Disruption of plant carotenoid biosynthesis through virus-induced gene silencing affects oviposition behaviour of the butterfly Pieris rapae


Author for correspondence:
Marcel Dicke
Tel: +31 317 484311
Email: marcel.dicke@wur.nl


  • Optical plant characteristics are important cues to plant-feeding insects. In this article, we demonstrate for the first time that silencing the phytoene desaturase (PDS) gene, encoding a key enzyme in plant carotenoid biosynthesis, affects insect oviposition site selection behaviour.
  • Virus-induced gene silencing employing tobacco rattle virus was used to knock down endogenous PDS expression in three plant species (Arabidopsis thaliana, Brassica nigra and Nicotiana benthamiana) by its heterologous gene sequence from Brassica oleracea.
  • We investigated the consequences of the silencing of PDS on oviposition behaviour by Pieris rapae butterflies on Arabidopsis and Brassica plants; first landing of the butterflies on Arabidopsis plants (to eliminate an effect of contact cues); first landing on Arabidopsis plants enclosed in containers (to eliminate an effect of volatiles); and caterpillar growth on Arabidopsis plants. Our results show unambiguously that P. rapae has an innate ability to visually discriminate between green and variegated green-whitish plants. Caterpillar growth was significantly lower on PDS-silenced than on empty vector control plants. This study presents the first analysis of PDS function in the interaction with an herbivorous insect.
  • We conclude that virus-induced gene silencing is a powerful tool for investigating insect–plant interactions in model and nonmodel plants.


A major challenge for current biology is to integrate research approaches that address different levels of biological organization, that is, from subcellular mechanisms to functions in ecological interactions (Zheng & Dicke, 2008). To achieve this, the determination of the function of genes is indispensable. To date, most reverse genetics approaches have relied on post-transcriptional gene silencing to study the function of plant genes (Watson et al., 2005). This is especially suitable for nonmodel species, which lack a rich abundance of well-characterized mutants, but is also a useful tool for model species (e.g. Alonso et al., 2003). Virus-induced gene silencing (VIGS) has been developed recently to characterize the function of plant genes through gene transcript suppression (Burch-Smith et al., 2006; Wang et al., 2006; Pflieger et al., 2008). The most widely used VIGS vectors are based on the tobacco rattle virus (TRV) (Ratcliff et al., 2001; Liu et al., 2002). TRV-based VIGS vectors have been mainly used to silence genes in a number of solanaceous plant species, including Nicotiana benthamiana, N. attenuata, Solanum lycopersicum and S. nigrum (Ratcliff et al., 2001; Liu et al., 2002; Bhattarai et al., 2007; Kandoth et al., 2007; Senthil-Kumar et al., 2007; Hartl et al., 2008; Wu et al., 2008). The cosilencing of LeMPK1 and LeMPK2 has been shown to reduce systemin-mediated resistance to Manduca sexta larvae in tomato (Kandoth et al., 2007). The silencing of methyl jasmonate (MeJA) esterase expression in N. attenuata has shown that the resistance elicited by MeJA treatment is not elicited by MeJA, but by its demethylated analogue jasmonic acid (Wu et al., 2008). The application of VIGS has also been used in reverse genetics of floral scent in petunia (Spitzer et al., 2007) and in the development of symbiotic root nodules in Pisum sativum (Constantin et al., 2008).

In this article, we investigate insect behavioural responses to plant pigmentation which is modified through TRV-based VIGS. Adult insects select plants as food for themselves or as host for their offspring. For example, many butterfly species select flowers for nectar feeding and green leaves for oviposition (Prokopy & Owens, 1983). During host plant selection, butterflies are known to utilize visual, tactile, olfactory and gustatory information (Prokopy & Owens, 1983; Schoonhoven et al., 2005). In many animal species, foraging and sexual behaviour is guided by visual stimuli (Logothetis & Sheinberg, 1996; Briscoe & Chittka, 2001). Investigations of the visual cues that mediate biological interactions usually employ differently coloured nonliving substrates to circumvent the difficulties in the experi-mental control of the plant or animal pigments involved. Oviposition site selection involves an important behavioural decision in the life cycle of an herbivorous insect, because neonate larvae have limited dispersal capacity (Chew & Renwick, 1995).

We employed VIGS to knock down endogenous phytoene desaturase (PDS) gene expression in three plant species (Arabidopsis thaliana, Brassica nigra and Nicotiana benthamiana) by its heterologous gene sequence from Brassica oleracea. Although PDS silencing has been used as a marker of successful gene silencing, studies on the biological function of PDS have been restricted to its role in carotenoid biosynthesis (Giuliano et al., 1993; Fang et al., 2008). In this article, we address the function of PDS in an ecological context. Defects in PDS activity result in green-whitish variegated plants that resemble virus-infected or leafminer-infested plants (Wetzel et al., 1994). The green-whitish variegated phenotype occurs in many plant species (Evenari, 1989; Lev-Yadun, 2006) and appears to influence the oviposition behaviour of insects (Soltau et al., 2009). In this study, we specifically investigated the effect of PDS expression on the interaction with an insect, the small cabbage white butterfly, Pieris rapae, as mediated by changes in colour pattern in the plant’s phenotype. This butterfly species is a specialist on plants in the Brassicaceae (syn. Cruciferae) family, and research on sensory cues relevant for oviposition behaviour has focused on plant chemistry, especially the stimulatory role of the characteristic secondary metabolites of Brassicaceae, i.e. glucosinolates (Van Loon et al., 1992; Renwick & Chew, 1994; de Vos et al., 2008). However, chemosensory information is integrated with visual information, as female P. rapae butterflies can learn to associate colour with the taste of glucosinolates (Traynier & Truscott, 1991). The closely related large cabbage white butterfly, Pieris brassicae, learns to associate a difference in green shades with the presence of glucosinolates on artificial substrates (Smallegange et al., 2006). This study is the first to use living plants in which the pigment pattern has been manipulated by disruption of the function of a single gene, that is, PDS, through VIGS. We hypothesized that plants with a pigmentation pattern deviating from that of wild-type plants would be avoided by the insect in making reproductive decisions. We designed specific experiments to eliminate an effect of potential chemical changes on butterfly behaviour. Furthermore, we tested the performance of the offspring of the butterflies on plants with different pigmentation patterns.

Materials and Methods

Plants and insects

Seeds of Arabidopsis thaliana (L.) Heynh ecotype Col-0 were sown and grown at the Laboratory of Entomology (Wageningen University) in a climate chamber at 60–70% relative humidity, 23 ± 1°C, 8 h : 16 h light : dark photoperiod under a light intensity of c. 45 μmol m−2 s−1 (lamps used: TLD 58W/840HF, type Reflex, Philips, Eindhoven, the Netherlands). Four- to five-leaf seedlings were used for VIGS, c. 21–25 d after sowing. Brassica nigra L. CGN06619 and Nicotiana benthamiana Domin were sown and grown at Wageningen University’s Unifarm facility (50–70% relative humidity, 20–30°C, 16 h : 8 h light : dark photoperiod) for 1 and 4 wk, respectively, before they were transferred to a climate chamber. Four- to five-leaf seedlings of N. benthamiana and seedlings of B. nigra in the cotyledon stage were used for VIGS in a climate chamber at 60–70% relative humidity, 23 ± 1°C, 16 h : 8 h light : dark photoperiod (c. 45 μmol m−2 s−1; lamps used: Philips, TLD 58W/840HF, type Reflex). Arabidopsis and Brassica plants 21 d post-infiltration with TRV were used in bioassays.

Pieris rapae L. (small cabbage white butterfly) were routinely reared at the Laboratory of Entomology (Wageningen University) on cabbage plants (B. oleracea var. gemmifera cv Cyrus) in a climatically controlled room at 21 ± 1°C, 50–70% relative humidity, 16 h light : 8 h dark (natural daylight supplemented with fluorescent strip lights, Philips TLD 58/840HF).

Isolation of B. oleracea full-length phytoene desaturase (BoPDS) cDNA

A fragment of the B. oleracea PDS (BoPDS) gene, 643 bp long, was obtained by reverse transcriptase-polymerase chain reaction (RT-PCR) with B. oleracea var. gemmifera cv Cyrus cDNA as template and the primer pair 5′-ATGGT-TGTGTTTGGGAATGTTTC-3′ and 5′-TCTCTGGCC-ATGTCAGCATCTCG-3′. The primer pair was designed on the basis of the A. thaliana PDS gene accession NM_117498 (At4g14210) and its sequence similarity with the B. napus expressed sequence tag sequence (CD827969). Total RNA isolation, cDNA synthesis and RT-PCR were performed according to the methods described previously (Zheng et al., 2007). The purification, ligation into pGEM-T Easy (Promega Corp., Madison, WI, USA) and bacterial transformation of the PCR products were performed according to the manufacturer’s protocol. Ampicillin-resistant colonies were identified and plasmid DNA was isolated using a miniprep kit (Qiagen, Venlo, the Netherlands) according to the manufacturer’s instructions. Plasmids were selected which showed inserts of the expected size after digestion with EcoRI. Expected inserts based on size were used for sequencing. Sequencing was performed in an ABI PRISM 310 automated DNA sequencer (Perkin Elmer, Foster City, CA, USA).

A full-length BoPDS cDNA was obtained by rapid amplification of cDNA ends (RACE) PCR. Cloning of the 5′ and 3′ ends of BoPDS cDNA was accomplished by the BD SMART RACE cDNA Amplification Kit (BD Bioscience Clontech, Palo Alto, CA, USA) following the manufacturer’s protocol. 3′-RACE-Ready cDNA and 5′-RACE-Ready cDNA synthesis, PCR and touchdown PCR were used as described previously (Zheng et al., 2007). BoPDS gene-specific primers (forward primer, 5′-AACAACGA-GATGCTGACATGGCCAGAG-3′; reverse primer, 5′-TTCTCTGGCCATGTCAGCATCTCGTTG-3′) were designed on the basis of the above-mentioned partial sequence. The 5′- and 3′-RACE products were purified, ligated and sequenced as explained above. The full nucleotide sequence of BoPDS has been deposited in GenBank (EU719210).

Construction of TRV plasmids

The pTRV1 (pYL192) and pTRV2 (pYL156) vectors described previously (Liu et al., 2002) were used in this study. To generate VIGS plasmids, two cDNA fragments were amplified by PCR using the above-mentioned B. oleracea var. gemmifera cv Cyrus cDNA and primers F1 (5′-TCAGAATTCCGATCTCTTCACAAGCGC-3′) and R1 (5′-TCTCTGGCCATGTCAGCATCTCGTTG-3′), and F2 (5′-CAACGAGATGCTGACATGGCCAGAGA-3′) and R2 (5′-GCGCCTTCCATGGAAGCTAAGTACTT-CTG-3′), respectively. The resulting PCR products, 552 and 969 bp in length, were cloned into EcoRI-cut pTRV2 and generated sense VIGS constructs pTRV2-BoPDS1 and pTRV2-BoPDS2, respectively (Fig. 1). The pYL156 recombinant constructs obtained were confirmed by PCR using insert-specific primers (F1 and R1 or F2 and R2) and vector-specific primers 156F (5′-GGTCAAGGTACGTAGT-AGAG-3′) and 156R (5′-CGAGAATGTCAATCTCGTAGG-3′), which span the multiple cloning site in TRV2 (Hileman et al., 2005). Both constructs were then purified and the identity of the final constructs was verified by sequencing with the 156F and 156R primers. Recombinant constructs were electroporated into competent cells of Agrobacterium tumefaciens strain AGL0 as described previously (Zheng et al., 2005) by a microPulser™ (Bio-Rad, Hercules, CA, USA). Colonies were screened by PCR using primers 156F and 156R for the presence of pTRV2-BoPDS1 or pTRV2-BoPDS2. Glycerol stocks were made from single positive transformants.

Figure 1.

 Tobacco rattle virus (TRV)-based virus-induced gene silencing (VIGS) constructs used in this study. (a) Map of the phytoene desaturase (PDS) gene from cabbage (Brassica oleracea) BoPDS. The light blue and dark blue areas are the 552 and 969 bp fragments cloned into TRV2 to generate pTRV2-BoPDS1 and pTRV2-BoPDS2, respectively. The primers (F1 and R1 vs F2 and R2) used for amplification are indicated below. The asterisk represents the stop codon. 3′-UTR represents the 3′-untranslated region. (b, c) TRV-based VIGS vector systems. 2 × 35S, duplicated CaMV 35S promoter; NOSt; nopaline synthase terminator; RdRp, RNA-dependent RNA polymerase; 16K, 16 kDa cysteine-rich protein; MP, movement protein; CP, coat protein; LB and RB, left and right borders of T-DNA; Rz, self-cleaving ribozyme; MCS, multiple cloning sites. (d) Map of the TRV2-BoPDS1 construct used for VIGS. The light blue area is the 552 bp BoPDS1 fragment. (e) Map of the TRV2-BoPDS2 construct used for VIGS. The dark blue area is the 969 bp BoPDS2 fragment.

Preparation of Agrobacterium inoculation

Cultures of A. tumefaciens strain AGL0 containing pTRV1, pTRV2 and pTRV2-BoPDS1 or pTRV2-BoPDS2 plasmid were cultured separately in 20 ml Luria–Bertani medium containing 50 μg ml−1 kanamycin and 50 μg ml−1 rifampicin, and shaken at 28°C overnight. After centrifugation at 2880 g for 15 min, bacterial cell pellets were resuspended in infiltration buffer containing 1 mM Mes (pH 5.8), 10 mM MgCl2 and 100 μM acetosyringone to an optical density at 600 nm (OD600) = 2, and allowed to stand at room temperature for 2 h before infiltration. Suspensions of A. tumefaciens carrying pTRV1 and pTRV2-BoPDS1 or pTRV2-BoPDS2 were mixed 1 : 1 and infiltrated into the underside of the upper leaves of 3–4-wk-old Arabidopsis plants, 4-wk-old N. benthamiana plants or 1-wk-old plants of B. nigra (the latter only carrying cotyledons) using a 2 ml syringe. Suspensions of A. tumefaciens carrying pTRV1 and pTRV2 were also mixed 1 : 1 and infiltrated in the same way as above as an empty vector control (referred to as EV control below).

Generation of PDS-silenced plants in three different plant species

Agroinfiltration with VIGS constructs pTRV2-BoPDS1 and pTRV2-BoPDS2, respectively, resulted in successful photobleaching in all three plant species studied. Either insert fragment BoPDS1 or BoPDS2 was sufficient to induce gene silencing in different species to produce photobleaching pds phenotypes. However, the timing and extent of photobleaching varied among plant species. Nicotiana benthamiana plants developed photobleaching symptoms 5 d after infiltration, whereas A. thaliana and B. napus showed symptoms 7–10 d after infiltration. We further examined whether growing conditions and plant stages for A. thaliana affected the expression of VIGS. We analysed the number of plants showing a pds phenotype after growth under long-day (16 h : 8 h light : dark photoperiod) vs short-day (8 h : 16 h light : dark photoperiod) conditions. Under both long-day and short-day conditions, 80–100% of the plants displayed photobleaching in 10 independent experiments. After testing differently aged seedlings, we found that the silencing of AtPDS was equally effective in seedlings inoculated at the two- to three-leaf stage, four- to five-leaf stage and six- to seven-leaf stage (data not shown). Therefore, four- to five-leaf stage seedlings were used in later oviposition experiments under the same conditions as used for the rearing of A. thaliana plants. After agroinfiltration, B. nigra and N. benthamiana plants were also grown under the same conditions as used for the rearing of the stock plants.

Real-time quantitative PCR

To quantify endogenous PDS transcript levels in PDS-silenced and EV control plants, real-time quantitative PCR was performed. Leaf samples for real-time quantitative PCRs were harvested from individual PDS-silenced and EV control Arabidopsis plants. Total RNA extraction, purification and cDNA synthesis and PCRs were the same as described previously (Zheng et al., 2007). Briefly, a 25 μl PCR was prepared containing 1 μl of template cDNA, 1 μl of forward primer (10 μM), 1 μl of reverse primer (10 μM), 9.5 μl of water and 12.5 μl of 2× iQ™SYBR green Supermix (Bio-Rad). The final amount of cDNA template assayed was equivalent to 10 ng of RNA. All samples were amplified in duplicate assays under the following conditions: 95°C for 3 min for one cycle, followed by 40 cycles of 95°C for 15 s and 56°C for 45 s, with data collection at 56°C. The PCR products for each primer set were also subjected to melt curve analysis. Melt curve analysis ensured that the resulting fluorescence originated from a single PCR product and did not represent primer dimers formed during PCR or nonspecific product. No-template controls, as water and minus RT (10 ng of RNA), were also included to detect any spurious signals arising from the amplification of any DNA contamination or primer dimer formed during the reaction. The gene-specific primers of AtPDS (gene accession NM_202816) and AtGAPDH (gene accession M64116 as reference; Shih et al., 1991) were designed with Beacon Designer software (Premier Biosoft International, Palo Alto, CA, USA) set to an annealing temperature of 56°C. AtPDS primers were F-AtPDS (5′-CAATGACGAT-GGCACGGTTAAGAG-3′) and R-AtPDS (5′-CTGGAGC-GGCAAACACATAAGC-3′). The predicted length was 83 bp. The AtPDS primers corresponded to regions in the plant mRNAs that were absent from the virus vectors. AtGAPDH primers were F-AtGAPDH (5′-AATGAAGGACTGGAGAGGTGGAAG-3′) and R-AtGAPDH (5′-ACGGTTGGGAC-ACGGAAAGAC-3′). The predicted length was 138 bp. These primer sequences were blasted against the National Center for Biotechnology Information nucleotide and expressed sequence tag database to ascertain that they were not homologous to other genes. The quantification of gene expression was performed with a Rotor-gene™ 6000 (Corbett, Sydney, Australia) machine. For the relative quantification of gene transcript abundances in PDS-silenced and EV control plants, a standard curve method was applied according to the manufacturer’s protocol (Corbett, Australia). AtPDS expression relative to AtGAPDH expression was quantified by comparing the threshold cycle for each PCR to their respective dilution series and dividing the resulting quantities.

Butterfly oviposition preference experiments on A. thaliana and B. nigra

Soon after pupation, adult butterflies were transferred into a small mesh cage (25 × 25 × 35 cm) and supplied with 10% sucrose solution in the climate chamber at 60–70% relative humidity, 23 ± 1°C, 16 h : 8 h light : dark photoperiod (c. 45 μmol m−2 s−1; lamps used: Philips TLD 58W/840HF, Reflex) to allow them to mate during 2–4 d. Butterflies that were used in the bioassays had not been exposed to plants before the experiment and were considered naive. For the experiments, one male and female butterfly were introduced per oviposition cage (50 × 50 × 50 cm). In these cages, the butterflies were also supplied with 10% sucrose solution. In the morning, the different PDS-silenced and EV control plants of A. thaliana or B. nigra were introduced into the cages, and the butterflies were allowed to oviposit for 6 or 24 h depending on the experiment. At the end of each experiment, the numbers of eggs on the plants were counted. The experiments were carried out in several cages per day, and this was repeated over at least 3 d per choice assay. The positions of the plants in the cages were randomized to avoid positional bias.

Oviposition preference experiment on artificial substrate

Green and white paper sheets were sprayed with sinigrin (Bruinsma et al., 2007), a glucosinolate that is a known oviposition stimulant for P. rapae (Traynier, 1986), to investigate whether female P. rapae butterflies discriminated between green and white paper. The green and white paper sheets (8 × 12 cm) were pierced and folded onto a T-shaped wire standard. The T-shaped wire (width, 15 cm; height, 19 cm) was positioned on a solid copper socket (Supporting Information Fig. S1). The T-shaped wire was placed in a flow cabinet and 1 ml of a 5 mM sinigrin solution was evenly distributed over the upper side of the paper with a chromatographic sprayer that produces a fine mist (Desaga, Heidelberg, Germany). Sinigrin is not a volatile, but a glucoside that has been amply demonstrated to be detected by gustatory receptors on the butterfly tarsi after landing, rendering paper acceptable for oviposition by P. rapae butterflies (Städler et al., 1995). The two standards with green and white paper sheets were placed next to each other at a distance of 20 cm in the oviposition cage (50 × 50 × 50 cm). In each cage, four female butterflies, each with a colour mark for individual identification on the underside of their hind wings (e.g. black, blue, green and red), were released. The landings by each marked butterfly were recorded for 2 h. After recording the landing frequencies, the butterflies were left for another 4 h, and the eggs deposited during this time interval were counted on both sides of the paper in each experiment. The experiments were carried out in two to three cages per day, and each experiment consisted of a total of nine independent replicates. The position of the paper sheets in the cages was randomized to avoid positional bias.

Butterfly first landing choice on EV green vs PDS-silenced green-whitish plants

The first landing response of P. rapae butterflies was investigated in a two-choice test. This experiment was set up to investigate the response of the butterflies before they could contact the plant, so as to eliminate an effect of contact chemicals. Pairs of 3-wk-old PDS-silenced and EV A. thaliana plants were used to investigate whether P. rapae butterflies discriminated between optically different plants during host plant selection for oviposition on freely accessible plants. The pot size was such that large parts of the rosette leaves were presented against the dark background of the potting soil. Five female butterflies, marked with different colours to allow individual recognition, were released in a cage (50 × 50 × 50 cm) with one pair of distinct test plants. The first choice of each butterfly on a green or green-whitish plant was recorded. After a butterfly had deposited an egg, both butterfly and egg were discarded (a video recording of butterfly choice behaviour is available as Supporting Information Movie S1). Another marked butterfly was subsequently introduced into the cage. After five independent first choices in response to one pair of plants, the plant pair was replaced by a new pair, switching positions between the treated and control plants to prevent potential positional bias. On each day, five pairs of plants were tested, and the experiment was repeated on four separate dates to accumulate data on 100 landing butterflies.

In order to rule out any metabolic changes potentially occurring in PDS-silenced green-whitish plants that might influence butterfly preference behaviour through olfactory cues, additional experiments were carried out with plants enclosed in air-tight sealed transparent boxes. PDS-silenced and EV plants of A. thaliana were individually placed in transparent cylinders (diameter, 12 cm; height, 6 cm). Each cylinder (one with a PDS-silenced plant and one with an EV plant) was covered with a Petri dish (145 mm × 20 mm) which was tightly sealed with Parafilm just before the two cylinders were placed into the cage. The release of the butterfly was performed in the same way as before. The butterfly’s first choice was recorded when it landed either on the Petri dish cover of the container with the PDS-silenced plant or that of the EV plant. The experiment was repeated on four separate days to obtain data on the first landing of 60 butterflies. The positions of the plants in the cages were randomized to avoid positional bias.

Pieris rapae larval performance

Three-week-old PDS-silenced and EV plants of A. thaliana were offered to P. rapae butterflies to oviposit five eggs per plant. If butterflies laid more than five eggs on the plant, the surplus eggs were removed. The eggs all hatched 4 d after oviposition on the same day on both plant types. The fresh body mass of individual larvae from 18 plant pairs was determined 8 d after oviposition using a Sartorius balance (accuracy, 0.1 mg; Göttingen, Germany).

Leaf reflectance measurements

Leaf reflectance was measured using an improved version of the system described by Soares et al. (2008). In our system, the USB-2000 spectroradiometers were replaced by USB-4000 spectroradiometers (Ocean Optics, Dunedin, FL, USA) with custom-enlarged slit width, which were cooled to 5°C in order to increase the signal strength and stability. Transmittance measurements are not shown. Two blue light-emitting diodes (405 and 435 nm peak wavelength) were added to both laboratory-built, stabilized, quartz-halogen measuring light sources to increase the signal in the blue part of the spectrum. Three plants from each type (i.e. EV control, PDS-silenced and wild-type plants) were used for reflectance measurements. Reflectance was measured on at least five leaf discs per plant, cut from visually healthy, young mature leaves. For the PDS-silenced plants, separate samples were taken from the green and from the whitish, chlorophyll-deficient leaf zones (five samples per tissue type per plant). Reflectance was determined per nanometre in the range 400–800 nm.

Statistical analysis

Each individual butterfly female was subjected to a dual-choice situation, in which individual butterflies were allowed to oviposit on PDS-silenced and EV plants. The data were analysed using the Wilcoxon matched-pair signed-ranks test employing SPSS 15.0 for Windows. Butterfly first choices in assays offering EV plants and PDS-silenced plants, both those freely accessible and those enclosed in air-tight transparent containers, were analysed by a binomial test. The performance of caterpillars of P. rapae on EV plants and PDS-silenced plants was analysed by Student’s t-test. The results for the experiments on freely accessible plants and enclosed plants were analysed using a chi-squared test.


Silencing of endogenous PDS gene in A. thaliana, B. nigra and N. benthamiana plants via its heterologous gene sequence BoPDS from B. oleracea

In the first step, we generated plants in which pigment expression was modified by targeting the gene encoding for phytoene desaturase (PDS), a key enzyme in the biosynthesis of carotenoids, pigments that prevent photobleaching (Ratcliff et al., 2001; Liu et al., 2002). The silencing of PDS, and thus the absence of the gene product, resulted in a typical green-whitish variegated pds phenotype. We used the B. oleracea PDS (BoPDS) gene sequence (EU719210) to obtain VIGS constructs via TRV (Fig. 1c–e) and efficiently silenced plants of three different species, that is A. thaliana (Fig. 2a,b), N. benthamiana (Fig. 3c–e) and B. nigra (Fig. 3h). The pds-silenced leaves in A. thaliana and N. benthamiana plants are completely white along the vascular areas of the veins. This pattern differs from that reported for the im mutant in A. thaliana where most edge areas of the leaves are white, whereas the vascular areas of the veins are still green (Aluru et al., 2006). However, in B. nigra, only a narrow area bordering the veins appeared bleached. EV control plants, which were agroinfiltrated with TRV1 and TRV2 vectors only (Liu et al., 2002), did not show any photobleaching symptoms and were indistinguishable from wild-type, noninfiltrated plants (Figs 2a, 3b,g), as confirmed by reflectance spectra in the 400–800 nm range (Fig. 2c). However, a dramatic difference in reflectance was observed in the whitish parts of PDS-silenced plants compared with EV control and wild-type plants (Fig. 2c). All three plant species studied developed photobleaching symptoms 5–15 d after agroinfiltration with TRV1 and TRV2-BoPDS1 or TRV1 and TRV2-BoPDS2 (Figs 2, 3). These results demonstrate that a heterologous gene sequence from B. oleracea can be used to silence its orthologues in A. thaliana, B. nigra and N. benthamiana. We included the Nicotiana plants here because they show that the PDS sequence from B. oleracea can even silence the PDS gene in a nonbrassicaceous plant. However, the timing and extent of photobleaching varied among plant species. Nicotiana benthamiana plants developed photobleaching symptoms 5 d after infiltration, whereas A. thaliana and B. napus showed symptoms 10 d after infiltration.

Figure 2.

 Silencing of the phytoene desaturase (PDS) gene from cabbage (Brassica oleracea) BoPDS using tobacco rattle virus (TRV)-based virus-induced gene silencing (VIGS) in Arabidopsis Col-0 plants. (a) Wild-type (left), empty vector negative control (middle) and PDS-silenced plant (right) agroinfiltrated with pTRV1 and pTRV2-BoPDS1. (b) Overview of PDS-silenced plants. (c) Reflectance spectra of the three different plant types and leaf parts of plants. Note that the curves for the empty vector control and the green leaf parts of the PDS-silenced plants are almost indistinguishable.

Figure 3.

 Silencing of the phytoene desaturase (PDS) gene using tobacco rattle virus (TRV)-based virus-induced gene silencing (VIGS) in Nicotiana benthamiana and Brassica nigra. (a–e) Nicotiana benthamiana: (a) wild-type; (b) empty vector (EV) control plant; (c, d) PDS-silenced plants of N. benthamiana agroinfiltrated with pTRV1 and pTRV2-BoPDS1; (e) PDS-silenced N. benthamiana plants agroinfiltrated with pTRV1 and pTRV2-BoPDS1 (top two plants), and pTRV1 and pTRV2-BoPDS2 (bottom two plants). (f–h) Brassica nigra: (f) wild-type; (g) EV control; (h) PDS-silenced plant of B. nigra agroinfiltrated with pTRV1 and pTRV2-BoPDS1.

We have shown the expected pds phenotype in all three plant species (Figs 2, 3). For A. thaliana, we supported this evidence with quantitative PCR data to demonstrate that the PDS transcript level was affected. Silencing the PDS gene in A. thaliana plants reduced significantly the abundance of transcripts of AtPDS (6.7-fold compared with the control) (Fig. 4).

Figure 4.

 Effect of agroinfiltration of Arabidopsis plants with specific constructs on the abundance of AtPDS transcripts as quantified by quantitative RT-PCR (empty vector (EV) control plants were treated with pTRV1 and pTRV2, whereas PDS-silenced plants were treated with pTRV1 and pTRV2-BoPDS1; mean ± SE, = 3).

We compared the numbers of A. thaliana plants showing the pds phenotype when growing under long-day (16 h : 8 h light : dark photoperiod) and short-day (8 h : 16 h light : dark photoperiod) conditions. In both long-day and short-day conditions, 80–100% of the plants displayed photobleaching in 10 separate experiments. Furthermore, silencing of AtPDS was equally effective in seedlings inoculated at the two- to three-leaf stage, four- to five-leaf stage or six- to seven-leaf stage (data not shown). Therefore, four- to five-leaf stage seedlings were used in later oviposition experiments under short-day conditions (8 h : 16 h light : dark photoperiod) because, under long-day conditions, A. thaliana plants initiate flowering early. Both B. nigra and N. benthamiana plants were grown under long-day conditions (16 h : 8 h light : dark photoperiod) after agroinfiltration.

Effects of PDS silencing on oviposition behaviour by P. rapae butterflies

We investigated the effects of modified leaf pigmentation on the oviposition behaviour of P. rapae butterflies in A. thaliana and B. nigra plants. Three-week-old PDS-silenced plants and EV control plants of A. thaliana or B. nigra were offered to P. rapae butterflies in dual-choice bioassays. During bioassays lasting either 6 or 24 h, the naive butterflies significantly preferred to oviposit on EV plants rather than PDS-silenced plants (A. thaliana in Fig. 5a, = 23, < 0.001; Fig. 5b, = 24, < 0.001; B. nigra in Fig. 5c, = 14, < 0.05). These results might also be explained by possible pleiotropic effects of pigment alteration on the metabolic pathways of silenced plants, because Pieris butterflies use the presence of secondary metabolites, such as glucosinolates or their volatile breakdown products, as stimuli for oviposition (Van Loon et al., 1992; Loivamäki et al., 2008; de Vos et al., 2008). Therefore, to assess the possible role of contact cues, of either gustatory or tactile nature, we recorded whether naive butterflies landed first on EV plants or on PDS-silenced plants. We observed that as many as 81% of the first landings were made on EV Arabidopsis plants (Fig. 6a, = 100, < 0.001; see Movie S1 for a video recording of butterfly behaviour in this dual-choice bioassay). Furthermore, to test the possible effect of differences in plant volatile production, we offered both EV and PDS-silenced plants in air-tight sealed transparent containers. In this bioassay, 90% of the first landings were on the contained EV Arabidopsis plants (Fig. 6b, = 100, < 0.001). The results for landing behaviour in the experiments on freely accessible plants (Fig. 6a) and enclosed plants (Fig. 6b) were similar (χ2 = 2.304, > 0.05), thus clearly demonstrating that the observed preference for EV plants was not influenced by a possible difference in plant volatiles. We have not performed experiments with B. nigra in containers because the A. thaliana rosette plants were much more amenable for such experiments than the B. nigra plants. All previous results demonstrate that the discrimination by butterflies is based on visual cues only. These results allow the unambiguous conclusion that Pieris butterflies possess an innate visual ability to discriminate green from green–white variegated plants. This is the first time that this ability has been demonstrated in response to living plant material by silencing only one functional gene. Our conclusion is supported by previous studies employing coloured papers (Traynier, 1986) and our own experiments with such papers (Figs S1, S2). Our present experiments with PDS-silenced plants allow the rigorous conclusion that this ability extends to the variable colour patterning of living plant leaves.

Figure 5.

 Oviposition (number of eggs per plant) by Pieris rapae butterflies in two-choice assays on empty vector (EV) control plants (closed boxes) and PDS-silenced plants (open boxes) after different exposure durations. (a) Arabidopsis thaliana exposed to P. rapae oviposition during 6 h (total number of eggs laid: 672 on 23 pairs of plants). (b) Arabidopsis thaliana exposed to P. rapae oviposition during 24 h (total number of eggs laid: 1642 on 24 pairs of plants). (c) Brassica nigra exposed to P. rapae oviposition during 6 h (total number of eggs laid: 593 on 14 pairs of plants). The boxes span the first to third quartile range; the median is indicated by a line across the box. Outliers are represented by dots. Asterisks indicate significant differences between oviposition responses to EV and PDS-silenced plants (*< 0.05; ***< 0.001; Wilcoxon matched-pair signed-ranks test).

Figure 6.

 First landing choice by Pieris rapae butterflies in a dual-choice bioassay on empty vector (EV) control and PDS-silenced Arabidopsis plants. (a) Plants exposed to butterfly landing (total number of butterflies observed to land was 100). (b) Plants were offered as optical stimuli to butterflies enclosed in air-tight sealed transparent boxes (total number of butterflies observed to land was 60; ***< 0.001, binomial test). There is no significant difference (ns) between experiment (a) and experiment (b) (χ2 = 2.304, > 0.05), which shows that volatiles played no significant role in butterfly landing preference.

Performance of Pieris caterpillars on PDS-silenced green-whitish plants

We further investigated how the silencing of PDS in A. thaliana influences the performance of Pieris caterpillars that hatch from the eggs. We predicted that caterpillars would perform less well on PDS-silenced than on EV A. thaliana plants. To test this hypothesis, we offered 3-wk-old PDS-silenced and EV A. thaliana plants to P. rapae butterflies to oviposit five eggs per plant. On both plant types, the eggs all hatched 4 d after oviposition. After another 4 d of feeding, the fresh body mass of individual larvae from 18 plant pairs was determined. Surprisingly, hatched larvae only fed on the green parts of pds-silenced plants and not on the white parts of the plants (Fig. 7b). Indeed, caterpillar performance was found to be reduced significantly on PDS-silenced plants compared with EV control plants. Pieris rapae larvae grew much more slowly on PDS-silenced compared with EV Arabidopsis plants (Fig. 7c; Student’s t-test, = 11.55, < 0.001). The reduction in caterpillar growth by 20% over the first 4 d will probably become larger during further development, as a caterpillar’s development is important for performance later on (Gols et al., 2008).

Figure 7.

 Performance of Pieris rapae caterpillars on empty vector (EV) control green and green-whitish PDS-silenced plants. (a) Arabidopsis plants before butterfly oviposition. (b) Plants 8 d after oviposition (left, control plants; right, silenced pds plants). (c) Fresh body mass of caterpillars on the two types of plant (18 plants were used for each plant type; mean ± SE; ***< 0.001, Student’s t-test).


The VIGS vectors that were originally developed in the model species N. benthamiana (Ratcliff et al., 2001; Burch-Smith et al., 2004Goodin et al., 2008;) have been used successfully to decipher the functions of several plant genes in this species, which yields high infection rates, and in other solanaceous species such as tomato (Liu et al., 2002;Bhattarai et al., 2007; Kandoth et al., 2007; Hartl et al., 2008; Shao et al., 2008; Wu et al., 2008). Recently, Senthil-Kumar et al. (2007) have shown that TRV-mediated VIGS can be performed in a wide range of solanaceous plant species, and that heterologous gene sequences from distantly related plant species can be used to silence their respective orthologues in the VIGS-efficient model plant N. benthamiana (Goodin et al., 2008). However, the use of VIGS in high-throughput functional genomics has thus far been limited, mostly as a result of the difficulties of introducing the VIGS vector into plant species other than solanaceous species. These limitations also exist for the model plant of molecular genetics, A. thaliana, although optimized procedures for the efficiency of agroinoculation of this species have been reported recently (Burch-Smith et al., 2006; Wang et al., 2006; Pflieger et al., 2008). However, it was unknown to date whether heterologous gene sequences from other plant species can be used to silence their respective orthologues in VIGS-refractory plant species, such as B. nigra. In A. thaliana, only seedlings in the two- to three-leaf stage grown under long-day conditions have been shown previously to display the photobleaching phenotype (Burch-Smith et al., 2006). Here, we report an efficient protocol for silencing the PDS gene in A. thaliana plants under both long-day (16 h light) and short-day (8 h light) conditions for plants in the two- to seven-leaf stages. Furthermore, our study shows that heterologous gene sequences from B. oleracea can not only silence its orthologues in the related brassicaceous plants, such as B. nigra and A. thaliana, but also in the nonrelated solanaceous model species N. benthamiana. In this study, we have exploited the unique opportunities that VIGS provides to investigate the role of a functional PDS in interactions of plants with herbivorous insects. By employing VIGS, we have unambiguously shown that Pieris butterflies discriminate between green plants and green–white variegated conspecifics on the basis of visual cues only. Although chemical cues are very important for foraging Lepidoptera in general, and for Pieris spp. in particular (Chew & Renwick, 1995; Schoonhoven et al., 2005), a potential role for such cues has been ruled out by our experiments in which the butterflies displayed the same preference for EV plants over PDS-silenced plants when the plants were enclosed in air-tight containers. To the best of our knowledge, this is the first time that pigment-manipulated living plants have been used to study the effects of variegated plant tissues on insect behaviour. So far, studies on the role of vision in host plant selection for oviposition by insects have employed the use of nonliving substrates (paper) and dyes. However, the spectral characteristics and patterns of artificial substrates never match exactly plant pigmentation patterns, as has also been shown experimentally for coloured papers vs Brassica plants (Smallegange et al., 2006). In the present study, we have effectively silenced one gene, i.e. PDS, and investigated the effect of this on butterfly behaviour. PDS is a crucial gene in carotenoid biosynthesis and thereby in the harvesting of light, the peroxidation of lipids, hormone synthesis and assembly of the photosystem. All of these functions of PDS have received ample attention in previous studies (Grotewold, 2006). However, one more phenotypical component that is influenced by PDS is the plant’s visual characteristics to animals, such as butterflies. It is known that the inhibition of PDS in the im mutant of A. thaliana results in white–green variegated plants (Wetzel et al., 1994; Aluru et al., 2009) and white–green variegated forms are also known for many other plant species (Evenari, 1989; Lev-Yadun, 2006). It is interesting to note that a recent study has shown that, in the araceous plant Caladium steudneriifolium, the degree of variegation, as artificially manipulated with white Tipp-Ex® correction fluid, is negatively correlated with caterpillar damage (Soltau et al., 2009). The authors have suggested that the white–green variegated plants deter ovipositing moths by mimicry of miner infestation (Soltau et al., 2009). Results supporting this suggestion have also been reported in a nonmanipulative study with naturally occurring Hydrophyllum virginianum (Campitelli et al., 2008). Thus, variegation might be beneficial for plants despite the implicated loss of photosynthetically active surface. It remains to be investigated what are the net fitness consequences, and this is likely to depend on local conditions, especially related to herbivore abundance. Indeed, in nature, natural variation in the degree of variegation was found in C. steudneriifolium (Soltau et al., 2009), and the role of plant colour variegation in plant anti-herbivore defence has been hypothesized previously (Lev-Yadun, 2006).

Pieris butterflies probably have an innate preference for green because most of their host plants are green (Snell-Rood & Papaj, 2009). Moreover, the green-whitish variegated optical appearance of the PDS-silenced plants mimics the generally observed yellow-whitish symptoms of virus infection or the patterns of leafminer infestation. The data of our larval growth experiment show that the PDS-silenced plants are nutritionally inferior: caterpillars grow more rapidly on EV plants than on PDS-silenced plants, and thus butterfly preference is positively correlated with offspring performance. It is important to note that, although not displaying symptoms that are optically detectable, EV plants are infected with TRV. One probable cause of the nutritional inferiority of PDS-silenced plants is that the abundance of transcripts of AtPDS is reduced significantly in PDS-silenced A. thaliana plants (Fig. 4). Low-carotenoid plants are likely to have many chemical differences from control plants. The absence of carotenoids obviously leads to oxidation of chlorophyll and probably other compounds (Aluru et al., 2006; Fang et al., 2008). This results in a lower carbohydrate content which, in turn, may result in changes in other nutrients (Aluru et al., 2006). However, in our oviposition experiments on enclosed plants, we explicitly excluded the possible effects of such chemical changes on butterfly behaviour: here only visual stimuli could be used by the butterflies. Carotenoid deficiency of plants is likely to have an impact on the visual capacity of insects developing on such plants. Rhodopsins are crucial visual pigment molecules in insect eyes, composed of an opsin moiety and a retinal-based chromophore (Stavenga & Arikawa, 2006). The biosynthetic precursors of 11-cis-retinal and 11-cis-3-hydroxyretinal are plant-derived carotene and xanthophylls (Seki & Vogt, 1998).

In this study, we have exploited, for the first time, a molecular genetic approach to investigate the role of vision in insect–plant interactions; this was performed by generating plants that are genetically identical, except for the functional expression of one gene, leading to a leaf colour pattern difference. EV plants may differ in more characteristics than only colour from PDS-silenced plants. However, in the current study, we were able to specifically address the optical aspects and eliminated other aspects through the experimental set-ups used. Our behavioural data show that only the colour change associated with PDS silencing influenced P. rapae’s host plant selection behaviour. The PDS gene has been mainly used previously as a marker of successful silencing. Mutation of PDS in rice leads to preharvest sprouting and photo-oxidation because of its influence on the carotenoid precursors of ABA biosynthesis (Fang et al., 2008). PDS is also differentially regulated by photo-oxidative stress and developmental mechanisms that control carotenoid biosynthesis in leaves, flowers and fruits in tomato (Giuliano et al., 1993). However, the effect of modified PDS expression on plant–animal interactions has never been investigated.

It is interesting to note that neonate P. rapae larvae only fed on the green parts of pds-silenced plants and avoided feeding on the white leaf parts. The white sectors of im phenotypes of A. thaliana accumulate low levels of sucrose in contrast with the green sectors (Aluru et al., 2006). Transcriptome studies have suggested that energy is deprived via aerobic and anaerobic metabolism of imported sugar for growth and development in im white leaf sectors, and that oxidative stress responses are induced in these sectors (Aluru et al., 2009). These and other effects of defective carotenoid biosynthesis may explain why the caterpillars avoid feeding on the white parts of the variegated plants through nutritional feedbacks. Thus, elimination of the functional expression of this particular gene not only affects animal behaviour through its effect on the production of visual cues, but also affects feeding site selection by the herbivorous stages. These data relate to laboratory studies and cannot yet be extrapolated to the field. However, it is interesting to see that field studies support our conclusion that variegated plants receive less herbivory than green plants (Soltau et al., 2009). Our approach allows the study of putative adaptive shifts in host plant selection and feeding behaviour as affected by plant carotenoids.

Although Arabidopsis can be used as a model plant species to study plant–insect interactions (Kappers et al., 2005), it is important to point out that our protocol can be applied to other, nonmodel, cruciferous species, such as B. nigra, which is a close relative of other economically important crops, such as B. oleracea and B. napus. In addition, the variegated phenotype of B. nigra PDS-silenced plants is subtle compared with that of A. thaliana (Figs 2b, 3h). It is indeed highly remarkable that the butterflies significantly avoided pds-silenced B. nigra plants (Fig. 5c). These findings imply that TRV-based VIGS holds promise as a powerful tool for molecular and behavioural ecology of plant–insect interactions for which mutant plants are not available. Quantitative variation in character expression is an important subject in ecology, and a next step is the development of VIGS to allow different levels of silencing of the target gene (Zheng & Dicke, 2008). The use of VIGS can be further extended to provide unique optical plant phenotypes by the targeted modification of the carotenoid pathway, allowing controlled modification of carotenoid types, levels and their patterns of expression within plant organs. Understanding how herbivorous insects exploit visual signals during host plant selection may contribute to improved insect pest management (Foster & Harris, 1997).


We thank L. Koopman, F. van Aggelen and A. Gidding for insect rearing and Unifarm of Wageningen University for supplying Brassica nigra and Nicotiana benthamiana plants. We are grateful to S. P. Dinesh-Kumar (Yale University, USA) for kindly providing the VIGS vectors, D. G. Stavenga and S. S. Liu for comments and discussion, T. Bukovinszky for help with data analysis, H. M. Smid for technical assistance, and V. C. F. Dupré, C. E. Jongeling and H. P. M. Thoens for help with the experiments. The comments on a previous version of the manuscript by four anonymous reviewers have been very helpful in improving the manuscript. This work was supported financially by a VICI grant from the Netherlands Organization for Scientific Research, NWO (865.03.002).