Geranyl diphosphate synthase is required for biosynthesis of gibberellins


  • Chris C. N. Van Schie,

    Corresponding author
    1. Department of Plant Physiology, Swammerdam Institute for Life Sciences, University of Amsterdam, 1098 SM Amsterdam, The Netherlands
      (fax +1 858 534 7262; e-mail
    Search for more papers by this author
    • Present address: Chris C. N. van Schie, University of California San Diego, 9500 Gilman Drive #380, La Jolla, CA 92093-0380, USA.

  • Kai Ament,

    1. Department of Plant Physiology, Swammerdam Institute for Life Sciences, University of Amsterdam, 1098 SM Amsterdam, The Netherlands
    Search for more papers by this author
  • Axel Schmidt,

    1. Max Planck Institiute for Chemical Ecology, Jena, Germany
    Search for more papers by this author
  • Theo Lange,

    1. Institut fur Pflanzenbiologie, Technische Universität, Braunschweig, Germany
    Search for more papers by this author
  • Michel A. Haring,

    1. Department of Plant Physiology, Swammerdam Institute for Life Sciences, University of Amsterdam, 1098 SM Amsterdam, The Netherlands
    Search for more papers by this author
  • Robert C. Schuurink

    1. Department of Plant Physiology, Swammerdam Institute for Life Sciences, University of Amsterdam, 1098 SM Amsterdam, The Netherlands
    Search for more papers by this author

(fax +1 858 534 7262; e-mail


Geranyl diphosphate synthase (GPS) is generally considered to be responsible for the biosynthesis of monoterpene precursors only. However, reduction of LeGPS expression in tomato (Lycopersicon esculentum) by virus-induced gene silencing resulted in severely dwarfed plants. Further analysis of these dwarfed plants revealed a decreased gibberellin content, whereas carotenoid and chlorophyll levels were unaltered. Accordingly, the phenotype could be rescued by application of gibberellic acid. The dwarfed phenotype was also obtained in Arabidopsis thaliana plants transformed with RNAi constructs of AtGPS. These results link geranyl diphosphate (GPP) to the gibberellin biosynthesis pathway. They also demand a re-evaluation of the role of GPS in precursor synthesis for other di-, tri-, tetra- and/or polyterpenes and their derivatives.


Terpenoids comprise a diverse class of plant metabolites; most of them are known as defense-related compounds, flavors or scents. In addition, primary metabolites like carotenoids, chlorophyll, abscisic acid, quinone electron carriers, steroids and gibberellins are also terpene-derived (Lange and Ghassemian, 2003; McGarvey and Croteau, 1995). The biochemistry of terpene precursor biosynthesis is well studied. Isopentenyl diphosphate (IPP, C5) and dimethylallyl diphosphate (DMAPP, C5) are the terpene building blocks that can both be made in plastids and cytosol by independent pathways (Lichtenthaler, 1999). IPP and DMAPP are used by prenyltransferases to catalyze the synthesis of the general terpene backbones. Geranyl diphosphate synthase (GPS) generates geranyl diphosphate (GPP) for monoterpenes (C10), farnesyl diphosphate synthase (FPS) generates farnesyl diphosphate (FPP) for sesquiterpenes (C15) and triterpenes (C30), and geranylgeranyl diphosphate synthase (GGPS) generates geranylgeranyl diphosphate (GGPP) for diterpenes (C20) and tetraterpenes (C40). Most genes encoding important enzymes for terpene precursor synthesis have been identified in several plant species. However, localization, regulation and sharing of precursor pools are only starting to be explored (Botella-Pavia et al., 2004; Chang et al., 2005; Dudareva et al., 2005; Laule et al., 2003; Leivar et al., 2005; Nogues et al., 2006; Tholl et al., 2004).

Geranyl diphosphate synthase is generally considered to be responsible for the generation of the monoterpene precursor GPP. GPSs are either homomeric or heteromeric; the Clarkia breweri, Antirrhinum majus and Mentha piperita GPSs consist of a small and a large subunit, which have very low homology to homomeric angiosperm GPSs (Burke et al., 2004; Tholl et al., 2004), see Figure 1. The Abies grandis (Burke and Croteau, 2002), Citrus sinensis (CAC16851), Vitis venifera (ABD77587), Quercus robur (CAC20852), Arabidopsis thaliana (At2g34630; Bouvier et al., 2000) and tomato GPS (DQ286930) (probably) function as homodimers. The Arabidopsis GPS has two possible transcription and translation initiation sites, potentially resulting in proteins localized either in plastids or cytosol (Bouvier et al., 2000). The Arabidopsis genome contains only one GPS sequence (Bouvier et al., 2000; Lange and Ghassemian, 2003), and this is probably the case in most angiosperms (based on sequence searches and blasts,,, whereas three GPSs are present in the gymnosperm A. grandis (Burke and Croteau, 2002), which are different from other GPSs because they have high sequence homology to GGPSs.

Figure 1.

 Alignment of deduced amino acid sequences of prenyldiphosphate synthases. Included proteins: tomato (Lycopersicon esculentum) geranyldiphosphate synthase (LeGPS, DQ286930), Arabidopsis thaliana GPS (AtGPS, At2g34630), Citrus sinensis GPS (CsGPS, CAC16851, probably not full-length), Mentha piperita GPS; the large subunit of the heterodimer (MpGPSlsu, AAF08793) and tomato geranylgeranyldiphosphate synthase 1 (LeGGPS1, DQ267902). Amino acids identical in at least three of five sequences are boxed black, and in addition amino acids identical in the tomato and the Arabidopsis GPS are boxed grey. The alternative translation initiation sites (second methionine) for tomato GPS and Arabidopsis GPS are indicated by an asterisk. The DDxxD motif, conserved among prenyltransferases and terpene synthases is underlined. The cDNA regions that were used for RNAi in Arabidopsis or virus-induced gene silencing (VIGS) in tomato are indicated by grey (At-RNAi) and black (Le-VIGS) triangles, respectively.

We took a gene-silencing approach in order to investigate the role of GPS in terpene biosynthesis. Surprisingly, this revealed that GPS is involved in the synthesis of the diterpene-derived gibberellins. The implications of this finding are discussed, and an alternative model of gibberellin precursor biosynthesis and sharing is proposed.


Isolation and characterization of LeGPS

Based on nucleotide sequence homology to known GPS sequences, two tomato expressed sequence tags (ESTs) were identified ( one with homology to the 5′ region (TC165735) and the other to the 3′ region (AI485604). Using primers on the 5′ and 3′ ends of these ESTs, respectively, a full-length cDNA was amplified from tomato leaf extract. The tomato GPS (LeGPS, DQ286930) sequence is highly homologous to the Arabidopsis GPS (AtGPS, At2 g34630) and the Citrus GPS (AJ243739) (Figure 1), which are both active as homomers. In contrast, the heteromeric GPSs of Snapdragon (A. majus) and Mint (M. piperita, Mp) consist of a small and a large subunit (ssu and lsu, respectively). The lsu is closely related to GGPS (Tholl et al., 2004), whereas the ssu has very low homology to other prenyltransferases. The amino acid alignment in Figure 1 shows that LeGPS1 has only partial homology to the Mint GPSlsu, to which the previously described LeGGPS1 (Ament et al., 2006) is very homologous. In analogy with the Arabidopsis GPS protein, the N-terminal sequence of LeGPS (amino acids 1–94) might allow dual targeting to plastid and cytosol (Bouvier et al., 2000); the full length Arabidopsis protein is presumably targeted to plastids, where the transit peptide is removed upon import, whereas the shorter one resides in the cytosol.

Activity of recombinant LeGPS protein

A truncated version of LeGPS, starting at 2 amino acids upstream of the second Met (amino acid 95, Figure 1) and containing a His-tag, was expressed in Escherichia coli and used for prenyltransferase assays. This truncation was based on the observation that the Arabidopsis GPS with a similar truncation was active in vitro, and that, in addition, this processed mature form of GPS was isolated from plastids (Bouvier et al., 2000). The full-length recombinant LeGPS protein was almost completely insoluble (data not shown). The purified LeGPS showed mainly GPS activity, but also showed FPS activity and minor GGPS activity (Figure 2). The total synthesized and hydrolyzed prenyldiphosphate products comprised 51% (± 2.5%) geraniol (GPS activity), 42% (± 3.0%) farnesol (FPS activity) and 7% (± 2.2%) geranylgeraniol (GGPS activity). Bacterial extracts containing the vector-control showed no prenyltransferase activity.

Figure 2.

 Catalytic activity of recombinant tomato (Lycopersicon esculentum) geranyldiphosphate synthase (LeGPS). Purified recombinant LeGPS was assayed with dimethylallyl diphosphate and [14C]-isopentenyl diphosphate. The enzyme activity was measured by radio-gas chromatography (middle panel) and was identified by co-injection of terpene standards via thermal conductivity detector measurement (bottom panel). The main (hydrolyzed) products for LeGPS are geraniol (C10), farnesol (C15) and minor quantities of geranylgeraniol (C20). Purified protein extracts from bacterial cultures containing a vector control showed no activity (top panel).

Virus-induced gene silencing of GPS in tomato

The tobacco rattle virus (TRV) gene-silencing system (Liu et al., 2002a,b) was used to transiently silence LeGPS expression in tomato. The TRV construct, carrying a 350-bp fragment of the LeGPS cDNA (Figure 1), was transfected to seedlings, which resulted in a growth retardation phenotype (Figure 3a). The approximately twofold decreased plant height (Figure 3b) was a result of shorter internodes, a typical gibberellin deficiency effect. Spraying LeGPS-silenced plants biweekly with 50 μm gibberellic acid (GA3) rescued shoot elongation back to control levels, confirming that LeGPS-silenced plants indeed have a classical reduced-gibberellin phenotype (Brian and Hemming, 1955; Phinney, 1956). GA3 spraying did not affect the general silencing mechanism, as phytoene desaturase (PDS)-silenced control plants kept developing bleached leaves after spraying (data not shown).

Figure 3.

 Phenotype of LeGPS-silenced tomato plants using virus-induced gene silencing.
(a) Pictures were taken 35 days after inoculation. The dwarf phenotype was rescued by spraying gibberellic acid (GA3) biweekly.
(b) Plant length (35 days post-inoculation) is displayed as the average of three plants and SD are included: mock, control plants; TRV, empty vector-inoculated plants; TRV:GPS, GPS-silenced plants.

We also checked the expression of LeGA3ox, as in Arabidopsis the expression of this ‘late’ gibberellin biosynthesis gene, GA-3β-hydroxylase (At1g15550, AtGA3ox1), is known to be feedback-regulated in a GA-dose-dependent manner, i.e. GA3ox expression increases when levels of active gibberellins are low (Cowling et al., 1998). Indeed, the expression of the tomato homolog (LeGA3ox2, AB010992; Yang et al., 1998) was increased threefold in leaves in which LeGPS expression was reduced (Figure 4), corroborating the fact that gibberellin levels were decreased.

Figure 4.

 The effect of TRV:LeGPS inoculation on gene expression.
Gene expression levels were determined relative to Actin2 expression by quantitative-reverse transcriptase-polymerase chain reaction and displayed as the average values from four to seven leaf samples, including SD. Gene expression was analyzed for LeGPS, the GA-marker gene GA-3β-hydroxylase (LeGA3ox) and LeGGPS1. Leaf samples were from control plants (mock), empty vector-inoculated plants (TRV) and GPS-silenced plants (TRV:GPS).

To verify LeGPS silencing, leaves from silenced tomato plants were harvested 5 weeks after inoculation and RNA was extracted. Quantitative-reverse transcriptase-polymerase chain reaction (Q-RT-PCRs) showed that LeGPS expression was clearly decreased (by twofold on average) in leaf samples of LeGPS-silenced plants, as compared with control plants (mock) or plants inoculated with the ‘empty TRV vector’ (Figure 4). Furthermore, to exclude cross-silencing of distantly related prenyltransferases (LeGPS is 18% identical to LeGGPS1 on the amino acid level, see Figure 1), the transcript level of LeGGPS1 (Ament et al., 2006) was analyzed. Expression of LeGGPS1 was not reduced (Figure 4), indicating that no cross-silencing of weakly similar LeGGPS genes had occurred (on the nucleotide level, LeGPS and the LeGGPSs have no apparent sequence homology; data not shown).

The effect of LeGPS silencing on gibberellin and pigment levels

To confirm that LeGPS-silenced plants are dwarfed as a result of gibberellin deficiency, the gibberellin content in leaves was analyzed. Gibberellin measurements were performed on the same leaf samples as those used for gene-expression analysis. Gibberellin contents were more than threefold reduced in LeGPS-silenced leaves (Figure 5a). As gibberellin, carotenoid and chlorophyll synthesis all require GGPP, carotenoid and chlorophyll contents were also measured in LeGPS-silenced leaves (Figure 5b). These pigment levels were unaffected in samples with decreased gibberellin levels. As a control, photobleached PDS-silenced leaves were analyzed (see also Liu et al., 2002a,b). These leaves contained eight- to tenfold less carotenoids and chlorophyll (data not shown).

Figure 5.

LeGPS silencing reduces gibberellin content without affecting levels of carotenoids or chlorophyll.
(a) Total gibberellin content was analyzed by SIM-GC/MS using leaf extracts from control plants (mock), empty vector-inoculated plants (TRV) and LeGPS-silenced plants (TRV:GPS). Total gibberellin content is displayed as the average value (and SD) of between three and seven leaves.
(b) Total chlorophyll and carotenoid contents were analyzed spectrophotometrically using leaf extracts from the same leaf samples as used for gibberellin measurements. Pigment contents are displayed as the average values (and SD) of between four and seven leaves.

Silencing Arabidopsis GPS with RNAi results in dwarfing

In order to test whether the reduction of GPS expression leads to a dwarfed phenotype in another plant species, AtGPS-RNAi Arabidopsis plants were generated. An RNAi construct was made using a 560-bp fragment of AtGPS (At2g34630; Bouvier et al., 2000), see Figure 1), which has no homology to other Arabidopsis genes. Independent RNAi lines showed a growth reduction compared with control plants, ranging from slight growth retardation to severely dwarfed plants with delayed flowering (Figures 6a and S1). Gene expression analysis by Q-RT-PCR showed that AtGPS expression was dramatically reduced in the Arabidopsis AtGPS-RNAi lines compared with the control plants (Figure 6b). A second RNAi construct, carrying a different part of AtGPS (see Figure 1), resulted in a similarly dwarfed phenotype (data not shown). Finally, the phenotype of the Arabidopsis GABI-Kat line 097G02 (Genbank, AL755431), which has a T-DNA inserted in the first intron of AtGPS, suggests that a GPS knock-out is embryo-lethal or unable to germinate. In a population of 80 plants, only wild-type and hemizygous individuals were present. In siliques from selfed plants that were hemizygous for the T-DNA insertion, 16% of the embryos were aborted, compared with only 2.4% in control plants (Figure S2). This indicates that AtGPS is an essential gene during embryo development, in which gibberellins are of vital importance.

Figure 6.

 The effect of GPS silencing in Arabidopsis.
(a) Phenotype of Arabidopsis plants transformed with an AtGPS-RNAi(1) construct. Three T2 plants of six independent lines are shown.
(b) Quantitative-reverse transcriptase-polymerase chain reaction analysis of AtGPS expression in the same RNAi lines. AtGPS expression levels were corrected for Actin expression and are displayed relative to expression in control plants (average control set to 100%). Standard errors are included.


Silencing LeGPS causes a prototype GA-related dwarf phenotype

The tomato LeGPS described in this paper has high homology (76% amino acid identity of the processed protein) to the previously described Arabidopsis GPS, AtGPS (Bouvier et al., 2000). The in vitro activity of the recombinant protein shows that LeGPS is a bona fide GPS, but suggests that LeGPS can generate substantial quantities of FPP and traces of GGPP (Figure 2). The question of to what extent the additional activity of this recombinant protein occurs in planta remains. The homologous AtGPS did not have clear FPS activity (Bouvier et al., 2000), whereas in grand fir (A. grandis) one of three GPSs showed significant (30%) FPS activity (Burke and Croteau, 2002). An FPS with additional GGPS activity was recently described (Cervantes-Cervantes et al., 2006), whereas a GGPS from Hevea brasiliensis produced some FPP (Takaya et al., 2003). Therefore, chain-length specificity of prenyltransferases does not appear to be highly stringent, but one should keep in mind that in vitro assay products do not necessarily reflect the in planta situation. It will be interesting to investigate to what extent possible prenyltransferase by-products contribute to the sometimes hypothesized presence of GPP in the cytosol or FPP in plastids (Aharoni et al., 2003; Wu et al., 2006).

It should be noted that the previously characterized heteromeric GPSs were isolated from highly specialized peltate glands and flower petals (Burke et al., 1999; Tholl et al., 2004), and that the small family of homomeric A. grandis GPSs were isolated from resin-producing needles of a gymnosperm (Burke and Croteau, 2002). Therefore, one cannot assume that the homomeric Arabidopsis and tomato GPS, which are expressed throughout the plant ( and data not shown, respectively), should have the same characteristics and functions as these other GPSs that more closely resemble GGPSs.

The identified tomato GPS has been implicated in gibberellin biosynthesis because LeGPS-silenced plants showed a dwarfed phenotype and reduced gibberellin levels (Figures 3a and 5a). We have observed the dwarfing phenotype after virus-induced transient LeGPS silencing, which did not affect pigment levels. Because of the nature of this technique certain plant organs or cell-types might not be silenced, influencing the possible phenotypes. However, it has been established that virus-induced gene silencing (VIGS) works in pigment synthesizing cells, as genes in carotenoid and chlorophyll synthesis have successfully been silenced (Hiriart et al., 2002; Liu et al., 2002a,b; Page et al., 2004).

In our VIGS experiments with tomato we only observed about 50% decreased LeGPS expression. This is probably a result of the patchiness of silencing. In addition, VIGS is transient and partial (Burch-Smith et al., 2004; Liu et al., 2002a,b). A characteristic of silencing approaches is the possibility of cross-silencing homologous genes. We wanted to rule out cross-silencing of GGPSs because these can be involved in gibberellin precursor biosynthesis, and because GGPSs have low sequence homology with GPS (max. 25% identity on the amino acid level). In our experiments, cross-silencing is highly unlikely because the LeGPS fragment used for VIGS (Figure 1) does not have homology to other tomato transcripts ( Moreover, Figure 4 shows that LeGGPS1 expression is not decreased in LeGPS-silenced tomato leaves. The Arabidopsis GPS (AtGPS) RNAi fragments (Figure 1) also have no homology on the nucleotide level to other Arabidopsis genes. In the complete AtGPS cDNA sequence, there are no identical stretches of 21 nt or more that would be sufficient for silencing with any of the AtGGPSs or any other gene. Therefore, we conclude that in both tomato and Arabidopsis the observed growth phenotype was caused only by the AtGPS silencing. Nonetheless, expression of five plastidial AtGGPSs was determined in the AtGPS-silenced plants. Expression of AtGGPS7, AtGGPS9, AtGGPS10, AtGGPS11 and AtGGPS12 was not clearly changed (Figure S3). The slightly reduced AtGGPS11 and AtGGPS12 transcript levels might be caused by unknown secondary effects of reduced GPP levels or gibberellin levels. AtGGPS11 and AtGGPS12 are by far the highest expressed GGPSs, and are likely to be involved in chlorophyll and carotenoid biosynthesis, as their expression strongly correlates with the expression of geranylgeranyl reductase and phytoene synthase (At1g74470, At5g17230;, the first dedicated steps in the biosynthesis of these pigments.

How to explain the unaffected pigments in gibberellin-deficient dwarfs after GPS silencing

Previously, GPS was considered to be involved only in the synthesis of monoterpene precursors (reviewed by Dudareva et al., 2004; McGarvey and Croteau, 1995; Pichersky et al., 2006; Tholl, 2006). To our knowledge, involvement of GPS in the synthesis of precursors for larger terpenes has not been investigated in plants so far. Some simplified schemes in reviews and textbooks draw GPP ‘en route’ to FPP or GGPP, whereas there are no existing reports that support this, other than in vitro data showing potential use of GPP as substrate by FPSs or GGPSs. Our results suggest that GPP feeds into the GGPP biosynthesis pathway that leads to gibberellins. GGPSs involved in gibberellin synthesis would require GPP as substrate for condensation of two additional IPP units (Figure 7). Interestingly, the precursor pool for carotenoid and chlorophyll synthesis is not dependent on LeGPS expression (Figure 5b). Therefore, pigments and gibberellins originate from separate GGPP pools, and GGPSs involved in pigment synthesis are not dependent on GPP as substrate (Figure 7).

Figure 7.

 Schematic representation of terpene biosynthesis, focused on synthesis of the prenyldiphosphates Isopentenyl diphosphate (IPP), dimethylallyl diphosphate (DMAPP), geranyl diphosphate (GPP) and geranylgeranyl diphosphate (GGPP) in plastids of the aerial parts of a plant. Three different cell types are schematically indicated, and the synthesis of IPP and DMAPP is placed outside these cells to indicate that this is a common pathway to all cells. Prenyldiphosphates are indicated by rounded boxes and terpene end-products are indicated by filled boxes. Dotted arrows indicate the possible requirement of other prenyldiphosphate synthases or uncertain substrate origins and intermediate steps. CPS, ent-copalyldiphosphate synthase; DOXP, 1-deoxy-d-xylulose 5-phosphate; Kox, ent-kaurene oxidase; KS, ent-kaurene synthase; MEP, methyl-d-erythritol.

Literature also indicates that in the aerial parts, gibberellins and pigments are made in different tissues. Localization of the first dedicated step in gibberellin synthesis (conversion of GGPP to ent-copalyl diphosphate, ent-CPP) was studied using ent-CPP synthase promoter-GUS fusions in Arabidopsis. This revealed that CPS is mostly expressed in provasculature or vasculature tissue of developing embryos, seedlings, and young as well as mature leaves (Sun and Kamiya, 1997; Yamaguchi et al., 2001). In contrast, chlorophyll and carotenoid synthesis occurs mainly in chloroplasts of mesophyll cells (chlorenchym) (Cookson et al., 2003; Reiter et al., 1994) (Figure 7). However, some data published earlier could suggest precursor sharing. Firstly, depletion of the plastidial terpene precursor pool upstream of GGPP by antisense or chemical suppression of IPP synthesis in Arabidopsis leads to both a pigment-deficient and a dwarfed phenotype (Okada et al., 2002). IPP is indeed required for both metabolites. However, in these experiments the IPP pool is lowered in both cell types (vascular and mesophyll tissue), and thus does not provide evidence for the sharing of a GGPP pool. Secondly, ectopic overexpression of a carotenoid biosynthesis gene, phytoene synthase, in tomato plants or Arabidopsis seeds causes precursor depletion, and leads to lower gibberellin levels and dwarfed plants or delayed germination, respectively (Fray et al., 1995; Lindgren et al., 2003). In these cases, the ectopic and therefore ‘mislocalized’ carotenoid biosynthesis apparently depleted the GGPP pool normally used for gibberellin biosynthesis. Therefore, the cell type-specific biosynthesis of gibberellin precursors, as we propose (Figure 7), is in agreement with previously published data.

Furthermore, it has been observed earlier that overexpression of a monoterpene synthase in Arabidopsis led to dwarfing, whereas chlorophyll and carotenoid levels were unaltered (Aharoni et al., 2003). It is known that overexpression of terpene synthases can lead to precursor depletion or product toxicity, and as product toxicity appears as yellowing of leaves with damaged regions (Aharoni et al., 2006), it can be concluded that the dwarfing in Arabidopsis was probably caused by precursor depletion affecting gibberellin levels (Aharoni et al., 2003). The observation that depletion of the GPP pool by the overexpressed monoterpene synthase did not affect pigment levels supports our statements that GPP is an intermediate in gibberellin biosynthesis, and that gibberellins and pigments do not share precursors.

We would like to emphasize that the Arabidopsis genome contains 12 putative GGPSs (Lange and Ghassemian, 2003), which have different organ-specific and developmentally regulated expression patterns (, and their encoded enzymes localize to various subcellular compartments (one mitochondrial, two cytosolic/endoplasmic reticulum and nine plastidial proteins; Lange and Ghassemian, 2003; Okada et al., 2000). Still, it is not known which of the GGPSs are involved in gibberellin synthesis (Figure 7). Synthesis of the gibberellin precursors ent-CPP and ent-kaurene occurs in plastids (Aach et al., 1997; Sun and Kamiya, 1994), thus all plastid-targeted GGPSs are candidates. Differential regulation of GGPS expression also occurs in tomato. The two tomato GGPSs characterized so far both encode putatively plastid-targeted proteins, whereas their expression patterns differ greatly; the LeGGPS1 gene expressed in leaves is specifically regulated in accordance with synthesis of the volatile diterpene-derived homoterpene 4,8,12-trimethyltrideca-1,3,7,11-tetraene (TMTT; Ament et al., 2006). This supports the hypothesis that independent and dedicated GGPP pools exist.

Importantly, we deduce from our results that different GGPSs utilize different allylic substrates with IPP. GGPS presumably generates GGPP by sequential coupling of three IPP molecules to DMAPP (Hedden and Phillips, 2000; Kasahara et al., 2002; Lange, 1998). However, in vitro data show that several GGPSs can also utilize or even have lower Kms for GPP or FPP as allylic substrate (Hefner et al., 1998; Okada et al., 2000; Takaya et al., 2003). Therefore, it is possible that GPP or FPP serve as intermediates for GGPP synthesis.

What other roles may GPS play?

Thus far, the exact role of GPS in di-, tri-, tetra- and/or polyterpene synthesis received little attention and is still unclear. It remains to be investigated whether GPS-silenced plants have other, more subtle phenotypes. We analyzed only gibberellin, carotenoid and chlorophyll levels, and therefore we do not know whether GPS is involved in synthesis of other large terpenes (and derivatives), like plastoquinone, phylloquinone, tocopherols or GGPP, for protein prenylation. Taking into account that GPS could also be localized to the cytosol (Bouvier et al., 2000), one could speculate that GPS is even involved in the synthesis of cytosolic terpenes like sesquiterpenes, steroids and dolichols. Therefore, feeding explants, seedlings or cell suspensions with labeled GPP and analysis of incorporation into various terpene products would be interesting, provided that all cells and subcellular compartments can be reached. It is still an open question if and how the GGPP precursor pools are shared among different terpene metabolites, and how this is regulated. Furthermore, we should take into account that GPS functions might differ between plant species. Firstly, it is not yet clear whether cytosolic localization of GPS or formation of FPP in plastids is common in plants. Secondly, the existence of homomeric as well as heteromeric GPSs suggests independent origins of this protein activity, which makes it likely that GPP is utilized for various classes of terpenes.

Experimental procedures

Plant material and chemicals

Tomato plants (Lycopersicon esculentum) cultivar Moneymaker and Arabidopsis (A. thaliana) ecotype Columbia (Col-0) were used for all experiments. GA3 was obtained from Duchefa ( The Arabidopsis T-DNA insertion line 097 G02 (AL755431; was obtained from the GABI-Kat collection (Rosso et al., 2003). The insertion in AtGPS (At2 g34630) was verified by PCR using the following primers: left-border 5′-CCCATTTGGACGTGAATGTAGACAC-3′ and AtGPS-forward 5′-AATTGGTGACCTAACCAATGTGT-3′.

Isolation of LeGPS and generation of silencing constructs

The full-length tomato LeGPS cDNA (DQ286930; was amplified from leaf cDNA using the primers 5′-ATATCATGATATTTTCAAAGGGTTTAGCTC-3′ and 5′-CCCCTATTTTGTTCTTGTGATGAC-3′, which were designed on the ESTs TC165735 and AI485604 ( For VIGS in tomato, a 350-bp LeGPS fragment was generated using primers 5′-GAGATGATCCATGTTGCTAGCC-3′ and 5′-CACTATGCCCAGCAAGTAGTGC-3′, which was ligated into the TRV-based VIGS vector pYL156 (Liu et al., 2002a) using standard techniques.

The AtGPS-RNAi constructs were made using fragments of At2 g34630 that were amplified from Arabidopsis cDNA with primers containing GatewayTM adapters: fragment 1 (568 bp), 5′-AAAAAGCAGGCTTAGAACAAGTCTTTGCCTCTTTGG-3′ and 5′-AGAAAGCTGGGTAGTGATTTCAGCAATACCCCG-3′; fragment 2 (378 bp), 5′-AAAAAGCAGGCTTAGGATCGTTGTCAGATATTCGC-3′ and 5′-AGAAAGCTGGGTAACGAATTGGACCAACAAAATCC-3′. The PCR products were recombined into the entry vector pDONRTM201 (Invitrogen, and subsequently into the binary RNAi vector pK7GWiWG2(I) (Karimi et al., 2002) using GatewayTM technology according to the manufacturer’s instructions (Invitrogen). All constructs were verified by sequencing. Arabidopsis plants were transformed with the pK7GWiWG2-AtGPS constructs using Agrobacterium tumefaciens (A. tum) GV3101 by means of floral dipping (Clough and Bent, 1998).

AtGPS homology searches

Searches with the AtGPS cDNA for 21-nt homology (identity) stretches in other genes was performed using Blastn (; threshold 10, word size 11, match/mismatch scores 1, –3, gap existence/extension costs 5, 2). The search was validated by creating an artificial sequence composed of AtGPS and several 21-nt sequence fragments of different AtGGPSs pasted in different locations. All 21-nt AtGGPS fragments were recognized by BLASTn. Lowering match stringency by changing the match/mismatch scores to 2, –3 retrieved only a 21-nt homology domain, with one mismatch to a cytochrome P450 of unknown function and a region of low homology (four or more mismatches in 21 nt) to solanesyldiphosphate synthases (required for the synthesis of phylloquinone, vitamin K1, involved in plastidial electron transfer).

Functional characterization of the recombinant LeGPS protein

A truncated LeGPS cDNA was generated and subcloned into pET32a (Invitrogen) using the primers 5′-ATACCATGGTAGTCGCGGAGGTCCC-3′ and 5′-CCCCTATTTTGTTCTTGTGATGAC-3′. The removal of the N-terminal part, which is likely to encode the plastid localization signal, was based on the functional, truncated Arabidopsis GPS protein described previously (Bouvier et al., 2000). Recombinant LeGPS was expressed in E. coli by the use of the Overnight ExpressTM Autoinduction System 1 (Novagen, at 18°C. After purification of the bacterial crude extract with Ni-NTA-agarose columns, determination of GPS enzyme activity, product analysis and product identification were carried out according to the method described by Tholl et al. (2001). The substrates DMAPP and IPP were obained from Echelon Research Laboratories ( and [1-14C]IPP (55 Ci mol−1) was purchased from Biotrend ( After alkaline hydrolysis followed by ether extraction, the products were measured by radio-gas chromatography. All products were analyzed and quantified after co-injection and thermal conductivity detector measurement of unlabeled standards.

VIGS in tomato

Agrobacterium tumefaciens GV3101 (containing pSoup; Hellens et al., 2000) transformed with the pYL156-LeGPS construct (RNA2) and the RNA1 plasmid (Liu et al., 2002a) were grown overnight at 28°C in Luria Bertani (LB) medium supplemented with 50 mg l−1 kanamycin and 25 mg l−1 rifampicin. Cultures carrying RNA2 or RNA1 were diluted 2000 or 500 times, respectively, in LB medium containing 0.5% sucrose, 2 mm MgSO4, 10 mm 2-(N-morpholine)-ethanesulphonic acid (MES) (pH 5.6), 50 mg l−1 kanamycin, 25 mg l−1 rifampicin and 20 μm acetosyringone, and were grown for 12–16 h at 28°C, shaking at 200 rpm until an OD600 of about 1.0. Cells were harvested by centrifugation for 15 min at 3000 g, and were resuspended in Murashige and Skoog medium (Duchefa) with 2% sucrose, 10 mm MES (pH 5.6) and 200 μm acetosyringone to an OD600 of 0.4. RNA2 and RNA1 containing A. tum suspensions were mixed in a 1:1 ratio and left at room temperature (20–25°C) for 1–4 h. Tomato seedlings (12-days old) were carefully taken from the soil and submerged upside down in the A. tum suspension in a tube or beaker and put in a vacuum desiccator. Seedlings were incubated for 1 min under vacuum (5 kPa) and after quickly releasing the vacuum, seedlings were potted and transferred to the greenhouse, which was set at 20°C for 2 days and 25°C thereafter. Internode lengths were measured and leaf samples were taken 35 days post-inoculation (dpi). Dwarf phenotypes were rescued by biweekly spraying with 50 μm GA3, starting from 14 dpi.

Real time Q-RT-PCR

Total RNA was isolated from tomato or Arabidopsis using Trizol (Invitrogen), and DNA contamination was removed with DNAse (Ambion, cDNA was synthesized from 2–5 μg of RNA using SuperscriptII (Invitrogen) in a 20-μl reaction volume that was diluted to 50 μl prior to using it for PCR. PCRs were performed in the ABI 7500 Real-Time PCR System (Applied Biosystems, using the Platinum SYBR Green qPCR SuperMix-UDG kit (Invitrogen). PCR reactions (20 μl) contained 0.25 μm of each primer, 0.1 μl of ROX reference dye and 1 μl of cDNA. The cycling program was set to 2 min 50°C, 5 min 95°C, 40 cycles of 15 sec 95°C and 1 min 60°C, followed by a melting curve analysis. Primer pairs were tested for specificity and for linearity with a standard cDNA dilution curve.



Gibberellin and pigment measurements

Gibberellin levels were analyzed by SIM-GC/MS as described elsewhere (Lange et al., 2005). Endogenous GA levels are given as the sum of C19– GAs GA1, GA4, GA8, GA20, GA29, GA34 and GA51.

Pigments were extracted from 50 mg of ground leaf tissue with 1 ml of 95% ethanol by incubation at room temperature for 10 min in dim light. Extracts were cleared by centrifugation for 1 min at 12 000 g, diluted 10 times in 95% ethanol and measured spectrophotometrically. Chlorophyll and carotenoid contents were calculated according to Lichtenthaler (1987). Non-saturating (and linear) extraction conditions were first determined using different tissue:ethanol ratios. Leaves of photobleached PDS-silenced tomato plants were used as positive controls, which contained about 10-fold less carotenoids and chlorophyll than non-silenced leaves.


The authors thank Peter Hedden and Harro Bouwmeester for useful discussions, Dorothea Tholl for advice on GPS enzyme assays and the lab of Savithramma Dinesh-Kumar for kindly providing the TRV-VIGS vectors.

Accession numbers: DQ286930, At2g34630, AL755431.