To enable the use of a sealed chamber biolistic apparatus, detached flower buds of Antirrhinum were used. We had previously determined that Antirrhinum buds could successfully develop in vitro, with buds 5–10 mm in length developing relatively normal pigmentation and expanding to open flowers, although the youngest buds did not always reach the normal size and developed more slowly.

To test transient, biolistic-based RNAi for determining gene function in flowers, two flower colour genes were targeted. The construct pPN187, based on pRNA69, was made for formation of hairpin RNA of an Antirrhinum chalcone synthase transgene (CHS). CHS has been used as a test gene in many gene suppression studies, and was one of the first targets of sense and antisense RNA experiments in plants 111228. CHS is one of the biosynthetic enzymes of the anthocyanin pathway (catalyzing the first step committed to the biosynthesis of all flavonoids), and inhibition of this step in anthocyanin-accumulating flowers results in the easily detected phenotype of white petals. Flower buds from a fully pigmented Antirrhinum line were picked when 3–10 mm in length (stages 1 and 2 as defined by Martin et al. 29), the sepals removed, and the outer epidermis of the exposed petals bombarded six times with pPN187. Buds at early developmental stages were used so that petal tissue was relatively un-pigmented when it was bombarded. As the buds subsequently developed in culture, notable differences in pigmentation were apparent in comparison to control buds, which were bombarded with pRT99GFP (Figure 1). While still at the unopened bud stage, extensive areas with markedly reduced pigmentation could be seen, with the chlorophyll of the young bud stage visible (Figure 1B and 1D). As the flowers developed, white patches appeared in the petals rather than the usual fully red colouration (Figure 1F, G and 1H), and these were seen in both the inner and outer epidermis, which are normally both pigmented in Antirrhinum. The areas lacking pigmentation did not have sharp boundaries, and mixed cell populations were apparent (Figure 1G and 1H).

The other colour gene targeted was Rosea1, which encodes a MYB-related transcription factor that regulates anthocyanin biosynthesis in Antirrhinum 30. The construct used was pRNA69-based pPN107, for formation of hairpin RNA of an Antirrhinum Rosea1 transgene. As with the CHS experiments, biolistic introduction of the plasmid into Antirrhinum buds resulted in the development of white, non-pigmented areas of the petals (Figure 2B). These did not occur in buds bombarded with the control construct (Figure 2A). By comparing the pattern of pigment loss between the inner and outer epidermis, it was clear that, at least in many cases, where pigment was absent in one epidermis, it was also absent from the corresponding cells in the other epidermis (Figure 2C and 2D). While cell division in the bud continues until the bud is approximately 10 mm long 31, the zones of inhibition observed across both epidermal surfaces are not due to cell division from the original transformed cells, as bombardment was at a stage past the formation of independent epidermal cell lineages. Therefore, the presence of the same pattern of gene inhibition in cells of both the inner and outer epidermis suggests that the RNAi inhibition signal is moving out from the original biolistically transformed cells and moving between cell layers. Such a finding is consistent with the mobility of the silencing signal and spreading of the silencing state that has been observed in other species and tissues 3233, and contrasts with the conclusions of Douchkov et al. 23, who assumed the method could be used to study only cell-autonomous traits. We do not know the extent to which the silencing signal was promulgated from an individual transformed cell, as the patterns of inhibition are most likely due to the merging of smaller zones of inhibition (due to the frequency of transformed cells that can be achieved in Antirrhinum petals using particle bombardment). The extent of inhibition observed means that it will be possible to use biolistics-based transient RNAi to produce significant amounts of transgenic tissue for subsequent analysis. With some genes of interest, it may be necessary or desirable to simultaneously inhibit expression of a marker gene to aid in the identification of areas of tissue for analysis. Chen et al. 34 used tandem constructs for CHS and an ACC oxidase gene in a viral vector to induce transient RNAi in a purple-flowered petunia. Silencing of CHS resulted in white flowers or flower sectors, and it was found that within these flowers or flower sectors transcript abundance from the target ACC oxidase gene (and a related ACC oxidase gene) was greatly reduced compared with abundance in purple tissue. Also, tandem inverted repeats within a vector to produce hairpin RNA for a selectable marker gene and a gene of interest were successful in triggering RNAi silencing of both genes in Chlamydomonas, thus allowing RNAi strains to be easily selected 35.

Although excised flower buds and a sealed chamber 'gene gun' were used in this study, it is assumed that the chamber-less guns would allow transient RNAi experiments with petals in situ.

Construction of hairpin gene constructs can be a rate-limiting step for high-throughput screens of gene function, and more recently developed constructs have used a hairpin of the transcript terminator region for easier construct building 24. To enable high-throughput construction of RNAi vectors, we made a new vector, pDAH1, that incorporates an antisense nos-sense nos hairpin and several other advantageous features (Figure 3). In particular; two XcmI restriction sites in the multiple cloning site (MCS) allow T-A cloning of the gene sequence of interest; NotI sites flank the promoter-hairpin sequence for cloning into pART27-based binary vectors 36; the promoter can be exchanged using the upstream PstI site and any of the sites in the MCS; and the antisense nos-sense nos hairpin is separated by SacI sites to allow it to be replaced if required. The spacer in pDAH1 is 97 bp, which is close to the minimum size that can be used to separate hairpin-producing sequences. To test the utility of pDAH1, the ORF of Antirrhinum CHS was PCR-amplified using Taq polymerase and directly ligated into the MCS using T-A cloning, resulting in pPN283. Biolistic-based RNAi was then carried out using the same method as used for pPN187 and pPN107. The pPN283 construct was effective in inhibiting pigment formation in Antirrhinum buds (Figure 4), with zones of silencing observed as white patches, similar to those seen after bombardment with the construct containing an inverted repeat of the CHS transgene (Figure 1).

Agroinfiltration for transient gene expression was tested in three species, Antirrhinum, petunia and tobacco, using the intron-GUS (IGUS) and GFP reporter genes. The intron in the GUS gene prevents Agrobacterium-derived GUS expression. Tobacco flowers developed normally after agroinfiltration of flowers in situ with 35S:IGUS, and GUS staining after 1.5 to 3 days clearly showed GUS expression throughout the epidermis of the petals (Figure 5). Infiltration was successful also for detached tobacco flowers (Figure 5C), and similar results were observed when using GFP as the reporter (Figure 6). Infiltrating detached petunia flowers with 35S:IGUS (Figure 5D) and 35S:GFP (data not shown) constructs showed similar results as for tobacco, with strong positive signal for samples ranging from 15 mm-length buds to fully opened flowers. The level of GFP detected in tobacco petals infiltrated with a 35S:GFP construct was visually similar to that of transgenic plants stably transformed with a 35S:GFP transgene (Figure 6D compared with 6E). Examinations based on petal sections revealed that the reporter genes were expressed in all cell layers of the agroinfiltrated petals (data not shown). GUS or GFP signal was not observed in any of the respective negative control plants (Figures 5A and 6F; data not shown).

Agroinfiltration (using strain LBA4404) of Antirrhinum petals was unsuccessful (data not shown). Both lack of Agrobacterium infection and induced necrosis in target tissues have been noted as problems when developing agroinfiltration protocols, and the identification of Agrobacterium strains more compatible with the host plant has been successful for some species 5. There are no previous reports on the use of floral tissue for transient assays with agroinfiltration, so it is not known whether lack of success with transient infiltration can be correlated to the infectability of species with Agrobacterium when generating stably transformed plants. Antirrhinum is readily infected by A. tumefaciens or A. rhizogenes 27; however, regeneration of plantlets from transformed tissues is difficult.

The 'Mitchell' petunia line (sometimes referred to as W115) was obtained from the University of Auckland, New Zealand, and is Petunia axillaris × (P. axillaris × P. hybrida) 37. Antirrhinum line 603 and wild type lines H75A and JI522 were obtained as seed from Prof Cathie Martin and Rosemary Carpenter of the John Innes Centre, Norwich (UK). The tobacco line used was N. tabacum cv Samsun.

The plasmid DAH1 (Figure 3) was derived from pGEM5Zf (In Vitro Technologies, Auckland, New Zealand). Linker1F and Linker1R (5'-NotI-PstI-MfeI-NotI-3') were inserted into the SphI/NsiI site of pGEM5Zf, destroying the SphI and NsiI sites, to produce pAF. A PstI-EcoRI fragment containing the CaMV 35S promoter from pART7 36, Linker2F and Linker2R (BamHI/SacI) and the nos terminator were then ligated into the PstI/MfeI sites of pAF to produce pAO. A second nos terminator fragment (with 97 bp of the 3'-end of the GFP sequence) was PCR-amplified with the primers BglII-SGFP-NOS and XbaI-EcoRI-ASNOS, digested with BglII and XbaI and cloned in the antisense orientation into the XbaI/BamHI sites of pAO to make the nos hairpin vector pAP. A multiple cloning site was then created by inserting Linker3F (Figure 3A, top strand) and Linker3R (Figure 3A, lower strand) into pAP digested with XbaI and EcoRI (destroying the upstream XbaI site) to produce pDAH1. The oligonucleotide sequences used were:

Linker1F 5'-AGCGGCCGCCTGCAGACGGACAATTGGCGGCCGCCTGCA-3'

Linker1R 5'-GGCGGCCGCCAATTGTCCGTCTGCAGGCGGCCGCTCATG-3'

Linker2F 5'-CCTGAG-3'

Linker2R 5'-GATCCTCAGGAGCT-3'

Primer BglII-SGFP-NOS 5'-CGCAGATCTCCACATGGTCCTTCTTGA-3'

Primer XbaI-EcoRI-ASNOS

5'-TGCTCTAGAACGAATTCCCGATCTAGTAACATA-3'

The binary vectors used were pBINm-gfp5-ER 3839 and p27IGUS, a vector based on pART27 but containing a GUS reporter gene with an intron (IGUS; 40). IGUS was PCR-amplified from pMOG410 and cloned into pART7 which had been digested with KpnI and SmaI. The cassette containing 35S:IGUS:OCS was released by NotI digestion and ligated into NotI-digested pART27.

cDNA encoding the ORF of CHS was PCR-amplified from a pool of first strand cDNA derived from floral RNA (isolated from Antirrhinum wild type line H75A) using FastStart Taq polymerase (Roche Applied Science) and gene-specific primers. pPN187 was constructed based on the vector pRNA69 41, containing the CaMV 35S promoter, multiple cloning sites (separated by the Yabby5 intron) for inserting sense and antisense sequences for the gene of interest, and the ocs terminator. The CHS ORF was ligated in a sense orientation into the XhoI site and in an antisense orientation using ClaI/XbaI sites. pPN283 was constructed by ligating the PCR-amplified CHS ORF into pDAH1 using T-A cloning. pPN107 was made by ligating the ORF of Rosea1 into pRNA69 in a sense orientation using the XhoI/BclI sites and in an antisense orientation using the ClaI/XbaI sites.

Whole buds (3–10 mm in length; minus sepals) were surface sterilized for 10 minutes using 10% (v/v) bleach containing 1–2 drops of Tween20/100 mL. Buds were then rinsed three times with sterile water and maintained on medium #2 (1/2 × MS macro salts/L; 1 × MS micro salt/L; 1 × MS iron/L; 1 × LS vitamins/L; 3% sucrose (w/v)/7.5% agar (w/v) during and after particle bombardment. The cultured buds were placed under artificial lights (16 h photoperiod) at 25°C after bombardment.

Particle bombardment used a helium-driven particle inflow gun based on Vain et al. 42, but modified by the addition of a high speed, direct current solenoid valve for accurate valve opening times down to 8 ms. The bombardment conditions were a solenoid valve opening time of 30 ms, a pressure setting of 400 kPa, a shooting distance of 13 cm, and a partial vacuum of approximately -95 kPa. Preparation of the DNA/gold suspension was essentially as in Schwinn et al. 30, with the gold in 50 μL water prior to precipitation of plasmid DNA onto the gold particles. A final DNA concentration (for each construct of interest) of 2 μg DNA per mg of 1.0 μm gold particles was used. Each bombardment used 5 μL of DNA/gold suspension. Buds were bombarded six times and then cultured. Transformation was monitored by including an internal control vector, pRT99GFP, which was co-precipitated onto the gold particles with the construct of interest (at one-fifth the concentration). Also, pRT99GFP alone (at the same concentration) was used for control bombardment experiments.

Attached tobacco flowers were infiltrated with A. tumefaciens strain LBA4404 harbouring either p27IGUS or pBINm-gfp5-ER. The LBA4404 cells were cultured in 10 ml LB broth with antibiotic overnight, pelleted and re-suspended in medium #1003 (AB media salts + NaH₂PO₄ 240 mg/L+ glucose 10 g/L + MES 14.693 g/L) supplemented with 100 μM acetosyringone, and cultured for 4 h. The cells were then pelleted and re-suspended to a concentration of A₆₀₀ = 0.5 in 1% (w/v) glucose solution (pH 5.3) supplemented with 100 μM acetosyringon. Flower buds or opened flowers were pierced with a needle and infiltrated with the A. tumefaciens culture using a syringe. When using detached flowers, the flowers were cut into half across the middle of the tube, agroinfiltrated using a vacuum chamber, blotted with Whatman paper and cultured in petri-dishes containing moistened Whatman paper. The agroinfiltrated, detached flowers were cultured at 25°C under artificial lights (16 h photoperiod) for 2 to 2.5 days before examining reporter gene activity.

1.5 to 3 days following agroinfiltration, flowers were histochemically assayed for GUS activity. Flower samples were incubated for 12 to 48 h at 37°C in X-gluc staining buffer (5-bromo-4-chloro-3-indoyl-β-D glucuronide dissolved in dimethyl formamide then diluted to 0.5 mg/L X-gluc in 50 mM phosphate buffer (pH 7.0) containing 1% (v/v) Triton X-100), and then placed in 70% (v/v) ethanol to remove the chlorophyll and preserve the sample. In some instances a brown colour developed due to tissue necrosis, and this was removed by treatment in a 5% acetic acid/ethanol (v/v) solution at 70°C for 30 min. To enable penetration of the GUS substrate into the petals they were either cut into pieces or wounds were made in the petals.

Light microscopy used an Olympus BH2 microscope and fluorescent microscopy used an Olympus SZX microscope. Images were recorded using a Leica DC 50 digital camera.