The AU-Rich RNA Recombination Hot Spot Sequence of Brome Mosaic Virus Is Functional in Tombusviruses: Implications for the Mechanism of RNA Recombination

Abstract

RNA recombination can be facilitated by recombination signals present in viral RNAs. Among such signals are short sequences with high AU contents that constitute recombination hot spots in Brome mosaic virus (BMV) and retroviruses. In this paper, we demonstrate that a defective interfering (DI) RNA, a model template associated with Tomato bushy stunt virus (TBSV), a tombusvirus, undergoes frequent recombination in plants and protoplast cells when it carries the AU-rich hot spot sequence from BMV. Similar to the situation with BMV, most of the recombination junction sites in the DI RNA recombinants were found within the AU-rich region. However, unlike BMV or retroviruses, where recombination usually occurred with precision between duplicated AU-rich sequences, the majority of TBSV DI RNA recombinants were imprecise. In addition, only one copy of the AU-rich sequence was essential to promote recombination in the DI RNA. The selection of junction sites was also influenced by a putative cis-acting element present in the DI RNA. We found that this RNA sequence bound to the TBSV replicase proteins more efficiently than did control nonviral sequences, suggesting that it might be involved in replicase “landing” during the template switching events. In summary, evidence is presented that a tombusvirus can use the recombination signal of BMV. This supports the idea that common AU-rich recombination signals might promote interviral recombination between unrelated viruses.

RNA recombination plays a major role during the evolution of plus-strand RNA viruses (1, 6, 10, 29, 35). New viruses or strains may emerge via recombination between different viral RNAs or between viral and host RNAs (29, 35). In addition to contributing to genetic variability, RNA recombination has also been proposed to function as a repair mechanism that can salvage damaged or mutated viral RNAs and use them to generate infectious viral RNAs (10, 20). RNA recombination also plays a role in the formation of defective interfering (DI) RNAs associated with many animal and plant viruses (30). The first hallmark feature of DI RNAs is that they are derived from the parent (helper) virus via sequence deletion(s). The second is that DI RNAs are dependent on the helper virus for their replication, survival, and/or spread. The best-known DI RNAs among plant viruses are those associated with tombusvirus infections (30).

The most popular model of RNA recombination is the template switching (copy choice) mechanism, which suggests that the viral RNA-dependent RNA polymerase (RdRp) switches templates during cRNA synthesis (8, 10, 20). After the jump from the donor to the acceptor RNA, the RdRp resumes RNA synthesis by using the nascent RNA (which has been made on the donor RNA) as a primer. On the basis of the role of base pairing between the acceptor RNA and the primer (nascent RNA), RNA recombination events are divided into three categories: base pairing dependent (similarity essential), base pairing assisted, and base pairing independent (similarity nonessential) (20). RNA recombination is probably a chance event; thus, each nucleotide in an RNA molecule may serve as a target for recombination. Most experimental data demonstrate that not all regions within an RNA are equally recombinogenic. The sequences that participate in RNA recombination at higher and lower frequencies are called hot and cold spots, respectively. Various models have been proposed to explain the occurrence of the observed hot and cold spots for different viruses (8, 10, 20).

Tombusviruses are single-component plus-stranded RNA viruses of plants, and they are known to support RNA recombination at high frequency (2, 30-33). Although the involvement of the genomic RNAs (gRNAs) in recombination is well documented for tombusviruses, DI RNAs associated with these viruses are the most popular templates in studies of RNA recombination. This is because (i) they are involved in RNA recombination with high frequencies (2, 30-33), (ii) they do not contribute essential protein factors to replication, and thus (iii) they have greater genetic plasticity than the viral gRNA. Importantly, recombination in DI RNAs is thought to occur by using the same mechanism as recombination involving the viral gRNA (30). The replication process of tombusviruses and the associated DI RNAs is carried out by the replicase complex, which includes two viral proteins and unknown host factors (17, 22, 28). The tombusvirus DI RNAs, such as Tomato bushy stunt virus (TBSV)-associated DI-73 (Fig. 1A), contain three or four noncontiguous genomic segments (30). The two or three sequence deletions leading to DI RNA formation are thought to be the consequence of viral replicase jumping on the genomic template, and the deletions may occur in a stepwise manner (30).

FIG. 1.

The AU-rich hot spot sequence of BMV promotes recombination in DI RNA of TBSV. (A) Schematic representation of the recombination vector and its derivatives. A 117-nt-long sequence from the 3′ end of dEGFP (termed FP), the 69-nt-long AU-rich sequence of BMV (12) (termed AU¹ or AU², depending on the location relative to the FP sequence), or a 152-nt-long region from the barstar gene (7) was inserted as shown between RI and RII of the prototypical DI-73 RNA of TBSV. The probe used in Northern blot assays and the primers for RT-PCR are depicted above the constructs. (B) Sequence and predicted structure of AU (shown in plus-strand orientation). Note that the AU sequence consists of two parts: (i) BMV-derived R′, which contains a 21-nt AU-rich region (underlined with a thick gray line), and (ii) the artificial highly AU-rich sequence termed AUs that is 25 nt long (encircled; note that the four nucleotides marked with asterisks are part of both the R′ and AUs regions). (C) Northern blot analysis of total RNA extracts obtained from protoplasts (zero passage). Protoplasts were electroporated with the gel-isolated DI RNAs shown, in the presence of the CNV gRNA. The parental DI RNAs are marked. (D) Ethidium bromide-stained gel of the total RNA extracts obtained from protoplasts (first passage) that were electroporated with the total RNA extract obtained from zero-passage protoplasts. The parental and recombinant DI RNAs are marked. Each experiment was based on 12 or more independent protoplast samples (4 representative samples are shown). (E) Northern blot analysis of the samples shown in panel D. An asterisk marks a unique recombinant (lane 6) that had the recombination site close to the 3′ end, away from the inserted sequences (thus, it could not be detected by RT-PCR). (F) RT-PCR analyses of samples shown in panel D. Lanes 5, 10, 15, and 20 contain the control RT-PCR products obtained with the same DI RNA transcripts that were used for electroporation. (G) Northern blot analysis of total RNA extracts obtained from uninoculated (systemically infected) N. benthamiana leaves 10 days after inoculation with the gel-isolated DI RNAs shown, in the presence of the CNV gRNA.

Since it is possible that various RNA viruses may utilize similar RNA recombination pathways or mechanisms, we wished to compare how different viruses recognize and use similar recombination-promoting signals. This is not only important from a mechanistic point of view, but it can also give practical observations about the possibility of interviral recombination, which could be facilitated by recognition of the same recombination-promoting signals by different viruses. To initiate these studies, in this paper, we analyzed the effect of a recombination-promoting signal (a short, AU-rich sequence), which is well defined for Brome mosaic virus (BMV) (12), on tombusvirus recombination. The model recombination template was based on a DI RNA associated with TBSV. We observed that the AU-rich sequence of BMV could indeed promote RNA recombination in the model tombusvirus. The data obtained support the idea that viruses belonging to different supergroups, such as BMV and TBSV, can recognize the same recombination-promoting signals. However, the distribution of recombination sites was different between BMV and TBSV recombinants. We propose that, in addition to the role of the AU-rich sequence, a putative cis-acting replication element might also affect the selection of the recombination sites in the tombusvirus DI RNA.

MATERIALS AND METHODS

Construction of cDNA clones for TBSV DI RNAs.

The DI-FP recombination vector was constructed by deleting the N-terminal segment from the dEGFP (BD Biosciences) open reading frame (ORF) by using NcoI and construct DI-73dEGFP (J. Pogany and P. D. Nagy, unpublished data), leaving only a 117-bp segment (termed FP; Fig. 1A) from the dEGFP ORF (Fig. 1A). The resulting DI-FP vector contains, in addition to the truncated FP sequence between RI and RII in DI-73 (33), the unique restriction sites XbaI, XhoI, and NcoI at the 5′ side of the FP sequence and MluI and BamHI sites at the 3′ side.

All constructs that contained AU, GC2, R′, or AUs sequences (Fig. 1A) were obtained by inserting the PCR-amplified (by using the primers listed in Table 1) AU sequence (including a segment from the 3′ end of BMV RNA3, between positions 197 and 242) (12), GC2 (11), R′ (13), or the artificial AUs (12) segments into DI-FP either at the 5′ side of the FP sequence (at the unique XbaI and NcoI restriction sites) or at the 3′ side (at the unique MluI and BamHI restriction sites). The RII deletion constructs (derivatives of DI-AU¹-FP-AU²) were obtained by amplifying the RII and RIII/IV sequences of DI-FP with the primer sets shown in Table 1, followed by treatment with BamHI and SalI, and cloning to the similarly treated DI-AU¹-FP-AU² or DI-FP (Fig. 1A) construct. All of the clones were sequenced to confirm the desired changes.

TABLE 1.

Primers used for PCR

Construct	Primer	Sequence	Template (reference)
DI-AU1-FP-AU2	114	GCAGTCTAGAGGACGCGTGTAGAGACCCTGTCCAG	AU1 (12)
	115	GCAGCTCGAGCCATGGATCCATAATTGATGG
DI-AU1-FP	114	GCAGTCTAGAGGACGCGTGTAGAGACCCTGTCCAG	AU1 (12)
	115	GCAGCTCGAGCCATGGATCCATAATTGATGG
DI-Bar	922	GGAGTCTAGATTGTCTGACCGGATGGGT	DI-73-BARS
	923	GGAGGGATCCGAAAGTATGATGGTGA
DI-FP-RII/Δ40	751	GGCGGATCCTTTGGACGTCTTTCC	DI-FP
	21	CCGCCATGGTCCATAATTGATGGA
DI-RII/Δ40	751	GGCGGATCCTTTGGACGTCTTTCC	DI-FP
	21	CCGCCATGGTCCATAATTGATGGA
DI-RII/Δ20	933	CCGGGATCCACAACCTACCAAAG	DI-FP
	21	CCGCCATGGTCCATAATTGATGGA
DI-RII/Δ60	934	CCGGGATCCCCGTTCAGGAAAGC	DI-FP
	21	CCGCCATGGTCCATAATTGATGGA
DI-RII/Δ100	935	CCGGGATCCCCGTTCAGGAAAGC	DI-FP
	21	CCGCCATGGTCCATAATTGATGGA
DI-AUs-FP-AU2	860	CGACTCTAGAGCTAATTTATAATTCCATCA	AU1 (12)
	753	CCGCCATGGTCCATAATTGATGGA
DI-R′/GC2-FP-AU2	752	CGACTCTAGACCCTGTCCAGGTAGG	GC2 (11)
	794	CCGCCATGGTCCGCCCGCCCTATTTG
DI-AU1-FP-R′/GC2	792	CGACACGCGTCCCTGTCCAGGTAGG	GC2 (11)
	798	CCGGGATCCTCCGCCCGCCCTATTTG
DI-R′-FP-AU2	752	CGACTCTAGACCCTGTCCAGGTAGG	AU1 (12)
	796	CCGCCATGGTAGCTTTTAACCTTAGCC
Competitor templates
#1	1077	CGCTGCCCATGTCTTGT	DI-AU1-FP
	1005	TAATACGACTCACTATAGGGACACATTGATCCTAGCA
#7	1009	TAATACGACTCACTATAGGGTCCACGGCTCGTGTGTA	DI-AU1-FP
	933	CCGGGATCCACAACCTACCAAAG
#18	1038	TCGTCTTATTGGACGACACCGACTTGGGTATGA	DI-AU1-FP
	14	GTAATACGACTCACTATAGGGTTCTCTGCTTTTACGAAG
Templates for primer extensions in vitro
#11	167	TCGTCTTATTGGACGAAGAAACGGGAAGCTCGC	DI-AU1-FP
	14	GTAATACGACTCACTATAGGGTTCTCTGCTTTTACGAAG
#16	1037	TCGTCTTATTGGACGAGGAGCCTTTGGACGTCT	DI-AU1-FP
	1009	TAATACGACTCACTATAGGGTCCACGGCTCGTGTGTA
#17	1037	TCGTCTTATTGGACGAGGAGCCTTTGGACGTCT	DI-AU1-FP
	14	GTAATACGACTCACTATAGGGTTCTCTGCTTTTACGAAG
#18	1038	TCGTCTTATTGGACGACACCGACTTGGGTATGA	DI-AU1-FP
	14	GTAATACGACTCACTATAGGGTTCTCTGCTTTTACGAAG
#20	1036	TCGTCTTATTGGACGACACGCTGCCCATGTCT	DI-AU1-FP
	14	GTAATACGACTCACTATAGGGTTCTCTGCTTTTACGAAG
#21	1035	TCGTCTTATTGGACGACAGGAGAGCGGGATGGA	DI-AU1-FP
	14	GTAATACGACTCACTATAGGGTTCTCTGCTTTTACGAAG

Construct DI-152Bar* was made by replacing the FP region in DI-FP (Fig. 1A) with the PCR-amplified 152-bp C-terminal segment of the barstar ORF (7) by using the unique XbaI and BamHI sites. The PCR primers used are presented in Table 1.

Preparation of CNV gRNA and DI RNA transcripts in vitro.

To generate the gRNA of Cucumber necrosis virus (CNV) and the DI RNA transcripts, we linearized pK2/M5p20STOP (27) and all of the DI RNA clones, respectively, with SmaI, followed by in vitro transcription with T7 RNA polymerase (18, 19). In vitro-generated transcripts of DI RNAs were purified from 1% agarose gel, phenol-chloroform extracted, precipitated in 95% ethanol, and washed three times with 70% ethanol to remove residual salts. The in vitro RNA transcripts were analyzed in 1% agarose gels and quantified with a UV spectrophotometer (Beckman).

For the in vitro RdRp experiments, RNA templates were obtained by in vitro transcription reaction with T7 RNA polymerase (see above and references 4 and 17). The templates and primers used for PCR are listed in Table 1. The unincorporated nucleotides were removed by phenol-chloroform extraction and repeated ammonium acetate-isopropanol precipitation (4, 17). The T7 transcription products were analyzed by 5% denaturing polyacrylamide gel electrophoresis (PAGE), and the amounts of RNA were measured with a UV spectrophotometer.

Preparation and electroporation of protoplasts.

Nicotiana benthamiana protoplasts were prepared as described before (24). Briefly, N. benthamiana callus was treated with 0.5 g of cellulysin and 0.1 g of macerase (Calbiochem) for 4.5 h in protoplast incubation medium (18) at 25°C and then washed twice with 0.5 M mannitol and once with electroporation buffer (10 mM HEPES, 10 mM NaCl, 120 mM KCl, 4 mM CaCl₂, 200 mM mannitol). For electroporation, we used 5 × 10⁵ protoplasts, 5 μg of CNV gRNA, and 1 μg of DI RNA. Electroporation was performed with a Gene Pulser II (Bio-Rad) at a voltage of 0.2 kV and a capacitance of 0.5 μF. After electroporation, the samples were left on ice for 30 min and then 1.8 ml of protoplast culture medium was added (18). Protoplasts were incubated in petri dishes (35 by 10 mm) in the dark for 24 to 48 h at 22°C.

N. benthamiana plants were inoculated with 3 μg of CNV gRNA and 1 μg of DI RNA (prepared as described above) by rub inoculation as described earlier (17). Plants were incubated for 10 days in a temperature-controlled (∼22°C) room.

Total RNA extraction from protoplasts and plants and RNA analysis.

Total RNA was extracted from protoplast by the phenol-chloroform method (18). Aliquots of total RNA were analyzed on 1.2% agarose gels. RNA samples were treated with formamide at 85°C before being loaded onto the gel. For Northern blot analysis, RNA was transferred to Hybond XL membrane (Amersham-Pharmacia) by electrotransfer and hybridized with DI-72-specific probes [RI(−)] (24). Hybridization with ³²P-labeled RNA probes was performed in ULTRAhyb hybridization buffer at 68°C under the recommended conditions (Ambion). The ³²P-labeled RNA probes were made in an in vitro transcription reaction with T7 RNA polymerase in the presence of [α-³²P]UTP and a PCR template obtained with primers #15 (5′-GTAATACGACTCACTATAGGGCATGTCGCTTGTTTGTTG-3′) and #20 (5′-GGAAATTCTCCAGGATTTCTC-3′) (24).

RT-PCR analysis, cloning, and sequencing of recombinant DI RNAs.

To obtain recombinant DI RNA clones, a reverse transcriptase (RT) reaction was done with primer #106 (5′-ACCTGGAAGCTTATGCCAGATTTACACTCATC-3′) and 2 μl of total RNA (16). This was followed by amplification by PCR with primers #106 and #380 (5′-GGACGAATTCCATAATTATTATCTTTAFTTG-3′).

The RT-PCR products were digested with EcoRI and HindIII, followed by gel isolation and ligation into a similarly treated pUC19 vector. The clones for sequencing were selected after restriction digestion with EcoRI and HindIII. The sequencing was done with a CEQ Cycle Sequencing Kit (Beckman) and primer #553 (5′-GTAAAACGACGGCCAGT-3′).

In vitro RNA binding studies.

The RNA probe in the gel shift experiments was RIII(−)/cPR11 (23), containing the 82-nucleotide (nt) minus-stranded RIII replication enhancer and the 11-nt minimal promoter for plus-strand synthesis (23). The labeling was done with [α-³²P]UTP and T7 RNA polymerase (26). Competitor RNAs (see Fig. 6) were prepared with T7 polymerase on PCR-amplified templates with primers described in Table 1. The gel shift experiments were performed as described by Rajendran and Nagy (26a). Briefly, various amounts of competitor RNAs (used in 5×, 15×, and 45× excesses over the constant amount of the [α-³²P]UTP-labeled RNA probe) were mixed with 1 μM recombinant p33 and p92 preparations in the presence of 50 mM Tris-HCl (pH 8.2), 10 mM MgCl₂, 10 mM dithiothreitol, 10% glycerol, 2.4 U of RNase inhibitor, and 100 ng of tRNA. After 10 min of preincubation at 25°C, ∼2 ng of [α-³²P]UTP-labeled RNA probe was added to each RNA-binding reaction mixture, which was further incubated for 25 min. The samples were then analyzed by electrophoresis on native 4% polyacrylamide gels run at 200 V for 60 min at 4°C in Tris-glycine buffer (25 mM Tris, 190 mM glycine, 5 mM EDTA, pH 8.5). Dried gels were analyzed with a PhosphorImager. The recombinant proteins were purified from Escherichia coli as described earlier (24, 26).

FIG. 6.

Efficient binding of RII(−) to the recombinant p33 and p92 replicase proteins of TBSV in vitro. (A) Schematic representation of the competitor RNAs used in gel mobility shift experiments. Note that the competitors represent negative-strand sequences. See Fig. 1A for further details. (B and C) Gel mobility shift assay showing the competition between the ³²P-labeled probe [i.e., the RIII(−) sequence fused to the cPR11 plus-strand initiation promoter (23)] and the unlabeled competitors shown. The unbound, free RNA probe and the shifted (bound) RNA-protein complexes are marked on the right. The competitor RNAs were used at 5×, 15×, and 45× excesses over the labeled probe. Quantification of the binding of the probe to p92 and p33 is shown below the gel image. Binding of the 92-nt-long ³²P-labeled probe to the recombinant p92 protein (1 μM) in panel B and to recombinant p33 (1 μM) in panel C in the absence of the competitor (lane 11) was taken as 100%. Each experiment was repeated three times. Note that lane 10 includes the probe in the absence of proteins and competitor RNAs.

RdRp assay.

The CNV RdRp preparations were obtained from systemically infected N. benthamiana leaves as described by Nagy and Pogany (17). The RdRp reaction mixtures contained ∼1 μg of RNA transcripts (the amounts of RNA templates were adjusted on the basis of their sizes to have similar molar amounts of templates in each reaction mixture). RdRp reactions were carried out as previously described (17). The reaction was terminated by adding 5 μl of 10% sodium dodecyl sulfate, followed by phenol-chloroform extraction and ammonium acetate-isopropanol precipitation (17). The RdRp products were analyzed by 5% denaturing PAGE in the presence of 8 M urea, followed by PhosphorImager analysis. The data for each sample were normalized on the basis of the amount of templated UTP incorporated (4, 17).

RESULTS

Rationale.

To test if different viruses could recognize the same recombination signal, we chose an AU-rich sequence, which is a well-defined recombination-promoting signal in BMV (12). This AU-rich sequence was tested in a recombination system based on DI RNA associated with TBSV, a tombusvirus. BMV and TBSV belong to different supergroups of viruses, and their RdRps are only distantly related. In addition, tombusviruses do not code for an RNA helicase-like protein, which was shown to participate in RNA recombination in BMV (15).

Previous works defined BMV-derived and artificial AU-rich sequences that served as recombination-promoting signals (Fig. 1) during BMV replication in vivo (12-14, 16). Frequent generation of recombinants was only observed when the two recombining RNA templates carried the same or similar AU-rich sequences. Because recombination sites in BMV were located within the AU-rich regions, it was proposed that the BMV replicase recognized the AU-rich signals during the recombination events (12-14, 16). We wanted to test if these AU-rich sequences could also promote RNA recombination in tombusviruses.

The control TBSV DI RNA-based recombination vector does not support RNA recombination.

To study RNA recombination in tombusviruses, first we developed a recombination vector, which facilitated the construction and testing of additional parental DI RNAs (Fig. 1A). The recombination vector, named DI-FP, is based on TBSV DI-73 RNA carrying a 117-nt insert of nonviral origin (derived from dEGFP [BD Biosciences]) between regions I and II (RI and RII; Fig. 1A). To avoid possible artifactual RNAs (i.e., unwanted recombinant-like RNAs) made during plasmid propagation in E. coli and/or during RNA transcription with T7 polymerase, we gel isolated DI-FP RNA transcripts prior to their use. In addition, to exclude those DI RNAs that might be generated spontaneously from the helper gRNA during the infection, a process known as de novo DI RNA formation (27, 30), we used the heterologous virus CNV (closely related to TBSV) as the helper virus. CNV can efficiently support the replication of TBSV DI-73 RNA and the DI-FP vector (24) (Fig. 1C).

The gel-isolated DI-FP transcripts were coelectroporated with CNV gRNA transcripts into N. benthamiana protoplasts as described earlier (24). We followed the accumulation of DI-FP RNA in protoplasts (termed zero passage; Fig. 1C) by Northern blotting with a DI RNA-specific probe [RI(−); Fig. 1A] and RT-PCR (primers #160 and #380; Fig. 1A). These experiments detected only the parental DI-FP RNAs in the zero-passage protoplast samples, suggesting a lack of recombinant accumulation (Fig. 1C). The total RNA obtained after 48 h of incubation in zero-passage protoplasts was used for electroporation into a new batch of protoplasts (termed first passage). Northern blot and RT-PCR analyses revealed that the DI-FP RNA recombination vector replicated efficiently and stably in the first-passage protoplasts (Fig. 1D to F, lanes 1 to 4). Note that we made sequential passages of the progeny DI RNAs from one batch of protoplasts to another, since there is no cell-to-cell spread of tombusviruses in protoplasts. Altogether, we performed three sequential passages, yet the DI-FP vector did not appear to support recombinant DI RNA accumulation (data not shown).

An AU-rich sequence supports RNA recombination efficiently in TBSV-associated DI RNA.

To test the effect of an AU-rich sequence on tombusvirus recombination, we inserted the 69-nt-long AU sequence (Fig. 1B; this sequence is identical to the AU1 sequence used in the study described in reference 12), which had been tested previously for recombination in BMV, into DI-FP in such a way that identical copies of the AU sequence flanked the dEGFP-derived FP region at the 5′ and 3′ sides, respectively (termed AU¹ and AU² in the construct DI-AU¹-FP-AU²; Fig. 1A), resulting in a repeated sequence. It was predicted, on the basis of the BMV results (12), that recombination might take place between the repeated copies of the AU sequence, thus resulting in deletion of one of the AU copies plus the FP sequence (Fig. 1A) located between the repeated AU sequences.

Incubation of the protoplast cells after coelectroporation of the gel-purified DI-AU¹-FP-AU² RNA with the CNV gRNA resulted in efficient amplification of the parental-size DI RNA, suggesting that the insertions did not debilitate DI RNA replication (Fig. 1C). No recombinants were detected in these protoplasts by Northern blotting, while RT-PCR analysis did show the occurrence of novel RNAs that were shorter than the input RNAs by ∼100 to 250 bp (not shown). This indicated that the putative recombinant DI RNAs were present in the zero-passage protoplasts at low levels that could only be detected by the more sensitive RT-PCR assay. However, a passage of the total RNA to a new batch of protoplasts, followed by incubation, resulted in recombinant-like DI RNAs in 100% (15 out of 15) of the experiments on the basis of total RNA (Fig. 1D, lanes 6 to 9), Northern blot (Fig. 1E), and RT-PCR (Fig. 1F) analyses. Interestingly, the sizes of the novel DI RNA recombinants were variable, suggesting that the recombination events were “imprecise” in nature. Indeed, cloning and sequencing of these putative recombinants confirmed that (i) they derived from DI-AU¹-FP-AU² RNA via deletions and (ii) the recombination junction sites were different in many of the recombinants (Fig. 2A). All of the 5′ deletion sites in the 15 recombinants sequenced were located within the AU sequence (i.e., AU¹), while the 3′ deletion sites were clustered mostly within RII, which flanks the inserted sequences in DI-AU¹-FP-AU² (Fig. 2A). Overall, we did not find the generation of precise (homologous) recombinants between the duplicated AU sequences, which were the most common recombinants in the BMV system (12).

FIG. 2.

Sequences of DI RNA recombinants around the junction sites. A representative number of recombinants obtained with construct DI-AU¹-FP-AU² (A and B) or with DI-AU¹-FP (C and D) were amplified by RT-PCR (see Fig. 1), cloned, and sequenced around the junction sites. The top sequence represents the sequence of the parental DI RNAs. The regions are shown schematically on the top. The underlined nucleotides represent marker mutations. Double slashes indicate sequences not shown, while the lengths of the sequences not shown are indicated by the values below the parental sequences. The recombinants are shown below the asterisk. The deleted sequences in the recombinants are marked by dotted lines, while nontemplated nucleotides are shown by small letters. +++ indicates the following unshown sequence: UGGACCGUCACCCUGCAGCCUGUGCUUCUG. Note that positive-strand sequences are shown in the 5′-to-3′ orientation. Two cDNA clones were sequenced per protoplast sample, but we counted only one recombinant per protoplast sample in the frequency column.

On the basis of the distribution of the recombination junctions, the AU² sequence did not seem to influence the selection of recombination sites in DI-AU¹-FP-AU². Therefore, we deleted AU² to generate construct DI-AU¹-FP (Fig. 1A). Testing of the recombination activity of DI-AU¹-FP, in a way similar to that described above for DI-AU¹-FP-AU², revealed that DI-AU¹-FP supported RNA recombination as efficiently as DI-AU¹-FP-AU² did (Fig. 1D to F, lanes 11 to 14). Even more important, the distribution of recombination sites for DI-AU¹-FP was comparable to that described for DI-AU¹-FP-AU² (Fig. 2C), confirming that the recombination hot spots include the AU¹ sequence and RII. This supports the idea that one copy of the AU sequence plays an important role in recombination, while the role of the second copy is less obvious.

The isolated DI RNA recombinants from protoplasts appear to be true recombinants, since they can be detected by gel analysis of the total RNA (Fig. 1D) and by Northern blotting (Fig. 1E). In addition, control RT-PCR assays performed on gel-isolated DI RNA transcripts (the same DI RNA transcripts that were used for electroporation to protoplasts) did not detect recombinant-sized DI RNAs for these constructs (Fig. 1F, lanes 5, 10, and 15).

To test if the length of the inserts is an important factor in recombination events, we generated a DI RNA carrying an ∼152-nt-long sequence from the barstar gene (Fig. 1A) (7). The resulting DI-152Bar* RNA was stable in protoplasts after one (Fig. 1D to F, lanes 16 to 19) or two (data not shown) passages. This result, together with that of DI-FP, supports the notion that DI RNAs with short inserts can be stable in protoplasts (under the conditions used) if they lack recombination-promoting signals.

To determine if recombination might occur in whole plants as well, we tested the accumulation of DI-FP, DI-AU¹-FP-AU², and DI-AU¹-FP RNAs (in the presence of the CNV gRNA) in N. benthamiana plants ∼10 days after inoculation with gel-purified transcripts. Northern blot (Fig. 1G) and RT-PCR (data not shown) analyses demonstrated that the systemically infected (uninoculated) leaves contained the parental-size DI RNAs for each DI RNA tested. In contrast, recombinant-like DI RNAs appeared only in DI-AU¹-FP-AU²- and DI-AU¹-FP RNA-containing plants (Fig. 1G, lanes 5 to 12) and not in DI-FP-containing plants (Fig. 1G, lanes 1 to 4). Cloning and sequencing of a representative number of recombinants confirmed that the accumulating recombinants in plants are similar to those observed in protoplasts, with most of the junction sites located within the 5′ AU¹ sequence and RII (Fig. 2B and D).

RII sequence affects the distribution of recombination sites.

Since many recombinants generated with DI-AU¹-FP-AU² and DI-AU¹-FP RNAs had the recombination sites within a 40-nt stretch close to the 5′ end of RII(+) (we term this region the hs40 hot spot), it is possible that sequences around hs40 might influence the selection of recombination sites. To test if hs40 is required for replication of DI RNA, we generated a series of DI constructs based on various 5′ deletions in RII. First, we tested construct DI-FP-RII/Δ40 (Fig. 3A), which has the same sequence as the DI-FP vector (Fig. 1A) but lacks the 5′ 40 nt from RII (i.e., hs40). DI-FP-RII/Δ40 RNA accumulated efficiently in protoplasts, and no recombinants were detected in protoplasts after the first passage (Fig. 3B and C, lanes 1 to 5). This suggests that the hs40 sequence is not essential for DI RNA accumulation in protoplasts.

FIG. 3.

Role of the RII sequence in DI RNA recombination. (A) Schematic representation of the parental constructs. See Fig. 1A for details. The deleted sequences are marked by broken lines. Northern blot (B) and RT-PCR (C) analyses of the total RNA extracts obtained from first-passage protoplasts are also shown. Lanes 6, 12, 18, 24, and 30 (marked with the letter T for transcripts) contain the control RT-PCR products obtained with the same DI RNA transcripts that were used for electroporation. For further details, see the legend to Fig. 1.

To test if hs40 is important for RNA recombination, we tested derivatives of DI-AU¹-FP-AU² with various 5′ deletions in RII. Construct DI-RII/Δ20, which lacked 20 nt from the 5′ end of RII(+) in DI-AU¹-FP-AU², generated recombinants as efficiently as DI-AU¹-FP-AU² did after the first passage (Fig. 3B and C, lanes 7 to 11). Interestingly, the distributions of recombination sites in the recombinants obtained with DI-AU¹-FP-AU² (Fig. 2A) and DI-RII/Δ20 (Fig. 4A) were similar, except that ∼50% (7 out of 15) of the recombinants obtained with DI-RII/Δ20 contained precise recombination sites between the duplicated AU sequences. The second construct tested, DI-RII/Δ40, which lacked 40 nt from the 5′ end of RII(+) in DI-AU¹-FP-AU², also generated recombinants in 100% of the samples (Fig. 3B and C, lanes 13 to 17). Interestingly, 87% (26 of 30) of the 3′ junctions were within the AU² copy and only 13% were within the RII sequence. We also observed precise recombinants between the repeated AU regions in ∼36% (11 of 30) of the recombinants (Fig. 4B). The occurrence of precise recombinants may suggest that primer realignment (base pairing between the primer and the acceptor region) assisted by the presence of duplicated sequences may take place during these recombination events (see Discussion).

FIG. 4.

Sequences of DI RNA recombinants obtained with RII deletion mutants. A representative number of recombinants (from first-passage protoplasts) were amplified by RT-PCR (see Fig. 3), cloned, and sequenced. The duplicated regions that participated in precise recombination are boxed. Note that because of sequence identity in the duplicated regions, the actual recombination sites cannot be determined in these precise recombinants. The actual sequence of the region marked with +++ is shown at the bottom of panel B. For further details, see the legend to Fig. 2.

Construct DI-RII/Δ60, which lacked 60 nt from the 5′ end of RII(+) in DI-AU¹-FP-AU², supported recombination as efficiently as DI-AU¹-FP-AU² (Fig. 3B and C, lanes 19 to 23). Cloning and sequencing revealed that all 15 recombinants, each derived from separate samples, were precise recombinants (Fig. 4C) with junctions between the duplicated AU sequences. In contrast to the above DI RNA constructs, DI-RII/Δ100, which lacked 100 nt from the 5′ end of RII(+), did not support the generation of recombinants in protoplasts (Fig. 3B and C, lanes 25 to 29). Moreover, DI-RII/Δ100 accumulated only inefficiently (tested after the first passage) in protoplasts, suggesting that an important cis-acting signal may have been deleted from this construct (see Discussion). Overall, these results suggest that the RII sequence may have two important effects on RNA recombination: first, it could affect the frequency of recombinant formation, and second, it could also influence the selection of recombination sites. In addition, we conclude that hs40 is not required for recombination to take place within the RII sequence.

Short, AU-rich sequences can also support RNA recombination in DI RNA.

To test if shorter, AU-rich sequences could also support RNA recombination in tombusviruses, first we deleted the BMV-derived R′ region (40 nt long; Fig. 1B) from the 69-nt AU¹ sequence in DI-RII/Δ40. The resulting construct, DI-AUs-FP-AU², contained only the 29-nt-long artificial AU-rich sequence (termed AUs, with 76% AU content), yet it still supported recombination efficiently with most of the junction sites located within the 5′ AUs and 3′ AU² regions (Fig. 5A). Note that we found that the distribution of the recombination sites was similar to that observed for DI-RII/Δ40 recombinants, with almost 50% (7 of 17) of the recombinants having precise junctions between the repeated AU-rich sequences (Fig. 5A).

FIG. 5.

Effects of short AU-rich and GC-rich sequences on DI RNA recombination. All of these constructs are derivatives of DI-RII/Δ40 (Fig. 3A). The AUs and R′ sequences are shown in Fig. 1B, while the complete GC2 sequence is shown in panels C and D (4, 11). A representative number of recombinants were amplified by RT-PCR (data not shown) from first-passage protoplast samples, cloned, and sequenced around the junction sites. The duplicated regions that participated in precise recombination are boxed. +++ indicates the following unshown sequence: 5′-GCCGGAGGUGGAGGAGCAGGAUGAU. For further details, see the legends to Fig. 2 and 4.

Second, we deleted the AUs portion from the 5′ copy of the AU sequence (Fig. 1B) in construct DI-R′-FP-AU² (Fig. 5B), which left the 44-nt-long BMV-derived R′ region at the 5′ location. Although the R′ sequence contained only a 21-nt AU-rich stretch (with ∼75% AU content; Fig. 1B), it was still active in recombination (Fig. 5B). Most of the recombinants obtained with DI-R′-FP-AU² (Fig. 5B) contained precise junctions between the repeated AU-rich sequences. Overall, this suggests that both the BMV-derived R′ and the artificial AUs portions of the original AU sequence are capable of supporting RNA recombination efficiently in TBSV. We conclude that an AU-rich sequence as short as 21 to 29 nt can promote recombination in TBSV DI RNA.

GC-rich sequences cannot “silence” RNA recombination in TBSV.

Since RNA recombination promoted by the AU sequence could efficiently be inhibited by the presence of GC-rich sequences located 3′ of the AU sequence (referred to here in the plus-strand orientation) in the BMV RNAs (11), we also tested the effect of one of these well-characterized sequences (termed GC2, with 55% GC content) (11) on DI RNA recombination. First, the GC2 sequence was inserted 3′ of the R′ sequence in DI-R′-FP-AU² (Fig. 5B). The resulting construct, DI-R′/GC2-FP-AU² (Fig. 5C), supported RNA recombination efficiently. Most of the junctions were located within the R′ sequence and the 3′ AU² sequence, including more than 60% (9 of 15) with precise junctions between the repeated AU-rich sequences (Fig. 5C). Second, placing the GC2 sequence behind the R′ sequence within the 3′ repeat (construct DI-AU¹-FP-R′/GC2) did not inhibit recombination that took place between the 5′ AU¹ sequence and either R′, GC2, or the RII sequences (Fig. 5C). Interestingly, none of the isolated recombinants contained precise junctions between the repeated AU-rich sequences, suggesting that the GC2 sequence interfered with precise recombination but not with imprecise recombination.

Minus-stranded RII binds efficiently to the TBSV RdRp.

Previous in vitro studies with the CNV RdRp (4) predicted that the same AU sequence tested as described above is a good donor, while it is a poor acceptor during template switching events. On the basis of these observations, it is possible that RNA recombination might occur during plus-strand synthesis when the RdRp would “jump” from the AU¹(−) region to the RII(−) region (around hs40) (see Discussion). Thus, this model predicts that RII(−) should contain a binding site(s) for the tombusvirus replicase proteins. In contrast, the heterologous dEGFP-derived FP sequence is expected to lack a high-affinity binding site(s), thus serving as a cold spot during the recombination events. This model was tested as described below in two different in vitro assays.

First, we have tested the ability of the purified recombinant TBSV replicase proteins (i.e., p33 and p92 expressed and purified from E. coli) (26a) to bind to RII(−) sequences or the FP sequence (also tested in the complementary orientation). To obtain quantitative results, we used template competition in a gel mobility shift assay (26a). The experiments included the same amount of ³²P-labeled template [derived from the RIII(−) replication enhancer] (23) and the same amount of purified recombinant p92 (Fig. 6B) or p33 (Fig. 6C) protein. The amounts of unlabeled competitors were used in 5×, 15×, and 45× excesses over the labeled RNA for each competitor RNA. These experiments revealed that the two competitor RNAs that contained overlapping portions of the RII(−) sequence (namely, construct #7 and #18, Fig. 6A) were two- to threefold better competitors for binding to p92 or p33 (Fig. 6B and C, lanes 4 to 9) than the FP control sequence (lanes 1 to 3). Therefore, this supports the model in which RII(−) is a hot spot region because of its increased binding to the replicase proteins (see Discussion).

Since binding to a template does not necessarily result in productive interaction between the particular sequence and the RdRp, we also tested the ability of the partially purified CNV RdRp to use templates containing various portions of the RII(−) sequence and of the FP(−) sequence in the presence of a short primer in an in vitro primer extension assay (Fig. 7) (4). As shown earlier, primer extension by the CNV RdRp in the in vitro assay depends on (i) the ability of the template region to interact with the CNV RdRp and (ii) the number of base pairs formed between the primer and the template (4). Since all of the constructs tested in this work contained the same 5-bp primer-template region (termed Art-5, which supported self-priming of RNA synthesis on the template by the CNV RdRp; see reference 4), we predicted that the activity of a particular template in the CNV RdRp reaction should depend on the ability of the template region to interact with the CNV RdRp. These experiments revealed that primer extension was about three- to ninefold more efficient when it started within the RII(−) sequence than the primer extension from the FP sequence (compare constructs 16, 17, and 18 with constructs 20 and 21 in Fig. 7A to C), while primer extension from the 3′ end of RII(−) was almost twofold more efficient (construct 11). Interestingly, the most efficient constructs were those that contained the primer within the hs40 hot spot region (see constructs 16 and 17) or in the middle of RII(−) (construct 17, Fig. 7), suggesting that this region contains a putative cis-acting element (see Discussion). Overall, data from the primer extension experiments do support the model in which RII(−) might be active in recombination owing to its enhanced binding to the RdRp.

FIG. 7.

Comparison of the level of primer extension obtained with RII(−)- and dEGFP-derived sequences. (A) Schematic representation of the constructs tested in an in vitro CNV RdRp assay. These templates contain the same artificial primer regions at the 3′ end of the templates that form 5-bp primers (termed Art-5, indicated by a solid arrow; see reference 4) and the acceptor regions shown. Note that the acceptor regions represent minus-strand sequences. (B) Representative 8 M urea PAGE analysis of CNV RdRp products with the series of templates shown in panel A. The primer extension products initiated from the Art-5 primers are marked with arrowheads. All of these RdRp products are partially RNase sensitive (data not shown). (C) Normalized (based on the amount of radiolabel incorporated) percent values of the levels of primer extension by the partially purified CNV RdRp compared to that of construct 21.

DISCUSSION

Comparison of the roles of AU-rich sequences in BMV, retrovirus, and tombusvirus recombination.

The emerging picture of RNA recombination is that the nonrandom distribution of RNA recombination sites observed with several plus-strand RNA viruses (reviewed in reference 20) is due to the presence of recombination-promoting signals in the RNA templates. The viral RdRp is postulated to recognize these recombination signals, which then leads to recombination (template switching) events with high frequencies (i.e., forming recombination hot spots) (20). It is unknown if recombination signals that have been described for a particular RNA virus can also be recognized by another RNA virus. We were particularly interested in AU-rich sequences, which not only are common in many RNA viruses but are known to promote recombination in BMV (12-14, 16) and retroviruses (5, 36) and possibly in poliovirus (25). Current models of the role of AU-rich sequences in promoting recombination are based on the assumption that the viral RdRp might pause during RNA synthesis within the AU-rich stretch of the template owing to the weak A-U base pairing between the growing nascent strand and the donor RNA template inside the RdRp (12-14, 20). Moreover, the weak base pairing between the nascent strand and the donor RNA within the AU-rich region might also facilitate dissociation of the 3′ end of the nascent strand from the complementary donor strand. The free 3′ end of the nascent strand may then anneal to the acceptor RNA at some frequencies, followed by resumption of the RNA synthesis by the viral RdRp. If the weak base pairing between the growing nascent strand and the donor RNA within the AU-rich region is indeed “forcing” RNA recombination, then we predict that different viral RdRps should recognize this type of signal, albeit with various frequencies, depending on the processivity of different viral RdRps.

Indeed, we found that an AU-rich sequence that promoted RNA recombination in BMV also facilitated RNA recombination in a TBSV-associated DI RNA. The supporting evidence includes the following: (i) the recombination sites were frequently located within the AU-rich region, and (ii) the recombination frequency of the DI RNAs containing AU-rich sequences was higher than that of the control DI RNAs carrying dEGFP-derived or barstar sequences (Fig. 1). We also obtained data that support a different role for the AU-rich sequence in tombusvirus recombination than in the case of BMV. These include the facts that (i) only one copy of the AU-rich sequence was enough for promoting tombusvirus recombination, while two copies were needed for efficient recombination in BMV; (ii) most of the recombination events occurred precisely or semiprecisely between the two copies of AU-rich sequences in BMV, while this type of recombinants was less frequent in the case of TBSV DI RNA; and (iii) an additional viral sequence (i.e., RII) was also involved in RNA recombination in TBSV DI RNA, while it is unknown if sequences other than the AU-rich regions are involved in facilitating the occurrence of recombination between the AU-rich sequences in BMV. Overall, the above similarities and differences between the two recombination systems suggest that the AU-rich sequences can promote recombination in both viruses, but the selection of recombination sites (which are probably determined during resolution of the putative recombination intermediates) is different for the majority of recombinants (mostly precise for BMV and largely imprecise for TBSV recombinants). This observation is also supported by the different effects of GC-rich sequences on recombination in the two viral systems. While GC-rich sequences located downstream on the acceptor template (on the basis of the progress of the viral replicase on the template) inhibited or silenced recombination in BMV (11), we did not observe similar effects of the same GC-rich sequence in the case of TBSV DI RNA (Fig. 5C and D).

In addition to the in vivo data discussed above, in vitro data obtained with the purified BMV (9) and CNV (4, 4a) replicases also support the role of AU-rich sequences when present in the donor RNA. For example, the end-to-end template switching assay developed by Kim and Kao (9) demonstrated that the AU-rich stretches when present at the 5′ end of the donor RNA, where the nascent strand must be released from the donor RNA before template switching occurs, facilitated end-to-end recombination events. Interestingly, the AU-rich sequence tested in this work also supported template switching by the partially purified CNV RdRp in vitro (4, 4a). The template RNA containing the AU-rich sequence served primarily as a donor RNA in the in vitro CNV RdRp assay (4, 4a), giving valuable insight into the significance of sequence context in RNA recombination in vitro, and possibly in vivo.

Further evidence of the possibility that AU-rich sequences could serve as common recombination signals comes from studies with retroviruses. With an in vitro template switching assay based on purified RT from Human immunodeficiency virus, it has been demonstrated that AU-rich sequences promoted RNA recombination (5, 36). Since the RT favors base pairing between the nascent strand and the acceptor strand prior to resumption of cDNA synthesis, most of the in vitro recombinants were precise (i.e., occurred precisely within homologous regions), although recombinants with extra nucleotides, mismatched nucleotides, or short deleted regions at the recombination sites were also isolated (5, 36). The observation that AU-rich sequences form hot spots in recombinants obtained with BMV, retrovirus, and TBSV suggests that the induction of recombination by AU-rich sequences is similar for these viruses. Moreover, the resolution of recombination intermediates may use somewhat similar mechanisms for BMV and retroviruses, but it is different in tombusviruses. We propose that the difference is due to two factors. (i) The BMV RdRp and the retrovirus RT favor a somewhat precise annealing step between the nascent strand (primer) and the acceptor strand prior to the resumption of RNA-DNA synthesis, while the tombusvirus RdRp does not seem to favor this step. Indeed, we have proposed that the CNV RdRp can easily resume RNA synthesis (primer extension) without the need for extensive base pairing between the primer and the acceptor strands (4). (ii) There might be differences among these viruses in the use of cis-acting elements to guide the jumping viral replicase to a new acceptor site before resumption of primer extension, resulting in recombination hot spots (see below). It is also important to note that we cannot completely rule out the possibility that selection for the best-fit recombinants also affects the types of recombinants isolated in these virus infections. In summary, in vitro and in vivo data suggest that AU-rich sequences might serve as common recombination signals. However, the likely differences in recombination between viruses are caused by (i) differences in template sequence recognition during the template switching event and (ii) various roles for base pairing between the primer and template during the template switching events.

Model of AU-rich sequence-driven recombination in tombusviruses.

Since both plus- and minus-strand synthesis takes place during DI RNA replication in protoplasts, it is difficult to establish whether plus- or minus-stranded RNAs are used as templates for recombination. However, on the basis of previous in vitro experiments with the partially purified CNV RdRp, the likely role of the AU-rich sequence is to promote recombination at the donor sites (4a). Moreover, the same AU sequence used in this study was found to be a relatively poor template (in comparison with known cis-acting elements of tombusviruses) in in vitro primer extension and template switching experiments with the CNV RdRp (4, 4a), suggesting that the AU sequence is unlikely to form a recombination hot spot as an acceptor site. If this is the case in protoplasts as well, then most of the recombination events might involve minus-stranded templates and thus, template switching would occur during plus-strand synthesis (Fig. 8A). This is because the first copy of the repeated AU regions (i.e., AU¹), which is observed as a hot spot with DI-AU¹-FP-AU² (Fig. 2A), would be at the 3′-proximal location (relative to the deletion junctions) in the minus-stranded DI-AU¹-FP-AU² RNA (Fig. 8A). Thus, the AU¹ repeat is favorably positioned to serve as a donor site for promoting jumping events by the tombusvirus replicase to a new location [which is hs40 within RII(−); see below]. The above model (Fig. 8A1 and A2) also predicts that the second AU repeat (i.e., AU²), located at a more 5′ position in the minus-stranded DI-AU¹-FP-AU² RNA, should be less favorable as a donor site, because this region could only be copied by the RdRp after the first AU¹ repeat had already been copied. Accordingly, the lack of an AU² repeat in DI-AU¹-FP RNA did not significantly alter the recombination sites or the frequency of the recombination events (Fig. 1 and 2). Interestingly, the AU sequences have also been predicted to support recombination during plus-strand synthesis when they were present in the minus-stranded BMV RNAs (12).

FIG. 8.

Models of AU-rich sequence-promoted RNA recombination in tombusviruses. Panel A1 shows the DI-AU¹-FP-AU² template (Fig. 1A) in the minus-strand orientation. hs40 is the recombination hot spot, while a triangle points to the putative cis-acting element in RII(−) that facilitates binding to the replicase proteins. The CNV RdRp (represented by an oval), while copying the minus-strand template in the 3′-to-5′ direction, is proposed to jump from the 3′ AU¹ sequence to the putative cis-acting element in RII(−), probably still carrying the primer nascent strand (solid line). (A2) After binding to RII(−), the RdRp resumes extension of the primer from the nearby hs40 sequence, thus resulting in imprecise recombinants. (B1) Alternative model explaining the possible formation of precise recombinants observed with DI-RII/Δ40 (shown schematically here) and its derivatives (Fig. 3 to 5). See panel A1 for further details. (B2) After binding to the putative cis-acting element in RII(−), the RdRp might resume extension of the primer from the nearby AU² sequence in the template. We propose that annealing between the primer (which is complementary to AU¹) and the AU² sequence in the template (represented by five short lines) prior to primer extension might facilitate the formation of precise recombinants.

In contrast to the proposed primary role in promoting replicase jumping, the AU-rich sequence may have no or only a limited role during RdRp landing (i.e., at the acceptor sites) if it is located “far” from the putative RII(−) cis-acting element. This model is supported by (i) primer extension experiments with CNV RdRp, which demonstrated that the AU sequence is a poor template (4), and (ii) the lack of effect on distribution of recombination sites in the absence of the AU² repeat (compare the data obtained with the DI-AU¹-FP-AU² and DI-AU¹-FP RNAs, Fig. 2A and C). Interestingly, the AU² sequence became a recombination hot spot when it was located in the vicinity of the putative RII(−) cis-acting element (such as in constructs DI-RII/Δ40 and DI-RII/Δ60; Fig. 4). To explain the formation of these precise recombinants, we propose that the RdRp might still bind to the RII(−) region during the jumping event, but annealing (base pairing) between the primer and the template might take place within the AU² repeat. This annealing step might then promote primer extension initiation from the AU² repeat, therefore resulting in precise or semiprecise recombination (Fig. 8B1 and B2). We propose that a similar annealing step between the primer and the acceptor region might be inhibited in the presence of extra sequences between AU² and RII(−) (see construct DI-AU¹-FP-AU², Fig. 1), thus favoring the occurrence of recombination within the hs40 region that flanks the putative cis element in RII(−) (Fig. 8A1 and A2).

A surprising observation in this work is the discovery of the major role of RII in TBSV recombination. The in vitro binding studies with the recombinant TBSV p33 and p92 replicase proteins and the primer extension studies with the CNV RdRp preparation suggest that there is a putative cis-acting element in RII(−) located 5′ of the hs40 sequence. Indeed, construct 18, which contained the 5′ half of RII(−) (between positions 100 and 239, Fig. 6 and 7), competed efficiently for binding to p33/p92 (Fig. 6) and was an efficient template for primer extension in the in vitro CNV RdRp assay (Fig. 7). We propose (Fig. 8A2) that the jumping viral replicase is likely guided by the RII(−) cis-acting element before resumption of primer extension, resulting in recombination hot spots. The role of this cis element is likely more important than the effect of base pairing between the primer and the template during tombusvirus recombination events. Accordingly, we observed earlier that the CNV RdRp could perform RNA synthesis (primer extension and template switching) without the need for extensive base pairing between the nascent strand and the acceptor strand (4, 4a). This observation can explain why there are not long stretches of sequence identity around the recombination sites generated in infections with DI-AU¹-FP-AU² (Fig. 2A).

The role of cis-acting replication elements in recombination has been proposed before in several viral systems (reviewed in reference 20). For example, a known replication enhancer element (19) in a satellite RNA (termed satC), which is associated with Turnip crinkle virus infections, promoted recombination between mutated satC and satD, another satellite RNA (3, 21). It has been proposed that the replication enhancer of satC is involved in binding to the jumping Turnip crinkle virus replicase (18, 19), a process similar to the role that was proposed above for RII(−) in TBSV DI RNA (Fig. 8). A different class of cis-acting element, namely, the subgenomic promoter region, has also been proposed to promote RNA recombination in luteoviruses (10a) and BMV (34). In summary, the discovery of the role of the RII(−) element in RNA recombination, in combination with published data cited above, suggests that cis-acting elements might play much wider roles in viral RNA recombination than previously anticipated.