Reprogramming the Cyclization Cascade of epi-Isozizaene Synthase to Generate Alternative Terpene Products

Abstract

The class I sesquiterpene cyclase epi-isozizaene synthase from Streptomyces coelicolor (EIZS) catalyzes the transformation of linear farnesyl diphosphate (FPP) into the tricyclic hydrocarbon epi-isozizaene in the biosynthesis of albaflavenone antibiotics. The active site cavity of EIZS is largely framed by four aromatic residues – F95, F96, F198, and W203 – that form a product-shaped contour, serving as a template to chaperone conformations of the flexible substrate and multiple carbocation intermediates leading to epi-isozizaene. Remolding the active site contour by mutagenesis can redirect the cyclization cascade away from epi-isozizaene biosynthesis to generate alternative sesquiterpene products. Here, we present the biochemical and structural characterization of four EIZS mutants in which aromatic residues have been substituted with polar residues (F95S, F96H, F198S, and F198T) to generate alternative cyclization products. Most notably, F95S EIZS generates a mixture of monocyclic sesquiterpene precursors of bisabolane, a D2 diesel fuel substitute. X-ray crystal structures of the characterized mutants reveal subtle changes in the active site contour showing how each aromatic residue influences the chemistry of a different carbocation intermediate in the cyclization cascade. We advance that EIZS may serve as a robust platform for the development of designer cyclases for the generation of high-value sesquiterpene products ranging from pharmaceuticals to biofuels in synthetic biology approaches.

Graphical Abstract

Introduction

Nature is the most powerful and talented synthetic organic chemist, as represented by the tremendous chemodiversity and biological activities of organic molecules in all domains of life.^1–3 Chief among these natural products are presently 102,306 terpenes (including steroids and carotenoids) that comprise the largest class of molecules indexed in the Dictionary of Natural Products (www.dnp.chemnetbase.com).⁴ Terpenes serve myriad specialized functions in nature,^5–9 and some exhibit useful pharmacological properties. For example, the taxane diterpene paclitaxel (Taxol) is used in cancer chemotherapy,^10–12 the sesquiterpene artemisinin is a well-known antimalarial drug,^14,15 and the monoterpene menthol is an analgesic that activates the thermosensitive ion channel protein TRPM8,^16,17 leading to the incorporation of menthol in cough drops, skin ointments, and other palliative formulations^18,19 (Figure 1, left). Terpenes are also prominent in the energy industry as renewable carbon-based fuels – the saturated sesquiterpenes farnesane and bisabolane are advanced biofuel alternatives to petrochemically-derived jet fuel and D2 diesel fuel, respectively.^20–23

Figure 1.

Biosynthesis of selected terpene natural products. Abbreviations: DMAPP, dimethylallyl diphosphate; IPP, isopentenyl diphosphate; OPP, diphosphate; GPP, geranyl diphosphate; FPP, farnesyl diphosphate; GGPP, geranylgeranyl diphosphate. The sesquiterpene cyclase epi-isozizaene synthase (EIZS) catalyzes the cyclization of FPP to form epi-isozizaene (green box).

All terpene natural products ultimately derive from the 5-carbon precursors isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP).²⁴ In primary metabolism, prenyltransferases catalyze the head-to-tail coupling of one or more IPP molecules to DMAPP to generate a linear isoprenoid diphosphate such as C₁₀-geranyl diphosphate (GPP), C₁₅-farnesyl diphosphate (FPP), or C₂₀-geranylgeranyl diphosphate (GGPP).^25–28 These isoprenoid diphosphates then serve as substrates for the enzymes of secondary (specialized) metabolism, terpene cyclases, which catalyze some of the most exquisitely complex carbon-carbon bond-forming reactions found in nature to generate products typically containing multiple rings and stereocenters (Figure 1).^27–32 This chemistry is achieved through a tightly-regulated reaction sequence, the hallmark of which is the precise manipulation of multiple carbocation intermediates. Cyclic terpenes then serve as scaffolds for oxidation by monooxygenases and further derivatization by downstream enzymes to yield the final natural product.

For example, the antibiotic albaflavenone derives from the cytochrome P450 170A1-catalyzed oxidation of epi-isozizaene, a tricyclic sesquiterpene generated by epi-isozizaene synthase (EIZS) from Streptomyces coelicolor (Figure 1, green box).^33–35 Wild-type EIZS is a high-fidelity cyclase, generating 99% epi-isozizaene at 4° C; biosynthetic fidelity relaxes with increasing temperature, with 93% yield at 20° C and 79% yield at 30° C.^35,36 The relaxation of fidelity enables the generation of alternative cyclization products, which in turn hints at the potential biosynthetic product chemodiversity that might be harnessed using EIZS as a platform for protein engineering.

The active site of EIZS is situated in the center of an α-helix bundle, characteristic of a type I terpene cyclase, and metal-binding motifs at the mouth of the active site coordinate to a trinuclear magnesium cluster responsible for triggering substrate ionization and generation of an allylic carbocation intermediate.^35,37 The active site contour, largely defined by aromatic residues F95, F96, F198, and W203, serves as a template as it chaperones the binding conformation of FPP.³⁸ The active site contour can be remolded by mutagenesis to redirect the cyclization cascade to generate new cyclization products. For example, substitution of F96 with serine, methionine, or glutamine converts EIZS into a sesquisabinene synthase.³⁹ The hydrophobic active site is surprisingly tolerant of polar mutagenesis, in that a serine hydroxyl group can be introduced in F96S EIZS without being alkylated by a reactive carbocation intermediates. Moreover, the F96H substitution converts EIZS into a high-fidelity nerolidol synthase.³⁹

Here, we report the structural and biochemical characterization of four EIZS mutants in which polar amino acids are substituted for aromatic residues in the active site. The crystal structures of unliganded F96H EIZS and its complexes with two different phosphonate inhibitors advance our understanding of how this mutant accommodates and manipulates water to quench the nerolidyl cation to generate an alcohol product. Crystal structures of F95S EIZS, F198S EIZS, and F198T EIZS complexed with the benzyltriethylammonium cation (BTAC) and inorganic pyrophosphate (PP_i) show how a carbocation mimic is stabilized in the active site. Additionally, kinetic analyses and product array determinations of these EIZS mutants reveal how polar amino acid substitutions in the active site redirect the FPP cyclization cascade to generate alternative products.

Materials and Methods

Reagents.

All reagents for protein expression and purification were purchased from Fisher, Sigma, or GoldBio. Crystallization reagents were purchased from Hampton Research. Ligands for cocrystallization (Na₄P₂O₇ (PP_i), benzyltriethylammonium chloride (BTAC), pamedronate, and risedronate) were purchased from Fisher. The EnzChek assay kit was purchased from Fisher, and FPP was purchased from Isoprenoids, LC. Unless noted otherwise, all reagents were used without further purification.

Cloning.

The gene encoding wild-type EIZS, codon-optimized for expression in E. coli, was purchased from Genscript in a pET28 vector. Mutagenic primers (Table S1) were designed using the NEBaseChanger tool, and mutations were introduced into the gene using a Q5 mutagenesis kit (New England Biolabs) according to the manufacturer’s protocol. The NEB-5α cells were transformed with mutant plasmid, plated on selective media, and individual colonies were selected and grown in 5 mL of LB media supplemented with 50 μg/mL kanamycin (overnight, 37 °C, 250 rpm). Plasmid DNA was extracted using a Qiagen Miniprep kit, and successful incorporation of the mutation was confirmed via sequencing performed at the DNA sequencing facility at the University of Pennsylvania.

Protein Expression and Purification.

Wild-type and mutant proteins were prepared as previously described.^35,36,38,39 Briefly, BL21(DE3) cells harboring sequence-confirmed EIZS plasmid DNA were grown in LB media supplemented with 50 μg/mL kanamycin until the OD₆₀₀ reached ~0.5. Cultures were cooled and protein expression was induced upon the addition of 0.1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) and incubated overnight. Cells were pelleted by centrifugation and stored at −80 °C until purification. Wild-type and mutant proteins were purified using a 5 mL TALON HiTrap column (GE healthcare) for affinity chromatography, followed by size-exclusion chromatography on a HiLoad 26/600 Superdex 200 gel filtration column. Proteins were considered >95% pure on the basis of SDS-PAGE. Purified enzymes were concentrated to 7–11 mg/mL using an Amicon Ultra-15 centrifugation concentrator (30 kDa MWCO), flash-cooled in liquid nitrogen, and stored at −80 °C until use.

Crystallization.

A seed stock of wild-type EIZS crystals was prepared according to previous methods.³⁸ Briefly, wild-type EIZS was co-crystallized with 2 mM benzyltriethylammonium chloride (BTAC), 10 mM MgCl₂, and 2 mM sodium pyrophosphate (PP_i) by the sitting drop vapor diffusion method. Crystals appeared overnight and were crushed, diluted with mother liquor, and stored at −80 °C until use in all subsequent crystallization experiments.

All crystals of EIZS mutants were grown by the sitting drop vapor diffusion method. All EIZS mutants were incubated with millimolar concentrations of ligands prior to crystallization (Table S2). Most commonly, cocrystallization with Mg²⁺, PP_i, and BTAC yielded EIZS structures with closed active site conformations stabilized by Mg²⁺₃-PP_i binding at the mouth of the active site, with BTAC encapsulated in the active site. Alternatively, the use of the bisphosphonate inhibitors pamidronate or risedronate yielded EIZS structures with closed active site conformations stabilized by Mg²⁺₃-bisphosphonate binding at the mouth of the active site, which mimics Mg²⁺₃-PP_i binding. By trial-and-error, we observed that each mutant crystallized more readily with one ligand than another.

A Mosquito crystallization robot was used to screen crystallization conditions. Typically, 300 nL of protein-ligand solution was mixed with 300 nL of precipitant solution and 25 nL of 1:100 seed solution, and equilibrated against 75 μL reservoir of precipitant solution. Large hexagonal plates appeared overnight and reached maximum dimensions within one week. The composition of protein and precipitant solutions yielding optimal crystals for X-ray data collection are detailed in Table S2. All crystals were preserved in a cryosolution consisting of mother liquor augmented with 20% (v/v) glycerol before being flash-cooled in liquid nitrogen.

Crystallographic Data Collection and Structure Determination.

Diffraction data were collected remotely at the 17-ID-2 FMX beamline at the National Synchrotron Light Source II (NSLS-II), Brookhaven National Laboratory (Upton, NY) and the Northeastern Collaborative Access Team (NE-CAT) beamlines 24-ID-C/E at the Advanced Photon Source (APS), Argonne National Laboratory (Lemont, IL). Datasets were indexed, integrated, and scaled using XDS⁴⁰ then reduced using AIMLESS.⁴¹ Initial electron density maps were phased using the Phaser module⁴² of PHENIX,⁴³ using wild-type EIZS (PDB 3KB9) or F96S EIZS (7KJE), less ligands and solvent molecules, as initial search models. Structure refinements were performed using PHENIX.refine⁴³ and manual model building was performed in COOT.^43,44 Polypeptide loops and residue side chains with limited or no electron density in 2mF_o – DF_c and composite omit maps were truncated or removed, and are listed in Table S3. Solvent molecules were manually placed in peaks ≥ 3σ in the |F_o|-|F_c| map. Ligands and other nonproteogenic molecules were added to the model during the final stage of refinement. Molprobity⁴⁵ was used to validate all final models. Data collection and refinement statistics are listed in Table 1. Structure figures were prepared with PyMOL (PyMOL Molecular Graphics System, version 2.0, Schrödinger, LLC). All structural superpositions and root-mean-square deviation (RMSD) calculations were performed using the Align feature in PyMOL, and all active site contours were generated using the GetCleft component of the NRGsuite plug-in for PyMOL.⁴⁶

Table 1.

Crystallographic Data Collection and Refinement Statistics

	F96H (unliganded)	F96H–Mg2+3–pamidronate	F96H–Mg2+3–risedronate	F95S–Mg2+3–PPi–BTAC	F198S–Mg2+3–PPi–BTAC	F198T–Mg2+3–PPi–BTAC
Unit Cell
space group	P 21 21 21	P 21	P 21	P 21	P 21	P 21
a, b, c (Å)	46.47, 76.16, 107.67	51.68, 46.98, 75.54	51.55, 46.84, 75.17	51.59, 46.81, 75.25	52.88, 46.99, 75.53	53.26, 47.24, 75.15
α, β, γ (deg)	90, 90, 90	90, 98.14, 90	90, 98.07, 90	90, 97.90, 90	90, 95.86, 90	90, 95.39, 90
Data Collection
laboratory, beamline	APS, 24-ID-E	APS, 24-ID-E	APS, 24-ID-E	APS, 24-ID-C	APS, 24-ID-C	NSLS-II, 17-ID-2
detector	Dectris EIGER 16M	Dectris EIGER 16M	Dectris EIGER 16M	Dectris Pilatus 6M-F	Dectris Pilatus 6M-F	Dectris EIGER 16M
resolution (Å)	1.99	1.37	1.33	1.29	1.43	1.48
total/unique no. of reflections	175,163/27,010	248,781/74,716	259,775/79,951	283,944/87,503	225,143/67,562	209,636/61,597
Rmergea,b	0.314 (0.801)	0.038 (0.249)	0.049 (0.292)	0.053 (0.439)	0.038 (0.806)	0.055 (0.632)
Rpima,c	0.135 (0.347)	0.025 (0.164)	0.031 (0.198)	0.034 (0.290)	0.025 (0.522)	0.035 (0.397)
CC1/2a,d	0.969 (0.551)	0.999 (0.927)	0.998 (0.894)	0.998 (0.805)	0.999 (0.719)	0.999 (0.720)
I/σ(I)a	13.8 (3.3)	19.3 (4.2)	14.1 (3.0)	12.8 (2.4)	15.7 (1.5)	12.4 (2.0)
Redundancya	6.5 (6.5)	3.3 (3.2)	3.2 (3.1)	3.2 (3.1)	3.3 (3.3)	3.4 (3.2)
completeness (%)a	99.8 (99.4)	98.9 (95.9)	98.0 (98.3)	97.7 (95.7)	98.9 (98.5)	99.5 (93.7)
Refinement
reflections used in refinement/test set	26,941/1,318	74,694/3,647	79,919/7,976	87,471/4,278	67,488/3,358	61,567/2,000
R work a,e	0.184 (0.216)	0.159 (0.180)	0.160 (0.186)	0.150 (0.232)	0.171 (0.350)	0.168 (0.270)
R free a,e	0.212 (0.236)	0.170 (0.200)	0.172 (0.201)	0.161 (0.244)	0.189 (0.353)	0.179 (0.296)
no. of protein chains	1	1	1	1	1	1
no. of nonhydrogen atoms	2,576	2,998	2,991	3,098	3,019	3,031
protein	2,410	2,709	2,713	2,756	2,765	2,707
ligand	5	21	31	43	49	37
solvent	161	268	247	299	205	287
average B factor (Å2)	26	17	17	17	31	21
protein	25	16	16	16	30	20
ligand	26	13	15	19	32	22
solvent	34	24	25	27	38	28
root-mean-square deviation from ideal geometry
bonds (Å)	0.010	0.010	0.008	0.009	0.009	0.012
angles (deg)	0.95	1.2	1.0	1.1	1.0	1.2
Ramachandran plotf
Favored (%)	99.67	99.11	99.11	99.41	99.41	99.41
Allowed (%)	0.33	0.89	0.89	0.59	0.59	0.59
Outliers (%)	0.00	0.00	0.00	0.00	0.00	0.00
Molprobity scoref	0.76	0.78	0.64	0.73	0.85	0.92
PDB entry	8SU0	8SU1	8SU2	8SU3	8SU4	8SU5

^aValues in parentheses refer to the highest-resolution shell of data.

^bR_merge = ∑_h∑_i|I_h,_i − ⟨I⟩_h|/∑_h∑_iI_h,i, where ⟨I⟩_h is the average intensity calculated for reflection h from i replicate measurements.

^cR_p.i.m. = (∑_h(1/(N-1))^1/2∑_i|I_h,i −⟨I⟩_h|)/∑_h∑_i I_h,_i, where N is the number of reflections and ⟨I⟩_h is the average intensity calculated for reflection h from replicate measurements.

^dPearson correlation coefficient between random half-datasets.

^eR_work = ∑||F_o| − |F_c||/∑|F_o| for reflections contained in the working set. |F_o| and |F_c| are the observed and calculated structure factor amplitudes, respectively. R_free is calculated using the same expression for reflections contained in the test set held aside during refinement.

^fCalculated with MolProbity.

Enzyme Kinetics.

The EnzChek Pyrophosphate Detection Kit (Invitrogen) was used to evaluate enzyme kinetics through the detection and quantification of inorganic pyrophosphate, the co-product of the cyclization reaction. For specific activity measurements, 100 μL reactions were conducted in assay buffer [50 mM Tris-HCl (pH 7.5), 1.0 mM MgCl₂, 0.1 mM NaN₃]. Added to each reaction was 200 μM 2-amino-6-mercapto-7-methylpurine ribonucleoside (MESG), 1 unit purine nucleoside phosphorylase (PNPase), 0.03 unit inorganic pyrophosphatase (IPPase), and 0.1–0.5 μM EIZS. The assay was performed in a 96-well clear flat-bottom plate. Reactions were initiated upon the addition of FPP to a final concentration of 100 μM and monitored at A₃₆₀ on a Tecan M1000 spectrophotometer. Reaction mixtures with no enzyme were included as a negative control, and each enzyme was evaluated in triplicate. Initial velocities were calculated using the ΔA₃₆₀ within the linear range.

For steady-state kinetics measurements, 100 μL reactions were conducted in assay buffer containing 200 μM MESG, 1 unit PNPase, 0.03 unit IPPase, and 500 nM F95S EIZS. A mixture containing all assay components, except FPP, was aliquoted into individual wells in a 96-well clear flat-bottomed plate and incubated at room temperature for 15 minutes. Reactions were initiated upon the addition of FPP to final concentrations ranging from 50 to 2000 μM and monitored on a Tecan M1000 spectrophotometer in triplicate. Reaction rates were determined as the increase of A₃₆₀ per second. Steady-state kinetic parameters were determined by fitting reactions rates to a nonlinear regression in Prism 9.

GC/MS Product Analysis.

Products of cyclase activity were analyzed by gas chromatography-mass spectrometry (GC/MS). 6 mL reactions were performed in assay buffer [50 mM Tris-HCl (pH 7.5), 1.0 mM MgCl₂, 0.1 mM NaN₃], with 2 μM enzyme. Reactions were initiated upon the addition of 200 μM FPP, and immediately overlaid with 2 mL of a 1:1 mixture of hexanes:ethyl acetate to collect volatile organic compounds. Reactions were incubated overnight at room temperature. After the addition of 1 mL of brine, reactions were extracted three times with a total of 6 mL organic solvent. Extracts were filtered, dried over a column of anhydrous Na₂SO₄, and concentrated to ~300 μL under a stream of N₂ gas. Samples were analyzed on an Agilent 7890A/5976C GC/MS system in EI positive mode using a temperature program of 60–240 °C with a gradient of 20 °C/min and a solvent delay of 3 min. Terpene products were identified by comparison of individual chromatographic retention indices and mass spectra with those of compounds in the MassFinder 4.0 Database.

Results

Catalytic activity and product diversity of EIZS mutants.

All EIZS mutants were assayed to confirm catalytic activity and to identify alternative cyclization products. The F96H substitution has the greatest impact on enzyme function, decreasing specific activity by ~10-fold. The sensitivity of EIZS activity to mutation of F96 likely arises from the role of this residue in the structural change between open and closed active site conformations; only the closed conformation is catalytically active, and F96 rotates so as to pack in an aromatic cluster that stabilizes the closed conformation.^36,38 The remaining polar mutants – F95S, F198S, and F198T – exhibit more modest 2–4-fold reductions in catalytic activity (Figure 2A). This observation is consistent with previous characterizations of EIZS mutants.^35,36,39

Figure 2. Catalytic activity.

(A) Specific activities of wild-type and mutant EIZS enzymes. Reactions were monitored for PP_i release using a coupled assay and rates were determined within the linear range. Error bars represent the standard deviation of four independent measurements. Numerical data are recorded in Table S4. (B) Steady-state kinetics of F95S EIZS, which generates a mixture of bisabolane precursors. As in (A), reaction rates were measured by quantifying the release of co-product PP_i. Numerical data are recorded in Table S5.

To identify the sesquiterpene products generated by each mutant, we analyzed organic extracts of each reaction by GC/MS (Table 2). The biosynthetic manifold of EIZS mutants in the current study is summarized in Figure 3. Mutations of residues defining the active site contour virtually abolish epi-isozizaene generation and divert the reaction sequence from specific carbocation intermediates to form alternative sesquiterpene products. Apart from the generation of nerolidol by F96H EIZS, only hydrocarbon products are generated by the EIZS mutants studied.

Table 2.

Sesquiterpene products generated by EIZS mutants.^a

product	RIb	wild-typec	F95S	F96Hd	F198S	F198T
α-cedrene	1418	2
β-cedrene	1424				58	49
sesquisabinene A	1435	2
epi-isozizaene	1444	79	2
(E)-β-farnesene	1446	5		18
zizaene	1456	9			20
β-acoradiene	1465					42
γ-curcumene	1475		38
β-curcumene	1503		47		2
β-bisabolene	1503			1
(Ζ)-γ-bisabolene	1505		13		9	2
β-sesquiphellandrene	1516	1				4
(Ε)-nerolidol	1553			73
unknown				8	11	3

^aProducts are reported as a percentage of total sesquiterpene products.

^bRetention indices as reported in the MassFinder 4.0 database.

^cRef. 35.

^dRef. 39.

Figure 3. Biosynthetic manifold of wild-type and mutant EIZS.

The wild-type enzyme catalyzes the cyclization of FPP (yellow box) to yield epi-isozizaene (green box). Active site mutations can divert the reaction at specific carbocation intermediates to yield myriad alternative products (blue boxes). Occasionally, a trapped solvent molecule can rapidly quench a carbocation intermediate to yield a hydroxylated product (orange box).

It is notable that 98% of the sesquiterpenes generated by F95S EIZS are monocyclic isomers, indicating that the cyclization cascade is diverted at monocyclic carbocation intermediates: β-bisabolene is formed by deprotonation of the bisabolyl cation, and β- and γ-curcumene are formed by deprotonation of the homobisabolyl cation. Furthermore, these monocyclic isomers can be hydrogenated to form the same saturated hydrocarbon product, bisabolane, which is a validated substitute for D2 diesel fuel.²⁰ Given the uniquely relevant product array generated by F95S EIZS, we measured steady-state kinetic parameters for this mutant (Figure 2B). With k_cat/K_M = 1700 M⁻¹s⁻¹, the catalytic efficiency of this mutant is 45-fold higher compared with the plant enzyme α-bisabolene synthase originally considered in a synthetic biology approach for bisabolane generation.²⁰

The F96H mutation, as reported previously³⁹, converts EIZS into a high-fidelity nerolidol synthase (73% yield) that also generates 18% (E)-β-farnesene plus trace amounts of other products. The F96H substitution predominantly diverts the reaction sequence from the first carbocation intermediates formed to generate mainly acyclic products: F96H EIZS enables the hydroxylation of the nerolidyl carbocation intermediate or proton elimination from the initially formed farnesyl carbocation intermediate. Hydroxylation requires a second substrate, water, that must be properly positioned to quench the nerolidyl carbocation. It is notable that engineering the contour of a terpene cyclase active site not only reprograms the cyclization cascade, but can also introduce new catalytic capabilities – here, generating a sesquiterpene alcohol instead of a sesquiterpene hydrocarbon.

The substitution of polar serine or threonine residues for F198 diverts the cyclization cascade from a bicyclic carbocation intermediate. Both the F198S and F198T mutations convert EIZS into β-cedrene synthases exhibiting 58% and 49% cyclization fidelities, respectively. This reflects diversion of the cyclization cascade after formation of the acorenyl carbocation. Other notable products generated by these mutants include 20% zizaene and 9% (Z)-γ-bisabolene by F198S EIZS, and 42% β-acoradiene by F198T EIZS.

Structure of unliganded F96H EIZS.

This mutant crystallizes in an unliganded state with an open active site conformation despite incubation with millimolar concentrations of Mg²⁺, PP_i, and BTAC. The 1.99 Å resolution crystal structure reveals significant disorder for polypeptide segments P56-Y69 and S336-N355, so these segments were not included in the final model. These segments comprise helices B and K, respectively, the movement and ordering of which are the primary induced-fit conformational changes that occur upon ligand binding and active site closure.^35,36,38

Superposition with the structure of unliganded wild-type EIZS (PDB 4LTV)³⁶ shows an overall similar architecture, with a root-mean-square deviation (RMSD) of 0.185 Å over 240 Cα atoms. Of note, active site aromatic residues F95, F96/H96, F198, and W203 adopt similar positions and conformations in the unliganded wild-type and mutant enzymes (Figures 4A, 4B). However, considerable structural differences are observed for helix B. In the unliganded wild-type enzyme, Y63 and Y69 on helix B form an aromatic cluster with F96 and F332. The F96H substitution destabilizes this cluster, leading to disorder of residues P56-C68 and the side chain of Y69, and thence the disorder of helix B.

Figure 4. Crystal structures of F96H EIZS.

(A) Polder omit map of H96 in F96H EIZS (contoured at 4σ). confirms the position and orientation of H96 in the mutant enzyme. (B) Overlay of unliganded wild-type EIZS and unliganded F96H EIZS. The F96H substitution results in the disorder of helix B (cyan ribbon) due to the disruption of packing interactions in the Y63-Y69-F96-F332 aromatic cluster (black dashed lines). (C) Comparison of key active site residues in F96H EIZS (magenta) and wild-type EIZS (cyan) as they accommodate the binding of pamidronate and three Mg²⁺ ions. Polder omit maps of H96 (light pink), the Mg²⁺₃ cluster (green), and pamidronate (light brown) are contoured at 4σ. (D) Overlay of pamidronate and active site water molecules in pamidronate-bound wild-type EIZS (cyan sticks, blue spheres, light blue dashed lines) and pamidronate-bound F96H EIZS (brown sticks, red spheres, red dashed lines). Water #129 forms a 2.8 Å hydrogen bond with H96.

Structure of the F96H EIZS–Mg²⁺₃–pamidronate complex.

Overall, the 1.37 Å-resolution structure of the F96H EIZS–Mg²⁺₃–pamidronate complex reveals a closed active site conformation and exhibits no major conformational changes relative to the wild-type EIZS–Mg²⁺₃–pamidronate complex (PDB 7KJ8),³⁸ with an RMSD of 0.137 Å over 315 Cα atoms. Notably, helix B becomes ordered upon ligand binding. The bisphosphonate moiety of pamidronate engages in metal coordination and hydrogen bond interactions identical to those observed in the complex with the wild-type enzyme (Figure 4C). The F96H substitution triggers subtle conformational changes in the aromatic cluster in the active site. Relative to F96 in the wild-type enzyme complex, the imidazole ring of H96 rotates approximately 20°. Even so, the H96 side chain is still positioned for an offset π-stacking interaction with the aromatic ring of Y69 on helix B. Additionally, the F95 side chain rotates 60° toward the face of the H96 residue, where it engages in an edge-to-face aromatic interaction, instead of the offset π-stacking interaction observed in the wild-type enzyme.

The most notable difference between the pamidronate-bound structures of F96H EIZS and wild-type EIZS is the network of water molecules trapped in the enclosed active site (Figure 4D). In the wild-type enzyme, three water molecules hydrogen bond with each other and the hydroxyl and terminal amino group of the bisphosphate inhibitor. As seen in the F96H EIZS complex, water molecules still form hydrogen bonds with the bisphosphonate inhibitor, but the introduction of the smaller, polar imidazole group of H96 allows for the solvent network to expand. In particular, water #129 forms a hydrogen bond with the newly introduced imidazole group of H96. Located in the upper active site cavity, this water molecule may be properly positioned to add to the nerolidyl carbocation to general nerolidol. Moreover, H96 could serve as a general base to activate this water molecule for catalysis.

F96H EIZS–Mg²⁺₃–risedronate complex.

Similar to the structure of pamidronate-bound F96H EIZS, the 1.33 Å-resolution structure of the F96H EIZS–Mg²⁺₃–risedronate complex shows no major differences in comparison with the corresponding wild-type EIZS complex (PDB 7KJ9),³⁸ with an RMSD of 0.217 Å over 315 Cα atoms. Additionally, the F96H EIZS–Mg²⁺₃–pamedronate and F96H EIZS–Mg²⁺₃–risedronate complexes are very similar, with an RMSD of only 0.089 Å over 309 Cα atoms between the two structures. The trinuclear magnesium cluster and bisphosphonate moiety of risedronate bind in the same fashion as in the wild-type enzyme complex, and the pyridine ring of the inhibitor adopts the same orientation (Figure 5A).

Figure 5. Crystal structure of the F96H EIZS–Mg²⁺₃–risedronate complex.

(A) Comparison of key active site residues in F96H EIZS (magenta) and wild-type EIZS (cyan) as they accommodate the binding of risedronate and three Mg²⁺ ions. Polder omit maps of H96 (light pink), the Mg²⁺₃ cluster (green), and risedronate (light brown) are contoured at 4σ. (B) Overlay of risedronate and active site water molecules in risedronate-bound wild-type EIZS (cyan sticks, blue spheres, light blue dashed lines) and risedronate-bound F96H EIZS (brown sticks, red spheres, red dashed lines). Water #247 forms a 2.8 Å hydrogen bond with H96.

As similarly noted in the comparison of the pamidronate-bound structures of F96H EIZS and wild-type EIZS, the most notable difference between the risedronate-bound structures of F96H EIZS and wild-type EIZS is that the F96H substitution allows for an expansion of the active site solvent network. While in both risedronate-bound structures there are two hydrogen bonded water molecules to the hydroxyl and pyridine ring of the bisphosphonate inhibitor, the smaller, polar imidazole group of F96H EIZS allows for the accommodation of two additional water molecules. Water #247 forms a hydrogen bond with the imidazole group of H96 (Figure 5B), similar to that noted for water #129 in the F96H EIZS–Mg²⁺₃–pamidronate complex (Figure 4D).

F95S EIZS–Mg²⁺₃–PP_i–BTAC Complex.

The 1.29 Å-resolution structure of the F95S EIZS–Mg²⁺₃–PP_i–BTAC complex reveals a closed active site conformation stabilized by a Mg²⁺₃-PP_i cluster. Intermolecular interactions are unchanged relative to those observed in the wild-type enzyme complex with Mg²⁺₃-PP_i (PDB 3KB9).³⁵ Overall, the BTAC-bound F95S EIZS structure exhibits an RMSD of 0.168 Å over 294 Cα atoms with the BTAC-bound wild-type enzyme structure. Electron density for S95 shows that the side chain adopts three conformations with varying occupancies, with the predominant conformation (62%) oriented toward the interior of the active site, positioning the hydroxyl oxygen 4.2 Å from the quaternary amino nitrogen of BTAC (Figure 6). The positive charge on BTAC appears to be stabilized by a cation-π interaction with the side chain of F96. The polder omit map of BTAC shows well-defined density for the benzyl ring, which is sandwiched between the side chains of F96 and F198, with less well-defined density for the ethyl substituent that is most proximal to S95.

Figure 6. Crystal structure of the F95S EIZS–Mg²⁺₃–PP_i–BTAC complex.

(A) Active site of BTAC-bound F95S EIZS. Polder omit maps of S95 (light pink), the Mg²⁺₃ cluster (green), PP_i (light brown), and BTAC (blue) are contoured at 4σ. (B) The active site cleft of F95S EIZS is shown as a light blue surface. S95 is shown along with an overlay of F95 as it appears in the wild-type enzyme. The F95S substitution substantially enlarges the enclosed active site cavity.

F198S EIZS–Mg²⁺₃–PP_i–BTAC Complex.

The 1.43-Å resolution structure of F198S EIZS complexed with Mg²⁺₃-PP_i and BTAC is very similar to that of the corresponding complex with the wild-type enzyme (PDB 3KB9),³⁵ as indicated by an RMSD of 0.125 Å across 307 Cα atoms. Conformations of active site residues are largely unchanged in the mutant enzyme. In contrast with F95S EIZS, where the side chain of the serine mutation adopts multiple conformations, S198 in F198S EIZS adopts a single conformation (Figure 7A). The side chain hydroxyl group of S198 forms a hydrogen bond with one of the water molecules trapped in the enclosed active site. F198S EIZS does not generate hydroxylated products (Table 2), so none of the trapped water molecules observed in the crystal structure are properly positioned to react with carbocation intermediates in catalysis if present when substrate is bound.

Figure 7. Crystal structures of F198S EIZS and F198T EIZS complexed with Mg²⁺₃–PP_i and BTAC.

Polder omit maps for all ligands are contoured at 4σ; atomic color-codes are as follows: Mg²⁺, green; PP_i, light brown; BTAC, blue; F198S mutation, light pink; F198T mutation, yellow; water molecules appear as small red spheres. (A) F198S EIZS (residues, magenta; BTAC, blue). (B) Overlay of F198S EIZS (residues, magenta; BTAC, blue) and wild-type EIZS (residues and BTAC, cyan). (C) F198T EIZS (residues, yellow; BTAC, pale green). (D) Overlay of F198S EIZS (residues, magenta; BTAC, blue) and F198T EIZS (residues, yellow; BTAC, pale green).

There are several features of note regarding the binding mode of BTAC. Just as in the active site of the wild-type enzyme, the positively charged quaternary ammonium group of BTAC is positioned between the aromatic side chains of F95 and F96 and stabilized by cation-π interactions. However, the benzyl ring of BTAC undergoes a ~90° rotation upon binding to F198S EIZS in comparison with binding to the wild-type enzyme (Figure 7B). This conformational difference in BTAC binding is sterically enabled by the smaller S198 side chain.

F198T EIZS–Mg²⁺₃–PP_i–BTAC Complex.

The 1.48 Å-resolution structure of F198T EIZS complexed with Mg²⁺₃-PP_i and BTAC shows minimal overall structural changes relative to the corresponding wild-type complex (PDB 3KB9),³⁵ with an RMSD of 0.108 Å over 309 Cα atoms. Indeed, the structure of the F198T EIZS complex (Figure 7C) is quite similar to that of the F198S complex, exhibiting an RMSD of 0.120 Å over 322 Cα atoms. Curiously, however, no ordered water molecules are observed in the active site of the F198T EIZS complex even though the structures are determined at comparable resolutions. As observed upon binding to F198S EIZS, the benzyl ring of BTAC is rotated by ~90° upon binding to F198T EIZS due to the loss of the bulky F198 side chain (Figure 7D).

Discussion

The first X-ray crystal structures of terpene cyclases, reported 26 years ago,^47–49 established a critical framework for understanding the exquisite mechanistic complexity of isoprenoid cyclization reactions, and this framework continues to expand with each new structure reported. The active site contour of a terpene cyclase typically exerts remarkable control over the reaction sequence to chaperone each chemical step to yield a complex cyclic product with structural and stereochemical precision. Crystal structures of terpene cyclases reveal that the active site contour is defined mainly by aromatic and aliphatic residues. Aromatic residues such as phenylalanine are particularly important because they are large, contributing substantial surface area to the active site contour, and because they are capable of stabilizing carbocation intermediates in catalysis through cation-π interactions.

Active site contours are generally product-like in shape, each serving as a template for binding the flexible substrate and chaperoning a cyclization cascade toward a product complementary in shape. For example, the active site contour of wild-type EIZS is complementary to the molecular shape of epi-isozizaene (Figure 8A). It stands to reason that remolding the active site contour through mutagenesis will divert the cyclization cascade, effectively reprogramming the enzyme to generate alternative products. In our previous studies of EIZS, we reported single-point mutants that diminish or abolish the generation of epi-isozizaene while establishing new pathways for the generation of new monocyclic, bicyclic, and tricyclic products. For example, F95H EIZS generates 50% β-curcumene and 44% epi-isozizaene,³⁶ and F96Q EIZS generates 97% sesquisabinene A and 3% β-bisabolene, but no epi-isozizaene.³⁹

Figure 8. Active site contours of wild-type and mutant EIZS enzymes with major cyclization product modeled.

In all panels, pyrophosphate is represented as red and orange sticks, the trinuclear magnesium cluster is represented by green spheres, and active site contours are shown as grey surfaces. Residues which define the contour of the active site are shown in cyan (wild-type, panel A) or magenta (mutants, panels B-E) sticks. Panels B, C, D, and E are mutant EIZS enzymes reported in this manuscript. In panel C, pyrophosphate is modeled into the F96H-risedronate complex and, for clarity, risedronate is not shown. Active site volumes are indicated at the bottom of each plate. (A) wild-type EIZS (PDB 3KB9) with epi-isozizaene (green sticks). (B) F95S EIZS with β-curcumene (light green sticks). (C) F96H EIZS with (E)-nerolidol (turquoise sticks). (D) F198S EIZS with β-cedrene (yellow sticks). (E) F198T EIZS with β-cedrene (yellow sticks). (F) F198T EIZS with β-acoradiene (tan sticks).

Here, we describe the structure and catalytic function of four new EIZS mutants in which polar residues are substituted for three phenylalanine residues that largely define the active site contour. These substitutions challenge conventional wisdom, because the introduction of a polar residue in an active site that has evolved to accommodate reactive carbocation intermediates risks alkylation of the polar residue and deactivation of the enzyme catalyst. However, no evidence for deactivation of any of these mutants is observed; indeed, substitution of active site phenylalanine residues with histidine, serine, or threonine not only diverts the cyclization reaction, but in one case endows the enzyme with a new catalytic capability, the generation of a sesquiterpene alcohol. In comparing the active site of the wild-type enzyme (Figure 8A) with the active sites of the characterized mutants (Figures 8B–8F), it is evident how each mutation appears to remold the active site contour so as to resemble the molecular structure of its major product.

Of particular interest is F95S EIZS, which generates a mixture of regioisomers, 47% β-curcumene, 38% γ-curcumene, and 12% (Z)-γ-bisabolene; only 2% epi-isozizaene is observed (Table 2, Figure 3). The substitution of the smaller serine side chain for the bulky aromatic ring of phenylalanine elongates the active site and enlarges its volume by ~40 ų (Figure 8B). When modeling β-curcumene into this active site contour, it becomes immediately apparent that this remolded template is nearly incapable of holding the prenyl “tail” of the homobisabolyl cation in place for the second cyclization reaction leading to the acorenyl cation enroute to epi-isozizaene formation (Figure 3). Consequently, 98% of the detected products are formed by quenching the bisabolyl or homobisabolyl carbocations.

The reprogrammed catalytic activity of F95S EIZS has implications for energy science, since β-curcumene, γ-curcumene, and (Z)-γ-bisabolene can be hydrogenated to yield bisabolane, a D2 diesel fuel substitute.²⁰ Previously, we reported that F95H EIZS generates 50% β-curcumene and 44% epi-isozizaene, and further proposed that the bacterial enzyme mutant could be used in place of the sluggish, naturally-occurring plant α-bisabolene synthase²⁰ in a process to generate bisabolane more efficiently.³⁶ Not only does F95 appear to be a “hotspot” for diverting the cyclization cascade at the bisabolyl or homobisabolyl carbocations with appropriate amino acid substitutions, but F95S EIZS represents a significant improvement over F95H in the generation of bisabolane precursors – 98% of the product mixture can be hydrogenated to yield bisabolane (Figure 9). Moreover, with k_cat/K_M = 1700 M⁻¹s⁻¹, F95S EIZS generates these bisabolane precursors with 45-fold greater catalytic efficiency than the plant α-bisabolene synthase (k_cat/K_M = 38 M⁻¹s⁻¹) originally considered for this process.²⁰ We previously reported that F95H EIZS exhibits higher catalytic efficiency with k_cat/K_M = 2600 M⁻¹s⁻¹, but this mutant generates only 50% α-bisabolene.³⁶ Thus, F95S EIZS exhibits higher catalytic efficiency for the generation of bisabolane precursors compared with F95H EIZS or plant α-bisabolene synthase. We envisage that F95S EIZS could accordingly serve as a module for sustainable in vivo biosynthesis in advanced biofuel production.^{21–23,50–52} More generally, we envisage that as a robust bacterial enzyme, EIZS could be engineered to generate other high-value sesquiterpenes and thereby serve as a useful plug-and-play component in various synthetic biology systems.

Figure 9.

F95S EIZS is more efficient at carbon management than F95H EIZS in the generation of bisabolane precursors. Even though F95H EIZS exhibits higher catalytic efficiency than F96S EIZS in the generation of bisabolane precursor, it retains substantial wild-type activity and generates 44% epi-isozizaene, which cannot be converted to bisabolane. Both of these bacterial enzyme mutants exhibit higher catalytic efficiency than plant α-bisabolene synthase.

Remolding the active site contour of EIZS with aromatic-polar substitutions not only diverts the cyclization cascade from specific carbocation intermediates, but it can also introduce new catalytic activities through these diversions, e.g., hydroxylation. We previously reported that F96H EIZS generates nerolidol as the major product (73%).³⁹ This is the only F96 mutant studied to date that yields such a high percentage of nerolidol; the F96A, F96V, F96L, F96Y, F96W, F96S, F96T, F96N, F96M, and F96Q mutants generate only 0–5% nerolidol.^36,39 This suggests that substitution of the polar histidine side chain for the aromatic side chain of F96 introduces functionality that can activate a trapped water molecule for catalysis (Figure 3).

The X-ray crystal structures of F96H EIZS in the unliganded and inhibitor-bound conformations suggest a possible explanation for water activation in catalysis, in that the newly introduced H96 side chain could serve as a general base to activate its hydrogen bonded water molecule for reaction with the nerolidyl carbocation. This water molecule appears to be properly positioned for reaction, based on a model of nerolidol bound in the active site of F96H EIZS in which the hydroxyl group of nerolidol nearly overlaps with the position of the water molecule hydrogen bonded to H96 (Figure 8C). While there are other water molecules trapped in the active site of F96H EIZS-inhibitor complexes, the water molecule hydrogen bonded with H96 appears to be best oriented for reaction with the nerolidyl carbocation.

Finally, it is striking that F198S EIZS and F198T EIZS favor β-cedrene formation, with side products that similarly reflect successful closure of the second ring in the formation of acoradiene (Table 2, Figure 3). Other amino acid substitutions for F198 likewise convert EIZS into β-cedrene synthases. Interestingly, F198V EIZS and F198L EIZS generate β-cedrene as their major product; F198A EIZS generates a broad product array, and F198Y EIZS generates epi-isozizaene with wild-type fidelity.^35,36 Thus, F198 appears to be a hotspot for diverting the cyclization cascade at the acorenyl carbocation. Inspection of the F198S EIZS and F198T EIZS active site contours hints at how this chemistry is accomplished. By remolding the F198 region of the active site contour, the active site contour of the mutant enzyme is still sufficiently constricted in the right place so as to hold the pendant prenyl tail of the homobisabolyl carbocation in place for acorenyl cation formation (Figure 8D, 8E, 8F). However, it cannot preferentially direct the formation of a single acorenyl stereoisomer or chemical steps thereafter.

Conclusions

The present work advances our structure-function studies of a library of nearly 50 single-point mutants of EIZS such that we can make correlations between key aromatic residues and their influence on specific steps of the cyclization cascade. Amino acid substitutions for F96 can divert the initially formed farnesyl or nerolidyl carbocations prior to the first ring closure reaction; substitutions for F95 can divert the monocyclic bisabolyl or homobisabolyl carbocations prior to the second ring closure reaction; and substitutions for F198 can divert the bicyclic acorenyl cation prior to the third ring closure reaction. Moreover, the substitution of a histidine residue at an appropriate position in the active site introduces new activity through the generation of a sesquiterpene alcohol; the newly introduced histidine could serve as a general base in activating water for this chemistry. These studies illustrate the potential of EIZS as a robust platform for synthetic biology, in that select EIZS mutants could serve as plug-and-play components in metabolic engineering approaches for the sustainable generation of high-value sesquiterpene products. For example, the efficient generation of bisabolane precursors by F95S EIZS could be coupled with chemical or enzymatic reduction to generate the D2 diesel fuel substitute bisabolane. Future work in this direction will be reported in due course.