Assembly of CPSMV RNA1 and RNA2 infectious clones in the binary vector pAIDE. CPSMV infects both N. benthamiana and soybean. Thus, a CPSMV-based VIGS and gene expression vector could be easily propagated in agro-infiltrated N. benthamiana, permitting cheap and convenient amplification of inoculums that can be subsequently used to mechanically inoculate soybean. However, to develop a CPSMV-based VIGS and gene expression vector, we first needed to generate infectious clones for both genomic RNAs of CPSMV. To this end, we chose to clone the RNA1 and RNA2 cDNAs into pAIDE, a binary vector modified from pCambia1300 by our lab [20–23]. The modified pAIDE plasmid no longer has the cassette that confers hygromycin resistance in transgenic plants, but acquired a new empty cassette with the duplicated 35S promoter (2X35S) and the 35S terminator (T35S), both originated from cauliflower mosaic virus, flanking a multiple cloning site (Fig. 1B). The multiple cloning site permits the insertion of CPSMV cDNAs, which upon transcription in plant cells produce CPSMV RNAs.
It should be noted that the long poly-A tail at the 3’ ends of CPSMV RNA1 and 2 play critical role in the translation of viral proteins, as well as in viral genome replication, hence must be part of infectious cDNA clones. However, a reverse primer with a long stretch of T residues made long-range PCR very challenging. To address this problem, a stepwise strategy was devised to clone the CPSMV cDNAs. As shown in Fig. 1B, the first cloning step entailed the generation of two relatively short PCR fragments (< 1 kb) corresponding to two termini of the full-length viral cDNAs, bridged to each other with a unique restriction enzyme site (SmaI). As a result, the long poly-dT tract was relatively easily incorporated into the short 3’ fragment with a poly-dT-containing reverse primer (GA-CPSMR12-Rb. Figure 1B; Supplementary Table 1). The second cloning step then inserted the middle portion of the viral cDNAs. In the case of RNA1 cDNA, the middle portion was assembled with two overlapping fragments (Fig. 1B).
Despite initial setbacks, we eventually succeeded in obtaining the cDNA constructs for both RNA1 and RNA2 of CPSMV, with their sequences fully verified through Sanger sequencing. These constructs were named pAIDE-CPSMR1 and pAIDE-CPSMR2, respectively (Figs. 2 and 3). Notably, while pAIDE-CPSMR2 replicated robustly in both E. coli and A. tumefaciens, pAIDE-CPSMR1 replicated extremely slowly in E. coli, and failed to form any visible colonies when transformed into A. tumefaciens. These results suggested that similar to many other (+) RNA viruses , the cDNA of CPSMV RNA1 might direct the production of protein(s) or peptide(s) toxic to both bacteria, and that additional manipulations were needed to alleviate the toxicity of the construct containing RNA1 cDNA.
Stepwise attenuation of the toxicity of pAIDE-CPSMR1. Previous studies by others showed that plasmids harboring cDNAs of (+) RNA viruses were often toxic to hosting bacteria, but this toxicity could be neutralized by inserting introns into the protein-coding region(s) of these cDNAs [24–26]. This is likely due to the inability of prokaryotes such as E. coli or A. tumefaciens to process introns, as introns are only present in eukaryotic genes. To determine whether introns could be used to mitigate the toxicity of pAIDE-CPSMR1, we first attempted to map the regions in CPSMV RNA1 cDNA most likely to cause bacterial toxicity by generating a series of deletions within the RNA1 cDNA. The resulting constructs were tested in E.coli, using the time needed to produce visible colonies as an inverse indicator of toxicity. Figure 2A lists some of the deletions, with their effects on E. coli growth shown to the right. Deleting most of the RNA1 cDNA (ΔSpeI, from nt position 1,100 to the very 3’ end) completely abrogated the toxicity, restoring normal growth to E.coli (colonies became visible in less than 16 hours). Two smaller deletions, ΔAcc65I and ΔBsp1407I, between positions 1,314-3,069 (1,706 bp) and 2,710-4,220 (1,511 bp) respectively, restored normal E. coli growth as well. However, a deletion encompassing positions 4,321-5,423 (ΔBglII, 1,103 bp), as well as a small in-frame deletion spanning positions 4,210-4,320 (111 bp), failed to restore normal E. coli growth. Thus, the 1,314-to-4,220 region appears to be responsible for most of the bacterial toxicity conferred by CPSMV RNA1 cDNA.
We then began to insert introns into various positions of the RNA1 cDNA. Since RNA1 encodes a single ORF, it is possible that introns inserted near its 5’ end, by disrupting the translation of the entire polyprotein near its N-terminus, could exert a strong mitigating effect on its toxicity. We hence chose two locations near the 5’ end of RNA1 cDNA for initial intron insertion attempts. These locations are nt positions 327/328, and 604/605, both harboring the consensus splicing border motif AG/GT(T corresponds to U in RNA) (Fig. 2B). The introns a and b were identical, being derived from an 82-nt intron of AtPDS3. They were inserted into RNA1 cDNA separately, resulting in two constructs – INTa and INTb. Surprisingly, neither of them restored normal growth to transformed E. coli.
Given the failure of INTa and INTb, we next turned to the region deleted in ΔBsp1407I (nt positions 2,710-4,220; Fig. 2A). Two introns of 82 and 93 nt in size, both derived from AtPDS3, were then simultaneously inserted into two locations in RNA1 cDNA – 2,870/2,871 and 4,166/4,167 (Fig. 2B, Intron #1 and #2). The resulting construct was named as TIN (for two introns). Upon transforming into E. coli, TIN caused substantially faster bacterial growth, indicating a dramatic reduction of plasmid toxicity in E. coli (Fig. 2B). A. tumefaciens was also able to survive from TIN, but the colonies grew extremely slowly, becoming visible only after 4 days of growth at 28 ºC. This indicated to us that CPSMV RNA1 cDNA harbored additional region(s) that conferred specific toxicity to A. tumefaciens.
To further attenuate the toxicity of CPSMV RNA1 cDNA, we next created two more constructs, TRIN and QUIN, that contained three and four introns, respectively (Fig. 2B). Specifically, TRIN was a TIN derivative with an additional, 83-nt intron (Intron #3) at the position 3,386/3,387; whereas QUIN had a fourth intron at the position 2,023/2,024 (Fig. 2B). While TRIN showed a modest improvement over TIN in both E. coli and A tumefaciens, QUIN propagated in both bacteria to levels equivalent to insert-free pAIDE (Fig. 2B). More importantly, when paired with pAIDE-CPSMR2 and delivered into N. benthamiana with agro-infiltration, QUIN initiated systemic, symptomatic CPSMV infections in all plants (see below). Therefore, inserting four introns at multiple positions of CPSMV RNA1 cDNA caused a near complete neutralization of pAIDE-CPSMR1 toxicity in both E. coli and A. tumefaciens. As a result, QUIN was used in subsequent experiments as the provider of CPSMV RNA1.
A CPSMV-based VIGS vector induces robust silencing of PDS genes in N. benthamiana and soybean. To develop a VIGS vector using CPSMV cDNA clones, we next modified pAIDE-CPSMR2. Specific modifications included: (i) the protease processing site between MP and L-CP were duplicated by repeating the 20 aa residues downstream, and then 10 aa residues upstream, of the processing site (gray letters in Fig. 3A); (ii) the nt sequence of the duplicated aa residues were extensively codon-shuffled to minimize the nt-level identity while maintaining the aa sequence, hence reducing the chance of homologous recombination (see Supplementary Table 2, bottom row for sequence details); (iii) two Eco72I sites were inserted within the duplicated processing site to permit the cloning of nonviral sequences (orange letters, lines, and box in Fig. 3A); (iv) a 300 bp fragment of N. benthamiana PDS (NbPDS) was inserted between the Eco72I sites to permit the testing of VIGS in N. benthamiana. The resulting construct was designated FZ-NbPDS (the vector itself was named as FZ).
We then mixed A. tumefaciens strains containing QUIN (RNA1) and FZ-NbPDS (RNA2), and delivered them into young N. benthamiana leaves through agro-infiltration. FZ-NC, a construct containing a 399-nt non-plant, non-virus insert (partial coding sequence of mNeonGreen, a green fluorescent protein) was included as a negative control (Fig. 3B). As shown in Fig. 3C, N. benthamiana plants treated with FZ-NbPDS (plus QUIN) exhibited extensive photobleaching that was absent in FZ-NC (plus QUIN) plants by 21 days after agro-infiltration. Consistent with the photobleaching phenotype, the NbPDS mRNA levels as determined by semi-quantitative (sq) RT-PCR decreased substantially (Fig. 3D). Together these results demonstrated FZ as a robust VIGS vector in N. benthamiana.
We next tested the effectiveness of FZ as a VIGS vector in soybean. To do this, we inserted a 300 bp fragment of the GmPDS1 (Glycine max PDS1) gene into FZ, resulting in FZ-GmPDS. Since soybean is recalcitrant to agro-infiltration, the FZ-GmPDS (plus QUIN) was first propagated in agro-infiltrated N. benthamiana plants into the corresponding CPSMV derivative. The symptomatic N. benthamiana leaves harvested at 14 days after agro-infiltration were grounded and used to inoculate young soybean plants (accession Williams 82). Figure 3E shows soybean plants at 21 days after inoculation. FZ-GmPDS infection consistently caused widespread photobleaching that was absent in the control plants inoculated with FZ-NC. These results were confirmed with sqRT-PCR, showing a near complete loss of GmPDS1 mRNA (Fig. 3F). Therefore, the FZ VIGS vector was also highly efficient at silencing soybean genes.
PDS genes in multiple soybean accessions, as well as cowpea, are consistently silenced by FZ-GmPDS. A previous study  found that an ALSV-based VIGS vector was effective in only a few plant introduction (PI) lines of soybean (e.g. PI567301B), but ineffective in most cultivated accessions, such as Conrad, Sloan, and Thorne. To assess how effective the new FZ vector was in these soybean accessions, FZ-GmPDS was tested in four additional soybean lines: PI408097, Conrad, Sloan, and Thorne. Since the GmPDS insert in FZ-GmPDS shared more than 95% sequence identity with a cowpea PDS gene, we also included cowpea (Vigna unguiculata, variety California Black-Eye) in this set of experiments. As shown in Fig. 4, all of the plants inoculated with FZ-GmPDS, but none inoculated with the FZ-NC control, showed extensive photobleaching in top young leaves by three weeks post inoculation. Therefore, this CPSMV-based vector achieved robust VIGS in all soybean accessions tested, and was also active in cowpea.
FZ supports strong and lasting expression of fluorescent proteins in both N. benthamiana and soybean. We next assessed whether FZ was suitable for expressing nonviral proteins. The coding sequences of the green-fluorescent protein mNeonGreen, and the red-fluorescent protein mCherry, were cloned into FZ, leading to constructs FZ-mNeonGreen and FZ-mCherry. Note that these inserts were both 702 nt (Fig. 5A), approximately twice as large as the NbPDS, GmPDS, or NC inserts. These constructs, along with the QUIN construct encoding CPSMV RNA1, were transformed into Agrobacterium and introduced into N. benthamiana leaves with agro-infiltration. Signs of systemic infection became apparent in N. benthamiana plants 7–8 days later. Accordingly, both green and red fluorescence were observed in these systemic leaves at about the same time. Consistent with the stable retention of intact mNeonGreen and mCherry coding sequences, fluorescent protein expression persisted until at least 21 days after agro-infiltration (Fig. 5B). These results indicated that the FZ vector was a potent vector for expressing nonviral proteins in N. benthamiana.
For inoculating soybean plants, systemic N. benthamiana leaves were collected 14 days after agro-infiltration, homogenized, and mechanically applied to young soybean leaves. Consistently, the inoculated soybean plants began to show signs of systemic CPSMV infection 6–8 days post inoculation, and the systemic symptoms became easily observable in all inoculated soybean plants by 14 days. The systemically infected soybean leaves were also inspected with confocal microscopy and found to contain large green and red fluorescent leaf areas that were easily detectable (Fig. 5B, bottom) until at least 21 days post inoculation. Together these results demonstrated that the FZ vector was also a potent vector for expressing nonviral proteins in N. benthamiana and soybean.
MYB75 of Arabidopsis expressed from FZ is functionally active in N. benthamiana. To test whether nonviral proteins expressed using the FZ vector retained their original function, The coding sequence of the MYB75 transcription factor of Arabidopsis was cloned into FZ. MYB75, also known as PAP1 (for PRODUCTION OF ANTHOCYANIN PIGMENT 1), activates the expression of genes in the metabolic pathways that synthesize multiple phenylpropanoid compounds, among them anthocyanins [27–29]. Overaccumulation of anthocyanins causes plant leaves and flowers to display purple colors, providing an easily tractable phenotype for MYB75 overexpression. Indeed, a faint purple color began to appear at 5 days after agro-infiltration on the N. benthamiana leaves infiltrated with FZ-AtMYB75 (plus QUIN). More prominently, the top young leaves took an intensely purple appearance starting 14 days after agro-infiltration (Fig. 5C). Additionally, flowers that emerged from these plants also became intensely purple (instead of white as in control plants). Therefore, FZ-expressed AtMYB75 retained its function as an activator of anthocyanin synthesis in the heterologous N. benthamiana. Interestingly, the flower pedals, which in N. benthamiana plants are fused to form a long, straight tube, became bent and twisted in FZ-AtMYB75-infected plants. This phenotype has not been reported in Arabidopsis, thus may suggest additional novel functions for AtMYB75.