• Open Access

Chrysanthemum expressing a linalool synthase gene ‘smells good’, but ‘tastes bad’ to western flower thrips

Authors


  • Gene accession number: FaNES1: CAD57083.

Correspondence (Tel +31 317 480 932; fax +31 317 41 80 94; email maarten.jongsma@wur.nl)

Summary

Herbivore-induced plant volatiles are often involved in direct and indirect plant defence against herbivores. Linalool is a common floral scent and found to be released from leaves by many plants after herbivore attack. In this study, a linalool/nerolidol synthase, FaNES1, was overexpressed in the plastids of chrysanthemum plants (Chrysanthemum morifolium). The volatiles of FaNES1 chrysanthemum leaves were strongly dominated by linalool, but they also emitted small amount of the C11-homoterpene, (3E)-4,8-dimethyl-1,3,7-nonatriene, a derivative of nerolidol. Four nonvolatile linalool glycosides in methanolic extracts were found to be significantly increased in the leaves of FaNES1 plants compared to wild-type plants. They were putatively identified by LC-MS-MS as two linalool–malonyl–hexoses, a linalool–pentose–hexose and a glycoside of hydroxy–linalool. A leaf-disc dual-choice assay with western flower thrips (WFT, Frankliniella occidentalis) showed, initially during the first 15 min of WFT release, that FaNES1 plants were significantly preferred. This gradually reversed into significant preference for the control, however, at 20–28 h after WFT release. The initial preference was shown to be based on the linalool odour of FaNES1 plants by olfactory dual-choice assays using paper discs emitting pure linalool at similar rates as leaf discs. The reversal of preference into deterrence could be explained by the initial nonvolatile composition of the FaNES1 plants, as methanolic extracts were less preferred by WFT. Considering the common occurrence of linalool and its glycosides in plant tissues, it suggests that plants may balance attractive fragrance with ‘poor taste’ using the same precursor compound.

Introduction

Plant volatiles play important roles in plant–insect interactions (Maffei et al., 2011; Pichersky and Gershenzon, 2002; Schoonhoven et al., 2005), and those induced by herbivore attack are often found to have roles in direct defence by repelling herbivores, and indirect defence by attracting predators or parasitoids (Clavijo McCormick et al., 2012; Dicke and Baldwin, 2010; Dicke and Van Loon, 2000; Paré and Tumlinson, 1999). Herbivore-induced plant volatiles are usually dominated by mono- and sesquiterpenes (Degenhardt et al., 2003), and enhanced emissions of these terpenes have led to improved plant defence against herbivores (Dudareva and Pichersky, 2008; Turlings and Ton, 2006).

Linalool is a monoterpene alcohol with a sweet fragrance occurring in the floral scent of a wide variety of plants (Kamatou and Viljoen, 2008). It is also reported to be induced in different plants by damage of a variety of herbivore species, suggesting that linalool may have a role in direct or indirect plant defence against several herbivores. For example, it is induced in crabapple damaged by Japanese beetles (Loughrin et al., 1995), cotton and peanut damaged by beet armyworms (Cardoza et al., 2002; Paré and Tumlinson, 1997), maize damaged by caterpillars (Turlings et al., 1998), Nicotiana attenuata damaged by caterpillars, leaf bugs or flea beetles (Kessler and Baldwin, 2001), spruce damaged by white pine weevils (Miller et al., 2005), lima bean damaged by caterpillars (Mithöfer et al., 2005) or spider mites (Dicke et al., 1990), and tobacco damaged by western flower thrips (WFT) or caterpillars (Delphia et al., 2007). Constitutive high emission of linalool has been engineered in several transgenic plant species by overexpressing the linalool synthase gene (Aharoni et al., 2003; Aharoni et al., 2006; Lavy et al., 2002; Lewinsohn et al., 2001; Yang, 2008). Among these, linalool-emitting Arabidopsis were found to be less attractive to aphids and diamondback moths than wild-type plants, but it remained unclear whether linalool or linalool derivatives were responsible for the observed effects (Aharoni et al., 2003; Yang, 2008).

In the context of a plant, the expression of monoterpene alcohol synthase genes often results in the synthesis of an array of additional volatile and nonvolatile derivatives resulting from conversions by endogenous enzymes. Those compounds may also, or exclusively, be accountable for the observed effects on plant defence. Linalool synthase overexpressing plants have been found, for example, to emit, apart from linalool, also linalool oxide, hydroxylinalool or hydroxy-dihydrolinalool (Lewinsohn et al., 2001; Lavy et al., 2002; Aharoni et al., 2003; Aharoni et al., 2006). By contrast, transgenic petunia expressing linalool synthase did not produce any volatile linalool or derivative, but instead efficiently converted linalool to a nonvolatile glycoside (Lücker et al., 2001). The glycosides of linalool and hydroxy–linalool were also reported in Arabidopsis and potato overexpressing linalool synthase (Aharoni et al., 2003; Aharoni et al., 2006). Transgenic plants overexpressing a similar linear monoterpene alcohol geraniol by means of a geraniol synthase produced both volatile and nonvolatile derivatives of geraniol such as geranial (volatile), geranic acid (volatile), geranyl acetate (volatile) in tomato fruits (Davidovich-Rikanati et al., 2007), and geranyl malonyl glucopyranoside and other derivatives (nonvolatile) in maize leaves (Yang et al., 2011). In some cases, these derivatives were not found in controls (wild-type or empty vector controls), but often they are also found there and therefore can be taken to fulfil putative biological roles for the plant species involved. The emerging picture is that terpene derivatives are species and genotype specific and dependent on the developmental and tissue- or cell-type specific variations in endogenous modifying enzymes but that their biological roles other than for storage of volatiles have remained elusive.

Western flower thrips, Frankliniella occidentalis, is a highly polyphagous insect and has become one of the most serious pests in several vegetable and flower crops worldwide (Reitz, 2009). It feeds by using its mouthparts to pierce plant cells and suck out their contents. Damaged plant cells collapse or fill with air, resulting in stunted plant growth, flower and fruit deformation, or ‘silver’ reflective patches and flecking on expanded leaves (Tommasini and Maini, 1995). It causes serious damage to a variety of vegetable and flower crops, including chrysanthemum, and transmits tomato spotted wilt virus (van Driesche et al., 1999). Plant resistance may be improved by overexpressing linalool, as shown for aphids and moths (Aharoni et al., 2003; Yang, 2008). To study this for WFT, the linalool synthase gene FaNES1 was introduced into chrysanthemum plants. Constitutively expressed volatile and nonvolatile metabolites were analysed by GC- and LC-MS, and the effects of the separate volatile and nonvolatile components on WFT behaviour were analysed.

Results

Expression of the linalool synthase gene in chrysanthemum

Chrysanthemum genotype 1581 was transformed with a linalool/nerolidol synthase gene from strawberry (Aharoni et al., 2004), FaNES1, under the control of the Rubisco small subunit promoter (Outchkourov et al., 2003). The FaNES1 protein was targeted to the plastids by fusion with a plastid-targeting signal. Wild-type plants were used as control. Transcript levels of FaNES1 in cuttings of two T0 transgenic lines, line 28 and 37, were determined by quantitative RT-PCR and found to be similar, ranging from 963 to 1107, relative to the household gene actin (Figure 1a). FaNES1 chrysanthemum plants were slightly shorter and lighter in leaf colour compared with wild-type plants (Figure 1c). This kind of phenotype was observed before for FaNES1 Arabidopsis and potato plants (Aharoni et al., 2003; Aharoni et al., 2006) and it could due to insufficient availability of isoprenoid precursors for other essential metabolites such as carotenoids, chlorophylls and gibberellins (Aharoni et al., 2003).

Figure 1.

Expression level of FaNES1 and its effect on linalool emission and phenotype of chrysanthemum plants. (a) expression level of FaNES1 relative to the expression level of actin gene in wild-type or transgenic chrysanthemum plants. Expression levels of the reference gene, actin gene, were set to 1. Transgene expression levels were determined by quantitative RT-PCR. Wt, wild type; T, transgenic. Error bars indicate SE from 3 biological replicates. (b) linalool emission of cut leaves of wild-type and transgenic chrysanthemum plants. Error bars indicate SE from 3 biological replicates. (c) phenotype of wild-type chrysanthemum plant (Wt-1) and transgenic chrysanthemum plants T28-1 and T37-1.

GC-MS analysis of volatile compounds

Volatiles were collected from the headspace of cut leaves at half height of the plants. Linalool, the primary product of FaNES1, was strongly dominant among the detected volatiles (Figure 2a–c). The linalool emission was quantified to be 1.41 to 1.82 μg/h/g-FW in FaNES1 plants, but not found in wild-type plants (Figure 1b). Besides linalool, an acyclic C11-homoterpene, (3E)-4,8-dimethyl-1,3,7-nonatriene (DMNT), was also found to be emitted from FaNES1 plants and not from wild-type plants (Figure 2d,e). The peak area (total ion current) of DMNT was 30- to 70-fold smaller than that of linalool in transgenic plants.

Figure 2.

GC-MS chromatograms obtained by dynamic headspace trapping of leaves of wild-type chrysanthemum plant Wt-1 (a) and transgenic chrysanthemum plant T37-3 (b). (c), GC-MS chromatogram of an authentic standard of racemic linalool. The product of FaNES1 was determined as (S)-linalool by Aharoni et al. (2003). (d), zoom-in chromatogram of Wt-1 from 8.2 to 8.7 min. (e), zoom-in chromatogram of T37-3 from 8.2 to 8.7 min.

LC-MS analysis of nonvolatile compounds

As linalool could also be stored in the form of glycosides, we analysed the nonvolatile metabolites in transgenic (n = 6) and control (n = 3) plants. Aqueous methanol extracts from young leaves were prepared and analysed by accurate mass LC-MS in negative mode (Figure 3a,b). To reveal differential compounds, the LC-MS profiles of transgenic plants and control plants were compared in an untargeted manner using MetAlign followed by multivariate mass spectra reconstruction (MMSR) clustering of extracted signals, as described in 'GC-MS analysis of volatile compounds'.

Figure 3.

Negative mode LC–QTOF–MS chromatograms of aqueous methanol extract of leaves of a wild-type chrysanthemum plant Wt-1 (a) and transgenic chrysanthemum plant T37-1 (b). The extracts were freeze-dried and redissolved in MQ water to make plant nonvolatile contents used for WFT dual-choice assays between nonvolatile contents from wilt-type (c) and transgenic (d) plants. The top 4 significantly different peaks are indicated with boxes. They are putatively identified as 2 linalool–malonyl–glucose 47.79 and 46.07 min), a linalool–pentose–hexose (37.46 min) and a conjugation of hydroxy–linalool–hexose (36.15 min) based on LC-MS-MS analysis (Table 1, Figure S1).

In total, 8968 mass signals were extracted, which grouped into 301 clusters of different metabolites. Among all 8968 masses, 2482 masses (i.e. 28%, distributed in 80 clusters) showed at least twofold intensity difference (< 0.05) between transgenic and control plants. More masses were found to be significantly increased (2312 masses, distributed in 74 clusters) than decreased (170 masses, distributed in six clusters) in the transgenic plants. Differential masses with a signal intensity higher than 5000 (i.e. about 500-fold higher than noise, 18 masses, distributed in four clusters) were subsequently analysed by LC-MS/MS (Table 1, Figures 3 and S1). These compounds had signals of 22 625–33 651 and were 8–457 times higher expressed than in control plants. According to their MS/MS spectra, all these compounds were putatively identified as derivatives of linalool and hydroxy-linalool: two different types of linalool–malonyl–glucose, a linalool–pentose–glucose and a glycoside of hydroxy–linalool.

Table 1. Nonvolatile metabolites significantly increased in FaNES1-expressing chrysanthemum plants putatively identified by LC-MS-MS
Ret (min)Av intensity (Wt)Av intensity (T)Ratio (T/Wt)Accurate mass foundMol form∆mass (ppm)MS-MS fragmentsPutative IDMM
  1. Significantly different metabolites between transgenic and wild-type plants with mass intensities higher than 500 were selected for analysis by LC-MS/MS.

  2. Ret (min), retention time, in minutes; Av, average; Wt, wild-type control plants; T, transgenic plants; Ratio (T/Wt), ratio of mass signal between transgenic plants (T) and wild-type plants (Wt); Mol form, molecular formula of the metabolite; ∆mass (ppm), deviation between the found accurate mass and real accurate mass, in ppm; Putative ID, putative identification of metabolite; MM, monoisotopic molecular mass of the metabolite.

  3. a

    The scheme of collision-induced MS/MS-fragmentation is shown in Figure S1; ()FA, formic acid adduct.

  4. b

    Hydroxy–linalool–hexose conjugate to a molecule with formula of C7H14O3, such as hydroxyheptanoic acid, etc.

47.795424 579457.7803.3721C19H30O92.5401, 357, 315, 161Linalool–malonyl–glucosea ([2M-H])401.1812
46.07267830 16111.3803.3718C19H30O92.1401, 357, 315, 161Linalool–malonyl–glucosea ([2M-H])401.1812
37.46391033 6518.6493.2278C22H38O12−1.4447, 315, 233, 191, 161, 149, 131(Linalool–pentose–hexose)a FA493.2285
36.1514522 625155.7505.2647C24H42O110.4459, 415, 399, 331, 289, 161(Hydroxy–linalool–hexose-b)a FA505.2649

Effects of FaNES1 plants on thrips behaviour

Leaves at similar leaf stage were picked from plants of transgenic line 37 or line 39 and used to study the effects of FaNES1 plants on WFT behaviour. The results of repeated dual-choice assays showed that WFT were significantly deterred by FaNES1 plants 20–28 h after WFT release with 65%–70% of WFT settling on wild-type leaf discs (Figure 4a). However, we also noticed that in the first 15 min of WFT release, WFT significantly preferred the transgenic plants, with 66% WFT initially settling on those leaf discs. We hypothesized that the modified fragrance—linalool emission—of the FaNES1 plants caused the initial attraction of WFT, and that nonvolatile compounds related to linalool in methanol extractable fractions caused the deterrence.

Figure 4.

Dual-choice assays of western flower thrips (WFT) on different food or odour sources. (a) dual-choice assays of WFT on wild-type versus transgenic chrysanthemum leaves. The presence of the thrips on either leaf disc was visually recorded 0.25, 1, 2, 4, 20 and 28 h post-thrips release. The x-axis represents 10log-transformed time data. Asterisks indicate significant differences to the control (*:P < 0.05; **: P < 0.01).Error bars indicate SE (= 120 per treatment). (b) dual-choice assays of WFT on plant nonvolatile contents from wild-type or transgenic chrysanthemum leaves. The thrips behaviour was recorded by video. In a period of 40 min, thrips spent significantly more time underneath the wild-type plant nonvolatile contents after 24 h (P = 0.047). Mean ± SE; for transgenic line 37: 0, 4 and 24 h measured with n = 10, 6 and 10, respectively; for wild-type plant: 0, 4 and 24 h measured with n = 20, 18 and 16, respectively. (c) effect of transgenic chrysanthemum plant and linalool on the olfactory response of WFT. Solvent, paraffin oil used to make 10% or 0.1% linalool. Ten microlitre solvent, 10% or 0.1% linalool was applied on filter paper. Asterisks indicate significant differences of the choices between odour sources (n = 60 per treatment, *: P < 0.05). Wt, wild type; T, transgenic. The dashed line indicates 50% level of the y-axis.

Effects of volatile emissions of FaNES1 plants on thrips attraction

To test whether volatile cues determined the initial attraction, we assayed the choice of WFT by placing individual insects on a wire separating two leaf discs and scoring their choice for either leaf when they left the wire. In this way we compared their response to olfactory cues from wild-type and FaNES1 chrysanthemum leaves. This assay demonstrated that WFT preferred the odour from FaNES1 chrysanthemum leaves (Figure 4c). As linalool was the major compound in the volatile profile of FaNES1 chrysanthemum leaves, we also tested whether pure linalool dissolved in paraffin oil was attractive to WFT and found that both 10% and 0.1% linalool in paraffin oil were similarly significantly attractive to WFT (Figure 4c). This was demonstrated by Koschier et al. (2000) as well, but only for a 10% solution using a different experimental set-up (Y-tube olfactometer). GC-MS analysis of headspace-collected volatiles showed that the linalool emission rate of 0.1% linalool in paraffin oil (0.21 ± 0.03 μg/h) was in the range of the linalool emission rate from leaf discs (0.04 ± 0.03 μg/h, calculated from the emission rate of 1.41 ± 1.15 μg/h/g-FW for leaves of line 37 and the leaf disc weight of 30 ± 0.3 mg).

Effects of nonvolatile contents of FaNES1 plants on thrips deterrence

To test whether deterrent effects of transgenic plants were due to a constitutive difference in chemical composition and not due to differentially induced volatiles or nonvolatiles, a screening in a dual-choice setting with plant nonvolatile contents from line 37 vs wild-type leaves that had not been exposed to thrips was carried out. The plant nonvolatile contents were prepared by freeze-drying the methanol extracts used for the LC-MS analysis and then redissolving the residue in MQ water. LC-MS analysis of the water-dissolved plant nonvolatile contents showed that their chemical composition was comparable to that of the original methanol extracts with again the major chemical differences represented by the four linalool glycosides (Figure 3c,d). After 24 h, thrips were spending significantly more time underneath the plant nonvolatile contents from wild-type plants (Figure 4b). The trend towards preference for wild-type plants could already be observed after four hours, but was not yet significant in this replication (P = 0.173).

Discussion

In this study, we aimed to introduce resistance to WFT into chrysanthemum by genetically engineering production of the monoterpene alcohol linalool in the aerial parts of the plant. A linalool/nerolidol synthase, FaNES1, was expressed in the plastids of chrysanthemum plants and resulted in linalool emissions and accumulation of several forms of linalool glycosides. We observed that WFT during the first 15 min significantly preferred these FaNES1 plants, but in the next 24 h gradually changed their preference to the wild type. We were interested to test the hypothesis that volatile emissions from FaNES1 chrysanthemum dominated by linalool were attractive to WFT, and the nonvolatiles dominated by linalool glycosides were deterrent. These two opposing forces of attraction and deterrence could possibly result in the observed gradual change from attraction to deterrence. To prove the basic premise of simultaneous attraction and deterrence, we needed to dismiss the possibility of an induced effect. For example, endogenous defence mechanisms could be induced earlier or more strongly in the transgenic plant by the higher initial damage of thrips resulting from the linalool-induced attraction. Experiments were therefore designed to score thrips behaviour in dual-choice assays, which did not allow for these alternative induced defence hypotheses.

Firstly, we tested the olfactory response of WFT on leaf discs. Olfactory bioassays typically avoid all direct physical and visual contact of the insect with the plant substrate. Our olfactory dual-choice assay fulfilled the first requirement, but did not fully exclude a role for visual cues. This point was relevant as the plants expressing linalool were slightly lighter in colour if you compared them next to each other. However, during the assays, the insects would walk back and forth on the metal wire before deciding where to go. This behaviour is typical of insects trying to observe concentration gradients of volatiles. Besides, a direct way to prove that linalool was the major factor for WFT to choose FaNES1 plants was by taking the pure compound in the same assay at similar concentrations. We observed that applying an amount of linalool (10 μL 0.1%, with emission rate at 0.21 ± 0.03 μg/h) similar to what was emitted by the leaf discs (leaf disc weight 30 mg, with emission rate at 0.04 ± 0.03 μg/h) would induce their choice, thereby confirming that the emission of the single dominant peak of linalool sufficiently explains the attraction of WFT to FaNES1 plants.

Secondly, we performed dual-choice assays of WFT on plant nonvolatile contents from transgenic or wild-type plants not in contact with WFT before. We hypothesized deterrence to be a property of any or all of the linalool glycosides constitutively expressed in FaNES1 plants. Those four linalool glycosides were the only compounds significantly different relative to the control extract by more than a factor 2 and above a signal threshold of 5000. The compounds had signals of 22 625–33 651 and were 8–457 times higher expressed than in control plants. Giving thrips plant nonvolatile contents containing these compounds prepared from leaves of plants that had not been in contact with thrips resulted in a similar deterrence as was observed on the leaf discs after 24 h. Also in this video-monitored assay, only the data at 24 h were significant, suggesting that thrips take some time to avoid these compounds, possibly due to an induced physiological response in thrips itself. ‘Bad taste’ therefore refers to any directly or indirectly sensed nonvolatiles affecting their feeding behaviour to the extent that they switch food source. These results make it likely that the compounds present in these plant nonvolatile contents dominated by the four linalool glycosides are responsible for the deterrence. Direct proof of this relationship will require purification of these compounds and direct feeding choice assays. It will be interesting then to also test thrips performance (e.g. oviposition, WFT growth rate) in nonchoice assays and to investigate the mechanism of action.

In other transgenic plants overexpressing different linalool/nerolidol synthase genes, linalool and nerolidol were reported to be stored as glycosides as well (Aharoni et al., 2003; Aharoni et al., 2006; Lücker et al., 2001). The glycone was determined in linalool synthase expressing petunia as glucose (Lücker et al., 2001). In our previous study of transgenic maize expressing a geraniol synthase, which also produces a monoterpene alcohol, the glycone was determined as malonyl–glucose (Yang et al., 2011). In this study, several glycosides of linalool or hydroxy–linalool were putatively identified by LC-MS/MS. The linalyl–glucopyranoside, reported as the only glycoside of linalool in the linalool synthase expressing petunia, was not detected as the major linalool glycoside in FaNES1 chrysanthemum. Among the major linalool glycosides listed in Table 1, two glycosides showed the same molecular mass and mass spectrum. They could be isomers and they were identified as linalool conjugated to malonyl–glucose by comparing their mass spectra to that of geranyl-6-O-malonyl-β-D-glucopyranoside, which was identified by NMR in geraniol synthase expressing maize. Another linalool glycoside was putatively identified as linalool conjugated to a pentose-glucose. Such a glycoside has been found to be naturally present in raspberry fruit as S-(+)-linalool 3-O-α-L-arabinopyranosyl-(1→6)-β-D-glucopyranoside (Pabst et al., 1991). A glycoside of hydroxy–linalool was also putatively identified in FaNES1 chrysanthemum. The hydroxy–linalool glycosides have also been reported in FaNES1 Arabidopsis and potato; however, the glycone parts were not determined (Aharoni et al., 2003; Aharoni et al., 2006).

Terpene glycosides are regarded as transport and storage forms of terpenes in plant tissues, and they have been recognized to play important roles as precursors of terpene release (Winterhalter et al., 1997). They may be involved in indirect plant resistance against insects by releasing terpene volatiles as signal compounds attracting predators and parasitoids upon attack by herbivores, or they may be directly toxic to the herbivores (Pankoke et al., 2010; Zou and Cates, 1997). As linalool was attractive to WFT, we propose that the major glycosides stored in FaNES1 chrysanthemum may explain the deterrence against WFT. It is interesting to note that a study of the natural distribution of linalool and its glycosides in several linalool-emitting plants showed that linalool glycosides accumulated much more in flowers than in leaves, and that linalool emission was only detected from flowers (Raguso Robert and Pichersky, 1999). Attraction of pollinators by emitted linalool and parallel deterrence of co-attracted herbivores by stored linalool glycosides may therefore represent an intricate tactic of flowers to balance ‘attractive smell’ with ‘poor taste’ to optimize seed yields using the same precursor compound.

Experimental procedures

Plant materials

The linalool/nerolidol synthase gene from strawberry (Aharoni et al., 2004), FaNES1, driven by the rubisco small subunit promoter from chrysanthemum (Outchkourov et al., 2003), was cloned into ImpactVector1.1 (www.impactvector.com) and introduced into chrysanthemum plants (Chrysanthemum morifolium Ramat.) cv. 1581. The N terminus of FaNES1 was fused to the plastidic targeting signal derived from FvNES1 to direct FaNES1 from cytosol to the plastids (Aharoni et al., 2004). Wild-type chrysanthemum plants were used as control. Plants were grown in a greenhouse at 25 ± 2 °C under long day conditions (16-h-light/8-h-dark photoperiod).

Two T0 transgenic plant lines 28 and 37 producing the highest levels of linalool (Figure S2) and a wild-type control line were propagated by cuttings. Three clones of each plant line were used in the following experiments to determine the transcript levels of FaNES1, total volatiles and total nonvolatiles of leaves.

Transcript analysis

The RNA transcript levels of FaNES1 were determined by real-time quantitative RT-PCR analysis as described by Schijlen et al. (2007). The actin gene from chrysanthemum was used as reference gene. The sequence of actin gene was obtained from GenBank accession AB205087. All primers were designed by the Beacon Designer software package (Palo Alto, CA). For FaNES1, forward primer 5′-ATCGTCCTCAGCAGCAATTCTTC-3′ and reverse primer 5′- CAGCCTTCATGTTCCTCTAAGTAGC-3′ with an expected product size of 116 bp were used, and for the actin gene, forward primer 5′- GGATTCTGGTGATGGTGTGAGTC-3′ and reverse primer 5′- GAATCTTCATCAACGCATCAGTCAG-3′ with an expected product size of 119 bp.

Volatile analysis by GC-MS and nonvolatile analysis by LC-MS

The 7th leaf from the top of each plant with 14–15 leaves was harvested for headspace trapping. The volatiles were sampled for half an hour and then analysed by GC-MS as described by Yang et al. (2011). The temperature programme of the gas chromatograph was 60 °C for 2.5 min, rising to 280 °C at 20 °C min−1 and 0.5 min at 280 °C. The mass spectrometer was set to scan from 35 to 300 m/z. The helium flow was constant at 1.0 mL min−1. Ionization potential was set at 70 eV.

For identification, the authentic standard of (R,S)-linalool (Fluka) was run under identical conditions. (3E)-4,8-dimethyl-1,3,7-nonatriene (DMNT) was identified by comparing mass spectra with the Wiley mass spectra library and by calculating the Kovats Index of each peak based on the retention time relative to alkane standards. Linalool emission from transgenic plants was quantified based on calibration curves with the authentic standard.

From the same plants, the 6th leaf from the top of each plant was harvested for nonvolatile analysis. The nonvolatiles were extracted and analysed by LC-MS as described by Yang et al. (2011).

GC-MS and LC-MS data processing

GC-MS data were acquired using Xcalibur 1.4 (Thermo Electron Corporation) and LC-MS data using MassLynx 4.0 (Waters). The data were then processed using MetAlign version 1.0 (www.metAlign.nl) for baseline correction, noise elimination and subsequent spectral data alignment (De Vos et al., 2007). The processing parameters of MetAlign for GC-MS data were set to analyse from scan number 168 to 1929 (corresponding to retention time 4.83 to 13.62 min) with a maximum amplitude of 4 × 107. The parameters for LC-MS data were set to analyse from scan number 84 to 2600 (corresponding to retention time 1.61 to 49.35 min) with a maximum amplitude of 35 000.

To elucidate which mass signals originate from the same metabolite, all detected masses were clustered by an in-house-developed software package based on a MMSR approach (Tikunov et al., 2005). The mass signal intensities (expressed as peak height using MetAlign) obtained from transgenic plants and wild-type plants were compared using the Student's t-test. Masses with a significant (P < 0.05) intensity change of at least twofold were verified manually in the original chromatograms.

To annotate significantly different compounds, accurate masses were manually calculated, taking into account only those scans with the proper intensity ratios of analyte and lock mass [between 0.25 and 2 (Moco et al., 2006)], and elemental formulae generated within 5 ppm deviation from the observed mass. In addition, mass-directed LC-MS/MS experiments were performed on differential compounds. To obtain proper MS/MS spectra, only molecular ions with signal intensities higher than 500 ion counts per scan were selected.

Thrips dual-choice assays with leaf discs

A population of WFT, Frankliniella occidentalis, was mass-reared on flowering chrysanthemum (Chrysanthemum morifolium Ramat.) cv. Sunny Casa in a greenhouse under a photoperiod of L16:D8 at 25 ± 2 °C. In this study, only adult female thrips were used. All bioassays were conducted in a climate room at 20–22 °C with a L16:D8 photoregime as described by Yang et al. (2012). Chrysanthemum flowers do not emit linalool (Manjunatha et al., 1998), and thus, WFT would not be affected in the olfactory choice assays in this study.

Leaf discs from wild-type chrysanthemum plants were used as control discs, and leaf discs from plants of transgenic line 37 or line 39 were used as test discs. Twelve replicates were used in this experiment. The number of WFT on each leaf disc was recorded 0.25, 1, 2, 4, 20 and 28 h after the release of the WFT. At each time point, a two-tailed Wilcoxon signed rank test was used to assess the significance of the differences in the mean number of WFT between test and control. The plants of line 39 emitted slightly lower amount of linalool than plants of lines 37 and 28 (Figure S2).

Thrips dual-choice assay based on olfactory cues

To dissect the component of thrips host choice based on olfactory cues only, a metal wire (0.5 mm diameter, ~2.5 cm long) was placed between two leaf discs (1.6 cm diameter) embedded, abaxial side up, on a 1.5% (w/v) agar bed in a Petri dish (7 cm diameter). One of the leaf discs was from wild-type chrysanthemum plants as control disc and the other leaf disc was from transgenic plants of line 37 as test disc. The metal wire was not in contact with any of the leaf discs, and there was about 0.5 cm distance between the end of the metal wire and the leaf disc. Every time, one ice-anaesthetized thrips was released in the middle of the metal wire. Once the thrips became active, it walked along the metal wire without going off it. After one or a couple of rounds of walking, the thrips would finally leave the wire at either end and walk towards the leaf disc of choice. The number of thrips reaching either leaf disc was recorded. Every pair of leaf discs was assayed with 10 individual thrips, and this experiment was replicated with six pairs of leaf discs. More than 90% WFT made their choices within 2 min in this assay. A two-tailed Wilcoxon signed rank test was used to assess the significance of the differences in the mean number of WFT between test and control.

In the experiment checking the olfactory response of thrips to the linalool standard, WFT were given choices between filter papers (~1.5 cm2) applied with 10 μL paraffin oil or with 10 μL 10% or 0.1% linalool dissolved in paraffin oil,

Thrips dual-choice assay with leaf nonvolatile contents

Female WFT (3 weeks old) were starved overnight. Six wells of 24-well plates with some pollen added were inoculated with 3 WFT per well and sealed with stretched parafilm. On top of the parafilm, two droplets (30 μL) of plant nonvolatile contents were added (Figure S3). Preference of WFT for the plant nonvolatile contents from either wild-type or transgenic plants was determined by placing the plate on top of a light source and video recording the behaviour for 40 min at the start, and after 4 and 24 h post-WFT release. The plant nonvolatile contents were prepared by freeze-drying the methanol extracts used for the LC-MS analysis and then redissolving the residue in MQ water. The chemical concentration in the plant nonvolatile contents is comparable to that in leaves, assuming the water content of leaves is 90%. These plant nonvolatile contents were also analysed by LC-MS to check whether the residue was dissolved well in MQ water.

The set-up for the video tracking was shown in Figure S3. The footage was analysed live, using EthoVision 8.5xt software. Droplets of wild-type and transgenic line 37 were interpreted as ‘zone 1’ and ‘zone 2’, respectively, accounting for left or right preference. WFT were only in focus and detected when walking on the parafilm ceiling of the wells. For data analyses, time spent in either zone was averaged per arena (across all observed WFT). Misinterpreted subject detection by EthoVision (e.g. subject detected was not a WFT, but an artefact like reflection of droplet) was filtered manually afterwards. The subjects (WFT) were regarded as artefacts and filtered out if their mean velocity was outside the range of 0.5 and 5 mm/s, or if they were not observed in the wells for more than 95% of the total recording time. Based on these filters, the numbers of replicates per time interval and treatment varied: for line 37: 0, 4 and 24 h measured with n = 10, 6 and 10, respectively; and for wild-type: 0, 4 and 24 h measured with n = 20, 18 and 16, respectively. Two-tailed Wilcoxon signed rank test (with absolute values of averaged time spent in zone per well) were used, and P values of 0.05 was set as the significant threshold.

Acknowledgements

We thank Prof. Marcel Dicke and Prof. Harro J. Bouwmeester for their critical reading of the manuscript, Ric de Vos for his help in metabolomics data analysis, Bert Schipper for assistance in LC-MS analysis and Yury Tikunov for his help in using MMSR approach. This research was supported by Technology Top Institute Green Genetics of the Netherlands (grant no. 1C001RP) and the Technology Foundation of the Netherlands Organization for Scientific Research (NWO) (grant STW10989, Perspectief Programme ‘Learning from Nature’).

Conflict of interest

The authors have no conflict of interest to declare.

Ancillary