Assembly-Line Catalysis in Bifunctional Terpene Synthases

CONSPECTUS:

The magnificent chemodiversity of nearly 95,000 terpenoid natural products identified to date largely derives from catalysis by two types of terpene synthases, prenyltransferases and cyclases. Prenyltransferases utilize 5-carbon building blocks in processive chain elongation reactions to generate linear C_5n isoprenoid diphosphates (n ≥ 2), which in turn serve as substrates for terpene cyclases that convert these linear precursors into structurally complex hydrocarbon products containing multiple rings and stereocenters. Terpene cyclization reactions are the most complex organic transformations found in nature, in that more than half of the substrate carbon atoms undergo changes in chemical bonding during a multi-step reaction sequence proceeding through several carbocation intermediates. Two general classes of cyclases are established based on the chemistry of initial carbocation formation, and structural studies from our laboratory and others show that three fundamental protein folds designated α, β, and γ govern this chemistry. Catalysis by a class I cyclase occurs in an α domain, where a trinuclear metal cluster activates the substrate diphosphate leaving group to generate an allylic cation. Catalysis by a class II cyclase occurs in a β domain or at the interface of β and γ domains, where an aspartic acid protonates the terminal π bond of the substrate to yield a tertiary carbocation. Crystal structures reveal domain architectures of α, αβ, αβγ, βγ, and β.

In some terpene synthases, these domains are combined to yield bifunctional enzymes that catalyze successive biosynthetic steps in assembly-line fashion. Structurally characterized examples include bacterial geosmin synthase, an αα domain enzyme that catalyzes a class I cyclization reaction of C₁₅ farnesyl diphosphate in one active site and a transannulation-fragmentation reaction in the other to yield C₁₂ geosmin and C₃ acetone products. In comparison, plant abietadiene synthase is an αβγ domain enzyme in which C₂₀ geranylgeranyl diphosphate undergoes tandem class II-class I cyclization reactions to yield the tricyclic product. Recent structural studies from our laboratory show that bifunctional fungal cyclases form oligomeric complexes for assembly-line catalysis. Bifunctional (+)-copalyl diphosphate synthase adopts (αβγ)₆ architecture in which the α domain generates geranylgeranyl diphosphate, which then undergoes class II cyclization in the βγ domains to yield the bicyclic product. Bifunctional fusicoccadiene synthase adopts (αα)₆ or (αα)₈ architecture in which one α domain generates geranylgeranyl diphosphate, which then undergoes class I cyclization in the other α domain to yield the tricyclic product. The prenyltransferase α domain mediates oligomerization in these systems. Attached by flexible polypeptide linkers, cyclase domains splay out from oligomeric prenyltransferase cores.

In this Account, we review structure-function relationships for these bifunctional terpene synthases, with a focus on the oligomeric systems studied in our laboratory. The observation of substrate channeling for fusicoccadiene synthase suggests a model for dynamic cluster channeling in catalysis by oligomeric assembly-line terpenoid synthases. Resulting efficiencies in carbon management suggest that such systems could be particularly attractive for use in synthetic biology approaches to generate high-value terpenoid natural products.

Graphical Abstract

Introduction

The family of terpenoid natural products, including steroids, contains 94,406 distinct compounds as currently documented in the Dictionary of Natural Products (http://dnp.chemnetbase.com). The incredible structural and stereochemical diversity of this chemical library belies its simple roots in the initial steps common to the biosynthesis of nearly all terpenoid natural products. First, a prenyltransferase catalyzes the coupling of dimethylallyl diphosphate (DMAPP) and one or more molecules of isopentenyl diphosphate (IPP) in processive, head-to-tail fashion to generate a linear isoprenoid such as geranyl diphosphate (GPP), farnesyl diphosphate (FPP), or geranylgeranyl diphosphate (GGPP) (Figure 1).^5,6 These isoprenoid diphosphates are then utilized as substrates by terpenoid cyclases to yield myriad hydrocarbon products generally containing multiple rings and stereocenters. Terpenoid cyclases catalyze remarkably complex cyclization cascades with structural and stereochemical precision and are largely responsible for the vast chemodiversity of terpenoid natural products.

Figure 1.

Prenyltransferases catalyze the coupling chemistry of primary metabolism to yield linear isoprenoid diphosphates, which then serve as cyclization substrates in secondary (specialized) metabolism. Abbreviations: BPPase, bornyl diphosphate synthase; CPPase, chrysanthemyl diphosphate synthase; DMAPP, dimethylallyl diphosphate; FPPase, farnesyl diphosphate synthase; GGPPase, geranylgeranyl diphosphate synthase; IPP, isopentenyl diphosphate; LSase, limonene synthase; SQSase, squalene synthase; TSase, trichodiene synthase; TxSase, taxadiene synthase. Reprinted with permission from ref. 6. Copyright 2007 American Association for the Advancement of Science.

The hallmark of an isoprenoid coupling or cyclization reaction is a multi-step mechanism proceeding through several carbocation intermediates. In a class I terpenoid synthase, initial carbocation formation is achieved by metal-triggered dissociation of the isoprenoid diphosphate group to yield an allylic cation–inorganic pyrophosphate (PP_i) ion pair. This is the catalytic strategy utilized by prenyltransferases such as FPP synthase or GGPP synthase (Figure 2),^7,8 where 3 Mg²⁺ ions activate the diphosphate leaving group for departure.⁹ Metal ligands are conserved in two identical sequence motifs, DDXXD (boldface indicates metal-coordinating residues). An identical catalytic strategy is utilized by class I terpenoid cyclases, which contain one DDXXD motif that coordinates to Mg²⁺_A and Mg²⁺_C, and the “NSE/DTE” motif (N,D)D(L,I,V)X(S,T)XXXE which coordinates to Mg²⁺_B (boldface indicates metal-coordinating residues).^10,11 Occasionally the (S,T) ligand of the NSE/DTE motif is replaced by a non-liganding residue such as glycine,¹² and there is at least one example of a cyclase in which the NSE/DTE motif is completely replaced by a DDXXE motif.¹³ In a class II terpenoid cyclase, initial carbocation formation is typically achieved by protonation of the terminal carbon-carbon double bond of the isoprenoid to yield a tertiary carbocation. The general acid in this chemistry is the central aspartate conserved in the signature sequence motif DXDD.^14,15 In some cases, e.g., oxidosqualene cyclase, the substrate is an isoprenoid epoxide which is protonated by an aspartic acid usually contained in a divergent general acid motif.¹⁶

Figure 2.

Processive chain elongation reactions catalyzed by the prenyltransferases FPP synthase and GGPP synthase. Substrate ionization is triggered by coordination of the substrate diphosphate group to a trinuclear metal cluster. The maximum length of the growing isoprenoid chain is governed by the depth of the active site cavity. OPP = diphosphate, ^–OPP = inorganic pyrophosphate.

The first crystal structure of a class I terpenoid synthase was that of avian FPP synthase, which exhibits a single-domain α-helical fold assembled with dimeric quaternary structure (Figure 3a).¹⁷ Subsequently reported X-ray crystallographic studies of GGPP synthase revealed similar dimeric quaternary structures,^18,19 although hexameric quaternary structures can result from trimerization of the dimer.^1,3,20 A recent cryo-electron microscopy (cryo-EM) study revealed tetramerization of a GGPP synthase dimer to form an octamer.⁴

Figure 3.

(a) The dimeric prenyltransferase farnesyl diphosphate synthase yielded the first crystal structure of a class I terpene synthase with an α fold (PDB 1FPS). Metal-binding DDXXD motifs are red. (b) The α fold of a class I terpene synthase is found in pentalenene synthase (PDB 1PS1), in the αβ domain assembly of epi-aristolochene synthase (PDB 5EAS), and in the αβγ domain assembly of taxadiene synthase (PDB 3P5P). Domains are color-coded as follows: α = blue, β = green, γ = yellow. Metal-binding motifs in each α domain (DDXXD = red, NSE/DTE = orange) coordinate to catalytically obligatory metal ions. The class II terpenoid synthase fold is found in the βγ domain assembly of squalene-hopene cyclase (PDB 1SQC). An aspartic acid motif (brown) is the general acid that initiates catalysis. The βγ domain assembly also appears in taxadiene synthase, where it is vestigial and has no role in catalysis other than stabilizing the fold of the α domain. Reprinted with permission from ref. 26. Copyright 2010 Nature Publishing Group.

The first crystal structures of class I terpenoid cyclases, bacterial pentalenene synthase and plant epi-aristolochene synthase (Figure 3b), revealed catalytic domains with α-helical folds topologically similar to that of FPP synthase despite insignificant amino acid sequence identity, indicating that isoprenoid coupling and cyclase enzymes diverged from a common ancestor early in the evolution of terpenoid biosynthesis.^21,22 The evolutionary origins of what is now designated²³ the α domain of a class I terpenoid synthase are rooted in gene duplication and fusion of an ancient 4-helix bundle protein.²⁴ Since the DDXXD metal-binding motifs are retained on topologically identical helices of the coupling enzymes, but the second one has usually diverged to the NSE/DTE motif in the cyclases, the evolutionary origin of the class I cyclases is most likely a primordial isoprenoid coupling enzyme.

The first crystal structure of a class II terpenoid cyclase was that of squalene-hopene cyclase, which revealed double-domain βγ architecture (Figure 3b).¹⁵ Since these domains exhibit 23% sequence identity based on three-dimensional alignment, this domain architecture likely resulted from gene duplication and fusion.¹⁴ The catalytically obligatory general acid motif DXDD is located in the β domain. Most recently, the crystal structure of a minimalist class II terpenoid cyclase was reported, merosterolic acid synthase, that consists of just a single β domain.²⁵

The most complex domain assemblies for both class I and class II cyclases exhibit αβγ architecture. The first such enzymes to yield crystal structures were the class I cyclase taxadiene synthase (Figure 3b),²⁶ the class II cyclase ent-copalyl diphosphate synthase,²⁷ and the bifunctional class I-class II cyclase abietadiene synthase.²⁸ Regardless of the cyclase class, the tertiary structure of αβγ domain assembly is identical in all three enzymes.²⁹

Curiously, examples of bifunctional terpenoid cyclases exhibiting αα domain architecture have been described. For example, geosmin synthase catalyzes a cyclization-transannulation-fragmentation sequence to yield acetone and the earthy odorant geosmin,^30,31 and fusicoccadiene synthase catalyzes the GGPP synthase reaction followed by a cyclization reaction to yield fusicoccadiene in the biosynthesis of Fusiccocin A.³² In the remainder of this Account, we outline structure-function relationship for bifunctional terpenoid cyclases with αβγ and αα domain architectures to provide a framework for understanding assembly-line reaction sequences in the first steps of terpenoid biosynthesis.

Abietadiene Synthase from Abies grandis

In response to physical or biological assault, plant conifers such as pine, fir, and spruce secrete oleoresin, a complex mixture of turpentine (C₁₀ monoterpenes and C₁₅ sesquiterpenes) and rosin (C₂₀ diterpenes) into the newly-formed wound.³³ The turpentine fraction solubilizes the higher-molecular weight diterpenoid resin acids comprising the rosin fraction, but the turpentine fraction evaporates upon exposure to the atmosphere. Oxidation and crosslinking of the remaining resin acids create a hardened shell of rosin that plugs the wound.³⁴ This is the source of rosin that is used to treat a horsehair bow to improve its traction on the strings of a ‘cello, or in powdered form the rosin bag used by baseball pitchers to keep their hands dry and improve their grip on the ball.

The principal component of rosin is the diterpenoid (–)-abietic acid. The first committed step in the biosynthesis of (–)-abietic acid is the generation of (–)-abieta-7(8),13(14)-diene (or simply “abietadiene”) from GGPP through consecutive cyclization reactions catalyzed by abietadiene synthase in two distinct active sites (Figure 4a).^35–37 The first reaction utilizes GGPP in a class II cyclization reaction to generate (+)-copalyl diphosphate, which in turn undergoes a class I cyclization reaction to yield three abietadiene regioisomers.^37,38 Truncation studies failed to produce independent and functional class I α and class II βγ domain constructs, indicating that protein stability and the structural integrity of each catalytic domain largely depend on significant interactions at the α-β domain interface.³⁸ The detection of (+)-copalyl diphosphate in solution during steady-state kinetic measurements argues against substrate channeling between active sites, leading to the conclusion that (+)-copalyl diphosphate diffuses out of the class II active site into solution and then rebinds in the class I active site for the final cyclization reaction.³⁷

Figure 4.

(a) Generation of abietadiene regioisomers in tandem cyclization reactions catalyzed by abietadiene synthase. The reaction is initiated in the class II active site by general acid D404 (B:H) to yield (+)-copalyl diphosphate (2), which dissociates from the class II active site and reassociates with the class I active site. There, it undergoes metal-triggered ionization and further cyclization to form abietadiene (5), levopimaradiene (6) and neoabietadiene (7). Reprinted with permission from ref. 38. Copyright 2003 American Chemical Society. (b) Comparison of class I and class II plant terpene cyclases with αβγ domain architecture. Catalytically-active domains are shown in color (α = blue, β = green, γ = yellow) and catalytically-inactive domains are shown in gray. The α-helical linker connecting the α and β domains is red. Reprinted with permission from ref. 3, Copyright 2020, with permission from Elsevier.

The 2.3 Å-resolution crystal structure of the abietadiene synthase monomer was the first reported crystal structure of a bifunctional terpene cyclase,²⁸ revealing αβγ domain architecture identical to that of the αβγ diterpene cyclases, taxadiene synthase (class I) and ent-copalyl diphosphate synthase (class II).^26,27 This structural analysis highlighted the N-terminal class II active site located at the βγ domain interface, the class I active site within the C-terminal α domain, the structural contacts between the α and β domains maintaining the stability of the αβγ fold, and the α-helix linking the α and β domains (Figure 4b).²⁸

(+)-Copalyl Diphosphate Synthase from Penicillium verruculosum

Genome mining in Penicillium species recently led to the identification of two bifunctional diterpene synthases, each of which catalyzes sequential GGPP synthase and (+)-copalyl diphosphate synthase reactions (Figure 5a).³⁹ Amino acid sequence analysis predicted αβγ domain architecture for each enzyme, with the C-terminal α domain comprising the class I prenyltransferase and the N-terminal βγ domains comprising the class II cyclase. However, in contrast with αβγ plant terpene synthases (Figure 4b), where the α and β domains are connected by a 33-residue α-helix, the α and β domains of bifunctional (+)-copalyl diphosphate synthases from P. verruculosum and P. fellutanum are connected by 170- and 110-residue linkers, respectively. Truncation of these long linker segments does not affect catalytic activity, in contrast with other αβγ cyclases where interdomain interactions are required to stabilize catalytic domains for optimal activity.^2,38,40 It has not been ascertained whether GGPP is released to solution and then rebinds to the cyclase domain, or if there is substrate channeling between the prenyltransferase and cyclase active sites.

Figure 5.

(a) Assembly-line reactions catalyzed by fungal (+)-copalyl diphosphate synthase enzymes. (b) Crystal structure of the prenyltransferase domain of (+)-copalyl diphosphate synthase from Penicillium verruculosum (PDB 6V0K). The hexamer assembles with D3 symmetry; subunits and their symmetry equivalents are labeled. (c) The molecular envelope of full-length (+)-copalyl diphosphate synthase calculated from SAXS data is fit with the crystal structure of the α₆ prenyltransferase core. Pendant βγ cyclase domains splay out from the core with starburst-like geometry. Reprinted with permission from ref. 3, Copyright 2020, with permission from Elsevier.

Recombinant expression and purification of (+)-copalyl diphosphate synthase from Penicillium verruculosum yields an assembly with hexameric quaternary structure previously unobserved for αβγ cyclases.³ However, oligomerization is concentration-dependent, such that the hexamer dissociates into lower-order species at more dilute concentrations. Prenyltransferase activity remains generally unchanged at varying concentrations and oligomeric states.³ It remains unknown whether concentration or oligomerization influences cyclase activity.

Crystallization using limited proteolysis afforded the 2.41 Å-resolution structure of the hexameric N-terminal prenyltransferase α domain (Figure 5b).³ This hexameric assembly is identical to that first observed in hexameric human GGPP synthase.²⁰ Small-angle X-ray scattering (SAXS) experiments conducted with full-length (+)-copalyl diphosphate synthase yield a low resolution molecular envelope into which a model of the hexamer could be fit (Figure 5c).³ In this model, the hexameric prenyltransferase core is surrounded by six pendant βγ cyclase domains with overall starburst-like geometry. The SAXS data indicated significant flexibility of the 170-residue linker segment.

Geosmin Synthase from Streptomyces coelicolor

Geosmin is a volatile sesquiterpene derivative that is responsible for musty odors in food and water contaminated with Streptomyces.⁴¹ Geosmin synthase from Streptomyces coelicolor is the first bifunctional terpene cyclase discovered in bacteria and adopts αα domain architecture: the N-terminal α domain catalyzes the class I ionization-dependent cyclization of FPP to form a mixture of germacradienol and germacrene D in a ratio of 85:1, and the C-terminal α domain catalyzes the protonation-dependent transannulation and fragmentation of germacradienol to form geosmin and acetone (Figure 6a).³¹ The two α domains are connected by a 55-residue linker. Truncated constructs of isolated α domains of geosmin synthase are active with only slight loss of catalytic efficiency compared to the full-length enzyme, indicating that the α domains do not require close association to maintain structural integrity.⁴² The detection of intermediates germacradienol and germacrene D in solution suggests that geosmin synthase does not employ substrate channeling in the transfer of germacradienol from one active site to the other.³¹

Figure 6.

(a) Cyclization-transannulation-fragmentation sequence catalyzed by geosmin synthase from S. coelicolor. Black reaction arrows indicate the cyclization reaction catalyzed by the N-terminal α domain, and green reaction arrows indicate the transannulation-fragmentation reaction catalyzed by the C-terminal α domain. (b) Crystal structure of the cyclization domain (blue; PDB 5DZ2) and homology model of the transannulation-fragmentation domain (forest green) fit into the molecular envelope of full-length geosmin synthase calculated from SAXS data. Reprinted with permission from ref. 43. Copyright 2015 American Chemical Society.

Full-length geosmin synthase is refractory to crystallization, but the crystal structure of the N-terminal germacradienol synthase domain was solved using limited proteolysis during crystallization.⁴³ Curiously, the C-terminal geosmin synthase domain requires divalent metal ions for catalytic activity and bears 36% sequence identity to the N-terminal domain, yet does not require pyrophosphate binding for catalysis and appears to rely on general acid catalysis for activity.³¹ Although the C-terminal α domain is refractory to crystallization, sequence similarity between the two α domains allowed for the construction of a homology model based on the crystal structure of the N-terminal α domain.⁴³

To gain further structural insight on full-length geosmin synthase, SAXS was employed to generate a molecular envelope into which the N- and C-terminal α domains could be plausibly fit, albeit with multiple orientations (Figure 6b).⁴³ As such, the precise structural elements mediating domain-domain interactions could not be reliably determined. However, the shape of the molecular envelope nonetheless indicates significant interaction between the two domains.

Fusicoccadiene synthase from Phomopsis amygdali

The phytotoxin Fusicoccin A is a tricyclic diterpene that binds to 14-3-3 proteins and stabilizes their protein-protein complexes.⁴⁴ In plants, Fusicoccin A binds to the plasma membrane H⁺-ATPase and irreversibly activates proton pumping, which disrupts the electrochemical proton gradient across the membrane and ultimately leads to plant wilting.^45,46 As 14-3-3 proteins comprise a broad class of adapter proteins involved in the regulation of cellular signaling pathways,⁴⁷ their binding interactions are linked with various pathologies in humans, including cancer and neurological disease.^48,49 Accordingly, Fusicoccin A and its derivatives are emerging as therapeutic leads.

The first committed steps of Fusicoccin A biosynthesis are catalyzed by fusicoccadiene synthase, a bifunctional enzyme that combines the chain-elongation chemistry of GGPP synthase with the GGPP cyclization cascade leading to fusicoccadiene (Figure 7a).³² Although full-length fusicoccadiene synthase is refractory to crystallization, X-ray crystal structures of the cyclase and prenyltransferase domains yield important mechanistic insights.¹ The structure of the cyclase domain reveals an active site cavity that is a cyclization template complementary in shape to the 5-8-5 ring system of fusicoccadiene. The structure of the prenyltransferase domain reveals hexameric quaternary structure. Analysis of the N333A/Q336A double variant using SAXS provides the low-resolution molecular envelope of a hexamer into which the prenyltransferase and cyclase crystal structures could be docked to yield a plausible model of the full-length enzyme.¹ Cryo-EM studies of the wild-type enzyme indicate a ca. 9:1 mixture of octamer:hexamer.⁴ Assuming a facile equilibrium between octamer and hexamer, this corresponds to a free energy difference of only 1.3 kcal/mol.

Figure 7.

(a) Prenyltransferase and cyclase reactions catalyzed by fusicoccadiene synthase from P. amygdali. (b) Cryo-EM reconstruction of the PaFS prenyltransferase octamer (PDB 7JTH). Monomers A and B in each dimer are blue and cyan, respectively. The visible N-termini are indicated in the top view of the structure (left). (c) Model of full-length fusicoccadiene synthase showing prenyltransferase domains (blue) and cyclase domains (green) connected by the flexible 70-residue linker (red spring). Cryo-EM studies show that cyclase domains can adopt splayed-out positions as well as more closely associated positions relative to the octameric prenyltransferase core. Possible trajectories of GGPP transit are illustrated in side view structures. Reprinted with permission from Ref. 4. Copyright 2021, Faylo, J. L.; van Eeuwen, T.; Kim, H. J.; Gorbea Colón, J. J.; Garcia, B. A.; Murakami, K.; Christianson, D. W.

Covalent linkage of prenyltransferase and cyclase domains results in an approximate two-fold advantage in the rate of diterpene generation,¹ similar to the catalytic advantage resulting from covalent linkage of diterpene or sesquiterpene prenyltransferases with other cyclases in protein engineering experiments.^50,51 These catalytic advantages suggest that a proximity effect or substrate channeling might be operative in bifunctional assembly-line catalysis. While the magnitude of rate enhancement appears modest, it is important to note that substrate channeling does not necessarily confer kinetic superiority but instead establishes an efficient carbon management strategy, ensuring that most of the product of one enzyme is transferred to the active site of the second enzyme in the assembly-line sequence of biosynthesis.⁵²

Substrate channeling in fusicoccadiene synthase was further studied in substrate competition experiments.² When an equimolar mixture of fusicoccadiene synthase and taxadiene synthase was incubated with GGPP, an approximately 4:1 mixture of fusicoccadiene:taxadiene results. However, when incubated with DMAPP and IPP, such that the only source of GGPP is that generated by the prenyltransferase domain of fusicoccadiene synthase, the fusicoccadiene:taxadiene product ratio increases to 46:1. Thus, most of the GGPP generated by fusicoccadiene synthase remains on the enzyme for cyclization to form fusicoccadiene, rather than diffusing into solution and rebinding to the cyclase domain. However, some of the GGPP can escape to solution, where it can bind to taxadiene synthase to generate the small quantity of taxadiene observed.

The recent structure determination of full-length fusicoccadiene synthase by cryo-electron microscopy (cryo-EM) reveals that substrate channeling between the prenyltransferase domain and the cyclase domain does not occur through a tunnel.⁴ Instead, a proximity effect appears to be operative. Cryo-EM analysis of the full-length enzyme yields the 3.99 Å-resolution structure of the prenyltransferase with octameric quaternary structure (Figure 7b), but cryo-EM density for the cyclase domain and the linker is absent. Strikingly, negative stain EM images indicate that fusicoccadiene synthase consists of a donut-shaped core of eight prenyltransferase domains surrounded by variably positioned cyclase domains splayed out from the central prenyltransferase core.

Low-resolution cryo-EM structures of fusicoccadiene synthase crosslinked with glutaraldehyde show that cyclase domains can associate more closely with the octameric prenyltransferase core, including a position that caps the central pore into which the GGPP product of the prenyltransferase domains is released.⁴ Thus, GGPP might diffuse out of the prenyltransferase active site, into the central pore, and then into the active site of the cyclase. Alternatively, or additionally, GGPP might diffuse out of the central pore and into the active site of one of the eight splayed-out cyclase domains in constant motion surrounding the central prenyltransferase octamer (Figure 7c). Either model for substrate channeling involves proximity, the latter involving dynamic proximity since the eight cyclase domains are in constant motion. Therefore, assembly-line catalysis in this system appears to involve dynamic cluster channeling⁵² – a modified version of cluster channeling⁵³ – in which the agglomeration of active sites resulting from oligomerization facilitates substrate transfer from one active site to the next in the biosynthetic sequence.

Interdomain linkers

It is easy to overlook the polypeptide linker that connects catalytic domains in bifunctional assembly-line cyclases, especially in systems where the linker is disordered. However, the linker can be a critical component of a bifunctional assembly regardless of the degree to which it is ordered or disordered. In abietadiene synthase, a canonical αβγ diterpene synthase, the linker connecting the α and β domains is ordered as a 33-residue α-helix (Figure 4b).²⁸ In contrast, the αβγ domain architecture of (+)-copalyl diphosphate synthase is perhaps better symbolized as α–βγ due to the 170-residue linker connecting the α domain and the β domain; 119/170 residues are predicted to be disordered.³ The linker governs the separation and relative positions of enzyme active sites catalyzing sequential biosynthetic reactions. If the linker segment is predominantly disordered, how do we better understand its ensemble of structures as it connects the functions of an assembly-line terpene synthase?

The linker segments of (+)-copalyl diphosphate synthase, geosmin synthase, and fusicoccadiene synthase correspond to intrinsically disordered regions.^54,55 As a fascinating class of biomolecules, intrinsically disordered regions, peptides, and proteins have been studied using bioinformatics analyses and computational approaches that indicate general properties of low hydrophobicity, a high fraction of charged residues, and no persistent secondary or tertiary structure.^56,57 Most intrinsically disordered proteins are polyampholytes. Computational studies indicate that weak polyampholytes tend to form globules, whereas the structural tendencies of strong polyampholytes tend to vary depending on the fraction of charged residues and their distribution in the linear amino acid sequence;⁵⁶ strong polyampholytes can even form well-solvated coils or even stiff rods.⁵⁷ In many cases, amino acid sequences of intrinsically disordered peptides are classified at the boundary between weak and strong polyampholytes and are termed Janus sequences.⁵⁶

The classification of the amino acid sequences of linker segments in 15 bifunctional assembly-line terpene synthases ranging from 53–166 residues in length using localCIDER⁵⁷ is shown in Figure 8. For comparison, the 33-residue linker segments of bifunctional abietadiene synthase and the class I terpene cyclase taxadiene synthase, each of which adopt well-ordered α-helical conformations (Figure 4b), are also indicated in Figure 8. Curiously, these α-helical linkers are predicted to exhibit widely dispersed properties. All linker segments, with the exception of that of abietadiene synthase, are classified either as weak polyampholytes predicted to adopt globule-like structures, or Janus sequences which can adopt collapsed or expanded conformations depending on context. These classifications are consistent with the classification of a wider array of disordered polypeptide sequences from the Database of Protein Disorder analyzed by Das and Pappu.⁵⁶

Figure 8:

Sequence classification of intrinsically disordered linker segments in bifunctional terpene synthases using localCIDER. The panel at right is a close-up view of the panel at left, with each bifunctional enzyme linker labeled.

Interestingly, analysis of intrinsically disordered peptides in different contexts suggests that amino acid sequences with segregated residues of opposite charges are more likely to adopt different conformations depending on the potential for interaction with a protein surface, whereas such residues in globule-like conformations are more likely to make intra-globule interactions irrespective of their position relative to a functional domain.⁵⁸

General structure-function trends for assembly-line terpene synthases

Although the domain architectures of terpene cyclases are conserved as α, αα, αβ, αβγ, βγ, and β in different taxonomic domains of life,^23,29 bifunctional terpene synthases are thus far restricted to αα or αβγ domain architectures as represented by the cyclase-cyclase fusions and the prenyltransferase-cyclase fusions discussed herein. Structurally characterized cyclase-cyclase fusions are monomers in solution, while prenyltransferase-cyclase fusions form hexamers and octamers, with the prenyltransferase domain driving oligomerization. Might oligomerization facilitate substrate channeling? As observed for fusicoccadiene synthase, although some of the GGPP generated by the prenyltransferase escapes to solution, most of it remains on the enzyme assembly for cyclization to form fusicoccadiene.² In principle, oligomerization could increase the probability of dynamic cluster channeling by establishing a continuously moving array of cyclases surrounding the prenyltransferase core. Notably, the release of some GGPP to solution does not preclude substrate channeling; instead, this indicates simply that the efficiency of substrate transfer from one active site to the other is imperfect. Reaction intermediates of monomeric cyclase-cyclase enzymes are similarly detected in solution,^31,37 but it is unknown if here, too, some percentage of the reaction intermediate might stay on the enzyme for the subsequent cyclization reaction.

Polypeptide linker segments connecting the two catalytic domains tend to be longer in prenyltransferase-cyclase than in cyclase-cyclase assemblies. Linker segments have minimal secondary structure as predicted by computational analyses and corroborated in biophysical studies. However, it is notable that the 33-residue linkers of αβγ terpene synthases such as abietadiene synthase adopt well-ordered α-helical conformations. Linker length governs the separation of catalytic domains and hence can influence proximity channeling. There is low amino acid sequence similarity in linker sequences even among homologous bifunctional αα or αβγ enzymes. As intrinsically disordered peptides, however, linker segments are similar in that they are generally predicted to adopt globule-like structures (Figure 8). Regardless, it is not clear how important the wide range of linker lengths is for assembly-line catalysis. The evolutionary retention of long unstructured interdomain linkers suggests that these linkers serve some structural or functional purpose, which may be as simple as holding the two catalytic domains within a certain proximity of each other.

Protein engineering experiments with bifunctional systems begin to reveal the influence of the interdomain linker on catalysis. Varying the interdomain linker length and rigidity in engineered FPP synthase-patchoulol synthase fusion proteins does not significantly affect product flux, although shorter linkers are generally more favorable.⁵⁹ Fusion alone is insufficient to enable substrate channeling.^60,61 Nevertheless, fusion enzymes sometimes exhibit modest ~2-fold catalytic advantages, as demonstrated for fusicoccadiene synthase¹ as well as engineered prenyltransferase-cyclase fusion proteins.^50,51

Metabolic engineering efforts to control product flux include a focus on facilitating protein-protein proximity, e.g., by utilizing protein or nucleic acid scaffolds, specific communication domains, or, most simply, the fusion of open reading frames. While fusion of catalytic domains alone generally appears to be insufficient for inducing substrate channeling, oligomerization appears to form a versatile platform for catalysis from which GGPP appears to resist diffusion into solution even in the absence of a direct channel.^4,29 Perhaps the structural features of a bifunctional oligomeric terpene synthase, e.g., complementary electrostatics in the prenyltransferase, the disordered linker, and/or the cyclase, facilitate the retention and transfer of GGPP from one active site to another. Moreover, the agglomeration of prenyltransferase and cyclase domains would increase the local concentration of intermediate GGPP, sequestering it from competing diterpene cyclization pathways and enabling more efficient biosynthesis of a specific terpene product. We expect that future studies in metabolic engineering and synthetic biology will capitalize on these principles to optimize the generation of high-value terpene natural products in heterologous hosts.

Biographies

Jacque L. Faylo

Jacque Faylo received her bachelor’s degree in Chemistry from the University of Pittsburgh in 2016. She then began her Ph.D. studies at the University of Pennsylvania, where she joined the Christianson laboratory to explore the structural biology of assembly-line terpene synthases. She will soon begin her postdoctoral studies in Prof. Michael Burkhart’s laboratory at the University of California, San Diego, where she will study natural products biosynthesis.

Trey A. Ronnebaum

Trey A. Ronnebaum obtained his Ph.D. in the laboratory of Prof. Audrey L. Lamb at the University of Kansas where he studied the structures and mechanisms of nonribosomal peptide synthetases involved in the biosynthesis of the siderophore pyochelin. Trey is fascinated with natural product biosynthesis and continued in the field joining Prof. David W. Christianson’s laboratory as a postdoc at the University of Pennsylvania studying terpene biosynthesis. At Penn, Trey has been involved in Penn Biotech Group Healthcare Consulting as the Vice President of Business Development and enjoys playing golf and being active in his free time.

David W. Christianson

David W. Christianson is the Roy and Diana Vagelos Professor in Chemistry and Chemical Biology and Chair of the Department of Chemistry at the University of Pennsylvania, where he arrived in 1988 after completing the A.B., A.M., and Ph.D. degrees in chemistry at Harvard University. Christianson’s research program has been recognized by the Pfizer Award in Enzyme Chemistry (1999), a Guggenheim Fellowship (2006), and the Repligen Award in Chemistry of Biological Processes (2013), and his teaching accomplishments have been recognized by the Lindback Award for Distinguished Teaching at Penn (2017) and the Rhodes Trust Inspirational Educator Award, Oxford University (2019). Christianson was also the Underwood Fellow in the Department of Biochemistry and Fellow in the Natural Sciences at Sidney Sussex College, University of Cambridge (2006), the Elizabeth S. and Richard M. Cashin Fellow at the Radcliffe Institute for Advanced Study at Harvard University (2015), and Visiting Professor of Chemistry and Chemical Biology at Harvard University (2016). Christianson has been studying enzymes of terpenoid biosynthesis ever since reporting the first bacterial terpene cyclase structure nearly 25 years ago.²¹