Structure of the Prenyltransferase in Bifunctional Copalyl Diphosphate Synthase from Penicillium fellutanum Reveals an Open Hexamer Conformation

Abstract

Copalyl diphosphate synthase from Penicillium fellutanum (PfCPS) is an assembly-line terpene synthase that contains both prenyltransferase and class II cyclase activities. The prenyltransferase catalyzes processive chain elongation reactions using dimethylallyl diphosphate and three equivalents of isopentenyl diphosphate to yield geranylgeranyl diphosphate, which is then utilized as a substrate by the class II cyclase domain to generate copalyl diphosphate. Here, we report the 2.81 Å-resolution cryo-EM structure of the hexameric prenyltransferase of full-length PfCPS, which is surrounded by randomly splayed-out class II cyclase domains connected by disordered polypeptide linkers. The hexamer can be described as a trimer of dimers; surprisingly, one of the three dimer-dimer interfaces is separated to yield an open hexamer conformation, thus breaking the D3 symmetry typically observed in crystal structures of other prenyltransferase hexamers such as wild-type human GGPP synthase (hGGPPS). Interestingly, however, an open hexamer conformation was previously observed in the crystal structure of D188Y hGGPPS, apparently facilitated by hexamer-hexamer packing in the crystal lattice. The cryo-EM structure of the PfCPS prenyltransferase hexamer is the first to reveal that an open conformation can be achieved even in the absence of a point mutation or interaction with another hexamer. Even though PfCPS octamers are not detected, we suggest that the open hexamer conformation represents an intermediate in the hexamer-octamer equilibrium for those prenyltransferases that do exhibit oligomeric heterogeneity.

Graphical Abstract

INTRODUCTION

Prenyltransferases are crucial biosynthetic enzymes found in all branches of life. In primary metabolism, prenyltransferases catalyze the head-to-tail coupling of the C₅ isoprenoid dimethylallyl diphosphate (DMAPP) and one or more molecules of C₅ isopentenyl diphosphate (IPP). Catalysis is initiated by metal-triggered ionization of DMAPP to yield an allyl carbocation that reacts with IPP, and the resulting tertiary carbocation intermediate is quenched by proton elimination to yield C₁₀ geranyl diphosphate (GPP) (Figure 1) (Davisson et al., 1993; Poulter, 2006). Additional rounds of IPP substitution and proton elimination yield increasingly longer isoprenoids such as C₁₅ farnesyl diphosphate (FPP) and C₂₀ geranylgeranyl diphosphate (GGPP) (Poulter & Rilling, 1978; Kellogg & Poulter, 1997). The length of the final chain-elongation product is determined by the depth of the active site pocket (Tarshis et al., 1996; Chang et al., 2006; Ronnebaum et al., 2021).

Figure 1. Prenyltransferase reaction mechanism.

The chain elongation reaction catalyzed by a prenyltransferase is initiated by the metal-triggered ionization of dimethylallyl diphosphate (DMAPP) to yield an allyl carbocation that alkylates the C=C bond of isopentenyl diphosphate (IPP) (OPP = diphosphate). The resulting tertiary carbocation is quenched by proton elimination to yield the chain elongation product, geranyl diphosphate (GPP). Prenyltransferases are processive enzymes and this reaction sequence will cycle with successive addition of IPP to generate increasingly longer isoprenoid diphosphates such as farnesyl diphosphate (FPP) and geranylgeranyl diphosphate (GGPP). Final product length is determined by the depth of the active site pocket.

In lower organisms such as bacteria, fungi, and plants, as well as simple invertebrate marine animals such as corals, GPP, FPP, and GGPP are utilized as substrates for terpene cyclases to generate aliphatic hydrocarbons, most often with intricate multi-ring structures (Christianson, 2006; Christianson, 2017; Burkhardt et al., 2022; Scesa et al., 2022). In humans, FPP and GGPP are employed in posttranslational modifications to regulate protein interactions with cell membranes, e.g., as found for nuclear lamins A and B, rhodopsin kinase, and the oncogenic GTPase Ras (Zhang & Casey, 1996). Moreover, enzymes involved in protein prenylation can serve as drug targets for the treatment of certain diseases, such as cancer and osteoporosis (Cox et al., 2014; Wang & Casey, 2016; Palsuledesai & Distefano, 2015). Thus, advances in our understanding of prenyltransferase structure-function relationships provide valuable context for current explorations in drug design.

The first prenyltransferase to yield a three-dimensional structure was dimeric avian FPP synthase (Tarshis et al., 1994). FPP synthase adopts an α fold, designated the terpenoid synthase fold (Lesburg et al., 1997; Sacchettini & Poulter, 1997), thought to have evolved from gene duplication and fusion of a primordial 4-helix bundle (Huang et al., 2014). This α fold is shared with class I terpene cyclases (Lesburg et al., 1997; Starks et al., 1997), which similarly initiate catalysis through metal-triggered ionization of an isoprenoid diphosphate substrate. All prenyltransferases contain two signature metal-binding motifs, DDXXD, that coordinate to the catalytically required Mg²⁺₃ cluster (Aaron & Christianson 2010).

X-ray crystal structures reveal that avian FPP synthase and yeast GGPP synthase form dimers with similar quaternary structures (Tarshis et al., 1994; Chang et al., 2006), but higher order oligomers are observed for prenyltransferases from other species. For example, human GGPP synthase, the GGPP synthase domain of fusicoccadiene synthase from Phomopsis amygdali (PaFS), and the GGPP synthase domain of copalyl diphosphate synthase from Penicillium verruculosum (PvCPS) crystallize as hexamers with D3 symmetry that can be described as trimers of dimers (Kavanagh et al., 2006; Chen et al., 2016; Ronnebaum et al., 2020). The cryo-EM structure of bifunctional macrophomene synthase similarly reveals a hexamer that can be described as a trimer of dimers (Tao et al., 2022). Intriguingly, however, both hexamers and octamers of human GGPP synthase (hGGPPS) are detected in solution (Kuzuguchi et al., 1999), and cryo-EM analysis reveals hexamers and octamers of the GGPP synthase domain of PaFS (Faylo et al., 2021a). Thus, the prenyltransferase dimer can exhibit oligomeric heterogeneity, particularly as hexamers can be in equilibrium with octamers.

Curiously, the crystal structure of D188Y hGGPPS reveals an open hexamer conformation in which one of the three dimer-dimer interfaces is separated, thereby breaking the D3 symmetry observed for the wild-type hexamer (Lisnyansky et al., 2018). The structural basis of this quaternary structural change appears to be a packing interaction with another hexamer in the crystal lattice. It is interesting to speculate that the structure of D188Y hGGPPS represents an intermediate conformation that would be encountered in the hexamer-octamer transition observed for the wild-type enzyme (Kuzuguchi et al., 1999), since the hexamer must open up to allow for insertion of another dimer to form an octamer.

Here, we report the 2.81 Å-resolution cryo-EM structure of the hexameric prenyltransferase (GGPP synthase) core of copalyl diphosphate synthase from Penicillium fellutanum (PfCPS). PfCPS and PvCPS are the first bifunctional terpene synthases to be discovered with prenyltransferase and class II cyclase activities (Mitsuhashi et al., 2017) and are designated as assembly-line terpene synthases (Faylo et al., 2021b, 2022). In PfCPS, the prenyltransferase α domain is connected to the βγ domains of the class II cyclase by a 69-residue linker (Figure 2). Surprisingly, the hexameric prenyltransferase core of PfCPS adopts an open conformation such that two dimer pairs are separated, thus breaking the D3 symmetry normally observed for wild-type prenyltransferase hexamers. This open hexamer conformation is similar to that reported for D188Y hGGPPS (Lisnyansky et al., 2018). Even though the class II cyclase domains of PfCPS are randomly splayed out and could not be included in reconstructions of the prenyltransferase core, the structure of the PfCPS prenyltransferase hexamer demonstrates that an open conformation can be achieved even in the absence of a point mutation or interaction with another hexamer.

Figure 2. Reaction sequence catalyzed by PfCPS.

Reactions catalyzed by each catalytic domain are indicated with the domain colors shown in the primary structure representation. Sequence numbers for each domain are indicated; the 69-residue linker segment is represented by a red line.

MATERIALS and METHODS

Reagents.

The chemicals used in buffers and grid preparation conditions were purchased from Fisher Scientific or Millipore Sigma and used without further purification unless otherwise specified. Isoprenoids were purchased from Isoprenoids, LC.

Overexpression and purification of PfCPS.

Overexpression plasmids of PfCPS genes were supplied by GenScript. The BL21 (DE3) Escherichia coli strain containing the pfcps overexpression plasmid was grown in Terrific Broth (TB) containing 50 μg/mL kanamycin at 37°C with shaking (250 rpm). Protein expression was induced when the OD₆₀₀ reached ~0.6 by the addition of isopropyl β-D-thiogalactopyranoside (IPTG) to a final concentration of 500 μM. The temperature was reduced to 20°C for 24 h before cells were harvested by centrifugation (6,000 × g, 20 minutes, 4°C); cell pellets were stored at −80°C until purification. Upon thawing, the cell pellet (70 g) was resuspended in 100 mL of buffer A [25 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES)•NaOH (pH 7.5), 250 mM NaCl, 5 mM imidazole, 1 mM tris(2-carboxyethyl)phosphine (TCEP), and 10% glycerol] also containing 100 mg lysozyme (GoldBio), 250 units benzonase (New England BioLabs), 1 EDTA-free cOmplete Mini Protease Inhibitor tablet, and 1 mM EDTA. Cells were additionally lysed by sonication at 30 Hz for 10 min in 1 s-on/2 s-off cycles at 4°C. The resulting lysate was clarified by centrifugation (18 000 × g, 45 min, 4°C) and the resulting supernatant applied to a HisTrap^™ 5-mL column (GE Healthcare) pre-equilibrated in buffer A at 3–5 mL/min. After loading, a 5-column-volume (CV) wash with buffer A was followed by a 6-CV wash with 10% buffer B [buffer A plus 500 mM imidazole] before a 10-CV linear gradient of 10–100% buffer B was applied. PfCPS eluted along with many other proteins at approximately 200 mM imidazole. Fractions containing PfCPS were pooled, dialyzed with a 10-kDa cutoff Slide-A-Lyzer^® (Thermo Scientific) for 1 h, and then overnight in buffer C [25 mM HEPES•NaOH (pH 7.5), 1 mM TCEP, 10% glycerol]. Pooled dialysis fractions were clarified by ultracentrifugation (18 000 × g, 20 min, 4°C) before loading onto a HiTrap Q HP^™ ion exchange column pre-equilibrated in buffer C. A step gradient with buffer D [buffer C plus 500 mM NaCl] was performed. Fractions containing PfCPS, eluting at 30% buffer D, were collected and directly applied to a 26/60 Superdex 200 size-exclusion column (Amersham Biosciences) equilibrated with buffer E [25 mM HEPES•NaOH (pH 7.5), 150 mM NaCl, 1 mM TCEP, 10% glycerol]. PfCPS was eluted isocratically, fractions were pooled, and glycerol was added to a final concentration of 30%, before being concentrated with a 50-kDa cutoff Amicon centrifugal filter (Millipore) to 2 mg/mL as determined by NanoDrop One (ThermoFisher Scientific) using calculated ε_280nm = 107 260 M⁻¹ cm⁻¹ and calculated molecular weight 105 667 Da (ProtParam) (Gasteiger et al., 2005). A total of 520 μg PfCPS was obtained from the 12-L cell growth. Protein was flash-cooled and stored at −80°C until further use.

Size-exclusion chromatography.

Purified PfCPS was diluted to 1.43 mg/mL with buffer E and dialyzed overnight at 4°C against 25 mM HEPES (pH 7.5), 150 mM NaCl, and 0.1 M TCEP. The dialyzed sample was filtered and injected (100 μL) onto a Superose 6 Increase 10/300 GL Column (GE Healthcare). The PfCPS sample eluted from the column at a volume of 13.32 mL. A standard curve was established through the use of four commercial protein standards including ovalbumin, aldolase, ferritin, and thyroglobulin. Based on the standard curve, the molecular weight of the PfCPS sample was determined to be 696 kD. The theoretical molecular weight of the hexamer is 634 kD, suggesting that PfCPS eluted as a hexamer.

Cryo-EM data collection.

Prior to grid preparation for cryo-EM, PfCPS (40 μL, 2 mg/mL) was dialyzed overnight in 1 L of EM buffer [25 mM HEPES•NaOH pH 7.5, 50 mM NaCl, 1 mM TCEP]. PfCPS was filtered with a 0.22-μm PVDF centrifugal filter (Millipore) and concentration was determined as previously described. Prior to blotting, a final concentration of 5 mM MgCl₂, 500 μM IPP, and 0.025% NP-40 were added to the PfCPS sample. A 3-μL sample (0.3–0.5 mg/mL), was applied to glow-discharged (30 s, easiGlow, Pelco) R1.2/1.3 300-mesh copper grids (QUANTIFOIL^™). Grids were blotted for 6 s at 4°C with 100% humidity prior to being flash-frozen with liquid ethane using a Vitrobot Mark I (FEI). Frozen grids were clipped and transferred to a Titan Krios G3i cryogenic transmission electron microscope (Thermo Fischer Scientific) operating at 300 keV (Beckman Center for Cryo-Electron Microscopy, University of Pennsylvania). Images were recorded with a K3 Summit detector at 81,000x magnification (0.55 Å/pixel) at a nominal defocus of −0.8 to −2.5 μm, using the faster acquisition EPU software; 40 frames were taken at a dose rate of 5.95–6.05 e⁻pixel/s (43 e⁻/Ų per frame).

Cryo-EM data processing.

Data processing for PfCPS was performed using cryoSPARC (version 4.2.1) (Punjani et al., 2017), summarized by the workflow shown in Figure S1. A total of 12,467 movies were imported into cryoSPARC and aligned using the Patch Motion Correction tool. Contrast transfer function (CTF) estimation was performed using PatchCTF with the Fourier-crop factor set to 1/2. Exposures were manually curated in cryoSPARC, resulting in a final selection of 11,984 micrographs. A selection of 500 micrographs were used to train blob-based particle picking and 2D classification for downstream template-based particle picking. A blob diameter of 200–400 Å was used, which provided for an initial selection of 154,901 particles. Particles were manually inspected for quality giving a final stack of 121,834 particles. Picked particles were extracted with a box size of 324 pixels Fourier-cropped to 162 pixels. Initial reference-free 2D classification yielded three 2D classes consisting of 5,927 particles as suitable templates. Two rounds of template-based particle picking were tested on the same initial set of 500 aligned micrographs. The final set of particles picked through this approach yielded 69,077 particles after inspection, which were extracted with a box size of 288 pixels Fourier-cropped to 144 pixels. An improved set of 10 templates were obtained from 2D classification of this particle stack, and these were used for template-based picking on all 11,984 micrographs. This approach yielded a particle stack with 3,639,075 particles after manual inspection which were also extracted with a box size of 288 pixels Fourier-cropped to 144 pixels.

From the final generated particle stack from all micrographs, a set of 200 2D classes was created. Two ab-initio reconstructions were generated using particles from the best eight 2D classes. The ab-initio class representative of a hexameric prenyltransferase reconstruction was subjected to two rounds of heterogeneous refinement. The resulting particle stack from the best heterogenous refinement volume was utilized for local motion correction with an extraction box size of 576 pixels. These particles were utilized for non-uniform refinement with optimization of per-group CTF parameters as well as fitting of spherical aberration, tetrafoil, and anisotropic magnification to yield a final 3D reconstruction with C1 symmetry. Further analysis in cryoSPARC indicated C2 symmetry, which was independently confirmed using proSHADE (Nicholls et al., 2018). The C2 3D reconstruction was then generated using the local-motion corrected particle stack in non-uniform refinement. Final maps were generated with sharpening through use of deepEMhancer (Sanchez-Garcia et al., 2021).

A model of a PfCPS prenyltransferase domain monomer was generated with AlphaFold2 (Jumper et al., 2021) and six copies were fit into the sharpened C1 cryo-EM map using ChimeraX (Meng et al., 2023). The fit model and map were then transferred to PHENIX (Adams et al., 2010), which was utilized for real-space refinement with secondary structure restraints and simulated annealing. The graphics program Coot 0.9.6 (Emsley et al., 2010) was used to inspect, build, and modify the structure. Subsequent real-space refinements were iteratively performed using PHENIX. The refined C1 structure was then fit into the C2 map and subject to real-space refinement using PHENIX, and the resulting model was again inspected using Coot. The final C2 structure was validated using the cryo-EM validation tool in PHENIX, as well as MolProbity (Chen et al., 2010). All data collection, map reconstruction, and structure refinement statistics are recorded in Table 1.

Table 1.

Cryo-EM Data Collection, Reconstruction, and Refinement Statistics

	WT-PfCPS (C2)	WT-PfCPS (C1)
Magnification	81000	81000
Voltage (kV)	300	300
Exposure (e − /Å 2 )	43	43
Defocus range (μm)	−0.8 to −2.5	−0.8 to −2.5
Pixel size (Å/pix)	0.55	0.55
Symmetry imposed	C2	C1
Initial particles (no.)	3,639,075	3,639,075
Final particles (no.)	267,283	267,283
Map resolution (FSC = 0.143) (Å)	2.81	2.94
Model
Initial model used	C1 Model	AlphaFold2
Model composition (#)
Chains	6
Atoms	14703
Residues	1827
Water	0
Ligands	0
Root-mean-squared deviations
Bond lengths (Å)	0.005
Bond angles (°)	1.0
Validation
MolProbity score	1.91
Clash score	5.57
Poor rotamers (%)	3.78
Ramachandran plot (%, MolProbity)
Favored	97.0
Allowed	3.0
Outliers	0.0
Peptide plane (%)
Cis proline/general	0.0/0.0
Twisted proline/general	0.0/0.0
B-factors (Å 2 )
(min/max/mean)	24/130/59
Model vs. Data
CC (mask)	0.83
CC (box)	0.82
CC (peaks)	0.82
CC (volume)	0.84
PDB accession code	8V0F
EMDB accession code	EMD-42853	EMD-42855

Results

Full-length wild-type PfCPS was expressed and purified to approximately 95% purity based on SDS-PAGE. PfCPS eluted as a hexamer in size exclusion chromatography (Figure S2), and initial 2D class averages calculated from cryo-EM data revealed exclusively hexameric quaternary structure resulting from oligomerization of the prenyltransferase α domains (Figure 3A). The βγ domains of the class II cyclase were mostly unobserved in class averages due to being randomly splayed-out from the hexameric prenyltransferase core. However, a few 2D class averages revealed a “fog” of density adjacent to the side of the hexamer, possibly corresponding to multiple positions of associated class II cyclase domains (Figure 3B); indeed, some 2D class averages revealed discrete circular density corresponding to an associated cyclase domain viewed down the βγ domain axis (Figure 3C). These results suggest that the cyclase domain is in equilibrium between splayed-out and more closely-associated positions relative to the hexameric prenyltransferase core. Structural variability is presumably facilitated by the disordered 69-residue linker connecting the prenyltransferase and cyclase domains.

Figure 3. 2D class averages of PfCPS.

(A) Prenyltransferase domain hexamers. (B) A “fog” of density is observed adjacent to some hexamers that may correspond to multiple positions of associated cyclase domains. (C) Clear density for an associated cyclase domain (small circle corresponding to a view looking down the βγ domain axis) is observed adjacent to some hexamers at various positions. 2D classes were generated using particles picked using Topaz (Bepler et al., 2019).

Through the workflow summarized in Figure S1, we determined the structure of the prenyltransferase hexamer of PfCPS at 2.81 Å resolution. Cryo-EM density for each subunit in the hexamer is visible in 3D reconstructions for residues A621–S830 and C840–E934 (the density for E934 is absent in monomers C, D, and E) (Figure 4A). The density corresponding to residues I609–G620 is absent, indicating that the N-terminus of the prenyltransferase domain is disordered (the N-terminus connects to the disordered interdomain linker segment). Density for the C-terminal residue Q935 is also absent, similarly consistent with disorder. Additionally, the density corresponding to residues T831–L839 is not well resolved, consistent with an empty, ligand-free active site in each monomer (IPP was added to the protein solution prior to grid preparation, but bound IPP is not observed). Residues corresponding to T831–L839 are typically disordered in structures of other unliganded prenyltransferases, including PaFS, PvCPS, and macrophomene synthase (Chen et al., 2016; Ronnebaum et al., 2020; Faylo et al., 2021a, 2022; Tao et al., 2022).

Figure 4. PfCPS prenyltransferase structure.

(A) Cryo-EM density for the prenyltransferase hexamer is shown with local resolution color-coded as indicated. Particle orientation distribution is shown. The structure was solved at 2.81 Å resolution at Gold-Standard Fourier shell correlation (FSC) = 0.143. (B) Prenyltransferase dimers in PfCPS and PvCPS exhibit similar quaternary structure, with Hα-2 and Hα-3 helices capping the active sites. (C) Dimer-dimer interactions in PfCPS are mediated by the Hα-2 and Hα-3 helices, the DE loop, and the FG loop.

The PfCPS hexamer is readily described as a trimer of dimers, where each dimer exhibits the general quaternary structure first observed in the crystal structure of avian FPP synthase (Tarshis et al., 1994). More recently, similar dimer quaternary structure was observed in the related hexameric prenyltransferase of PvCPS (Ronnebaum et al., 2020) (Figure 4B). The subunit-subunit interface in each dimer is mediated through contacts between helices B, E, and F in each subunit. Dimer-dimer interactions are mediated by the Hα-2 and Hα-3 helices of one subunit and the DE loop of one subunit in an adjacent dimer along with the FG loop of the second subunit of the same adjacent dimer (Figure 4C). This interface is similar to that found in crystal structures of other hexameric prenyltransferases.

Surprisingly, although some 2D classes are observed that suggest a closed conformation for the PfCPS prenyltransferase hexamer, most 2D classes reveal an open conformation in which one dimer-dimer interface has separated (Figure 3A). Inspection of a sample raw micrograph indicates an approximate ratio of 47:3 for open and closed hexamers, respectively, that can be clearly viewed in a top-down orientation (Figure S3). Presuming that open and closed hexamers are in thermodynamic equilibrium, this ratio corresponds to a Gibbs free energy difference of only 1.6 kcal/mol. A side-by-side comparison of closed and open dimer-dimer interfaces in the cryo-EM structure of the PfCPS hexamer is shown in Figure 5. The overall three-dimensional complementarity of the dimer-dimer interface is readily apparent.

Figure 5. Comparison of closed and open dimer-dimer interfaces in PfCPS.

At left is the AB-EF dimer-dimer interface in the closed conformation; at right is the EF-CD dimer-dimer interface in the open conformation.

A side-by-side comparison of the PfCPS hexamer with the PvCPS and D188Y hGGPPS hexamers highlights overall differences in quaternary structure between the closed and open conformations (Figure 6). Notably, the open conformation of the PfCPS hexamer is similar to that of D188Y hGGPPS, where the hexamer adopts a slightly more open conformation (Lisnyansky et al., 2018). Thus, the cryo-EM structure of the PfCPS prenyltransferase is the first to show that an open hexamer conformation can be achieved for a wild-type prenyltransferase even in the absence of crystal packing interactions.

Figure 6. Prenyltransferase hexamer conformations.

Comparison of the crystal structure of the PvCPS prenyltransferase hexamer with D3 symmetry (PDB 6V0K), the cryo-EM structure of the PfCPS prenyltransferase hexamer with C2 symmetry (PDB 8V0F), and the crystal structure of D188Y hGGPPS with C2 symmetry (PDB 6G32). A 30° rotation of each hexamer around a vertical axis highlights the open conformations observed for the PfCPS and hGGPPS hexamers. Red arrows indicate symmetry axes.

It is instructive to compare the structural role of helices A and B at the N-terminus of the PfCPS prenyltransferase dimer with the N-termini of other prenyltransferase dimers (Figure 7). As first noted in the structure of dimeric avian FPP synthase (Tarshis et al., 1994), helices A and B form a “latch” that mediates subunit-subunit interactions in the prenyltransferase dimer. In hexameric prenyltransferases, the corresponding helix A is not visible, but helix B is uniformly present (though sometimes broken) and maintains the molecular latch function between dimer subunits, for example, as observed in the hexameric prenyltransferases of PvCPS, PaFS, macrophomene synthase, and hGGPPS (Kavanagh et al., 2006; Chen et al., 2016; Ronnebaum et al., 2020; Tao et al., 2022). Intriguingly, helix A is observed in the structure of the PfCPS prenyltransferase, but it adopts an extended orientation in a much different position relative to helix A in FPP synthase. This helix A conformation is observed uniformly in all dimer subunits regardless of whether the dimer-dimer interface is intact or not, and presumably augments the latch function established by helix B. Notably, helix A leads to the disordered linker that connects the prenyltransferase to the class II cyclase domain in PfCPS.

Figure 7. Subunit-subunit interactions in prenyltransferase dimers.

The conformations of helices A and B at the N-termini of prenyltransferase domains are highlighted in red in each dimer. The conformation of helix A in PfCPS is unique in comparison with other prenyltransferases. In PfCPS, the class II cyclase domain and the disordered polypeptide linker are connected to helix A. Neither the linker nor the cyclase domain is visible in reconstructions of the PfCPS prenyltransferase hexamer due to disorder.

While the 61-kD class II cyclase domains are predominantly splayed-out from prenyltransferase hexamers and mostly invisible in reconstructions of the prenyltransferase hexamer, individual cyclase domains can be spotted in micrographs and yield moderate to low resolution 2D reconstructions (Figure S4). However, these data were not sufficient for the generation of high-quality 3D reconstructions of the cyclase domain. Even so, it is clear from analysis of 2D reconstructions that the βγ architecture of the cyclase domain is very similar to that first observed in the class II terpene cyclase, squalene-hopene cyclase (Wendt et al., 1997).

The 69-residue disordered linker segment connecting the cyclase domain to the prenyltransferase domain, A539–P608, is of course not visible in cryo-EM maps. However, amino acid composition analysis of this segment using localCIDER (Holehouse et al., 2017) indicates that this segment corresponds to a Janus sequence, an intrinsically disordered peptide that can adopt a collapsed or expanded conformation depending on its context. Such conformational flexibility might enable transient prenyltransferase-cyclase association.

Discussion

The first bifunctional terpene synthase to be identified with αβγ domain architecture was abietadiene synthase from the grand fir tree, Abies grandis (Vogel et al., 1996; Zhou et al., 2012). This plant enzyme catalyzes the class II cyclization of GGPP at the βγ domain interface to form copalyl diphosphate, which then undergoes a class I cyclization reaction in the α domain to yield abietadiene regioisomers. As for all αβγ diterpene cyclases, the α and βγ domains of abietadiene synthase are closely associated and linked by a ~40-residue helical segment (Köksal et al., 2011a,b; Zhou et al., 2012). Chemical function and domain architecture are modified in class II assembly-line terpene synthases such as PfCPS, in that the α domain is a prenyltransferase and it is connected to the βγ domains of the class II cyclase through a ~70-residue disordered linker. Here, such systems are represented as α~βγ.

Our cryo-EM studies show that the prenyltransferase domain of PfCPS drives oligomerization to form a hexamer, which is readily described as a trimer of dimers. The fundamental quaternary structural unit of a prenyltransferase is a dimer, as first observed in the crystal structure of FPP synthase (Tarshis et al., 1994), but dimers can assemble to form hexamers with D3 symmetry as first observed in the crystal structure of hGGPPS (Kavanagh et al., 2006). Since a mixture of hexamers and octamers is observed for hGGPPS in solution (Kuzuguchi et al., 1999), the D3 symmetry of the hexameric prenyltransferase must be broken to enable insertion of another dimer to form an octamer.

While 1.5 mg/mL PfCPS exhibits exclusively hexameric quaternary structure in solution based on size-exclusion chromatography, the related assembly-line terpene synthase, PvCPS, exhibits significant concentration-dependent oligomerization – monomers, dimers, tetramers, hexamers, and higher-order aggregates are observed in solution at concentrations ranging from 0.1–9.4 mg/mL (Ronnebaum et al., 2020). Of note, quaternary structure does not influence prenyltransferase activity over the protein concentration range 0.156–1.25 mg/mL (the approximate upper limit of the assay). It is reasonable to expect that this is the case for other oligomeric prenyltransferases such as PfCPS.

We suggest that the open conformation of the PfCPS prenyltransferase observed by cryo-EM represents an intermediate conformation that would be encountered by a prenyltransferase hexamer as it undergoes oligomerization from a hexamer to an octamer. Interestingly, the X-ray crystal structure of D188Y hGGPPS reveals an even more open hexamer conformation (Figure 6); it is further notable that this open conformation results from hexamer-hexamer interactions in the crystal lattice (Figure S5). This suggests a simple model for potential oligomerization of the prenyltransferase domain in systems that exhibit oligomeric heterogeneity:

$$\begin{array}{l}{\left(\alpha ~\beta \gamma \right)}_6+{\left(\alpha ~\beta \gamma \right)}_6\rightleftharpoons {\left(\alpha ~\beta \gamma \right)}_8+{\left(\alpha ~\beta \gamma \right)}_4\\ {\left(\alpha ~\beta \gamma \right)}_4+{\left(\alpha ~\beta \gamma \right)}_4\rightleftharpoons {\left(\alpha ~\beta \gamma \right)}_8\end{array}$$

Even though PfCPS is not observed to form octamers at concentrations up to 1.5 mg/mL, octamers are observed in another bifunctional assembly-line terpene synthase, fusicoccadiene synthase, where an approximate ratio of 10:1 is observed for octamer:hexamer in raw cryo-EM micrographs (Faylo et al., 2021a). If this ratio reflects a thermodynamic equilibrium, the Gibbs free energy difference between octamer and hexamer is only ~1.3 kcal/mol. Thus, relatively small changes in sequence or conditions may influence the oligomerization behavior of an assembly-line terpene synthase.

What catalytic advantage might result from oligomerization of a bifunctional α~βγ terpene synthase? Substrate competition experiments with the bifunctional α~α terpene synthase PaFS demonstrate that most of the GGPP generated by the prenyltransferase stays on the enzyme for cyclization to form fusicoccadiene (Pemberton et al., 2017). The close proximity of multiple prenyltransferase and cyclase domains resulting from oligomerization may enhance carbon management through cluster channeling (Castellana et al., 2014), in which case a bigger cluster, i.e., an octamer instead of a hexamer, might enhance this effect. However, it has not yet been determined whether substrate channeling is operative in PfCPS. Even so, 2D class averages show that the class II cyclase domain can occupy discrete splayed-out and more closely-associated positions around the central prenyltransferase hexamer (Figure 3C). The resulting proximity of catalytic domains might facilitate GGPP transit from the prenyltransferase to the cyclase. It appears that cyclase domains are capable of transient association with the prenyltransferase hexamer, and are in equilibrium between splayed-out and associated positions as modeled in Figure 8.

Figure 8. PfCPS conformational model.

The proposed equilibrium between splayed-out and more closely-associated positions of cyclase domains is illustrated for the PfCPS hexamer using the cryo-EM structure of the prenyltransferase hexamer (PDB 8V0F) and an AlphaFold2 model of the class II cyclase domain. The model of the class II cyclase domain is very similar to the crystal structure of the class II terpene cyclase squalene-hopene cyclase (PDB 1SQC). Cyclase positions are based on those observed in 2D class averages (Figure 3C). The disordered linker segment (red) is represented schematically and can adopt extended or condensed conformations.

In summary, the 2.81 Å resolution cryo-EM structure of the hexameric prenyltransferase core of PfCPS reveals an open conformation in which one of the three dimer-dimer interfaces is separated so as to break the D3 symmetry exclusively observed for wild-type prenyltransferase hexamers in prior studies. A slightly larger open conformation is observed for the D188Y hGGPPS hexamer (Lisnyansky et al., 2018), but this appears to be caused by hexamer-hexamer contacts in the crystal lattice. The cryo-EM structure of PfCPS is the first to reveal that an open hexamer conformation can occur when the enzyme is free in flash-cooled solution, and it suggests a possible pathway for hexamer-octamer quaternary structural transitions as observed for other oligomeric prenyltransferases. Future studies will explore the influence of these structural transitions on catalytic function.

Highlights.

• P. fellutanum copalyl diphosphate synthase is an assembly-line terpene synthase

• This bifunctional enzyme catalyzes prenyltransferase and cyclase reactions

• Cryo-EM analysis reveals an open hexamer conformation for the prenyltransferase

• The open hexamer conformation may facilitate hexamer-octamer equilibration

• Splayed-out and more closely-associated positions are found for cyclase domains

Data availability

Cryo-EM maps of the PfCPS prenyltransferase hexamer with C1 and C2 symmetry have been deposited in the Electron Microscopy Data Bank (EMDB, www.ebi.ac.uk/pdbe/emdb) with accession codes EMD-42855 and EMD-42853, respectively. Atomic coordinates of the PfCPS prenyltransferase hexamer have been deposited in the Protein Data Bank (www.rcsb.org) with accession code 8V0F.