Reverse genetics of floral scent: application of tobacco rattle virus-based gene silencing in Petunia.

Floral fragrance is responsible for attracting pollinators as well as repelling pathogens and pests. As such, it is of immense biological importance. Molecular dissection of the mechanisms underlying scent production would benefit from the use of model plant systems with big floral organs that generate an array of volatiles and that are amenable to methods of forward and reverse genetics. One candidate is petunia (Petunia hybrida), which has emerged as a convenient model system, and both RNAi and overexpression approaches using transgenes have been harnessed for the study of floral volatiles. Virus-induced gene silencing (VIGS) is characterized by a simple inoculation procedure and rapid results relative to transgenesis. Here, we demonstrate the applicability of the tobacco rattle virus-based VIGS system to studies of floral scent. Suppression of the anthocyanin pathway via chalcone synthase silencing was used as a reporter, allowing easy visual identification of anthocyaninless silenced flowers/tissues with no effect on the level of volatile emissions. Use of tobacco rattle virus constructs containing target genes involved in phenylpropanoid volatile production, fused to the chalcone synthase reporter, allowed simple identification of flowers with suppressed activity of the target genes. The applicability of VIGS was exemplified with genes encoding S-adenosyl-l-methionine:benzoic acid/salicylic acid carboxyl methyltransferase, phenylacetaldehyde synthase, and the myb transcription factor ODORANT1. Because this high-throughput reverse-genetics approach was applicable to both structural and regulatory genes responsible for volatile production, it is expected to be highly instrumental for large-scale scanning and functional characterization of novel scent genes.

Successful pollination is a key factor in the evolutionary success of plants. Survival of many species depends on the plant's ability to attract pollinators to its flowers at anthesis. Flower architecture, color, and scent are major strategies harnessed by plants for interaction with pollinators (Pichersky and Gershenzon, 2002;Raguso and Willis, 2002). Floral scent is a mixture of various volatile molecules, the final scent being determined by the combination of compositions and levels of each compound . This is a highly dynamic trait that exhibits strong dependence on genetic background (Iijima et al., 2004), developmental status (Dudareva et al., 2000), and physiological condition (Underwood et al., 2005), as well as time of day (Helsper et al., 1998;Kolosova et al., 2001;Simkin et al., 2004), light (Hendel-Rahmanim et al., 2007), and other environmental factors (Verdonk et al., 2003;van Schie et al., 2006). The dependence of scent on various endogenous and external signals greatly complicates studies of the factors involved in the production/emission of volatiles and this is one of the main reasons that, for many years, research into flower scent focused on its chemical rather than biological aspects.
Only a handful of metabolic pathways are responsible for generation of the hundreds of volatile compounds that determine floral scent (Croteau and Karp, 1991). The best characterized of these is probably the terpenoid pathway, which leads to the production of hundreds of different mono-and sesquiterpene volatiles. Volatile fatty acid derivatives are generated through the concerted action of lipoxygenases, hydroperoxide lyases, isomerases, and dehydrogenases. Volatile benzenoids and phenylpropanoids are produced by the general phenylpropanoid pathway, responsible for the synthesis of several essential compounds such as lignins, polyphenols, and flavonoids, to name a few . Whereas the flavonoid branch leading to pigment production is extremely well detailed (Koes et al., 2005), numerous steps in volatile phenylpropanoid formation have yet to be deciphered. Nevertheless, some hydroxylation, acetylation, and methylation reactions involved in its production have been characterized . Enzymes catalyzing the latter reaction have been characterized in several plant systems: These include the S-adenosyl-L-Met (SAM)-dependent methyltransferases that methylate nitrogen, sulfur, oxygen, or carbon by transferring the methyl group of SAM to their substrate (Effmert et al., 2005). In petunia (Petunia hybrida), for example, two genes coding for SAM:benzoic acid/salicylic acid carboxyl methyltransferases (BSMT1 and BSMT2), with 99% identity of their amino acid sequences, were isolated. In vitro studies established the ability of this enzyme to methylate not only benzoic acid, but also salicylic acid, yielding methylbenzoate (MeBA) and methylsalicylate (MeSA), respectively-hence, the name of the enzyme (Negre et al., 2003). With respect to the mechanisms regulating floral scent production/ emission, very little is known. Only one transcription factor, ODORANT1 (ODO1), belonging to the myb family, has been isolated: Suppression of its expression in petunia led to a decrease in the emission of benzenoids (Verdonk et al., 2005).
Whereas Arabidopsis (Arabidopsis thaliana) has proven to be a highly useful system for studies of various traits, floral scent analyses would benefit from the use of alternative plant systems, those possessing large flowers that produce a wealth of volatiles belonging to diverse biochemical pathways/branches (Vainstein et al., 2001). Furthermore, in recent years, a large body of sequence information has been accumulated from diverse groups of plants, not only the classical models (Aharoni and O'Connell, 2002;Guterman et al., 2002;Mueller et al., 2005). The virus-induced gene silencing (VIGS) approach enables the utilization of this sequence information for the study of traits of interest in nonmodel plants (Kumagai et al., 1995;Burch-Smith et al., 2004;Robertson, 2004). Using VIGS, genes involved in flower and plant development, pest and virus resistance, and senescence have been characterized in tomato (Lycopersicon esculentum), Nicotiana benthamiana, Arabidopsis, opium poppy (Papaver somniferum), wheat (Triticum aestivum), and Aquilegia, among others (Liu et al., 2002a(Liu et al., , 2002bDinesh-Kumar et al., 2003;Robertson, 2004;Ryu et al., 2004;Chen et al., 2005;Hileman et al., 2005;Scofield et al., 2005;Burch-Smith et al., 2006;Fu et al., 2006;Lou and Baldwin, 2006;Gould and Kramer, 2007). The VIGS mechanism is based on posttranscriptional gene silencing, which targets viral RNA in a sequence-specific manner (Baulcombe, 2004). Passenger sequences within the virus genome, identical to plant sequences, will also be targets for posttranscriptional gene silencing. Therefore, plant DNA fragments introduced into the viral genome induce silencing of the corresponding genes in the infected plant cells (Ruiz et al., 1998;Baulcombe, 1999;Ratcliff et al., 1999;Robertson, 2004).
The tobacco rattle virus (TRV)-based vector is commonly used for VIGS in a variety of species, including petunia, due to its strong silencing efficiency, its ability to silence genes in different organs, including meristems, and the mild symptoms caused by infection (Ratcliff et al., 2001;Liu et al., 2002aLiu et al., , 2002bDinesh-Kumar et al., 2003;Ryu et al., 2004;Chen et al., 2005;Hileman et al., 2005;Burch-Smith et al., 2006;Lou and Baldwin, 2006;Gould and Kramer, 2007). TRV belongs to the genus Tobravirus and has a bipartite, positive-sense RNA genome. cDNA clones of both RNA1 and RNA2, under the control of two 35S promoters, were cloned into T-DNA vectors (pTRV1 and 2): RNA1 of the TRV vector encodes replication proteins, movement protein, and a 16-kD Cys-rich protein; RNA2 of the TRV vector encodes the coat protein and contains multiple cloning sites for insertion of foreign sequences. Both pTRV1 and pTRV2 are required for VIGS of target sequences (Liu et al., 2002b).
The field of floral scent would certainly benefit from an experimental system that combines forward-and reverse-genetics approaches. The lack of studies applying viral vectors to scent research is probably due to the fact that the character is invisible, dynamic, and consists of numerous low-molecular-weight compounds that are detrimentally affected by background noise from nonsilenced cells/tissues. Here, we present the applicability of the TRV-based VIGS system for studying floral scent in petunia. We used anthocyanin pathway suppression as a reporter allowing easily visible identification of silenced flowers/tissues . Because this high-throughput reversegenetics approach was found applicable to both structural and regulatory genes responsible for volatile production, it is expected to be highly instrumental for large-scale delineation of floral scent.

Screening of Petunia Lines for Both Scent and Sensitivity to TRV Infection/Silencing
Studies of volatile production in flowers using VIGS require a suitable model plant that possesses a wide infrastructure for the generation of volatiles, as well as high sensitivity to viral infection. For this purpose, rooted 3-to 7-week-old plantlets of numerous petunia lines were infected with Agrobacterium tumefaciens cultures containing pTRV1 and pTRV2 DNA constructs via various inoculation methods, including leaf infiltration, stem injection, and soaking cut apical meristems. To assess the systemic spread of TRV, RNA prepared from corollas of plants approximately 1 month after infection was used for reverse transcription (RT)-PCR analyses of TRV2 transcript. Meristem infection yielded the most reproducible results and corollas of plantlets inoculated with pTRV vectors by soaking, as shown in Figure 1A for petunia line B1, contained TRV2 transcript, which was absent from control mock-inoculated plants. Flowers of lines positive for systemic spread of the virus (W115, B1, P720; Fig. 1, B-D, respectively) were also screened for volatile production using dynamic headspace analyses (Fig.  1E). Volatile compounds emitted by pigmented flowers of petunia lines B1 and P720 were similar to those emitted by the white flowers of line W115, employed here as a reference because this line is commonly used in studies of floral scent. The level of most volatiles emitted by line B1 flowers was approximately 2 times higher than those emitted by line W115. Petunia lines B1 and P720, with pigmented flowers that emitted diverse volatile phenylpropanoid compounds and were sensitive to viral spread, were chosen for gene-silencing analyses. Suppression of endogenous PHYTOENE DESATURASE (PDS) expression in petunia leaves was used as a marker because its silencing has been shown to cause an easily recognizable photobleached phenotype due to inhibition of carotenoid biosynthesis (Kumagai et al., 1995). Leaves of petunia plantlets infected with a mixture of Agrobacterium transformed with pTRV1 and pTRV2 carrying the 369-bp PDS fragment (pTRV2-PDS) showed a strongly bleached phenotype throughout the plant for at least 3 months, first appearing approximately 2 to 3 weeks after inoculation (Fig. 2). The silenced phenotype was detected in over 85% of the pTRV2-PDS-inoculated plantlets and in none of the control mock-inoculated plants or those inoculated with a mixture of Agrobacterium transformed with pTRV1 and pTRV2. Note that the silenced phenotype could also be obtained when a tandem repeat of a 21-bp-long PDS fragment was used in pTRV2 instead of the 369-bp PDS fragment (Fig. 2F); however, in this case, the bleached phenotype, as compared to that generated by pTRV2 carrying the 369-bp PDS fragment, was not as long lasting and disappeared after approximately 1 month. Inoculation of petunia plantlets with pTRV2 carrying the silencing suppressor p19 (pTRV2-p19), together with pTRV2-PDS, prevented the bleaching that had resulted from PDS suppression (Fig. 2C). To confirm PDS silencing, semiquantitative RT-PCR analysis was performed using primers to the regions not used in the pTRV2-PDS construct. PDS expression was strongly suppressed in plants infected with pTRV2-PDS as compared to control plants infected with pTRV2 ( Fig. 2G).
To allow visual examination of silencing in floral tissues with no adverse effect on scent production, we evaluated the suitability of using CHALCONE SYN-THASE (CHS), a key anthocyanin gene, as a marker, as described in a previous report . Petunia lines with pigmented flowers, B1 and P720, were used for suppression of CHS expression using pTRV2-CHS. As expected, suppression of CHS led to a clearly identifiable white-flowering phenotype (Fig. 3, A and C). An over 95% decrease in the level of accumulated anthocyanins was revealed in suppressed petunia flowers from lines B1 and P720 (Fig. 3, B and D). Petals of some of these flowers were white, with a few scattered patches of pigmented tissue, whereas others were completely white. RT-PCR analyses confirmed the suppression of CHS expression in the white tissues (Fig. 3E).
To further detail the effects of CHS suppression on the level of downstream flavonoids, we used ultraperformance liquid chromatography (UPLC) coupled to either photodiode array (PDA) or mass spectrometry (MS) detectors. The levels of 10 different derivatives of quercetin and kaempferol were detected and their levels were dramatically reduced in the pTRV2-CHSinfected plants (Fig. 3F, a and b). Online UV and Total RNA was prepared from corollas of petunia approximately 1 month after infection with Agrobacterium carrying pTRV1 and pTRV2 (TRV2) and mock-inoculated plants (Mock). TRV2 plasmid was used as a control (pTRV2). B to D, Flowers of petunia lines W115, B1, and P720, respectively. E, Levels of volatiles emitted by flowers of petunia lines W115, B1, and P720. Dynamic headspace analysis was conducted for 12 h with flowers collected 2 DPA. Each graph represents the average of three independent experiments with SEs indicated by vertical lines. tandem MS (MS/MS) spectra were acquired for the most abundant differential flavonoids. All UV spectra (Fig. 3F, e) showed the characteristic absorbance bands of the compounds from the flavonol group, with two maximum absorption bands (lmax) at 240 to 280 and 300 to 380 nm. The absorbance maxima of prior absorption bands enabled assigning the most abundant flavonoid peaks to quercetin (lmax about 255 nm) or kaempferol (lmax 265 nm) derivatives. MS/MS fragmentation of the molecular ions revealed fragments corresponding to quercetin (mass-to-charge ratio [m/z] 5 303.05 D) or kaempferol (m/z 5 287.05 D) aglycones. Acid hydrolysis of the control petal extracts followed by chromatography further confirmed their identification (data not shown). Because phenylpropanoid compounds, including quercetin and kaempferol, are widely used by plants as part of their antimicrobial and antiherbivore defense arsenal (Simmonds, 2003;Gomez-Vasquez et al., 2004), we also performed comparative analyses of pTRV2-CHS-and control pTRV2infected plants' attractiveness to the whitefly Bemisia tabaci. About 3 times more B. tabaci adults landed and fed on the pTRV2-CHS plants ( x 2 5 13.1; degrees of freedom 5 1; P , 0.0003) when released at the center of the cage and allowed to choose between CHS-suppressed and control petunia plants. These VIGS-based choice experiments further support the involvement of flavonoids in plant-insect interactions.
Analyses of floral scent compounds in petunia B1 and P720 plants with suppressed CHS revealed that the levels of the emitted volatiles, with the exception of MeSA in line B1, were essentially unaffected by TRV infection (Fig. 3, G and H). The level of MeSA emitted by flowers of pTRV2-CHS-inoculated B1 plants was approximately 2-fold higher than that emitted by control mock-inoculated plants. Because MeSA levels in flowers of pTRV2-CHS-infected plants were similar to those in plants infected with pTRV2, we suggest that scent production was not affected by CHS suppression in line B1 either; rather, increased emission of MeSA was probably due to the plant's response (as shown before in Deng et al., 2004) to viral infection.

Application of TRV VIGS to the Study of Floral Scent
To directly assess the applicability of the TRV-based VIGS system to the study of floral scent, we generated pTRV2 constructs consisting of CHS fused to previously characterized floral scent genes. Easily identifiable flowers with strongly reduced pigmentation were targets for analyses of floral volatiles. Inoculation of petunia plantlets with pTRV2 carrying a fragment of BSMT upstream of CHS (pTRV-CHS:BSMT) led to suppression of both CHS and BSMT expression, as revealed by RT-PCR analyses (Fig. 4B). These analyses were performed using primers complementary to the region of BSMT that is identical in both BSMT1 and BSMT2. The gas chromatography (GC)-MS analysis of white flowers with suppressed BSMT expression revealed an approximately 7-fold reduction in MeBA . Line P720 plantlets coinfected with Agrobacterium transformed with pTRV1, pTRV2-PDS, and pTRV2 carrying the silencing suppressor p19 approximately 1 month after infection (C). Line P720 inoculated with Agrobacterium transformed with pTRV1 and pTRV2 carrying tandem repeats of a 21-bp-long PDS fragment approximately 1 month after infection (F). Semiquantitative RT-PCR analysis was used to determine PDS silencing in leaves of line B1 plants infected with pTRV2-PDS (G). For each sample, four amplification products (following 21, 24, 27, and 30 cycles of PCR) are shown. Ethidium-bromide-stained agarose gels show RT-PCR products. UBIQUITIN product (UBI) was used as a reference. and an approximately 10-fold reduction in MeSA levels as compared to control pTRV-CHS-inoculated flowers (Fig. 4A). The reduction in MeSA levels was clear despite activation of its production as a result of pathogen infection (Fig. 3G). Emission levels of other volatiles remained unaffected by BSMT suppression.
The effect of PHENYLACETALDEHYDE SYNTH-ASE (PAAS) suppression on floral scent production was also analyzed. Petunia plantlets inoculated with pTRV2 carrying CHS fused to a PAAS fragment (pTRV-CHS:PAAS) generated, as expected, white flowers with strongly reduced levels of phenylacetaldehyde (Fig.  4C). Levels of phenylethyl alcohol were also strongly reduced in these flowers, which is not surprising because phenylacetaldehyde serves as a substrate for its production. RT-PCR analyses confirmed suppression of PAAS expression in white flowers of pTRV-CHS:PAAS-inoculated plants (Fig. 4D).
To test whether VIGS might also be applicable for reverse-genetics analyses of regulators of floral scent production/emission, pTRV2 carrying CHS fused to a fragment of the myb-like transcription factor ODO1 (pTRV2-CHS:ODO1) was used to inoculate petunia plantlets. Transgenic petunia with silenced ODO1 has been shown to produce strongly reduced levels of several volatile phenylpropanoid compounds (Verdonk et al., 2005). Volatile levels generated by flowers of pTRV2-CHS:ODO1-infected plants were also strongly reduced compared to control pTRV2-CHS-infected plants: With the exception of benzylbenzoate, the level of all other volatiles was reduced by over 50% (Fig. 4E). Note that similar to our pTRV2-CHS:ODO1infected plants (Fig. 4E), the level of benzylbenzoate was not affected by suppression of ODO1 in transgenic flowers (Verdonk et al., 2005).

DISCUSSION
Floral fragrance is highly important, both biologically and commercially. However, the complexity and dynamic nature of this trait has hindered its molecular dissection; consequently, genes responsible for the synthesis of ,5% of reported volatile compounds have been isolated and characterized (Vainstein et al., 2001;Dudareva et al., 2006). Construction of cDNA libraries from tissues rich in scent compounds, such as petals and fruits, and large-scale expression-profiling analysis of genes active in these tissues has yielded a growing reservoir of EST collections with potential relevance to floral scent production (Guterman et al., 2002;Boatright et al., 2004;Mueller et al., 2005). Thus, the establishment of a sequence-specific reverse-genetics system that would allow high-throughput functional analysis of target sequences is called for and several characteristics make petunia a well-suited model system for this purpose, including large flowers exhibiting a rich spectrum of scent compounds and amenability to biochemical and molecular analyses, including stable genetic transformation (Gerats and Vandenbussche, 2005). Indeed, petunia is emerging as a convenient model system for floral scent studies and both RNAi and overexpression approaches using transgenes have been harnessed for the study of volatile phenylpropanoid/benzenoid production in petunia (Underwood et al., 2005;Kaminaga et al., 2006;Orlova et al., 2006;Dexter et al., 2007). Consequently, several petunia genes, which were initially selected based on their annotation, have been functionally characterized either in planta or in vitro and assigned a role in floral scent production. For example, PAAS, isolated from petunia flowers following screening of a petal EST database, is a bifunctional enzyme shown to catalyze the formation of phenylacetaldehyde by efficient coupling of Phe decarboxylation to oxidation in a single step (Kaminaga et al., 2006). Its functionality in phenylacetaldehyde formation was ultimately validated via RNAi suppression of its expression, which resulted in decreased phenylacetaldehyde emission from flowers of silenced plants (Kaminaga et al., 2006). In contrast to this one-enzyme-driven synthesis of phenylacetaldehyde from Phe, many volatile phenylpropanoids/benzenoids, such as MeSA, MeBA, benzylbenzoate, phenylethyl benzoate, and isoeugenol, result from several consecutive steps within the phenylpropanoid pathway. Whereas numerous biochemical reactions within the pathway are still unknown, several genes catalyzing the last steps of these reactions, such as BSMT, benzoyl-CoA:benzyl alcohol/ phenylethanol benzoyltransferase, and isoeugenol synthase (PhIGS1), and one transcriptional regulator, ODORANT1, have been characterized using transgenic petunia plants (Negre et al., 2003;Boatright et al., 2004;Verdonk et al., 2005;Dexter et al., 2007).
Despite the feasibility of using genetic transformation tools in petunia for scent research, this still constitutes a laborious and time-consuming procedure and is therefore not optimal for the efficient large-scale functional analyses required to delineate the extended pool of as-yet uncharacterized sequences of interest. VIGS offers an alternative approach for targeted functional analysis, characterized by its simple inoculation procedure and rapid results (approximately 1 month from inoculation). Among the viruses used to elicit VIGS, the TRV-based constructs have been most extensively used, producing consistent results for studies of various processes in many plant species, including petunia (Ratcliff et al., 2001;Liu et al., 2002aLiu et al., , 2002bDinesh-Kumar et al., 2003;Chen et al., 2004;Ryu et al., 2004). Here, we present the applicability of the TRV-based VIGS system to floral scent research. Whereas PDS suppression is commonly used as a silencing indicator in vegetative tissues (Robertson, 2004), we employed CHS as a marker gene because it provides a visual indication of silencing in flowers. Correspondingly, we used petunia lines P720 and B1 because their flowers accumulate anthocyanins in contrast to the white-flowering W115 petunia that is commonly used in scent research. Efficient CHS silencing in flowers of pTRV2-CHS-inoculated plants was visually evidenced by the appearance of white flowers or flowers with large white sectors. A strong reduction of CHS transcript level in the white tissues relative to either mock-inoculated or pTRV2-infected controls validated VIGS of CHS expression. CHS silencing was further supported by quantitative analyses of anthocyanin and flavonoid content, showing reduced levels in silenced pTRV2-CHS flowers. CHS silencing did not modify the qualitative or quantitative profile of volatiles emitted from flowers, with the exception of MeSA, for which increased levels emitted by TRV-infected plants relative to mock-inoculated ones were expected. This volatile compound has been shown to be emitted in response to various other biotic and abiotic stresses (Chen et al., 2003;Deng et al., 2004;Dudareva et al., 2006) and to function as an airborne signal activating disease resistance in neighboring plants and in the healthy tissues of plants following tobacco mosaic virus inoculation (Shulaev et al., 1997).
The possibility of inducing simultaneous silencing of both the CHS marker and a target gene (exemplified here with BSMT, PAAS, and ODORANT1), as evidenced by analyzed levels of the respective transcripts in corollas, further demonstrates the compatibility of this system for the study of floral-specific traits. The reliability of this system for scent studies is apparent from the consistency of the VIGS phenotypes relative to those previously described following RNAi silencing of the respective genes: PAAS silencing led to the expected (Kaminaga et al., 2006) strong reduction of phenylacetaldehyde and phenylethyl alcohol levels, and silencing of ODORANT1 to the expected (Verdonk et al., 2005) reduced levels of several benzenoid compounds emitted by flowers. Interestingly, analyses of B1 flowers with silenced BSMT revealed that, in planta, BSMT is involved in the production of not only MeBA (Negre et al., 2003;Underwood et al., 2005), but also MeSA. This promiscuous habit toward various substrates is typical of enzymes catalyzing scent production (Guterman et al., 2006;Pichersky et al., 2006) and is considered a biological mechanism that allows the formation of an enormous diversity of volatile compounds produced by relatively small numbers of genes throughout the plant kingdom (Schwab, 2003). It is therefore apparent that a wide perspective on the in vivo function of scent-related genes should be established, preferably based on their functional characterization in several plant systems. The already established protocols enabling the application of VIGS in an array of plants and the ease and short time needed for this approach has the potential to greatly advance the field of scent research allowing, for example, even high-throughput nontargeted functional characterization of floral scent regulators. Recent advances in the generation of virusbased systems for the expression of foreign genes should be highly instrumental in exposing the intricate mechanisms underlying scent production/emission.

Plant Material
Rooted plantlets of petunia (Petunia hybrida) lines B1, P720, and W115 used for VIGS infection were obtained from the Danziger-''Dan' ' Flower Farm (Mishmar Hashiva, Israel). Plants were grown in the greenhouse under 25°C day/20°C night temperatures and natural photoperiod. For dynamic headspace analysis, flowers were collected 2 DPA at 5 PM.

Agroinoculation of TRV Vectors
Agrobacterium tumefaciens (strain AGLO) transformed with pTRV1, pTRV2, and pTRV2 derivatives were prepared as described previously (Liu et al., 2002b). The Agrobacterium culture was grown overnight at 28°C in Luria-Bertani medium with 50 mg/L kanamycin and 200 mM acetosyringone. The cells were harvested and resuspended in inoculation buffer containing 10 mM MES, 200 mM acetosyringone, and 10 mM MgCl 2 to an OD 550 of 10. Following an additional 3 h of incubation at 28°C, the bacteria with pTRV1 was mixed with the bacteria containing the pTRV2 derivates in a 1:1 ratio; 200 to 400 mL of this mixture was injected into the stem or applied on the cut surface after removing the apical meristems of petunia plantlets.

Flavonoid Extraction and Analysis
For anthocyanin analyses, corolla limbs were extracted in methanol containing 1% (v/v) HCl (200 mg tissue/10 mL). Absorption values of the extract at A 530 and A 657 were measured to determine anthocyanin content using the formula A 530 2 0.25A 657 , which allows for subtraction of chlorophyll interference. For extraction of flavonoids, 100 mg of frozen corolla powder were extracted with 400 mL of MeOH:water:formic acid (90:10:0.1 [v/v/v]). The mixture was sonicated (20 min), centrifuged, and filtered through a 0.22mm polyvinylidene difluoride filter (Millex-GV; Millipore). For acid hydrolysis, filtered petal extract (200 mL) was mixed with 200 mL of 2.2 N HCl, heated at 90°C (1 h, dry bath), diluted with 200 mL of MeOH, and filtered (0.2-mm polyvinylidene difluoride filter). UPLC-PDA analysis of flavonoids was performed on a Waters UPLC Acquity instrument equipped with an Acquity 2996 PDA detector and a 150 3 2.1-mm i.d., 1.7-mm reverse-phase UPLC BEH C18 column (Waters Acquity). The mobile phase consisted of 0.1% formic acid in acetonitrile:water (5:95 [v/v]; phase A), and 0.1% formic acid in acetonitrile (phase B), and the following gradient program was used: 0% to 28% B in 22 min, 28% to 40% B in 0.5 min, 40% to 100% B in 0.5 min, followed by 100% B for 1.5 min, then returned to the initial conditions (100% A) in 0.5 min, and conditioning at 100% A (1 min); the column was kept at 35°C and the flow rate was 0.3 mL/min. UV spectra were recorded between 200 and 600 nm and the UV trace was measured at 400 nm. Absorbance spectra and retention times were compared with those of quercetin dihydrate and kaempferol standards (Sigma). For liquid chromatography (LC)-MS analyses, we used a UPLCquantitative time-of-flight instrument equipped with an electrospray ionization source (Premier, Waters-Micromass) and LC conditions were the same as described above for UPLC-PDA analysis. LC-MS/MS runs with electrospray ionization-positive mode settings and were as follows (100-1,500 m/z ions): capillary spray at 3.0 kV; cone voltage at 30 eV; collision energy ramp from 5 to 30 eV; argon was used as a collision gas. The MS was calibrated using sodium formate and Leu enkephalin was used as the lock mass. MassLynx software version 4.0 was used to control all instruments and calculate accurate mass.

Collection and GC-MS Analysis of Volatile Compounds
Volatiles emitted from detached petunia flowers were collected using an adsorbent trap consisting of a glass tube containing 200 mg of Porapak Type Q polymer (80/100 mesh; Alltech) held in place with plugs of silanized glass wool (Guterman et al., 2006). Trapped volatiles were eluted using 3 mL hexane and 2 mg isobutylbenzene were added to each sample as an internal standard. GC-MS analysis was performed using a device composed of a Pal autosampler (CTC Analytic), a TRACE GC 2000 equipped with an Rtx-5SIL MS (Restek; i.d. 0.25 mm, 30 m 3 0.25 mm) fused-silica capillary column, and a TRACE DSQ quadrupole mass spectrometer (ThermoFinnigan). Helium was used as the carrier gas at a flow rate of 0.9 mL/min. The injection temperature was set to 250°C (splitless mode), the interface to 280°C, and the ion source adjusted to 200°C. The analysis was performed under the following temperature program: 2 min of isothermal heating at 40°C followed by a 10°C/min oven temperature ramp to 250°C. The transfer line temperature was 280°C. The system was equilibrated for 1 min at 70°C before injection of the next sample. Mass spectra were recorded at 2.65 scan/s with a scanning range of 40 to 450 m/z and electron energy of 70 eV. Compounds were tentatively identified (.95% match) based on the NIST/EPA/NIH Mass Spectral Library (data version: NIST 05; software version 2.0 d) using the XCALIBUR version 1.3 program (ThermoFinnigan) library. Further identification of major compounds was based on comparison of mass spectra and retention times with those of authentic standards (Sigma) analyzed under similar conditions.

Choice Assay
About 200 to 300 adults of the whitefly Bemisia tabaci (Gennadius; Hemiptera: Aleyrodidae) were released at the center of a cage containing pTRV2 and pTRV2-CHS-infected petunia plants. The number of adult insects on each plant was determined after 48 h. This experiment was repeated twice and results were analyzed by a x 2 test (JMP statistical software, version 6.0.0; SAS Institute).