For quantification purposes, a standard curve was generated by amplification of SACMV DNA-A in pBluescript (24) spiked into healthy tobacco total DNA, and the amount of replicating viral DNA per sample was calculated from the standard curve and expressed as molecules/ng DNA using the following equation, based on the assumption that one plasmid (DNA-A in pBS) has one copy or molecule of DNA-A, therefore if 100 ng plasmid DNA = 0.026 pmol plasmid, then using Avogadro's number, 1 pg plasmid DNA = 1.56 × 10⁵ molecules:

Molecules/ng DNA =[calculated amount of viral DNA (pg) × 1.56 × 10⁵ molecules/pg plasmid DNA]/[DNA added to PCR (ng)]

Data was analyzed using Microsoft Excel and GraphPad InStat 3.01 software (GraphPad Software, San Diego, CA, USA). Standard errors of the means were calculated for four bombardment events (28). Control included bombardment with pART7 vector (no viral insert).

Results and discussion

In this study, mutagenesis was used to disrupt base pairing in the IR in transgene DNA constructs to prevent formation of cruciform structures. Bisulfite-induced mutagenesis resulted in G:U mismatches in the hairpin RNA sequence (Figure 2B), leaving the antisense strand unmodified. We hypothesized that IR or hpRNA secondary structure would not be affected and would be more stable than nonmismatched IR constructs. G:U base pairs are unusually stable in comparison to other non-Watson and Crick base pairings; in fact hairpins closed by G:U base pairs are approximately 0.7 Cal/mol more stable than loops closed by G:C or C:G base pairs (31). It has previously been shown that such G:U mismatches have no impact on RNAi knockdown efficacy in animal cells (32). Moreover, Krol et al. (33) describe Dicer processing of hairpins generated by complementary annealing of CNG repeats, suggesting a tolerance for duplex dsRNA substrates with wide-ranging stabilities.

Figure 2. Properties of the SACMV BC1 mismatched inverted repeat construct. (Click to enlarge)

Since geminivirus transcripts are known to have strong IR fold-back structures (34), we examined the secondary structure of the SACMV BC1 transcript and identified a region consisting of unstructured linear stretches of RNA (Figure 2A) that would be highly accessible to the RNA-induced silencing complex (RISC), thereby allowing efficient degradation of target transcripts. Hairpin constructs were designed to produce a population of an estimated 10–12 different small interfering RNAs (siRNAs) with affinity to the target transcript. A 222-bp fragment of the BC1 ORF was amplified by PCR and separated into two pools. One PCR pool was treated with sodium bisulfite while the second pool remained untreated (Figure 1). Bisulfite-catalyzed deamination only affects cytosine residues, therefore products amplified from the positive strand will contain cytosine to thymine mutations when read in the sense orientation, while negative strand products will contain guanine to adenine mutations when read in the sense orientation. Since we are interested in engineering C to T mutations, which lead to G:U mismatches in the RNA hairpin, primers were designed to allow for strand-specific PCR amplification of the positive strand of the mutated template DNA. Primer design ensured that the 3'-terminal nucleotide of the forward primer would be mismatched if bound to a mutated template strand, whereas the 3'-end of the reverse primer would be unaffected by mutations on the template strand, ensuring more efficient amplification of the positive strand. In this way, a single amplification product is produced with C to T mutations when read in the sense orientation. A single DNA strand of desired polarity was preferenpreferentially amplified in 75% of the selected clones (data not shown). The antisense strand of the IR constructs was not mutated, as this may compromise the binding of subsequently formed siRNAs to target viral sequences.

In order to determine optimal exposure time to induce mutations in the sense arm, a range of methylation times was tested. Ten minutes exposure to bisulfite was found to catalyze an adequate mutation rate (C to T conversion of 56%). A final BC1 clone (BC1hp7) was selected with this C to T conversion rate (with a total of 17.5% mismatched base pairs). There was no notable difference in C to T conversion percentage introduced in the 222-nucleotide SACMV BC1 sequence between 5 (60%), 10 (56%), and 15 (55%) min (Table 1). A further study was conducted in our laboratory using the monopartite geminivirus Maize streak virus (MSV) and SACMV AC1 sequences, which highlighted the possibility of variation in conversion rate for different sequences. After 2.5 h, 80% of the cytosine residues in the MSV sequence (246 nucleotides) were converted to thymine, compared with all of the cytosines being mutated in the SACMV AC1 fragment (193 nucleotides) after the same exposure time (Table 1). It has previously been established that bisulfite reacts strongly with ssDNA and unpaired bases (35). Since the secondary structure of these viral sequences differ at 64°C (data not shown), we predict that the relative stabilities of secondary structures present at 64°C in a target DNA region will directly affect the degree of mutation by sodium bisulfite. While 2.5 h (manufacturer's recommendation) applies to screening for methylation in genomic DNA, our results indicate that for this application, shorter incubation times are adequate to introduce an acceptable number of mutations to stabilize the silencing construct during cloning; however, this still may have to be determined empirically, as well as the threshold number of mutations.