INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

It is generally accepted that RNA recombination contributes significantly to the diversity of viruses with RNA genomes (37). However, little experimental evidence supports the occurrence of high-frequency recombination in the virus life cycle. The processes of RNA replication and RNA recombination have been studied extensively in brome mosaic bromovirus (BMV), a tripartite positive-strand RNA virus (12). BMV RNA1 and RNA2 code, respectively, for the 1a and 2a proteins (the viral components of the replicase complex), while RNA3 encodes the 3a (movement) and coat proteins (2). Both homologous and nonhomologous recombination events have been observed among different BMV RNAs (24). Homology-supported crossovers can occur between two nearly identical RNAs (or within nearly identical regions), while nonhomologous crosses can occur between nonrelated RNAs or dissimilar regions (8, 16, 29). The frequency of homologous intersegmental crosses in BMV is approximately 10-fold higher than that of the nonhomologous crosses (24). In addition to BMV, homologous RNA recombination has been demonstrated for picornaviruses (18-20, 35), coronaviruses (21, 23, 42), for cowpea chlorotic mottle bromovirus (3), tombusviruses (41), and bacteriophages (31). Homology-driven recombination of non-replicative RNA precursors has been reported for Sindbis virus within the overlapping sequences (34).

Homologous crossovers among different BMV RNA segments appear to require common 15- to 60-nucleotide (nt) sequences (26) which are composed of GC-rich regions followed by AU-rich regions (27, 28). A proposed template-switching mechanism (27, 28) predicts that the replicase enzyme pauses (stalls) at the AU-rich sequence on a donor BMV RNA molecule and switches to the acceptor template while the upstream GC-rich region facilitates the hybridization of the nascent RNA (27, 28).

There is little information about homologous recombination among a population of viral RNA molecules during virus replication, because most of the recombinants may not differ from the parental RNA. The crossovers in poliovirus RNAs were observed at various locations within the 190-nt region between two selectable marker mutations (20, 35). No striking sequence specificity of crossover sites was found, but the crossovers tended to occur within the potential inter- and intramolecular double-stranded (heteroduplex) regions (1). Homologous crossovers were also observed among genomic RNAs of strains of the coronavirus mouse hepatitis virus (22, 40, 42). The RNA crosses have occurred at apparent hot spots (6), but later data revealed that some of the hot spots resulted from selection of mouse hepatitis virus variants (5).

To investigate the contribution of homologous recombination in the RNA virus life cycle, we have studied in vivo the crossovers between molecules of BMV RNA3. Pairs of RNA3, carrying marker mutations at distant locations, were used to measure the frequency of recombination in Chenopodium quinoa, a BMV local-lesion host. The homologous RNA3-RNA3 recombinants accumulated at high frequency in local lesions that were doubly infected with different combinations of the RNA3 variants. The frequency of crossovers between one pair of RNA3 variants which possessed the closely located markers was similar to that of another pair of RNA3 variants with more distant markers, suggesting the existence of an internal recombination hot spot. We discuss the implications of these results in terms of the mechanism of RNA recombination, viral RNA genetics, and evolution.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Materials. Plasmids pB1TP3, pB2TP5, and pB3TP7 (17) were used as templates to synthesize in vitro the infectious capped transcripts of wild-type (wt) BMV RNA1, RNA2, and RNA3, by using the MEGAscript T7 kit (Ambion, Austin, Tex.). Plasmids SF23, SF25, SF29, and SF30 (see the next paragraph) were used to synthesize the RNA3 mutants in the in vitro transcription reactions. The Moloney murine leukemia virus reverse transcriptase (RT) and the restriction enzymes were from Promega Corp. (Madison, Wis.).

Generation of RNA3 mutants. The four BMV RNA3 mutants SF23, SF25, SF29, and SF30 were derived from the wt cDNA clone of RNA3, pB3TP7 (17). Each of the four mutated RNA3 components was obtained using different procedures. Mutant SF23 was created by site-directed mutagenesis, during which the nucleotide U-121 was replaced with an A. This mutation created a BamHI restriction site and was translationally silent. For mutant SF25, two nucleotides (AU) were introduced in the 3' untranslated region (position 1862) in the RNA3 molecule, and this resulted in a new restriction site (BamHI). The RNA3 mutant SF29 (also named BX4) was selected from infected Chenopodium hybridum leaves (as described in reference 9). The use of C. hybridum plants in BMV recombination was originally reported by Rao et al. (33). SF29 has a 12-nt [(GAUC)₃] insertion at position 860, resulting in suppression of the unique BclI restriction site and addition of 4 amino acids (Pro-Ser-Ile-Asp) between the residues Asp-257 and Gln-258. Mutant SF30 (previously named D1 [13]) was selected as a pseudorevertant from a single local lesion on C. hybridum after inoculation with a frameshift mutant D (9). Mutant SF30 contains three additional U residues inserted at two separate locations, resulting in replacement of Trp-22-Thr-23 with Phe-Ser-Gly in the coat protein. Both SF29 and SF30 are derived from SF25 by introducing appropriate mutations into the parental RNA3 variants.

In vivo recombination assays. Equal amounts of pairs of RNA3 mutants were coinoculated with wt RNA1 and wt RNA2 on C. quinoa leaves, as previously described (25). A mixture of 1 µg of each transcript in 15 µl of the inoculation buffer solution (10 mM Tris [pH 8.0], 1 mM EDTA, 0.1% Celite) was inoculated onto a fully expanded leaf. In each experiment, four separate leaves per plant (two to three plants) were inoculated. Each inoculation experiment was repeated two or three times. The inoculated C. quinoa plants were maintained in a standard greenhouse. Local lesions were counted, collected 14 days postinoculation, and stored at 80°C.

Cloning and analysis of recombinants. For each parental RNA3-RNA3 mutant combination, total RNA was isolated from separate local lesions as described previously (25) and the full-length RNA3 recombinants were amplified as cDNA, using RT-PCR, as described previously (25).

Control RT-PCR amplifications. Control experiments were performed to rule out the possibility of recombinant artifacts being generated by the RT-PCR amplifications. The transcribed BMV RNA3 mutants were mixed in pairs (SF23 × SF25, SF23 × SF29, SF23 × SF30, and SF29 × SF30), 100 ng each, and used directly as templates for RT-PCR amplification with primers AB6 and AB270. These PCR fragments were then cloned and analyzed, as described above.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Infectivity and in vivo stability of RNA3 mutants. Figure 1 shows the location of the sequence modifications present in the four parental BMV RNA3 variants SF23, SF25, SF29, and SF30. The infectivity of the SF mutants was determined on C. quinoa, a local-lesion host for BMV. The transcribed RNA3s for each mutant were mixed with transcribed RNA1 and RNA2 and inoculated on the leaves. The time necessary for the appearance of the local lesions and the number and size of the local lesions did not differ between the mutants and wt infections (data not shown). However, based on virus purification yield, mutants SF25, SF29, and SF30 accumulated five to six times less than did wt virus, while SF23 accumulated two to three times less than did the wt virus.

View larger version (40K): [in this window] [in a new window]   FIG. 1.   Location of marker mutations in the BMV RNA3 variants. Lines represent the noncoding regions, while shaded boxes represent open reading frames for 3a and for the coat protein (CP). The modified nucleotides and amino acids are shown below, in bold type. The numbers on top represent nucleotide positions in the wt sequence, marking the beginning and end of each of the noncoding and the coding regions. The disabled BclI site marker mutation in SF29 is indicated within square brackets (the third line).

Coinfection with mutants SF23 and SF25. In RNA3 variant combination SF23 and SF25, the marker restriction sites (BamHI) were introduced at the 3' and 5' ends, providing a 1.73-kb window for recombination (Fig. 1). The coinoculation with variants SF23 and SF25 could theoretically lead to the appearance of two kinds of recombinants, R1 and R2 (Fig. 2), since each of the marker mutants could serve as a donor or as an acceptor molecule. The restriction analysis of the RT-PCR cDNA products from 85 lesions (of four independent experiments) revealed that 75% of the lesions contained a single parental RNA3 variant (Table 1). Within this group of local lesions, the individual parental RNA3 variants were represented evenly, reflecting similarities between SF23 and SF25 in their ability to induce infection.

View larger version (22K): [in this window] [in a new window]   FIG. 2.   Location of marker mutations in pairs of BMV RNA3 variants and in the projected homologous recombinants (designated R1 through R8). Solid triangles and circles represent BamHI and disabled BclI sites, respectively. The numbers on the top scale give the location of marker mutations. The predicted RNA3 recombinants include only those that arose due to a single crossover event between the coinfecting parental RNA3 variants. Recombinants R5, R6, and R8 were not identified in these experiments.

Coinfection with mutants SF29 and SF30. Since marker mutations in the combination SF23 × SF25 were far apart from one another, no specific region(s) active in recombination could be mapped on the sequence of RNA3. To reduce the size of the overlapping region, another pair of mutated RNA3 recombinants (SF29 × SF30) was used, where the marker mutations (the lost BclI and the new BamHI sites) were separated by only 460 nt (Fig. 1). The expected recombinants R3 and R4 are shown in Fig. 2.

Control coinfections: SF23 × SF29 and SF23 × SF30. To further map the recombinationally active sequences, we have tested the reciprocal pairs of RNA3 recombinants: SF23 × SF29 and SF23 × SF30 (Fig. 1). As shown in Table 1, for the pair SF23 × SF29, all 50 local lesions tested contained only variant SF23, indicating that SF23 outcompetes SF29, which makes the appearance of recombinants difficult. However, for the pair SF23 × SF30, 61 lesions accumulated variant SF23 while 6 contained potential recombinants. Oddly, none of the 67 local lesions analyzed contained the mutant SF30. This result indicates that, similar to SF29, SF30 can be overcome by SF23 (Table 1).

Control RT-PCR amplifications. The in vitro-transcribed RNA3 variants (SF23, SF25, SF29, and SF30) were amplified separately by RT-PCR, using the same conditions and the same primers described above. As shown in Table 3, all the generated fragments had the expected parental restriction profiles and none revealed abolished BamHI or BclI sites, indicating that under the experimental conditions used, the RT-PCR amplifications did not generate sequence alterations affecting the restriction sites.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this paper we report the generation of several distinct BMV RNA3 variants by homologous recombination. We found that pairs of the RNA3 molecules, which coreplicate and coaccumulate in local lesions, can recombine with each other and thus provide a useful homologous recombination assay system. The recombinants can accumulate in the RNA3 populations with no or low selection pressure (mutants SF23 × SF25). In fact, the recombinants accumulated even when a competing pair (SF29 × SF30) of RNA3 was used: the recombinants were observed in 90% of doubly infected lesions, compared to 88% for the less competing SF23 × SF25 combination. This demonstrates that recombination between two BMV RNA3 variants can be readily identified in local lesions.

The recombinant, double-marker, full-length RNA3 cDNA clones were not the RT-PCR artifacts but were formed in planta during the life cycle of the virus. This suggests that neither the RT nor the Taq DNA polymerase can support the template exchanges in vitro at comparable levels. However, the in vivo formation of RNA recombinants may be promoted by a much larger number of RNA replication cycles during infection compared to that during the RT-PCR amplifications. Therefore, to compare the template-switching abilities among these three polymerases, the frequency of the BMV RNA crossovers must be studied in vitro.

The difference in the recombination frequency between pairs SF23 × SF25 and SF29 × SF30 is relatively small and cannot be attributed to the distance between marker mutations (1.73 and 0.46 kb, respectively). This result, as well as the fact that infection with the highly competitive pair SF29 × SF30 generated recombinants, suggests the existence of a homologous recombination hot spot within the internal region between the SF29 and SF30 markers. The nucleotide alterations in the progeny RNA3 recombinants, which have occurred close to the genomic region, may reflect the template-switching positions in the imprecise homologous crossovers. However, these alterations may have also arisen due to copying errors by BMV replicase. Further experiments are required to differentiate between these possibilities.

Interestingly, a related observation was made during the in vitro synthesis of full-length transcripts from the cDNA of RNA3. The occurrence of a shorter transcript (nearly 850 nt) was demonstrated (data not shown) using either the T7 or SP6 RNA polymerase or the enzyme mix of the MEGAscript T7 Kit (Ambion). This was not observed when the cDNA of RNA1 and RNA2 were used as templates. The shorter RNA3 transcripts most probably correspond to a fall-off product at the intercistronic region and suggest the existence of a pausing signal, which may operate not only with the T7 or SP6 RNA polymerases but also with the BMV RdRp replicase.

The following mechanisms can be responsible for promoting homologous crossovers within the intercistronic region. A poly(A) [poly(U)] track may facilitate the RdRp slippage, so that the enzyme could change templates during either plus- or minus-strand synthesis. The polymerase slippage has been described for several nucleic acid polymerases (4, 7, 10, 11, 15, 32). Another factor enhancing recombination might be the predicted secondary-structure elements that may act as pausing signals for the RdRp, thus facilitating the template switching. Relatively stable hairpins are predicted within the intercistronic regions of the RNA3 component of BMV, broad bean mottle virus, and cowpea chlorotic mottle virus (36). Secondary-structure-dependent evolution was described for cymbidium ringspot virus defective interfering RNAs (16). Also, a large portion of the intercistronic region constitutes the subgenomic RNA promoter, which could act as an internal reattachment site for the RdRp complex (38). Finally, plus strands of RNA3 may function as intergenic replication enhancers, which probably recruit the RNA3 templates into the replication complex (39). Along these lines, Gennadiy et al. have suggested that in barley yellow dwarf virus, the subgenomic promoter of the barley yellow dwarf virus RNA1 could act as a recombination hot spot in the family Luteoviridae (14).

Previously, we have demonstrated that homologous recombination crossovers among different BMV RNA segments could occur at the AU-rich/GC-rich sequences (27, 28, 30). It would be interesting to show whether similar trends could exist during crossovers between the same type (i.e., RNA3) of RNA molecules. Mapping of the homologous recombination hot spots on all three BMV RNA segments is required to find common characteristics of the recombinationally active sequences and the mechanism(s) of these processes.

Most of the supporting data for the role of recombination in RNA virus evolution came from phylogenetic analysis of different virus genes (37). Our results provide new perspectives on the role of homologous RNA recombination in RNA virus evolution. The frequent occurrence of RNA3-RNA3 crosses suggests that the progeny viral RNAs represent a mosaic composed of fragments of different parental RNA templates. According to this model, viral RdRp could jump among the homologous RNAs during replication before completing the replication cycle. Genetically, then, RNA viruses contain multiple copies of the genome that participate in recombination. Such a polyploidal nature not only can provide an efficient pathway to the variability of the viral RNA genome but also can randomize RNA sequences, thereby smoothing out the effects of the errors of RNA replication. Besides genetic recombination, other phenomena that require at least two parental genomes per cell include gene complementation and phenotypic mixing.

Overall, we have shown that the crossovers can occur frequently between homologous RNAs during BMV RNA replication in vivo. Homologous recombination can play an important role in generation the diversity in the RNA virus genome and can provide variants for natural selection that warrants rapid virus evolution. Homologous crossovers are not easily detectable in nature, since they require marker mutations. Our approach of using the coreplicating RNA variants should allow researchers to study such factors of viral RNA recombination as host dependence, virus dependence, and RNA sequence dependence under natural conditions.