A full-length cDNA clone of N. benthamiana PDS (NbPDS) was obtained using RACE PCR. The NbPDS cDNA was 2046 bp in length corresponding to a 1758 bp coding region, 223 bp 5' untranslated region (UTR) and a 65 bp 3' UTR (Figure 1a). The open reading frame of NbPDS showed high nucleotide identity with tobacco PDS (97%) and tomato PDS (90%). Our strategy for producing TRV-NbPDS constructs was to amplify fragments of NbPDS by PCR, clone them into a Gateway entry vector and recombine them into the TRV-RNA2 vector pYL279 21. The sequences of oligonucleotide primers used in this study and the primer combinations for amplifying VIGS inserts are shown in Table 1 and Table 2, respectively. All NbPDS cDNA inserts were in the antisense orientation relative to the TRV coat protein. Since all cloning procedures herein are performed with TRV cDNAs, in this paper we refer to the inserts as cDNAs. Following agro-infiltration and transcription driven by the 35S-promoter, the inserts exert their silencing effects as part of the single-stranded RNA genome of TRV. Four-week-old N. benthamiana plants were infiltrated with TRV-NbPDS constructs and TRV-RNA1. Plants that received infiltration buffer and those infected with a TRV-RNA2 vector lacking an insert served as negative controls. Control plants did not exhibit bleaching symptoms in any of our experiments although infection with empty TRV and TRV-GFP produced slightly stunted plants with some crinkled leaves.

The effect of silencing PDS was determined by analysis of chlorophyll a levels in pooled leaf and petiole tissue sampled from above the infiltrated leaves, with the reduction in pigment content through photobleaching used as a measure of silencing efficiency. In the case of PDS silencing, leaves may be green, white, pale yellow, pale green, or white with regions of dark green tissue. Since we did not separate green (non-silenced, unbleached) from white (silenced, bleached) tissue, measuring chlorophyll a levels summed the ability of the TRV construct to both spread within the plant and to silence PDS within infected tissue.

TRV-NbPDS constructs containing full-length NbPDS (2046 bp) or NbPDS cDNAs truncated at the 5' end (1661 bp, 1304 bp, 959 bp, 610 bp and 257 bp) were amplified with gene specific primers. Figure 1a shows a schematic summary of NbPDS constructs used in this experiment. The length of the NbPDS insert influenced silencing efficiency as demonstrated by quantification of chlorophyll a and visual observation of leaf bleaching (Figure 1b and 1c). TRV-NbPDS (2046 bp) infected plants showed limited bleaching, with small regions of white tissue occasionally forming near major leaf veins, and chlorophyll a levels were only slightly lower than mock or empty TRV infected plants. Silencing efficiency increased in plants infected with TRV-NbPDS (1661 bp), although chlorophyll a levels were about 86% of empty TRV control levels and much leaf tissue remained green. Further increases in bleaching were found with the 1304 bp, 959 bp and 610 bp inserts, all of which produced plants with completely white leaves. TRV-NbPDS (257 bp), the shortest insert in this series, was a less effective silencer than the longer inserts and produced leaves that were mottled and pale green.

To further probe the effect of insert size on silencing, TRV-NbPDS constructs with inserts of 345 bp, 192 bp, 103 bp and 54 bp from the middle of the NbPDS cDNA were generated. TRV-NbPDS (345 bp) and TRV-NbPDS (192 bp) were capable of silencing PDS expression. However the TRV constructs with 103 bp and 54 bp inserts did not lead to effective silencing; indeed plants infected with TRV-NbPDS (54 bp) had chlorophyll levels that were slightly higher than controls. The bleached leaves formed by silencing with the 345 bp, 192 bp and 103 bp inserts all contained small regions of green tissue near leaf veins, giving them a mottled appearance (Figure 1c).

We next tested if the position of the insert with respect to the full-length cDNA affected silencing. TRV constructs were synthesized using primers designed to amplify a non-overlapping series of short NbPDS fragments (Figure 2a). In this manner we synthesized constructs of TRV-NbPDS 1–385 (385 bp), TRV-NbPDS 386–742 (357 bp), TRV-NbPDS 743–1087 (345 bp), TRV-NbPDS 1088–1436 (349 bp) and TRV-NbPDS 1437–1789 (353 bp), which together with the NbPDS 1788–2046 (257 bp) construct, gave complete coverage of NbPDS. Plants infected with these constructs showed that all were effective in silencing but slightly more bleaching was produced by constructs located in the middle of the cDNA compared to the 5' and 3' ends (Figure 2b). This effect was apparent in the fine green mottling on the leaves of TRV-NbPDS 1–385 (385 bp) and TRV-NbPDS 1788–2046 (257 bp) infected plants while silencing inserts from the middle of the NbPDS cDNA gave more uniformly bleached leaves.

Since cDNAs often contain 3' homopolymeric regions (i.e. poly(A) tails) that are incorporated as part of the cDNA synthesis process, we tested the effect that inclusion of homopolymeric regions has on VIGS efficiency. Four silencing inserts that had been shown to be effective silencers in our first experiment (1304 bp, 959 bp, 610 bp, 257 bp) were amplified using an oligo(dT) primer that added a 3' 24 bp homopolymeric region. This gave inserts of 1328 bp, 983 bp, 634 bp and 281 bp. Plants infected with TRV constructs containing polyadenylated inserts showed reduced silencing efficiency compared to constructs lacking this region (Figure 3). The inclusion of the 24 bp poly(A/T) region had a particularly dramatic effect on the silencing effectiveness of the 1304 bp and 257 bp inserts. Silencing with the 1304 bp insert led to greater than 95% reduction in chlorophyll a levels compared to empty TRV control plants while a 21% decrease was found with the 1328 bp homopolymeric insert. We also added a poly(G) region to the 345 bp insert. The silencing effectiveness of this insert, which gave a 94% reduction in chlorophyll a levels compared with empty virus controls, was strongly reduced by the addition of the poly(G) sequence (Figure 3b).

Based on the optimal properties for VIGS inserts defined by the above experiments, we developed a cDNA library synthesis method suitable for constructing VIGS libraries. This method, which is outlined in Figure 4, utilizes suppression subtractive hybridization (SSH) as a means to produce subtracted libraries of short cDNAs that differ in their expression between a reference tissue (driver) and one with an altered gene expression profile (tester) 23. We used magnetic bead-linked oligo(dT) primers to capture mRNA and then synthesized cDNA on this solid-phase support. Subsequent digestion of the cDNA with RsaI freed the cDNA in ~600 bp fragments leaving the 3' ends and their poly(A/T) regions immobilized on the magnetic beads. Subsequent SSH cloning and ligation procedures were performed as described by Diatchenko et al. 24. In order to introduce cDNA fragments into the TRV-RNA2 VIGS vector, we modified the final PCR amplification procedure by using primers that introduced EcoRI and BamHI recognition sites onto the ends of the subtracted amplicons. Subsequent digestion, ligation into the TRV-RNA2 VIGS vector pYL156 and transformation into E. coli gave a high-titre cDNA library in a plasmid vector. The library was amplified in E. coli before transformation into Agrobacterium.

To demonstrate the utility of the VIGS-cDNA library synthesis method, and with the objective of using VIGS to study alkaloid metabolism in the Solanaceae, we constructed two subtractive VIGS-cDNA libraries from Nicotiana benthamiana roots. Nicotine is synthesized in the roots of Nicotiana species and transported to the leaves where it serves as a defense against insect herbivory 25. Nicotine biosynthesis is up-regulated in response to insect feeding, wounding or the application of jasmonates 26. We treated the roots of hydroponically grown N. benthamiana plants with methyl jasmonate and harvested roots 1, 3, 7 and 10 hours after treatment. Quantitative real-time PCR was use to confirm that the expression several genes involved in nicotine biosynthesis, including ornithine decarboxylase, putrescine N-methyltransferase (PMT), and quinolinate phosphoribosyl transferase, increased with methyljasmonate treatment (data not shown). Two subtracted cDNA libraries were constructed: (1) NBREL1 using by mRNA pooled from methyljasmonate-treated roots as tester and untreated root mRNA as driver, and (2) NBLEL2, using by mRNA pooled from methyljasmonate-treated roots as tester and untreated leaf mRNA as driver. The libraries were amplified in E. coli, transformed into Agrobacterium and plated on agar. For each library, 1920 colonies were transferred to 96-well plates, grown overnight and used as templates for PCR. The resulting amplicons were sequenced from the 5' end. The Agrobacterium cultures were stored as glycerol stocks for use in VIGS experiments.

The properties of the VIGS-cDNA libraries are shown in Table 3. The sequencing success rate of 77% was relatively low owing to variation in the amount of PCR products amplified from Agrobacterium colonies. Analysis of 150 cDNAs from NBREL1 and 77 from NBLEL2 showed that insert sizes ranged from 150 bp to 1150 bp with the majority (30%) in the size range of 401–500 bp (Figure 5). Less than 1% of NBREL1 and about 9% of NBLEL2 contained inserts smaller than 200 bp. We detected 18 cDNAs (0.5% of the total number sequenced) with poly(A) tails confirming that our strategy of immobilizing cDNAs before RsaI digestion was successful. One problem that was apparent from EST sequencing was that some inserts consisted of fragments of two or more genes. Such chimeric clones were of relatively low frequency (e.g. five of 150 cDNAs in NBREL1).

To test the effectiveness of gene silencing using TRV constructs derived from the subtracted VIGS-cDNA library, we silenced PMT, a gene encoding a rate-limiting enzyme in the nicotine biosynthetic pathway. VIGS of PMT has previously been shown to reduce nicotine levels in Nicotiana attenuata 27. Fifteen PMT ESTs, which corresponded to a fourteen-membered contig and one singleton, were present in the subtracted library and we selected ten representative constructs for VIGS. The ten inserts we distributed across the PMT cDNA but most aligned to the 3' end (Figure 6). They ranged in size from 122 bp to 517 bp and were in both the sense and antisense orientation relative to the TRV coat protein.

The TRV-NbPMT constructs were agro-infiltrated into three-week-old N. benthamiana plants. Control plants received infiltration buffer or a TRV-GFP construct. We used the latter because it reduced virus symptoms such as stunting and leaf curling compared with the empty TRV-RNA2 control. Plants were grown for three weeks and sprayed with a solution of methyljasmonate to stimulate nicotine production. After five days leaf nicotine levels were determined by reversed-phase ion-pair HPLC. Nicotine levels in buffer control of TRV-GFP infected plants were 326 μg/mg fresh weight (FW) and 416 μg/mg FW, respectively. Nicotine levels were reduced by greater than 90% in plants infected with all ten TRV-NbPMT constructs (Figure 5b). The levels of leaf nicotine in silenced plants ranged from undetectable to 37 μg/mg FW (mean of 10 μg/mg FW, n = 10). Both sense and antisense orientated inserts performed with equal efficiency.

The influence of host insert size on VIGS efficiency has received little experimental attention. Our results show that cDNAs longer than about 1500 bp, such as the 1661 bp NbPDS insert, gave little or no silencing of PDS. One explanation for this result is that viral replication is impaired when a large foreign sequence is inserted in its genome. Another possibility is that large inserts are removed by recombination and that, although TRV is spreading systemically, the silencing insert has been lost. The replication and movement of RNA viruses is often impaired when foreign sequences are inserted into the viral genome and such sequences are readily deleted through recombination. Insertion of a 1.6 kb bacterial β-glucoronidase (GUS) gene in plum pox potyvirus delayed virus accumulation and led to deletion of GUS after passaging in N. benthamiana 29. Similarly, recombinant poliovirus containing GFP (~750 bp) showed impaired replication and rapid deletion of the GFP gene in HeLa cells 30. GFP and GUS tagging of lettuce mosaic virus attenuated virus symptoms 31. Tobraviruses including TRV show a high degree of genome alteration through recombination 32, which suggests that non-optimal inserts may be quickly deleted or altered through recombination. Limitations on insert size due to packaging constraints have been observed for some viruses such as adeno-associated virus 33. In designing the TRV vectors used in this study, Liu et al. 721 deleted genes encoding the 28.7 k and 32.8 k proteins from a wildtype Ppk20 virus and inserted multiple cloning and Gateway sites between the coat protein and the 3' untranslated region. This reduced the size of TRV-RNA2 from 3855 bp to 2103 bp, or about 1.7 kb, for pYL156 and 2.3 kb for pYL279. This is near the upper limit that we determined, and suggests that there may be natural constraints for the size of the TRV genome. The size constraints uncovered in our study likely only apply to TRV and not other VIGS viruses such as those based on geminiviruses. Further experiments will be needed to determine the optimal inserts properties in other VIGS vectors.

The lower limit for efficient PDS silencing using TRV was determined to be about 190 bp and 122 bp for NbPMT. The difference in these lengths makes it clear that there is variation in the silencing of different genes. However, the fact that PDS required a ~200 bp fragment suggests that other researchers should use this as a minimum length rather than the 122 bp observed with PMT. The lower limit sizes (190 and 122 bp) are larger than the 23 nucleotides that was obtained using potato virus X (PVX), although such short fragments gave patchy phenotypes 34. This discrepancy may be due to the use of different VIGS viruses (PVX vs. TRV) in the experiments. In addition, the PVX study used a GFP transgene, rather than an endogenous gene, as a silencing target. Inserts longer than 190 bp likely lead to the generation of more and diverse siRNAs, and these silence more effectively than short inserts.

The position of the VIGS fragment relative to the full-length cDNA had a small effect on silencing efficiency (Figure 2). The 385 bp 5' fragment and the 3' 257 bp fragment, both of which included both coding sequence and untranslated regions, showed increased chlorophyll a levels compared to the other four constructs tested. The 257 bp insert from the extreme 3' end in particular was less effective. Using PVX, Ruiz et al. 35 assayed PDS silencing in N. benthamiana and found that inserts derived from the middle of the NbPDS cDNA (415 bp and 212 bp fragments), at the 5' end (377 bp fragment) were able to silence PDS. Inserts targeting 5' and 3' located introns (167 bp and 223 bp, respectively) did not trigger silencing. In general, other papers have recommended situating the VIGS fragment in the middle of the coding region of the transcript, except were targeting the 5' or 3' ends gives more specific silencing of gene family members (e.g. 36). Our study indicates almost any part of the cDNA can be used to trigger silencing but that avoiding the extreme 5' and 3' ends may increase silencing efficiency.

We observed that the presence of a 24 bp poly(A) region dramatically reduced the efficiency of silencing (Figure 3). A possible explanation for this effect is that poly(A) binding proteins interacting with this region interfere with viral replication. However, addition of both poly(A) and poly(G) sequences to the 345 bp insert, which previous experiments had shown acts as an efficient silencer, compromised silencing. These data indicate the effect was not specific to polyadenylated regions but rather homopolymeric regions in general. The TRV genome does not naturally contain polyadenylated regions 32.

Expressed sequence tags (ESTs) have proven to be one of the most useful types of genomic information, especially in eukaryotic organisms with large, unsequenced genomes. By constructing cDNA libraries directly in a VIGS vector and then sequencing randomly selected cDNAs, we aimed to produce a collection of VIGS-ready cDNAs in Agrobacterium and a corresponding EST dataset that could be used for functional genomics. The cDNA library method we developed produces short cDNA inserts lacking poly(A/T) regions. In addition, the use of subtractive cDNA library methodology allows for the enrichment of clones from certain tissues or those expressed in response to treatments (e.g. methyljasmonate treated N. benthamiana roots). Using this method we constructed two subtractive VIGS-cDNA libraries, analyzed their corresponding EST datasets for cDNAs matching a key enzyme in the nicotine pathway, and demonstrated efficient silencing using ten different TRV-NbPMT constructs. We are currently applying this approach to identify additional enzymes and regulatory proteins involved in alkaloid metabolism in the Solanaceae. This VIGS-EST approach should be useful for both fast forward and reverse genetics experiments in plants for which genomic information is limited.

A full-length NbPDS cDNA was obtained by RACE-PCR (GeneRacer kit, Invitrogen) using gene-specific primers based on a published NbPDS cDNA fragment. The full-length cDNA was amplified from RACE-ready cDNA with primers corresponding to the 5' end of the NbPDS cDNA and the GeneRacer 3' nested primer using Pfu polymerase and sequenced. NbPDS cDNA fragments from N. benthamiana were PCR amplified with Taq polymerase using the oligonucleotide primers shown in Table 1 and the full-length NbPDS cDNA as template. Due to errors in the draft NbPDS sequence, two primers (NbPDSfor2 and NbPDSrev6) had sequences that differed from the final sequence. PCR products were gel purified and ligated into pENTR-D-TOPO (Invitrogen). The TRV-GFP construct contained a non-functional 363 bp fragment of GFP. After sequencing, cDNA inserts were recombined into the Gateway TRV-RNA2 vector pYL279 21 using LR recombinase (Invitrogen). Constructs were confirmed by sequencing before transformation into A. tumefaciens C58 via electroporation. The nucleotide sequences of NbPDS [GenBank: EU165355] and NbPMT [GenBank: EU165356] have been deposited in GenBank.

N. benthamiana plants were grown in soil in a controlled environment chamber with 16 hour/23° days and 8 hour/20° nights at 100 μmol/m²/s light intensity. Cultures of A. tumefaciens containing TRV-RNA1 or TRV-RNA2 plasmid were cultured separately overnight at 28°C. After centrifugation, bacterial cell pellets were resuspended in infiltration buffer containing 1 mM MES (pH 5), 10 mM MgCl₂ and 100 μM acetosyringone to OD₆₀₀ = 1 and allowed to stand at room temperature for 3–6 hours before infiltration. Suspensions of TRV-RNA1 and TRV-RNA2 constructs were mixed 1:1 and infiltrated into the underside of the upper leaves of 3–4 week old plants using a 1 ml syringe. Mock infected plants received infiltration buffer.

N. benthamiana leaves and petioles were sampled 3–4 weeks after infiltration, frozen in liquid nitrogen and ground to a fine powder. Portions of approximately 150 mg were transferred to a microcentrifuge tube and accurately weighed. One-ml of 90% (v/v) aqueous acetone was added and the sample vortexed for 1 min. The extract was centrifuged for 1 min at 13,200 g and aliquots of the supernatant measured at 630, 645, 663 and 750 nm using a UV-vis spectrophotometer (Lambda 35, PerkinElmer) and chlorophyll a amounts calculated 37. Values were calculated as mean and standard deviation for three plants (n = 3) of one infiltration experiment. Representative plants or leaves were photographed at the time of chlorophyll a analysis using a digital camera.

N. benthamiana seedlings were grown hydroponically in 0.25× Hoagland's solution supplemented with iron chelate solution and oxygenated using an aquarium bubbler. Roots from three-week old plants were sampled immediately before and at 1, 4, and 7 h after addition of methyljasmonate to a final concentration of 11 μM. Total RNA was isolated from 450 mg each of untreated leaves, untreated roots, and a combined root sample composed of 150 mg roots each from the 1, 4 and 7 h time points using a RNeasy midi kit (Qiagen). A PCR-select subtractive cDNA library kit (Clontech) was used for cDNA synthesis with some modifications. Briefly, about 250 μg of total RNA was mixed with 300 μl of Oligo(dT)₂₅ Dynabeads (Dynal Biotech) in binding buffer (20 mM Tris-HCl pH 7.5, 1 M LiCl, 2 mM EDTA). After a 10 min incubation, the beads were washed three times with washing buffer B (10 mM Tris-HCl pH 7.5, 0.15 M LiCl, 1 mM EDTA), followed by washing twice with first strand buffer. The washed beads containing mRNA were resuspended in 40 μl of cDNA synthesis cocktail (8 μl 5× first strand buffer, 4 μl 10 mM dNTPs, 24 μl RNase-free water and 4 μl (8 U) AMV reverse transcriptase) and incubated at 42°C for 1.5 h. The second strand synthesis was completed by addition of 120 μl of second strand synthesis cocktail (32 μl of 5× second strand buffer, 3.2 μl of 10 mM dNTPs, 8 μl of 20× enzyme cocktail and 76.8 μl RNase free water) and incubation at 16°C for 2 h, followed by addition of 4 μl (12 U) T4 DNA polymerase and incubation for 30 min. The reaction was stopped by addition of 20 μl 0.5 M EDTA. The beads were magnetically separated, the supernatant removed, the beads resuspended in 500 μl of wash buffer (5 mM Tris-HCl pH 7.5, 0.5 mM EDTA, 1 M NaCl, 1% SDS and 10 μg/ml glycogen) and heated at 75°C for 15 min. The beads were then washed three times with wash buffer (5 mM Tris-HCl pH7.5, 0.5 mM EDTA, 1 M NaCl and 200 μg/ml BSA), followed by two more washes with RsaI buffer. The beads were resuspended in 84 μl water, 10 μl 10× RsaI buffer, 3 μl (30 U) RsaI, and incubated at 37°C overnight. The freed cDNA was isolated by magnetic separation of the beads and was used for adapter ligation, hybridizations and primary PCR as described in the manufacturer's protocol (Clontech). Secondary PCR was performed using primers 5'-CG

GGATCC

GAATTC

Three weeks after agro-infiltration, the leaves of N. benthamiana plants were sprayed with a solution of 0.1% (v/v) of methyljasmonate in water containing 0.1% Tween-20. Five days after spraying, three leaf disks (~30 mg) were collected from the upper leaves directly into a 2 ml screw top vial. The tubes were briefly centrifuged to pellet the leaf disks and 50 μl of 1 mm zirconium beads (BioSpec Products) and 100 μL of 50 mM citrate buffer (pH 3)-methanol (70:30) were added. The tubes were homogenized for three minutes in a tissue disrupter (Mini-Beadbeater-96, BioSpec Products), followed by sonication for 10 minutes in an ultrasonic bath. The homogenate was centrifuged at 4° for five minutes at 14,000 rpm in a microcentrifuge and the supernatant transferred to a 96-well filter plate (0.45 μm, MultiScreen Solvinert, Millipore). The plate was centrifuged at 4°C for 2 min at 4,000 rpm. The filtrate was transferred to an autosampler vial and a 20 μl aliquot analyzed by HPLC. HPLC analysis was performed on a Waters 2695 separations module equipped with a Waters XTerra C18 reversed phase column (4.6 × 150 mm, 5 μm) with precolumn at a column temperature of 60°C. The mobile phase consisted of 50 mM citrate buffer (pH adjusted to 3.0 with triethylamine) containing 10 mM octanesulfonic acid-MeOH (76:24). The separation was performed for 15 min at a flow rate of 1 ml/min. Nicotine was detected at 261 nm via photodiode array detection. Quantification was performed using peak area by comparison to a standard curve (r² 0.999) derived from injection of nicotine solutions ranging in concentration from 1 μg/ml to 1040 μg/ml.