Two nearly identical terpene synthases catalyze the formation of nerolidol and linalool in snapdragon flowers


*(fax +1 765 494 0391; e-mail


Terpenoids emitted from snapdragon flowers include three monoterpenes derived from geranyl diphosphate (GPP), myrcene, (E)-β-ocimene and linalool, and a sesquiterpene, nerolidol, derived from farnesyl diphosphate (FPP). Using a functional genomics approach, we have isolated and biochemically characterized two nearly identical nerolidol/linalool synthases, AmNES/LIS-1 and AmNES/LIS-2, two enzymes responsible for the terpenoid profile of snapdragon scent remaining to be characterized. The AmNES/LIS-2 protein has an additional 30 amino acids in the N-terminus, and shares 95% amino acid sequence identity with AmNES/LIS-1, with only 23 amino acid substitutions distributed across the homologous regions of the proteins. Although these two terpene synthases have very similar catalytic properties, and synthesize linalool and nerolidol as specific products from GPP and FPP, respectively, they are compartmentally segregated. GFP localization studies and analysis of enzyme activities in purified leucoplasts, together with our previous feeding experiments, revealed that AmNES/LIS-1 is localized in cytosol, and is responsible for nerolidol biosynthesis, whereas AmNES/LIS-2 is located in plastids, and accounts for linalool formation. Our results show that subcellular localization of bifunctional enzymes, in addition to the availability of substrate, controls the type of product formed. By directing nearly identical bifunctional enzymes to more than one cellular compartment, plants extend the range of available substrates for enzyme utilization, thus increasing the diversity of the metabolites produced.


Terpenoids are the largest and most functionally and structurally diverse class of plant secondary metabolites. They are not only vital for numerous basic plant processes, such as photosynthesis, respiration, growth and development, but also for the fitness of a plant (Gershenzon and Kreis, 1999; Rodriguez-Concepcion and Boronat, 2002). Volatile terpenoids released from vegetative tissues defend plants against herbivore and pathogen attacks, and protect against reactive oxygen species, whereas their emission from flowers promotes plant–pollinator interactions (Dudareva et al., 2006). To date, 556 volatile terpenoids have been identified in the floral scent of many plant species, with monoterpenes representing the most abundant constituents, followed by sesquiterpenes, irregular terpenes and diterpenes (Knudsen and Gershenzon, 2006).

All terpenoids originate from isopentenyl diphosphate (IPP) and its allylic isomer dimethylallyl diphosphate (DMAPP), which are derived from two alternative pathways (Figure 1). In the cytosol, IPP is synthesized from acetyl-CoA (Newman and Chappell, 1999) by the mevalonic acid (MVA) pathway (McCaskill and Croteau, 1995), whereas in plastids, it is derived from pyruvate and glyceraldehyde-3-phosphate via the methyl-erythritol-phosphate (MEP) pathway (Lichtenthaler, 1999; Rohmer, 1999). Although the subcellular compartmentation allows these pathways to operate independently, metabolic ‘crosstalk’ between them was previously discovered (Schuhr et al., 2003), particularly from plastids to the cytosol (Dudareva et al., 2005; Hemmerlin et al., 2003; Laule et al., 2003).

Figure 1.

 Metabolic pathways leading to terpenoid compound biosynthesis in snapdragon flowers.
Abbreviations: DMAPP, dimethylallyl diphosphate; DXP, 1-deoxy-d-xylulose-5-phosphate; DXR, DXP reductoisomerase; DXS, DXP synthase; FPP, farnesyl diphosphate; FPPS, FPP synthase; GA-3P, glyceraldehyde-3-phosphate; GGPP, geranylgeranyl diphosphate; GGPPS, GGPP synthase; GPP, geranyl diphosphate; GPPS, GPP synthase; HMG-CoA, 3-hydroxy-3-methylglutaryl-CoA; HMGR, 3-hydroxy-3-methylglutaryl-CoA reductase; IPP, isopentenyl diphosphate; MEP, 2-C-methyl-d-erythritol-4-phosphate; MVA, mevalonic acid. The solid arrow from the plastid to the cytosol shows trafficking of IPP in snapdragon (Dudareva et al., 2005).

In both subcellular compartments, prenyltransferases use IPP and DMAPP to produce prenyl diphosphates. In the cytosol, the condensation of two IPP molecules and one DMAPP molecule catalyzed by the enzyme farnesyl diphosphate synthase (FPPS) results in FPP formation (C15), the natural precursor of sesquiterpenes (McGarvey and Croteau, 1995). In plastids, a head-to-tail condensation of one IPP molecule and one DMAPP molecule catalyzed by geranyl diphosphate synthase (GPPS) forms GPP (C10), the universal precursor of all monoterpenes (Ogura and Koyama, 1998; Poulter and Rilling, 1981). Condensation of one DMAPP molecule with three IPP molecules by the action of geranylgeranyl diphosphate synthase (GGPPS) yields GGPP, the C20 diphosphate precursor of diterpenes (Koyama and Ogura, 1999). Following the formation of the acyclic precursors GPP, FPP and GGPP, a wide range of structurally diverse cyclic and acyclic monoterpenes, sesquiterpenes and diterpenes is generated through the action of a large family of terpene synthases/cyclases (TPSs; Cane, 1999; Wise and Croteau, 1999).

Snapdragon flowers emit three monoterpenes derived from GPP, myrcene, (E)-β-ocimene and linalool, as well as the sesquiterpene (E)-nerolidol, derived from FPP (Dudareva et al., 2003, 2005). Emission of these compounds follows diurnal rhythms, and is controlled by a circadian clock. Two specialized monoterpene synthases responsible for synthesizing myrcene and (E)-β-ocimene, respectively, have been isolated and functionally characterized (Dudareva et al., 2003). However, there is no information about linalool and nerolidol biosynthesis in snapdragon flowers. The formation of these compounds could follow at least two scenarios, and require two specialized TPSs, linalool synthase and nerolidol synthase, or be the result of the action of a bifunctional nerolidol/linalool synthase, as was recently reported in strawberry (Aharoni et al., 2004). It was also proposed that linalool biosynthesis in ripe strawberry fruit occurs in the cytosol (Aharoni et al., 2004; Hampel et al., 2006), in contrast to the generally accepted plastidial origin of monoterpenes, thereby questioning the subcellular compartmentalization of linalool formation in the petal tissue of snapdragon flowers. Here, we show that snapdragon possesses two nearly identical bifunctional TPSs, which are responsible for linalool and nerolidol biosynthesis in flowers. Biosynthesis of these compounds is compartmentally separated, with linalool biosynthesis occuring in plastids and nerolidol biosynthesis occuring in the cytosol. Our results also provide evidence that the subcellular localization of the enzyme and the availability of the substrate determine the type of product formed.


Isolation of terpene synthase genes responsible for linalool and nerolidol biosynthesis in snapdragon flowers

A search of a snapdragon expressed sequence tag (EST) database, containing >25 000 random sequences from different plant organs (Bey et al., 2004), for clones with sequence similarity to known TPSs resulted in five contigs corresponding to transcripts of two similar genes. Two clones, ama1i06 and ama2j20, designated as AmNES/LIS-1 and AmNES/LIS-2 (for Antirrhinum majus nerolidol/linalool synthase; see below), respectively, represent full-length cDNAs of 1885 and 1898 nucleotides, and share 97% identity. These cDNAs encode proteins of 566 and 596 amino acids (aa), with calculated molecular weights of 65 218 and 68 466 Da, respectively (Figure 2a). The protein encoded by AmNES/LIS-2 has an additional 30 aa in its N-terminus, and shares 95% aa sequence identity with AmNES/LIS-1. Both proteins exhibit ∼62% aa identity to snapdragon myrcene and (E)-β-ocimene synthases, as well as ∼50% and 43% identity to strawberry nerolidol/linalool synthases (Aharoni et al., 2004) and Arabidopsis linalool synthase (Chen et al., 2003), respectively (Figure 2a). Like other monoterpene synthases from snapdragon, AmNES/LIS proteins belong to the TPS-g subfamily (Figure 2b; Dudareva et al., 2003), which lacks the RRx8W motif, a key feature of all monoterpene synthases of the Tps-b and Tps-d groups (Aubourg et al., 2002; Bohlmann et al., 1998b).

Figure 2.

 Relatedness of AmNES/LIS proteins to other terpene synthases.
(a) Alignment of AmNES/LIS-1 (GenBank accession no. EF433761) and AmNES/LIS-2 (GenBank accession no. EF433762) predicted amino acid sequences with snapdragon myrcene synthase (AAO41726), (E)-β-ocimene synthase (AAO42614), strawberry nerolidol/linalool synthase (FaNES; CAD57106) and Arabidopsis linalool synthase (AAO85533). The alignment was performed using ClustalW, and was shaded with BoxShade version 3.21 software (Human Genome Sequencing Center, Residues in black are conserved (identical in at least four out of six sequences shown), whereas residues in gray are similar in at least two of the six sequences shown. Dashes indicate gaps inserted for optimal alignment. The horizontal line indicates the N-terminal peptide, which was shown to be involved in AmNES/LIS-2 protein targeting to plastids. The conserved DDxxD motif is shown in the box. Arrowheads indicate positions of Met-31 and Met-79 in AmNES/LIS-2 where truncations were made. Asterisks show 23 amino acid differences between AmNES/LIS-1 and AmNES/LIS-2.
(b) Phylogenetic tree illustrating AmNES/LIS-1 and AmNES/LIS-2 as members of the Tps-g subfamily. Subfamilies are divided based on cluster analysis (Aubourg et al., 2002; Bohlmann et al., 1998b). The unrooted neighbor-joining tree was created using ClustalX and TreeView for visualization. Only a subset of the known plant terpene synthases (TPSs) is shown: Artemisia annua linalool synthase (AAF13356); Abies grandis (−)-4S-limonene synthase (AF006193), myrcene synthase (U87908), pinene synthase (U87909), terpinolene synthase (AF139206), β-phellandrene synthase (AF139205), γ-humulene synthase (U92267) and δ-selinene synthase (U92266); Arabidopsis thaliana myrcene/ocimene synthase (AF178535); Clarkia breweri linalool synthase (U58314); Cichorium intybus germacrene A synthase (AF497999); Citrus limon (−)-β-pinene synthase (AF514288), (+)-limonene synthase (AF514287) and γ-terpinene synthase (AF514286); Cucurbita maxima copalyl diphosphate synthase (AF049905) and ent-kaurene synthase (U43904); Gossypium arboreum (+)-δ-cadinene synthase (U23205); Lycopersicon esculentum copalyl diphosphate synthase (AB015675) and germacrene C synthase (AF035630); Mentha piperita (E)-β-farnesene synthase (AF024615); Mentha spicata 4S-limonene synthase (L13459); Nicotiana tabacum 5-epi-aristolochene synthase (L04680); Picus abies linalool synthase (AAS47693); Perilla citridora limonene synthase (AF241790) and linalool synthase (AY917193); Perilla frutescens linalool synthase (AF444798); Salvia officinalis (+)-bornyl diphosphate synthase (AF051900), (+)-sabinene synthase (AF051901) and 1,8-cineole synthase (AF051899); Solanum tuberosum vetispiradiene synthase (AB022598); Zea mays TPS1 (AAO18435).

Functional characterization of snapdragon AmNES/LIS proteins

For functional characterization of AmNES/LIS proteins, the coding region of each gene was subcloned into an expression vector, and encoded proteins were expressed in Escherichia coli and analyzed for TPS activities. Although the crude bacterial extract containing AmNES/LIS-1 protein showed activity with both GPP and FPP, no activity was found in AmNES/LIS-2. The absence of activity could be caused by inclusion body formation as a result of transit peptide presence, a typical feature of monoterpene synthases (Bohlmann et al., 1998b; Williams et al., 1998). Although in silico analysis of the AmNES/LIS-2 N-terminal sequence did not give a clear indication of the presence of a transit peptide, truncations were made before the second (Met-31) and third (Met-79) methionines (indicated by arrowheads in Figure 2a), and the corresponding proteins were expressed in E. coli. As both truncated AmNES/LIS-2 proteins efficiently converted GPP and FPP to products, it is possible that the N-terminal 30 aa are responsible for inclusion body formation.

Recombinant purified AmNES/LIS-1, Met-31-AmNES/LIS-2 and Met-79-AmNES/LIS-2 proteins predominantly catalyzed the formation of linalool from GPP (Figure 3a,c, data for Met-79-AmNES/LIS-2 not shown), and of (E)-nerolidol from FPP (Figure 3b,d). In addition, trace levels of (E)-β-ocimene (<1%), myrcene (<0.5%) and α-pinene (<0.15%) were formed from GPP, and β-farnesene (<0.4%), α-farnesene (<0.3%) and α-bisabolol (<0.12%) were formed from FPP. To determine the absolute configuration of linalool and nerolidol produced by AmNES/LIS proteins as well as snapdragon flowers, enantioselective GC-MS analysis was performed. Both enzymes exclusively synthesized (S)-linalool from GPP (Figure 4c,e) and (3S)(E)-nerolidol from FPP (Figure 4d,f), with the same conformation of terpenoids found for both in the scent of the flower (Figure 4g,h).

Figure 3.

 Analyses of products formed by AmNES/LIS enzymes from geranyl diphosphate (GPP) and farnesyl diphosphate (FPP).
(a–f) GC-MS analysis of products formed by recombinant AmNES/LIS-1 (a, b) and AmNES/LIS-2 (c, d) from GPP (a, c) and FPP (b, d), respectively. Reaction products were obtained by solid-phase microextraction (SPME). Inserts with numbered peaks in a–f represent mass spectra of corresponding products. Retention times and mass spectra of synthesized linalool (a, c) and nerolidol (b, d) are identical to those of corresponding authentic standards (e, f, respectively); m/z, mass-to-charge ratio.
(g, h) GC-MS chromatograms of the products formed by extracts of Escherichia coli expressing the pET-32 vector without an insert (control) in the presence of GPP (g) and FPP (h), respectively.

Figure 4.

 Chiral analysis of linalool and nerolidol emitted by snapdragon flowers and formed by AmNES/LIS enzymes.
(a, b) Racemic mixtures of linalool and nerolidol. Peaks 1 and 2, (3R)- and (3S)(Z)-nerolidol (the elution order is unknown); peaks 3 and 4, (3R)- and (3S)(E)-nerolidol, respectively.
(c, e) (S)-linalool produced by AmNES/LIS-1 (c) and AmNES/LIS-2 (E) from GPP.
(d, f) (3S)-(E)-nerolidol produced by AmNES/LIS-1 (d) and AmNES/LIS-2 (f) from FPP.
(g, h) (S)-linalool and (3S)-(E)-nerolidol emitted by snapdragon flowers.

To test for the presence of nerolidol and linalool synthase activities in snapdragon, crude protein extracts from upper and lower petal lobes of 5-day-old flowers, a developmental stage with the highest terpenoid emission (Dudareva et al., 2003), were assayed with GPP and FPP. GC-MS analysis revealed that the sesquiterpene (E)-nerolidol was produced upon incubation with FPP (Figure 5b), whereas three monoterpenes, linalool (12.9% of total product), (E)-β-ocimene (71.9%) and myrcene (9.3%), along with minor quantities of (Z)-β-ocimene, were formed from GPP (Figure 5a).

Figure 5.

 Products formed by snapdragon petal crude extracts and purified leucoplasts from geranyl diphosphate (GPP) and farnesyl diphosphate (FPP).
Solid-phase microextraction (SPME)-gas chromatographs of products formed from GPP (a) and FPP (b) by petal crude extracts of 5-day-old flowers. Scanning electron microscopy of purified leucoplasts (c and d). Scale bars: 0.5 μm (c, d). SPME-gas chromatographs of products produced from GPP (e) and FPP (f) by purified leucoplasts. Peaks are myrcene (a), (Z)-β-ocimene (b), (E)-β-ocimene (c), linalool (d) and nerolidol (e).

Biochemical characterization of purified recombinant AmNES/LIS1 and Met-31-AmNES/LIS-2 proteins revealed that they exhibit similar properties and require, for activity, a divalent cation co-factor, Mg2+or Mn2+. Both linalool and nerolidol synthase activities showed a preference for Mg2+, and were 2.9- and 6.7-fold higher in the presence of Mg2+ than Mn2+, respectively. Mn2+ inhibited corresponding activities by 44% and 70%, respectively, in the presence of Mg2+ (6 mm). Although maximum TPS activity was achieved at 2.0–8.0 mm Mg2+, further increases in Mg2+ to 20 mm led to a 20% reduction in the maximum velocity. The apparent Km values for Mg2+ were 1.16 and 1.06 mm in the presence of GPP, and 0.32 and 0.34 mm in the presence of FPP, for AmNES/LIS-1 and AmNES/LIS-2, respectively (Table 1).

Table 1.   Kinetic parameters of AmNES/LIS proteins
GeneSubstrateKmm)Vmax (pcat mg−1)kcat (sec−1)kcat/Km (mm−1 sec−1)
  1. All values represent mean ± SE, n = 3. Km, the Michaelis constant; Vmax, the maximal velocity; Kcat, turnover number of enzyme.

  2. aAt 6 mm Mg2+.

  3. bAt 8 mm Mg2+.

  4. cAt 43 μm GPP.

  5. dAt 72 μm FPP.

AmNES/LIS-1GPPa7.57 ± 2.76314.68 ± 12.900.021 ± 0.00082.75 ± 0.076
FPPb24.13 ± 1.231493.12 ± 41.330.099 ± 0.0034.11 ± 0.19
Mg2+ (GPP)c1156 ± 62265.18 ± 3.160.017 ± 0.0020.015 ± 0.007
Mg2+ (FPP)d316 ± 2.91395.12 ± 18.120.092 ± 0.0010.292 ± 0.006
AmNES/LIS-2GPPa7.68 ± 2.53341.65 ± 8.210.022 ± 0.00052.94 ± 0.061
FPPb22.68 ± 0.661257.19 ± 26.750.083 ± 0.0023.64 ± 0.03
Mg2+ (GPP)c1066 ± 16294.58 ± 9.540.019 ± 0.00060.018 ± 0.003
Mg2+ (FPP)d339 ± 5.41247.32 ± 8.390.082 ± 0.0050.243 ± 0.005

AmNES/LIS proteins were active over a broad range of pH values from 5.0 to 9.0, with maximum activity at pH 8.0. The molecular mass of 66.1 kDa for the recombinant AmNES/LIS-1 was determined by gel filtration chromatography, and was consistent with the predicted molecular mass (65.2 kDa), indicating that the active enzyme is a monomer. Kinetic characterization of AmNES/LIS-1 and AmNES/LIS-2 using a range of FPP (3–110 μm) and GPP (2–65 μm) concentrations resulted in very similar apparent Km values of 7.57 and 7.68 μm for GPP, and 24.13 and 22.68 μm for FPP, respectively (Table 1). Both enzymes show similar catalytic efficiencies (kcat/Km ratio) with GPP, which were slightly lower than with FPP (Table 1).

Identification of AmNES/LIS in snapdragon flowers

To verify the involvement of AmNES/LIS enzymes in linalool and (E)-nerolidol formation in snapdragon flowers, proteins were partially purified from petal lobes of 4- to 6-day-old flowers using several chromatographic steps. As AmNES/LIS proteins use GPP and FPP as substrates, fractions during purification were analyzed for both monoterpene and sesquiterpene synthase activities, which we were unable to separate, as they eluted together. Three final fractions with the highest TPS activities were combined and run on SDS-PAGE, which revealed a single protein band of ∼60 kDa. (Figure S1). Protein sequencing of this band and surrounding proteins within 1 cm by LC-MS/MS, followed by screening a snapdragon EST database comprising >25 000 sequences, revealed six peptides corresponding to the predicted AmNES/LIS protein sequences, and fourteen peptides matching the protein sequences of the previously characterized myrcene and ocimene synthases (Dudareva et al., 2003). No peptides corresponding to other TPSs were found in the sequencing output, providing clear evidence that these TPSs are responsible for terpenoid emission in snapdragon flowers.

Subcellular localization of AmNES/LIS proteins

The biochemical characterization of AmNES/LIS proteins showed that they can potentially participate in linalool formation, a process that is believed to take place in the plastids, based on the presence of a transit peptide in the protein (Crowell et al., 2002; Jia et al., 1999), or in the cytosol, as was proposed in strawberry through labeling experiments (Hampel et al., 2006). To determine the site of linalool biosynthesis in snapdragon, leucoplasts were purified from petal lobes of 5-day-old flowers (Figure 5c,d), and were assayed with GPP. The isolated leucoplasts were free from cytosolic contamination, as assessed using marker enzyme assays specific for subcellular compartments (Figure S2), and did not contain detectable levels of linalool and nerolidol prior to the addition of prenyl diphosphate substrates. GC-MS analysis of the products revealed that linalool was one of the monoterpenes produced from supplied GPP (Figure 5e), suggesting that it is made in plastids.

Software (ChloroP,; TargetP,; PWoLF PSORT,; Predotar,; SignalP, was used to predict the likelihood of targeting signals in the N-termini of AmNES/LIS proteins. Based on all program predictions, AmNES/LIS-1 did not contain a targeting signal and was localized in the cytosol. However, no consistent conclusions could be drawn about the subcellular localization of AmNES/LIS-2, as three of the search algorithms predicted its cytosolic localization, whereas the ChloroP and Predotar programs predicted the presence of a plastid-targeting sequence with 49% and 77% probability, respectively. To experimentally determine the subcellular localization of AmNES/LIS proteins, coding regions of each gene were fused to a GFP reporter gene, and were then transferred to Arabidopsis protoplasts, where corresponding transient GFP expression was analyzed by confocal laser scanning microscopy (Figure 6). GFP fluorescence was observed in the cytosol of protoplasts expressing AmNES/LIS-1-GFP (Figure 6e–l), confirming AmNES/LIS-1 cytosolic localization.

Figure 6.

 Confocal laser scanning microscopy of transiently expressed AmNES/LIS-GFP fusions in Arabidopsis protoplasts.
GFP fusion constructs are shown on the left, and the corresponding transient expression in the protoplast is shown on the right; TP, transit peptide. Green fluorescent protein fluorescence detected in the green channel is shown in the ‘Green’ column; chlorophyll autofluorescence detected in the red channel is shown in the ‘Red’ column; the ‘Merged’ column shows combined red and green channels; the ‘Transmission’ column shows light-microscopy images of intact protoplasts. The results of two independent experiments are shown for each fusion construct, except for the controls (GFP and RbTP-GFP). RbTP (plastidial RuBisCO target peptide)-GFP is a chloroplast marker. Scale bars: 50 μm. The numbers in the fusion constructs correspond to amino acid positions.

Unlike AmNES/LIS-1-GFP, the AmNES/LIS-2-GFP fusion protein exhibited a punctate and discrete pattern associated with the chloroplasts (Figure 6m–t), indicating its plastidial location. Elimination of the first 30 aa in AmNES/LIS-2 changed the subcellular localization of the protein from plastid to cytosol (Figure 6u–b1). When the first 30 aa of AmNES/LIS-2 were fused to GFP, the fluorescence was detected only in the cytosol (data not shown), suggesting that these 30 aa are essential but not sufficient for targeting the protein to the chloroplast. However, when the first 69 aa of this protein were fused to GFP, the GFP was directed to chloroplasts, and its fluorescence was localized evenly throughout the plastids (Figure 6c1–j1). The pattern of accumulation of the 69-aa(AmNES/LIS-2)-GFP in chloroplasts resembled that of GFP fused to the transit peptide of RuBisCO small subunit, which is a stromal protein (Figure 6k1–n1). These results clearly show that the AmNES/LIS-2 protein contains a transit peptide that mediates its chloroplast targeting. In vitro protein import experiments further confirmed the subcellular localization of AmNES/LIS proteins (data not shown).

As AmNES/LIS-2 is localized in plastids and can catalyze the formation of both linalool and nerolidol (Figure 3c,d), we checked the presence of sesquiterpene synthase activity in leucoplasts. Purified leucoplasts produced nerolidol from FPP (Figure 5f), suggesting that in addition to linalool synthase activity (Figure 5e), nerolidol synthase activity exists in plastids. Both activities were also found in the cytosol.

Analysis of expression of AmNES/LIS-1 and AmNES/LIS-2 genes

The mRNA expression of AmNES/LIS genes was analyzed in different floral tissues and petals during flower development by quantitative real-time PCR with gene-specific primers. The expression of both genes was highest in the upper and lower petal lobes (Figure 7a), which are the parts of the flower that have previously been shown to be primarily responsible for scent emission in snapdragon (Dudareva et al., 2000, 2003). AmNES/LIS transcripts were also detected at lower levels in tubes, followed by pistils, stamens and ovaries (Figure 7a). A negligible level of transcripts was found in leaves and sepals.

Figure 7.

AmNES/LIS mRNA expression, linalool and nerolidol synthase activities, and emission of corresponding compounds in snapdragon flowers.
Semiquantitative (gels) and quantitative RT-PCR (graph) analysis of AmNES/LIS-1 and AmNES/LIS-2 expression in different floral tissues of 5-day-old snapdragon flowers (a) and petal lobes during flower development from mature buds from 1 day before opening to 12 days postanthesis (b). Abbreviations: L, young leaves; UL and LL, upper and lower petal lobes, respectively; T, tubes; P, pistils; S, stamens; O, ovaries; SE, sepals. AmNES/LIS-1 and AmNES/LIS-2 mRNA expression in leaves (a) and buds (b) was set to 1 to determine the relative abundances of mRNA transcripts. Each point is an average of between three and five independent experiments. Vertical bars indicate standard error values. 18S rDNA was used as an endogenous reference in semiquantitative and quantitative RT-PCR.
(c) Changes in linalool and nerolidol synthase activities in snapdragon petal lobes during flower development. For each point, enzyme assays were run in duplicate on two independent crude extracts and standard deviations were obtained.
(d) Nerolidol and linalool emission from snapdragon flowers. Headspace collections were performed at 24-h intervals. Each point represents an average of between three and five independent experiments. Vertical bars indicate standard deviations.

During flower development, mRNA levels for both genes in petal lobes gradually increased during the first 3 or 4 days, peaking on days 3 and 5 postanthesis for AmNES/LIS-2 and AmNES/LIS-1, respectively, and they remained relatively stable thereafter for AmNES/LIS-2, but slightly decreased for AmNES/LIS-1 (Figure 7b). TPS activities in petal lobes were also developmentally regulated. Nerolidol synthase activity peaked on day 5 and decreased by 50% in old flowers (10–12 days postanthesis), whereas linalool synthase activity peaked on day 5 and remained stable (Figure 7c). Emission of linalool and nerolidol positively correlated with the corresponding TPS activities (P < 0.0001 for both activities) only in young flowers, in which emission of both compounds peaked sharply on day 5–6 postanthesis (Figure 7d). Although the positive correlation between linalool synthase activity and AmNES/LIS-2 transcript abundances was observed over the entire flower development, nerolidol synthase activity correlated with AmNES/LIS-1 expression only up to day 9 postanthesis (Figure 7b,c). These results suggest that regulation of enzyme activities occurs largely at the level of gene expression, and, in the case of nerolidol synthase activity, post-transcriptional or translational/post-translational controls also contribute to regulation in 10- to 12-day-old flowers. High levels of TPS activities in mature and old flowers (day 7–12 postanthesis), with low to no concomitant emission of volatile products, indicate that the levels of available substrates in the cell are also involved in the regulation of the biosynthesis and emission of linalool and nerolidol.


Two bifunctional nerolidol/linalool synthases are involved in linalool and nerolidol biosynthesis in snapdragon flowers

Linalool and nerolidol formation in plant cells could either require two specialized TPSs, linalool and nerolidol synthase, or be the result of a bifunctional nerolidol/linalool synthase. Linalool synthases have been isolated and functionally characterized from a wide range of plant species, including the angiosperms Clarkia breweri (Pichersky et al., 1995), Artemisia annua (Jia et al., 1999), Mentha citrata (Crowell et al., 2002), Arabidopsis thaliana (Chen et al., 2003), Lycopersicon esculentum (van Schie et al., 2007) and the gymnosperm Picea abies (Martin et al., 2004). These linalool synthases produce linalool as the sole or predominant product from GPP, and possess an N-terminal transit peptide that directs their import into plastids. Four linalool synthases were examined for their ability to use FPP as a substrate. In contrast to Artemisia and Clarkia linalool synthases, which were inactive with FPP (Aharoni et al., 2004; Jia et al., 1999), Arabidopsis and tomato linalool synthases were able to convert FPP to nerolidol, which was not found in their tissues. Whereas in Arabidopsis nerolidol formation occurred at a rate that was 50-fold lower than the rate of formation of linalool from GPP (Chen et al., 2003), no information about the efficiency of nerolidol biosynthesis relative to linalool formation was provided in tomato (van Schie et al., 2007).

Enzymes reported to be bona fide nerolidol synthases have only been characterized from maize and strawberry. (E)-Nerolidol was the only product formed from FPP in vitro by strawberry FaNES enzymes (Fragaria ananassa nerolidol synthase; Aharoni et al., 2004). Maize TPS1 produced nerolidol as one of the products (29% of total product), in addition to (E)-β-farnesene (26%) and (E,E)-farnesol (45%; Schnee et al., 2002). Both enzymes were able to catalyze linalool formation from GPP in vitro, with FaNES forming both linalool and nerolidol with equal efficiency (Aharoni et al., 2004), and with TPS1 being four times less active with GPP than with FPP (Schnee et al., 2002).

To discover genes responsible for scent production in snapdragon flowers, and to understand the molecular mechanisms that control volatile emission, we have identified and characterized two nearly identical (95% aa identity) nerolidol/linalool synthases, AmNES/LIS-1 and AmNES/LIS-2, which were the two remaining uncharacterized enzymes responsible for the terpenoid profile of snapdragon scent. These enzymes share ∼50% identity with FaNES proteins, and, similar to FaNES, catalyze the formation of linalool or nerolidol, depending on the type of precursor used (GPP for monoterpenes and FPP for sesquiterpenes; Figure 3).

The recombinant AmNES/LIS proteins produced in E. coli exhibit very similar catalytic efficiency towards GPP and FPP, although with a slight preference for FPP (Table 1). The apparent Km value of snapdragon nerolidol/linalool synthases for GPP is three times lower than that for FPP (∼7.6 versus ∼24 μm for GPP and FPP, respectively; Table 1), in contrast to FaNES1, which has an apparent Km for GPP that is 3.6 times higher than that for FPP (29.0 versus 8.1 μm for GPP and FPP, respectively; Aharoni et al., 2004). Although the apparent Km values of the AmNES/LIS enzymes for GPP and FPP are relatively high, when compared with those of monoterpene synthases for GPP (the low μm range) and sesquiterpene synthases for FPP (0.1–10 μm), the turnover rates (kcat) with both substrates are within the range reported for monoterpene (0.01–1.0 sec−1) and sesquiterpene (0.03–0.5 sec−1) synthases (Cane, 1999; Wise and Croteau, 1999). Moreover, apparent Km values of AmNES/LIS enzymes for GPP were lower than those reported for linalool synthases from M. citrata (56 μm; Crowell et al., 2002) and A. annua (64 μm; Jia et al., 1999). The ability of TPSs to accept both GPP and FPP is not unique to linalool/nerolidol synthases, and has been well documented in sesquiterpene synthases from Mentha piperita (Crock et al., 1997), Malus domestica (Green et al., 2007; Pechous and Whitaker, 2004), Abies grandis (Bohlmann et al., 1998a; Steele et al., 1998), L. esculentum (Colby et al., 1998), A. annua L (Mercke et al., 1999), A. thaliana (Tholl et al., 2005), O. basilicum (Iijima et al., 2004) and Zea mays (Kollner et al., 2004). However, their sesquiterpene synthase activities significantly exceeded their monoterpene synthase activities.

Linalool and nerolidol biosynthesis in snapdragon is compartmentally segregated

It is generally accepted that monoterpene biosynthesis takes place in plastids, where GPP is synthesized, whereas sesquiterpene formation occurs in the cytosol, where FPP is formed (Aharoni et al., 2005). Thus, compartmental separation of precursors for monoterpene and sesquiterpene biosynthesis would prevent a bifunctional monoterpene/sesquiterpene synthase from using both of its activities, unless it is localized in both compartments. However, recent metabolic engineering in Arabidopsis, tobacco and potato revealed the existence of small levels of FPP in plastids and GPP in the cytosol (Aharoni et al., 2003, 2006; Wu et al., 2006). The presence of substrate for monoterpene and sesquiterpene biosynthesis in both plastids and cytosol, although at different levels (high GPP/trace FPP in plastid and vice versa in cytosol), would allow a dual functional enzyme to produce monoterpenes and sesquiterpenes in both compartments, but at a ratio representative of that of the precursors. Indeed, overexpression of strawberry FaNES1 in plastids of Arabidopsis and potato resulted in the formation of small quantities of nerolidol, which was 100- to 300-fold lower than the quantity of linalool that was produced (Aharoni et al., 2003, 2006).

Here, we show that snapdragon flowers emit linalool and nerolidol, the levels of which are developmentally regulated, and are of a similar order of magnitude (Figure 7d). These compounds are synthesized by two nearly identical AmNES/LIS proteins that exhibit different subcellular localization. AmNES/LIS-1 is localized in the cytosol and is responsible for nerolidol biosynthesis, whereas AmNES/LIS-2 has a transit peptide directing it to plastids, and accounts for linalool formation (Figure 6). Nerolidol formation from FPP and linalool from GPP by purified leucoplasts also confirmed the presence of both monoterpene/sesquiterpene synthase activities in plastids (Figure 5e,f). The coexistence of AmNES/LIS-1 and AmNES/LIS-2 enzymes with dual monoterpene/sesquiterpene activities in the cytoplasm and leucoplasts does not exclude the possibility that minute quantities of linalool and nerolidol can be synthesized in the cytosol and plastids, respectively, as a result of the possible trace levels of GPP and FPP derived from promiscuous side reactions of FPP and GGPP synthases in the corresponding cellular compartments. However, exogenously supplied [2H2]mevalolactone ([2H2]MVL), to cut snapdragon flowers efficiently, labeled nerolidol and showed no detectable incorporation in linalool (Dudareva et al., 2005), suggesting that if a GPP pool exists in the cytosol, it is small and does not contribute significantly to linalool formation. Moreover, snapdragon GPP synthase was localized in plastids, indicating that GPP biosynthesis takes place predominantly in leucoplasts (Tholl et al., 2004); there is no direct evidence for plastidial FPP synthase.

Although linalool and nerolidol are components of snapdragon floral scent, they are also major volatiles of fruit flavor in ripe strawberry (Aharoni et al., 2004). Similar to snapdragon, the cultivated strawberry genome contains two genes, FaNES1 and FaNES2, each encoding an enzyme with dual linalool/nerolidol synthase activities in vitro. It has been shown that FaNES1 and FaNES2 are localized in the cytosol and plastids, respectively. Although the FaNES1 gene was highly expressed in ripe strawberry fruits, FaNES2 expression was barely detectable. Based on these results, Aharoni et al. (2004) concluded that both linalool and nerolidol are formed by FaNES1 in the cytosol, and that strawberry cytoplasm contains sufficient levels of GPP and FPP for the biosynthesis of roughly similar quantities of linalool and nerolidol, although the pools of available substrates in fruit tissue are not known (Aharoni et al., 2004). Moreover, this work did not disprove plastidial formation of linalool, as linalool synthase activity was not checked in purified plastids isolated from strawberry fruits, and nor were any genetic or proteomic proofs provided.

Recently, Hampel et al. (2006) suggested that strawberry fruits synthesize terpenoids exclusively via the cytosolic MVA pathway, based on [2H2]MVL incorporation and a lack of labeling with deuterium-labeled 1-deoxy-[5,5-2H2]-D-xylulose ([2H2]DOX), in contrast to snapdragon flowers, where both plastidial monoterpene and cytosolic sesquiterpene biosyntheses rely on MEP pathway derived precursors (Dudareva et al., 2005). Although the MVA pathway is localized in the cytosol, the results obtained in fruits do not exclude the possible transport of IPP formed from labeled MVL to plastids for linalool biosynthesis. The absence of detectable [2H2]DOX labeling of terpenoids in fruit tissue could result from a lack of d-xylulose kinase, which catalyzes the phosphorylation of exogenously supplied DOX (Hemmerlin et al., 2006) in order for it to enter the pathway and be utilized. Unfortunately, d-xylulose kinase activity was not checked in fruit tissue, leaving the actual contributions of the MEP pathway to terpenoid biosynthesis in strawberry fruits yet to be determined.

Similar to snapdragon, nearly identical but compartmentally separated enzymes exist in monocotyledons. Maize plants contain TPS1, which is capable of producing (E)-nerolidol, (E)-β-farnesene and (E,E)-farnesol from FPP, and linalool from GPP in vitro (Schnee et al., 2002). Attack on maize cv. B73 by Egyptian cotton leafworms (Spodoptera littoralis) increases tps1 transcript levels by around eightfold over the basal level, and induces emission of (E)-β-farnesene, linalool and a metabolite of (E)-nerolidol, (3E)-4,8-dimethyl-1,3,7-nonatriene (DMNT). Although Schnee et al. (2002) reported TPS1 to be localized in the cytosol and to be responsible for sesquiterpene formation, our computer analysis of its subcellular localization using five search algorithms predicted its plastidial location, suggesting it could be involved in linalool formation. In order for TPS1 to catalyze (E)-nerolidol and (E)-β-farnesene formation, similar enzyme(s) must exist in the cytosol. Indeed, genes encoding proteins closely related to TPS1 have recently been identified in maize, and may be localized in the cytosol (J. Degenhardt and J. Gershenzon, personal communication).

Substrate availability in the cell determines the product type formed, and regulates developmental linalool/nerolidol emission in snapdragon

Linalool and nerolidol emission during flower development follows similar patterns to other terpenoids and phenylpropanoids within the snapdragon scent bouquet (Dudareva et al., 2000, 2003): peaking on day 5–6 postanthesis and declining thereafter (Figure 7d). Nerolidol and linalool synthase activities follow two different profiles, and belong to two different groups of scent biosynthetic enzymes based on developmental activity (Dudareva and Pichersky, 2000). Despite high levels of nerolidol/linalool synthase activities in old snapdragon flowers, emission of the corresponding compounds declines (Figure 7c,d), suggesting that substrate availability plays a role in the regulation of formation and emission of these compounds. Moreover, our previous labeling experiments showed that the exogenous supply of [2H2]MVL increased nerolidol emission by 2.4- and 10-fold during the day and night, respectively, suggesting that the capacity of AmNES/LIS-1 for nerolidol production has not been exceeded, even in mature snapdragon flowers (Dudareva et al., 2005). In these flowers, both monoterpenes and the sesquiterpene nerolidol are synthesized from MEP pathway derived IPP/DMAPP (Dudareva et al., 2005). However, exogenously supplied [2H2]DOX had no effect on their emission profiles, indicating that flux towards GPP and FPP is tightly regulated.

Overall, our results show that two nearly identical TPSs with similar catalytic properties are responsible for the formation of different terpenoid products in different subcellular compartments. The subcellular localization of these bifunctional enzymes and substrate availability determine the type and quantity of product formed. Since many TPSs accept more than one substrate (Tholl, 2006), the direction of nearly identical bifunctional enzymes to different cellular compartments allows plants to extend the range of available substrates, thus increasing the diversity of synthesized metabolites. However, there is no evidence for the production of high levels of monoterpenes in cytosol, which is most likely the result of a very small pool of GPP in this cellular compartment.

Experimental procedures

Plant material and headspace analysis of floral volatiles

Snapdragon (A. majus) cv. Maryland True Pink (Ball Seed, was grown in a greenhouse (Dudareva et al., 2000). Emitted volatiles were collected every 24 h at 09:00 h by the closed-loop stripping method (Donath and Boland, 1995) and were then analyzed following the methods described by Dudareva et al. (2000) and Kolosova et al. (2001).

Chemicals and radiochemicals

[1-3H]-Geranyl diphosphate (GPP; 3.7 kBq mol−1) and [1-3H]-farnesyl diphosphate (FPP; 3.7 kBq mol−1) were purchased from American Radiolabeled Chemicals ( Unlabeled GPP (1 mg ml−1) and FPP (1 mg ml−1) were purchased from Echelon Research Laboratories ( Terpenoid standards and reagents were purchased from Sigma-Aldrich ( unless otherwise noted.

Sequence analysis

Amino acid sequence alignments were generated using the ClustalX computer program ( Sequence relatedness was analyzed using the neighbor-joining method, and an unrooted tree was visualized using TreeView (

Expression of AmNES/LIS proteins in E. coli and purification of recombinant proteins

As both full-length AmNES/LIS cDNA clones from a snapdragon EST collection had an intron, the corresponding open reading frames were obtained by reverse transcriptase (RT) PCR amplification of RNA from 5-day-old snapdragon petal tissues. Forward primers were 5′-CATATGAGTAACCTTCATGTAAAG-3′ for AmNES/LIS-1 and 5′-CATATGGCATTTTATATTGGCAACAAC-3′ for AmNES/LIS-2, and introduced an NdeI site (underlined) at the initiating ATG codon. Reverse primers were 5′-CTCGAGATCAAACAACTAAATCTTAACATAC-3′ for AmNES/LIS-1 and 5′-CTCGAGATCAAACAACATAAACTTAACATAC-3′ for AmNES/LIS-2, and contained an XhoI site (underlined), which allowed a translational fusion to a C-terminal (His)6-tag. Truncated versions of AmNES/LIS-2 were amplified from AmNES/LIS-2 cDNA by PCR using the AmNES/LIS-1 forward primer for Met-31-AmNES/LIS-2 and the 5′-CATATGGACTTGAATGGACAAAG-3′ forward primer for Met-79-AmNES/LIS-2, both in combination with the AmNES/LIS-2 reverse primer. The amplified coding regions were subcloned in-frame and upstream of the (His)6-tag of the pET32a expression vector (Novagen, Sequencing revealed that no errors had been introduced.

For functional expression, E. coli BL21 Rosetta competent cells were transformed with recombinant plasmids and pET32a lacking an insert (control). Induction, harvesting and protein purification by affinity chromatography on nickel-nitrilotriacetic acid-agarose (Qiagen, were performed as described by Negre et al. (2002) and Kaminaga et al. (2006). The purity of the proteins was determined by densitometry of the SDS-PAGE gels after Coomassie Brilliant Blue staining, and ranged from 65% to 72%. The protein concentration was determined by the Bradford method (Bradford, 1976).

Terpene synthase assays and product identification

Assays for TPS activity were performed in a 100-μl assay buffer (30 mm HEPES, pH 7.5, 5 mm DTT, and 1 mm or 40 mm MgCl2) containing recombinant protein (0.58–0.66 μg) or plant crude extract (50–100 μg of protein), and 41/72 μm [1-3H]GPP/FPP (55 mCi mmol−1). The mixture was immediately overlaid with 1 ml of hexane. After a 15-min incubation at 30°C, reaction products (and alcohols produced by endogenous phosphohydrolase activity) were extracted into the hexane phase and a volume of 600 μl was counted (Boatright et al., 2004). Extracts from E. coli transformed with pET-32a lacking an insert and heat-denatured enzymes served as controls.

For identification of reaction products, large-scale assays (1000 μl) were performed using purified AmNES/LIS proteins, crude extracts from flowers, and purified leucoplasts from petals using unlabeled FPP and GPP (60 μm) with the solid-phase microextraction (SPME) system. After incubation for 1 h at 30°C and 15 min at 42°C, synthesized volatiles were analyzed by GC-MS. Chiral GC-MS analysis was performed on the same instrument equipped with 2,3-di-O-methyl-6-O-tert-butyl dimethylsilyl beta cyclodextrin doped into a 14% cyanopropylphenyl/86% dimethyl polysiloxane (RtTM-βDEXsm) column (30-m × 0.25-mm inner diameter; Restek, under conditions specified by the manufacturer.

For detecting linalool synthase activity, the total monoterpene synthase activity determined using radiolabeled GPP was split amongst three monoterpene compounds, myrcene, (E)-β-ocimene, and linalool, based on the level of products produced by the same crude extracts from unlabeled GPP as determined by SPME.

Purification of native AmNES/LIS proteins from snapdragon petals

Protein purification was performed as described in Murfitt et al. (2000), except for the extraction buffer that contained 30 mm HEPES, pH 8.0, 1 mm DTT, 1 mm phenylmethanesulfonyl fluoride and 10% (v/v) glycerol (buffer A) with 1% (w/v) polyvinylpyrrolidone (PVP-40). After (DEAE)–cellulose (Whatman, column chromatography, fractions with the highest monoterpene and sesquiterpene synthase activities in the 230–280 mm KCl range were loaded on a HiTrap-Q FF column (Pharmacia Biotech Inc., pre-equilibrated with buffer A. TPSs were eluted with the linear gradient of 0–1 m KCl, and fractions with the highest monoterpene and sesquiterpene synthase activities were subjected to size-exclusion gel filtration chromatography on a HiLoadTM 16/60 SuperdexTM 200 column (Pharmacia Biotech Inc.), which resulted in three 1-ml active fractions. These fractions were precipitated with trichloroacetic acid (20% final concentration; Kohlstaedt, 2006), and were examined by SDS-PAGE followed by gel staining with Coomassie brilliant blue.

Protein sequencing

The ∼60-kDa protein band from the combined size-exclusion fractions with the highest TPS activities as well as surrounding proteins within 1 cm were excised from the SDS-PAGE gel, digested with trypsin and sequenced by LC-MS/MS as described by Chen et al. (2005), followed by the screening of a snapdragon EST database comprising more than 25 000 sequences using the mascot program (Perkins et al., 1999). Five unique peptides obtained from the protein band (SKREEEPIDR, LMFVDAIQR, FRGELQR, GLMELYEASQLR and LQHPYR) matched predicted protein sequences of both AmNES/LIS-1 and AmNES/LIS-2 with a significance threshold of 0.05. A sixth peptide (LGVNHNFEELIETILR) matched only AmNES/LIS-1. Eight peptides (HLLHLK, LGVNHHFQK, EIEEILRK, YVSADVFNNFK, FKEELKR, NYDDDLGGMYEWGK, ESFNLNHDSVSSFK and GPVLEEYVK) matched protein sequences of previously characterized myrcene and ocimene synthases (Dudareva et al., 2003), whereas six peptides (GENDPIESLIFVDATLR, YPFHTSIAR, HGYNPIDSLK, GSIELNDTQELMSSIAIIFR, QAALNLAR and MVPLMYSYDHNQR) matched only myrcene synthase. At this stringency, no peptide hits to other TPSs in the EST database were found.

Determination of AmNES/LIS kinetic properties

For kinetic analysis, an appropriate enzyme concentration was chosen so that the reaction velocity was proportional to the enzyme concentration, and was linear with respect to incubation. Kinetic data were evaluated using Lineweaver–Burk, Eadie–Hofstee and Hanes plots.

RNA isolation and real-time RT-PCR

Total RNA from leaves and floral tissues was isolated and analyzed as described previously (Dudareva et al., 1996, 1998, 2000; Kolosova et al., 2001). One microgram of total RNA was used to prepare cDNA according to the manufacturer’s instructions (High-Capacity cDNA Archive Kit; Applied Biosystems, A linear range of cDNA was used for real-time PCR. Real-time quantitative reverse transcriptase polymerase chain reaction (real-time RT-PCR) was performed in a 96-well plate using a Gene-Amp 5700 sequence detection system (PE Applied Biosystems, with SYBR green fluorescent dye. A 250-bp fragment in the 3′ region of the coding sequence was amplified using gene-specific primers: forward 5′-AAAGAACATATAAGCTTGTCGAC-3′ and reverse 5′-GTTCTTTAATTGGTAGTGCTTCG-3′ for AmNES/LIS-1, forward 5′-AAAGAACATACAAGCTTGTCGCT-3′ and reverse 5′-GTTCTTTAATTGGTAGTGCTCGA-3′ for AmNES/LIS-2, forward 5′-TGGTCGCAAGGCTGAAACTTAAAGGA-3′ and reverse 5′-ACCAGACAAATCGCTCCACCAACTAA-3′ for 18S cDNA. Real-time RT-PCRs were performed in a 20-μl reaction mixture containing forward and reverse primers according to the manufacturer’s protocol (PE Applied Biosystems). Real-time RT-PCR conditions were as follows: 94°C for 10 min for one cycle, followed by 40 cycles of 94°C for 15 sec, 54°C for 15 sec and 72°C for 15 sec. Several real-time RT-PCRs were run per sample in triplicate per run. The quantification of gene expression was performed using the comparative cycle threshold (Ct) method (Applied Biosystems).

Subcellular localization of AmNES/LIS proteins

The open reading frames of AmNES/LIS-1, AmNES/LIS-2, Met-31-AmNES/LIS-2 and the first 69 aa of AmNES/LIS-2 were fused upstream of, and in frame with, GFP in the cloning sites XbaI and BamHI of the p326-SGFP vector containing the CaMV 35S promoter (a gift from Dr I. Hwang, Pohang University of Science and Technology, Korea). Sequencing confirmed the accuracy of the fusion. Arabidopsis protoplasts were transformed as described by Sheen (2002). DNA from each construct (20 μg) was used for PEG-mediated transformation of 100 μl of ice-cold protoplasts. p326-SGFP and p326-RbTP-SGFP were markers for cytosolic and plastidial localization. Transient expression of GFP fusion proteins was observed 16–20 h after transformation using MRC-1024 UV (Bio-Rad, on a Diaphot 300 (Nikon, inverted microscope with a 60 × 1.4 numerical aperture lens. The 488-nm line of the krypton–argon laser (Ion Laser Technology; was used to excite GFP. Fluorescence was collected with a 522/35 bandpass filter. Chlorophyll fluorescence was excited using the 568-nm line of the krypton–argon laser, and emission was collected with a 680DF32 filter.

Leucoplast isolation

Leucoplasts were isolated according to the method described by Kang and Rawsthorne (1994), and were purified using a discontinuous Percoll gradient (Smith et al., 1992). The gradient consisted of 2 ml 70% Percoll, 4 ml of each of 50, 35, 22 and 10% PBF-Percoll, and leucoplasts were collected from the 22–35% interface after 8 min of centrifugation at 9200 g using a Sorvall HB-6 swing-out rotor. The integrity of the resulting leucoplasts was confirmed by scanning electron microscopy, and yield and cytosolic contamination were assessed by measuring NADP-glyceraldehyde-3-phosphate dehydrogenase and alcohol dehydrogenase activities (Kang and Rawsthorne, 1994). No alcohol dehydrogenase activity was detected in purified plastids.

Statistical analysis

The correlation between TPS activities and the emission of corresponding terpenoid compounds was analyzed by anova using the jmp statistical package (version 6.0.2;


We thank Dr Inhwan Hwang for GFP vectors. Confocal microscopy data was acquired from Purdue Cancer Center Analytical Cytometry Laboratories, supported by the Cancer Center NCI core grant # NIH NCI-2P30CA23168. We thank Jennifer Sturgis and Chia-Ping Huang for their confocal and scanning electron microscopy assistance, respectively, Dr Yasuhisa Kaminaga for protein purification help and Anthony Qually for his help with enantioselective GC-MS analysis. This work was supported by grants from the National Science Foundation (grant no. MCB-0615700) and the Fred Gloeckner Foundation, Inc. to ND, and the state Sachsen-Anhalt (Landesexzellenznetzwerk Biowissenschaften) to MG. This article is contribution no. 2007-18089 from Purdue University Agricultural Experimental Station.