To test whether it would be possible to exploit the mechanism of cap snatching for the mapping of begomoviral TSSs to their corresponding genome DNA templates, we first verified the occurrence of mixed infections of tenuiviruses and begomoviruses. To this end, we selected the rice stripe tenuivirus (RSV) as the cap-snatching sNSV (Falk and Tsai, 1998), as this virus was earlier shown to snatch capped RNA fragments from a co-infecting cucumber mosaic virus (CMV) in N. benthamiana and from a co-infecting rice ragged stunt virus (RRSV) in rice (Yao et al., 2012; Liu et al., 2016). For begomoviruses, two evolutionarily distant species, tomato yellow leaf curl virus (TYLCV) and ramie mosaic virus (RaMV), were selected (Navot et al., 1991; Li et al., 2010).

RSV-infected N. benthamiana were obtained by mechanical inoculation and those that became infected were next super-infected with a begomoviral infection by agro-infiltration of infectious clones. In all cases, we observed a mixed infection of RSV and the begomovirus of interest, although the time of appearance of the symptom varied (Figure 1A). Total RNA and DNA were extracted from upper leaves of these plants and the presence of RSV and the begomovirus in all mixed infections was confirmed using RT-PCR/PCR (Figure 1B).

Having observed that RSV was able to co-infect N. benthamiana with TYLCV and RaMV, we next verified whether RSV mRNAs contained 5′ capped leader sequences derived from the mRNAs of the two begomoviruses. Considering that RSV, besides begomoviral mRNAs, uses a large variety of host-cellular capped RNA leaders to prime production of its viral mRNAs (Liu et al., 2017), we aimed to collect RSV viral transcript sequences as much as possible. To this end, a high-throughput sequencing procedure as diagramed in Figure 2 was established and aimed at the 5′ ends of RSV noncapsid protein (NCP) transcripts, a mRNA that is being produced in relatively large amounts in RSV infected plant cells (Zhu et al., 1992). Two plants, one co-infected with RSV and RaMV (Sample 1) and the other one co-infected with RSV and TYLCV (Sample 2) were used for the sequencing. Another sample (Sample 3), infected with RSV only, was used as a (negative) control.

For each sample, about 1.2 million sequences were obtained. The sequences were compared with the genome of the co-infecting begomoviruses to extract NCP cDNA clones with a 5′ terminal capped leader snatched from the begomovirus. In total, 53 and 30 distinct NCP sequences with their 5′ termini matching the genome of RaMV and TYLCV, respectively, were collected. These sequences, along with their read numbers are presented in Table 1 and Supplemental Table 1. Collectively, RaMV- and TYLCV-matching sequences accounted for 2 and 0.5% of the total reads of Sample 1 and Sample 2, respectively. Most of these sequences were found exclusively in either Sample 1 or Sample 2, collected from leaf material containing a mixed-infection, and were absent from Sample 3, collected from singly, RSV-infected leaf material, indicating that these 5′ capped-RNA leader sequences were derived from transcripts of the co-infecting begomovirus (Table 1 and Supplemental Table 1).

According to a current model supported by an increasing number of studies, sNSVs snatch capped-RNA leader sequences from cellular mRNAs in a certain size range to prime viral genome transcription by base-pairing with the 3′ ultimate residues of the viral RNA template. In some cases, the transcription stalls and the elongated capped-RNA leader shifts backwards to realign on the 3′ ultimate residues of the viral RNA template, a process commonly known as prime-and-realign (Garcin et al., 1995; Liu et al., 2016). All the NCP transcript sequences in Table 1 and Supplemental Table 1 clearly fit this model. E.g., the first NCP sequence can be explained by a cleavage of a putative RaMV mRNA after the residue C₁₂. The resulting capped-RNA leader, 5′-G₁UAGGCGGCCC₁₁C₁₂, base-pairs by its 3′-ultimate C₁₂ residue with G₂ of the RSV NCP template (which has a 3′-U₁G₂U₃G₄U₅U₆UCAG- sequence at its 3′ terminus, see the bottom row of Table 1). After an extension of two residues according to the template, the extended capped-RNA leader, 5′-G₁UAGGCGGCCC₁₁C₁₂A₁₃C₁₄, is shifted back to pair with its A₁₂C₁₃ residues to U₁G₂ of the RSV template. After that, a processive elongation starting from U₃ gives rise to the NCP mRNA. The second sequence can be explained similarly, except that the putative mRNA was cleaved after C₁₁ and two cycles of priming and realignment occurred before processive transcription elongation resulting in the addition of a repetitive AC dinucleotide after the snatched capped-RNA leader. Other sequences can be explained in a similar way except that different RaMV mRNAs have been used as donors cap leaders.

Following this result, we next set out to map the 5′ ends of these putative begomoviral mRNAs to the genome of RaMV or TVLCV. As shown and summarized in Figure 3, mapping of the snatched RaMV and TYLCV 5′ capped-RNA leaders on their corresponding DNA genomes revealed transcriptional start sites (TSSs) of geminiviral genes that were largely in agreement with those reported from previous studies on related viruses. Notably, the heterogeneity of the TSSs identified here was considerably higher than that previously observed with primer extension experiments (Townsend et al., 1985; Petty et al., 1988; Hanley-Bowdoin et al., 1989; Sunter and Bisaro, 1989; Sunter et al., 1989; Mullineaux et al., 1993), and more or less comparable to (or slightly higher than) that identified recently with 5′ RACE experiments (Shivaprasad et al., 2005; Akbar et al., 2012). In addition, a putative TSS at nt 1,634 of RaMV was found. This TSS suggested the presence of a monocistronic AC3 mRNA (Figure 3A and Supplemental Table 1). This is somewhat surprising since AC3 is considered to be expressed from a polycistronic mRNA according to the results obtained previously for all begomoviruses except mungbean yellow mosaic virus (MYMV, Shivaprasad et al., 2005).

To confirm the 5′ termini of the begomoviral mRNAs obtained from the cap-snatching based cloning strategy, an RNA ligase mediated 5′ RACE method was applied to collect RSV transcript sequences and analyze their heterogeneous 5′ leader sequences (Liu and Gorovsky, 1993). The procedure of this method is similar to that shown in Figure 2. In brief, an adaptor was added to the 5′ termini of capped RNAs. The 5′ oligo-tagged RNAs were reverse transcribed with gene-specific primers and subsequently PCR amplified with a primer specific to the adaptor (forward primer) and two different reverse primers (the positions of which are indicated in Figure 3B) specific to virion- or complementary-sense transcripts of TYLCV, respectively. Both PCR reactions yielded multiple fragments of different sizes that were all purified and cloned for sequence analyses. As shown in Figures 4A,B, 12 sequences were obtained for each PCR reaction, and mapping them on the viral DNA genome of TYLCV pointed toward four TSSs for virion-sense transcripts, i.e., at nt 141, 143, 145, and 148 of the viral genome (Figure 4A), and also four for the complementary-sense transcripts, i.e., at nt 1,659, 1,661, 1,712, and 1,714, respectively (Figure 4B). Of the four virion-sense TSSs, three (at nt 143, 145, and 148, respectively) were found in the cap-snatching based cloning strategy (Figure 3B and Supplemental Table 1). For the RACE, 11 of the 12 (92%) cDNA clones pointed toward these three TSSs. For the cap-snatching based cloning strategy, 5,912 of the 7,326 (81%) expected reads (reads matching TYLCV genome regions upstream of the primer used in the RACE) pointed toward these three TSSs. Of the four complementary-sense TSSs, two (at nt 1,712 and 1,661, respectively) were identical to those identified by the cap-snatching based cloning strategy (Figure 3B and Supplemental Table 1). For the RACE, 9 of the 12 (75%) cDNA clones pointed toward these three TSSs. For the cap-snatching based cloning strategy, 1,304 of the 2,082 (63%) expected reads pointed toward these three TSSs. While one virion- and two complementary-sense TSSs were identified exclusively by RACE, two virion- and four complementary-sense TSSs were identified only by the cap-snatching based cloning strategy. These results altogether suggested that the cap-snatching based cloning strategy enabled the mapping and identification of not only most commonly used TSSs but also some rare ones.

A Jiangsu isolate of RSV (LS-JSJJ03, EF198702, EF198682, EF198733, EF198702) was used in this study. This virus was maintained in rice by transmission with viruliferous Laodelphax striatellus Fallen. The inoculation of RSV to N. benthamiana was done as described previously (Zheng et al., 2015). TYLCV and RaMV were isolated from Fujian and maintained with their infectious clones. The RSV infected N. benthamiana was super infected with a begomoviral infection by agro-infiltration of infectious clones (10 days after the inoculation of RSV). To do this, Agrobacterium tumefaciens strain EHA105 carrying the plasmid of interest were grown to an optical density at 600 nm (OD₆₀₀) of 0.8 at 28°C on LB liquid medium supplemented with rifampicin (50 μg/mL) and kanamycin (50 μg/mL). The cultures were centrifuged at 12,000 g for 1 min followed by resuspension with an induction medium [10 mM 2-(N-morpholino) ethanesulfonic acid (MES), pH 5.6, 10 mM MgCl₂ and 150 μM acetosyringone]. After virus inoculation, N. benthamiana plants were grown under standard greenhouse conditions (16-h day length; 22 to 24°C)

The high throughput sequencing was done with a procedure diagramed in Figure 2. Briefly, total RNA was extracted from N. benthamiana 15 days after inoculation of begomoviruses using Triziol (Invitrogen). After being dephosphorylated using alkaline phosphatase (NEB), the total RNA was treated with RppH (NEB), which removes the cap structure of mRNAs and leaves a monophosphate at their 5′ ends. T4 RNA Ligase 1 (NEB) was used to ligate a RNA oligo (TCTACrArGrUrCrCrGrArCrGrArUrC) to the 5′-monophosphate-bearing mRNAs. SuperScript® III First-Strand Synthesis System (Invotrogen) was used to reverse transcribe the oligo-tagged RNAs with a primer named NCPR1, which is specific to NCP of RSV. Two primers, one designed according to the RNA oligo (GTTCTACAGTCCGACGATC) and another specific to NCP but 5′ to NCPR1, were used to amplify the cDNA. After 30 (Sample 3) or 34 (Sample 1 and 2) cycles of amplification, the PCR products were resolved by agarose gel electrophoresis. PCR bands about 200 bp in size were recovered and sent to Biomarker Technologies (Beijing, China) for high throughput sequencing with a HiSeq 2500 platform.

A RNA ligase mediated RACE (RLM-RACE) was used to verify the 5′ termini of begomoviral mRNAs (Liu and Gorovsky, 1993). Total RNA extraction, dephosphorylation, cap-removing and RNA adaptor ligation were done as described above. Primers TYTAV2 and TYTAC2 were used to reverse transcribe putative virion- and complementary-sense transcripts of TYLCV, respectively (Figure 3B). Primers TYTAV1 and TYTAC1, which are inner to TYTAV1 and TYTAC2, respectively, on corresponding transcripts, were used to PCR amplify the cDNA obtained above, together with a forward primer designed according to the RNA adaptor. The PCR products were resolved by agarose gel electrophoresis and discrete bands were recovered, cloned and sequenced.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.