Crystal Structure of Type-III Geranylgeranyl Pyrophosphate Synthase from Saccharomyces cerevisiae and the Mechanism of Product Chain Length Determination*

Geranylgeranyl pyrophosphate synthase (GGPPs) catalyzes a condensation reaction of farnesyl pyrophosphate with isopentenyl pyrophosphate to generate C20 geranylgeranyl pyrophosphate, which is a precursor for carotenoids, chlorophylls, geranylgeranylated proteins, and archaeal ether-linked lipid. For short-chain trans-prenyltransferases that synthesize C10-C25 products, bulky amino acid residues generally occupy the fourth or fifth position upstream from the first DDXXD motif to block further elongation of the final products. However, the short-chain type-III GGPPs in eukaryotes lack any large amino acid at these positions. In this study, the first structure of type-III GGPPs from Saccharomyces cerevisiae has been determined to 1.98 Å resolution. The structure is composed entirely of 15 α-helices joined by connecting loops and is arranged with α-helices around a large central cavity. Distinct from other known structures of trans-prenyltransferases, the N-terminal 17 amino acids (9-amino acid helix A and the following loop) of this GGPPs protrude from the helix core into the other subunit and contribute to the tight dimer formation. Deletion of the first 9 or 17 amino acids caused the dissociation of dimer into monomer, and the Δ(1-17) mutant showed abolished enzyme activity. In each subunit, an elongated hydrophobic crevice surrounded by D, F, G, H, and I α-helices contains two DDXXD motifs at the top for substrate binding with one Mg2+ coordinated by Asp75, Asp79, and four water molecules. It is sealed at the bottom with three large residues of Tyr107, Phe108, and His139. Compared with the major product C30 synthesized by mutant H139A, the products generated by mutant Y107A and F108A are predominantly C40 and C30, respectively, suggesting the most important role of Tyr107 in determining the product chain length.

IPP to the growing chain by a group of prenyltransferases (5). These prenyltransferases either form a cis or trans double bond during each IPP condensation (6). Two conserved aspartate-rich motifs DDXXD, where X encodes any amino acid, found in all trans-prenyltransferases, coordinate with Mg 2ϩ for substrate binding and the subsequent reaction (7)(8)(9)(10), whereas an Asp in a P-loop of the cis-prenyltransferases is responsible for the similar function (11)(12)(13).
In general, trans-prenyltransferases synthesize products up to C 50 in chain lengths, which can be further divided into short-chain (C 10 -C 25 ), medium-chain (C 30 -C 35 ), and long-chain (C 40 -C 50 ) (14) and cis-prenyltransferases generate much longer products (15). In the co-crystal structure of FPP synthase (FPPs) with nonreactive thiol analog of the allylic substrate DMAPP and the homoallylic substrate IPP, the first DDXXD binds the pyrophosphate head group of the allylic substrate via Mg 2ϩ and the second DDXXD coordinates with two other Mg 2ϩ ions to form a trinuclear cluster for pulling the pyrophosphate leaving group of the allylic substrate (16). A bulky amino acid residue at the fourth or fifth position before the first DDXXD motif in helix D (see Fig. 1 for sequence homology) forms a blockage underneath the allylic substrate site to avoid the further elongation of FPP product. Replacement of the large amino acid with a small one removes the blockage and leads to longer products (17)(18)(19). For the long-chain C 40 -octaprenyl pyrophosphate synthase (OPPs), its crystal structure shows a small residue at that position and a large amino acid Phe 132 from another helix (helix G) seals the bottom of the active site crevice (20). Similar to OPPs, medium-chain C 30 hexaprenyl pyrophosphate synthase (HexPPs) also utilizes L164 from helix G for shielding the final product in the middle of the crevice (10).
It is interesting that the short-chain geranylgeranyl pyrophosphate synthases (GGPPs) from different species contain either large or small residues at the fourth or fifth position before the first DDXXD (Fig. 1). By virtue of one IPP condensation reaction with FPP, GGPPs synthesizes C 20 GGPP, which serves as a precursor for carotenoids, geranylgeranylated proteins, chlorophylls, and archaeal ether-linked lipid for the cytoplasmic membrane (21). These GGPPs are classified into three types based on their amino acid sequences; type-I GGPPs contains a large amino acid at the fourth or fifth position prior to the DDXXD motif, but type-II (with insertion of two amino acids within the first Asp-rich motif) and type-III GGPPs found in eukaryotes (except plants) have a small amino acid at the corresponding position (see Fig. 1) (22). To rationalize the mechanism of product chain length determination for the type-III GGPPs, we solved the crystal structure of Saccharomyces cerevisiae GGPPs to 1.98 Å resolution as reported here. While bearing a common structural feature with other trans-prenyltransferases, GGPPs structure reveals a unique orientation of its N terminus. The helix A of type-III GGPPs protrudes into the other subunit, which seems to be involved in the tight dimer formation. Mutants truncated by deleting this helix and the following loop region were examined for dimer formation as presented here. On the basis of the structure, we carried out site-directed mutagenesis studies to identify the amino acids critical for product chain length determination. Together with recently The stars indicate the amino acid residues mutated in this study. The box denotes the fourth and fifth amino acid residues, which are small in long-chain OPPs, medium-chain HexPPs, and type-II and -III GGPPs but large in short-chain FPPs and type-I GGPPs.
solved type-I GGPPs, 4 our structures enhance the understanding of the chain elongation mechanisms for the short-chain transprenyltransferases.

EXPERIMENTAL PROCEDURES
Materials-Radiolabeled [ 14 C]IPP (55 mCi/mmol) was purchased from Amersham Biosciences. Nonlabeled FPP and seleno-DL-methionine were obtained from Sigma. Reverse-phase TLC plates were purchased from Merck. PfuTurbo DNA polymerase was obtained from Invitrogen. The plasmid miniprep kit, DNA gel extraction kit, and Ni 2ϩ -NTA resin were purchased from Qiagen. Potato acid phosphatase (2 units/mg) was purchased from Roche Applied Science. Factor Xa and the protein expression kit (including the pET32Xa/LIC vector and competent JM109 and BL21 cells) were obtained from Novagen. The QuikChange site-directed mutagenesis kit was obtained from Stratagene. All commercial buffers and reagents were of the highest grade.
Expression and Purification of GGPPs-The gene encoding GGPPs was cloned from S. cerevisiae genomic DNA by using PCR and the forward primer 5Ј-ggtattgagggtcgcatggaggccaagatagatg-3Ј and the reverse primer 5Ј-agaggagagttagagcctcacaattcggataagtgg-3Ј into the pET32Xa/LIC vector. In this construct, we included a five-residue sequence (MTKNK) from Thermotoga maritima OPPs between the N terminus of GGPPs and the Factor Xa cleavage sequence to expose the protease site for tag removal. The recombinant GGPPs plasmid was then used to transform Escherichia coli JM109 competent cells that were streaked on a Luria-Bertani (LB) agar plate containing 100 g/ml ampicillin. Ampicillin-resistant colonies were selected from the agar plate and grown in 5 ml of LB culture containing 100 g/ml ampicillin overnight at 37°C. The correct construct by sequencing was subsequently transformed to E. coli BL21 (DE3) for protein expression. The 60-ml overnight culture of a single transformant was used to inoculate 6 liters of fresh LB medium containing 100 g/ml ampicillin. The cells were grown to A 600 ϭ 0.6 and induced with 1 mM isopropyl ␤-thiogalactopyranoside at 16°C. After 16 h, the cells were harvested by centrifugation at 7,000 ϫ g for 15 min to collect the cell paste.
The enzyme purification was conducted at 4°C. Cell paste was suspended in 75 ml of lysis buffer containing 25 mM Tris-HCl, pH 7.5, and 150 mM NaCl. Cell lysate was prepared with a French pressure cell press (AIM-AMINCO Spectronic Instruments) and centrifuged at 17,000 ϫ g to remove cell debris. The cell-free extract was loaded onto a Ni 2ϩ -NTA column, which had been previously equilibrated with lysis buffer. The column was washed with 10 mM imidazole followed by 20 mM imidazole-containing buffer. His-tagged GGPPs eluted with 100 mM imidazole was dialyzed twice against 3 liters of buffer (25 mM Tris-HCl, pH 7.5, and 150 mM NaCl) and then subjected to Factor Xa digestion to remove the tag. The mixture was then passed through another Ni 2ϩ -NTA column, and subsequently untagged GGPPs was eluted with 10 mM imidazole-containing buffer and then dialyzed twice against 3 liters of buffer (25 mM Tris-HCl, pH 7.5, and 150 mM NaCl) for storage. SDS-PAGE analysis was used to check the purity (Ͼ95%) of GGPPs and its mutants.
Preparation of Selenomethionine-labeled GGPPs-To produce selenomethionine (SeMet)-labeled GGPPs for x-ray phase solving, the plasmid that GGPPs gene was transformed into a BL21 (DE3)-competent cell. The single transformant was first grown overnight at 37°C in 50 ml of LB medium containing 100 g/ml ampicillin. The cells were harvested by centrifugation at 7,000 ϫ g for 15 min. The paste sus-pended in SeMet minimal medium (24) was then used to inoculate 6 liters of SeMet minimal medium with the supplement of 100 g/ml seleno-DL-methionine. Isopropyl ␤-thiogalactopyranoside was added to a final concentration of 1 mM for induction after the cell density reached A 600 ϭ 0.4. The cells were grown for another 96 h at 16°C and then harvested by centrifugation. The subsequent purification procedure for SeMet GGPPs was the same as for the native GGPPs.
Crystallization and Data Collection-Native and SeMet-labeled GGPPs were crystallized using the hanging drop method from Hampton Research (Laguna Niguel, CA) by mixing 2 l of the GGPPs solution (10 -12 mg/ml in 25 mM Tris-HCl, pH 7.5, 150 mM NaCl, and 0.1% Triton X-100) with 2 l of the mother liquor (0.08 M CH 3 COONa, 0.145 M (NH 4 ) 2 SO 4 , 13% polyethylene glycol 4000, 7-9% glycerol, and 7-9% 1,2-propanediol) and equilibrating with 500 l of the mother liquor. Within 7 days at room temperature, crystals grew to dimensions of about 0.4 ϫ 0.15 ϫ 0.15 mm. The x-ray diffraction data for the native GGPPs were collected to 1.98 Å resolution at beam line BL17B2 of the National Synchrotron Radiation Research Center (NSRRC, Hsinchu, Taiwan), and the SeMet-labeled GGPPs data were collected to 2.4 Å at 100 K on the Taiwan Contract BL12B2 station at SPring-8 (Hyogo, Japan) by employing the multiple-wavelength anomalous diffraction method. The diffraction data were processed and scaled by using the program HKL2000 (25). These GGPPs crystals belong to the space group P2 1 2 1 2 1 , with typical unit cell parameters of a ϭ 48 Å, b ϭ 116 Å, and c ϭ 129 Å. Each asymmetric unit contained a dimeric GGPPs molecule.
Structure Determination and Refinement-The multiple-wavelength anomalous diffraction data sets in the resolution range of 30 to 2.4 Å were collected at wavelengths of 1.0332 Å (low remote), 0.9799 Å (inflection point), 0.9796 Å (peak), and 0.9537 Å (high remote) at and near the selenium absorption edge. Using SOLVE (26), 20 selenium sites Residues in additionally allowed regions (%)

5.9
a The values in the parentheses are for the highest resolution shells. MAY 26, 2006 • VOLUME 281 • NUMBER 21

Structure and Mechanism of Type-III GGPPs
in the asymmetric unit were located, and the phase angles were determined in the single wavelength anomalous diffraction method using the peak wavelength data. Further iterative cycles of density modification, model building, and refinement were carried out with RESOLVE (27), Xtalview (28), and the CNS program (29), respectively. R free values were calculated by using 5% reflections. The stereochemical quality of the refined structure was checked with the program PROCHECK (30). Each monomer of the refined structure encompassed 322 of the 335 total residues of the GGPPs sequence, with a small disordered region of residues 315-327. 578 well ordered water molecules were also located and included in the model. Refinement statistics are summarized in Table 1.
All of the structural diagrams were drawn by using the PyMol software (available on the World Wide Web at pymol.sourceforge.net/) and GRASP (31).
Site-directed Mutagenesis of GGPPs-GGPPs mutants were prepared by using the QuikChange site-directed mutagenesis kit in conjunction with the S. cerevisiae GGPPs gene template in the pET32Xa/LIC vector. The mutagenic primers used were prepared by MDBio Inc. (Taiwan). The mutagenic oligonucleotides for performing site-directed mutagenesis are 5Ј-caccgcaaattatatgGCGttcagagccatgcaacttg-3Ј for Y107A, 5Ј-caccgcaaattatatgtatGCGagagccatgcaacttg-3Ј for F108A, 5Ј-caccg-caaattatatgGCGGCGagagccatgcaacttg-3Ј for Y107A/F108A, 5Ј-cgattt-tcaacgaagaaGCGatcaatctacatagg-3Ј for L135A, 5Ј-gaagaattgatcaatcta-GCGaggggacaaggcttgg-3Ј for H139A, and 5Ј-gctcttgcataattccTACcttttaatcgacgatatag-3Ј for S71Y (the mutated nucleotides are underlined and shown in capital letters). The multiple mutants were constructed with the combination of the above mutagenic primers. The mutations were confirmed by sequencing the entire GGPPs mutant gene of the plasmid obtained from the overnight culture. The correct construct was subsequently transformed to E. coli BL21(DE3) for protein expression and purification. The purity (Ͼ95%) of each purified mutant GGPPs was checked by SDS-PAGE.
Kinetic  To measure the initial rate, 40-l portions of the reaction mixture were periodically withdrawn within 10% substrate depletion and then mixed with 10 mM EDTA for reaction termination. The radiolabeled products were then extracted with 1-butanol, and the radioactivities associated with aqueous and butanol phases were separately quantitated by using a Beckman LS6500 scintillation counter. Data of initial rates versus substrate concentrations were analyzed by nonlinear regression of the Michaelis-Menten equation using the KaleidaGraph computer program (Synergy software) to obtain K m and V max values. The k cat was calculated from V max /[E]. Steady-state activity of the S71Y mutant was assayed under the same condition except that GPP instead of FPP was used.
Product Analysis-The GGPPs reaction containing 1 M enzyme (wild-type or mutant GGPPs), 10 M FPP, 150 M [ 14 C]IPP, 0.1% Triton X-100, 0.5 mM MgCl 2 , and 50 mM KCl in 100 mM Hepes buffer (pH 7.5) was incubated for 6 h at 25°C. Ten mM EDTA was used to terminate the enzyme reaction. For identification of S71Y product, GPP instead of FPP was used as the allylic substrate. The radiolabeled products were extracted with 1-butanol. The 1-butanol was then evaporated, and the 20% 1-propanol solution containing 4.4 units/ml acidic phosphatase, 0.1% Triton X-100, 50 mM sodium acetate (pH 4.7) was used to convert the products to the corresponding alcohols according to the reported procedure (32). After the pyrophosphate hydrolysis catalyzed by acidic phosphatase was completed, the polyprenols were extracted with n-hexane. The polyprenols were separated on reversed-phase TLC using acetone/water (19:1) as mobile phase. The radiolabeled products were identified by autoradiography using a bioimaging analyzer FUJIFILM BAS-1500 (Japan) according to their R f (retention factor) values reported (22).
Construction of N-terminal Truncated Mutants-For constructing the genes to express the truncated GGPPs without the first N-terminal 17 amino acids ⌬(1-17) or without the first 9 amino acids ⌬(1-9), a common reverse primer, 5Ј-agaggagagttagagcctcacaattcggataagtgg-3Ј, was used with the forward primer 5Ј-ggtattgagggtcgccaaaatgaaagcttgatttc-3Ј for ⌬(1-17) or 5Ј-ggtattgagggtcgcaataatgatcctgtttggtc-3Ј for ⌬(1-9). The PCR-amplified products from the template were inserted into the pET32Xa/LIC vector as described above. The deletion was confirmed by sequencing, and correct construct was subsequently transformed to E. coli BL21 (DE3) for protein expression. The truncated proteins were purified using a Ni 2ϩ -NTA column, and their activities were measured as the other mutant enzymes described above.
Gel Filtration Experiments-The molecular mass of GGPPs was determined on a Superdex 75 10/300 GL High Performance column (1 ϫ 30 cm; Amersham Biosciences) by comparing the elution volume of the GGPPs (1 and 0.2 mg/ml) with those of protein molecular mass standards, including catalase (232 kDa), aldolase (158 kDa), albumin (67 kDa), ovalbumin (43 kDa), and ribonuclease A (13.7 kDa). A buffer of 25 mM Tris-HCl, pH 7.5, and 150 mM NaCl was used to elute the proteins at a flow rate of 0.5 ml/min.

RESULTS
Overall Structure of S. cerevisiae GGPPs-The crystal structure of S. cerevisiae GGPPs has been determined to 1.98 Å resolution. As shown in Fig. 2, A and B, each asymmetric unit of the crystal unit cell contains one GGPPs dimer, and two identical subunits are associated into a dimer by forming a four-layer helix bundle using helices F and G and a four-layer helix arm using helices A and B. The refined structure of type-III GGPPs in complex with one magnesium ion in each subunit contains amino acid residues 2-314 and 328 -340. The electron density map between residues 315 and 327 at the C terminus is not clearly visible. The structure contains 15 ␣ helices, nine of them (helices A-I) surrounding a large central cavity. Helices F and G are involved in the dimer formation, with the major stabilization coming from the helices F (␣ chain)-F (␤ chain) and F (␣ chain)-G (␤ chain) intersubunit hydrophobic interactions and hydrogen bonding. The side chain of Met 111 is stacked with that of Met 111 from the other subunit, so it is different from the T. maritima OPPs, which uses aromatic amino acid of Phe 117 for hydrophobic interaction (9), and Sulfolobus solfataricus HexPPs, which uses the crossing stacking interaction between Trp 136 from monomer A and Pro 114 from monomer B (10).
Active Site-Two conserved DDXXD motifs located on helices D and I, respectively, near the opening of a deep cleft, are proposed to be the  MAY 26, 2006 • VOLUME 281 • NUMBER 21 substrate-binding pocket ( Fig. 2A). In the superimposed three-dimensional structures of this GGPPs with other trans-prenyltransferases (Fig.  3A), this region is coincident with the active site proposed for FPPs. From the potential surface diagram (not shown), this proposed active site pocket is surrounded by the polar (mostly positively charged) residues, including Arg 84 , Arg 85 , His 68 , Lys 37 , and Arg 40 , which are probably responsible for attracting if not directly interacting with the negatively charged pyrophosphate head groups of the substrates. As revealed by a more focused view of this active site (Fig. 3B), a Mg 2ϩ ion is coordinated by Asp 75 and Asp 79 from the first DDXXD and four water molecules. The distances of coordination bonds from the Mg 2ϩ to Asp 75 , Asp 79 , and four water molecules are 2.06, 2.0, 2.5, 3.08, 3.07, and 2.12 Å, respectively, in ␣ subunit and 2.03, 2.06, 2.3, 2.33, 2.8, and 3.04 Å, respectively, in ␤ subunit. This Mg 2ϩ supposedly coordinates with the pyrophosphate group of FPP. There are two short ␣ helices, helix E (Ser 95 -Phe 99 ) and helix J (Glu 231 -Glu 235 ) on the outer surface of the ␣-cone. The region including the loop between D and E, the loop between I and J, helix J, and the loop between J and K are not part of the active site, but it may switch the top region of the active site between open and closed forms, as shown in FPPs (16) and HexPPs structures (10). The open form may allow the substrates to enter and the final product to release, and the closed form makes the tight binding of the substrates with the enzyme (16).

Structure and Mechanism of Type-III GGPPs
The Amino Acids Responsible for Controlling Product Chain Length-The active site crevice is surrounded by five ␣-helices (helices D, F, G, H, and I). Underneath the substrate-binding site, three large amino acid residues, including Tyr 107 and Phe 108 on the side of helix F and His 139 on another side of helix G, occupy the bottom portion of the elongated crevice (Fig. 4A). These residues are probably responsible for blocking the further chain elongation of the C 20 products. In fact, His 139 has been previously suggested to be important in controlling the product chain length for the type-III GGPPs, since H139A formed larger products than C 20 (22).

Reaction Kinetics and Final Products of Different
Mutants-To test the above hypothesis, we replaced these large residues Tyr 107 , Phe 108 , and His 139 with Ala and examined the chain lengths of their products. Less than 5-fold changes of enzyme k cat values for the mutants as compared with the wild type (see Table 2) were observed. However, the FPP K m of Y107A/F108A/H139A was decreased 9-fold, indicating that this triple mutant has higher FPP affinity. None of these single mutations or the mutations on other sites has caused significantly reduced FPP K m . The increase of IPP K m values by 120, 65, 22, 340, and 170-fold compared with the wild type were observed in the mutants of Y107A/ F108A/H139A, Y107A/F108A/L135A/H139A, L135A/H139A, Y107A/ H139A, and F108A/H139A, respectively, indicating significantly lower affinity of these mutants with IPP. However, under the high concentrations of FPP (10 M) and IPP (150 M), which were used to yield the products by all of the mutants as shown below, the activities were not significantly changed (less than 5-fold). Further increase of the substrate type-I GGPPs shown in a stick-and-ball model with green color have similar spatial location as Tyr 107 and His 139 shown in red of type-III GGPPs, respectively. Trp 74 at the fifth position before the first DDXXD motif in helix D of type-I GGPPs can be predicted to be a critical residue for chain length determination. B, products synthesized by wild-type and mutant GGPPs. Y107A, F108A, and H139A generated C 40 and C 30 products, which are longer than the C 20 synthesized by the wild type. Tyr 107 plays a more important role in chain length determination. S71Y synthesizes shorter product C 15 , a single condensation between the bound GPP and IPP. Their product distributions are summarized in Table 3. C, the molecular ruler mechanism for chain length determination catalyzed by the type-III GGPPs. The final product of C 20 GGPP shown in black is shielded by Tyr 107 and His 139 at the bottom. The H139A mutation has yielded smaller space for product elongation through helix G (in purple). However, Y107A provides more space to accumulate longer product (C 40 ) along helix F (in green). The substitution of Ser 71 , the fourth amino acid prior to the first DDXXD motif, with a large residue of Tyr caused the formation of C 15 . concentrations (up to 20 M FPP and 320 M [ 14 C]IPP) did not lead to different final products (data not shown).
Compared with the previously identified His 139 from mutagenesis study (22), Tyr 107 plays a more dominant role in determining product chain length. According to our crystal structure, the side chains of Tyr 107 and His 139 point into the tunnel interior from opposite ␣-helixes (Fig. 4A, top, for the side view). From the top view (bottom of Fig. 4A), Tyr 107 and His 139 are well positioned to act as the "double floor" in the hollow tunnel, with Tyr 107 occupying the major part.
Ser 71 (the fourth amino acid before the first DDXXD in type-III GGPPs) is found in the upper portion of the active site crevice. When this residue was substituted with a larger residue of Tyr, S71Y GGPPs became an FPPs to produce C 15 FPP as a final product when using C 10 The wild type (1 and 0.2 mg/ml) was eluted as a dimer, but ⌬(1-9) and ⌬(1-17) GGPPs (1 mg/ml) were both eluted later, corresponding to a monomer. Only the wild type contained an extra five amino acids in the N terminus for tag cleavage (see "Experimental Procedures"). mAU, 10 Ϫ3 ϫ absorbance unit.

TABLE 2 Kinetic parameters of wild-type and mutant S. cerevisiae GGPPs
The steady-state kinetics of wild-type and mutant GGPPs were assayed with FPP and ͓ 14 C͔IPP, and S71Y was assayed with GPP and ͓ 14 C͔IPP at pH 7.5 and 25°C.

TABLE 3 Product distribution of the wild-type and mutant type-III GGPPs from S. cerevisiae
The enzyme-catalyzed condensation reactions were performed with 10 M FPP and 150 M ͓ 14 C͔IPP at pH 7.5 and 25°C. The radioactivity of each product was normalized by the number of ͓ 14 C͔ IPP incorporated.  GPP as the allylic substrate to react with one IPP (Fig. 4B, right). Under the same reaction conditions, wild-type enzyme catalyzes C 20 GGPP formation. This is consistent with the fourth amino acid before the first conserved DDXXD is at the correct position to shield the product of FPPs. The Molecular Ruler Mechanism-As shown in the superimposed structures (Fig. 4A), Tyr 107 and His 139 of type-III GGPPs in fact reside at the almost identical spatial positions as Trp 74 (the fifth amino acid before the first DDXXD) and Ala 143 of type-I GGPPs, respectively. It can be expected that Trp 74 in type-I GGPPs controls the product chain length, although this has not yet been experimentally proven. This supports our hypothesis that Tyr 107 is the most critical amino acid in determining the product chain length of type-III GGPPs, since Ala 143 in the type-I GGPPs is a small amino acid that could not form a sufficient blockage, and the large Trp 74 (corresponding to Tyr 107 in type-III GGPPs) should play a dominant role in chain length determination. However, it is the "double-floor" by Tyr 107 and His 139 for the type-III GGPPs but a "single floor" for the type-I GGPPs. According to our data, a molecular ruler mechanism for regulation of product chain lengths of type-III GGPPs is proposed in Fig. 4C. When the large side chain of Tyr 107 is removed, the chain length of Y107A reaches C 40 at the position of Phe 108 . By the double replacement of Tyr 107 and Phe 108 with small Ala, C 50 product blocked by Val 114 is generated. This chain elongation pathway is shown in green (Fig. 4C). On the other way of chain elongation shown in purple, a C 30 product blocked by Leu 135 is formed when His 139 is substituted with Ala. The mutants with removal of both His 139 and Leu 135 side chains still generated C 30 , although C 40 can be reached, probably due to the partial blockage by Tyr 107 and Phe 131 . On the other hand, mutation S71Y (the fourth amino acid prior to the first DDXXD motif of type-III GGPPs) created a blockage at the upper part of the active site crevice, and the mutant produced shorter C 15 FPP as the final product. Our data also suggest that the fourth amino acid before the first DDXXD is critical for the product chain length of C 15 -FPPs, and the fifth amino acid is important for type-I C 20 -GGPPs.

S. cerevisiae GGPPs
The Function of N-terminal Helix A in Dimer Formation-As shown in Fig. 3A, an unexpected feature in the crystal structure of type-III GGPPs is the orientation of the first N-terminal 17 amino acids, which protrude into and hold the other subunit. In order to elucidate the possible function of the N-terminal residues in dimer formation, the truncated ⌬(1-9) and ⌬(1-17) GGPPs, by deleting the first 9 amino acids (helix A) and 17 amino acids (helix A plus a loop), were constructed. According to size exclusion chromatography (Fig. 5), both of the truncated proteins were a monomer at 1 mg/ml, whereas the wildtype GGPPs formed a dimer at 1 mg/ml and lower protein concentration (0.2 mg/ml). ⌬(1-9) showed a 325-fold lower activity compared with the wild type, but ⌬(1-17) had no detectable activity. The CD spectra of wild type, ⌬(1-9), and ⌬(1-17) were similar (data not shown), indicating the deletion did not change the secondary structures. A subtle change in active site conformation must occur to account for the activity reduction for the truncated mutants.

DISCUSSION
In this study, we solved the three-dimensional structure of type-III GGPPs from S. cerevisiae. This enzyme shares mostly the same folding as the available structures of avian FPPs (8), Thermus thermophilus type-I GGPPs, 4 S. solfataricus HexPPs (10), and T. maritima OPPs (9), although it shares only 22, 23, 22, and 25% sequence identity and 39, 40, 45, and 45% similarity, respectively, with those prenyltransferases. Compared with the type-I GGPPs, the superposition is better in helices D, E, F, G, H, I, K, and N with root mean square deviation ϭ 1.53 Å for the fitting of 648 C␣ atoms of the dimeric type-III GGPPs. The most different feature in the crystal structures of type-III GGPPs is the location of the N-terminal helix A, which embraces the other subunit. Deletion of the first nine amino acids (the helical region) is sufficient to dissociate the dimer into monomer, supporting the importance of the N-terminal helix (helix A) in dimer formation. For other trans-prenyltransferases, however, their N-terminal helix is located within its own subunit, and the dimer is formed mainly through the interfacial interaction from helices E and F (e.g. OPPs and HexPPs) without the contribution from this N-terminal helix. Our data also indicate the monomer is remarkably less active, although each monomer contains a separate active site. To our knowledge, this is the first case in which a dimeric prenyltransferase can be disrupted into a monomer by a mutation (a deletion here). The structure of the monomer will be further studied.
It was mysterious that unlike the other short-chain trans-prenyltransferases, the fourth and the fifth amino acids prior to the first DDXXD motif (the FPP binding site) of type-III GGPPs are not large amino acids. On the basis of crystal structure, we have replaced some large amino acids at the bottom of the active site crevice to test their role in chain length determination. Some mutants show different kinetic parameters (smaller FPP K m and larger IPP K m values) compared with the wild type. This was also observed for the mutants of FPPs that became GGPPs by random mutagenesis (18). The calculation of kinetic parameters followed our previous reports on the kinetics of prenyltransferases (32,33). Although the prenyltransferases and their mutants produced various products, all of the products can be extracted with 1-butanol. We also demonstrated that each IPP condensation step has a similar rate constant (32,33), so the k cat could represent the IPP condensation rate constant. Here the k cat values of the GGPP mutants, which made multiple products, were calculated from the initial rates of the substrate depletion and should represent the steady-state activity for forming their favorite elongated products due to the removal of the blockage by replacing the large amino acids at the bottom of active crevice with a small Ala. The final products of the mutants described here were also generated under sufficient quantities of FPP and [ 14 C]IPP.
From the structure and mutagenesis analysis presented above, we discovered an important residue, Tyr 107 , in addition to the previously identified His 139 (22), which seals the bottom of the active site crevice to control the product chain length. The substitution of the large Tyr 107 and His 139 with a smaller Ala removes the "floor" of the crevice, thereby allowing for the formation of longer chain length products. This kind of "floor" has been found in other trans-prenyltransferases and also the protein prenyltransferases, which catalyze the transfer of farnesyl or geranylgeranyl group into the Cys of the CAAX motif (where A is an aliphatic amino acid) of the protein substrates, indicating a common mechanism of molecular ruler (34). However, distinct from other prenyltransferases, the "floor" in type-III GGPPs is a "double floor" constituted of two large amino acids. As illustrated in Fig. 6A, the "floor" is located at helix D for FPPs and type-I GGPPs but in helix F and G for type-III GGPPs and in helix G for OPPs and HexPPs. The key residues for chain length determination are Phe 113 for C 15 -avian FPPs, Trp 74 for C 20 -T. thermophilus type-I GGPPs, Tyr 107 and His 139 for C 20 -S. cerevisiae type-III GGPPs, Leu 164 for C 30 -S. solfataricus HexPPs, and Phe 132 for C 40 -T. maritima OPPs (Fig. 6A). According to these known threedimensional structures, the distances between the first Asp of the first DDXXD motif and Phe 113 , Trp 74 , Tyr 107 /His 139 , Leu 164 , and Phe 132 for FPPs, Type-I GGPPs, Type-III GGPPs, HexPPs, and OPPs, respectively, are increasing with the longer product chain lengths (Fig. 6B, left). The areas inside the circled cavity of these prenyltransferases are also larger with the increasing size of the ultimate products catalyzed by these enzymes (Fig. 6B, right).
Type-I GGPPs utilizes Trp 74 at helix D, and type-III GGPPs uses Tyr 107 and His 139 at different helices, F and G, to control the product chain length although both enzymes make the same product. Based on the different strategies in sealing the active site, our data suggest that type-I GGPPs may evolve from FPPs by shifting the large amino acid from the fourth to fifth position before the first DDXXD to shield the product, whereas type-III GGPPs and medium-chain and long-chain prenyltransferases may evolve from the common ancestor. Together, the data significantly extend our understanding on the biosynthesis of short-chain polyprenyl molecules.