A Ligation-Independent Cloning Tobacco Rattle Virus Vector for High-Throughput Virus-Induced Gene Silencing Identiﬁes Roles for NbMADS4 - 1 and - 2 in Floral Development

Virus-induced gene silencing (VIGS) is a widely used, powerful technique for reverse genetics. VIGS vectors derived from the Tobacco rattle virus ( TRV ) are among the most popular for VIGS. We have developed a TRV RNA2 vector that allows the insertion of gene silencing fragments by ligation-independent cloning (LIC). This new vector has several advantages over pre- vious vectors, particularly for applications involving the analysis of large numbers of sequences, since TRV-LIC vectors containing the desired insert are obtained with 100% efﬁciency. Importantly, this vector allows the high-throughput cloning of silencing fragments without the use of costly enzymes required for recombination, as is the case with GATEWAY-based vectors. We generated a collection of silencing vectors based on 400 tomato ( Solanum lycopersicum ) expressed sequence tags in this TRV-LIC background. We have used this vector to identify roles for SlMADS1 and its Nicotiana benthamiana homologs, NbMADS4 - 1 and - 2 in ﬂowering. We ﬁnd that NbMADS4 - 1 and - 2 act nonredundantly in ﬂoral development and silencing of either gene results in loss of organ identity. This TRV-LIC vector should be a valuable resource to the plant community.

The last decade has seen an explosion in the availability of plant gene sequences. The genomes of the model species Arabidopsis (Arabidopsis thaliana) and rice (Oryza sativa) have both been sequenced, while those of tomato (Solanum lycopersicum) and maize (Zea mays) are currently being sequenced (Mueller et al., 2005; http://www.sgn.cornell.edu/about/tomato_project/; http://www.maizesequence.org/index.html). Large collections of ESTs have also been generated for a variety of species that are widely used for research purposes. Concomitant with the availability of this sequence information, many important aspects of plant growth and development have been analyzed by DNA micro-arrays, leading to the identification of numerous genes potentially involved in these processes. At this time then, the challenge to most plant biologists is to effectively mine this data to identify and characterize the genes and gene products that are critical to the crucial processes that have been investigated. This calls for techniques that start with a known DNA sequence and allow the determination of biological function. This approach is called reverse genetics and some of the most common methods for performing reverse genetic studies are based on RNA silencing.
Although recently discovered, RNA silencing is a wellcharacterized, endogenous system for monitoring RNA inside a cell and eliminating foreign molecules or inhibiting mRNA translation (for review, see Brodersen and Voinnet, 2006). It is a homology-based process that uses small RNA fragments to identify targets for destruction or inhibition. RNA silencing is also indispensable for normal plant growth and development, regulating the expression of central genes in flowering, meristem identity, and other processes (Meins et al., 2005). In plants, RNA silencing plays critical roles in viral defense, generating small interfering RNA sequences that target the foreign viral RNA for degradation. It is this capability of RNA silencing in plants that has been harnessed for use as a tool in reverse genetics. Virus-induced gene silencing (VIGS) is an efficient tool based on RNA silencing. VIGS uses viral vectors to introduce gene fragments homologous to a gene of interest into a plant host (for review, see Lu et al., 2003;Burch-Smith et al., 2004). Endogenous RNA silencing machinery then acts to eliminate viral RNA sequences and inadvertently also targets the mRNA of the gene of interest for removal. In this way, the gene of interest is silenced, through the removal of its mRNA.
VIGS has been widely adapted for functional gene studies (Lu et al., 2003;Burch-Smith et al., 2004;Robertson, 2004). It was originally described for use in tobacco (Nicotiana tabacum) and its relative Nicotiana benthamiana (Kumagai et al., 1995;Ruiz et al., 1998), but VIGS has now been adopted for gene analysis in many dicotyledonous species including Arabidopsis (Burch-Smith et al., 2006;Wang et al., 2006), potato (Solanum tuberosum; Brigneti et al., 2004), tomato (Liu et al., 2002a), and pepper (Capsicum annum; Chung et al., 2004). Even plant species that are not core eudicots have been analyzed with VIGS, including opium poppy (Papaver somniferum; Hileman et al., 2005), and Aquilega (Gould and Kramer, 2007). VIGS has also been developed for use in monocots and it has been described in barley (Hordeum vulgare; Holzberg et al., 2002;Hein et al., 2005), wheat (Triticum aestivum; Scofield et al., 2005), and rice (Ding et al., 2006(Ding et al., , 2007. VIGS has become a widely used technique for reverse genetics because it is transient and does not require the generation of stable transgenics. In addition, it is rapid with phenotypes observed with 10 d to 3 weeks of silencing. Another advantage of VIGS is that it can be designed to silence either a single member or multiple members of a gene family, addressing the problem of functional redundancy. The most widely used VIGS vectors are derived from the Tobacco rattle virus (TRV). It has been used for VIGS in the Solanaceae (Ratcliff et al., 2001;Liu et al., 2002aLiu et al., , 2002bBrigneti et al., 2004;Chung et al., 2004), Arabidopsis (Burch-Smith et al., 2006;Wang et al., 2006), opium poppy (Hileman et al., 2005), California poppy (Eschscolzia californica; Wege et al., 2007), and Aquilega (Gould and Kramer, 2007). TRV is a bipartite, singlestranded, positive sense RNA virus with a large host range (Hull, 2002). Some of the features of this virus that are believed to contribute to its success as a VIGS vector include its ability to enter growing parts of the plant, its toleration of DNA inserts up to about 1 kb in size, and mild symptoms of infection (Lu et al., 2003;Burch-Smith et al., 2004).
We have previously generated and described TRVbased VIGS vectors (Liu et al., 2002a(Liu et al., , 2002b. Our first generation TRV VIGS vector used restriction digestion and ligation for cloning of inserts (Liu et al., 2002b). We then developed a second-generation vector that carried GATEWAY recombination sites (Liu et al., 2002a). However, prohibitive features of this vector include time-consuming and complicated cloning procedures. In addition, the comparatively high cost of the enzymes used for GATEWAY-based cloning and proprietary issues are also a consideration for large-scale gene cloning and VIGS screening. To overcome these limitations, we have adopted modified ligationindependent cloning (LIC; Aslanidis and de Jong, 1990;Dieckman et al., 2002) as an approach for highthroughput cloning into the TRV VIGS vector. LIC is much faster and more accurate than other cloning strategies as only a single transformation is required. We used this new TRV-LIC vector to clone 400 tomato ESTs and we assessed its effectiveness in N. benthamiana. Thus, the TRV-LIC vector is an improved vector for high-throughput VIGS.
In addition, we observed several interesting phenotypes and identified N. benthamiana MADS-box genes with important, nonredundant roles in flowering. MADS-box genes represent a large family of transcription factors with critical roles in floral and general development. In particular, they function in conferring floral organ identity, floral determinancy, and the timing of the vegetative-floral transition (Ng and Yanofsky, 2001;Henderson and Dean, 2004). We used our TRV-LIC vector to silence NbMADS4-1 and NbMADS4-2 simultaneously or individually. In all instances we observed a loss of floral organ identity and determinancy, resulting in green, bushy flowers. These data suggest that NbMADS4-1 and -2, despite their high homology, have critical, nonredundant roles in floral development.

Development of a LIC Strategy for TRV-Based VIGS
One limitation of currently available VIGS vectors is the multiple steps involved in the cloning of the silencing fragment into the vector. This can become quite expensive when using GATEWAY-based recombination vectors. To overcome these limitations, we adopted a modified LIC strategy (Aslanidis and de Jong, 1990;Dieckman et al., 2002) for the high-throughput cloning of inserts into our TRV VIGS vector. To do this we added two LIC adaptors to the TRV2 vector, pYL170 (Burch-Smith et al., 2006), to generate pYY12. These adaptors consist of 15 bp sequences containing central PstI restriction sites (Fig. 1A). We then inserted a ccdB gene between the two LIC sites by PstI digestion to create pYY13, referred to as TRV2-LIC (Fig. 1A). The ccdB gene allows us to screen putative recombinant colonies rapidly and accurately, a strategy employed by the widely used GATEWAY technology. Digestion of TRV2-LIC with PstI linearizes the vector, and subsequent treatment with T4 DNA polymerase generates sticky ends. PCR products carrying ends homologous to the adaptors and similarly treated with T4 DNA polymerase can form covalent bonds with ends of the linearized vector (see ''Materials and Methods'' for details). Our new TRV2 vector, TRV2-LIC, can thus be used for high-throughput cloning since any number of sequences carrying compatible ends can be combined with TRV2-LIC.
We confirmed that the changes we had made to TRV RNA2 in generating TRV2-LIC did not compromise its infectivity by examining whether the virus was able to spread to the upper leaves of plants. We infected N. benthamiana plants with the TRV RNA1 and TRV2-LIC vectors only and collected upper leaves and inflorescence tissue 20 d later. The presence of TRV RNA2 in these tissues was confirmed by reverse transcription (RT)-PCR (Fig. 1C). Thus, the modifications made to TRV RNA2 to produce the TRV2-LIC plasmid does not hinder the ability of the virus to replicate and spread in plants. We also confirmed the ability of the new TRV-LIC vector to initiate gene silencing by inserting a fragment of the NbPHYTOENE DESATURASE (NbPDS) gene into this vector. We performed PDS silencing experiments according to standard protocols  and found that after 10 d, we observed photobleaching of upper leaves in presumably silenced plants (Fig. 1D). The timing and appearance of photobleaching was indistinguishable from that observed with earlier versions of our TRV VIGS vectors (data not shown). Thus, our TRV2-LIC vector functions in VIGS in a manner similar to other TRVbased VIGS vectors.

High-Throughput Cloning of Tomato ESTs
While N. benthamiana has become extremely popular for gene studies, the limited availability of sequence collections hinders the study of gene function. In contrast, large EST collections from tomato are readily accessible for large-scale gene function analyses (D'Agostino et al., 2007). Tomato and N. benthamiana are both members of the Solanaceae family and there is usually a high degree of similarity in gene coding sequences. Indeed, tomato gene fragments have been successfully used to silence homologs in N. benthamiana (Liu et al., 2004b) or vice versa (Ekengren et al., 2003). To demonstrate the efficacy of the TRV-LIC vector for high-throughput screening for gene function, we cloned 400 tomato ESTs into this vector (Supplemental Table S3; see ''Materials and Methods''). These ESTs predominantly belonged to one of the following three  classes: kinases, transcription factors, and phosphatases. The efficiency of this large-scale cloning approach was confirmed by the presence of inserts in the vector (representative gel is shown in Fig. 1E). The presence of the desired insert was detected in 100% of all clones examined.

Developmental Phenotypes Revealed by Silencing 400 Tomato ESTs
Our TRV-LIC EST clone collection was then screened for silencing phenotypes in N. benthamiana plants. For this, we infiltrated four-leaf stage plants with a 1:1 mixture of TRV1 and TRV2-LIC-EST fragment for each of the clones in our collection, and monitored the infiltrated plants for 8 weeks. Eventually, we observed several developmental phenotypes in our silenced plants, and some examples are shown (Fig. 2). For each EST, we infiltrated four plants per trial and did at least three trials.
One of the most common phenotypes resulting from silencing was lethality ( Fig. 2A). This is exemplified by the silencing of SlRibosomal protein L11 like, Sla-tubulin, SlHSP70, and SlPolyubiquitin ( Fig. 2A). Another phenotype we observed was the deformation of the upper, presumably silenced, leaves as shown for SlMAP3K epsilon (Fig. 2B). These plants were ultimately much shorter than control plants. Interestingly, SlMAP3K epsilon-silenced plants did not produce inflorescences, suggesting severe disruption of development beyond leaf structure (Fig. 2B, center section). The silencing of several constructs led to loss of pigmentation or chlo-rosis, as shown for silencing with SlGeranylgeranyl reductase (Fig. 2C) and SlPlastidic aldolase (Fig. 2D). The phenotypes that we observed in these two cases are in close agreement with previous reports on the effects of reduction of the levels of these gene products (Boldt et al., 1992;Tanaka et al., 1999).
Interestingly, silencing of some clones led to developmental defects in the meristems (Fig. 2E) or inflorescences (Fig. 2F) of plants. Silencing SlCyclin-dependent protein kinase, p34cdc2 (Fig. 2E), resulted in an attenuation of apical growth and severe stunting. The leaves of silenced plants also showed chlorosis. Silencing initiated by the tomato AGAMOUS homolog, SlAG1, produced phenotypes previously described for knock down in tomato (Pnueli et al., 1994). As in tomato, we observed the homeotic transformation of the third whorl stamens into petals, and the replacement of fourth whorl carpels with other flowers or inflorescences (Fig.  2F). These results reinforce our previous conclusion that TRV-mediated VIGS is an effective tool for assessing the functions of genes in the reproductive tissues of plants Liu et al., 2004a).
For 21 EST clones representing 18 unique genes that showed interesting developmental phenotypes when silenced, we attempted to clone their N. benthamiana homologs into the TRV2-LIC silencing vector (Supplemental Table S1). We failed to obtain N. benthamiana homologs corresponding to four tomato ESTs. We repeated the silencing using 14 N. benthamiana sequences and observed similar developmental phenotypes as   had been obtained with the tomato sequences (data not shown). We then determined the degree of silencing of each of the 14 genes by RT followed by semiquantitative RT-PCR using primers annealing outside the sequence used for silencing ( Fig. 2G; Supplemental Table S2; ''Materials and Methods''). In all cases we observed a greater than 60% reduction in transcript levels in the silenced plants. Thus, the approach used for the high-throughput cloning and silencing of tomato ESTs to assess their function using N. benthamiana as a heterologous system was successful.

Silencing SlMADS1 Produces Homeotic and Other Transformations in Flowers
Plants infiltrated with TRV carrying a fragment of SlMADS1 displayed very interesting developmental phenotypes, particularly in the floral inflorescences and individual flowers (Fig. 3). The altered development is first manifested as increased branching, first of the primary stem and then of the secondary stems, giving rise to a bushy plant compared to the control plants (Fig. 3A). Most of the branches of the SlMADS1-silenced Figure 5. Silencing NbMADS4-1 and NbMADS4-2. A, Control N. benthamiana flower. B and C, Silencing NbMADS4-1 using fragment 1 (B) or fragment 2 (C) causes enlargement of sepals and the loss of large white sepals. D and E, Silencing NbMADS4-2 causes identical phenotypes to silencing with either NbMADS4-1 fragment. F, The reduction of NbMADS4-1 transcript levels was confirmed by semiquantitative RT-PCR in tissue that had been silenced with fragment 1 (row 2) or fragment 2 (row 3). The levels of NbMADS4-2 mRNA were not altered in NbMADS4-1 fragment 1 silenced plants (row 4) and NbMADS4-1 fragment 2 silenced plants (row 5). The transcript levels of the related genes NbMADS5 and NbMADS11 were also examined to confirm specificity of silencing (rows 6 and 7, respectively). The phenotypes shown are not due to the silencing of another MADS-box transcription factor NbAG1 (row 8). Lanes to the left of the marker (M) contain control tissue and those on the right contain tissue from silenced tissue. C is no RT control. Numbers at the top of each lane indicate number of PCR cycles.
plants went on to produce inflorescences, such that the final number of flowers in the silenced plants is almost double that of the control plants ( Fig. 3A; data not shown). The timing of the floral transition did not differ between control and SlMADS1-silenced plants.
Wild-type N. benthamiana flowers typically possess a first whorl of five green sepals and a second whorl of five white petals that fuse to form a tube that surrounds the reproductive organs (Fig. 3B). The third whorl usually consists of four stamens that surround the central gynoecium of the fourth whorl (Fig. 2F, left  section). SlMADS1-silenced flowers carried enlarged sepals that were about 5-times bigger than the sepals of control flowers (Fig. 3, C and in D compare left and right sections). Further, petals and stamens converted into green leaf-like structures (Fig. 3, C-E). Interestingly, in some cases the fourth whorl carpel was replaced by another flower or even inflorescence (Fig. 3, D-G), resulting in indeterminate flowers.
We examined the ultrastructure of these modified petals and stamens by confocal microscopy. Wild-type petals and stamens normally do not possess stomata and trichomes on their abaxial surface (Fig. 4, left section). Both these structures were present in the leaflike structures of SlMADS1-silenced flowers (Fig. 4, right section). Thus, the cells on the leaf-like petals and stamens in SlMADS1-silenced plants looked very similar to pavement cells of leaf epidermis.

NbMADS4-1 and 4-2 Have Nonredundant, Critical Roles in Floral Development
To further characterize SlMADS1's role in plant development, we cloned the N. benthamiana sequences showing highest similarity to SlMADS1. A search of the GenBank database revealed that the tobacco sequence, NtMADS4, showed highest similarity (85%) to SlMADS1at the amino acid level (Supplemental Fig.  S1). Using RT-PCR with primers based on NtMADS4, we obtained two sequences from N. benthamiana that we named NbMADS4-1 and NbMADS4-2 (Supplemental Fig. S2). These two genes are 82% identical to each other at the nucleotide and amino acid levels. NbMADS4-1 and NbMADS4-2 show 83% and 97% identity to NtMADS4, respectively (Supplemental Fig. S1).
We then cloned two different fragments of each of these genes separately into the TRV2-LIC vector for silencing (Supplemental Figs. S3 and S4). The phenotypes obtained on silencing NbMADS4-1 and NbMADS4-2, each with either fragment, were identical to those observed on silencing SlMADS1 (Figs. 3 and 5). This suggests that NbMADS4-1 and NbMADS4-2 have closely related functions in flower development. To determine whether NbMADS4-1 and NbMADS4-2 were functionally redundant, we examined the mRNA levels of these genes in plants where silencing using the alternate gene had been initiated. In NbMADS4-1silenced plants, NbMADS4-1 RNA transcript levels decreased by about 90% (Fig. 5F), while levels of NbMADS4-2 mRNA were virtually unaffected (Fig. 5F).
Similarly, while NbMADS4-2 silencing produced a drastic reduction in NbMADS4-2 transcript levels, NbMADS4-1 was unaffected (Fig. 5F). We confirmed that the expression of other MADS-box genes was not affected when silencing NbMADS4-1 and -2 by monitoring levels of two closely related genes, NbMADS5 and NbMADS11 (Fig. 5F). We also confirmed that other genes involved in floral organ identity were not disrupted in NdMADS4-1 or -2-silenced tissue by determining that levels of NbAG1, the homolog of AtAGAMOUS, were unchanged (Fig. 5F). These data suggest that NbMADS4-1 and NbMADS4-2 have important, nonredundant roles in floral development in N. benthamiana.
Given the effects on floral development we observed, we examined the expression of NbMADS4-1 and NbMADS4-2 in different organs. Both genes were expressed in all whorls of floral organs (Fig. 6), while neither was highly expressed in any of the vegetative organs sampled (Fig. 6).

DISCUSSION
We report here the construction of a TRV-LIC vector that facilitates the high-throughput cloning of silencing fragments. This vector allows rapid, efficient cloning of inserts and has the added advantage of doing so at markedly reduced costs. While gene-specific primers are still required for each insert of interest, these are shorter than those used with other systems like GATEWAY, which have been adopted for cloning large numbers of inserts (Liu et al., 2002a). Further, there are fewer cloning steps and neither ligases nor recombinases are required, further reducing the tediousness and cost of the procedure. In addition, we have demonstrated that for the 400 tomato ESTs cloned here, cloning efficiency approaches 100%; thus, there is less need for screening colonies obtained after the cloning process to confirm the presence of inserts.
The TRV-LIC vector achieves silencing of target genes with the same timing and efficiency as previous Figure 6. Analysis of NbMADS4 tissue-specific expression. NbMADS4-1 and NbMADS4-2 transcripts were detected by semiquantitative RT-PCR. Lane 1 is tissue from sepals, 2 is petals, 3 is petioles, 4 is carpels, 5 is roots, 6 is stems, 7 is leaves, and 8 is seedlings. Lane M is the size marker. EF1a was used as a loading control.
TRV-based VIGS vectors. We observed greater than 60% reduction in target transcript levels in all cases examined. The TRV-LIC vector is also able to silence genes expressed in the growing parts of the plant as demonstrated by the flowering defects we observed in several instances (Figs. 2 and 3). This is consistent with other reports of VIGS of meristematic and flowering genes (Ratcliff et al., 2001;Chen et al., 2004;Liu et al., 2004a;Hileman et al., 2005;Gould and Kramer, 2007;Wege et al., 2007).
We have used the TRV-LIC vector to generate a collection based on 400 tomato ESTs. This collection includes kinases, phosphstases, and transcription factors. It should therefore prove to be a useful collection for researchers interested in investigating the function of members of these protein families. TRV-VIGS is effective in several Solanaceous species and our tomato collection could be used for VIGS in tomato, potato, pepper, or N. benthamiana. Here, we used tomato clones to silence several N. benthamiana genes. Based on the tomato sequence, we cloned the corresponding N. benthamiana sequence and examined whether the phenotypes we obtained with the tomato clone were an accurate representation of the endogenous gene function. In all cloned 14 homologs of N. benthamiana, we found that the results from tomato clones held true for the N. benthamiana sequence.
In the course of our studies with this silencing collection we observed a range of developmental phenotypes. Of particular interest was that generated by silencing SlMADS1. Sequence analysis shows that SlMADS1 belongs to the MADS transcription factor family. Members of this family have critical roles in flowering time and floral organ identity, in addition to roles in other important aspects of plant development (Ng and Yanofsky, 2001;Henderson and Dean, 2004). SlMADS1 belongs to the MIKC class of MADS transcription factors and its domain architecture consists of the MADS box (M) that is important for DNA binding and dimerization, the Intervening (I) domain for functional specialization, the Keratin-like (K) domain, and the C-terminal (C) domain for transcriptional activation (Kaufmann et al., 2005). The flowers of SlMADS1-silenced plants displayed enlarged sepals, the conversion of second-whorl petals to leaves in some cases, and the replacement of the fourth-whorl carpel with other flowers that also showed largerthan-normal sepals and petal conversion. Some of these phenotypes are reminiscent of the sepallata1 (sep1) sep2 sep3 triple mutant (Pelaz et al., 2000). However, SlMADS1-silenced plants also displayed a loss of apical dominance and the production of many secondary branches. SlMADS1 is most similar to NtMADS4, a member of the AGAMOUS-like2 subfamily of MADS transcription factors (Jang et al., 2002). NtMADS4 is known to have a role in flowering in tobacco, where ectopic overexpression leads to early flowering, although the flowers showed no defects in organization or tissue identity (Jang et al., 2002).
We cloned the N. benthamiana homolog of SlMADS1 and found two very similar sequences, NbMADS4-1 and NbMADS4-2. Both NbMADS4-1 and NbMADS4-2 show typical MIKC-type MADS transcription factor domain architecture (Supplemental Fig. S1) and they are highly similar to other MADS proteins, including Arabidopsis SEP proteins. Similar to NtMADS4 and other AGL2 family members, NbMADS4-1 and -2 were expressed almost exclusively in floral tissue (Jang et al., 2002). We used the TRV2-LIC vector to silence both NbMADS4-1 and NbMADS4-2, either simultaneously or individually. It should be noted that the sequences we used for specific silencing contained one or two regions of greater than 21 bp of homology between . Despite this, we were able to generate specific silencing, as shown by our RT-PCR analysis. One possible explanation for this is that insufficient small interfering RNAs were derived from these regions to initiate silencing of the homolog not being targeted.
Silencing NbMADS4-1 and NbMADS4-2, either simultaneously or separately, resulted in a loss of apical dominance and defects in floral organ identity. However, we did not observe a change in the timing of floral initiation. Interestingly, a loss of apical dominance was also observed when NtMADS4 was overexpressed (Jang et al., 2002). Together, these phenotypes point to an important role for NtMADS4/NbMADS4-1/2 in flowering. Interestingly, NbMADS4-1 and -2 genes are not redundant as specific silencing of either homolog produced all the phenotypes observed when SlMADS1 was silenced. This is quite different from the situation with the AtSEP genes, which are functionally redundant (Pelaz et al., 2000). Thus while NbMADS4-1 and NbMADS4-2 function in the same process, they appear to have independent, essential roles. MADS proteins are known to function as dimers (Kaufmann et al., 2005). We speculate the NbMADS4-1 and -2 may act together as a dimer and loss of either protein results in loss of activity.

Cloning Tomato EST VIGS Collection
Four hundred tomato (Solanum lycopersicum) ESTs were amplified with primers: 5#-CGACGACAAGACCCT-plasmid-specific sequence-3# and 5#-GAG-GAGAAGAGCCCT-plasmid-specific sequence-3#. The PCR products were purified with polyethylene glycol/MgCl 2 to remove any nonspecific PCR products and primers. A total of 50 ng of purified PCR product was treated with T4 DNA polymerase (New England Biolabs) in 13 reaction buffer containing 5 mM dATP and dithiothreitol at 22°C for 30 min followed by 20 min of inactivation of T4 DNA polymerase at 70°C. The TRV2-LIC vector was then digested with PstI and similarly treated with T4 DNA polymerase but dTTP replaced dATP. A total of 50 ng of treated PCR product and TRV2-LIC vector were mixed and incubated at 65°C for 2 min and then 22°C for 10 min. Then 6 mL of the mixture was transformed into Escherichia coli DH10B competent cells (Fig. 1B). Transformants were tested by PCR amplification using primers 5#-TGTTACTCAAGGAAGCACGATGAGCT-3# and 5#-GAG-GAGAAGAGCCCTGCCGCTCTAGAACTAGTGGATCC-3#. Plasmids from positive clones were purified and sequenced. To generate TRV2-LIC-NbPDS, NbPDS was amplified from Nicotiana benthamiana genomic DNA using primers 5#-CGACGACAAGACCCTCGGTCTAGAGGCACTCAACTTTATAAACC-3# and 5#-GAGGAGAAGAGCCCTTCCCTTCAGTTTTCTGTCAAACC-3#.

Plant Growth and Agroinfiltration
N. benthamiana plants were grown in pots at 25°C on light carts under continuous light. For VIGS TRV1 (Liu et al., 2002b), TRV2 or TRV2-LIC and its derivatives were introduced into Agrobacterium tumefaciens strain GV2260 by heat shock. For screening the TRV2-LIC tomato EST collection, 2 mL overnight cultures were grown at 28°C in the appropriate antibiotic selection medium in a 96-well culture box. The next day, cultures were spun down and cells were resuspended in infiltration medium (10 mM MES, 10 mM MgCl 2 , 200 mM acetosyringone), adjusted to OD 600 of 1, and incubated at room temperature for 3 h. N. benthamiana infiltration was performed as previously reported (Liu et al., 2002b). Infiltrated plants were observed for 8 weeks.

RNA Extraction and RT-PCR Analysis
Total RNA was extracted from leaves or flowers of wild-type N. benthamiana plants using RNeasy plant minikit (Qiagen). First-strand cDNA was synthesized using 1 mg of total RNA, gene-specific primers, and SuperScript reverse transcriptase (Invitrogen) according to the manufacturer's protocol. Primers are listed in Supplemental Table S1. RT-PCR products were cloned into TRV-LIC vector (pYY13) and performed VIGS in N. benthamiana. The expression levels of 14 genes whose silencing resulted in severe developmental phenotypes were monitored by semiquantitative RT-PCR using gene-specific primers that anneal outside the region targeted for silencing (Supplemental Table S2).

Microscopy
Tissue was vacuum infiltrated with 0.1 mg/mL propidium iodide and incubated for 30 min at 4°C in the dark. Confocal images were acquired on a Zeiss Axiovert 200 M light microscope equipped with a Zeiss LSM 510 NLO laser scanning microscope using a 103 C-Apochromat (NA 1.2) watercorrected objective lens. Scale bar is 50 mm.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. Alignment of amino acids of N. benthamiana MADS4, and their tobacco and tomato homologs.