Silencing geranylgeranyl diphosphate synthase in Nicotiana attenuata dramatically impairs resistance to tobacco hornworm.

In bioassays with artificial diets, the 17-hydroxygeranyllinalool diterpenoid glycosides (HGL-DTGs) of Nicotiana attenuata function as antifeedants for the plant's adapted herbivore, tobacco hornworm (Manduca sexta). To determine whether HGL-DTGs have a defensive function in planta, we suppressed HGL-DTG production by silencing the source of the geranylgeranyl diphosphates (GGPPs) required for geranyllinalool biosynthesis, a key intermediate. We used virus-induced gene silencing to suppress transcript levels of GGPP synthase gene (Naggpps) and farnesyl diphosphate (FPP) synthase gene (Nafpps), northern blotting and real-time polymerase chain reaction to quantify transcript accumulations, and radio gas chromatography to analyze prenyltransferase specificity. Silencing Nafpps had no effect on the accumulation of HGL-DTGs but decreased leaf steroid content, demonstrating that DTG-synthesizing enzymes do not use GGPP derived from FPP and confirming FPP's role as a steroid precursor. Unlike plants silenced in the phytoene desaturase gene (Napds), which rapidly bleached, Naggpps-silenced plants had reduced HGL-DTG but not carotenoids or chlorophyll contents, demonstrating that Naggpps supplies substrates for GGPP biosynthesis for HGL-DTGs, but not for phytoene or phytol. Expression of Naggpps in Escherichia coli revealed that the recombinant protein catalyzes the GGPP synthesis from isopentenyl diphosphate and dimethylallyl diphosphate. When fed on silenced plants, hornworm larvae gained up to 3 times more mass than those that fed on empty vector control plants or plants silenced in Nafpps, the trypsin protease inhibitor gene, or the putrescine N-methyltransferase gene. We conclude that HGL-DTGs or other minor undetected diterpenoids derived from GGPP function as direct defenses for N. attenuata and are more potent than nicotine or trypsin protease inhibitors against attack by hornworm larvae.

In bioassays with artificial diets, the 17-hydroxygeranyllinalool diterpenoid glycosides (HGL-DTGs) of Nicotiana attenuata function as antifeedants for the plant's adapted herbivore, tobacco hornworm (Manduca sexta). To determine whether HGL-DTGs have a defensive function in planta, we suppressed HGL-DTG production by silencing the source of the geranylgeranyl diphosphates (GGPPs) required for geranyllinalool biosynthesis, a key intermediate. We used virus-induced gene silencing to suppress transcript levels of GGPP synthase gene (Naggpps) and farnesyl diphosphate (FPP) synthase gene (Nafpps), northern blotting and real-time polymerase chain reaction to quantify transcript accumulations, and radio gas chromatography to analyze prenyltransferase specificity. Silencing Nafpps had no effect on the accumulation of HGL-DTGs but decreased leaf steroid content, demonstrating that DTG-synthesizing enzymes do not use GGPP derived from FPP and confirming FPP's role as a steroid precursor. Unlike plants silenced in the phytoene desaturase gene (Napds), which rapidly bleached, Naggppssilenced plants had reduced HGL-DTG but not carotenoids or chlorophyll contents, demonstrating that Naggpps supplies substrates for GGPP biosynthesis for HGL-DTGs, but not for phytoene or phytol. Expression of Naggpps in Escherichia coli revealed that the recombinant protein catalyzes the GGPP synthesis from isopentenyl diphosphate and dimethylallyl diphosphate. When fed on silenced plants, hornworm larvae gained up to 3 times more mass than those that fed on empty vector control plants or plants silenced in Nafpps, the trypsin protease inhibitor gene, or the putrescine N-methyltransferase gene. We conclude that HGL-DTGs or other minor undetected diterpenoids derived from GGPP function as direct defenses for N. attenuata and are more potent than nicotine or trypsin protease inhibitors against attack by hornworm larvae.
Plants employ both direct and indirect defenses against herbivores. Direct defenses may provide physical barriers, act to poison herbivores or impede their digestion, or make plant tissues otherwise unappetizing. There are many ways of identifying the metabolites that function as defenses in plants. The effects on insect herbivores of secondary and primary metabolites such as polyphenols, quinones, alkaloids, some saponines, diterpenoid glycosides (DTGs), phytic acid, and trypsin protease inhibitors (TPIs) have been examined by supplementing artificial diets (De Boer and Hanson, 1987;Broadway and Duffey, 1988;Felton et al., 1989;Snook et al., 1997;Green et al., 2001;Pohlon and Baldwin, 2001;Jassbi et al., 2006). Unfortunately, some chemicals may be altered in the artificial diets (Usher et al., 1989) and the effectiveness of metabolites as defenses may depend on their chemical context (Felton et al., 1989).
Proteins are thought to be the most important nutrients for Lepidoptera herbivores, and some secondary metabolites, such as tannins, gossypol, and TPIs are thought to inhibit digestion of proteins (Damaty and Hudson, 1975;Feeny, 1976;Rhoades and Cates, 1976;Broadway and Duffey, 1988;Barbehenn and Martin, 1994); by binding to proteins, including digestive enzymes, tannins reduce the digestibility of proteins in larval midguts (Feeny, 1976;Rhoades and Cates, 1976;Downs et al., 2003). Many compounds have diverse modes of action. In addition to acting as an antifeedant, phytic acid may reduce the digestibility of foliage and fruits by forming complexes with dietary proteins and prevent the detoxification of toxins by inhibiting cytochrome P450 monooxygenases (Green et al., 2001). Moreover, the defensive function of some metabolites may depend on the herbivore's behavioral responses to other metabolites. In an example of a defensive synergism between metabolites, nicotine was found to be a particularly effective defense in Nicotiana attenuata against the larvae of Spodoptera exigua because the alkaloid prevented the larvae from compensating for the ingestion of TPIs by increasing consumption of leaf material (Steppuhn and Baldwin, 2007).
These complex interactions represent a principal reason why studies of metabolite defensive function in planta are thought to provide more reliable results compared to the use of artificial diet bioassays (Felton et al., 1989). Heterologous gene expression provides a means of examining the defensive role of metabolites in planta (Duan et al., 1996;Smigocki and Wilson, 2004;Zavala et al., 2004b). Introducing a Ser-type PI gene from potato (Solanum tuberosum) into rice (Oryza sativa) resulted in a new insect-resistant rice cultivar (Duan et al., 1996). Although the heterologous expression of a defensive gene in a plant can reduce herbivore performance, two complications can confound the extrapolation of the results to a natural host plant-herbivore interaction. First, specialist herbivores are often resistant to the metabolites of their host plant. Second, the newly introduced metabolite encounters a new chemical environment in the transgenic plant and may interact differently with these metabolites (Broadway, 1997;Winterer and Bergelson, 2001;Zavala et al., 2004b). Endogenous overexpression or silencing of genes that control the accumulation of metabolites provides the best tools for examining defensive function of a metabolite. Silencing a gland-specific P450 hydroxylase gene in the trichomes of Nicotiana tabacum resulted in suppression of biosynthesis of the predominant exudate component, cembratriene (CBT)-diol and increases in its precursor, CBT-ol. Transgenic plants with higher concentrations of the precursor diterpenoid CBT-ol showed greatly diminished aphid colonization, suggesting that the precursor was a more important defense against aphids than the (CBT)-diol (Wang et al., 2001).
Multiple direct defenses are known to function in the native tobacco N. attenuata; the alkaloid nicotine, a neurotoxin, functions as a potent direct defense, while antidigestive TPIs decrease the nutritional value of plants' tissues and slow insects' growth rates (Steppuhn et al., 2004;Zavala et al., 2004aZavala et al., , 2004b. N. attenuata emits volatile terpenoids, among other compounds, when attacked by herbivores; these terpenoids have been shown to attract the herbivores' natural predators and deter herbivore oviposition, serving as effective indirect defenses and reducing herbivore loads by up to 90% (Kessler and Baldwin, 2001). The defensive functions of nicotine and TPIs in N. attenuata against the specialist lepidopteran herbivore, tobacco hornworm (Manduca sexta), have been extensively examined by silencing the expression of key biosynthetic genes in planta (Steppuhn et al., 2004;Zavala et al., 2004b;Steppuhn and Baldwin, 2007). N. attenuata is a phytochemically diverse plant containing alkaloids (nicotine, anabasine, and anatabine), proteins (TPIs), phenolics (rutin, chlorogenic acids, and caffeoylputrescine), volatile terpenoids, and DTGs (Halitschke et al., 2000;Keinänen et al., 2001;van Dam et al., 2001;Lou and Baldwin, 2003;Steppuhn et al., 2004). Here, we examine the role of DTGs in the context of the other major secondary metabolites within the plant.
DTGs with geranyllinalool (GL) carbon skeletons (GL-DTGs) have been reported to be abundant secondary metabolites in several species of Nicotiana (Shinozaki et al., 1996;Snook et al., 1997;Keinänen et al., 2001;Jassbi et al., 2006). Two of the characterized 17-hydroxygeranyllinalool (HGL)-DTGs in N. tabacum showed antibiosis activity and acted as direct defenses inhibiting the growth of tobacco budworm (Heliothis virescens), a major pest on tobacco (Snook et al., 1997). The defensive role of HGL-DTGs against hornworm in Nicotiana was first proposed in a study that compared larval performance on three different species of Nicotiana, N. attenuata, N. bigelovii, and N. clevelandii, and found that larvae gained less mass when they fed on plants containing high HGL-DTG levels (Lou and Baldwin, 2003).
To investigate the defensive role of DTGs in planta and to determine the biosynthetic pathway of HGL-DTGs in N. attenuata, we used virus-induced gene silencing (VIGS) to suppress the expression of the geranylgeranyl diphosphate synthase (GGPPS) gene (Naggpps) and the farnesyl diphosphate synthase (FPPS) gene (Nafpps). VIGS has been extensively used to investigate the function of genes (Liu et al., 2002;Lu et al., 2003;Benedito et al., 2004). It was successfully used to silence both the TPI gene (Natpi) in the leaves and the putrescine N-methyltransferase gene (Napmt) in the roots, suppressing the production of direct defenses TPI and nicotine, respectively (Saedler and Baldwin, 2004). We used Natpi-and Napmt-silenced plants as positive controls for our in planta feeding assays. Phytoene desaturase gene (Napds)-silenced plants bleach rapidly due to suppressed carotenoid production, and these were used as positive controls for the VIGS process (Liu et al., 2002). In addition to its role in the synthesis of diterpenoids, GGPP is also known to be the precursor of carotenoids and the phytol side chain of chlorophylls (Fray et al., 1995;Keller et al., 1998), and we consider the effects of silencing Naggpps on the total chlorophyll and carotenoid contents. Because FPP from Taxus baccata cell cultures is reported to be the substrate for GGPPS (Laskaris et al., 2000), we silenced Nafpps to clarify its role in the biosynthesis of HGL-DTGs.

VIGS of Naggpps and Nafpps
Approximately 20 d after inoculating early rosettestage N. attenuata plants with Agrobacterium tumefaciens GV3101 harboring pTVFPPS, pTVGGPPS, pTV00 (empty vector control [EV]), and pTVPD plasmids, Napds-silenced plants became completely bleached, demonstrating successful gene silencing. To determine the silencing efficiency of Naggpps and Nafpps, total RNA was isolated from the leaves of the transformed plants and subjected to northern analysis (Fig. 2, insets); transcripts of both genes were substantially decreased in gene specific-silenced lines in comparison to EV controls.
Because the silencing efficiency of the Nafpps gene was only 69% and several homologs of fpps and ggpps have been reported in Artemisia tridentata and Arabidopsis (Arabidopsis thaliana; Hemmerlin et al., 2003;Lange and Ghassemian, 2003), we tested whether N. attenuata has more than one homologous copy of each gene. If so, this could influence the silencing, and if this gene has a redundant function, all other experiments might also be affected.
After construction of pTVFPPS and pUCFPPS, the Nafpps inserts of at least four clones of each plasmid were sequenced. There was complete identity in the amplified Nafpps sequence between the PCR primers among the clones of each plasmid. For primary cloning of three different Naggpps PCR fragments in pTVGGPPS, pTVGGPPS2, and pUCGGPPS and of the full-length Naggpps on pJETGGPPS, four independent PCRs with eight different primers on cDNA of N. attenuata as template were performed. After successful cloning, the Naggpps inserts of 14 clones (two of pTVGGPPS, two of pTVGGPPS2, six of pUCGGPPS, and four of pJETGGPPS) were sequenced. There were no differences in the amplified sequences between the respective PCR primers among all 14 clones. Because the primers were designed from sequence information from other Solanaceous plants (tomato [Solanum lycopersicum], Capsicum annuum, and N. tabacum), complete identity between primers and template could not be expected. For instance, a comparison of the full-length Solanaceous ggpps sequences available in GenBank reveals an 85% identity of the 3# 900 bp of Naggpps (EF382626) to tomato ggpps (DQ267903) and a 79% identity to C. annuum ggpps (X80267). As a consequence, a few sequence differences between Naggpps and the clones in the primer regions occurred, showing that in our PCR system, primer/Naggpps template mismatches still allowed for efficient PCR amplification. Therefore, the amplification of only one Nafpps and only one Naggpps gene under conditions that allow primer/ template mismatches represents strong evidence that there is only one copy of Nafpps and of Naggpps and no family of genes homologous to Nafpps and Naggpps in the genome of N. attenuata.
Southern-blot analysis confirmed that the N. attenuata genome harbored only one close homolog of fpps and ggpps (Supplemental Fig. S1, A and B); using a fulllength Naggpps probe and a 350-bp Nafpps probe, all four digestions revealed only a single band. Together with the PCR fragment-sequencing results, these results lead us to conclude that there is no other close homolog of Nafpps and Naggpps in N. attenuata.
Plant prenyltransferases are only highly similar in their amino acid sequences of the catalytic domain. In a phylogenetic tree of all prenyltransferases comparing the amino acid sequences of the complete open reading frame, FPPS cluster separately from the more related GGPPS or geranyl pyrophosphate synthase (GPPS; Burke and Croteau, 2002;Tholl et al., 2004;Bouvier et al., 2005;Schmidt and Gershenzon, 2007b). In contrast to FPPS, where it is possible to predict the putative enzyme activity, functional analyses of GGPPS are necessary to confirm enzyme activity. Because silencing Naggpps did not significantly change the levels of synthesized GGPP in the plants' total protein, we heterologously expressed full-length cDNA in Escherichia coli to determine if the cDNA of Naggpps codes for a functional GGPPS. After the putative chloroplast transit peptide was truncated, NaGGPPS functioned mainly as a GGPPS and showed only a minor amount of FPPS activity. The total synthesized and hydrolyzed prenyldiphosphate products were composed of 85% 6 3.1% GG and 15% 6 2.7% farnesol. The recombinant protein showed no GPPS activity (Fig. 3).
To determine the effects of silencing Naggpps and Nafpps on secondary metabolite composition, a methanol extract of leaf material from silenced plants was analyzed by reverse-phase (RP)18 HPLC (Supplemental Fig. S2). Levels of HGL-DTGs were lower in pTVGGPPS plants than in pTVFPPS and pTV00 plants (Supplemental Fig. S2; Fig. 4, dotted bars). The average levels of HGL-DTGs in the herbivore-attacked leaves of pTVGGPPS and pTVFPPS plants were 33.6% (n 5 10; mean relative area [MRA] % 5 99.2 6 12.3; t 5 10.118; Figure 2. VIGS efficiency of Naggpps (A) and Nafpps (B). RT-PCR was run on cDNA from leaf samples of plants treated with pTV00, pTVFPPS, and pTVGGPPS. Transcript accumulation levels were determined relative to the N. attenuata actin gene and presented as the mean value (6SE) of five biological replicates. ***, Significantly different from pTV00 at P , 0.0001. Insets: RNA was isolated and pooled from leaves of two biological replicates for each line pTV00, pTVFPPS, and pTVGGPPS, hybridized with radio-labeled probes for Naggpps (A) and Nafpps (B), washed and analyzed with a Fujifilm FLA-3000 Phosphoimager. Lower lanes: Ethidium bromide-stained agarose gel with rRNA as a control for equal RNA loading. To compare the performance of larvae that fed on plants deficient in HGL-DTGs with those deficient in TPIs and nicotine, we repeated the experiments with the addition of Natpi-and Napmt-silenced plants. By day 16, the masses of larvae that fed on pTVGGPPS were significantly (n 5 16; AM 5 232.9 6 9.4 mg; t 5 7.003; df 5 28; P , 0.0001) higher than the masses of those that fed on EV plants (n 5 14; AM 5 100.2 6 17.2 mg). Masses of larvae that fed on pTVFPPS (n 5 9; AM 5 93.3 6 13.4 mg), pTVPMT (n 5 14; AM 5 128.2 6 32.0 mg), and pTVTPI (n 5 14; AM 5 102.0 6 17.6 mg) were not significantly different from those that fed on EV plants (Fig. 5A). As in the first experiment, the number of larvae that survived to the end of the feeding bioassay was higher on pTVGGPPS plants (16, or 59%) than on the EV plants, where only 14 (52%) of 27 larvae survived.
Analysis of TPI activity revealed similar levels and kinetics of inhibition among pTVGGPPS, pTVFPPS, pTVPMT, and pTV00 plants; as expected, TPI activity was negligible in the pTVTPI plants (Fig. 5D). Like levels of the other defensive substances, levels of TPIs increased 2.3-to 8.3-fold in the silenced and EV plants after 11 d of larval feeding.

Silencing Nafpps Decreases Free Sterol Levels But Not Volatile Terpenoids
To investigate the effect of silencing Nafpps on terpenoid biosynthesis, the head spaces of whole plants were analyzed for volatile sesquiterpenoids. No significant differences (all P values . 0.1) were detected in their levels in comparison to other lines (data not shown). We assumed that Nafpps was involved in steroid biosynthesis, so the concentrations of the free steroids, b-sitosterol and stigmasterol, were measured in dichloromethane extracts of fresh leaves by GC-mass spectrometry (MS). Significant decreases were found in the levels of b-sitosterol (73% of the pTV00 plants; t 5 3.099; df 5 14; P 5 0.0079) and stigmasterol (79% of the pTV00 plants; t 5 2.721; df 5 14; P 5 0.0165) in pTVFPPS plants compared to their levels in EV plants (Fig. 6). Silencing Naggpps had no significant effect on the levels of free steroids in pTVGGPPS plants compared to those in EV plants.

Effects of Silencing Prenyltransferases on Photosynthetic Pigments
To evaluate the effect of silencing prenyltransferases on photosynthetic pigments, leaf chlorophyll contents were measured during the feeding bioassay with a portable chlorophyll meter (Supplemental Fig. S3). Total extractable chlorophylls (Chl.a 1 Chl.b) and carotenoids were measured in 80% aqueous acetone extracts of leaves with a UV/VIS spectrophotometer (Supplemental Fig. S4, A and B). Total chlorophyll and carotenoid content did not differ between silenced and pTV00 plants in the same treatment group (Supplemental Figs. S3 and S4). However, the total carotenoid levels of herbivore-attacked plants were significantly lower than in unattacked plants (Supplemental Fig.  S4B).

DISCUSSION
Previous work with N. attenuata plants silenced in their ability to produce nicotine and TPIs suggested that the plants produce additional potent defense metabolites (Steppuhn and Baldwin, 2007). When inverted-repeat putrescine N-methyltransferase-silenced N. attenuata plants growing in a field plantation in the Great Basin Desert were elicited by methyl jasmonate (MeJA), the damage these plants received from native herbivores was as low as that of wild-type plants, suggesting the presence of other potent MeJA-inducible defenses (Steppuhn et al., 2004). The results from this study suggest that HGL-DTGs or other minor undetected diterpenoids derived from GGPP are these powerful defenses. Although artificial diet bioassays have indicated that HGL-DTGs have a defensive function (Snook et al., 1997;Jassbi et al., 2006), their defensive function had not been previously explored by manipulating in planta production, the gold standard of proof of defensive function. We silenced two prenyltransferase genes that were potentially involved in HGL-DTG biosynthesis, Naggpps and Nafpps, and compared the growth of hornworm larvae on the silenced lines with their growth on EV plants. In a second study, we used Natpi-and Napmt-silenced plants to compare the antifeedant potential of the well-studied defensive compounds, PIs and nicotine, with that of HGL-DTGs. Because silencing ggpps can reduce the levels of all diterpenoids, we examined the presence of other diterpenoids such as cembrane-and labdane-type diterpenoids, which are found in N. tabacum (Wang et al., 2001(Wang et al., , 2004Wang and Wagner, 2003) by means of GC-MS, HPLC, and NMR analyses of the chloroform extract of the plant, but we could not detect any; the different GL-DTGs were the major diterpenoid constituents of the plant that were extractable by methanol and water (Jassbi et al., 2006).
N. attenuata plants, which were inoculated with A. tumefaciens GV3101 carrying pTVFPPS, pTVGGPPS, pTVPMT, and pTVTPI plasmids, grew the same as EV plants. RT-PCR analyses of the total RNA extracted from leaves of the gene-silenced plants revealed that Naggpps and Nafpps are 92% and 69% silenced in their transcript levels, respectively (Fig. 2, A and B). However, the unexpected up-regulation of Nafpps after silencing Naggpps (Fig. 2B) can only be explained when our understanding of the pathways for the synthesis of different metabolites is more complete. Apparently, silencing prenyltransferase synthesis leads to the accumulation of certain metabolites and activates different pathways that increase the level of FPPS transcripts.
Radio-GC analyses of the radiolabeled synthesized and hydrolyzed prenyldiphosphates in plants' protein extracts confirmed the reduction of FPPS activity in pTVFPPS plants. Although the levels of synthesized GG were quite low and not significantly reduced after silencing Naggpps, heterologous expression of Naggpps in E. coli and analyses of its expressed product confirmed that NaGGPPS functions exclusively as a GGPPS (Fig. 3). The lower levels of synthesized GG in plant protein extracts indicated that the products of NaGGPPS do not accumulate and are instead rapidly used as a substrate for subsequent reactions such as the synthesis of HGL-DTGs.
We found that silencing Naggpps dramatically reduced HGL-DTG levels (Figs. 4, dotted bars, and 5B) but not levels of the other defensive metabolites, nicotine, and PIs (Fig. 5, C and D). Similarly, silencing Napmt or Natpi did not strongly affect the concentration of HGL-DTGs (Fig. 5B). Although the levels of induced HGL-DTGs in pTVFPPS and pTVPMT plants were less than those of EV plants after 11 and 14 d of herbivory (Fig. 5B), the differences in the levels of the suppressed metabolites (HGL-DTGs and nicotine) in the respective silenced plants (pTVGGPPS and pTVPMT) compared to those in EV plants were always significantly higher than those in the other silenced lines (Fig. 5, B and C; P , 0.0001). Silencing Natpi resulted in a line (pTVTPI) that was free of TPI activity. Overall, we conclude that transforming N. attenuata with pTVGGPPS, pTVPMT, and pTVTPI plasmids successfully silenced the production of HGL-DTGs, nicotine, and TPIs, respectively, without affecting the other plant metabolites, including carotenoids and chlorophylls. Therefore, these plants could be used to compare the defensive value of each metabolite in a plant-herbivore interaction. The dramatically increased performance of hornworm larvae feeding on pTVGGPPS plants argues that DTGs are a more effective defense against this herbivore than are TPIs and nicotine.
Although we did not detect significant effects of TPI and nicotine on larval growth in our experiment (Fig.  5A), the defensive function of nicotine and TPIs against hornworm herbivory has been well established by gene-silencing experiments with stably transformed lines of N. attenuata and Nicotiana sylvestris (Voelckel et al., 2001;Steppuhn et al., 2004;Zavala et al., 2004b). The experiments reported here use VIGS to reduce nicotine and TPI levels, and VIGS is known to elevate salicylic acid levels. Because both nicotine and TPIs utilize the jasmonic acid (JA)-signaling cascade for their induction after herbivore attack (Halitschke and Baldwin, 2003), it is possible that salicylic acid/JA antagonism and the resulting decrease in these defense metabolites could have reduced our ability to detect Figure 5. VIGS of Naggpps dramatically reduces HGL-DTG contents and increases hornworm larvae performance to a greater degree than the silencing of TPI and nicotine. A, hornworm larval mass (milligrams; mean 6 SE; n 5 27 in the beginning of the assay) from N. attenuata plants inoculated with A. tumefaciens GV3101 carrying pTV00, pTVGGPPS, pTVFPPS, pTVTPI, and pTVPMT. B, Relative area percentage of HGL-DTGs (mean 6 SE; n 5 6; RP-HPLC-UV peak area at 22-27 min/area of thymols' peak 3 100, l 5 210 nm), nicotine (mean 6 SE; n 5 6; mg g 21 leaf FW; C), and TPI activity (mean 6 SE; n 5 5; nmol mg 21 protein; D). Nicotine and DTGs were measured in 40% methanolic extracts of the leaves of VIGS plants. *, Significantly different values from pTV00 with P , 0.05. negative effects on larval growth. Moreover, hornworm tolerates doses of nicotine that unadapted insects find fatal (Snyder et al., 1994), and it is possible that other primary metabolites, such as proteins or carbohydrate levels altered by the VIGS system, may have minimized the influence of nicotine and TPIs on larval growth. Low nitrogen (N) supply attenuated JAinduced TPI and nicotine levels in N. attenuata (Lou and Baldwin, 2004), which points to the importance of plant growth conditions in influencing defensive function of particular metabolites.
N. attenuata germinates in the postfire environment and copes with large changes in soil N during postfire succession. As the N content of the soil in natural habitat of N. attenuata remains high for up to 3 years after a fire (Lynds and Baldwin, 1998), plants can take advantage of the high N levels to support the biosynthesis of N-intensive defenses, such as TPIs and nicotine; when the N content of soil decreases, plants can use another effective carbon-intensive defense, such as HGL-DTGs. Analyses of HGL-DTGs, nicotine, and TPIs (Fig. 5, B-D) revealed that after 11 d, all of these direct defenses were highly induced by larval feeding and reached their maximum concentration in the leaves of the plant. Increasing the levels of direct defenses encourages herbivores to move to new plants or leaves, helping the plant not to lose its entire canopy, and perhaps reduces competition of conspecific neighboring plants (Van Dam et al., 2000;Lou and Baldwin, 2003;Zavala and Baldwin, 2004;Paschold et al., 2007).
The lower levels of HGL-DTGs only in Naggppssilenced plants suggest that GGPP is a precursor of HGL-DTG diterpenoids in N. attenuata (Fig. 1). In addition to its role in the biosynthesis of diterpenoids and carotenoids, GGPP is required to produce phytol, the side chain of chlorophylls. Carotenoids provide accessory pigments for photosynthesis; when they degrade, plants become photobleached (Fray et al., 1995;Keller et al., 1998). Silencing Naggpps, in contrast to silencing Napds, did not result in bleaching. The product of the silenced Naggpps gene is therefore likely involved in the synthesis of the precursor of diterpenoids rather than in the synthesis of carotenoids or chlorophylls' prenyl side chain. Levels of carotenoids were, however, decreased in all plants after herbivore attack (Supplemental Fig. S4B), consistent with the commonly observed down-regulation of photosynthesis-related genes in N. attenuata and other plants in response to herbivore attack (Bi and Felton, 1995;Hermsmeier et al., 2001;Heidel and Baldwin, 2004). Down-regulation of the carotenoids after herbivore attack is clearly not Naggpps dependent and may be controlled by other genes involved in carotenoid biosynthesis.
Because the amplification of full-length Naggpps and three different fragments of it from N. attenuata cDNA always yielded exactly the same Naggpps sequence and because a Southern blot with chromosomal N. attenuata DNA (Supplemental Fig. S1) suggested the presence of only one copy of Naggpps, we conclude that the N. attenuata genome harbors only one copy of Naggpps. This conclusion does not exclude the existence of additional nonhomologous ggpps genes in N. attenuata or the possibility of complex regulation or compartmentalization of the synthesis of GGPP, as is seen in Arabidopsis and tomato (Bartley and Scolnik, 1995;Keller et al., 1998;Lange and Ghassemian, 2003;Ament et al., 2006). This separation may be vital for plants that use different GGPPS to synthesize either primary metabolites, such as chlorophyll and carotenoids, necessary for plant growth, or induced defensive secondary metabolites. Some GGPPS may additionally catalyze the formation of GPP, as inferred from the high homology between the two enzymes (Gershenzon and Kreis, 1999;Burke and Croteau, 2002;Tholl et al., 2004;Bouvier et al., 2005). However, heterologous expression of plant GGPPS or GPPS in E. coli provides a means of clearly differentiating among product profiles (Engprasert et al., 2004;Gershenzon, 2007a, 2007b). The expression of Naggpps in E. coli and the analyses of its expressed product confirmed that Naggpps functions exclusively as a GGPPS (Fig. 3).
The ggpps gene in tomato is reportedly induced by both JA and methyl salicylate. However, the product of the gene, (E,E)-4,8,12-trimethyltrideca-1,3,7,11tetraene, is induced mainly by JA (Ament et al., 2006). (E,E)-4,8,12-trimethyltrideca-1,3,7,11-tetraene is the dehydration product of GL, and because GL synthase levels in plants induced by JA are higher than in plants induced with methyl salicylate, GL synthase is also thought to be induced mainly by JA, especially because GL levels increase dramatically when spider mites attack the plant (Ament et al., 2006). As in tomato, the induction of HGL-DTGs in N. attenuata by MeJA and insect feeding indicates that the GL biosynthesis and induction pathways may be similar in both species. Like linalool and nerolidol, which are the allylically rearranged products of geraniol and farnesol, respectively, GL can be derived from an allylic rearrangement in GG. This conversion has been shown to be catalyzed by GL synthase, which was assayed in tomato (Ament et al., 2006). As reported for plaunotol (18-hydroxy geranylgeraniol), further hydroxylation at C-17 followed by glycosylation at C-3 and C-17 hydroxyls can produce HGL-DTGs in Nicotiana (Nualkaew et al., 2006; Fig. 1). From the above observations, we can conclude that at least four kinds of enzymes are involved in the biosynthesis of HGL-DTGs from GGPP in N. attenuata: GGPP phosphatase, which converts GGPP to GG; GL synthase, which converts GG to GL; GL-17hydroxylase, which hydroxylates the Me-17 of the GL; and glycosylating enzymes, which convert HGL to different HGL-DTGs (Ament et al., 2006;Nualkaew et al., 2006).
Silencing Nafpps did not significantly affect the levels of HGL-DTGs, the performance of feeding larvae, or the chlorophyll and the carotenoid contents of the plants (Figs. 4 and 5;Supplemental Figs. S3 and S4), suggesting that NaGGPPS does not use FPP as a substrate and most probably catalyzes the condensation of three IPPs together with one DMAPP to produce GGPP ( Fig. 1) and that the FPP produced by the Nafpps gene product is not involved in forming GGPP for carotenoid synthesis. To explore the effect of Nafpps and Naggpps silencing on volatile terpenoids, the headspaces of all of the silenced plants were analyzed, but no significant differences were observed between volatile sesquiterpenoids levels in any lines (data not shown).
The lower levels of free b-sitosterol and stigmasterol in dichloromethane extracts of leaves from pTVFPPS plants confirmed that Nafpps is involved in steroid synthesis in N. attenuata ( Fig. 6; Brown, 1998). The presence of multiple fpps genes in A. tridentata, each with about 80% identity, suggested that the encoded isozymes function specifically. For example, one FPPS (FDS-1) has a major role in sesquiterpene production, while another (FDS-2) maintains a pool of FPP for other branches of the pathway (Hemmerlin et al., 2003). A Southern blot with chromosomal N. attenuata DNA (Supplemental Fig. S1) revealed the existence of only one copy of Nafpps. N. attenuata may contain additional but nonhomologous fpps genes that were not silenced by our VIGS construct. This may be one reason why we did not detect differences in volatile sesquiterpene production between pTV00 and pTVFPPS plants.
In summary, we found that silencing Naggpps had the greatest effect on larval performance and we conclude that it is N. attenuata's most effective defense gene against hornworm larvae. The gene product showed only GGPPS activity and synthesizes a precursor of HGL-DTGs. The agreement of these results with prior artificial diet feeding assays confirms the utility of a combination of in planta and bioassay approaches for identifying new defense metabolites. As Naggpps is not involved in primary metabolite synthesis and HGL-DTGs do not incorporate nutrients such as N that commonly limit growth, overexpressing Naggpps could increase the resistance of N. attenuata to hornworm herbivory without affecting the growth of the plant. Given the evidence from this and other studies for a strict separation of prenyltransferases involved in primary metabolism from those involved in secondary metabolism, and given the ubiquity of terpene synthesis, such an approach may also prove useful in agricultural plants. However, as has been shown in N. tabacum where the silencing of a P450 hydroxylase gene suppressed the accumulation of hydroxylated cembranoids (Wang et al., 2001(Wang et al., , 2004Wang and Wagner, 2003), silencing geranyllinalool synthase and GL-17hydroxylase in N. attenuata may represent a more specific means of suppressing HGL-DTGs.

Construction of Plasmids
The primer sequences used are listed in Supplemental Table S1. Cloning was done in Escherichia coli strains TOP10 and TOP10F' (Invitrogen, http:// www.invitrogen.com). For the VIGS experiments, pTV00 derivatives were used to transform Agrobacterium tumefaciens strain GV3101.
To silence Nafpps, the PCR fragment obtained with primer pair FPPS3-34 and FPPS4-33 was cloned as a BamHI/SalI fragment in pTV00, leading to silencing vector pTVFPPS (5.9 kb). Plasmid pUCFPPS was constructed by cloning the fragment synthesized with primers FPPS1-33 and FPPS2-34 in the NcoI and PstI sites of pUCPDS5. The GenBank accession numbers of the cloned Nafpps fragments are EF382631 and EF382632. The construction of the vectors pTVTPI (5.7 kb) and pTVPMT (6.2 kb) used for silencing the N. attenuata genes Napmt and Natpi was performed as described previously (pTVPMT2 and pTVPI2 in Saedler and Baldwin, 2004) by cloning the PCR fragments obtained with primer pairs TRV5-31/TRV6-31 (0.7-kb Napmt1 product) and TRV9-34/ TRV10-36 (0.2-kb Natpi product) in the BamHI and HindIII sites of pTV00. pTVPD used for silencing Napds is a pTV00 derivative carrying a 206-bp fragment of the Nicotiana benthamiana phytoene desaturase (AJ571700) and was kindly provided by D. Baulcombe of the Sainsbury laboratory, Norwich, UK.

Plant Growth and VIGS
Smoke-treated N. attenuata seeds originating from glasshouse-grown generations of a collection originating from Utah were germinated on phytagel agar (Sigma, http://www.sigmaaldrich.com) as described previously and transferred to soil after about 10 d (Baldwin et al., 1994). All plants were grown under a 28°C/24°C 16-/8-h light/dark regime until the early rosette stage. The plants were transferred to a growth chamber under 20°C/17°C 6 1°C 16-/8-h light/dark at 70% 6 10% humidity for the VIGS inoculation. VIGS was used to silence the Naggpps gene, utilizing the TRV genome resident on the pTV00derived plasmids and on plasmid pBINTRA6 (Ratcliff et al., 2001;Saedler and Baldwin, 2004). Approximately 6 d after plants were transferred to the growth chamber, they were coinoculated with A. tumefaciens GV3101 carrying pTVPD, pTVFPPS, pTVGGPPS, pTVPMT, or pTVTPI, and with the same Agrobacterium strain carrying pBINTRA6 according to the modified Ratcliff method (Ratcliff et al., 2001;Saedler and Baldwin, 2004). About 20 d after inoculation, the Napdssilenced plants bleached thoroughly, demonstrating successful silencing.

Larval Rearing and Feeding Bioassay
Larvae of the tobacco hornworm (Manduca sexta; Lepidoptera, Sphingidae) were hatched overnight at 28°C from eggs received from Carolina Biological Supply (http://www.carolina.com). One newly hatched larva of hornworm was released on a leaf of each silenced and nonsilenced (empty vector) plant when the corresponding leaf in the Napds-silenced plants (positive control) had completely bleached. A total of 40 and 27 replicate plants from each line were used in the bioassay in the first and second VIGS experiments, respectively (Figs. 4 and 5A). To measure larval performance, larvae were weighed after 4, 14, and 18 d in the first experiment (larval masses after 14 d of feeding are presented in Fig. 4) and after 4, 8, 11, 14, and 16 d of feeding in the second experiment (Fig. 5A). The insect-feeding bioassays were performed in the chambers and under the same conditions as the VIGS experiments. In the first VIGS experiment, Naggpps-and Nafpps-silenced plants were used; in the second experiment, Napmt-, Natpi-, Naggpps-, and Nafpps-silenced plants were used.

RNA Extraction, Northern Blotting, and RT-PCR
Samples were ground in liquid N with a mortar and pestle. Total RNA was extracted with TRI ReagentTM (Sigma) according to the manufacturer's instructions. RNA quality was checked on a denaturing 1% agarose gel (5% formaldehyde) and concentrations were measured spectrophotometrically at 260, 280, and 320 nm. RNA samples were stored at 280°C until use.
To determine the efficiency of silencing Naggpps and Nafpps genes, we conducted northern-blot analysis and used quantitative PCR analysis. RNA samples (10 mg, pooled from three individually grown plants) were size fractionated by 1.2% (w/v) agarose-formaldehyde gel electrophoresis and blotted onto a nylon membrane (GeneScreenPlus; PerkinElmer, http:// www.perkinelmer.com) according to the manufacturer's protocol. Ethidium bromide fluorescence was used as the loading control. After blotting and UV cross-linking, blots were prehybridized for 1 h at 42°C in Ultrahyb hybridization buffer (Ambion, http://www.ambion.com) followed by overnight hybridization with the 32 P-labeled probes. The blots were washed once, 15 min at 60°C with 23 SSC, 0.1% (w/v) SDS, followed by twice at 60°C in 0.13 SSC and 0.1% (w/v) SDS, 20 min each. Blots were exposed for 24 h on a phosphoimage film (Fujifilm, http://www.fujifilm.com). The signals were read by fluorescent image analyzer FLA-3000 (Fujifilm) and quantified with Aida Image Analyzer software (Raytest, http://www.raytest.com). The 313-bp PCR fragment obtained with primer pair GGPPS7-32/GGPPS8-33 and pUCGGPPS as template served as a probe for Naggpps (Supplemental Table S1). The level of Nafpps transcript was detected using probe PCR synthesized with primer pair FPPS1-33/FPPS2-34 and template pUCFPPS (Supplemental Table S1).
For RT-PCR analysis, five replicated biological samples were used. One microgram of total RNA obtained from each sample was reverse transcribed using oligo(dT) and Superscript II reverse transcriptase (Invitrogen) for a total volume of 20 mL. cDNA samples were further diluted with water to 40 mL, and 1 mL of the diluted cDNA was used for RT-PCR, carried out on a ABI PRISM 7700 Sequence Detection system (Applied Biosystems, http://www.appliedbiosystems.com) using qPCRTM Core kits (Eurogentec, http://www.eurogentec.com). For each analysis, a linear standard curve, threshold cycle number versus log (designated transcript level), was constructed using a series dilution of a specific cDNA standard; the transcript levels in all unknown samples were determined according to the standard curve. The N. attenuata actin gene was used as an internal standard for normalizing cDNA concentration variations. The sequences of the primers used for SYBR Green-based RT-PCR (the N. attenuata actin gene, Actin-F1 and Actin-R1; Nafpps, FPPS FOR and FPPS REV; Naggpps, GGPPS FOR, GGPPS REV) are provided in Supplemental Table S1.

Southern Blotting
Genomic DNA was extracted from N. attenuata young leaves using the cetyl-trimethyl-ammonium bromide method (Doyle and Doyle, 1987). Quality and concentration were checked by agarose gel electrophoresis. Genomic DNA samples (5 mg) were digested with AseI, BamHI, BclI, or DraI (New England Biolabs, http://www.neb.com), size fractionated on a 0.8% (w/v) agarose gel, and blotted onto a nylon membrane (GeneScreenPlus; Perkin Elmer). Labeling, washing, and analytical procedures were done according to those used for northern blotting. The Naggpps probe was PCR synthesized with primer pair GGPPS10-26/GGPPS11-31 and pJETGGPPS as template and comprised the complete coding sequence of the gene. The Nafpps probe was the same used for northern blotting.

Heterologous Expression of Naggpps in E. coli
To express Naggpps in E. coli, a fragment of this gene was amplified using the Expand High Fidelity PCR system (Roche Applied Science, http:// www.roche-applied-science.com) with pJETGGPPS as template, thereby removing the coding sequence of the potential chloroplast transit peptide. The cleavage site of this peptide was predicted using the chloroP program (http:// www.cbs.dtu.dk). With primer pair GGPPS14-25 and GGPPS13-24, a 0.9-kb fragment was synthesized; after an ATG start codon, the fragment contained codon 54 of EF382626. The fragment was cloned with T-overhangs in plasmid pCR-T7 CT TOPO (2.7 kb; Invitrogen) yielding Naggpps expression vector pGGPPS (3.6 kb). Expression strain E. coli BL 21(DE3) pLys S (Invitrogen) was transformed with pGGPPS. Bacterial cultures expressing Naggpps were grown using Overnight Express Autoinduction system 1 (Novagen, http://www. novagen.com) as described by the manufacturer's instructions but at 18°C. Bacterial pellets were resuspended in assay buffer without DTT and sonicated. The His-tagged recombinant protein was purified with nickel-nitrilotriacetic acid agarose columns (Qiagen, http://www.qiagen.com) according to the manufacturer's instructions. The recombinant protein was eluted with 250 mM imidazole in the assay buffer. After adding DTT, fractions were checked by SDS-PAGE and used to determine enzyme activity.

Functional Characterization of NaGGPPS
NaGGPPS assays were carried out in a final volume of 500 mL containing 20 mM 3-(N-Morpholino)-2-hydroxypropanesulfonic acid, pH 7.0, 10 mM of MgCl 2 , 10% (v/v) glycerol, and 2 mM DTT. Assays were carried out in triplicates of at least three biologically independent experiments with 40 mM [1-14 C]IPP (2 MBq/mmol; Biotrend, http://www.biotrend.com) and 40 mM DMAPP (Echelon Biosciences, http://www.echelon-inc.com). After the reaction was initiated by adding recombinant protein, the assay mixture was covered with 1 mL pentane and incubated overnight at 30°C. To stop the assay and hydrolyze all diphosphate esters, a 1-mL solution with 2 units of calf intestine alkaline phosphatase (Sigma) and 2 units of potato (Solanum tuberosum) apyrase (Sigma) in 0.2 M Tris-HCl, pH 9.5 was added to each assay and incubated at 30°C overnight. After enzymatic hydrolysis, the resulting prenyl alcohols were extracted into 2 mL of diethyl ether and, after the addition of a standard terpene mixture (geraniol, farnesol, and geranylgeraniol), the organic extracts were evaporated under N 2 and used for radio-GC measurements.
Radio-GC analysis was performed on a Hewlett-Packard (http://www. hp.com) HP6890 gas chromatograph equipped with a thermal conductivity detector (TCD) and Raga radioactivity detector (Raytest) and using a DB5-MS capillary column (J&W Scientific, http://www.jandw.com; 30 m 3 0.25 mm i.d., 0.25-mm film thickness). The oven temperature was set 3 min at 70°C, followed by a gradient from 70°C to 240°C at 6°C min 21 and kept 3 min at the final temperature. The carrier gas was H 2 with a flow rate of 2 mL min 21 . Injector and TCD detector temperatures were maintained at 220°C and 250°C. The injection volume of the concentrated organic phase was 1 mL. The synthesized products, measured by radio-GC, were identified by comparing their retention times with those of coinjected authentic nonradioactive terpene standards, detected via the TCD. Protein concentrations in enzyme assays were measured according to Bradford (Bradford, 1976) using the Bio-Rad reagent with bovine serum albumin as standard. The protein concentration in each assay was equilibrated to a range of 50 to 100 mg mL 21 .

Analysis of Protease Inhibitors
A total of 150 mg of fresh leaf material of the plants was harvested, ground in liquid N, and extracted with 0.3 mL extracting buffer; subsequently, samples were subjected to radial diffusion assays as described previously  and expressed as nmol mg 21 of leaf protein.

Analysis of Free Sterols
A total of 300 mg of the systemic leaf plant material, collected 14 d after herbivore treatment in the first VIGS experiment, was extracted as above but with 2 mL dichloromethane. The extracts (nine replicates for each silencing construct) were dried under an N 2 stream, redissolved in 100 mL of dichloromethane, and subjected to GC-MS analyses. The GC-MS was carried out on a Varian 3800 gas chromatograph coupled to a Varian Saturn 2000 mass spectrometer operating in EI mode at 70 eV (http://www.varianinc.com). The GC equipped to a DB-5 MS (J&W Scientific column, 30-m 3 0.25-mm i.d., 0.25-mm film thickness). The oven temperature was programmed from 285°C to 310°C at 5°C min 21 and kept for 10 min at 310°C. The carrier gas was helium with a flow rate of 1 mL min 21 . The injector temperature was set at 310°C in splitless mode. The injection volume was 1 mL for all of the samples. Two major steroids, b-sitosterol and stigmasterol, were identified by comparing their retention times and mass spectrums with those of the authentic standards (from Fluka, http://www.sigmaaldrich.com, and Sigma).

Total Chlorophyll and Carotenoid Measurements
The chlorophyll contents of the plants' leaves were monitored by a portable SPAD-502 chlorophyll meter (Konica Minolta, http://www.konicaminolta. com) and expressed as a SPAD value (Netto et al., 2002). To measure the extractable chlorophylls and carotenoids, 100 mg of N 2 -frozen leaf tissues, collected 14 d after herbivore treatment, were extracted with 1 mL of 80% acetone for 30 min and diluted 10 times. Absorptions of the solutions at 664, 647, and 470 nm were measured using an Ultraspec 300 UV/VIS spectrophotometer. Concentrations of total carotenoids and chlorophylls were calculated according to the method described in Lichtenthaler (1987).

Statistical Analyses
The Student's t tests were performed using the algorithm embedded in StatView (http://www.imp.com).
Sequence data from this article can be found in the GenBank/EMBL data libraries under accession numbers EF 382626, EF 382631, and EF 382632.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Southern-blot analysis of Naggpps and Nafpps genes in N. attenuata.
Supplemental Figure S2. RP-HPLC chromatograms of the methanolic extracts of the fresh leaves of N. attenuata plants silenced with A. tumefaciens GV3101 containing pTV00 (A), pTVGGPPS (B), and pTVFPPS (C), and structures of the HGL-DTGs (D) found in N. attenuata.
Supplemental Figure S3. Total chlorophyll contents expressed as SPAD (named after the company's Soil-Plant Analysis Division, Konica Minolta) mean values of three replicate leaves of nine plants (6SE, n 5 9) at different time intervals after seedlings were measured with a portable chlorophyll meter.
Supplemental Figure S4. Spectrophotometer-calculated concentration (mean 6 SE, n 5 5, mg g 21 FW of leaves) of total chlorophyll (A), Chl b (l 5 647 nm), and Chl a (l 5 664 nm), and total carotenoids (B; l 5 470 nm) in 80% acetone extract of the leaves of VIGS plants after 14 d of herbivory (herbivore-attacked plants) and the plants of the same age but not attacked by herbivores.
Supplemental Table S1. Sequences of the primers used for the construction of plasmids, northern blotting, Southern blotting, and RT-PCR.