L. Sellés Vidal, M. Isalan, J. T. Heap and R. Ledesma-Amaro, RSC Chem. Biol., 2023, 4, 271–291.
164. M. J. Kschowak, F. Maier, H. Wortmann and M. Buchhaupt, ACS Synth. Biol., 2020, 9, 981–986.
165. P. H. Celie, A. H. Parret and A. Perrakis, Curr. Opin. Struct. Biol., 2016, 38, 145–154.
166. E. S. Wenger, K. Schultz, R. Marmorstein and D. W. Christianson, Proc. Natl. Acad. Sci. U.S.A., 2024, 121, e2408064121.
167. A. Di Girolamo, J. Durairaj, A. van Houwelingen, F. Verstappen, D. Bosch, K. Cankar, H. Bouwmeester, D. de Ridder, A. D. J. van Dijk and J. Beekwilder, Arch. Biochem. Biophys., 2020, 695, 108647.
168. C. R. Roach, D. E. Hall, P. Zerbe and J. Bohlmann, J. Biol. Chem., 2014, 289, 23859–23869.
169. T. E. O'Brien, S. J. Bertolani, Y. Zhang, J. B. Siegel and D. J. Tantillo, ACS Catal., 2018, 8, 3322–3330.
170. I. S. Torrence, T. E. O'Brien, J. B. Siegel and D. J. Tantillo, Methods Enzymol., 2024, 699, 231–263.
171. M. Jia, Y. Zhang, J. B. Siegel, D. J. Tantillo and R. J. Peters, ACS Catal., 2019, 9, 8867–8871.
172. F. Zhang, T. Zeng and R. Wu, J. Chem. Inf. Model., 2023, 63, 5018–5034.
173. N. Chen, S. Wang, L. Smentek, B. A. Hess and R. Wu, Angew. Chem., Int. Ed., 2015, 54, 8693–8696.
174. N. G. H. Leferink, K. E. Ranaghan, V. Karuppiah, A. Currin, M. W. Van Der Kamp, A. J. Mulholland and N. S. Scrutton, ACS Catal., 2018, 8, 3780–3791.
175. Y. Luo, X. Ma, Y. Qiu, Y. Lu, S. Shen, Y. Li, H. Gao, K. Chen, J. Zhou, T. Hu, L. Tu, H. Zhao, D. Li, F. Leng, W. Gao, T. Jiang, C. Liu, L. Huang, R. Wu and Y. Tong, Angew. Chem., Int. Ed., 2023, 62, e202313429.
176. M. Furubayashi, M. Ikezumi, J. Kajiwara, M. Iwasaki, A. Fujii, L. Li, K. Saito and D. Umeno, PLoS ONE, 2014, 9, e93317.
177. R. Lauchli, K. S. Rabe, K. Z. Kalbarczyk, A. Tata, T. Heel, R. Z. Kitto and F. H. Arnold, Angew. Chem., Int. Ed., 2013, 52, 5571–5574.
178. M. Vardakou, M. Salmon, J. A. Faraldos and P. E. O'Maille, MethodsX, 2014, 1, 187–196.
179. S. K. Kim, S. H. Kim, B. Subhadra, S.-G. Woo, E. Rha, S.-W. Kim, H. Kim, D.-H. Lee and S.-G. Lee, ACS Synth. Biol., 2018, 7, 2379–2390.
180. N. G. H. Leferink, M. S. Dunstan, K. A. Hollywood, N. Swainston, A. Currin, A. J. Jervis, E. Takano and N. S. Scrutton, Sci. Rep., 2019, 9, 11936.
181. J. Yang, F.-Z. Li and F. H. Arnold, ACS Cent. Sci., 2024, 10, 226–241.
182. T. Yu, H. Cui, J. C. Li, Y. Luo, G. Jiang and H. Zhao, Science, 2023, 379, 1358–1363.
183. A. H.-W. Yeh, C. Norn, Y. Kipnis, D. Tischer, S. J. Pellock, D. Evans, P. Ma, G. R. Lee, J. Z. Zhang, I. Anishchenko, B. Coventry, L. Cao, J. Dauparas, S. Halabiya, M. DeWitt, L. Carter, K. N. Houk and D. Baker, Nature, 2023, 614, 774–780.
184. J. L. Watson, D. Juergens, N. R. Bennett, B. L. Trippe, J. Yim, H. E. Eisenach, W. Ahern, A. J. Borst, R. J. Ragotte, L. F. Milles, B. I. M. Wicky, N. Hanikel, S. J. Pellock, A. Courbet, W. Sheffler, J. Wang, P. Venkatesh, I. Sappington, S. V. Torres, A. Lauko, V. De Bortoli, E. Mathieu, S. Ovchinnikov, R. Barzilay, T. S. Jaakkola, F. DiMaio, M. Baek and D. Baker, Nature, 2023, 620, 1089–1100.
185. J. Durairaj, E. Melillo, H. J. Bouwmeester, J. Beekwilder, D. De Ridder and A. D. J. Van Dijk, PLoS Comput. Biol., 2021, 17, e1008197.
186. X. Yan, J. Zhou, J. Ge, W. Li, D. Liang, W. Singh, G. Black, S. Nie, J. Liu, M. Sun, J. Qiao and M. Huang, ACS Catal., 2022, 12, 4037–4045.
187. J. Zhou and M. Huang, Chem. Soc. Rev., 2024, 53, 8202–8239.
188. R. A. Silverstein, N. Kim, A.-S. Kroell, R. T. Walton, J. Delano, R. M. Butcher, M. Pacesa, B. K. Smith, K. A. Christie, L. L. Ha, R. J. Meis, A. B. Clark, A. D. Spinner, C. R. Lazzarotto, Y. Li, A. Matsubara, E. O. Urbina, G. A. Dahl, B. E. Correia, D. S. Marks, S. Q. Tsai, L. Pinello, S. S. De Ravin, Q. Liu and B. P. Kleinstiver, Nature, 2025, 643, 539–550.
189. W. Du, Z. Cheng, X. Pan, C. Liu, M. Yue, T. Li, Z. Xiao, L. Li, X. Zeng, X. Lin, F. Li and L.-B. Dong, Angew. Chem., Int. Ed., 2025, 64, e202419463.
190. Y. Ao, S. Pei, C. Xiang, M. J. Menke, L. Shen, C. Sun, M. Dörr, S. Born, M. Höhne and U. T. Bornscheuer, Angew. Chem., Int. Ed., 2023, 62, e202301660.

---

Footnote

† These authors contributed equally to this work.

---