cis–prenyltransferases (CPTs) are predicted to be involved in the synthesis of long-chain polyisoprenoids, all with five or more isoprene (C5) units. Recently, we identified a short-chain CPT, neryl diphosphate synthase (NDPS1), in tomato (Solanum lycopersicum). Here, we searched the tomato genome and identified and characterized its entire CPT gene family, which comprises seven members (SlCPT1–7, with NDPS1 designated as SlCPT1). Six of the SlCPT genes encode proteins with N–terminal targeting sequences, which, when fused to GFP, mediated GFP transport to the plastids of Arabidopsis protoplasts. The SlCPT3–GFP fusion protein was localized to the cytosol. Enzymatic characterization of recombinant SlCPT proteins demonstrated that SlCPT6 produces Z,Z–FPP, and SlCPT2 catalyzes the formation of nerylneryl diphosphate while SlCPT4, SlCPT5 and SlCPT7 synthesize longer-chain products (C25–C55). Although no in vitro activity was demonstrated for SlCPT3, its expression in the Saccharomyces cerevisiae dolichol biosynthesis mutant (rer2) complemented the temperature-sensitive growth defect. Transcripts of SlCPT2, SlCPT4, SlCPT5 and SlCPT7 are present at low levels in multiple tissues, SlCPT6 is exclusively expressed in red fruit and roots, and SlCPT1, SlCPT3 and SlCPT7 are highly expressed in trichomes. RNAi-mediated suppression of NDPS1 led to a large decrease in β–phellandrene (which is produced from neryl diphosphate), with greater reductions achieved with the general 35S promoter compared to the trichome-specific MKS1 promoter. Phylogenetic analysis revealed CPT gene families in both eudicots and monocots, and showed that all the short-chain CPT genes from tomato (SlCPT1, SlCPT2 and SlCPT6) are closely linked to terpene synthase gene clusters.
Plant isoprenoids represent a group of structurally diverse compounds that include photosynthetic pigments, sterols, hormones, redox co-factors in electron transport, and various specialized metabolites that are unique to particular plant lineages (Kirby and Keasling, 2009). These compounds are synthesized from common five-carbon (C5) isoprene building blocks that originate from two distinct pathways (McGarvey and Croteau, 1995; Rodríguez-Concepción and Boronat, 2002). The mevalonate (MVA) pathway in the cytosol produces isopentenyl diphosphate (IPP), while the methylerythritol phosphate (MEP) pathway in the plastids produces both IPP and its isomer, dimethylallyl diphosphate (DMAPP).
Isoprenoid synthesis is initiated by condensation of IPP with DMAPP. Successive head-to-tail additions of IPP generate longer-chain polyisoprenoids that typically range in size from C15 to C120. The sequential transfer of IPP units to allylic diphosphate acceptors is catalyzed by the class of enzymes known as prenyltransferases. Depending on the stereochemistry of the polyisoprenoid product, these enzymes are classified as either trans–prenyltransferases (TPTs) or cis–prenyltransferases (CPTs). Despite similarities in substrate preference and reaction products, TPTs and CPTs utilize different catalytic mechanisms and may be readily distinguished from one another by their primary amino acid sequences (Liang et al., 2002; Kharel and Koyama, 2003).
Several plant TPTs involved in the synthesis of geranyl diphosphate (GPP, C10), trans,trans–farnesyl diphosphate (FPP, C15), all-trans–geranylgeranyl diphosphate (GGPP, C20) and solanesyl diphosphate (SPP, C45) have been identified and characterized (Burke et al., 1999; Hirooka et al., 2003; Lange and Ghassemian, 2003; Ducluzeau et al., 2012). The products of these enzymes provide the precursors for brassinosteroids, carotenoids, gibberellins, prenylquinones, sterols and a variety of terpenes. Comparatively little is known about plant CPTs. On the other hand, the roles of bacterial, yeast and mammalian CPTs are well established. In bacteria, CPTs synthesize long-chain polyisoprenoid diphosphates (C50–C55) that serve as lipid carriers in cell-wall peptidoglycan biosynthesis. In yeast and mammals, CPTs produce dehydrodolichyl diphosphate, the precursor of dolichol, a mixed cis,trans–polyisoprenoid (C75–C95) with a reduced double bond in the α isoprene unit that has an indispensible role in the post-translational modification of proteins (Figure 1).
Polyisoprenoids have been identified in numerous plants, and broadly fall into one of two classes: dolichols and polyprenols (Swiezewska and Danikiewicz, 2005; Skorupinska-Tudek et al., 2008). As in animals, plant dehydrodolichyl diphosphates are synthesized by CPTs that elongate trans–FPP with C5 units in a cis orientation resulting in a final linear polymer that is typically 70–120 carbons in length. Polyprenyl diphosphates are structurally similar but are unsaturated at the terminal isoprene position (Figure 1). Moreover, polyprenols from plant sources appear to be initiated using trans–substrates of varying length. The so-called ‘betulaprenols’ and ‘ficaprenols’ are synthesized from trans–FPP and trans–GGPP, respectively.
Clear genomic evidence exists for the distribution of CPT-like proteins throughout the plant kingdom. For instance, in Arabidopsis thaliana, a CPT gene family with nine members has been reported (Surmacz and Swiezewska, 2011). One member of this group, ACPT/DPS (At2g23410), contributes toward the biosynthesis of long-chain polyisoprenoids (Cunillera et al., 2000; Oh et al., 2000), while another member, AtHEPS, was recently identified as a heptaprenyl diphosphate synthase (Kera et al., 2012). However, functions have not been assigned for the remaining members of the Arabidopsis CPT family. Other reports of plant CPTs include several genes from Hevea brasiliensis and Taraxacum brevicorniculatum that are associated with synthesis of natural rubber particles, and the LLA66 gene expressed in the anthers of Lilium longiflorum that appears to be involved in microspore development (Asawatreratanakul et al., 2003; Schmidt et al., 2010; Liu et al., 2011; Post et al., 2012). Finally, CPTs provide the precursors for monoterpene and sesquiterpene biosynthesis, respectively, in the glandular (type VI) trichomes of two Solanum species, cultivated tomato (S. lycopersicum) and its wild relative, S. habrochaites. In tomato, neryl diphosphate synthase 1 (NDPS1) condenses IPP and DMAPP to neryldiphosphate (NPP), and a monoterpene synthase, β–phellandrene synthase 1 (PHS1), uses NPP to produce β–phellandrene and several other monoterpenes (Schilmiller et al., 2009). In S. habrochaites, a cis,cis–farnesyl diphosphate synthase (zFPS) catalyzes two sequential additions of IPP to DMAPP to produce cis,cis–farnesyl diphosphate (Z,Z–FPP), and a sesquiterpene synthase, santalene/bergamotene synthase (SBS), converts Z,Z–FPP to santalene and bergamotene (van Der Hoeven et al., 2000; Sallaud et al., 2009).
Although the function of the vast majority of plant CPT genes whose sequences are found in various databases is not well established, their homology to dehydrodolichyl diphosphate synthases from non-plant species suggests that they are involved in the synthesis of long-chain polyisoprenoid diphosphates. At present, NDPS1 and zFPS are the only CPTs from plants that are known to catalyze the formation of short-chain prenyl diphosphates (≤C20). To understand the evolutionary origin of CPTs that synthesize short-chain prenyl diphosphates, we have analyzed the entire CPT gene family present in the genome of S. lycopersicum, taking advantage of the recently released tomato genome sequence.
Identification of the tomato CPT gene family
To identify all genes with homology to NDPS1, we searched the recently released tomato genome sequence (http://solgenomics.net). This search identified six additional CPT-like gene sequences encoding proteins with homology to NDPS1 (SlCPT1), which we named SlCPT2–7 (Figure 2a and Figure S1). In cases where a gene sequence was incomplete, the regions spanning the gaps were amplified from genomic DNA to complete the sequence of the gene. We also screened EST databases (http://solgenomics.net/), and found representatives of all seven genes, but no additional sequences. We conclude that the tomato CPT gene family contains seven members, including NDPS1. SlCPT2 is located in the same cluster of terpene synthase genes on chromosome 8 as NDPS1 (Falara et al., 2011), SlCPT3 is located on chromosome 3, SlCPT4 and SlCPT5 are tandemly linked (in the same direction) and are present on chromosome 10, and SlCPT6 and SlCPT7 are both located on chromosome 6 but 9.6 Mb apart, with SlCPT6 embedded within a terpene synthase gene cluster (Falara et al., 2011).
Based on sequence homology to NDPS1 and related sequences from other species, it was predicted that SlCPT2 contains three introns (same number and position as in NDPS1), SlCPT3 contains no introns, and SlCPT4–7 each contain two introns, in positions equivalent to introns 2 and 3 in NDPS1 (Figure 2a and Figure S1). To verify the predicted exon/intron boundaries for each gene model (Figure 2a and Figure S1), we amplified full-length coding sequences for each gene by RT–PCR. Full-length cDNAs with all introns removed were obtained for all SlCPT genes. However, fully processed SlCPT6 cDNAs were only obtained when using RNA prepared from root to red fruits; RT–PCR products obtained from all other source tissues revealed transcripts with an unspliced first intron (and therefore probably non-functional because of stop codons introduced in-frame).
Alignment of the tomato deduced protein sequences with previously characterized CPTs from Escherichia coli, Saccharomyces cerevisiae and A. thaliana revealed four general features (Figure 2b). First, the five characteristically conserved regions (regions I–V) of CPTs (Kharel and Koyama, 2003) are also conserved in the tomato CPT proteins, including the Asp residue in region I that critically affects catalysis. Second, the five residues implicated in substrate binding (F70, S71, R194, R200 and E213; amino acid numbers corresponding to the E. coli enzyme) are present in all SlCPTs, except for SlCPT3, which contains an L residue in place of E. Interestingly, in the yeast dehydrodolichyl diphosphate synthase, the residue at position 213 is not an E but an R. Third, the SlCPT3 protein is the only member of the tomato CPT family without an N–terminal extension, a feature that is shared with yeast and E. coli proteins. Lastly, the region inferred from crystallographic and mutagenesis studies to impact product chain length (indicated by the dashed blue line between regions III and IV in Figure 2b) varies significantly among SlCPTs. The presence of additional residues in this region is normally associated with long-chain elongating enzymes, such as the CPT from yeast. Only SlCPT3 has a full complement of amino acids in this region, while the other tomato proteins lack between one and seven residues in this stretch.
To explore the evolutionary relationships between the tomato CPT family and those found in other plants, an expanded search of six plant genomes [three dicots (Arabidopsis thaliana, Vitis vinifera and Populus trichocarpa) and three monocots (Oryza sativa, Sorghum bicolor and Zea mays)] was performed. A total of 30 additional CPT-like sequences from these six species (AtCPT1–9, VvCPT1–3, PtCPT1–5, OsCPT1–3, SbCPT1–5 and ZmCPT1–5) were identified and used to construct an unrooted phylogenetic tree together with representative CPTs from animals, eubacteria and yeast (Figure 2c). This analysis placed CPTs into four phylogenetic groups, with group 1 and group 2 representing dicot- and monocot-specific branches, respectively. Six of the seven tomato CPT genes – SlCPT1, SlCPT2, SlCPT4, SlCPT5, SlCPT6 and SlCPT7 – are in group 1. The CPTs from cyanobacteria form a distinct clade, group 3, with other bacterial CPTs from the phyla of actinobacteria, proteobacteria and firmicutes. Group 4, in which SlCPT3 is placed, includes CPTs known to synthesize dolichols in yeast, humans and zebrafish, the long-chain CPTs from Hevea brasiliensis involved in natural rubber biosynthesis, as well as one to three sequences found in the genomes of each of the six representative plant species. The majority of plant sequences in group 4 (with the exception of VvCPT1 and ZmCPT3) also appear to lack N–terminal extensions.
SlCPT gene expression
The widespread expression of the SlCPT genes is apparent from searches of the publically available tomato EST collections. To determine the relative amount and tissue distribution of the SlCPT transcripts, the expression of each SlCPT gene was measured in a selection of tomato tissues by real-time PCR. Expression was quantified by the standard curve method in order to compare absolute expression levels between each gene (Figure 3). As previously reported, NDPS1 was expressed predominantly in stem and leaf trichomes at very high levels (Schilmiller et al., 2009), and a similar pattern of expression was observed for SlCPT7. Although transcripts for SlCPT3 and SlCPT5 were also found to be enriched in trichomes, expression was detected to varying levels in all other tissues examined. SlCPT2 and SlCPT4 were also expressed in all tissues, but at relatively low levels compared with other members of the family, while SlCPT6 was expressed almost exclusively in roots and red fruits.
Subcellular localization of SlCPTs
The subcellular distributions predicted by various algorithms (TargetP (http://www.cbs.dtu.dk/services/TargetP/), WoLF PSORT (http://wolfpsort.org/) and Predator (http://urgi.versailles.inra.fr/predotar/predotar.html)) were inconclusive in determining whether the N–terminal extensions present on each SlCPT gene (with the exception of SlCPT3) encoded a bona fide transit peptide. Therefore, the first approximately 120 amino acids of each SlCPT gene were fused to GFP and transiently expressed in Arabidopsis mesophyll protoplasts to determine whether these N–terminal extensions mediate transport of GFP to distinct subcellular locations. For SlCPT3, which lacks an obvious targeting peptide, the entire coding region was fused to GFP. Consistent with the absence of a targeting sequence, SlCPT3 appeared to be localized to the non-organellar area of the cell, presumably the cytosol (Figure 4). All other members of the CPT family were found to be associated with plastids, with two distinct patterns of GFP fluorescence observed. For SlCPT1, 2 and 5, a uniform GFP fluorescence signal was observed evenly throughout the plastids. On the other hand, a punctate pattern of GFP fluorescence was observed for SlCPT4, 6 and 7 that coincided with chlorophyll autofluorescence (Figure 4). To rule out the possibility that these proteins localize to the mitochondria, the experiments were performed in transgenic protoplasts constitutively expressing a mitochondrial-localized CFP reporter. The punctate GFP fluorescence in these experiments did not coincide with the mitochondrial CFP signal (see Figure 4 for results with SlCPT6 and SlCPT7).
Enzymatic characterization of SlCPTs
To date, all plant CPTs that have been enzymatically characterized are involved in either dolichol biosynthesis (>C75) or synthesis of the short-chain prenyl diphosphates NPP and Z,Z–FPP (Figure 1). However, CPTs associated with the production of medium-chain polyisoprenoids (C25–C75) have yet to be identified, despite the reported occurrence of these compounds as well as the enzymatic activities responsible for their synthesis in plants. To examine the enzymatic properties of each SlCPT, the cDNAs for each gene were expressed in E. coli to give truncated proteins (without transit peptide) with C–terminal His tags, as previously done for NDPS1 (Schilmiller et al., 2009). SlCPT3, which lacks a transit peptide, was produced as the full-length protein. The recombinant proteins were purified by Ni2+ affinity chromatography and assayed for CPT activity using 14C–IPP and a variety of available initiator substrates, including DMAPP, GPP, NPP, all-cis– and all-trans–FPP, and all-trans–GGPP. Products were analyzed by radio-TLC (and, in the case of SlCPT2 and SlCPT6, also by GC–MS). The tested tomato CPTs (with the exception of SlCPT3) were broadly characterized as either short-chain CPTs (SlCPT2 and SlCPT6) or medium-chain CPTs (SlCPT4, SlCPT5 and SlCPT7). The range of substrates used by these enzymes is summarized in Table 1, and the products produced from the substrate for which the highest activity was observed are shown in Figure 5(a). SlCPT6′s preferred acceptor substrate was NPP, and the product obtained in this reaction was Z,Z–FPP (Figure 5a and Figure S3). SlCPT2 preferred DMAPP, and produced a product tentatively identified as nerylneryl diphosphate (NNPP, equivalent to Z,Z,Z–GGPP) (Figure S4), SlCPT4 and SlCPT5 produced polyprenyl diphosphates with a range of C5 subunits but with the predominant products being C50–55 and C55–65, respectively (Figure 5a). Both of these enzymes were able to elongate all substrates that were assayed, but SlCPT4 preferred NPP and GPP while SlCPT5 utilized trans–FPP and Z,Z–FPP most efficiently (Table 1). SlCPT7 was the only member to produce ‘ficaprenols’, as it uses GGPP exclusively and gives a range of polyprenols, with C25–C35 products being dominant (Figure 5a).
Table 1. Enzyme properties of recombinant SlCPTs
aNPP and Z,Z–FPP, btrans–FPP and Z,Z–FPP, and cNPP were utilized with slightly lower efficiency.
No in vitro activity with any substrate was demonstrated for E. coli-produced SlCPT3. As SlCPT3 is most closely related to dehydrodolichyl diphosphate synthases from eukaryotes (Figure 2c), and no other member of the SlCPT family appears to produce the long-chain (C75–C90) products that are typical of plant dolichols, we tested whether SlCPT3 complimented the growth defect of the Saccharomyces cerevisiae dolichol biosynthesis mutant, rer2, which has a severely retarded growth rate at temperatures above 30°C (Sato et al., 1999) as a positive control, the E. coli undecaprenyl pyrophosphate synthase (UPPS) that compliments the rer2 growth defect (Rush et al., 2010), was also introduced into the mutant cells. When SlCPT3 was introduced into rer2 mutant cells (rer2∆/SlCPT3), the temperature-sensitive growth defect was rescued, as cells readily grew at 33°C whereas mutant yeast cells did not (Figure 5b). The restoration of growth was associated with significantly higher total CPT enzyme activity on microsomal membranes (the site of dolichol biosynthesis) from cells of rer2∆/SlCPT3 compared to the rer2 mutant (Figure 5c). The enzymatic products of these CPT assays were resolved by radio-TLC and shown to include a long-chain polyprenyl diphosphate (C65) that was synthesized by rer2∆/SlCPT3 microsomes but not by microsomes from the rer2 mutant (Figure 5d).
Constitutive and trichome-specific RNAi-mediated knockdown of NDPS1
NDPS1 catalyzes the formation of NPP, which is used as a substrate by PHS1 to produce β–phellandrene and several other monoterpenes (Schilmiller et al., 2009). NDPS1 is expressed several hundred times more highly in type VI trichomes than in the rest of the plant (Figure 3) (Schilmiller et al., 2009), and its contribution to monterpene synthesis (mostly β–phellandrene) in these trichomes has been identified previously (Schilmiller et al., 2009). However, the trichomes constitute only a very small part of the mass of the cultivated tomato plant. To determine whether the low-level expression of NDPS1 in non-trichome tissue contributes to monoterpene biosynthesis in tomato, we performed RNAi-mediated knockdown of NDPS1 using two independent RNAi constructs, one driven by the promoter of methylketone synthase 1 (MKS1), a 3–ketoacid decarboxylase gene that is highly expressed in type VI trichomes (Figure 6a) (Fridman et al., 2005), and one driven by the CaMV 35S promoter.
We generated >40 transgenic tomato lines carrying ProMKS::NDPS1RNAi and Pro35S::NDPS1RNAi constructs. From these lines, we identified those in which NDPS1 expression is most highly suppressed by performing solid-phase microextraction (SPME) on leaf volatiles and examining the β–phellandrene/β–caryophyllene ratio, as β–caryophyllene, a sesquiterpene, is a prominent tomato leaf volatile and is synthesized from the substrate trans–trans–FPP (Schilmiller et al., 2010), and therefore its synthesis is not affected by suppressing NDPS1 transcript levels (Figure S4). For selected lines showing the greatest reduction in the β–phellandrene/β–caryophyllene ratio, a comparison of NDPS1 transcript levels in total RNA extracted from whole leaves with that of control plants indicated a mean knockdown of 75 and 87% with the MKS1 and 35S promoters, respectively (Figure 6b). For RNA isolated from leaf trichomes of these transgenic lines, 76% NDPS1 knockdown was observed in the ProMKS::NDPS1RNAi lines, while only 65% NDPS1 knockdown was measured in Pro35S::NDPS1RNAi lines (Figure 6b). On the other hand, for RNA isolated from leaves from which the glandular trichomes had been removed (Figure 6b,c), there was no reduction in NDPS1 transcript levels compared to control in the ProMKS::NDPS1RNAi lines, but a 90% decline in NDPS1 transcript abundance was observed in the Pro35S::NDPS1RNAi lines. Taken together, these results indicate that, while NDPS1 is predominantly expressed in glandular trichomes, it is also expressed elsewhere in the leaf, and that NDPS1 transcript reduction in non-trichome cells is achieved in Pro35S::NDPS1RNAi but not ProMKS::NDPS1RNAi lines. However, the latter construct was somewhat better at suppressing NDPS1 gene expression in the trichomes (Figure 5b, right).
To examine the effect on β–phellandrene synthesis of reduction in NDPS1 transcript levels by the two RNAi constructs, total volatile compounds were extracted from leaves of plants showing high levels of NDPS1 transcript reduction. The amounts of β–phellandrene and β–caryophyllene were quantified and compared to those found in wild-type leaf tissue. On average, the ProMKS::NDPS1RNAi lines exhibited a 56% decrease in extractable β–phellandrene, while a decrease of more than 92% was observed in Pro35S::NDPS1RNAi lines (Figure 7). There was no significant difference in the amount of β–caryophyllene between wild-type and either RNAi line, or in the total sesquiterpene content (Figure 7). When trichomes were removed from wild-type leaves, extractable β–phellandrene and β–caryophyllene levels fell by 72 and 89%, respectively, indicating that the majority of each of these compounds, but not all, is found in the trichome, as previously observed (Schilmiller et al., 2009). Consistent with this result, when β–phellandrene levels were compared in plants from which trichomes had been removed, these levels were the same for trichome-less wild type leaves and the ProMKS::NDPS1RNAi line, but the Pro35S::NDPS1RNAi line showed an 84% reduction in extractable β–phellandrene levels compared with trichome-less wild-type leaves, with no significant changes in β–caryophyllene. Taken together, these results indicate that a small but significant proportion of β–phellandrene is found outside the glandular trichomes, and that impaired NPP synthesis significantly affects the amount of β–phellandrene synthesis throughout the leaf tissue.
The tomato CPT family contains seven members, six of which encode plastidic proteins
To date, complete bioinformatics analysis of a plant CPT family has only been reported for Arabidopsis (Surmacz and Swiezewska, 2011). However, of the nine Arabidopsis CPT genes identified, the function of only one gene (AtCPT1) has been described (Cunillera et al., 2000; Oh et al., 2000). This gene was demonstrated to partially rescue the yeast dehydrodolichyl diphosphate synthase mutant, rer2, and was therefore deduced to encode the Arabidopsis equivalent. The subcellular localizations of the Arabidopsis proteins have also not been reported. Here we show that tomato has seven CPT genes, one of which, SlCPT3, is closely related to dehydrodolichyl diphosphate synthases from eukaryotes (group 4, three genes in Arabidopsis), with the other six tomato CPT genes residing in a group comprised of dicot-specific CPTs (group 1; six genes in Arabidopsis, including AtCPT1; Figure 1). Examination of the phylogenetic tree shown in Figure 1 indicates that a recent series of gene duplications occurred in both tomato and Arabidopsis after the split of the two lineages, with SlCPT4 and SlCPT5 being one example of genes related to each other by such a duplication, and SlCPT1, SlCPT2, SlCPT6 and SlCPT7 being a second example in tomato, while AtCPT1, AtCPT2, AtCPT6, AtCPT8 and AtCPT9 were created by duplications in Arabidopsis after the split of the tomato and Arabidopsis lineages from each other.
Bioinformatic analysis and subcellular localization experiments indicate that SlCPT3 does not have an N–terminal extension and appears to reside in the cytosol, while the other six tomato CPT proteins contain a transit peptide at their N–terminus and are localized to the plastids (Figures 1 and 3). The observation that most tomato cis–prenyltransferases reside in the plastids fits well with previous reports indicating that the bulk of polyisoprenoid biosynthesis occurs in this compartment (Spurgeon et al., 1984; Sakaihara et al., 2000; Skorupinska-Tudek et al., 2008), a compartment that also contains an active MEP pathway that leads to the synthesis of IPP, DMAPP, GPP, all-trans–FPP and GGPP (McGarvey and Croteau, 1995; Bonk et al., 1997; Sanmiya et al., 1999; Phillips et al., 2008), and also NPP and Z,Z–FPP, at least in Solanum (Sallaud et al., 2009; Schilmiller et al., 2009).
SlCPT4, SlCPT5 and SlCPT7 produce medium-length chain polyisoprenyl diphosphates
A significant proportion of the polyisoprenoids that are found in plant tissues are of medium chain length (C25–C65); however, no CPTs responsible for their synthesis have been identified (Swiezewska and Danikiewicz, 2005). We show here that SlCPT4, 5 and 7 synthesize polyprenols in that range. SlCPT4 is expressed relatively uniformly in all tissues, albeit at a much lower level than either SlCPT5 or SlCPT7 (Figure 2). It is noteworthy that SlCPT4 appears to prefer GPP and NPP as substrates, given that all plant polyisoprenoids that have been identified so far contain either two (ω–trans2–cisn) or three (ω–trans3–cisn) trans isoprene units located at the ω–terminus of the molecule (Skorupinska-Tudek et al., 2008). In the case of SlCPT4, elongation of a C10 substrate (such as GPP or NPP) results in a mono-trans type (ω–trans1–cisn), which has only been reported in bacteria (Schulbach et al., 2000). On the other hand, SlCPT7 is the only tomato CPT member that strictly uses GGPP as the substrate, to give C25–35 products, thus appearing to be the only CPT in the tomato genome that synthesizes ficaprenols (ω–trans3–cisn), which have been identified in many other plants (Swiezewska et al., 1994).
Despite the widespread occurrence of medium-chain products (C25–C65) such as those synthesized by SlCPT4, 5 and 7, their functional significance remains unclear. Medium- and long-chain polyisoprenoids are known to accumulate in senescing photosynthetic tissues, and their synthesis is stimulated by light (Chojnacki and Vogtman, 1984; Bajda et al., 2005). Indirect evidence for the association of polyisoprenoids with plant cell walls comes from a recently characterized group of saprophytic bacteria that possess a unique polyisoprenoid-binding protein that aids in the dismantling of plant cell-wall material (Vincent et al., 2010). It has been shown in vitro that polyisoprenoids alter the permeability of cellular membranes by increasing membrane fluidity (Vigo et al., 1984; Ciepichal et al., 2011). Also, the association of polyisoprenoids with specific membrane proteins is thought to influence translocation processes (Zhou and Troy, 2003). The incorporation of medium-chain lipophilic polyisoprenoids into membranes is plausible given the thickness of biological membranes (40–60 Å), although most plant polyisoprenoids are considerably shorter in length and only penetrate to the mid-bilayer region from one side of the membrane, much like the orientation of dolichol on the ER membrane (Schenk et al., 2001; Zhou and Troy, 2003; Swiezewska and Danikiewicz, 2005). It is possible that plastid-derived polyisoprenoids associate with plastoglobuli (Austin et al., 2006), the lipid-rich structures in plastids that also contain proteins involved in isoprenoid metabolism such as GGPP synthase (Cheniclet et al., 1992; Ytterberg et al., 2006). Intriguingly, proteins localized to this area of the plastid exhibit a punctate GFP fluorescence pattern within the chloroplast (Eugeni-Piller et al., 2011), very similar to the pattern observed for SlCPT4, 6 and 7 in this study. The finding that all-trans–GGPP is the substrate for SlCPT7 further supports such an arrangement.
SlCPT3 is involved in dolichol biosynthesis
Phylogenetic analysis places SlCPT3 in a distinct group of CPTs (group 4) that include dolichol synthases from animals to yeast. When SlCPT3 was expressed in the yeast dolichol biosynthesis rer2 mutant, growth was restored and isolated microsomoal fractions synthesized a C65 isoprenyl diphosphate, further supporting the identification of SlCPT3 as a dehydrodolichyl diphosphate synthase. The localization of SlCPT3 to the cytosol, the site of dolichol biosynthesis on the ER membrane, is also consistent with this functional assignment.
The CPTs that synthesize dolichols in animals and yeast do not appear to function autonomously, but may require a partner protein known as NgBR (NUS1 in yeast) that also has a short CPT domain and another domain that is anchored to the ER membrane (Harrison et al., 2011). The association between the CPT and this ER-anchored protein has been suggested to form the functional dolichol synthase (Schenk et al., 2001; Harrison et al., 2011). Indeed RNAi-mediated knockdown of NgBR results in defective N–linked protein glycosylation and dolichol synthesis in animals, while knockout mutants of the yeast and plant orthologs are lethal (Yu et al., 2006; Zhang et al., 2010; Harrison et al., 2011). The fact that dehydrodolichyl diphosphate biosynthesis by a single eukaryotic protein has not been demonstrated in vitro, and that such an activity has only been recovered from microsomal membranes from animals, plants and yeast (Sakaihara et al., 2000; Shridas et al., 2003; Rush et al., 2010), suggests that eukaryotic dehydrodolichyl diphosphate synthases are evolutionarily conserved. The observation that rer2 yeast mutant cells complemented with SlCPT3 synthesized an isoprenyldiphosphate of only 65 carbons suggests that the SlCPT3 protein may not work as well with NUS1 as the endogenous RER2 protein does.
Function and evolution of SlCPT1, SlCPT2 and SlCPT6, the short-chain cis–prenyldiphosphate synthases
Until recently, only short prenyl diphosphates (C10, C15 and C20) produced by trans–prenyl diphosphate synthases were known in plants. These trans–prenyl diphosphates (GPP, trans–FPP and all-trans–GGPP) are used to create volatile and non-volatile terpenes by mono-, sesqui- and diterpene synthases, respectively, to create sterols and carotenoids by head-to-head condensation of trans–FPP and all-trans-GGPP, respectively, and to prenylate other molecules such as proteins and flavonols. In addition, they served as starter molecules for cis–prenyltransferases. The discovery of the synthesis of NPP and Z,Z–FPP by NDPS1 (SlCPT1) and zFPS in the trichomes of S. lycopersicum and S. habroachaites as the substrates of monoterpene and sesquiterpene synthases, respectively (Sallaud et al., 2009; Schilmiller et al., 2009) indicated additional biochemical functions for cis–prenyltransferases in plants.
Here we show that, in addition to NDPS1 (SlCPT1), tomato has two other CPTs that create short cis–prenyl diphosphates. SlCPT6′s preferred acceptor substrate is NPP, and its product is Z,Z–FPP, a cis–prenyl diphosphate that has already been reported to be synthesized in another Solanum species (Sallaud et al., 2009). SlCPT2 utilizes DMAPP as the acceptor molecule, and adds three IPP molecules to make NNPP, a compound (Figure 2) that has not previously been reported in plants. As previously reported (Falara et al., 2011), SlCPT1 and SlCPT2 are embedded in a cluster of terpene synthase genes on chromosome 8, one of which (PHS1) uses the product of SlCPT1 (NPP) to produce β–phellandrene, the major monoterpene found in tomato trichomes (Schilmiller et al., 2009). Here we show, via comparisons between NDPS1 RNAi lines driven by trichome-specific and CaMV 35S promoters, that NDPS1 is expressed in non-trichome cells of the leaves as well, and contributes to monoterpene production there. On chromosome 6, a suite of terpene synthase genes also surround SlCPT6 (Falara et al., 2011). It is not uncommon to find clusters of non-homologous but functionally related genes in plants (Osbourn 2010), and it is likely that certain terpene synthases may have co-evolved with the short-chain CPTs.
These three short-chain SlCPTs from tomato are closely related to each other and to zFPS from S. habrochaites (Figure 2). Together they form a distinct sub-clade within the dicot-specific CPT group 1 that is most closely related to SlCPT7, the ficaprenol synthase. The phylogenetic analysis shows that the short-chain CPTs evolved by gene duplications that occurred after the split between the tomato and Arabidopsis lineages. It further shows that the progenitor of the short-chain tomato CPTs is a ficaprenol synthase, because this enzymatic activity is widespread in plants (Swiezewska and Danikiewicz, 2005) and therefore likely to be the ancestral function in this sub-clade.
Apart from the Solanum short-chain CPTs, such synthases have been identified only from bacteria (Schulbach et al., 2000; Ambo et al., 2008). Given that most CPTs of animal, bacteria and plant origin synthesize long-chain products, the identification of short-chain CPTs has stimulated interest in the structural basis for this activity (Kharel and Koyama, 2003). Based on the crystal structure of the long-chain CPT from Micrococcus luteus (UPPS), a large hydrophobic cleft near region III of this protein is believed to accommodate the enzymatic product during prenyl chain elongation (Fujihashi et al., 2001). Kharel et al. (2006) noted a conspicuous reduction in the number of residues in this particular region of zFPP synthase from Mycobacterium tuberculosis, and insertion of only three residues into this region increased the product chain length from C15 to C70. Strikingly, the corresponding regions of SlCPT1, 2 and 6 are also lacking these residues, suggesting that this domain of CPT proteins critically affects chain length determination. It is noteworthy that SlCPT3, 4, 5 and 7, which synthesize longer chain products, all have more residues in this region (Figure 2b).
Arabidopsis thaliana (Col–0 ecotype) and tomato (cv. MP–1) plants (wild-type and transgenic) were grown in potting soil supplemented with Osmocote® (http://;www.scotts.com/;) in a growth chamber maintained at 22°C under a 12 h photoperiod (approximately 200 μmol m−2 sec−1 fluence rate generated by mixed cool-white fluorescent and incandescent bulbs). Red ripe tomato fruit was defined as breaker stage plus 7 days. Tomato stem and leaf trichomes were isolated by vigorously vortexing approximately 10 g of frozen plant material in a 50 ml centrifuge tube, and trichomes were collected from the bottom of the tube. Type VI tomato leaf trichomes were removed from 3-week-old tomato leaves by gently rolling a cotton swab over the leaf surface. The Arabidopsis thaliana line expressing a mitochondrial-localized CFP marker (Nelson et al., 2007) was obtained from the Arabidopsis Biological Resource Center (stock number CS16262).
cDNA isolation and phylogenetic analysis of CPT sequences
Primers for cDNA isolation were designed based on the predicted open reading frame of each gene, and used to amplify the full-length cDNA by RT–PCR. All primers used in this study are listed in Table S1. RNA was extracted from various tomato tissues using and EZNA plant RNA mini kit (Omega Biotek) (http://;www.omegabiotek.com/;), reverse-transcribed using Superscript II reverse transcriptase (Invitrogen) (http://;www.invitrogen.com/;), and used directly for PCR amplification with KOD hot start DNA polymerase (Novagen) (http://;www.emdmillipore.com/;). PCR products were cloned into pGEM-T-Easy (Promega) (http://;www.promega.com/;) and their sequences were verified. Using the SlCPTs, all CPT-like sequences in six additional plant genomes were identified (http://www.phytozome.net), and phylogenetic analysis of these sequences was performed using mega version 4 (Tamura et al., 2007).
To quantify the mRNA abundance of each SlCPT, quantitative RT–PCR was performed as previously described (Schilmiller et al., 2009).
The open reading frame of SlCPT3 and a region corresponding to the first approximately 120 amino acids of the remaining SlCPT gene products were amplified, digested accordingly, and ligated into pSAT6A (Yu et al., 2010), creating an in-frame C–terminal fusion protein with GFP. The constructs were mobilized into Arabidopsis protoplasts according to the ‘tape sandwich’ method (Wu et al., 2009), and GFP fluorescence was visualized 16 h after transfection using a Leica SP5 laser scanning confocal microscope (http://;www.leica-microsystems.com/;) as previously described (Falara et al., 2011). Spectral detection of CFP fluorescence was performed between 465 and 495 nm using a double dichroic 458/514 beam splitter.
Expression and purification of recombinant SlCPTs
Tomato CPTs were expressed and purified from E. coli cells as described previously (Schilmiller et al., 2009).
cis–prenyltransferases activity assays were performed as described previously (Schilmiller et al., 2009), and contained approximately 1–3 μg purified recombinant protein and 20 μM acceptor substrate, and were initiated by adding 14C–IPP (50 μCi mmol−1) at a final concentration of 80 μm (200 nCi), and incubated at room temperature for 30 min. To identify the products, 1 volume of 1 m HCl was added to the reaction mixture, and the hydrolyzed products were extracted with ethyl acetate and quantified by scintillation counting. Product analysis was also performed by thin-layer chromatography on reverse-phase 60 Å silica plates (Whatman), which were developed in acetone/water (39:1) and visualized by phosphorimaging. Authentic standards (C15–C120) were co-chromatographed and visualized by iodine vapor staining. For assays with SlCPT2 and SlCPT6 (the short-chain SlCPTs), reaction products were hydrolyzed by the more time-consuming method of alkaline phosphatase treatment rather than incubation with acid prior to TLC analysis, because acid hydrolysis of short-chain polyisoprenoids results in chemical rearrangements of some of the product (but the structures of longer chain isoprenoids are not affected). The preferred substrate for each enzyme was defined as the substrate that resulted in the highest specific activity under defined sub-saturating concentrations.
Yeast culture conditions, transformation and microsomal CPTase activity assays
The yeast strain YG932 (rer2∆ mutant) (MATα rer2∆::kanMX4 ade2–101 ura3–52 his3∆200 lys2–801) has been described previously (Schenk et al., 2001), and was routinely cultured in yeast peptone dextrose media (YPD) at 23°C. The SlCPT3 gene was transferred to pYEP352, as previously described (Rush et al., 2010), and yeast transformation was performed as described by Gietz and Schiestl (2007). Transformants were selected on 0.67% yeast nitrogen base without amino acids, 50 mm sodium succinate, pH 5.5, 2% glucose and all necessary auxotrophic requirements, except uracil. Preparation of microsomal membranes for CPT activity assays was performed as described by Rush et al. (2010).
RNAi-mediated knockdown of NDPS1 (SlCPT1)
A 204 bp fragment of NDPS1 (corresponding to base pairs 19–223) was amplified by PCR and ligated in a sense/antisense orientation into pRNA69 (Foster et al., 2002) between the XhoI/EcoRI and BamHI/XbaI restriction sites, respectively. The hairpin cassette was released by SpeI/SacI digestion and transferred to the pZP212 binary vector (Hajdukiewicz et al., 1994) between the XbaI/SacI restriction sites. For trichome-specific RNAi-mediated knockdown of NDPS1, the CaMV 35S promoter in pRNA69 was released by SacI/XhoI digestion and replaced by the methyl ketone synthase 1 promoter (MKS1, Genbank accession number GU98105.1), which was obtained from S. habrochaites genomic DNA by PCR. The specificity of the MKS1 promoter was confirmed by placing it upstream of GFP between the AgeI and BamHI sites of pSAT-EGFP–N1 (Tzfira et al., 2005). The expression cassette was digested using PI-PspI, and ligated into the corresponding sites of the pPZP-RCS2 binary vector (Tzfira et al., 2005). The binary vectors were introduced into tomato by the University of Nebraska Plant Transformation Facility (http://biotech.unl.edu/plant-transformation). GFP fluorescence in transgenic tomato type VI trichomes was visualized using a Nikon Eclipse E600 Y–FL fluorescence microscope (http://;www.nikon-instruments.jp) equipped with a 100 W mercury arc lamp, 505 nm dichroic mirror and using filter settings of 450–490 nm and 520–560 nm for GFP excitation and emission, respectively.
Terpene extraction and analysis
For analysis of total terpenes, approximately 200 mg of 3-week old leaf material was ground to a pulp in 2 ml hexane containing 34 μg ml−1 1,2–dimethoxyphenol (internal standard) using a glass mortar and pestle, and vortexed vigorously for 1 min. The extract was filtered through glass wool, dried using anhydrous MgSO4, and concentrated to a final volume of 0.2 ml. A 2 μl aliquot was injected into a QP–5000 GC–MS system (Shimadzu) (http://;www.shimadzu.com/;) equipped with an EC–WAX column (Grace Davison) (http://;www.discoverysciences.com/;) and analyzed as described by Falara et al. (2011). Terpenes were quantified by constructing standard concentration curves for β–phellandrene and β–caryophyllene.
We thank Thomas Clemente and the University of Nebraska plant transformation facility for generating transgenic tomato plants. We also thank Charles Waechter and Jeffrey Rush (University of Kentucky, Department of Molecular and Cellular Biochemistry, University of Kentucky College of Medicine, Lexington, Kentucky) for providing the yeast rer2∆ mutant and the E. coli UPPS construct, the members of the Solanum Trichome Project (http://www.trichome.msu.edu) for many fruitful discussions, and Greg Sobocinski (University of Michigan, Department of Molecular, Cellular and Developmental Biology, Ann Arbor, Michigan) for technical advice on confocal microscopy. This work was supported by US National Science Foundation awards DBI-0604336 and IOS-1025636 and by the Agricultural and Food Research Initiative Competitive grant number 2008-35318-04541.