Formation of the Unusual Semivolatile Diterpene Rhizathalene by the Arabidopsis Class I Terpene Synthase TPS08 in the Root Stele Is Involved in Defense against Belowground Herbivory

This work reports that Arabidopsis roots produce at least four related semivolatile diterpenes named rhizathalenes that have not been previously identified in this or any other plant species. It shows that rhizathalenes act as antifeedants against root-feeding insects.

Abstract

Secondary metabolites are major constituents of plant defense against herbivore attack. Relatively little is known about the cell type–specific formation and antiherbivore activities of secondary compounds in roots despite the substantial impact of root herbivory on plant performance and fitness. Here, we describe the constitutive formation of semivolatile diterpenes called rhizathalenes by the class I terpene synthase (TPS) 08 in roots of Arabidopsis thaliana. The primary enzymatic product of TPS08, rhizathalene A, which is produced from the substrate all-trans geranylgeranyl diphosphate, represents a so far unidentified class of tricyclic diterpene carbon skeletons with an unusual tricyclic spiro-hydrindane structure. Protein targeting and administration of stable isotope precursors indicate that rhizathalenes are biosynthesized in root leucoplasts. TPS08 expression is largely localized to the root stele, suggesting a centric and gradual release of its diterpene products into the peripheral root cell layers. We demonstrate that roots of Arabidopsis tps08 mutant plants, grown aeroponically and in potting substrate, are more susceptible to herbivory by the opportunistic root herbivore fungus gnat (Bradysia spp) and suffer substantial removal of peripheral tissue at larval feeding sites. Our work provides evidence for the in vivo role of semivolatile diterpene metabolites as local antifeedants in belowground direct defense against root-feeding insects.

INTRODUCTION

Nearly all terrestrial plants depend on roots as life-supporting organs. Consequently, more than 50% of the net primary production of plants is commonly allocated to roots (Fogel, 1985). Root herbivores represent a constant belowground threat to plants by disrupting the uptake of water and mineral nutrients or limiting the storage of carbohydrates (Blossey and Hunt-Joshi, 2003; van Dam, 2009). Moreover, root damage by herbivores influences organismal communities both below- and aboveground (van Dam and Heil, 2011).

Despite the importance of roots for plant growth and fitness, knowledge of root defense mechanisms in the interaction between plants and belowground herbivores is still limited. Plants deploy chemical defense or specialized metabolites that directly or indirectly affect herbivore performance (Kessler and Baldwin, 2001; Kaplan et al., 2008). Terpenes constitute the largest class of plant defense chemicals, the functions of which have been investigated primarily in aboveground plant organs. For example, terpenes of low molecular weight, such as monoterpenes (C₁₀) and sesquiterpenes (C₁₅), are released as volatile compounds upon herbivore feeding damage to deter or intoxicate insects or to serve as signals of indirect defense by attracting parasitoids and predators of arthropod herbivores (Unsicker et al., 2009; Dicke and Baldwin, 2010). Root-specific production of terpenes has been reported in gymnosperms (Huber et al., 2005) and several angiosperms and has been associated primarily with antimicrobial and allelopathic activities. In rice (Oryza sativa), several nonvolatile diterpenes (C₂₀), such as momilactones, with functions as phytoalexins or allelochemicals are found in root exudates (Toyomasu et al., 2008; Xu et al., 2012). Similarly, cotton (Gossypium hirsutum) roots produce gossypol terpene aldehydes, and oat (Avena sativa) roots release triterpene saponins (C₃₀) exudates with antimicrobial activity in response to pathogen infection (Hunter et al., 1978; Osbourn et al., 2003). Belowground herbivory has been shown to induce the emission of volatile terpenes from plant roots, and these compounds can act in indirect defense by attracting entomopathogenic nematodes (Rasmann et al., 2005; Ali et al., 2010; Lawo et al., 2011). Despite these findings, possible antiherbivore activities of many root-specific terpene or other specialized metabolites have not been described in vivo, and little is known about how the function of these compounds relates to their cell type–specific formation in plant roots.

Arabidopsis thaliana represents a suitable model system for such studies since its genome contains 32 terpene synthase (TPS) genes (Aubourg et al., 2002), 15 of which are expressed primarily or exclusively in the roots (Birnbaum et al., 2003; Chen et al., 2003; Tholl and Lee, 2011). TPS enzymes catalyze the formation of monoterpenes, sesquiterpenes, or diterpenes from the central prenyl diphosphate precursors, geranyl diphosphate (GPP), farnesyl diphosphate, and geranylgeranyl diphosphate (GGPP), respectively (Bohlmann et al., 1998; Tholl et al., 2006; Chen et al., 2011). The prenyl diphosphate substrates are synthesized by sequential condensation of the five-carbon building blocks isopentenyl diphosphate and its isomer dimethylallyl diphosphate that are derived from the mevalonate pathway in the cytosol or the methylerythritol phosphate (MEP) pathway in plastids (Dudareva et al., 2006). Most studies investigating the subcellular localization of terpene biosynthesis in aboveground tissues have demonstrated a primary formation of sesquiterpenes by terpene synthases located in the cytosol, while monoterpenes and diterpenes are synthesized by TPS enzymes predominantly targeted to plastids (Dudareva et al., 2006; Chen et al., 2011). By contrast, the subcellular compartmentation of terpene formation in root-specific cells is less well understood.

Of the 15 root-specific Arabidopsis TPSs, four have been characterized: Two identical TPSs, TPS23 and TPS27, produce the monoterpene 1,8-cineole and are expressed in epidermal cells of Arabidopsis roots (Chen et al., 2004). The two other highly homologous genes, TPS12 and TPS13, encoding (Z)-γ-bisabolene sesquiterpene synthases, are constitutively expressed in the root cortex and subepidermal layers (Ro et al., 2006). None of the remaining 11 root-expressed TPSs, which exhibit cell type–specific expression patterns according to fine-scale transcript analyses (Birnbaum et al., 2003; Brady et al., 2007), have been characterized in terms of their biochemical activities and the biological function of their terpene products.

Here, we show that the terpene synthase TPS08 is responsible for the formation of the semivolatile diterpene hydrocarbons named rhizathalenes in Arabidopsis roots. The predominant diterpene product, rhizathalene A, which represents a so far uncharacterized class of tricyclic spiro[5.5]undecane diterpene carbon skeletons, is constitutively released from roots of plants grown in potting substrate and under nonsoil culture conditions. We further show that TPS08 is a modern diterpene synthase of the TPS-a subfamily and produces rhizathalenes by a class I type carbocationic reaction mechanism. Rhizathalene synthase is largely expressed in the vascular tissue and is targeted to root plastids. To determine possible defensive activities of rhizathalenes, we established bioassays with larvae of the common opportunistic root herbivore fungus gnat (Bradysia spp) (Harris et al., 1996). We demonstrate that roots of tps08 gene knockout plants, which are deficient in rhizathalene formation, are more susceptible to insect herbivory and the removal of peripheral cell layers. Our studies show that semivolatile diterpenes are involved in direct belowground plant defense.

RESULTS

Arabidopsis Roots Emit Volatile Diterpenes

From the 15 TPS genes that are constitutively expressed in Arabidopsis roots (Birnbaum et al., 2003; Chen et al., 2003; Ro et al., 2006; Tholl and Lee, 2011), five (TPS08, TPS17, TPS20, TPS26, and TPS29) were previously predicted to function as diterpene synthases (Aubourg et al., 2002). Therefore, we investigated the possible formation of diterpenes in Arabidopsis roots grown under different culture conditions. Columbia (Col-0 CS60000) wild-type plants were grown in potting mix, hydroponic culture, axenic culture supplemented with 1% Suc, and aeroponic culture (Vaughan et al., 2011) (Figure 1A). Under the assumption that some of the diterpenes produced in Arabidopsis roots were volatile, we applied automated solid phase microextraction (SPME) to collect volatiles from roots that were detached from plants grown under the different culture conditions. Four diterpene olefins (A to D) were detected by gas chromatography–mass spectrometry (GC-MS) analysis of volatiles emitted from hydroponically grown roots (Figure 1B). Only the predominant diterpene compound A was detected from roots grown in the other culture conditions. The mass spectrum of this compound resembled that of a tricyclic hydrocarbon diterpene (Figure 1C) but did not match the spectra of any other known diterpene compounds listed in the National Institute of Standards and Technology/Wiley libraries, other databases, or references. Therefore, we named these diterpene compounds rhizathalene A to D (rhiza, originating from the Greek word for “root,” and thale for the Arabidopsis common name thale cress).

Figure 1.

Diterpene Volatile Emission from Arabidopsis Roots.

(A) Arabidopsis culture systems used for root volatile analysis.

(B) Total ion GC-MS chromatogram of volatiles collected from 1 g of hydroponically grown roots using SPME. Four diterpene compounds were detected designated rhizathalene A to D (peaks A to D).

(C) Electron impact (EI) mass spectrum of rhizathalene A. m/z, mass-to-charge ratio.

(D) Concentrations of rhizathalene A in roots grown under the different culture conditions. Diterpenes were extracted with organic solvent and quantified by GC-FID. Letters above bars indicate significant differences (one-way ANOVA and Tukey-Kramer HSD, n = 3, P < 0.01). Numbers are averages ± se. FW, fresh weight.

Quantitative analysis of the detected diterpenes by solvent extraction of root tissue and subsequent gas chromatography–flame ionization detection (GC-FID) showed highest concentrations of rhizathalene A in hydroponically grown roots of ∼880 ng g⁻¹ fresh weight (Figure 1D). By comparison, 550 ng g⁻¹ of rhizathalene was found in roots grown in sterile liquid culture and ∼150 to 170 ng g⁻¹ fresh weight was detected in roots from plants grown in soil or under aeroponic culture conditions (Figure 1D). Other volatile terpenes that were detected in small amounts from Arabidopsis roots were the monoterpene 1,8-cineole and the sesquiterpene (Z)-γ-bisabolene, which have been described as products of the root-expressed terpene synthases TPS23/27 and TPS12/13, respectively (Chen et al., 2004; Ro et al., 2006).

A tps08 T-DNA Insertion Line Lacks the Formation of Rhizathalene

To identify the TPS gene responsible for rhizathalene formation, we analyzed roots from two independent T-DNA insertions in or near the genes of the five putative root-specific diterpene synthases (see Plant Material in Methods) for the absence of rhizathalene A. Only a single line, SALK_125194, with a T-DNA insertion in the sixth exon of the gene TPS08 (tps08-1) (Figure 2A) did not produce rhizathalene A (Figure 2B) nor was the TPS08 transcript detected by RT-PCR (Figure 2C). By contrast, a T-DNA insertion in the 3′-untranslated region of TPS08 in line SALK_112521 (tps08-2) (Figure 2A) did not affect the formation of rhizathalene A (Figure 2B) or TPS08 transcript levels (Figure 2C). tps08-2 was therefore used as a T-DNA control line. Since only a single tps08 null mutant was available, we conducted DNA gel blot analysis to confirm that the lack of rhizathalene formation in the SALK_125194 mutant was caused by a single T-DNA insertion in the TPS08 gene (see Supplemental Figure 1 online). To complement the TPS08 loss-of-function phenotype, we transformed the tps08-1 mutant with the TPS08 gene under the control of a 2.2-kb TPS08 promoter fragment. TPS08 was found to be expressed and rhizathalene emitted from roots of at least three independently established transgenic plants (see Supplemental Figure 2 online). However, the TPS08 transcript and diterpene compound were observed at lower abundance in the transgenic lines than in the wild type (see Supplemental Figure 2 online), indicating that the phenotype was only partially restored because of a possible lack of full activity of the selected promoter region.

Figure 2.

A tps08 T-DNA Insertion Line Lacks the Formation of Rhizathalene A.

(A) Positions of T-DNA insertions in the TPS08 (At4g20210) gene. Exons are represented by the gray boxes. Introns and untranslated regions (UTR) are represented by the black line. The two independent T-DNA insertions are indicated as triangles.

(B) Total ion GC-MS chromatograms of rhizathalene A emitted from roots of wild type (WT) and T-DNA insertion lines grown in axenic culture.

(C) Analysis of TPS08 transcripts in roots of wild-type and T-DNA insertion lines.

Biochemical Characterization of TPS08

To confirm the catalytic activity of the TPS08 enzyme, we cloned a 1678-bp cDNA of TPS08 lacking the first 123 bp, which encode a putative 41–amino acid plastidial transit peptide, into the Escherichia coli expression vector pET102/D-TOPO. The partially purified recombinant TPS08 enzyme converted all-trans GGPP into several diterpene compounds (Figures 3A and 3B). The retention time and mass spectrum of the major enzymatic product were identical with that of rhizathalene A. Accordingly, the mass spectra of the three less abundant compounds were identical to those of rhizathalene B, C, and D, which had been detected in roots of hydroponically grown plants (Figure 1A). Incubation of the TPS08 recombinant enzyme with the substrates GPP and (E,E)-farnesyl diphosphate did not result in the formation of enzyme-specific products in comparison to assays with protein extracts prepared from bacterial cells carrying the empty expression vector.

Figure 3.

TPS08 Enzymatic Products and Structure of Rhizathalene A.

(A) GC-MS chromatogram of diterpene compounds produced from all-trans GGPP with partially purified recombinant TPS08 enzyme. TPS08 catalyzes the formation of rhizathalene A (peak A) as the major product and seven minor diterpene products. Peaks A to D correspond to the same compounds as shown in Figure 1B.

(B) GC-MS chromatogram obtained with an extract from E. coli carrying the empty expression vector. The extract was subjected to the same purification and assay procedure.

(C) The structure of rhizathalene A derived from NMR data. The relative stereochemistry is consistent with the ROESY (see Supplemental Table 1 and Supplemental Methods online).

Kinetic characterization of the partially purified recombinant TPS08 protein showed a low apparent K_m value for all-trans GGPP of 0.8 ± 0.2 µM, which is in the range of those reported from other plant diterpene synthases, including the class-I type casbene and taxadiene synthases (Hill et al., 1996; Williams et al., 2000). The V_max was 4.6 ± 0.6 pkat mg⁻¹, and the k_cat was 3 × 10⁻⁴ ± 4 × 10⁻⁵ s⁻¹, resulting in a k_cat/K_m value of 0.5 ± 0.1 s⁻¹ µM⁻¹, which is 10- to 50-fold lower than those found for taxadiene and casbene synthase. Consistent with other TPS enzymes, the activity of TPS08 was dependent on the presence of Mg²⁺ as a divalent metal ion.

Structural Elucidation of Rhizathalene A

To elucidate the molecular structure of rhizathalene A, we produced larger amounts of the diterpene compound by transformation of the E. coli C41 (DE3) strain with the pET102/D-TOPO-TPS08 vector construct. Strain C41(DE3) contained an additional pACYCDuet-rAgGGPS plasmid carrying the pseudomature GGPP synthase from grand fir (Abies grandis), rAgGGPS (Cyr et al., 2007). After collection of rhizathalene A from the headspace of 2 liters of bacterial cultures coexpressing the GGPP synthase and TPS08 enzymes and subsequent purification of the compound by liquid chromatography–mass spectrometry, the molecular formula and structure were verified and obtained using NMR experiments that included proton (¹H) NMR (see Supplemental Figure 3 online), Double quantum filtered-correlated spectroscopy (DQF-COSY) (see Supplemental Figure 4 online), Rotating frame overhauser enhancement spectroscopy (ROESY) (see Supplemental Figure 5 online), Total correlated spectroscopy (TOCSY) (see Supplemental Figure 6 online), Heteronuclear Single Quantum Coherence-with distortionless enhancement by polarization transfer (HSQC-DEPT) (see Supplemental Figure 7 online), and Heteronuclear multiple bond coherence (HMBC) (see Supplemental Figure 8 online). Proton and carbon chemical shifts are shown in Supplemental Table 1 online. The analysis revealed a novel tricyclic diterpene carbon structure with a cyclopenta-spiro[5.5]undecane skeleton (Figure 3C). The relative stereochemistry was deduced from the ROESY spectrum (see Supplemental Figure 5 online). Complete listings of correlations are presented in Supplemental Table 1 online. Assignment of the relative stereochemistry was supported by the results of quantum chemical calculations. ¹³C and ¹H chemical shifts for all possible diastereomers were computed using density functional theory (Lodewyk et al., 2012), and the assigned structure of rhizathalene A was the diastereomer whose computed chemical shifts best matched (Smith and Goodman, 2010) the experimentally determined chemical shifts, with mean average deviations between computed and experimental shifts of only 2.1 ppm for ¹³C and 0.06 ppm for ¹H (see Supplemental Figures 9 to 12 online). The relative stereochemistry proposed for rhizathalene A at C6 and C7 (6S* and 7R*) corresponds to that of the structurally related spiro[5.4]decane sesquiterpene (−)-α-acoradiene (C6 epimer of β-acoradiene; see Supplemental Figure 13 online) (Tashiro et al., 2004). The absolute configuration of rhizathalene A has not yet been determined.

Subcellular Localization of TPS08 and Rhizathalene Formation

Since diterpenes are primarily synthesized in chloroplasts, we examined whether TPS08 was targeted to plastids of Arabidopsis root cells. TargetP and ChloroP algorithms predicted a plastidial transit peptide of TPS08 of 41 amino acids. To determine the subcellular localization of the TPS08 protein, we generated transgenic plants in the tps08-1 knockout background expressing the TPS08 transit peptide in C-terminal fusion with fluorescent proteins under the control of the cauliflower mosaic virus (CaMV) 35S and the native TPS08 promoters. Analysis of hypocotyl tissue from independent Pro35S:TPS08-eGFP (for enhanced green fluorescent protein) lines showed a subcellular localization of green fluorescence in plastids confirming that the TPS08 protein is targeted to these organelles (Figures 4A to 4D; see Supplemental Figure 14 online). In yellow fluorescent protein (YFP) fusion lines carrying the TPS08 promoter (ProTPS08:TPS08-eYFP), fluorescence was observed in leucoplasts of the root stele, indicating a primary expression of TPS08 in this tissue (Figures 4E to 4H).

Figure 4.

Plastidial Localization of the TPS08 Protein.

(A) to (D) Images of the hypocotyl of plants expressing eGFP fused to a 41–amino acid N-terminal TPS08 peptide under the control of the CaMV 35S promoter.

(A) and (B) Chlorophyll autofluorescence and GFP fluorescence.

(C) Bright-field image.

(D) Overlay of (A) to (C).

(E) to (H) Images of roots from plants expressing enhanced YFP fused to the N-terminal TPS08 peptide under the control of the TPS08 promoter. Untransformed wild-type roots did not show any fluorescence and untargeted GFP protein was primarily located in the cytosol (see Supplemental Figure 14 online).

(E) Red channel.

(F) YFP fluorescence.

(G) Bright-field image.

(H) Overlay of (F) and (G).

Bars = 20 μm in (A) to (D) and 10 μm in (E) to (H).

To investigate to what extent rhizathalene A is produced from precursors synthesized by the plastidial MEP pathway, we administered different concentrations of deuterium-labeled deoxyxylulose ([5,5-²H₂]-DOX, DOX-d₂) to axenic Arabidopsis hairy root cultures and determined how much of the labeled precursor was incorporated into rhizathalene A after 24 and 48 h of incubation (Figures 5A to 5C). Headspace volatile analysis of roots incubated with 0.2 or 2 mg/mL DOX-d₂ for 48 h indicated that ∼50% of the detected rhizathalene A was labeled. Occurrence of the fully labeled isomer is apparent by an increase in the molecular mass of 8 from 272 to 280. We also analyzed the incorporation of deuterated mevalonolactone ([2,2-²H₂]-MVA, MVL-d₂) at a concentration of 2 mg/mL under the same incubation times. Under these conditions, only a trace amount of labeled rhizathalene A was detected (Figure 5A), indicating a minimal contribution of precursors from the MVA pathway to the formation of the diterpene compound.

Figure 5.

Production of Deuterium-Labeled Rhizathalene A by Administration of Deuterated Precursors of the MEP and MVA Pathways.

(A) GC-MS chromatograms (TIC) of labeled and unlabeled rhizathalene obtained from different feeding experiments. a, 24-h incubation with [5,5-²H₂]-DOX; b, 48-h incubation with [5,5-²H₂]-DOX; c, 24-h incubation with [2,2-²H₂]-MVA; d, 48-h incubation with [2,2-²H₂]-MVA; e, control. 1, Unlabeled rhizathalene A; 2, deuterated rhizathalene A upon [5,5-²H₂]-DOX feeding; tr, trace of deuterated rhizathalene A upon [2,2-²H₂]-MVA feeding. All deuterium-labeled precursors were used at a concentration of 2 mg/mL. The retention time of rhizathalene A is shorter than that shown in Figure 1B because of small changes in column conditions.

(B) Mass spectrum of peak 2 in (A). m/z, mass-to-charge ratio.

(C) Labeling degree of rhizathalene A with [5,5-²H₂]-DOX after 48 h of incubation. Bars represent the mean ± sd of three replicates.

TPS08 Promoter Tissue-Specific Activity in the Root Stele

Tissue-specific activity of the TPS08 promoter was examined in planta by staining of transformed Arabidopsis plants carrying a ProTPS08:β-glucuronidase (GUS) fusion. GUS staining was observed exclusively in the roots of at least three independent lines in the T2 generation. In seedlings grown on sterile Murashige and Skoog medium, weak GUS staining was detected in the stele and at the tips of both primary and secondary roots (Figures 6A to 6C). In comparison, hydroponically grown ProTPS08:GUS plants showed considerably stronger GUS activity in the same tissues and cell types (Figures 6D to 6F). This result is consistent with the higher transcript levels of TPS08 observed under hydroponic culture conditions (see below).

Figure 6.

Tissue-Specific Expression of GUS Activity in Roots of ProTPS08-GUS Plants.

Histochemical GUS staining was observed in the stele and in emerging secondary roots and root tips. Results are representative of at least four independent lines. Bars = 200 µm.

(A) to (C) Images of ProTPS08-GUS seedlings grown on half-strength Murashige and Skoog plates for 15 d.

(D) to (H) Staining of roots from hydroponically grown ProTPS08-GUS plants ([D] to [F]) and wild-type plants ([G] and [H]).

Tps08 Knockout Mutants Are More Susceptible to Bradysia Larval Feeding

To investigate the defensive activity of rhizathalene in planta, we established an aeroponic clay pellet culture system that maintained a suitable belowground environment for a root herbivore while providing easy access to the root tissue (Figures 1A and 7A) (Vaughan et al., 2011). We selected Bradysia as a root herbivore since Bradysia larvae were shown to be destructive to Arabidopsis seedlings and jasmonate-insensitive mutants with impaired root defense (McConn et al., 1997). Wild-type and tps08 mutant plants were grown in aeroponic culture for 4 weeks and subsequently treated with 200 to 300 Bradysia larvae of the second and third instar, which were allowed to feed on the roots for 5 d. The number of larvae is equivalent to the offspring of a single fungus gnat female laying up to 1000 eggs (Jagdale et al., 2007).

Figure 7.

Arabidopsis Plants Lacking Rhizathalene Production Are Significantly More Susceptible to Bradysia Root Herbivory.

(A) Image of Bradysia larva (indicated by arrow) feeding on Arabidopsis roots in aeroponic culture.

(B) Mature wild-type (WT) and tps08-1 mutant plants grown in aeroponic culture were infested with 200 to 300 Bradysia larvae for 5 d. Average dry root mass remaining after Bradysia feeding (gray bars) compared with untreated control plants (white bars). Significantly less root mass was recovered from the tps08-1 knockout mutant (SALK_125194) in comparison to wild-type plants and the tps08-2 T-DNA insertion control line (SALK_112521) (Figure 2A), which produces rhizathalene at wild-type levels. Cumulative log-transformed data from two independent experiments were analyzed with two-way ANOVA and Tukey-Kramer HSD; bars represent the mean ± se; n = 24 for the wild type and tps08-1; n = 12 for tps08-2; P < 0.01.

(C) Arabidopsis seedlings grown in soil were infested with 10 larvae per plant for 7 d. Average seedling mass of plants infested with Bradysia larvae (gray bars) compared with noninfested control plants (white bars). Seedlings treated with Bradysia larvae showed reduced growth. Seedlings of tps08-1 plants were significantly smaller than wild-type and tps08-2 plants subjected to the same stress. Cumulative log-transformed data from two independent experiments were analyzed with two-way ANOVA and Tukey-Kramer HSD; bars represent the mean ± se; n = 20; P < 0.01.

The amount of root mass remaining after Bradysia feeding was compared with that of noninfested control plants (Figure 7B). Wild-type and mutant plants showed a reduction of root mass, but significantly less root tissue was recovered from tps08-1 plants in comparison to wild-type plants and the tps08-2 T-DNA control, which produced rhizathalene at wild-type levels (P < 0.0001 and P < 0.01, respectively) (Figure 7B). The percentage of root mass consumed from control plants ranged between 32 and 48%, while 56 to 63% of root tissue was consumed from tps08-1 plants, amounting to a 10 to 20% higher loss of root tissue than that of controls.

We also tested the effect of Bradysia feeding on wild-type and mutant plants at the seedling stage. Fourteen-day-old seedlings planted in potting mix were subjected to a more moderate infestation of 10 larvae per plant. The seedling biomass of the wild type and tps08-1 mutants challenged by larval feeding was compared with that of nondamaged controls (Figure 7C). All seedlings treated with Bradysia larvae showed reduced growth. However, seedlings of the tps08-1 line consistently had significantly less biomass after larval feeding than wild-type seedlings and those of the tps08-2 line (P < 0.01 and P < 0.001) (Figure 7C). The percentage of reduction in seedling biomass was between 31 and 50% for wild-type and tps08-2 plants. By comparison, reduction of biomass was significantly higher for seedlings of the tps08-1 line, ranging between 62 and 72%. We did not determine differences in larval weight in both bioassays since a quantitative recovery and weight measurement of the small larvae from both substrates proved to be too difficult.

Bradysia Larval Feeding Damage in Situ and Feeding on Artificial Diet Containing Rhizathalene A

Since there was a significant difference in the amount of root tissue remaining after Bradysia larval feeding, we investigated possible differences in larval feeding behavior by comparing the larval root damage of aeroponically grown wild-type and tps08-1 plants. After removal of the clay granules, undamaged and damaged roots were examined under a dissecting microscope (Figure 8). Undamaged tps08-1 roots were phenotypically indistinguishable from wild-type roots (Figures 8A to 8D). Wild-type roots were visibly damaged by larval feeding. Portions of the root epidermis and cortex were removed and several root tips were severed. Generally, larvae avoided feeding on the vascular tissue and parts of the endodermis (Figures 8E to 8H). By contrast, damage of roots of the tps08-1 line was much more severe per feeding site, resulting in the consumption of approximately three times more tissue per site (see Supplemental Figure 15 online). Large areas of the root tissue were completely stripped of the epidermis and cortex, and the endodermis and phloem were consumed, leaving only a string of lignified xylem cells and resulting in Figures 8I to 8L. The difference between the total amount of tissue removed from tps08-1 and wild-type roots (Figures 7B and 7C) was somewhat smaller than that observed per feeding site, indicating that larvae feeding on wild-type roots moved more frequently from one site to the next, whereas they spent more time per feeding site on roots of the tps08-1 mutant.

Figure 8.

Bradysia Larval Feeding Damage on Roots of Wild-Type and tps08-1 plants in Aeroponic Culture.

(A) to (D) Undamaged roots of wild-type plants.

(E) to (H) Bradysia feeding damage on wild-type (WT) roots. Parts of the root epidermis, cortex, and endodermis were removed and tips of several secondary roots were severed.

(I) to (L) Herbivore damage on roots of tps08-1 mutant line. Large areas of the root were severely damaged. In addition to the epidermis and cortex, the endodermis and phloem were largely consumed, leaving only a string of lignified xylem. Bar = 250 µm.

To further corroborate the effect of rhizathalene on Bradysia larval feeding behavior, we designed a nonchoice in vitro experiment according to Schmelz et al. (2002). Two fungus gnat larvae were placed in glass cylinders containing a mushroom and potato agar diet with various concentrations of rhizathalene A. After 48 h of larval feeding, the average volume of diet containing 500 to 1000 ng mL⁻¹ of rhizathalene A was significantly less than control diets containing DMSO (Figure 9).

Figure 9.

Bradysia Larval Feeding on Artificial Diet Containing Rhizathalene.

Average volume of artificial diet consumed by two fungus gnat larvae in 48 h (two-way ANOVA and Tukey-Kramer HSD, P < 0.05). Bars are the mean ± se. Gray bars represent the volume consumed in the presence of different concentrations of rhizathalene shown on the x axis in comparison to DMSO controls (white bars). In a nonchoice experiment, significantly less volume was consumed of diets containing 500 to 1000 ng mL⁻¹ of rhizathalene A.

Expression of TPS08 in Response to Larval Feeding and Jasmonate Treatment

Many terpene synthases that are involved in the defense against herbivores show induced expression in response to mechanical wounding or feeding damage (Schnee et al., 2002; Huang et al., 2010). Using real-time quantitative PCR, TPS08 expression levels from aeroponically grown wild-type Arabidopsis roots exposed to 2, 4, and 5 d of Bradysia larval feeding were compared with the expression in uninfested roots. No increased TPS08 mRNA levels were found in response to larval feeding for these particular time points (see Supplemental Figure 16A online). We also tested whether TPS08 expression could be induced by treatment with the defense hormone jasmonic acid (JA). TPS08 transcript levels were slightly induced upon treatment with 100 μM JA for 24 h in axenic culture (see Supplemental Figure 16B online), but this increase did not correspond to a significant change in the amount of rhizathalene produced. However, a more stringent correlation between diterpene production and the extent of TPS08 transcription was observed in roots of plants grown under different culture conditions. The highest TPS08 transcript levels were found in roots of hydroponically grown plants corresponding to the highest concentrations of rhizathalene A in these roots, whereas the lowest accumulation of TPS08 mRNA and rhizathalene A were present in roots of plants grown in potting mix (see Supplemental Figure 16C online).

DISCUSSION

Arabidopsis Roots Produce Semivolatile Diterpenes

We found that roots of the Arabidopsis ecotype Columbia produce at least four related semivolatile diterpenes that, to our knowledge, have not been previously reported in this or any other plant species. Rhizathalene A was detected at a concentration of at most 1 μg per g Arabidopsis root fresh weight. Similar constitutive concentrations of momilactone diterpenes have been reported in the roots of rice seedlings (Kato-Noguchi et al., 2008; Toyomasu et al., 2008) with accumulation at higher levels upon fungal pathogen infection (Hasegawa et al., 2010). In the absence of secretory tissues, such as oil and resin ducts that are present, for example, in terpene-accumulating carrot (Daucus carota) and conifer roots (Hampel et al., 2005; Huber et al., 2005), diterpene production in Arabidopsis roots might be limited by potential inhibitory or autotoxic effects of the produced compounds. Moreover, it appears that rhizathalenes are not further modified by hydroxylation and glycosylation, which excludes the accumulation and sequestration of these compounds at higher concentrations within the cell as shown for other root-specific terpenes, such as triterpene saponins (Osbourn et al., 2003).

The amount of rhizathalene produced in Arabidopsis roots is dependent on the culture conditions. The increased formation of diterpenes in liquid culture in comparison to soil or aeroponically grown roots may be caused by the continuous nutrient availability under these conditions. Nutrient-dependent effects on terpene production in soil and effects of fertilizer on terpene accumulation in hydroponically grown aromatic plants have previously been reported (Davtyan, 1976; Ormeño et al., 2008). Moreover, differences in aeration and oxygen availability to roots may contribute to the differential formation of rhizathalene, although only a small increase in transcript levels of TPS08 has been reported under hypoxic conditions in microarray data sets (Lee et al., 2011).

TPS08 Catalyzes the Formation of Rhizathalenes in a Class I Type Mechanism

We demonstrated that the Arabidopsis gene TPS08 encodes an enzyme that produces rhizathalene diterpenes from the substrate all-trans GGPP in vitro and is responsible for the root-specific formation of these compounds in vivo. The primary enzymatic product, rhizathalene A, is a tricyclic diterpene with an unusual cyclopenta-spiro[5.5]undecane skeleton. Previous phylogenetic comparisons of Arabidopsis TPSs have demonstrated that TPS08 clusters with 21 Arabidopsis TPSs that belong to the large subfamily of plant TPS a-type sesquiterpene and diterpene synthases (Aubourg et al., 2002; Tholl et al., 2005; Tholl and Lee, 2011). Four type-a Arabidopsis sesquiterpene synthases have been identified: the two root-specific (Z)-γ-bisabolene synthases, TPS12 and 13 (Ro et al., 2006), and the flower-expressed sesquiterpene synthases TPS11 and TPS21 (Tholl et al., 2005). Within the Arabidopsis type-a clade, TPS08 is most closely related to six putative diterpene synthases (TPS 07, 09, 15, 16, 26, and 28; sharing 53 to 61% sequence identities) (Aubourg et al., 2002; Tholl and Lee, 2011), two of which (TPS09 and 26) are also expressed in Columbia roots and the other four of which appear to be primarily expressed in seeds according to microarray data sets (Tholl and Lee, 2011). Whether these TPSs actively produce terpenes in the Columbia accession remains to be determined since no other diterpene products have so far been reported in Arabidopsis roots and seeds.

When compared with diterpene synthases from other plants, TPS08 shows closest similarity to enzymes of the TPS-a subfamily, such as casbene diterpene synthase (32% sequence identity) from Ricinus communis (Mau and West, 1994; Tholl et al., 2005). According to models for the structural evolution of plant TPS proteins (Cao et al., 2010), TPS08 and other TPS-a type enzymes represent modern class I TPSs with a functionally active C-terminal or α-domain containing the highly conserved Asp-rich DDXXD sequence motif, which is essential in positioning the diphosphate substrate for initial ionization (Davis and Croteau, 2000). We therefore propose that the formation of rhizathalene A follows a class I type carbocationic reaction mechanism (Figure 10). The mechanism shown is consistent with the results of quantum chemical calculations on all proposed intermediates (A, C1/2, D, and E), transition state structures, and the reaction coordinates connecting them (details are provided in Supplemental Figures 11 and 12, Supplemental Tables 2 to 5, and Supplemental Data Set 1 online; Tantillo, 2011). The reaction is initiated by the ionization of the diphosphate followed by an isomerization of the 2,3 double bond to form a tertiary allylic geranyllinalyl diphosphate intermediate (GLPP). The isomerization step allows for a subsequent ring closure between C1 and C6 (numbered as in GGPP) to generate the cyclohexene moiety of the first carbocation intermediate (A). Cation A is then converted directly to cation C (conformer C1) via a concerted hydride shift/spirocyclization reaction (Tantillo, 2008, 2010). This step requires a rotation around the C6–C7 bond, which allows for a favorable alignment of the C6–H bond with the formally vacant p-orbital at C7 (see Supplemental Figure 12 online). Cation B was not found to be an intermediate in the conversion of A to C. Following a change in conformation (C1-to-C2), a ring-expanding 9,11 Wagner-Meerwein rearrangement leads to cation D, which is not itself an intermediate but is encountered along the reaction coordinate from C2 to E (Tantillo, 2010, 2011), and subsequent 14,10 ring-closure, completing the construction of the polycyclic framework of rhizathalene A. A final deprotonation step forms the 15,16 double bond. The overall A-to-E reaction is predicted to be exothermic by ∼4 kcal/mol, and the highest energy transition state structure (that connects A to C1) is predicted to lie only ∼7 kcal/mol above carbocation A.

Figure 10.

Proposed Reaction Pathway from GGPP to Rhizathalene A.

Atom numbering refers to that of GGPP. Note that the discontinuous numbering of C9, C10, and C11 in rhizathalene A is a result of the rearrangement step C2→D. Quantum mechanical calculations predict that B and D are not energy minima. The relative stereochemistry of rhizathalene A predicted from the ROESY plot is shown in Figure 3C.

Whereas the formation of carbocation A via GLPP has not been established experimentally as a reaction catalyzed by other plant diterpene synthases, the initial steps in the reaction pathway are analogous to those that generate the bisabolyl cation from a nerolidyl diphosphate intermediate in the biosynthesis of cyclic sesquiterpenes (Cane, 1999; Hong and Tantillo, 2009), and the spirocyclization event is analogous to that leading to the acoradiene sesquiterpenes (Hong and Tantillo, 2009). Furthermore, the natural occurrence of the spiro[4.5]decane diterpene viscida-3,11(18),14-triene (see Supplemental Figure 13 online; Tesso et al., 2005), clearly an isoprenylog of the well-known β-acoradiene (see Supplemental Figure 13 online) (Ghisalberti et al., 1984), affords additional precedent for the initial steps in the mechanism in Figure 10 (i.e., GGPP → GLPP → A → B → C). Interestingly, the flower-specific Arabidopsis sesquiterpene synthase TPS11, which clusters with TPS08 in a TPS-a type clade, produces the spiro[5.5]undecane type sesquiterpenes α-chamigrene (see Supplemental Figure 13 online) and β-chamigrene, among several other cyclic sesquiterpenes by a reaction cascade from a central bisabolyl cation (Tholl et al., 2005), indicating mechanistic similarities of TPS11 and TPS08 in the formation of their spiro terpene products. However, when tested with the substrate (E,E)-farnesyl diphosphate, the recombinant TPS08 protein does not produce sesquiterpenes nor does the enzyme make monoterpenes from GPP.

As indicated by the proposed reaction pathway, TPS08 does not initiate carbocation formation by protonation, a mechanism typical for class II TPSs, such as copalyl diphosphate (CPP) synthases of the TPS c-subfamily, and bifunctional class I/class II TPSs, such as abietadiene synthase that convert all-trans GGPP to CPP as the final product or intermediate, respectively (Sun and Kamiya, 1994; Peters and Croteau, 2002). Accordingly, the N-terminal or β-domain of TPS08 and other TPS a-type proteins lack the DxDD motif characteristic for class II type enzymes. Noticeably, the architecture of the TPS08 protein differs from those of class I type diterpene synthases of the gymnosperm-specific TPS d-subfamily, such as taxadiene synthase from Taxus brevifolia (Köksal et al., 2011) and diterpene synthases of the TPS e/f-subfamily by the absence of an ancestral-type 200–amino acid γ-domain. Furthermore, diterpene synthases in the TPS e/f-subfamily, such as kaurene synthases involved in gibberellin biosynthesis (Yamaguchi et al., 1998) and labdane-related rice diterpene synthases, do not accept all-trans GGPP but only use CPP, ent-CPP, or syn-CPP as substrates (Peters, 2006).

Although TPS08 is related to sesquiterpene synthases in the Arabidopsis TPS-a type clade and its reaction mechanism shows obvious similarities to that of the sesquiterpene cyclases in this group, the TPS08 enzyme does not produce sesquiterpenes. It can be assumed that structural differences at the active site of sesquiterpene and diterpene synthases in this clade are responsible for the enzymatic discrimination in the conversion of the different sized prenyl diphosphate substrates. Identifying these differences will provide insight into the structural differentiation and evolutionary plasticity in sesquiterpene and diterpene formation in the TPS-a subfamily.

Rhizathalene Synthase Is targeted to Root Leucoplasts

With the exception of Arabidopsis geranyllinalool synthase, which is located in the cytosol (Herde et al., 2008), all plant diterpene synthases characterized so far carry an N-terminal transit peptide for a presumed protein transport to plastids (Bohlmann et al., 1998). Similarly, rhizathalene synthase contains a 41–amino acid predicted plastidial transit peptide and fluorescent protein fusion experiments indicated that TPS08 produces rhizathalenes from GGPP in root leucoplasts. The formation of rhizathalene A in plastids via the plastid-specific MEP pathway is further supported by the incorporation of the deuterium-labeled precursor DOX-d₂ into the diterpene compound (Figure 5). Similar observations have been made, for example, for the incorporation of DOX-d₂ into plastid-specific monoterpenes produced by elicitor-induced leaves of lima bean (Phaseolus lunatus; Bartram et al., 2006) and into the monoterpene terpinolene in the phloem of carrot roots (Hampel et al., 2005). In contrast with these results, the degree of labeling of rhizathalene did not exceed 50% when DOX-d₂ concentrations were increased from 0.2 to 2 mg/mL of DOX-d₂, which could be attributed to possible feedback regulatory mechanisms at higher DOX concentrations. Alternatively, elevated DOX levels may lead to an enhanced export of isopentenyl diphosphate-d₂ from plastids into the cytosol, or the limited degree of labeling could be due to the contribution of nonplastidial isopentenyl diphosphate pools to the biosynthesis of rhizathalene. Given the low expression levels of several MEP pathway enzymes in Arabidopsis roots (Estévez et al., 2000; Guevara-García et al., 2005), we hypothesized that precursors derived from cytosolic MVA could contribute to the formation of rhizathalene. However, the low incorporation rate of labeled mevalonolactone (MVL-d₂) into rhizathalene (Figure 5A) indicated a limited role of the MVA pathway in providing C₅-precursors for the leucoplast-specific formation of rhizathalene.

Rhizathalene Is Produced in the Stele of Arabidopsis Roots and Contributes to the Direct Defense against a Root Herbivore

Analysis of GUS activity and YFP expressed under control of the TPS08 promoter suggested specific expression of TPS08 in the root stele and to some extent in the tips of primary and secondary roots (Figure 6). These observations are largely in agreement with results obtained from high-resolution, cell type–specific gene transcript maps of Arabidopsis roots showing highest expression of the TPS08 transcript in the procambium and the xylem-pole pericycle (Brady et al., 2007) of all root developmental zones (see Supplemental Figure 17 online).

Cell type or tissue specificity is a common feature in the organization of terpene metabolism or other plant specialized metabolites and has been discussed with regard to its significance in the optimal defense against herbivores and pathogens or the attraction of beneficial organisms. For example, the accumulation of terpenes and alkaloids in secretory defense tissues or cells, such as resin ducts, laticifers, and glandular trichomes, allows plants to mount potent chemical defenses in stems and in the epidermis of leaves and fruits (Hagel et al., 2008; Schilmiller et al., 2008). To facilitate the attraction of pollinating insects, snapdragon (Antirrhinum majus) flowers synthesize volatile benzenoids in the cells of the inner epidermal layer of petal lobes and the corolla tube that are exposed to the pollinator during landing (Kolosova et al., 2001). In Arabidopsis roots, the formation of terpenes appears to be largely confined to specific cell types. Besides the expression of TPS08 in the stele tissue, cell-type specificity is also apparent for other TPS genes, such as the 1,8-cineole synthases, TPS23 and TPS27, with activity in epidermal cells, and the (Z)-γ-bisabolene synthases, TPS12 and TPS13, in the cortex. A highly coordinated expression of the genes of triterpene biosynthesis gene clusters was reported in Arabidopsis root epidermal cells (Field and Osbourn, 2008). TPS08 appears to be coexpressed with other genes encoding cytochrome P450 enzymes (At4g15380, At3g20110, At5g4258, and At5g3611) (http://www-ibmp.u-strasbg.fr/~CYPedia/). However, it seems unlikely that these enzymes catalyze a conversion of rhizathalenes to potential alcohol or acid derivatives, since no such derivatives have been detected in organic extracts of Arabidopsis (Columbia) roots.

A comparison of root tissue damage on wild-type plants and the tps08 gene knockout line caused by feeding of Bradysia larvae revealed clear differences in the extent of damage with regard to the removal of epidermal, cortex, and endodermal cells at feeding sites (Figures 8E to 8H and 8I to 8L). Xylem tissue was not consumed, most likely because of the higher degree of cell lignification. The observed difference in feeding damage and the results from in vitro feeding experiments suggest that rhizathalenes function as antifeedants. Potent antiherbivore activities of diterpenes have been reported in other plants such as for 17-hydroxygeranyllinalool glycosides in leaves of Nicotiana attenuata (Jassbi et al., 2010) and more recently for kauralexins produced in maize (Zea mays; Schmelz et al., 2011). To allow for an optimal defense against a chewing insect, one would have expected that rhizathalenes are produced in the outer cell layers of the root rather than in the central root tissue. The formation of the diterpene compounds at effective concentrations in the pericycle might critically depend on an active terpenoid metabolism in the leucoplasts of these cells with immediate carbon supply by the vascular tissue. By contrast, only trace amounts of the monoterpene 1,8-cineole are produced by the Arabidopsis root epidermis despite substantial expression levels of the respective 1,8-cineole synthases in these cells (Chen et al., 2004). Moreover, we assume that the produced diterpenes could serve multiple functions by protecting roots from soil-borne microbial pathogens that invade plants by penetrating the root vasculature, such as root knot nematodes (Sijmons et al., 1994) and the root-infecting fungi Fusarium and Verticillium (Parry et al., 1995; Klosterman et al., 2009). The stele-specific formation of terpenes as reported in the phloem and xylem of carrot roots might serve similar functions (Hampel et al., 2005). Kauralexin diterpenes produced in maize have been characterized both for their phytoalexin and antifeedant activities (Schmelz et al., 2011), and diterpene phytoalexins synthesized in rice roots might have similar multitargeted effects. Multifunctionality is also apparent for volatile terpenes emitted from specific floral tissues in the way that they serve as antimicrobials as well as possible pollinator attractants (Huang et al., 2012).

Given the semivolatile nature of rhizathalenes, diffusion of these compounds from the root stele most likely circumvents costly transport mechanisms that are required for the secretion of many nonvolatile metabolites into the rhizosphere (Badri and Vivanco, 2009). Moreover, a central radial release of diterpenes may have effects on the expression of defense genes or their priming in the surrounding cell layers. Priming activities by volatile compounds including terpenes have been observed in induced inter- and intraplant signaling responses in several plants (Engelberth et al., 2004; Frost et al., 2007). While these interactions have been exclusively documented between different tissues, volatile compounds may also play a role as priming signals within the same plant organ.

The production of diterpenes in Arabidopsis roots most likely contributes to other more prominent chemical defenses, such as glucosinolates. Root-specific defensive activities of glucosinolates against herbivores are known from studies with Brassicaceae (Hopkins et al., 2009). Adding to these activities, our results show that the formation of relatively low levels of volatile terpene metabolites, even when confined to specific cell types in plant roots, can considerably affect feeding patterns of root herbivores.

METHODS

Plant Material and Insect Culture

Arabidopsis thaliana plants were grown under controlled conditions at 22 to 25°C, 150 μmol m⁻² s⁻¹ PAR, and a 10-h-light/14-h-dark photoperiod. All wild-type plants were ecotype Columbia (Col-0 CS60000). Seeds of the following T-DNA insertion lines were ordered from the ABRC stock center (Alonso et al., 2003): SALK 125194 (tsp08-1), SALK 112521 (tps08-2), SALK_102259C (tps17-1), SALK 065023C (tps17-2), SALK 112196 (tps20-1), SALK 112203 (tps20-2), SALK 57381 (tps26-1), SALK 133064 (tps26-2), SALK 138882 (tps29-1), and SAIL 728 E10 (tps29-2). Plants were grown in potting mix (90% Sunshine mix #1 [Sun Gro Horticulture] and 10% sand) for 2 to 4 weeks.

Hydroponic cultures were established by transferring 4-week-old plants grown in potting mix to plastic containers containing Hoagland solution under constant aeration as previously described (Gibeaut et al., 1997). Plants were grown for 3 to 4 weeks in hydroponic culture before volatile analysis. Axenic cultures were prepared by following the procedures described by Hétu et al. (2005). Two days before harvesting the root tissue, the medium was changed to reduce the Suc content to 1%. Hormone treatments with JA were applied to axenically grown roots by adding 100 µM JA (Sigma-Aldrich) directly to the liquid medium. Treated tissue was collected 24 h after application of the hormone. Aeroponic cultures were prepared as described by Vaughan et al. (2011) using 50-mL plastic conical tubes (Fisher Scientific) filled with Seramis clay granules. Plants were grown for 4 weeks until the initiation of primary bolt production.

Fungus gnat larvae were reared on 4 liters of moist Sunshine mix #1 enriched with 1.5 kg of shredded potato as described by Vaughan et al. (2011). Specimens were identified as a mixed colony of Bradysia coprophila and Bradysia impatiens.

Reagents and Radiochemicals

Unlabeled GPP, farnesyl diphosphate, and GGPP were purchased from Echelon Biosciences. Tritium labeled GGPP ([1-³H]-GGPP ∼0.74 TBq mmol⁻¹) was purchased from American Radiolabeled Chemicals. [5,5-²H₂]-DOX and MVL-d₂ were kindly provided by Wilhelm Boland (Max Planck Institute for Chemical Ecology, Jena, Germany).

Bradysia Feeding Experiments with Plants Grown in Aeroponic Culture

Larvae were isolated from the culture medium using a modified flotation/extraction procedure as described by Vaughan et al. (2011). Larvae of all four instars were placed in a Petri dish on moist filter paper without food for 20 to 24 h prior to feeding assays. Application of larvae to aeroponically grown plants followed the protocol developed by Vaughan et al. (2011). Approximately 200 to 300 second- and third-instar larvae, collected as described above, were transferred to a single culture tubes in 1 mL of Hoagland solution pipetted into the clay substrate. First and fourth instar larvae were excluded from the experiments because they were too small to count and too close to pupation, respectively. During the feeding experiment, plants were watered every other day by applying 10 mL of Hoagland solution through the top of the tube. Cultures were kept in a screened box under the same conditions described above. Five days after larval feeding, roots were removed from the tubes and the clay granules by submerging the tubes in water. Feeding damage on roots was examined under a dissecting microscope, and the average area of tissue consumed per feeding site was determined using ImageJ software (Collins, 2007). Roots were dried with Kimwipes (Fisher Scientific) and the fresh weight recorded. After drying for 2 d at room temperature, the dry weight was measured. For data analysis, the weights from two independent experiments were combined and log transformed before performing a two-way analysis of variance (ANOVA) followed by Tukey-Kramer Honestly Significant Difference (HSD), n = 24 for the wild type and tps08-1 knockout, and n = 12 for T-DNA control. The percentage of root mass consumed was calculated for each individual plant in comparison to its untreated control.

Bradysia Feeding Experiments with Seedlings Grown in Potting Mix

Seedlings were grown in individual pots (8 cm × 8 cm × 10 cm) under the described conditions. Fourteen-day-old seedlings were challenged with 10 second- and third-instar Bradysia larvae isolated and transferred to the pots as described above. Larvae that did not move into the potting substrate within 5 min of the transfer were removed from the experiment and replaced. Pots with treated and control seedlings were then randomly placed in separate net enclosures and kept under growth room conditions as described above. After 7 d of larval feeding, seedlings were removed from the soil and roots were rinsed in water. Seedlings were allowed to air dry for 2 d to record dry weight. Cumulative log-transformed data from two independent experiments were analyzed with two-way ANOVA and Tukey-Kramer HSD, n = 20. The percentage of biomass reduced was calculated for each individual plant in comparison to its noninfested control.

Bradysia Feeding Experiments on Artificial Diet Containing Purified Rhizathalene A

An artificial diet of lyophilized and powdered mushroom and potato agar was prepared using a slightly modified protocol previously described (Schmelz et al., 2002). Diets were composed of 0.75 g of lyophilized and powdered mushroom (Agarcus bisporus) and potato (1:1), 0.25 g of agar, and 18 mL of distilled water. The mixture was microwaved to a boil, allowed to cool, and then divided into aliquots to which different amounts of DMSO or purified rhizathalene A (dissolved in DMSO) were added. Diets contained final concentrations of 0, 50, 100, 500, and 1000 ng mL⁻¹ of rhizathalene A. The agar was poured into Petri plates (60 × 15 mm) and allowed to solidify. Glass cylinders (height = 60 mm; diameter = 6 mm) were used to punch out 282 mm² of mushroom and potato agar (height =10 mm; radius = 3 mm). Diet feeding assays were performed by placing two second- and third-instar larvae into the opened end of the glass cylinder. The cylinders containing the larvae were then placed in covered Petri dishes (100 × 15 mm) on moistened filter paper. After the larvae had fed for 24 h, the volume of diet consumed was calculated by measuring the remaining height and diameter of the diet in the cylinder. The average diet consumed was analyzed for differences by two-way ANOVA and Tukey-Kramer HSD.

Volatile Collection and Analysis

Volatiles were collected from roots by automated SPME using a AOC-5000 Shimadzu autosampler. One gram of root tissue was detached and placed in a 20-mL screw cap glass vial (Supelco). Volatiles were adsorbed in the headspace with a 100-µM polydimethylsiloxane fiber (Supelco) for 30 min at 30°C. Volatile analysis was performed using a gas chromatograph (GC-2010; Shimadzu) coupled with a quadrupole mass spectrometer (GC-MS-QP2010S; Shimadzu). Samples were separated with a 2:1 split injection using the same conditions and mass spectral analysis as described by Huang et al. (2010). Quantification of rhizathalene A was conducted by GC-FID analysis of organic root extracts (see below) using the authentic diterpene standard, cembrene, for external calibration. A Shimadzu 2010 GC-FID was used with the same column and temperature program as for GC-MS.

Organic Extraction and Quantitative Analysis of Arabidopsis Root Terpenes

Volatile terpenes were extracted from 2 g of roots grown under the different culture conditions. Root tissue was rinsed with water, dried with tissue paper, and immediately submerged in 10 mL of ethyl acetate. Roots were then ground in the presence of the organic solvent and 10 mL of deionized water. The ground material was mixed with 90 mL each of ethyl acetate and deionized water in a separatory funnel for 10 min. After phase separation, the organic phase was collected and concentrated to ∼100 μL under a gentle air flow. Following the addition of 5 mL of hexane, the extract was passed through a small silica column overlaid with magnesium sulfate and preconditioned with hexane. The flow-through and a second 4-mL fraction were combined and concentrated again to ∼250 μL. One microliter of the extract was analyzed by GC-MS and GC-FID analysis.

Deuterium Labeling of Rhizathalene A

For feeding experiments with deuterium-labeled precursors, ∼200 mg of 2-week-old hairy roots grown at room temperature in Gamborg’s B5 liquid media with 2% Suc under semidark conditions (70 rpm) were transferred into a 10-mL sterile screw cap vial. One milliliter medium containing 2 mg of [5,5-²H₂]-DOX or [2,2-²H₂]-MVA was added to the vial to incubate for 24 or 48 h under the same conditions described above. Since a higher degree of labeling was achieved at 48 h of incubation, this time was chosen for incubations with additional concentrations of 0.02 and 0.2 mg/mL [5,5-²H₂]-DOX. Rhizathalene A production was analyzed by SPME–GC-MS analysis. The degree of labeling was calculated based on the ratio of the peak area of labeled rhizathalene A to the total peak area of labeled and unlabeled product. All treatments were replicated three times.

Isolation and Cloning of TPS08 cDNA

Total RNA was isolated from 150 mg of axenically grown Arabidopsis roots using the TRIzol method according to Huang et al. (2010). Two micrograms of total RNA were reverse transcribed into cDNA according to the manufacturer’s protocol (Invitrogen) and primers A and B (see Supplemental Table 6 online) corresponding to the start and the end of the coding region of TPS08 were used for PCR amplification of the TPS08 cDNA. The resulting PCR product was cloned into the pET102/D-TOPO vector (Invitrogen) following the manufacturer’s protocol.

Genotyping of Plant Material and Complementation Analysis

Mutants carrying a T-DNA insertion in the TPS08 gene were identified in the insertional mutant population obtained from the ABRC stock center (Alonso et al., 2003). The presence of the T-DNA insertion in line SALK_12594 (tps08-1) was confirmed by PCR and sequencing of the right and left border PCR products. To determine the copy number of T-DNA insertions in the SALK_12594 line, genomic DNA was isolated from leaves. DNA gel blot analysis was performed with 50 µg of DNA digested with BamHI and HindIII and a digoxigenin-labeled T-DNA–specific probe designed with primers C and D (see Supplemental Table 6 online) following the hybridization protocol by Smith and Summers (1980).

For complementation of the tps08 mutant, the TPS08 open reading frame including the stop codon was amplified from TPS08 cDNA using primers A and B and a 2.2-kb TPS08 promoter fragment was amplified from genomic DNA using primers E and F (see Supplemental Table 6 online). The amplicons were each cloned into the Gateway entry vector pENTR/D-TOPO (Invitrogen) and recombined into the binary vector pB7Y24WG. pB7Y24WG was derived from pB7YWG2 (Karimi et al., 2002) by replacing the attR1 site with an attR4 site. The ProTPS08:TPS08 construct was introduced into Agrobacterium tumefaciens strain GV3101 and transformed into SALK_125194 plants by floral vacuum infiltration. Transformants were screened on half-strength Murashige and Skoog plates with 1% (w/v) Suc and 0.1 mg/mL kanamycin.

Heterologous Expression in Escherichia coli, Purification of the TPS08 Protein, and Terpene Synthase Assays

A truncated version of TPS08 without 123 bp encoding a predicted 41–amino acid N-terminal plastidial transit peptide was amplified from the TPS08-pET102/D-TOPO construct by PCR using primers G and H (see Supplemental Table 6 online). The PCR product was cloned into the pET102/D-TOPO expression vector (Invitrogen) to generate a translational N-terminal fusion to thioredoxin and a C-terminal fusion to a His₆-tag. The resulting plasmid was transformed into E. coli BL21 Codon Plus cells (Stratagene). Expression and partial purification of the TPS08 protein using nickel-nitrilotriacetic acid (Ni-NTA) agarose was performed as described by Chen et al. (2003) and Tholl et al. (2005).

Enzyme assays were conducted as described previously by Tholl et al. (2005) in a total volume of 1 mL for 30 min at 30°C in a 10-mL screw cap vial (Supelco). Enzyme products were collected during assay incubation by automated SPME and subsequently analyzed by GC-MS as described above. For enzyme characterization, assays were performed in a final volume of 50 μL with 1 µg partially purified TPS08 enzyme and 10 µM [1-³H]-GGPP (0.74 TBq mmol⁻¹). Assay conditions and quantification of the radioactive product were as described by Tholl et al. (2005). Six different concentrations of [1-³H]-GGPP were applied to determine the K_m value for GGPP in triplicate assays. Calculation of K_m and V_max values was performed by Hanes plot analysis using the Hyperbolic Regression Analysis (HYPER 1.01) software (J.S. Easterby, University of Liverpool). The divalent metal ion dependency of TPS08 was determined by performing assays without MgCl₂ and MnCl₂.

Structural Elucidation of Rhizathalene A

Larger amounts of rhizathalene A were produced in vivo to conduct structural analysis of the diterpene hydrocarbon. The truncated TPS08-pET102/D-TOPO construct was transformed into E. coli strain C41(DE3) (Invitrogen), which carried a recombinant pseudomature GGPP synthase (GGPS) from grand fir (Abies grandis) rAgGGPS in the pACYCDuet vector (Cyr et al., 2007) (kindly provided by Reuben Peters, Iowa State University, Ames, IA). Bacteria were grown at 37°C to an OD₆₀₀ of 0.7 to 0.9 and induced with 0.5 mM Isopropyl β-D-1-thiogalactopyranoside (IPTG) overnight at 24°C. The induced bacterial culture (250 mL) was transferred to a desiccator, and the diterpene volatile was collected in the culture headspace using 25 mg of Super-Q traps and a closed-loop stripping procedure described previously (Huang et al., 2010). Compounds were eluted from the volatile traps twice with 100 μL of CH₂Cl₂, and the eluate was concentrated under nitrogen to a few microliters and then resuspended in 1 mL of 100% methanol. The procedure was repeated 10 times to obtain sufficient quantities of diterpene compound for NMR analysis. Rhizathalene A was then purified by liquid chromatography–mass spectrometry (Walters) using an X-Bridge column (4.6 × 250 mm, 5-µm particle size). A solvent gradient of 80 to 90% methanol in 1% (v/v) formic acid was applied over 5 min, and the final concentration was held for 20 min; the flow rate was 0.8 mL min⁻¹. The injection volume varied from 20 to 100 μL. The fraction containing rhizathalene A at 95 to 99% purity was collected between 16 and 17.5 min as determined by mass spectrometry analysis.

All NMR spectra were acquired at 30°C with a Bruker AVANCE 600 spectrometer (600-MHz ¹H frequency) equipped with a 5-mm triple resonance z-gradient probe (see additional information in Supplemental Methods 1 online). NMR chemical shift calculations were performed as described by Lodewyk et al. (2012). The Gaussian03 software suite (Frisch et al., 2004; Gaussian03, revision D.01) was used for all calculations. Structures were optimized in the gas phase using the B3LYP/6-31+G(d,p) method. For NMR chemical shift calculations, single point calculations (GIAO) at the B3LYP/6-311+G(2d,p) level in an implicit chloroform solvent continuum (CPCM, UACK radii) were performed. All reported chemical shifts were scaled as described at http://cheshirenmr.info. The scaled chemical shifts were also subjected to DP4 probability analysis (Smith and Goodman, 2010). Further details and references are provided in Supplemental Methods 1 and in Supplemental Data Set 1 online.

Construction and Analysis of ProTPS08:GUS and TPS08-GFP/YFP Fusions

A ProTPS08:GUS construct was generated by recombining the 2.2-kb TPS08 promoter fragment from the Gateway entry vector pENTR/D-TOPO into the binary vector pKGWFS7 (Karimi et al., 2002). Following transformation using Agrobacterium GV3101, histochemical GUS assays were performed as previously described (Jefferson et al., 1987). To determine the subcellular localization of the TPS08 protein, C-terminal translational fusions of the 41–amino acid N-terminal TPS08 peptide were established with eGFP and enhanced YFP under the control of the CaMV 35S and the native TPS08 promoters, respectively. eGFP fusion constructs were generated by PCR amplification using primers I and J (see Supplemental Table 6 online) and cloning of the 123-bp N-terminal transit peptide into the pENTR/D-TOPO vector and recombination into the binary vector pK7FWG2 carrying the 35S promoter (Karimi et al., 2002). TPS08 promoter constructs were established by recombining the 123-bp amplicon into the Gateway-compatible pB7Y24WG vector carrying the 2.2-kb TPS08 promoter fragment and the enhanced YFP coding sequence. Generation of transgenic plants was completed as described above. Two-week-old transformants grown on half-strength Murashige and Skoog, 1% (w/v) Suc, and 50 µM glufosinate were used for fluorescence imaging by confocal laser scanning microscopy using a LSM 510 microscope (Carl Zeiss) equipped with a HeNe laser. Tissue autofluorescence was excited at 543 nm and GFP/YFP fluorescence at 488 nm. Band-pass was set to 500 to 550 nm, and long-pass was set to 560 nm. Bright-field images were acquired with the differential interference contrast channel. Images were processed using the Zeiss image software ZEN 2009 light edition.

Transcript Analysis by RT-PCR and Quantitative Real-Time PCR

RT-PCR was performed with crude RNA extracted from 150 mg of root tissue as described by Huang et al. (2010) using TPS08 internal primers K and L and Actin 8 primers M and N (see Supplemental Table 6 online). Primers A and B were used to verify TPS08 expression in Pro35S:TPS08 plants.

Quantitative real-time PCRs followed the procedure described by Lee et al. (2010) using TPS08-specific primers O and P (see Supplemental Table 6 online). Relative expression levels were determined for three independent biological replicates with three technical replicates each. Threshold cycle (Ct) values for TPS08 were normalized to protein phosphatase 2A subunit A3 (PP2AA3) using primers Q and R (see Supplemental Table 6 online).

Quantum Chemical Calculations on the Mechanism of Formation of Rhizathalene A

Mechanistic calculations were performed with Gaussian03 (Frisch et al., 2004; Gaussian03, revision D.01, Gaussian). Structures were optimized in the gas phase using the B3LYP/6-31+G(d,p) method, and mechanistic conclusions are based primarily on mPW1PW91/6-31+G(d,p)//B3LYP/6-31+G(d,p) single point energies. All stationary points were characterized by frequency calculations, and reported energies include B3LYP/6-31+G(d,p) zero-point energy corrections (unscaled). Intrinsic reaction coordinate calculations were used for further characterization of all transition state structures. See Supplemental Methods 1 online for additional details and references to computational methods.

Statistical Analysis

Specific data analysis performed has been described with the individual experiments. ANOVA was accomplished using JMP statistical software (SAS Institute).

Accession Numbers

Sequence data from this article can be found in the Arabidopis Genome Initiative or GenBank/EMBL databases under the following accession numbers: TPS08 (At4g20210), TPS12 (At4g13280), TPS13 (At4g13300), TPS17 (At3g14490), TPS20 (At5g48110), TPS23 (At3g25820), TPS26 (At1g66020), TPS27 (At3g25830), TPS29 (At1g31950), Actin 8 (At1g49240), and PP2AA3 (At1g13320).

Supplemental Data

The following materials are available in the online version of this article.

• Supplemental Figure 1. DNA Gel Blot Analysis of T-DNA Copy Numbers in tps08-1.

• Supplemental Figure 2. Complementation Analysis of the tps08-1 Line.

• Supplemental Figure 3. ¹H NMR spectrum of Rhizathalene A.

• Supplemental Figure 4. COSY Spectrum of Rhizathalene A.

• Supplemental Figure 5. ROESY NMR Spectrum of Rhizathalene A.

• Supplemental Figure 6. TOCSY NMR Spectrum of Rhizathalene A.

• Supplemental Figure 7. HSQC Spectrum of Rhizathalene A.

• Supplemental Figure 8. HMBC Spectrum of Rhizathalene A.

• Supplemental Figure 9. Proposed Structure and Experimental Chemical Shift Assignments.

• Supplemental Figure 10. All Possible Diastereomers (without Enantiomers) for the Structure Shown in Supplemental Figure 9.

• Supplemental Figure 11. Assignment of the Relative Stereochemistry to the Compound (See Supplemental Figure 9 Online) Based on GIAO NMR Calculations and DP4 Probability.

• Supplemental Figure 12. Rearrangements of Cation A to the Cation E That Proceeds to 1’.

• Supplemental Figure 13. The Structures of the Spiro Terpenes α-Chamigrene, β-Acoradiene, and Viscida-3,11(18),14-Triene.

• Supplemental Figure 14. Confocal Laser Scanning Microscopy of Wild-Type and GFP-Transformed Roots.

• Supplemental Figure 15. Bradysia Larval Feeding Damage on Roots of Aeroponically Grown Wild Type and the tps08-1 Mutant.

• Supplemental Figure 16. TPS08 Expression under Different Culture Conditions and in Response to Insect and Hormone Treatment.

• Supplemental Figure 17. Relative Abundance of the TPS08 Transcript According to the High-Resolution Spatiotemporal Map by Brady et al. (2007).

• Supplemental Table 1. Proton and Carbon Chemical Shifts for Rhizathalene A and Key Correlations from the 2D NMR Spectra.

• Supplemental Table 2. Computed ¹³C Chemical Shifts (in ppm).

• Supplemental Table 3. Computed ¹H Chemical Shifts (in ppm).

• Supplemental Table 4. Statistical Analysis on ¹³C Chemical Shift Data.

• Supplemental Table 5. Statistical Analysis on ¹H Chemical Shift Data.

• Supplemental Table 6. Sequences of Primers Used in the Different Experiments.

• Supplemental Data Set 1. Coordinates, Energies, NMR Isotropic Shielding Constants, and GIAO NMR Calculations.

• Supplemental Methods 1.

Glossary

GPP

geranyl diphosphate

GGPP

geranylgeranyl diphosphate

MEP

methylerythritol phosphate

SPME

solid phase microextraction

GC-MS

gas chromatography–mass spectrometry

GC-FID

gas chromatography–flame ionization detection

ROESY

rotating frame overhauser enhancement spectroscopy

CaMV

cauliflower mosaic virus

eGFP

enhanced green fluorescent protein

GFP

green fluorescent protein

YFP

yellow fluorescent protein

GUS

β-glucuronidase

JA

jasmonic acid

GLPP

geranyllinalool diphosphate

CPP

copalyl diphosphate

ANOVA

analysis of variance

HSD

honestly significant difference