Interaction with the small subunit of geranyl diphosphate synthase modifies the chain length specificity of geranylgeranyl diphosphate synthase to produce geranyl diphosphate.

Geranyl diphosphate synthase belongs to a subgroup of prenyltransferases, including farnesyl diphosphate synthase and geranylgeranyl diphosphate synthase, that catalyzes the specific formation, from C(5) units, of the respective C(10), C(15), and C(20) precursors of monoterpenes, sesquiterpenes, and diterpenes. Unlike farnesyl diphosphate synthase and geranylgeranyl diphosphate synthase, which are homodimers, geranyl diphosphate synthase from Mentha is a heterotetramer in which the large subunit shares functional motifs and a high level of amino acid sequence identity (56-75%) with geranylgeranyl diphosphate synthases of plant origin. The small subunit, however, shares little sequence identity with other isoprenyl diphosphate synthases; yet it is absolutely required for geranyl diphosphate synthase catalysis. Coexpression in Escherichia coli of the Mentha geranyl diphosphate synthase small subunit with the phylogenetically distant geranylgeranyl diphosphate synthases from Taxus canadensis and Abies grandis yielded a functional hybrid heterodimer that generated geranyl diphosphate as product in each case. These results indicate that the geranyl diphosphate synthase small subunit is capable of modifying the chain length specificity of geranylgeranyl diphosphate synthase (but not, apparently, farnesyl diphosphate synthase) to favor the production of C(10) chains. Comparison of the kinetic behavior of the parent prenyltransferases with that of the hybrid enzyme revealed that the hybrid possesses characteristics of both geranyl diphosphate synthase and geranylgeranyl diphosphate synthase.

A subgroup of isoprenyl diphosphate synthases, referred to as the "short-chain prenyltransferases," consists of geranyl diphosphate (GPP 1 ; C 10 ) synthase, farnesyl diphosphate (FPP; C 15 ) synthase, and geranylgeranyl diphosphate (GGPP; C 20 ) synthase. These enzymes provide the acyclic branch point intermediates for the biosynthesis of numerous terpenoids, including monoterpenes, sesquiterpenes, diterpenes, triterpenes, tetraterpenes, and polyterpenes such as natural rubber. GGPP synthase and FPP synthase occur nearly ubiquitously in plants, animals, and bacteria (1). GPP synthase appears to be of much more limited distribution in nature, having been identified most frequently in essential oil (monoterpene)-producing plants (2). Most isoprenyl diphosphate synthases, including the short-chain prenyltransferases, catalyze the divalent metal iondependent 1Ј-4 condensation of isopentenyl diphosphate (IPP) with an allylic prenyl diphosphate cosubstrate (3), and they are distinguished by the specific chain length and double bond geometry at C2-C3 of the prenyl diphosphate product generated (Fig. 1). Thus, GPP synthase catalyzes a single condensation of IPP with dimethylallyl diphosphate (DMAPP) to form, specifically, GPP (C 10 ). FPP synthase and GGPP synthase catalyze sequential condensations of IPP with an allylic primer (i.e. DMAPP, GPP, or FPP, as appropriate) to form the respective C 15 and C 20 elongation products. Reaction parameters, such as substrate concentration (4,5) and metal ion cofactor (6), are known to modify chain length specificity of some prenyltransferases. Evaluation of the crystal structure of homodimeric avian FPP synthase (7), coupled with prenyltransferase sequence alignment data, has led to the directed mutagenesis of FPP synthase to alter the contour of the active site, thereby generating mutant enzymes that synthesize GGPP (8) or GPP (9,10). A random chemical mutagenesis approach has also provided an altered FPP synthase capable of producing GGPP (11) and has yielded homodimeric GGPP synthase mutants capable of producing FPP (5) and polyprenols greater than C 20 (12). This work eventually led to the direct mutagenesis of an archaeal GGPP synthase so as to produce FPP by the modified enzyme (13). These experiments clearly indicate that amino acid substitutions to alter the contour of the active site may either restrict or allow sequential elongation of the polyprenyl chain.
cDNAs encoding the large and small subunits of GPP synthase were isolated from a peppermint (Mentha piperita) oil gland library and were confirmed by functional coexpression of the heteromeric enzyme (14). Thus, GPP synthase, in which both subunits are absolutely required for prenyltransferase activity, is unlike both FPP synthases and GGPP synthases that are functional homodimers (1). The GPP synthase large subunit (GPPS.lsu) shows a high level of deduced amino acid sequence identity (56 -75%) with GGPP synthases of plant origin and a lower level of deduced amino acid identity (21-37%) with FPP synthases of plant origin. The large subunit sequence also contains two highly conserved aspartate-rich motifs found in other prenyltransferases (Fig. 2). Numerous lines of evidence have indicated these aspartate-rich clusters to be involved in substrate binding and product determination (15,16) by forming a salt bridge to the divalent metal ions that are coordinated to the diphosphate group of the reacting ester (8). These aspartate-rich domains are noticeably absent in the GPP synthase small subunit (GPPS.ssu), which shares only 24 -29% deduced amino acid sequence identity with plant GGPP synthases and shows no significant homology with FPP synthases. Because the small subunit lacks the aspartate-rich functional motifs and is alone inactive, it was hypothesized that the small subunit may bind to and modify the otherwise inactive large subunit to promote transfer catalysis while restricting chain length elongation specifically to the C 10 product. If the GPPS.ssu plays such a role, then it also seemed possible that the small subunit might be capable of influencing the chain length specificity of other types of prenyltransferases. To test this possibility, a His 8 -tagged version of the small subunit (GPPS.ssu.his) was coexpressed in Escherichia coli with FPP synthase or GGPP synthase to permit affinity-based purification of any resulting chimeric species. Product analysis and kinetic evaluation of the resulting hybrid heterodimers, and comparison to the parent homodimeric transferases, demonstrated that the GPPS small subunit modifies the specificity of GGPP synthase by promoting the kinetically competent production of C 10 chains.

EXPERIMENTAL PROCEDURES
Substrates and Reagents-[4-14 C]IPP (54 Ci/mol) was purchased from PerkinElmer Life Sciences. Unlabeled IPP, DMAPP, GPP, and FPP were purchased from Echelon Research Laboratories (Salt Lake City, UT). Authentic terpenol standards were from our own collection. Restriction enzymes were purchased from New England Biolabs. T4 ligase, PfuTurbo DNA polymerase, and CodonPlus E. coli competent cells were purchased from Stratagene. Synthesis of oligonucleotide primers was performed by Invitrogen. pET vectors were from Novagen. Alkaline phosphatase, apyrase, and protein molecular weight standards were purchased from Sigma.
General Procedures-Standard molecular biology protocols were followed (17). Protein concentrations were determined by UV absorbance using molecular weights and extinction coefficients calculated from the PeptideSort program (18). Proteins were analyzed by SDS-PAGE (19), followed by staining with Coomassie Brilliant Blue R-250 (20). Antibody preparation and immunoblotting protocols were described earlier (14), as were radio-gas chromatography methods (21).
Generation of Plasmid Constructs-A truncated version of the GPP synthase small subunit (designated GPPS.ssu), in which the plastidial transit peptide was deleted, was prepared from the original pS-BET13.18 cDNA clone (14) for transfer into pET-37b using forward primer (5Ј-CC GCC GCC CAT ATG CAG CCG-3Ј) to create an NdeI site and code for a starting methionine (in place of amino acid Ser-48 of the original protein) and reverse primer (5Ј-G GAT CCG AAT AAG CTT CTA AGC CG-3Ј) to generate a HindIII restriction site downstream of the stop codon. The resulting amplicon was digested with NdeI and HindIII, gel-purified, and directionally ligated into the NdeI/HindIIIdigested pET-37b vector to yield pETGPPS.ssu. Mutation of the stop codon, to permit translation through the carboxyl-terminal His 8 -tag provided on the pET-37b vector, was accomplished by PCR amplification using pETGPPS.ssu as template and forward primer (5Ј-GGA GAT ATA CAT ATG CAG CCG-3Ј) and reverse primer (5Ј-CC CGC AAG CTT CCC AGC CGC G-3Ј), thereby resulting in a lysine substitution for the stop codon six residues from the His 8 -tag. The resulting amplicon (designated GPPS.ssu.his) was digested and gel-purified as before and directionally ligated into pET-37b that had been digested with NdeI and HindIII.
A similarly truncated version of the GPP synthase large subunit was prepared from the original pMp23.10 cDNA clone (14) using forward primer (5Ј-GCG CCT TCG ACT TCG CAT ATG TTC GAT TTC GAC GG-3Ј) to create an NdeI site and code for a starting methionine (in place of amino acid Ala-83 of the original protein) and reverse primer (5Ј-CAC TAT AGG GCG AAT TGG GAT CCG GGC CCC CCC TCG AG-3Ј) to convert the downstream KpnI site (GGTACC) in the original pBluescript SK Ϫ version (14) into a BamHI site (GGATCC). The amplified sequence (designated GPPS.lsu) was digested with NdeI and BamHI, and the gel-purified fragment was ligated into pET-32a that had been similarly digested to yield pETGPPS.lsu.
A truncated version of the Taxus canadensis GGPP synthase, from which the amino-terminal plastid targeting peptide was also deleted, was prepared from the original full-length cDNA clone (22) using forward primer (5Ј-CC CGA AGA CAT ATG TTT GAT TTC AAC G-3Ј), to install an NdeI site and thus mutate Glu-98 to Met-1, and reverse primer (5Ј-CTA GCC CGG TCG ACC TCA GTT TTG CCT GAA TGC-3Ј), to create a SalI restriction site downstream of the stop codon. The amplified sequence (designated GGPPS) was digested with NdeI and SalI, gel-purified, and ligated into the pET-32c vector that had been identically digested to yield pETGGPPS. A truncated version of the Abies grandis GGPP synthase was similarly prepared.
An FPP synthase cDNA clone was prepared from a full-length sequence located in a peppermint (M. piperita) oil gland expressed sequence tag library (23) using primer (5Ј-GGG TGA TTA CAT ATG GCG AAT C-3Ј) to create an NdeI restriction site at the starting methionine and reverse primer (5Ј-GTA TTT CAA AGC TCG AGT TTA TTT CTG C-3Ј) to create an XhoI site downstream of the stop codon. The resulting amplicon was digested with NdeI and XhoI, gel-purified, and ligated into similarly digested pET-32b to yield pETFPPS.
Transformations and Expression in E. coli-Cotransformations with two plasmids (e.g. both GPP synthase subunits and pETGPPS.ssu.his with pETFPPS or pETGGPPS) were performed in a single transformation event. Positive transformants were screened for multiple resistance with kanamycin (pETGPPS.ssu and pETGPPS.ssu.his), carbenicillin (pETGPPS.lsu, pETGGPPS, and pETFPPS), and chloramphenicol (conferred by the pACYC-based plasmid of the BL21-Codon Plus host cells that also contains extra copies of the argU, ileY, and leuW tRNA genes). Transformants were initially grown in 5 ml of Luria-Bertani medium and then transferred to 1 liter of the same medium and grown at 20°C until A 600 reached 0.6. The temperature of the culture was then lowered to 15°C prior to induction with 0.4 mM isopropyl-1-thio-␤-Dgalactopyranoside, and incubation was continued for an additional 16 h.
Protein Purification-Following induction and incubation, the transformed cells were pelleted by centrifugation (20 min at 2500 ϫ g) and resuspended in either 20 ml of buffer A (see buffer description below) or in 10 ml of His-tagged lysis buffer (50 mM NaH 2 PO 4 (pH 8.0), 300 mM NaCl, 1 mM dithiothreitol (DTT), 1 mM benzamidine, and 10 mM histidine). The suspended cells were disrupted by brief sonication (VirSonic, 100% power, two 30-s bursts at 4°C), and the homogenate was centrifuged at 12,000 ϫ g (30 min) to pellet debris and then at 195,000 ϫ g (1.5 h) to provide the soluble enzyme fraction that was filtered through a cellulose acetate membrane (Nalgene, 0.2 m).
Crude recombinant GPP synthase was purified by fast protein liquid chromatography (Amersham Biosciences) using an HR 10/10 column (Amersham Biosciences) containing Poros PI anion-exchange medium (PerSeptive Biosystems) that was previously equilibrated with buffer A (25 mM Mopso (pH 7.2), 10% (v/v) glycerol, 1 mM DTT and 1 mM benzamidine) and eluted with a 30-ml linear gradient (0 -50%) with buffer B (buffer A containing 2 M NaCl). Recombinant GPP synthase eluted at 200 -300 mM NaCl, and the pooled fractions were adjusted to 2 M NaCl and applied to an HR 10/10 column containing phenyl-Sepharose (Amersham Biosciences) previously equilibrated with buffer B and eluted with a 30-ml linear gradient (0 -100%) to buffer A. GPP synthase eluted in fractions containing less than 500 mM NaCl, and the combined material was desalted on Econo-Pac10DG Columns (Bio-Rad) and applied to an HR 5/5 column containing a strong anion-exchange matrix (Source 15, Amersham Biosciences) that was previously equilibrated with buffer A. A 30-ml linear gradient (0 -50% buffer B) eluted recombinant GPP synthase at 250 -270 mM NaCl to provide the protein at Ͼ90% purity as determined by SDS-PAGE. Recombinant GPPS.ssu, GGPP synthase, and FPP synthase were similarly purified.
For metal ion-affinity chromatography of the His 8 -tag recombinant proteins, the initial extracts (10 ml; ϳ150 mg of protein) were combined with 0.5 ml of nickel-nitrilotriacetic acid matrix (Qiagen), mixed on a rotary shaker for 1 h at 4°C, and then poured into a 10-ml Poly-Prep column (Bio-Rad) to gravity drain. The nickel-nitrilotriacetic acid matrix was then washed with 16 column volumes of 50 mM NaH 2 PO 4 buffer (pH 8.0) containing 300 mM NaCl and 20 mM histidine, followed by elution with 2 ml of the same buffer containing 250 mM histidine. The protein fraction eluted with 250 mM histidine was desalted (Bio-Rad Econo-Pac10DG) into buffer A, loaded onto a HR 5/5 column containing Source 15 strong anion-exchange matrix that was equilibrated with buffer A, and then eluted with a 10-ml linear gradient (0 -50%) of buffer B.
Sedimentation equilibrium experiments were performed at 4°C on a Beckman Optima XL-A analytical ultracentrifuge equipped with electronic speed control and photoelectronic scanning. GPP synthase (purity Ͼ90%) in 25 mM NaH 2 PO 4 (pH 7.5) containing 150 mM NaCl and 5 mM MgCl 2 was loaded into a 12-mm double-sector cell containing 3-mm columns, and molecular weight was determined from the average of scans at three different velocities and three radii using two enzyme concentrations, by calculation in SEDNTERP (24) using sigma ( Ͼ95% confidence level, determined using NONLIN (25)) and a partial specific volume (0.7427 ml/g determined from amino acid sequence) calculated in SEDNTERP (24).
Prenyltransferase Assays-Assays used to determine steady-state kinetic constants and to evaluate product chain length distribution were based on the acid-lability procedure (26,27) in which the product allylic diphosphates are solvolyzed to mixtures of the corresponding cis, trans, and tertiary allylic alcohols that are partitioned from the remaining unincorporated (acid-stable) IPP by hexane extraction. For kinetic assays, 50 -200 nmol of enzyme was combined with 100 mM Mopso (pH 6.7) containing the appropriate amounts of MgCl 2 , [4-14 C]IPP, and the allylic cosubstrate (DMAPP, GPP, or FPP) in a total volume of 100 l. The mixtures were incubated for 2 min, and the linear reaction was then quenched with 100 l of a 9:1 (v/v) methanol:3 N HCl solution. The acidified mixture was then overlaid with 1 ml of hexane, shaken for 10 min to allow completion of solvolysis, and then centrifuged briefly to separate the aqueous layer from the hexane layer (containing allylic alcohols), an aliquot of which was measured by liquid scintillation counting. Kinetic data were evaluated using Enzyme Kinetics software (Trinity Software) and the Hanes-Woolf algorithms. Reported values for substrate K M are the means of three experiments; the reported K M values for MgCl 2 are the means of two experiments.
Preparative assays used for product analysis were similar to those described above with the exceptions that 400 ng of protein was employed at saturating substrate concentrations, and the reaction time was extended to 10 min followed by the addition of 1 ml of 100 mM Tris buffer (pH 9.5) containing 10 units each of bovine alkaline phosphatase and potato apyrase to hydrolyze the diphosphate esters. The reaction mixture was overlaid with pentane and allowed to incubate for 3 h at 31°C, after which the contents were vigorously mixed and centrifuged to separate phases. The extraction procedure was repeated twice with 1-ml portions of diethyl ether, and the combined organic extract was dried over Na 2 SO 4 , diluted with internal standards, and concentrated under N 2 (to 20 l) for radio-gas chromatographic analysis.

RESULTS
Expression, Purification, and Characterization of GPP Synthase-Unlike other short-chain prenyltransferases that are homodimers (1), GPP synthase is composed of two subunits, a large subunit (GPPS.lsu) that resembles GGPPS (but when expressed alone is inactive in prenyltransfer catalysis) and a small, similarly inactive, subunit (GPPS.ssu) that does not closely resemble any known prenyltransferase but appears to confer function and chain length specificity in the subunit interaction (14). To investigate whether GPPS.ssu could interact with other prenyltransferases (FPPS and GGPPS) to influence chain length distribution of the prenyl diphosphate products, it was first necessary to develop suitable expression systems and purification methods for characterizing the target proteins. Heterologous expression in E. coli of functional GPPS had been demonstrated previously (14), yet the amount of soluble recombinant protein produced was low due to the propensity to form inclusion bodies.
Both GPPS (28,29) and GGPPS (30, 31) of plant origin have been localized to plastids. Thus, these nuclear gene products are expected to be translated as preproteins bearing an aminoterminal plastidial transit peptide for directing organellar import and subsequent proteolytic processing to the mature form (32,33). Both GPPS.ssu and GPPS.lsu appear to be translated as such preproteins based on the comparison of the deduced amino-terminal sequences (14) to those of other transit peptides (34). Previous studies with other plastidial enzymes, including monoterpene and diterpene synthases (35,36) and GGPPS (22), have demonstrated higher levels of heterologous expression of soluble, more readily purified proteins when truncated to resemble the mature forms. Therefore, both the GPPS small and large subunits were truncated for expression from the pET vector. The truncation sites selected (Fig. 2) yielded proteins of a size consistent with the molecular weights of the native subunits previously determined by SDS-PAGE (14).
Coexpression in E. coli of the truncated versions (pETG-PPS.ssu and pETGPPS.lsu) demonstrated substantially improved expression over the previously employed preprotein forms and a higher net yield of recombinant enzyme when both subunits were expressed simultaneously (data not shown). Purification of the recombinant GPPS (from coexpressed pETG-PPS.ssu and pETGPPS.lsu) by the procedure described yielded greater than 5 mg of protein (Ͼ90% purity) per liter of culture, as judged by SDS-PAGE (Fig. 3). Polyclonal antibodies raised against the mature subunit versions of GPPS (14) readily detected the corresponding truncated subunit versions of the enzyme in these preparations, at a ratio of 1:1 (Fig. 4, B and C). Radio-gas chromatographic analysis of the enzymatically dephosphorylated reaction products confirmed that the recombinant, truncated GPPS produced exclusively geranyl diphosphate (Fig. 5B).
Gel permeation chromatography indicated that the size of GPP synthase was 135 Ϯ 10 kDa (Fig. 6A), thus suggesting a tetrameric protein. To confirm this result, the recombinant GPPS was also evaluated by analytical ultracentrifugation, from which a deduced size of 134 Ϯ 5 kDa was obtained (data not shown). These measurements, combined with the results of SDS-PAGE, which show equal amounts of the two subunits, indicate that recombinant GPP synthase is a tetramer composed of two small subunits (of 27 kDa each) and two large subunits (of 35 kDa each); the empirically determined molecular weight is within 9% of the calculated value. Because previous analysis of the recombinant, preprotein form of GPPS had indicated a size of 68 Ϯ 6 kDa (14), the native version of GPPS from peppermint oil glands was also analyzed. By activity assay, GPP synthase was observed in this instance to elute in column fractions corresponding to sizes of 140 (ϳ70% of total activity) and 66 kDa (ϳ30% of total activity) (Fig. 6B), and both molecular weight forms produced exclusively GPP as product (in proportion to the amount of synthase protein) as determined by radio-gas chromatographic analysis (data not shown). These results demonstrate that heterodimeric and heterotetrameric (a dimer of the heterodimer) forms of GPPS exist naturally and that both are catalytically active; the concentration dependence of the distribution of dimer and tetramer, which may underlie the earlier observation of only the dimeric form (14), is being examined. It is interesting to note that GPPS.ssu, when expressed alone, forms homodimers of 54 kDa as determined by gel permeation (Fig. 6A); the large subunit, when expressed alone, apparently does not self-associate (data not shown).
Kinetic evaluation of the recombinant, truncated version of GPPS (tetrameric form) provided an apparent K M (DMAPP) value of 54 M and a K M (MgCl 2 ) value of 2.1 mM, which are within the range of values previously reported for partially purified, native GPP synthases from other sources; however, the apparent K M (IPP) value of 26 M is 2-4 times higher than values reported for other prenyltransferases of this type (Table  I) (21,(37)(38)(39). GPPS was unable to utilize GPP or FPP as the allylic cosubstrate with IPP, consistent with the observed chain length specificity of this enzyme. A k cat value of 4.8 s Ϫ1 was determined for GPPS. This appears to be the first reported turnover rate for any GPPS and is 10-fold higher than the rate reported for the well characterized avian FPPS (9).
Expression, Purification, and Characterization of GGPP Synthase-A cDNA encoding GGPP synthase from Canadian yew (T. canadensis) was previously isolated and functionally expressed in yeast (22). This prenyltransferase was chosen to evaluate the possible interaction with the GPPS small subunit because Taxus is phylogenetically distant from Mentha, and the Taxus GGPP synthase bears the lowest level of deduced sequence identity to the GPPS large subunit than does any other functionally documented GGPPS of plant origin (57% at the amino acid level; see Fig. 2), thereby providing a challenging and most useful test of small subunit function.
Both the preprotein and truncated versions of GGPPS (Fig.  2) were expressed in E. coli from the pET vector, and protein production was monitored by SDS-PAGE. Expression of the preprotein yielded predominantly inclusion bodies, whereas the truncated cDNA provided mostly soluble protein (Fig. 3B). Purification of the truncated form of GGPPS provided up to 10 mg of protein (at Ͼ90% purity) per liter of culture (Fig. 4A). The single protein band indicated by SDS-PAGE (at 32 kDa), combined with an estimated size of 66 kDa by gel permeation chromatography (Fig. 6A), confirms that this GGPP synthase is a functional homodimer. Polyclonal antibodies raised against the Mentha GPPS large subunit detected the Taxus GGPPS (Fig. 4C). Radio-gas chromatographic analysis demonstrated that geranylgeranyl diphosphate was the exclusive product of this enzyme when DMAPP and IPP were employed as cosubstrates ( Fig. 5E; note that geranyllinalool eluting at 30 -40 min is an artifact of the rearrangement of geranylgeraniol during the analysis as observed previously (22,40)).
Kinetic evaluation of the purified Taxus GGPPS (Table I) (1, 40 -42); turnover rates for other GGPP synthases have not been reported. It is notable that k cat increases in direct proportion to the size of the allylic cosubstrate (i.e. the rate of formation of GGPP from DMAPP (involving three condensation steps) is one-third that from FPP (involving a single condensation step)). The K M values for GPP and FPP are also substantially lower than that of DMAPP, consistent with the inability to detect either GPP or FPP as free intermediates in the elongation sequence when DMAPP is used as the cosubstrate.
Expression, Purification, and Characterization of FPP Synthase-A cDNA encoding a full-length FPP synthase was previously identified in a peppermint (M. piperita) oil gland-expressed sequence tag project (23). Unlike GPP synthase and GGPP synthase, this FPP synthase is devoid of an organellar targeting sequence and, consequently, is cytosolic. Expression from the pET vector in E. coli, and purification of the recombinant protein, yielded 20 mg/liter FPP synthase at greater than 90% purity by SDS-PAGE (Fig. 4A); radio-gas chromatographic analysis confirmed that FPP was the predominant product (Fig. 5D). The enzyme was similar in properties to other FPP synthases of plant origin (21).
Expression, Purification, and Characterization of Hybrid Prenyltransferases-To determine whether the GPPS small subunit could interact with FPP synthase or GGPP synthase to alter chain length distribution of the product, a purification procedure was developed that could separate heteromultimeric forms containing the GPPS.ssu from recombinant FPP synthase and GGPP synthase homodimers. A histidine tag (His 8tag) was appended to the carboxyl terminus of GPPS.ssu (designated GPPS.ssu.his) to permit separation of heteromeric species (GPPS.ssu.his/GGPPS and GPPS.ssu.his/FPPS) from the homodimeric GGPPS and FPPS by metal ion affinity chromatography. Similar histidine tagging has been used successfully in the purification of human GGPP synthase (43). To test the binding efficiency for GPPS.ssu.his on the affinity matrix and to ensure that the carboxyl-terminal tag did not interfere with the function of GPPS.ssu, GPPS.ssu.his and GPPS.lsu were coexpressed (the resulting protein is designated GPPS.his), and the derived soluble protein fraction was applied to the affinity column. Following washing, the bound material was eluted with 250 mM histidine and was shown by SDS-PAGE to consist of two proteins in equal amounts and of sizes expected for the GPP synthase subunits (Fig. 4A). Immunoblotting demonstrated that rabbit anti-GPPS.ssu and anti-GPPS.lsu recognized the His 8 -tagged protein version and confirmed that both large and small subunits were present in the complex (Fig. 4, B  and C). These results indicated that any heteromeric subunit assemblies containing the His 8 -tagged small subunit could be purified by immobilized Ni 2ϩ -affinity chromatography. Gel FIG. 5. Qualitative radio-gas chromatographic analysis of the prenols derived by enzymatic hydrolysis of prenyl diphosphate products of the various prenyltransferases. A is the thermal conductivity detector response to authentic standards of isopentenol (peak 1), dimethylallyl alcohol (peak 2), linalool (peak 3), nerol (peak 4), geraniol (peak 5), cis-nerolidol (peak 6), trans-nerolidol (peak 7), cis-and trans-farnesol (peak 8), and geranylgeraniol (peak 9). The remaining panels are the radioactivity detector responses to the derived products generated from [ 14 C]IPP by GPPS with DMAPP (B), by GPPS.his with DMAPP (C), by FPPS with DMAPP (D), by GGPPS with DMAPP (E), and by the GGPPS/GPPS.ssu.his hybrid with DMAPP (F), GPP (G), and FPP (H). See text for description of the various prenyltransferases. For the concentrations of reactants used, see Table I; a concentration of 80 M IPP was employed in the hybrid assay (H). Equal amounts of radioactivity were injected. Because the level of residual [ 14 C]IPP (and thus enzymatically liberated [ 14 C]isopentenol) varied with the type of assay, the detector response for E, G, and H was attenuated to emphasize the longer chain products.
permeation chromatography (Fig. 6A) demonstrated that the GPPS.his species had a size of 140 Ϯ 10 kDa, thereby indicating that the appended His 8 -tag did not interfere with tetramer assembly. Kinetic evaluation (Table I) and product analysis (Fig. 5C) demonstrated that GPPS.his was comparable in catalytic properties to the original recombinant GPPS and that the carboxyl-terminal tag did not compromise GPPS.ssu function.
To determine whether GPPS.ssu.his could interact productively with the Taxus GGPP synthase, the corresponding constructs (pETGGPPS and pETGPPS.ssu.his) were coexpressed in E. coli, and the resulting soluble recombinant protein was purified by affinity chromatography as before. The similarly purified enzyme preparation from E. coli transformed with pETGGPPS alone was employed as the control. In both cases, the flow-through and wash fractions from the affinity column

TABLE I Apparent K M values and rate constants for recombinant prenyltransferases
The abbreviations used in this table are as follows: GPPS, geranyl diphosphate synthase (truncated versions of both subunits); GPPS.his, geranyl diphosphate synthase (truncated versions of both subunits with the His-tagged small subunit); GGPPS, geranylgeranyl diphosphate synthase (truncated version); GPPS.ssu.his/GGPPS, chimeric prenyltransferase composed of truncated version of GGPPS and truncated, His-tagged version of GPPS small subunit.
were shown by SDS-PAGE to contain a prominent 32-kDa band (Fig. 7, C and D) that was shown by immunoblotting with anti-GPPS.lsu to be the Taxus GGPPS (data not shown). Affinity elution with 250 mM histidine, in the case of the coexpression system, revealed the presence of equal amounts of prominent proteins at 32 and 27 kDa that were entirely absent in the control and that were identified by immunoblotting as GGPPS and GPPS.ssu.his (Fig. 4, B and C). This heteromeric protein was subsequently purified by strong anion-exchange chromatography, through which the associated GGPPS and GPPS.ssu.his remains intact (Fig. 7D). Gel permeation chromatography of the associated species provided a size of 62 kDa (Fig. 6A), indicating that, unlike GPPS, the GGPPS/GPPS.ssu. his is a heterodimer. Radio-gas chromatographic analysis demonstrated that the GGPPS/GPPS.ssu.his heterodimer produced predominantly geranyl diphosphate as product, with trace amounts of farnesyl diphosphate and geranylgeranyl diphosphate, when supplied with DMAPP as the allylic cosubstrate (Fig. 5F); however, unlike GPPS, the GGPPS/GPPS.ssu.his hybrid dimer also accepts GPP and FPP as allylic substrates to yield, respectively, FPP with lesser amounts of GGPP (Fig. 5G), and exclusively GGPP (Fig. 5H), as products. Kinetic evaluation of the hybrid dimer (GGPPS/GPPS.ssu.his; Table I) indicated that the apparent K M values for DMAPP and for IPP (with DMAPP as cosubstrate) were very similar to those of GPP synthase, and that K M values for GPP and FPP were remarkably similar to those of GGPP synthase. With GPP or FPP as cosubstrates, however, the K M values for IPP with the hybrid dimer differed from those of GGPP synthase (one lower and one higher; Table I). The overall rate of prenyltransferase activity was highest for the hybrid dimer when supplied with DMAPP as cosubstrate (i.e. for GPP synthesis) and thus opposite to that observed for the native GGPPS (Table I).
To probe further this unusual phenomenon, another phylogenetically distant GGPPS was tested. In this case, the cDNA (GenBank TM accession no. AF425235; 56% identical to the GPPS large subunit) encoding GGPPS from grand fir (A. grandis) (21) was first expressed alone, as before, purified to homogeneity, and confirmed to produce essentially only GGPP (trace level of GPP) with IPP and DMAPP as cosubstrates. The Abies GGPPS was then coexpressed with pETGPPS.ssu.his, and the resulting soluble recombinant protein was purified by affinity chromatography. Analysis of the protein eluted with 250 mM histidine, by SDS-PAGE as before, revealed the presence of equal amounts of the Abies GGPPS and GPPS.ssu.his, thus indicating an equimolar association as observed previously with the Taxus GGPPS and the Mentha GPPS small subunit. Assay of this heteromeric enzyme with IPP and DMAPP yielded essentially only GPP as product (trace of FPP), and at a rate comparable to the formation of the C 20 product by the parent GGPPS homodimer. This second example confirmed the ability of the GPPS small subunit to modify the specificity of GGPPS for production of the shorter prenyl chain.
To determine whether GPPS.ssu.his could associate with and catalytically influence FPP synthase, an experiment of identical format to that described above was performed, in which pETGPPS.ssu.his and pETFPPS were coexpressed in E. coli, and the resulting recombinant protein was separated by affinity chromatography. In this case, the purified enzyme preparation from E. coli transformed with pETFPPS alone served as the control. SDS-PAGE analyses of both preparations showed that the bulk of the FPPS (at 45 kDa) eluted in the flow-through and wash fractions from the affinity column (Fig.  7, A and B). Elution of the affinity column with 250 mM histidine yielded only traces of the recombinant FPPS (at 45 kDa) in both cases, and the bulk of the GPPS.ssu.his (at 27 kDa) in the case of the coexpression experiment; this protein, as anticipated, was absent in the control experiment in which FPPS was expressed alone (Fig. 7, A and B). Assay of each protein fraction with DMAPP plus IPP showed by radio-gas chromatographic analysis that the only product formed in all instances was FPP (data not shown). These data indicated that, although GPPS. ssu.his and FPPS were successfully coexpressed, no interaction between the two occurred that was sufficient to permit isolation of an associated species or to promote a detectable alteration in product formation by FPPS. DISCUSSION The observation that Mentha geranyl diphosphate synthase is composed of a large subunit that resembles geranylgeranyl diphosphate synthase and a small subunit that does not resemble any known prenyltransferase suggested a role for the small subunit in interacting with the prenyltransferase-like large subunit to promote catalysis while restricting the reaction to a single condensation cycle (with DMAPP and IPP) to produce specifically geranyl diphosphate. To determine whether the small subunit could interact with other prenyltransferases to influence product distribution, a strategy was developed that employed a histidine-tagged version of the GPPS small subunit to permit affinity-based purification of this subunit along with any associated protein. This approach was tested with the GGPPS from Taxus which bears the least resemblance of any extant GGPPS to the Mentha GPPS large subunit. Following coexpression, the resulting chimeric prenyltransferase was purified and shown to synthesize primarily GPP with competent kinetics. These results, which were confirmed with the GGPPS from A. grandis, clearly demonstrate that the small subunit of GPPS is capable of binding to and modifying GGPPS to promote the efficient production of GPP by compromising the subsequent elongation steps to FPP and GGPP. The physical basis of subunit interactions that underlies this catalytic alteration is unknown and is presently under investigation. Because the small subunit lacks the aspartate-rich motifs associated with prenyltransferase chemistry, a direct role in catalysis seems unlikely but rather suggests that the small subunit serves to bind to and constrict the GGPPS active site cavity, thereby restricting the ability to conduct elongation beyond the C 10 stage. Comparing hydropathy plots for GPPS.ssu, GPPS.lsu, FPPS, and GGPPS reveals only that these proteins have similar profiles, and modeling against existing crystal structures of prenyltransferases provides no guidance as to how GPPS.ssu modifies the active site of its bound partner.
Both monoterpene biosynthesis and diterpene biosynthesis are localized in plastids of higher plants (and both GPPS and GGPPS are plastid-targeted). Because diterpene biosynthesis is considered to evolutionarily predate monoterpene biosynthesis (44), it seems plausible that a means for supply of GPP for monoterpene biosynthesis could have evolved by modification of GGPP synthase involving the interaction with another protein (i.e. the small subunit). Both GPPS and GGPPS operate in the same subcellular locale and utilize the same precursor pools of IPP and DMAPP; thus, the derivation of GPP and GGPP by functionally very different enzymes may be of regulatory significance in the control of precursor flux toward monoterpene or diterpene biosynthesis. Because there are few prenyltransferases that appear capable of synthesizing multiple products of different chain lengths (1), evolution appears to disfavor this approach to controlling supply of precursors for the synthesis of the different terpenoid families. The fact that GPPS is heteromeric (both dimer and tetramer are active) is suggestive of functional regulation, although possible bases of such control by allosterism or dissociation are presently unknown. It is of interest in this connection to note that some mammalian GGPP synthases appear to possess tetrameric architecture (45,46).
A related set of experiments directed to evaluating GPPS small subunit interaction with Mentha FPPS failed to demonstrate association of the two proteins or any catalytic alteration in this case. The failure of the GPPS small subunit to interact with FPPS may also have evolutionary origins of regulatory significance. Thus, FPPS operates in the cytosol to provide precursor for sesquiterpene and triterpene biosynthesis, and there is no specific requirement for GPP production at this locale. Therefore, there is no functional utility for the interaction of the GPPS small subunit and FPPS, and any interaction of the two would appear to be counterproductive during the period of transit of the small subunit from cytosol to plastid.
In addition to the Mentha GPPS, two bacterial prenyltransferases, hexaprenyl-diphosphate synthase of Micrococcus luteus (47) and heptaprenyl-diphosphate synthase of Bacillus subtilis (48), possess heteromeric structures consisting of a large subunit that resembles a prenyltransferase, including the catalytically functional motifs, and a small subunit that shares little sequence identity to prenyltransferases. The three small subunits of these prenyltransferases, however, share little sequence identity with each other (Ͻ20% identity), suggesting that, although these heteromeric enzymes are conceptually similar in design, the "effector" small subunits have different evolutionary origins.
Recently, Bouvier et al. (49) have isolated a cDNA from Arabidopsis thaliana that resembles a prenyltransferase and that when expressed in E. coli generates GPP from DMAPP and IPP. This enzyme is seemingly plastid-directed, but the kinetic competence and subunit architecture are unknown; the enzyme is certainly not heteromeric. Because Arabidopsis does not accumulate appreciable amounts of monoterpenes, the physiological significance of this GPPS is uncertain; however, the presence of this prenyltransferase along with a recently discovered monoterpene synthase in this species (50) suggests the operation of a monoterpene-based signaling system as a response to pathogens or herbivore attack, as found in other plants (51,52). Further studies with essential oil-producing plants and other (non-producing) species should allow determination of the significance of these two fundamentally different types of geranyl diphosphate synthases.