X-ray Crystallographic Studies of Substrate Binding to Aristolochene Synthase Suggest a Metal Ion Binding Sequence for Catalysis*

The universal sesquiterpene precursor, farnesyl diphosphate (FPP), is cyclized in an Mg2+-dependent reaction catalyzed by the tetrameric aristolochene synthase from Aspergillus terreus to form the bicyclic hydrocarbon aristolochene and a pyrophosphate anion (PPi) coproduct. The 2.1-Å resolution crystal structure determined from crystals soaked with FPP reveals the binding of intact FPP to monomers A-C, and the binding of PPi and Mg2+B to monomer D. The 1.89-Å resolution structure of the complex with 2-fluorofarnesyl diphosphate (2F-FPP) reveals 2F-FPP binding to all subunits of the tetramer, with Mg2+Baccompanying the binding of this analogue only in monomer D. All monomers adopt open activesite conformations in these complexes, but slight structural changes in monomers C and D of each complex reflect the very initial stages of a conformational transition to the closed state. Finally, the 2.4-Å resolution structure of the complex with 12,13-difluorofarnesyl diphosphate (DF-FPP) reveals the binding of intact DF-FPP to monomers A-C in the open conformation and the binding of PPi, Mg2+B, and Mg2+C to monomer D in a predominantly closed conformation. Taken together, these structures provide 12 independent “snapshots” of substrate or product complexes that suggest a possible sequence for metal ion binding and conformational changes required for catalysis.

Farnesyl diphosphate (FPP), 2 a flexible 15-carbon isoprenoid, is the universal precursor of Ͼ300 different cyclic sesqui-terpenes found in numerous plants, bacteria, and fungi (1,2). The cyclization of FPP is catalyzed by a sesquiterpene cyclase that utilizes a trinuclear magnesium cluster to trigger the departure of the pyrophosphate (PP i ) leaving group, thereby forming an allylic carbocation that typically reacts with one of the remaining bonds of the substrate (3)(4)(5)(6)(7). The remarkable diversity of sesquiterpene structure and stereochemistry is the consequence of precise control exerted by the cyclase over the conformations of the flexible substrate and carbocation intermediates in the cyclization cascade.
Aristolochene synthase from Aspergillus terreus is a sesquiterpene cyclase that catalyzes the cyclization of FPP to form aristolochene (Fig. 1a), the parent hydrocarbon of a large group of fungal toxins such as gigantenone, PR-toxin, and bipolaroxin (8). In contrast to aristolochene synthase from Penicillium roqueforti, which generates aristolochene predominantly (Ͼ90%) but also small amounts of germacrene A and valencene (9,10), aristolochene synthase from A. terreus generates aristolochene exclusively (9). Each cyclase adopts the common ␣-helical fold of a class I terpenoid cyclase and contains two conserved metal binding motifs: the "aspartate-rich" motif D 90 DLLE that coordinates to Mg 2ϩ A and Mg 2ϩ C , and the "NSE/ DTE" motif N 219 DIYSYEKE that chelates Mg 2ϩ B (boldface residues are metal ligands) (11). X-ray crystallographic studies show that the binding of Mg 2ϩ 3 ⅐PP i stabilizes the active site in a closed conformation that is completely sequestered from bulk solvent (Fig. 1b) (11). In addition to multiple metal coordination interactions, the PP i anion accepts hydrogen bonds from conserved residues Arg 175 , Lys 226 , Arg 314 , and Tyr 315 when bound to the closed conformation (Fig. 1c). It is likely that the diphosphate group of FPP makes identical metal coordination and hydrogen bond interactions in the Michaelis complex, i.e. the complex between the enzyme and the productively bound substrate that immediately precedes the initiation of the cyclization cascade.
Substrate conformation is a crucial determinant of the biosynthetic outcome of the terpenoid cyclase reaction. The active site of aristolochene synthase from A. terreus serves as a high fidelity template that fixes FPP in a single, productive conformation in the Michaelis complex; otherwise, aberrant cyclization products would result. To study the conformational control of FPP in the active site of aristolochene synthase from A. terreus, we now report the structures of crystalline complexes with the intact substrates FPP and 2-fluorofarnesyl diphosphate (2F-FPP) (12), and the inhibitor 12,13-difluorofarnesyl diphosphate (DF-FPP) (13). Differences observed in active site conformations and substrate conformations appear to be linked to differences in metal binding, analysis of which suggests a possible sequence for metal ion binding and conformational changes required for catalysis.

MATERIALS AND METHODS
Enzyme Incubations with 2F-FPP and DF-FPP in Solution-The syntheses of 2F-FPP and DF-FPP have been reported (12)(13)(14), and an additional synthesis of 2F-FPP is outlined in the supplemental materials. Recombinant A. terreus aristolochene synthase was expressed in Escherichia coli BL21(DE3)pLysS and purified as previously described (11,15). Analysis of the products generated by enzyme incubation with 2F-FPP was performed using gas chromatography and mass spectroscopy (GC-MS) using a similar procedure as previously described for the reaction with substrate FPP (9).
Briefly, a 4-ml solution of 500 nM aristolochene synthase, 20 mM Hepes (pH 7.6), 10% glycerol, 5 mM MgCl 2 , and 5 mM ␤-mercaptoethanol was incubated with 50 -100 M 2F-FPP or DF-FPP and overlaid with high-performance liquid chromatography-grade n-pentane (Fisher P399-1) in a glass centrifuge tube at 30°C for about 30 h. After the initial 10-h incubation, 500 nM enzyme and 50 M 2F-FPP or DF-FPP were added twice into the aqueous reaction layer at 6-h intervals. Reaction products were extracted with high-performance liquid chromatography-grade n-pentane, and the organic extract was passed through a 3-cm 230 -400 mesh silica gel column. The purified extract was concentrated on an ice-water mixture under reduced pressure until the volume was decreased to ϳ50 l. The resulting concentrate was analyzed using a Hewlett-Packard 6890 gas chromatograph coupled to a 5973 mass selective detector and equipped with an HP-5MS capillary column (0.25 mm inner diameter ϫ 30 m with 0.25-m film, Agilent Technologies). Typically, a 2-l sample was injected in splitless mode into the GS-MS. The oven temperature was at 35°C for 1 min, then raised to 230°C at the rate of 5°C/min, and 3 ⅐PP i complex (green) (11), and the FPP/PP i complex (blue). The binding of Mg 2ϩ 3 ⅐PP i triggers significant conformational changes in helices G1 and H and loops F-G1 and A-C1. Additionally, the N-terminal portion of the H-␣1 loop is incorporated into an extension of helix H; helices C1, D, D1, F, G1, H, and I shift inward, and the H-␣1 loop (Ser 231 -Gly 239 ), disordered in the unliganded enzyme, becomes ordered and caps the active site. In the FPP complex (blue), helices C1 and H and loops A-C1 and H-␣1 shift slightly toward their expected positions in the closed conformation, reflecting the very initial stages of a transition from the open active site conformation to the closed active site conformation. c, aristolochene synthase-Mg 2ϩ 3 ⅐PP i complex (11); PP i metal coordination and hydrogen bond interactions are shown as red dotted lines. then followed by a 20°C/min increase to 280°C. The temperature was maintained at 280°C for 20 min.
Crystal Structure Determinations of Enzyme Complexes with FPP, 2F-FPP, and DF-FPP-Recombinant A. terreus aristolochene synthase was crystallized by the hanging drop, vapor diffusion method at 4°C as previously described (11). The structures of complexes with substrate FPP and substrate analogues 2F-FPP and DF-FPP were determined by soaking crystals of the unliganded enzyme for 20 -24 h in a buffer solution containing 100 mM Tris-HCl (pH 8.4), 18% polyethylene glycol 6000, 0.5 M NaCl, 1 mM MgCl 2 , and 2 mM substrate or substrate analogue. X-ray diffraction data were collected from crystals of each complex at the Brookhaven National Synchrotron Light Source and processed using HKL (16) or Mosflm (17). Crystals were isomorphous with those of the unliganded enzyme (space group P2 1 , a ϭ 61.2 Å, b ϭ 147.2 Å, c ϭ 83.7 Å, ␤ ϭ 97.9°, with four molecules in the asymmetric unit (11)). Iterative rounds of refinement and model adjustment for each complex were performed using CNS and O, respectively (18,19). Individual B-factors were refined in each complex.
The N termini Met 1 -Ser 12 , C termini Val 318 -Asp 320 , and loop segments Ser 231 -Gly 239 were disordered in all monomers of each structure. Additionally, loop segments Phe 46 -Pro 47 and Arg 160 -Thr 161 were disordered in monomer D of the FPP complex, loop segment Leu 286 -Glu 287 was disordered in monomer C of the 2F-FPP complex, and loop segments Asn 45 -Asn 48 , Lys 54 -Phe 55 , and Tyr 95 -Ser 97 were disordered in monomer D of the DF-FPP complex. Disordered residues were excluded from the final model of each complex. The exclusion of 111/ 1280 residues from each tetramer (8.7% of the scattering matter) likely contributed to the somewhat higher than expected R and R free values recorded in Table 1. Continuous electron den-sity peaks adjacent to 5, 3, and 2 cysteine residues, respectively, were modeled as disulfide-linked ␤-mercaptoethanol molecules in the FPP, 2F-FPP, and DF-FPP complexes. Data collection and refinement statistics for each complex are recorded in Table 1. The figures were generated using the program PyMOL. 3

Reactivity of Fluorinated
Substrates-Aristolochene synthase from A. terreus generates (ϩ)-aristolochene as its sole product from reaction with substrate FPP (9). However, following incubation with 2F-FPP in solution, two products are identified in a 95:5 ratio by GS-MS (Fig. 2). The chemical ionization mass spectrum of the major peak eluting at 25.1 min corresponds to a fluorinated sesquiterpene (C 15 H 23 F) with m/z ϭ 222 (data not shown). This result is consistent with the generation of 2-fluorogermacrene A, as observed after incubating aristolochene synthase from P. roqueforti with 2F-FPP (12). The small peak at 23.37 min has a chemical ionization spectrum indicative of a non-fluorinated sesquiterpene (C 15 H 24 ) with M-1 ϭ 204 (data not shown); this is presumed to be a degradation product. A control experiment conducted in the absence of enzyme indicates that 2F-FPP itself does not react to yield any extractable products based on GC-MS analysis.
Incubation of A. terreus aristolochene synthase with DF-FPP for 30 h using the approach described above for 2F-FPP does not generate any pentane-extractable products based on GC-MS analysis. This is consistent with the lack of activity with DF-FPP reported for aristolochene synthase from P. roqueforti,   against which DF-FPP is a reversible competitive inhibitor with K i ϭ 0.8 M (13).
Crystal Structure of the Complex with FPP-The structure of tetrameric aristolochene synthase determined from crystals soaked in a buffer solution containing FPP reveals that monomers A-D adopt open active site conformations similar to those of the unliganded enzyme (11). Intact FPP molecules with partially similar folded conformations bind in the active site clefts of monomers A, B, and C (Fig. 3, a  and b), whereas Mg 2ϩ B ⅐PP i binds in the active site cleft of monomer D (Fig. 3c). The r.m.s. deviations of 295 C␣ atoms in monomers A and B with the corresponding monomers in the unliganded enzyme are 0.28 Å and 0.26 Å, respectively, indicating that these monomers remain in predominantly open conformations. However, some slight structural changes were observed in monomers C and D that reflect the very initial stages of a transition to the closed conformation: helices C1, D, and D1 and loops A-C1 and D-D1 shift slightly toward their expected positions in the closed conformation of monomer C, and helices C1 and H do likewise in monomer D (for reference, the locations of these structural elements are indicated in Fig. 1b). Accordingly, the r.m.s. deviations with the unliganded enzyme are larger for these monomers: 0.59 Å for 295 C␣ atoms in monomer C, and 0.51 Å for 290 C␣ atoms in monomer D. Even so, the active sites of monomers C and D remain in predominantly open, solvent-exposed conformations.
In each of monomers A, B, and C, the terminal diphosphate group of FPP accepts hydrogen bonds from Arg 314 and Tyr 315 , and the terminal isoprenoid group of FPP is nestled between Phe 87 and Phe 153 . The conformation of the middle isoprenoid moiety of FPP in monomer B is flipped, however, compared with that observed in monomers A and C, both of which have the predicted relative and absolute orientation of the C3 and C7 methyl groups (Fig.  3b). Notably, none of the observed FPP conformations correspond to that required for the formation of germacrene A (the first intermediate in aristolochene biosynthesis), because the electrophilic C1 atom is positioned ϳ6 Å away from the C10 -C11 bond and is not properly oriented for the C1-C10 bondforming step. The binding of FPP to monomers A, B, and C thus may provide "snapshots" of possible substrate binding conformations in the open active site prior to metal ion binding and active site closure.
The active site cleft of monomer D contains only the Mg 2ϩ B ⅐PP i complex (Fig. 3c). Notably, only monomer D appears to be capable of achieving the closed conformation required for catalysis in complex with Mg 2ϩ 3 ⅐PP i (11). Although aristolochene synthase from A. terreus is a dimer in solution (11), the assembly of two dimers to form the tetramer observed in the crystal structure may hinder the conformational transition from the open, inactive state to the closed, active state, thereby accounting for the attenuation of catalytic activity with increasing protein concentrations (9). Additionally, catalytically required conformational changes may be hindered in monomers A-C due to crystal lattice contacts. In monomer D, the observed single-metal ion complex Mg 2ϩ B ⅐PP i is presumably a remnant of the FPP cyclization reaction catalyzed by this monomer in the crystal. This structure represents the first observation of product PP i binding to the predominantly open active site conformation of a wild-type terpenoid cyclase.
Because the orientation of PP i bound in the presence of Mg 2ϩ B in monomer D differs from the orientations of the FPP diphosphate groups in monomers A-C (Fig. 3d) Crystal Structure of the Complex with 2F-FPP-Despite the evidence for catalytic activity with 2F-FPP in solution (Fig. 2), the crystal structure of the aristolochene synthase tetramer complexed with 2F-FPP reveals the binding of intact 2F-FPP to all four monomers with open active site conformations. As observed in the FPP complex, slight structural changes are observed in monomers C and D that appear to reflect the very initial stages of a transition to the closed conformation. The r.m.s. deviations with the unliganded enzyme are 0.26 Å, 0.29 Å, 0.57 Å, and 0.52 Å for monomers A, B, C, and D, respectively.
Electron density corresponding to the diphosphate group of 2F-FPP is well defined in monomers A and D (Fig. 4a) but ambiguous in monomers B and C (data not shown). Only one metal ion, Mg 2ϩ B , is bound in this complex, and it binds only to the enzyme⅐substrate complex in monomer D. Here, the substrate diphosphate group chelates Mg 2ϩ B and accepts hydrogen bonds from Arg 314 , Tyr 315 , and Lys 226 . The position and orientation of the diphosphate group of 2F-FPP is roughly comparable to that of PP i in the Mg 2ϩ B ⅐PP i complex (Fig. 4b). Strikingly, comparison of the metal-free and metal-bound conformations of 2F-FPP in monomers A and D, respectively, reveals a significant Mg 2ϩ B -dependent reorientation of the substrate (Fig. 4c). Thus, it appears that Mg 2ϩ B is the key metal ion governing the proper binding orientation of the substrate diphosphate group.
In each monomer, the fluorine atom of 2F-FPP accepts a hydrogen bond from Lys 181 . As observed for FPP, the terminal isoprenoid group of 2F-FPP is nestled between residues Phe 87 and Phe 153 , but the 2F-FPP molecules adopt somewhat varying folded conformations in each monomer. Because none of the observed 2F-FPP conformations in monomers A-D are consistent with that required for C1-C10 bond formation in the first step of catalysis, the observed binding conformations of 2F-FPP in the presence or absence of Mg 2ϩ B suggest that additional metal ion binding and active site closure are necessary to chaperone the substrate into a cyclization-competent conformation. Regardless  Fig. 5a. No metal ions accompany DF-FPP binding, and the observed isoprenoid and diphosphate conformations are somewhat variable, suggesting some degree of disorder (Fig. 5b). The binding of DF-FPP to monomers A-C provides additional snapshots of substrate binding conformations that, in the absence of metal ion binding and full active site closure, are not cyclization-competent. In contrast, monomer D adopts a predominantly closed conformation and contains a PP i anion complexed with Mg 2ϩ B and Mg 2ϩ C (Fig. 5c). There is no residual electron density in the active site to indicate that an intact DF-FPP molecule is present but disordered. A key conformational change evident in this structure is that the aspartate-rich and NSE/DTE metal ion binding motifs on helices D and H, respectively, are brought closer together to accommodate the binding of the binuclear magnesium cluster. Reflecting this significant conformational change, the r.m.s. deviation of 278 Ca atoms with the unliganded structure is 0.83 Å.
Although the incubation of DF-FPP with the A. terreus enzyme (data not shown) or the P. roqueforti enzyme (13) yields no observable products, it is possible that a slow reaction with this substrate analogue occurs over the time course of the x-ray crystallographic experiment to yield the observed PP i product. Notably, the O␦1 and O␦2 atoms of Asp 90 coordinate to Mg 2ϩ C with bidentate geometry, whereas in the complex of aristolochene synthase with Mg 2ϩ 3 ⅐PP i , Asp 90 coordinates to both Mg 2ϩ A and Mg 2ϩ C with synsyn bidentate geometry (i.e. Asp 90 O␦1 coordinates to Mg 2ϩ A and Asp 90 O␦2 coordinates to Mg 2ϩ C ) (11). Some, but not all, hydrogen bond interactions observed in the Mg 2ϩ 3 ⅐PP i cluster (11) are observed in the Mg 2ϩ B ⅐Mg 2ϩ C ⅐PP i complex: Arg 314 and Tyr 315 donate hydrogen bonds to PP i , but Lys 226 is characterized by poor electron density and appears to be somewhat disordered, and Arg 175 is too far from PP i (5.7 Å) to be considered a formal hydrogen bond donor.
Importantly, this structure reveals that only 2 Mg 2ϩ ions are needed to stabilize the predominantly closed active site conformation. Specifically, the binding of Mg 2ϩ C to the FPP/PP i ⅐Mg 2ϩ B complex appears to be the key molecular event that triggers the conformational transition of the active site from the open state to the closed state. The binding of Mg 2ϩ A to complete the trinuclear magnesium cluster also completes these conformational transitions such that the r.m.s. deviation of 295 C␣ atoms between open and closed conformations is 1.5 Å.

DISCUSSION
The three crystal structures reported herein contain 12 independent structures of A. terreus aristolochene synthase monomers that illuminate an unprece-   Fig. 3c. The interaction of the 2F-FPP diphosphate group with Mg 2ϩ B is roughly similar to that observed for PP i . Metal coordination interactions are indicated by dashed lines; water molecules are omitted for clarity. c, superposition of 2F-FPP bound to monomer A (red) and monomer D (blue) illustrates the Mg 2ϩ B -triggered reorientation of the substrate diphosphate group (for clarity, only the protein atoms of monomer D are shown).

Mg 2ϩ
C (11), these structures suggest a possible sequence for metal ion binding and conformational changes required for catalysis.
Substrate Conformation-The observed conformations of intact FPP, 2F-FPP, and DF-FPP are not cyclization-competent due to the fact that they are bound to the open active site conformation. This is a direct consequence of incomplete metal binding, because a complete trinuclear magnesium cluster is required to stabilize the fully closed active site conformation. It is only in the closed conformation that the active site is sequestered from bulk solvent, thereby forming a protected environment in which reactive carbocation intermediates can be formed without the risk of premature quenching by solvent.
Interestingly, most of the crystal structures reported to date of class I terpenoid cyclases complexed with isoprenoid substrate analogues similarly reveal isoprenoid conformations that are not catalytically competent (21)(22)(23). As with aristolochene synthase, this often appears to be a consequence of incomplete or compromised metal binding, e.g. as observed in the complex of 5-epi aristolochene synthase with trifluorofarnesyl diphosphate (22). However, isoprenoid conformations that are not cyclization-competent are occasionally observed in monoterpene cyclase active sites with closed conformations stabilized by complete trinuclear magnesium clusters. For example, the binding of 3-aza-2,3dihydrogeranyl diphosphate and 3 Mg 2ϩ ions to bornyl diphosphate synthase triggers active site closure, and the position, orientation, and intermolecular contacts of the analogue diphosphate group are nearly identical to those observed in the Mg 2ϩ 3 ⅐PP i complex (21). Even so, the conformation of the substrate analogue is inconsistent with that required for cyclization. More recently, cocrystallization of limonene synthase with 2-fluorogeranyl diphosphate yields the structure of the complex with 2-fluoro-(3S)linalyl diphosphate, a catalytic intermediate, bound with an extended conformation that is not cyclization-competent; in contrast, cocrystallization with racemic 2-fluorolinalyl diphosphate yields the helical, cyclization-competent isoprenoid conformation (23). This suggests the possibility that isomerization of geranyl diphosphate to linalyl diphosphate does not require a cyclization-competent substrate conformation, such a conformation could instead be achieved after formation of the linalyl diphosphate intermediate.
Metal Binding-Analyses of enzyme⅐substrate and enzyme⅐ PP i complexes reported herein and comparisons with the structure of the enzyme⅐Mg 2ϩ 3 ⅐PP i complex reported previously (11) lead to a proposed model for substrate and metal binding (Fig. 6)  by the Mg 2ϩ B -dependent reorientation of the 2F-FPP diphosphate group shown in Fig. 4c. Because it is only Mg 2ϩ B that is chelated by three protein residues, it is reasonable to hypothesize that this is the first of the three metal ions to bind (its binding site is largely pre-organized). Some slight conformational changes reflecting the very initial stages of active site closure may be triggered by Mg 2ϩ B and/or substrate binding, e.g. as observed in monomers C and D of the FPP/PP i and 2F-FPP complexes.
Subsequently, the binding of Mg 2ϩ C by both Asp 90 and the substrate diphosphate group would trigger more substantial conformational changes leading to active site closure, as illustrated by the Mg 2ϩ B ⅐Mg 2ϩ C ⅐PP i complex in Fig. 5c. Finally, the binding of Mg 2ϩ A would complete the formation of the trinuclear metal ion cluster, causing Mg 2ϩ C to shift such that Asp 90 would coordinate to both Mg 2ϩ A and Mg 2ϩ C with synsyn bidentate geometry.
The formation of a complete trinuclear magnesium cluster would complete conformational changes required for full active site closure. In the closed conformation, substrate diphosphate metal ion coordination interactions and hydrogen bond interactions with Arg 175 , Lys 226 , Arg 314 , and Tyr 315 would be fully formed, as observed in the Mg 2ϩ 3 ⅐PP i complex (Fig. 1c) (11). Upon achieving a cyclization-competent substrate conformation in the closed active site, with the three negative charges of the diphosphate leaving group fully neutralized, substrate ionization (24,25) and cyclization would be initiated.
At the conclusion of the cyclization cascade, we suggest that metal ion dissociation and the transition to the open active site conformation occurs in reverse sequence: Mg 2ϩ A would disso-ciate first, but Mg 2ϩ B and Mg 2ϩ C would stabilize the active site conformation in a predominantly closed state, e.g. as observed in the Mg 2ϩ B ⅐Mg 2ϩ C ⅐PP i complex in Fig.  5c. The subsequent dissociation of Mg 2ϩ C would then facilitate the transition to an open active site conformation with the release of product aristolochene, and the release of Mg 2ϩ B ⅐PP i would conclude the catalytic sequence.
Additional support for the proposed role of metal binding in the conformational transition of the terpenoid synthase active site derives from site-directed mutagenesis studies of another fungal sesquiterpene cyclase, trichodiene synthase, related to aristolochene synthase by only 15% amino acid sequence identity. The r.m.s. deviation of 349 C␣ atoms between unliganded and Mg 2ϩ 3 ⅐PP i complexed wild-type trichodiene synthase is 1.4 Å (26). In contrast, D100E trichodiene synthase is defective in Mg 2ϩ C binding and does not achieve a fully closed active site conformation in its complex with Mg 2ϩ A , Mg 2ϩ B , and PP i ; the r.m.s. deviation of 354 C␣ atoms between the unliganded and liganded enzymes is only 0.44 Å (27). Therefore, Mg 2ϩ C must play a critical role in triggering the structural transition of the trichodiene synthase active site from the open to the closed conformation, just as Mg 2ϩ C is proposed to play a critical role in triggering the structural transition of the aristolochene synthase active site from the open to the closed conformation.
In accord with the metal binding behavior observed with aristolochene synthase and trichodiene synthase, we conclude that the metal binding sequence summarized in Fig. 6 suggests specific functions for the three metal ions in terpenoid cyclasecatalyzed ionization-cyclization reactions: Mg 2ϩ B governs initial enzyme⅐substrate association; Mg 2ϩ C triggers the conformational transition of the active site from the open state to a predominantly closed state in which metal ion binding motifs on helices D and H are brought closer together; and Mg 2ϩ A completes the trinuclear magnesium cluster and thereby completes the conformational transition to the closed state and initiates the cyclization cascade. Future studies of site-specific mutants in the metal binding site of aristolochene synthase and other terpenoid cyclases will allow us to test this model of metal ion function.
Acknowledgments-We thank the National Synchrotron Light Source at Brookhaven National Laboratory for access to x-ray crystallographic data collection facilities, and we thank Drs. Luigi Di Costanzo and Heather Gennadios for helpful scientific discussions as well as their assistance in preparing the figures.