INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

RNA recombination contributes significantly to the high level of genetic diversity in animal, plant, and bacterial RNA viruses (5, 8, 20, 35). Studies conducted during the last decade clearly indicate that the exchange of RNA genetic information can occur between viral strains (18, 22), viral species (21), or viral and cellular RNAs (17, 23). In addition to having a role in the evolution of the viral RNA genome and in the generation of new viral strains, RNA recombination can correct errors that arise during RNA replication (5).

In spite of extensive studies, the molecular mechanism of RNA recombination is not well understood. The prevailing model for RNA recombination posits that recombination occurs when the RNA replicase switches strands during RNA synthesis (4, 8, 10, 14, 35). Kirkegaard and Baltimore provided the first experimental evidence in support of this copy-choice model by demonstrating that poliovirus RNA recombination depends on RNA replication (19). For brome mosaic virus (BMV), areas of local complementarity and similarity between viral RNAs have been shown to promote, respectively, nonhomologous and homologous crossovers (24-26). A similar heteroduplex-mediated mechanism may be responsible for homologous recombination in the poliovirus genome (33).

The BMV genome is composed of three RNAs named RNA1 to RNA3 (Fig. 1A). All BMV RNAs share a highly structured 200-nucleotide (nt) tRNA-like sequence at their 3' ends which directs the synthesis of complementary minus-strand RNAs (2, 7). RNA1 and -2 encode the viral replication proteins and are sufficient for replication activities in the absence of RNA3 in protoplasts (1). RNA1 encodes the 1a protein, whose N-terminal half has homology to known methyltransferases and whose C-terminal half contains all the motifs found in RNA helicases (1). RNA2 encodes the 2a protein, which has a central polymerase-like domain flanked by nonconserved N- and C-terminal domains (Fig. 1B). The C terminus can be deleted without any apparent effect on RNA replication in barley protoplasts (39).

View larger version (19K): [in this window] [in a new window]   FIG. 1.   Organization of the BMV genome, functional domains of 1a and 2a proteins, and locations of MF mutations. (A) Molecular organization (not to scale) of the BMV genome. RNA1, RNA2, and RNA3 components encode four proteins: 1a, 2a, 3a, and coat protein (CP). The CP is expressed from subgenomic RNA4. Open reading frames are represented by boxes, while terminal and intercistronic noncoding regions are represented by lines. The folding of the tRNA-like structure is shown at the 3' terminus. (B) Domain organization and interaction between BMV 1a and 2a replicase proteins. Domains for putative methyltransferase (capping), helicase, and RNA polymerase activities are indicated by cross-hatched boxes. Open boxes depict nonconserved regions. The core polymerase domain is marked by a filled box. The negative and positive net charges of 2a N-terminal and helicase domains are indicated by  and +, respectively. (C) Sequence of the first 150 amino acids of the N-terminal domain in 2a protein containing MF mutations. Acidic amino acids bearing negative charges are underlined. The clusters of amino acids that replace the original residues in the MF1 to MF5 mutations are marked by the single-letter code above the wt sequence. The infectivity and 1a-2a interaction data are indicated. For more details, see Table 1.

The BMV replicase is an enzymatic complex that contains the BMV-encoded proteins 1a and 2a, as well as a factor eIF3 associated with the host (32) and possibly other still undefined host factors (16, 32). The interactions in vitro between 1a and 2a have previously been analyzed by coimmunoprecipitation and the yeast two-hybrid assay (15, 16). A portion of the 2a protein N terminus encompassing amino acids 25 to 140 has been demonstrated to bind the 1a helicase-like domain in vitro and by the yeast two-hybrid assay (30a). The domain in 1a that interacts with 2a has been mapped to a protease-resistant structure in the helicase-like region (31). The capacity for protein-protein interaction in vitro is strongly correlated with the ability of the proteins to direct RNA replication in protoplasts (15). The in vitro interaction between 1a and 2a is disrupted by concentrations of KCl greater than 0.5 M (15), suggesting that ionic forces are involved in the interaction of these proteins.

BMV is the first plant RNA virus for which RNA recombination was observed (6). The RNA3 component bearing a 20-nt deletion at the 3' end was repaired in infected plants by replacement of the mutated region with wild-type (wt) RNA1 or RNA2 sequences. However, the frequency of recombination (defined as the fraction of local lesions that accumulated recombinant cells) was lower than 5%. Biologically active BMV RNA3 constructs were designed to increase the recombination frequency and to identify the recombinationally active sequences within the 3' regions of all three BMV RNAs and the intercistronic region of RNA3 (25, 26). Depending on whether or not similar or complementary sequences were required for recombination during BMV RNA replication, as well as whether crossovers occurred within homologous or other regions, the process was described as, respectively, homologous or nonhomologous (30).

For both types of recombination in BMV, experimental systems based on either potential heteroduplex formation or the existence of local homologous regions have been described (30). Results obtained with these systems can be interpreted by assuming the presence of a copy-choice mechanism, which necessitates the participation of viral RNA replication complexes. Consistent with this, mutations introduced into the helicase domain of the 1a protein of BMV can influence the locations of crossover sites and increase the fraction of imprecise heteroduplex-induced recombinants (27) and a single amino acid mutation in the central conserved domain of 2a can inhibit nonhomologous recombination without influencing homologous recombination (12).

We are systematically mapping the domains of BMV replicase proteins involved in recombination. In this work we examine how replacements of clusters of acidic amino acids with their neutral counterparts at five different locations within the 2a N terminus influence 1a-2a interactions and recombinant formation. The lack of 1a-2a interaction is correlated with the loss of virus infectivity, confirming previous observations regarding the role of 1a-2a interaction in BMV replication. Other mutations within the N terminus which do not detectably affect 1a-2a interaction can change the frequencies of homologous and nonhomologous crossovers, as well as the positions at which homologous crossovers occur. One mutant, although noninfectious, exhibited a nearly wt level of 1a-2a interaction. This level of interaction suggests that, in addition to 1a-2a interaction, other features of the 2a N terminus may determine the biological activity of BMV. Overall, our study reveals the multifunctional character of the 2a N terminus and provides new evidence that engineering of modified recombination levels is possible by mutagenizing viral replicase proteins.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Materials. Plasmids pB1TP3, pB2TP5, and pB3TP7 (13) were used to synthesize in vitro infectious transcripts of the wt BMV RNA components 1, 2, and 3. Plasmids pB2 MF1 to -5 were used to synthesize RNA2 transcripts bearing the MF1 to -5 mutations. Plasmid pPN8() (25) was used to obtain in vitro the transcripts of the PN8() derivative of BMV RNA3, which is suitable for study of nonhomologous recombination. Plasmids pPN-H39 and pPN-H66 were used to synthesize transcripts of the PN-H39 and PN-H66 derivatives of BMV RNA3, which are suitable for study of homologous recombination (26). T7 RNA polymerase, Moloney murine leukemia virus reverse transcriptase, and restriction enzymes were from Gibco BRL (Gaithersburg, Md.), and our Sequenase kit was from United States Biochemical Corporation (Cleveland, Ohio).

Preparation of MF RNA2 mutants. Mutated RNA2 was derived from pB2TP5 containing wt RNA2 cDNA. Appropriate alterations within the cDNA were introduced by using a PCR-based site-directed mutagenesis method with the primers listed below. The primers were used to generate appropriate PCR products that were then used to replace portions of plasmid pB2TP5 by cutting it with specific restriction enzymes. Finally, each new fragment introduced into pB2TP5 was sequenced to verify the introduced changes. Primers and restriction sites used to obtain plasmids pB2MF1 to -5 were as follows (where restriction enzyme sites appear in boldface letters and underlined bases indicate changes introduced into the wt 2a sequence). For construct pB2MF1, primers 1 and 2 were used. Primer 1, 5'-TGAAGGCTAGCAGCCTCCACCTGGTTTTGTAAGGATTG-3' (38 nt), was complementary to the plus-strand sequence of pB2TP5 between positions 170 and 207 and had an NheI restriction site; primer 2, 5'-CATGCCTGCAGGTCGAC-3' (17 nt), had a unique PstI restriction site. For construct pB2MF2, primers 3 and 4 were used. Primer 3, 5'-GGCTGCTAGCCTTCAGGTGCAGCAGCCCGCAAACGGAGTTGCC-3' (43 nt), represented the plus-strand sequence of pB2TP5 between positions 193 and 235 and had an NheI restriction site; primer 4, 5'-GTTGAACCATTTGTTGGACGGTG-3' (23 nt), represented the plus-strand sequence of pB2TP5 between positions 336 and 362 and had a PflMI restriction site. For construct pB2MF3, primers 5 and 6 were used. Primer 5, 5'-GGGATACCAGTTATCAATTTGATCTTGGAGCACGAAAGAG-3' (40 nt) (of negative polarity), represented the pB2TP5 sequence between positions 430 and 469; primer 6, 5'-GGGGCTCTATTTGCGACAC-3' (23 nt) (of positive polarity), represented the pB2TP5 sequence between positions 324 and 342. The PCR product obtained with primers 5 and 6 (about 250 nt long) was purified and used as a primer for the next PCR together with primer 7, 5'-CAAAGTCCATGGAATAATCACC-3' (22 nt), which was complementary to the pB2TP5 sequence between positions 872 and 898 and had an NcoI restriction site. pB2MF4 and -5 were obtained in the same way except that different mutagenic primers were applied. For construct pBMF4, primer 8, 5'-CCGTAACCATTACTAGTATTCTGGGGATACC-3' (31 nt) (of negative polarity), representing the pB2TP5 sequence between positions 462 and 492, was used. For construct pBMF5, primer 9, 5'-CGCTCGCATGATTTTGATTGGCGGCAAACG-3' (30 nt) (of negative polarity) representing the pB2TP5 sequence between positions 498 and 527, was used.

In vivo recombination assays. Full-length, capped RNA transcripts were made from EcoRI-linearized plasmids, according to previously published procedures (12). Chenopodium quinoa leaves were inoculated with a mixture of the transcribed BMV RNAs, as described before (24). Briefly, a mixture of 1 µg of each transcript in 15 µl of inoculation buffer (10 mM Tris [pH 8.0], 1 mM EDTA, 0.1% Celite, 0.1% bentonite) was inoculated on a fully expanded leaf. In each experiment, six to nine separate leaves (on 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 14 days postinoculation.

Northern blot hybridization. The accumulation of recombinant RNAs was analyzed by Northern blot hybridization. Total RNA extracts from separate local lesions were separated by electrophoresis in a 1% agarose gel and blotted to a Nylon membrane (Hybond N+; Amersham). The membranes were probed with a ³²P-labeled RNA complementary to the 3' end of the plus strand of BMV RNA3. Based on the differences in the sizes of the parental and recombinant RNA species, we were able to confirm that the recombinants were not generated by RT-PCR but were formed during infection in planta.

1a-2a interaction in the yeast two-hybrid system. DNAs encoding the MF1 to -5 mutations were generated by PCR with a 5' primer (5'-AGAATTCGTCTTTCCAATGGATC-3') containing an EcoRI site and BMV cDNA2 (nt 148 to 163) and a 3' primer (5'-AGGATCCATCAGATCGCTCGCATGA-3') containing a BamHI site and sequence complementary to BMV cDNA2 (nt 531 to 517). After PCR with templates containing the MF1 to -5 mutations, the products were first cloned into the pCRII vector (Stratagene Inc.) and then digested with EcoRI and BamHI and cloned behind the GAL4 activation domain in plasmid pGAD424 (kind gift of Stan Fields [11]). The resultant plasmids, pGAD2aN-MF1 to -MF5, were transformed into Saccharomyces cerevisiae Y835 (lys2::lexAop-HIS3 ura3::lexAop-lacZ trp1-901 his3 leu2-3 ade2 gal14 gal80) by electroporation (3) along with 812 H (31). Qualitative and quantitative assays for -galactosidase activity were performed as previously described (31).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Infectivity and stability of 2a mutants. Several clusters of acidic amino acids can be distinguished within the 2a N-terminal domain (Fig. 1C). To examine their role in 1a-2a interaction and in RNA recombination, we decided to mutagenize the 2a N terminus. The mutations within the N terminus were expected to affect not only the interaction of 1a and 2a but also, indirectly, the activities of 1a and 2a. The MF series of mutants (MF1 to MF5) were made in the N terminus of 2a by site-directed mutagenesis, as described in Materials and Methods. The amino acids changed are between residues 26 and 28 (MF1), 38 and 41 (MF2), 114 and 123 (MF3), 123 and 127 (MF4), and 136 and 138 (MF5). Each mutant contains changes of two to four amino acids, with aspartate usually being changed to asparagine and glutamate usually being changed to glutamine (Table 1; Fig. 1C). These changes were predicted to reduce the overall negative charge within the N-terminal domain.

The effect of MF mutations on 1a-2a interactions. Whether mutations in the acidic residues of 2a will affect the 1a-2a protein-protein interaction was tested by the yeast two-hybrid assay (31). For these assays, DNA encoding the N-terminal 140 residues of wt or mutant 2a were fused to DNA encoding the GAL4 activation domain in a LEU2⁺ plasmid (see Materials and Methods). The 1a helicase-like sequence was fused with the LexA protein in a TRP1⁺ plasmid, p1aH (31). Plasmids were then transformed into yeast, and the interactions between both proteins were detected and quantified as previously described (31). The levels of specific activity obtained for MF1, MF2, and MF5 were similar to that of the wt sequence (Table 2), but there was no detectable interaction obtained for MF3 and MF4 with the 1a helicase-like domain. This demonstrates that the 2a amino acids within residues 114 and 127 are important for 1a-2a interaction. Interestingly, MF1 was not infectious despite the wt level of 1a-2a interaction. Apparently, factors other than the level of 1a-2a interaction are also important for virus infectivity.

The effect of MF mutations on the frequency of recombination. Since MF2 and MF5 were infectious, they were used for further studies of RNA recombination. For this reason, two previously elaborated in vivo recombination systems were employed (Fig. 2). We have found previously that, with both systems, practically all local lesions accumulated only one type (though different types among lesions) of recombinant.

View larger version (20K): [in this window] [in a new window]   FIG. 2.   Schematic diagrams (not to scale) of BMV RNA3 recombination vectors used in this study. (A) General organizations of PN8() RNA3 (used for the study of nonhomologous recombination) and PN-H39 and PN-H66 RNA3 (used for the study of homologous recombination). Each RNA3 consists of four regions, A, B, C, and D. In PN8() RNA3, region A is 216 nt long and consists of the conserved BMV 3' end (positions 1 to 236 from the 3' end, counting an internal 20-nt deletion [see below]), and region B constitutes a partial duplication of region A between positions 7 and 200. Additionally, both regions A and B have 20-nt deletions between positions 81 and 100 (marked with small boxes). The 197-nt region C (black box) is derived from all but the last 23 nt of the 3'-terminal sequence of cowpea chlorotic mottle virus (CCMV) RNA3, and region D is the 141-nt sequence complementary to the RNA1 3' noncoding fragment between positions 249 and 389 (counted from the 5' end). PN-H39 and PN-H66 RNA3s have the same regions A and B as PN8() RNA3. However, their region C has a 765-nt-long CCMV RNA3 sequence (positions 24 to 788, counted from the 3' end). Also, region D includes wt RNA2 sequence between positions 160 and 219 (for PN-H39) or between positions 197 and 219 (for PN-H66). (B) During nonhomologous (heteroduplex-mediated) recombination, PN8() RNA3 and wt RNA1 can form a heteroduplex structure. The proposed mechanism of nonhomologous recombination assumes that crossovers take place during minus-strand synthesis initially of RNA1 and that when viral replicase reaches the stable double-stranded region, the enzyme complex switches from one RNA template to another. Repaired recombinant RNA3 is generated if replicase starts RNA synthesis from the nonmodified RNA1 3' end and after crossover resumes nascent-strand elongation with RNA3 as a template. (C) The PN-H39 and PN-H66 RNA3-based vectors, which have 23 to 60 nt of sequence homology with wt RNA2, can efficiently direct homologous RNA2-RNA3 crossovers. In such instances, the generated recombinant has a nonmodified 3' end derived from RNA2 and the coding region and noncoding 5' end derived from RNA3. According to a proposed mechanism (29, 30), the enzyme complex initiates on the minus-strand RNA3 template and continues the synthesis of the nascent (plus) strand on the RNA2 template. CP, coat protein.

View this table: [in this window] [in a new window]   TABLE 3.   Influence of MF2 and MF5 mutations on nonhomologous [parental PN8() RNA3] and homologous (parental PN-H39 and PN-H66 RNA3s) recombination frequencies

The effect of MF mutations on the distribution of crossover sites. Locations of the crossover sites were determined by sequencing of the corresponding RT-PCR-generated cDNA clones. With the nonhomologous recombinants it was possible to determine precisely the locations of the junction sites. Sequencing revealed that the nonhomologous junction sites obtained with MF2 and MF5 occurred within the same portion of the heteroduplex as those obtained with wt RNA2 (Fig. 3). For MF2 RNA2 we found only three recombinants. Interestingly, all of them were identical even though they were from independent infections. At present we do not know whether this is fortuitous, reflects changes in overall selection pressure, or is an effect of the particular 2a mutation on selection of crossover sites. For MF5 RNA2, six recombinants were identified; five of them were located within the recombination hotspot region, while only one recombinant (b1) was highly asymmetrical, i.e., the crossovers occurred at distant locations on both recombining RNA components (Fig. 3).

View larger version (17K): [in this window] [in a new window]   FIG. 3.   Distribution of crossover sites in nonhomologous recombinants generated between PN8() RNA3 and nonmutated RNA2 (A) or with MF2 and MF5 RNA2 mutants (B). Each recombinant was isolated from a separate local lesion. The upper lines of sequences represent the positive strand of PN8() RNA3 that contains the RNA1 sequences of negative polarity (segment D), while the lower lines represent the corresponding (complementary) sequences of the positive strand of wt RNA1. Junctions are marked by arrows with letters (uppercase letters for wt RNA2 and lowercase letters for MF mutants). Asterisks indicate recombinants found with the MF2 mutant. The numbers show how many recombinants of a given type have been identified. The numbers 36 and 17 in large brackets indicate the numbers of bases in unknown wt BMV sequences.

View larger version (32K): [in this window] [in a new window]   FIG. 4.   Diagram summarizing the recombination frequencies (Fr) and distribution of crossover sites in homologous recombinants. (A) Locations of crossover sites and of marker nucleotide substitutions within the common sequence in homologous RNA2-RNA3 recombinants isolated from infections with wt RNA1; either wt, MF2, or MF5 RNA2; and PN-H39 RNA3 (A) and with PN-H66 RNA3 (B). Uppercase letters indicate homologous nucleotides. Marker mutations are indicated by asterisks, whereas regions between marker mutations are named R1 to R4 or R'1 to R'3. Each recombinant contains 3'-terminal sequences derived from RNA2 on the right side and 5'-terminal sequences from RNA3 on the left side. The incidence of each RNA3 recombinant is shown in the boxes immediately to the right of the arrows (the Fr columns give the numbers of particular recombinants over the total numbers of identified recombinants, and the % columns give those values expressed as percentages). Each entry represents an RNA2-RNA3 recombinant isolated from a separate local lesion. The names of individual recombinants are shown on the left. Each leftward-pointing arrowhead marks the last nucleotide derived from RNA2, and each rightward-pointing arrowhead marks the first nucleotide derived from RNA3. Dotted lines show ambiguous regions that may be derived from either RNA2 or RNA3 in the precise homologous recombinants. Nontemplated nucleotides and nucleotide substitutions generated during the crossover events are shown by lowercase letters between the arrows. The nucleotide sequences in the imprecise recombinants with ambiguous crossovers were arbitrarily placed with the upstream junction.

Precision of homologous recombination. Analysis of recombinant sequences allowed us to determine whether mutations introduced into the N terminus of 2a protein can influence the precision of homologous crossovers. Four imprecise homologous recombinants (types A5 to A9) were identified during infection with wt RNA1, wt RNA2, and PN-H39 RNA3, while one imprecise recombinant (type B4) was observed with PN-H66 RNA3 (Fig. 4). Imprecise recombinants constituted 12 and 5% of the total number of recombinants generated with PN-H39 and PN-H66, respectively. Similar results were obtained with MF5 RNA2. Here, two imprecise recombinants were obtained with PN-H39 RNA3 and two others were obtained with PN-H66 RNA3 (respectively, 6 and 9% of identified recombinants). The situation markedly changed if, instead of wt RNA2, MF2 RNA2 was used. MF2 RNA2 gave eight (35%) imprecise recombinants when it was used with wt RNA1 and PN-H39 RNA3 (Fig. 4A) or 2 (33%) imprecise recombinants with wt RNA1 and PN-H66 RNA3 (Fig. 4B). As with precise recombinants, MF2 mutation (Fig. 4A) did not influence the locations of imprecise crossover sites for PN-H39 RNA3 (situated within region R2).

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

Several lines of evidence demonstrate that RNA recombination in BMV is a process intimately associated with viral RNA synthesis. Mutations in the BMV-encoded 1a protein, and especially a temperature-sensitive mutation in its helicase-like domain, were shown to influence the frequency and positions of nonhomologous crossovers (27). Also, we demonstrated previously that a single amino acid substitution (designated DR7) within the central core polymerase domain of 2a can inhibit nonhomologous RNA-RNA recombination (12). DR7 also influenced the location but not the frequency of homologous crossovers, suggesting that homologous and nonhomologous recombination are the result of different mechanisms.

To assess the role of various functional domains on BMV replicase proteins in recombination, in this work we study mutations within the N-terminal domain of 2a. Our findings have several implications. First, we show that replacements of selected clusters of negatively charged amino acids into their neutral counterparts reduce the levels of 1a-2a interactions. The N-terminal domain has been shown previously to participate in 1a-2a interactions (15, 16). Our data not only map the participating regions of the N terminus but also suggest that these interactions have an electrostatic character. Also, the role of 1a-2a interactions in virus infectivity is confirmed by observing a correlation between the presence of negatively charged amino acids and the electrostatic character of the regions in which they are found.

Second, one purpose of our study was to test whether 1a-2a interactions were involved in guiding recombination events. We expected to observe a quantitative correlation between levels of 1a-2a interaction and recombination. However, of the two infectious MF mutations, neither displayed conditional properties. Thus, it was not possible to test the influence of the 1a-2a interaction on recombination, even though it was possible to test that on replication. Further mutagenesis efforts are necessary.

Third, we now confirm a previously reported property of DR7 (see above), namely, that it is possible to decouple replication and recombination events. While the replication levels of MF2 and MF5 reach those of the wt virus, the ability of both mutants to support crossovers is deeply reduced. Apparently, there are some regions within the viral RNA polymerase which are responsible for recombinant generation and at the same time are not involved in RNA replication. By studying the progeny BMV RNAs obtained with mutants of different recombination activities, one should be able to determine the role of recombination in maintaining viral genomic sequences during infection.

Fourth, it is certain that decoupling of replication and recombination functions can be achieved not only by mutagenizing directly the putative catalytic domain of RNA polymerase (mutant DR7) but also by changing more distant regions of the 2a protein. This finding suggests that the N-terminal domain is directly involved in the process of RNA recombination. Alternatively, indirect steric effects of N-terminal mutations on the active centers of the enzyme cannot be excluded.

Fifth, MF2 and MF5 display markedly reduced frequencies of nonhomologous recombination, and, in addition, MF2 shows a reduced frequency of homologous recombination. Mutant DR7 can affect only the frequency of nonhomologous recombination (12). Thus, we show that different sites on 2a participate in the two kinds of recombination events during the virus life cycle. Our data further confirm that the mechanisms responsible for both types of recombination are different.

Considering these findings together, we have shown that the N-terminal domain has a multifunctional character, as it can influence the levels of RNA replication, the levels of nonhomologous RNA recombination, and the levels of homologous RNA recombination. Whether this domain has other functions and whether or how all these functions are interrelated remain to be determined. Below we discuss more details regarding the role of the N-terminal domain in 1a-2a interaction and in RNA recombination.

Role of the 2a N terminus in 1a-2a interaction. It was shown previously by using both the in vitro and the yeast two-hybrid system (11, 15, 16, 30a, 31) that the nonconserved 2a N terminus is at least partially responsible for interaction with 1a through the helicase domain. In this work we confirm the involvement of this domain in 1a-2a interactions. Specifically, we observe that the acidic residues within positions 114 and 127 are important for 1a-2a interaction. However, since MF1, MF2, and MF5 did not affect 1a-2a interactions while MF3 and MF4 did, we conclude that acidic amino acids located in regions both downstream and upstream from MF3 and MF4 sites are not needed for interaction with 1a. Other data show that single acidic amino acid changes within residues 70 and 124 can abolish 1a-2a interaction in the yeast two-hybrid system (38). Residues 26 to 28, which are modified in MF1, are known to be dispensable for 1a-2a interaction (31), and deletions in this region can reduce, but do not abolish, RNA replication in barley protoplasts (39). It is then interesting that MF1 is not infectious. Perhaps this mutation affects directly the structure of 2a (or indirectly the structure of 1a), which can then affect the RNA replication process. Thus, MF1 mutation may cause reduction of putative interactions of negatively charged amino acids with the phosphate backbones of replicating RNAs, which are mediated by metal cations (37).

The role of the N-terminal domain in recombination. Mutants MF2 and MF5 altered the frequency of nonhomologous recombination, but no effect on the locations of junction sites was observed. We speculate that such properties of the MF2 and MF5 mutants in nonhomologous recombination can be explained by a mechanism presented in Fig. 2, according to which the existence of a strong heteroduplex (140 nt long) that forms between wt RNA1 and PH8() RNA3 may ensure that the crossovers occur at specific locations (12, 25). This might explain the observations that the locations of crossovers were affected only by mutations within the helicase domain of 1a (27) and not within the domains of 2a (this paper and reference 12). Mutated helicase may have lower efficiency in unwinding the double-stranded RNA, which, in turn, may cause the changes in the positions of crossovers (27). On the other hand, the stability of the overall recombination complex (designated the recombinosome complex [see references 12 and 29]), which is formed by viral replicase, the RNA template (donor), the nascent strand, and the donor-associated acceptor RNA, might be important for the ability of both viral replicase and the nascent RNA strand to approach the acceptor RNA molecule during the switching event. This, in turn, may affect the frequency of crossover events at heteroduplex structures. Accordingly, mutants MF2 and MF5 are expected to influence the stability of the recombinosome complex. At present, we cannot unambigously determine whether changes introduced to the 2a N-terminal domain affect protein-protein or protein-RNA interactions. The data obtained with the yeast two-hybrid system may indicate that protein-RNA binding is disrupted, as both MF2 and MF5 mutants display the wt level of 1a-2a interaction.