Heterodimeric geranyl(geranyl)diphosphate synthase from hop (Humulus lupulus) and the evolution of monoterpene biosynthesis

Abstract

Myrcene, which accounts for 30–50% of the essential oil in hop (Humulus lupulus L.) trichomes, derives from geranyl diphosphate (GPP), the common precursor of monoterpenes. Full-length sequences of heterodimeric GPP synthase small subunit (GPPS.SSU, belonging to the SSU I subfamily) and large subunit (LSU) cDNAs were mined from a hop trichome cDNA library. The SSU was inactive, whereas the LSU produced GPP, farnesyl diphosphate, and geranylgeranyl diphosphate (GGPP) from dimethylallyl diphosphate and isopentenyl diphosphate in vitro. Coexpression of both subunits in Escherichia coli yielded a heterodimeric enzyme exhibiting altered ratios of GPP and GGPP synthase activities and greatly enhanced catalytic efficiency. Transcript analysis suggested that the heterodimeric geranyl(geranyl)diphosphate synthase [G(G)PPS] is involved in myrcene biosynthesis in hop trichomes. The critical role of the conserved CxxxC motif (where “x” can be any hydrophobic amino acid residue) in physical interactions between the 2 subunits was demonstrated by using site-directed mutagenesis, and this motif was used in informatic searches to reveal a previously undescribed SSU subfamily (SSU II) present in both angiosperms and gymnosperms. The evolution and physiological roles of SSUs are discussed.

Monoterpenes represent a large family of plant natural products. They can be released from vegetative tissues to serve directly as toxic agents against herbivores or pathogens, or they can indirectly protect plants by attracting predators of attacking herbivores (1–3). Monoterpenes are also emitted from floral tissues to attract pollinators (4).

Geranyl diphosphate (GPP), the common precursor for monoterpenes, is formed by head-to-tail condensation of 1 molecule of isopentenyl diphosphate (IPP) and 1 molecule of dimethylallyl diphosphate (DMAPP) via the action of GPP synthase (GPPS; EC 2.5.1.1) in plastids of plant cells. Heterodimeric GPPSs have been described from angiosperms (5, 6), and homodimeric GPPSs have been functionally characterized from both angiosperms (7, 8) and gymnosperms (9, 10). Both subunits of homodimeric GPPSs and the large subunit of heterodimeric GPPSs (GPPS.LSU) contain 2 aspartate-rich motifs, DD(X)_2–4D (where “X” is any amino acid), which are important in prenyl–substrate binding (11). However, the small subunit of heterodimeric GPPS (GPPS.SSU) lacks this motif and is catalytically inactive alone. The LSU of Mentha × piperita GPPS is also inactive alone, whereas that from Antirrhinum majus can convert DMAPP and IPP to geranylgeranyl diphosphate (GGPP) in vitro (5, 6). In both cases, the inactive SSU functions as a “modifier” to change the chain length of the product of the LSU from GGPP(C₂₀) to GPP(C₁₀). However, the effect of SSU on the catalytic efficiency of an active LSU, and the mechanism of physical interaction between the 2 subunits of heterodimeric GPPSs have yet to be investigated.

Mentha SSU can bind to the phylogenetically distant geranylgeranyl diphosphate synthases (GGPPSs) from the gymnosperms Taxus canadensis and Abies grandis to produce GPP as the main product (12), and it has been suggested that both heterodimeric GPPS subunits might have evolved from a GGPPS based on the fact that diterpene biosynthesis evolutionarily predates monoterpene biosynthesis (13). However, failure to clone SSU I subfamily homologs from A. grandis or Picea abies via homology-based strategies (8, 9) led to the assumption that heterodimeric GPPS might exist only in angiosperm species.

Plant monoterpenes are derived from the plastid-localized methylerythritol phosphate pathway (14, 15), and monoterpene synthases of plant origin have an N-terminal plastid-localization signal peptide (16). On the basis of colocalization, it has been assumed that the homodimeric GPPS is responsible for the formation of GPP as the precursor of monoterpenes in Arabidopsis, whereas this function has been ascribed to the heterodimeric GPPS in M. × piperita and A. majus. Tissue-specific, developmental, and rhythmic changes in the mRNA and protein levels of GPPS.SSU, but not LSU, correlate with monoterpene emission (6), suggesting that SSU might play a key role in the regulation of monoterpene biosynthesis in A. majus flowers.

Cultivated hop (Humulus lupulus) is a dioecious perennial vine, and female flowers (commonly called cones) are used to add flavor during the brewing of beer. The monoterpene myrcene accounts for 30–50% of the essential oil, depending on variety (17). Myrcene is synthesized and stored exclusively in lupulin glands (trichomes on the inner surface of the bracts of the female flowers). The tissue-specific and developmental expression profiles of a hop monoterpene synthase suggested a primary role in myrcene formation (18).

We report here the functional expression of heterodimeric GPPS large and small subunits from hop. The hop SSU possesses no prenyltransferase activity alone, but it alters both the kinetics and product specificity of the LSU. Expression of the SSU, but not the LSU, is directly correlated with myrcene production. The CxxxC motif (where “x” can be a hydrophobic amino acid, such as alanine, leucine, isoleucine, valine, glycine, or serine) found in both subunits of the hop heterodimeric geranyl(geranyl)diphosphate synthase [G(G)PPS] is crucial for physical interactions between the 2 subunits. Finally, biochemical and phylogenetic analyses revealed the existence of a previously undescribed class of heterodimeric G(G)PPS.SSUs with broad distribution in both angiosperms and gymnosperms.

Results and Discussion

Isolation and Analysis of Hop GPPS.SSU and GPPS.LSU cDNA Clones.

Searching the previously described hop trichome EST database (18) consisting of 3,619 unigenes led to the identification of a unigene (represented by a single EST) that encoded a protein with 39.4–44.8% sequence identity to the GPPS.SSUs from A. majus (6), M. × piperita (5), and Clarkia breweri, respectively (Fig. S1A). The full-length hop cDNA (1,307 bp) encoded a peptide of 279 aa with a calculated molecular mass of 30,708 Da and a pI of 6.27. The hop GPPS.SSU-like protein did not contain the DD(X)_2–4D motif, and the predicted score for plastid localization (http://www.cbs.dtu.dk/services/TargetP) was only 0.414.

Another hop trichome unigene (assembled from 2 ESTs to give a 1,342-bp full-length cDNA) encoded a protein of 369 aa (M_r, 40,088; pI, 5.63) with 67.5% and 65.9% amino acid sequence identity to GPPS.LSUs from A. majus (6) and M. × piperita (5), respectively (Fig. S1B). Like the functionally characterized GPPS.LSUs from A. majus and M. × piperita, the N terminus of the hop GPPS.LSU-like protein had the characteristic features of a plastid-targeting peptide, with a predicted score for plastid localization of 0.807.

Neither the hop SSU nor the LSU genomic sequences contained introns. DNA gel blot analysis identified a single copy of the LSU and 2 copies of the SSU in the hop genome (Fig. S2).

Biochemical Characterization of Hop Heterodimeric G(G)PPS.

The hop SSU and LSU genes were N-terminally truncated (Fig. S1A) to aid solubility (9, 12) and were subcloned into appropriate vectors (see Materials and Methods, SI Materials and Methods, and Table S1), followed by expression in appropriate Escherichia coli strains for analysis of the biochemical properties of the recombinant proteins.

The purified SSU protein (expressed in the pMAL-C2X vector) was inactive when assayed with [1-¹⁴C]-IPP and any of the 3 potential cosubstrates DMAPP, GPP, or farnesyl diphosphate (FPP; Fig. 1A). The purified His-tagged LSU protein showed GGPPS activity in assays using DMAPP and [1-¹⁴C]-IPP as substrates. GPP (26.9%; each molecule contains one ¹⁴C-labeled carbon from [1-¹⁴C]-IPP) and FPP (4.9%; two ¹⁴C-labeled carbons) were also produced, although GGPP (68.2%; three ¹⁴C-labeled carbons) was the major product (Fig. 1A). The LSU protein could also recognize GPP or FPP as cosubstrates with [1-¹⁴C]-IPP to produce FPP (14.0%)/GGPP or GGPP alone, respectively (Fig. 1A).

Fig. 1.

In vitro assay of hop heterodimeric G(G)PPS with allylic substrates and [1-¹⁴C]-IPP. (A) Product profiles. Reaction products were hydrolyzed to the corresponding alcohols, extracted with hexane, and separated by TLC. Authentic standards (G, geranol; F, farnesol; GG, geranylgeranol) were visualized by spraying the plates with anisaldehyde/concentrated sulfuric acid/absolute ethanol (2:4:94 vol/vol/vol). The 4 groups of 4 lanes show reactions with different proportions of allylic substrates (D, dimethylallyl diphosphate; I, isopentenyl diphosphate; F, farnesyl diphosphate; GG, geranylgeranyl diphosphate). Lane 1 shows activity with SSU alone; lane 2, LSU alone, lane 3, SSU plus LSU; and lane 4, SSU plus LSU boiled control. (B) Time course analysis of products formed by hop LSU and heterodimeric GPPS. For each time, lanes 1–4 are as described above. All assays contained 1 μg of purified protein, 30 μM [1-¹⁴C]-IPP, and 33 μM DMAPP.

Although His-tagged SSU was insoluble when expressed alone in E. coli, it could be purified together with the untagged LSU by nickel affinity chromatography when coexpressed in E. coli together with the LSU. The purified heterodimeric protein was active with [1-¹⁴C]-IPP and any of the 3 cosubstrates DMAPP, GPP, or FPP (Fig. 1A). Surprisingly, GPP (59.5%) and GGPP (40.5%) were the products of the heterodimer with DMAPP as cosubstrate, and GGPP was the sole product with GPP or FPP as cosubstrate. Negligible amounts of FPP were produced in any of these reactions, consistent with the fact that there is no specific requirement for FPP production in plastids, and that FPP synthase and sequiterpene synthase are cytosolic enzymes (19, 20). In contrast, the heterodimeric GPPSs from M. × piperita and A. majus are unable to use GPP or FPP as cosubstrates with IPP, and GPP is the sole product with DMAPP as cosubstrate with IPP (5, 6). Longer incubations favored production of GGPP at the expense of GPP (Fig. 1B). Thus, the hop protein complex was designated as a heterodimeric G(G)PPS.

Kinetic analysis revealed that the heterodimer exhibited around 12-fold lower affinity for GPP and 17-fold higher affinity for FPP than the LSU alone (Table S2). The k_cat values for the heterodimeric G(G)PPS at saturating concentrations of IPP were considerably higher than for the LSU alone (Table S2). Thus, in addition to altering product specificity, the SSU promoted enzymatic efficiency of the LSU by more than an order of magnitude (Table 1).

Table 1.

Catalytic efficiency of short-chain prenyltransferase

	kcat/Km (s−1M−1)
	DMAPP	GPP	FPP
LSU	13	70	19
LSU/SSU	354	51	1,073
GGPPS11	67	338	35
GGPPS11/GGR	132	177	177
GGPPS11/SSU	356	220	650
LSU/ GGR	254	59	220

LSU and SSU are from hop, and GGPPS11 and GGR are from Arabidopsis. All kinetic parameters are determined from triplicate experiments (for more detailed information, see Table S2). IPP was cosubstrate for all reactions.

Spatial and Temporal Expression of Heterodimeric G(G)PPS.

Quantitative real-time PCR was used to examine transcript levels in different tissues and in trichomes at different developmental stages. G(G)PPS.LSU primers (Table S1) were designed to target the 5′ UTR plus the N-terminal signal peptide region (399 bp) to distinguish G(G)PPS.LSU from GGPPS genes. G(G)PPS.SSU transcripts were only significantly detected in trichomes where myrcene is synthesized (18) (Fig. 2A). G(G)PPS.LSU transcript levels were highest in trichomes from cones harvested at 4 weeks after flowering, although expression was detected in all tissues tested (Fig. 2B).

Fig. 2.

Tissue specificity of hop G(G)PPS.SSU and G(G)PPS.LSU transcript expression. (A) Quantitative RT-PCR analysis of SSU transcript levels in different hop tissues and at different developmental stages of cones and trichomes. No trichome indicates female bracteole after trichome removal. (B) As in A, but for LSU transcripts. Transcript levels are expressed relative to glyceraldehyde-3-phosphate dehydrogenase. WAF indicates weeks after flowering.

To test the cellular localization of hop SSU and LSU, the ORF of the enhanced green fluorescent protein (EGFP) gene was fused to the C termini of the SSU and LSU ORFs under control of the CaMV 35S promoter. 35S:SSU-EGFP and 35S:LSU-EGFP were introduced by particle bombardment into tobacco leaves. The LSU-EGFP fusion protein accumulated exclusively in chloroplasts of tobacco leaf epidermal cells. In contrast, the SSU-EGFP fusion protein aggregated around chloroplasts and could not be translocated into chloroplasts in tobacco guard cells (Fig. S3). A protein construct in which the first 50 aa of GPPS.SSU were fused with EGFP exhibited similar localization to the complete SSU-EGFP fusion protein (Fig. S3). It is possible that the hop SSU is targeted to nonpigmented leucoplasts (21) in hop trichomes.

The heterodimeric G(G)PPS could compete with myrcene synthase for the GPP pool in hop lupulin glands. GPP could be used for monoterpene biosynthesis, and GGPP could be used for biosynthesis of diterpenes, chlorophylls, plastoquinone, and phylloquinone (22), or be involved in protein prenylation. However, myrcene synthase is expressed at a significantly higher level than the heterodimeric GPPS subunits in hop lupulin glands (18), and it has a higher affinity for GPP than does the heterodimeric GPPS. Furthermore, there are no ESTs that hit other GPPS/GGPPS-like genes in a new database (http://trichome.noble.org/trichomedb) comprising 22,959 sequences from hop trichomes. Myrcene/(E)-β-ocimene biosynthesis and emission are regulated by both GPPS and monoterpene synthase in snapdragon (6, 23).

The CxxxC Motifs Are Critical for Interaction Between G(G)PPS.SSU and G(G)PPS.LSU.

BLAST comparisons of all available SSU, LSU, and GGPPS sequences revealed the presence of 2 conserved CxxxC motives (where “x” can be alanine, leucine, isoleucine, valine, glycine, or serine) in all SSU sequences and 1 CxxxC motif that is conserved in all LSU and GGPPS sequences (Fig. S4). To determine the role of the CxxxC motifs, we made Cys→Gly mutations in the 2 motifs of the hop SSU, either separately or together, and made similarly mutated LSUs (Cys to Gly or Ser; Fig. 3A). The SSU and LSU pairs were subcloned into the pETDuet1 expression vector for coexpression of the N-terminally His-tagged SSU and C-terminally S-tagged LSU in E. coli. This allowed us to use immunoblotting with specific anti-His-tag and S-tag antibodies to determine whether coexpressed subunits associated together. Neither epitope tag affected G(G)PPS activity or binding ability between wild-type SSU and LSU. When coexpressed with wild-type LSU in E. coli, the mutated SSUs lost the ability to interact with the LSU, and only LSU could be detected by immunoblot of soluble fractions (Fig. 3A and 3B, lanes 2–4). Likewise, mutated LSUs became mostly insoluble and lost their ability to physically interact with wild-type SSU (Fig. 3A and 3B, lanes 5 and 6). Moreover, no heterodimeric G(G)PPS could be purified by nickel affinity chromatography from any combination of coexpressed LSU and SSU in which either subunit was mutated, whereas active enzyme was recovered from the combination of wild-type SSU and LSU.

Fig. 3.

Effects of mutation of Cys residues and reducing agent on the subunit-binding ability of hop G(G)PPS.SSU and LSU. (A) Sequence comparisons of wild type (no. 1) and mutant SSU/LSU (nos. 2–6). Only the CxxxC motifs are shown, and the mutated amino acids are in italics. (B) Protein samples (2 μg of crude protein, with lane numbers corresponding to the mutants in A) were separated by SDS/PAGE, transferred onto nitrocellulose membrane, and blotted with anti-His-tag (for SSU detection) and anti-S-tag (for LSU detection) monoclonal antibodies. Lane a shows total protein of N-terminally His-tagged G(G)PPS.SSU (calculated M_r 30,230 kDa); lane b, soluble protein of C-terminally S-tagged G(G)PPS.LSU (calculated M_r 37,000 kDa). T, total protein of whole-cell extract; S, soluble supernatant separated by centrifugation at 25,000 × g for 30 min. (C) Lanes 1 and 2 show total soluble protein without and with addition of 50 mM β-mercaptoethanol during sample preparation; lanes 3–5, purified protein after Ni-NTA affinity chromatography with addition of Triton X-100 (0, 0.2%, and 0.5% for lanes 3, 4, and 5, respectively) during protein purification. Ponceau-S-stained proteins on the nitrocellulose membrane were used as loading control (Upper in B and C).

The above results suggest that disruption of disulfide bonds might affect physical interactions between the 2 subunits. However, the G(G)PPS complex could be dissociated without addition of a reducing agent in the sample preparation for gel analysis (boiling for 10 min in loading buffer containing 1% lithium dodecyl sulfate, which would not dissociate disulfide bonds), and no band with the calculated molecular mass of the hop G(G)PPS complex (67.23 kDa) could be detected by immunoblotting with either anti-His-tag or anti-S-tag antibody. Second, addition of 50 mM DTT or β-mercaptoethanol, a concentration that should dissociate all disulfide bonds (24) during purification, did not affect the interaction between the 2 subunits, and the presence of a reducing agent enhanced or stabilized GPPS activity (Fig. S5). The hop heterodimeric G(G)PPS was dissociated by detergent (0.2% and 0.5% Triton X-100) during purification in the cold, and only the inactive G(G)PPS.SSU was then recovered (Fig. 3C, lanes 4 and 5). This suggests that the subunits bind together via hydrophobic interactions (25). The CxxxC motif is necessary but not sufficient for physical interaction between the SSU and LSU.

A Previously Undescribed SSU Subfamily (SSU II) Is Widely Distributed in Seed Plant Taxa.

An Arabidopsis gene annotated as a geranygeranyl reductase (GGR; At4g38460) encodes a protein with 29.7% amino acid sequence identity to the hop SSU. This enzyme lacks GGPPS activity in vitro because of its lacking the second conserved DDxxD motif (26), and its gene is phylogenetically distinct from other GGPPS-like genes in Arabidopsis. However, the existence of 2 conserved CxxxC motifs in AtGGR and its sequence identity with all known SSUs led us to speculate that AtGGR may have a similar biochemical function to known SSUs. We found homologs of AtGGR in the genome sequences of rice and grape, and in EST collections from both angiosperms and gymnosperms (http://compbio.dfci.harvard.edu/tgi/). None of these genes, which form a phylogenetically distinct clade, has been functionally identified. We designate this previously undescribed clade of genes, protein sequences of which share from 41.3% to 93.2% amino acid identity, as SSU II genes (Fig. 4).

Fig. 4.

Phylogenetic analysis of homodimeric and heterodimeric GPPS and GGPPSs. Posterior probabilities (>0.5) generated by using PhyML 3.0 aLRT (500 replicates; left side of slash) and TNT1.1 (1,000 replicates; right side of slash) are indicated near nodes (e.g., 0.97/92). GenBank accession numbers are: A. majus_SSU AAS82859; piperita_SSU AF182827; C. breweri_SSU AAS82870; H. brasiliensis_SSU BAF98300; V. vinifera_SSU CAO38946; A. majus_LSU AAS82860; M. × piperita_LSU AAF08793; V. vinifera_GPPS CAO17862; Q. robur_GPPS CAC20852; S. lycopersicum_GPPS ABB88703; P. abies_GPPS ACA21459 (A, angiosperm; B, bryophyte; C, chlorophyte; F/P, fungi/protisa; G, gymnosperm; L, lycophyte; genes marked with star are partial protein sequences derived from EST sequences). All Arabidopsis GGPPS-like proteins with predicted plastid-localization transit peptides were included in subsequent expression analysis (arrows).

All members of the SSU II clade contain 2 CxxxC motifs, with the exception of those from spruce and pine (both annotated as GGPPS-like proteins), in which a Thr replaces the second Cys in the second motif. Comparison of primary protein sequences indicates that both SSU clades probably evolved from GGPPS (Fig. 4). Hierarchical cluster analysis of the tissue-specific expression patterns of 12 putative GGPPSs (22) and 4 functionally identified monoterpene synthases using GeneInvestigator (27) showed that AtGGR, AtGGPPS11 (At4g36810), and 1 monoterpene synthase [At4g16740, an (E)-β-ocimene synthase; ref. 28] are tightly coexpressed (Fig. S6), suggesting that AtGGR and GGPPS11 might together function as a heterodimeric GPPS involved in monoterpene biosynthesis in Arabidopsis.

To test the hypothesis that AtGGR has a similar biochemical function to functionally characterized SSUs from other plant species, we coexpressed the N-terminally truncated hop SSU or AtGGR fused with a His-tag at the C terminus with each of 4 potential LSU partners: hop G(G)PPS.LSU, AtGGPPS2 (At2g18620; a functionally uncharacterized protein with 70.9% amino acid sequence identity to AtGGPPS11), AtGGPPS6 (At3g14530; functionally uncharacterized) and AtGGPPS11. AtGGR, like hop G(G)PPS.SSU, could interact not only with Arabidopsis GGPPS2 and GGPPS11, but also with hop G(G)PPS.LSU (Fig. 5A). However, it could not interact with the GGPPS-like protein encoded by At3g14530 (AtGGPPS6).

Fig. 5.

Purification and assay of recombinant heterodimeric G(G)PPSs expressed in E. coli. (A) SDS/PAGE analyses of purified hybrid heterodimeric GPPS and GGPPS. The calculated sizes of the truncated AtSSU and hop SSU with N-terminal His tags are 32.1 and 30.5 kDa, respectively; truncated AtGGPPS11 (with His tag, 32.4 kDa; without His tag, 31.3 kDa); truncated AtGGPPS2 (with His tag, 31.9 kDa; without His tag, 30.8 kDa); truncated hop LSU (with His tag, 34.9 kDa; without His tag, 33.2 kDa). Note that the purified hop LSU showed 2 distinct bands. (B) In vitro functional assay of the above hybrid heterodimers with DMAPP and [1-¹⁴C]-IPP (see Fig. 1 legend for details). All reactions were incubated at 30 °C for 30 min except for the reactions with AtSSUII/GGPPS2 and HlSSU/GGPPS2, which were incubated for 2 h or overnight (O/N).

GGPP was the sole product of GGPS2, whereas GGPPS11 produced GGPP as the main product (94.5%; n = 2) along with small amounts of GPP (3.4%; n = 2) and FPP (2.1%; n = 2) (Fig. 5B). The enzymatic product of AtGGPPS6 was a polyprenyl diphosphate with a chain of more than 20 carbons (Fig. S7). Assays using DMAPP and [1-¹⁴C]-IPP showed that AtGGR is not an active G(G)PPS enzyme by itself. However, when coexpressed with AtGGPPS11, AtGGR and GPPS.SSU could change the product profile, with GGPP production decreasing and GPP accounting for 31.6% (n = 2) of the total product for the AtGGR/AtGGPPS11 combination and 25.3% (n = 2) of the total product for hop SSU/AtGGPPS11; no FPP was detected in either case. AtGGR also interacted with hop LSU to form an active heterodimeric G(G)PPS, with GPP accounting for 65.2% (n = 2) of the total product (Fig. 5B). Although AtGGR and hop SSU could interact with AtGGPPS2, the heterodimeric proteins had product profiles similar to that of AtGGPPS2 alone and lower GGPPS activity than AtGGPPS2 itself (<10%); GGPP was the predominant product, with a trace amount of GPP (Fig. 5B). Thus, SSU proteins might bind to specific GGPPS-like proteins to increase the GPPS catalytic efficiency, and not all GGPPS-like proteins may partner with SSUs, at least in Arabidopsis.

Purified AtGGPPS11 could accept DMAPP, GPP, and FPP as allylic substrates. Both AtGGR/AtGGPPS11 and hop SSU/AtGGPPS11 had lower affinities toward GPP and higher affinities for FPP than those of AtGGPPS11 alone. AtGGR/hop LSU had similar kinetic properties to hop heterodimeric G(G)PPS (Table 1 and Table S2). Overall, our results indicate that AtGGR and its homologs should be reclassified as class II SSUs that, like the members of the SSU I subfamily, function as not only “modifiers” to alter the product profile, but also as “accelerators” to promote the GPPS activity of the LSU, at least in hop and Arabidopsis.

Recently, Schmidt and Gershenzon (10) reported that they could not find the GPPS.SSU homologs from P. abies by PCR using primers designed from the SSU sequences described from M. × piperita (5), A. majus, and C. breweri (6). This might be because the SSU II homolog in P. abies (TC1244) has very low nucleotide identity with the SSU I sequences from M. × piperita (28.6%), A. majus (35.8%), and C. breweri (6.8%), and there is no stretch of 12 identical nucleotides or more between them.

Evolutionary Origins and in Planta Functions of Heterodimeric G(G)PPSs.

Plant-derived monoterpenes play roles in pollination, biotic defense, and seed dispersal, and the monoterpene synthases have been well-characterized in both gymnosperms and angiosperms (16, 29). However, the functional roles of plant G(G)PPSs are poorly understood. Our phylogenetic analysis shows that both SSU lineages evolved from a common ancestral gene, probably after the emergence of seed plants, because no SSU homologs from more ancient plant species were identified. It is interesting that the SSU genes from monoterpene-rich plant species fall into 1 clade (SSU I), and their expression appears to be restricted to monoterpene storage/emission organs (e.g., myrcene in the glandular trichome of the female flower in hop, ref. 18; linalool in the flower of Clarkia, ref. 30; menthol in the leaf glandular trichome of peppermint, ref. 31; and myrcene/ocimene in snapdragon flower ref. 23) and parallel expression of the corresponding monoterpene synthase gene. In contrast, based on available microarray evidence from Arabidopsis, Medicago truncatula (TC128421; ref. 32), and rice (Os02g0668100; http://bioinformatics.med.yale.edu/riceatlas), genes in the SSU II subfamily are likely to be constitutively expressed (Figs. S6 and S8). Thus, plants may have evolved the ability to produce monoterpenes by acquiring GPPS enzymes with higher efficiency (with the SSU increasing the efficiency of production of GPP for monoterpene synthesis) in specific organs, thereby gaining a selective advantage through increased ability to attract pollinators by floral scents, or to defend against potential herbivores or pathogens. Clarkia and snapdragon, both of which possess a flower-specific SSU I, cannot self-pollinate.

The wide distribution of homodimeric GPPS and GGPPS suggests functions in plant processes, such as gibberellin biosynthesis, and a recent study indicates that at least one homodimeric GPPS that was initially thought to be responsible for the formation of GPP as a precursor for monoterpene biosynthesis (7) is in fact involved in gibberellin biosynthesis (8). The exact physiological roles of homodimeric and heterodimeric GPPSs can now be elucidated by genetic gain- and loss-of-function approaches.

Materials and Methods

Plant Materials.

Rhizomes of hop (H. lumulus) cv Nugget were purchased from Northern Brewer Company and grown under greenhouse conditions as described previously (18).

Chemical Reagents.

All chemicals were purchased from Sigma–Aldrich, except for radiolabeled [1-¹⁴C]-IPP (55 mCi/mmol), which was purchased from American Radiolabeled Chemicals.

Hop Trichome Isolation, cDNA Library Construction, and EST Sequence Analysis.

Hop trichome isolation, RNA isolation from trichomes, cDNA library construction, and EST sequence analysis were performed as described previously (18).

Generation, Expression, and Assay of Recombinant G(G)PPS.

No soluble full-length or truncated GPP.SSU protein could be obtained when expressed in pET or pEXP5-CT/TOPO vectors. The GPPS.SSU insert was therefore subcloned into pMAL-C2X vector (New England Biolabs) by a PCR-mediated method using the primers SSUC2XFor and SSUC2XRev (Table S1). For coexpression of G(G)PPS.SSU and G(G)PPS.LSU, the truncated ORFs were subcloned into pET26(a) and pET32(a) vectors, respectively. Further details of cloning and expression are given in SI Materials and Methods.

GPPS assays were performed as described previously (5, 23). Product ratios using [1-¹⁴C]-IPP as cosubstrate were calculated by determining the radioactivity of product spots on TLC plates after development with benzene/ethyl acetate (9:1 vol/vol). A total of 200 ng of purified enzyme was used in each reaction to determine steady-state kinetic parameters. K_m and k_cat values were calculated from initial rate data by using the hyperbolic regression analysis method in Hyper32 software (version 1.0.0; http://homepage.ntlworld.com/john.easterby/hyper32.html).

Analysis of Protein, DNA, and RNA.

Protein and DNA gel blot analysis and quantitative real-time RT-PCR were performed as described in SI Materials and Methods.

In Vitro Mutagenesis.

The CxxxC motifs in SSU and LSU were mutated to GxxxG or SxxxS by using a previously described PCR-mediated method (33). Primers are listed in Table S1.

Subcellular Localization of G(G)PPS Small and Large Subunits.

SSU:EGFP and LSU:EGFP constructs were generated by using a PCR-mediated gene splicing by overlap extension technique (34) (see Table S1 for details of primers). Preparation of gold particles, particle bombardment into tobacco leaf tissue, and image collection by laser scanning confocal microscopy were preformed as described previously (35).

Phylogenetic Analysis of the G(G)PPS Family.

A multiple alignment of the deduced amino acid sequences of both subunits of heterodimeric G(G)PPSs, homodimeric GPPSs and GGPPSs, and a neighbor-joining phylogenetic tree were constructed by using the “A la Carte” mode [Muscle 3.7 for multiple alignment, Gblocks 0.91b for alignment refinement, PhyML 3.0 aLRT for phylogeny using maximum likelihood (500 replicates for bootstrap value) or TNT 1.1 for phylogeny using parsimony (1,000 replicates for bootstrap value), TreeDyn 198.3 for Tree rending] of the Phylogeny.fr program (36).