Interactions of the RNA Polymerase with the Viral Genome at the 5′- and 3′-Ends Contribute to 20S RNA Narnavirus Persistence in Yeast*

20S RNA narnavirus is a positive strand RNA virus found in the yeast Saccharomyces cerevisiae. The viral genome (2514 nucleotides) only encodes a single protein (p91), the RNA-dependent RNA polymerase and does not have capsid proteins to form intracellular virions. The genomic RNA has no 3′ poly(A) tail and perhaps no cap structure at the 5′-end; thus resembling an intermediate of mRNA degradation. The virus, however, escapes the host surveillance and replicates in the yeast cytoplasm persistently. The viral genome is not naked but exists in the form of a ribonucleoprotein complex with p91 in a 1:1 stoichiometry. Here we investigated interactions between p91 and the viral genome. Our results indicate that p91 directly or indirectly interacts with the RNA at the 5′- and 3′-end regions and to a lesser extent at a central part. The 3′-end site is identical to or overlaps with the 3′ cis signal for replication identified previously. The 5′-site is at the second stem loop structure from the 5′-end (nucleotides 72–104), and this structure also contains a cis signal for replication. Analysis of mutants in the structure revealed a tight correlation between replication and formation of complexes. These results highlight the importance of ribonucleoprotein complexes for the viral life cycle. We will discuss implications of these findings especially on how the virus escapes from mRNA degradation pathways and resides in the cytoplasm persistently despite the lack of a protective capsid.

A wide variety of viruses (DNA, RNA, and retroviruses) can establish persistent infections in the host (1). For long term infections, the virus must not only be invisible to immune surveillance but also replicate stably inside of the host cells. DNA viruses and retroviruses can integrate their DNA genomes into host chromosomes or keep their genomes in an episomal state and utilize the host machinery to replicate them in a cell cycledependent manner. RNA viruses, including hepatitis C virus, also can establish persistent infections. Although clinical aspects of their infection such as the host immune responses have extensively been studied, intracellular events leading to the maintenance and stable replication of the viral genome remain to be elucidated.
20S RNA narnavirus is a persistent positive strand RNA virus found in most laboratory strains of the yeast Saccharomyces cerevisiae (2,3). The virus has no extracellular transmission pathway. The small and simple genome (2514 nucleotides (nt) 2 ) that encodes a single 91-kDa protein (p91), the RNA-dependent RNA polymerase (4 -6), makes this virus an attractive model system to investigate the mechanism of RNA virus persistence. Especially the absence of intracellular virions makes it easier to analyze the viral RNA inside of the cell. Replication proceeds from genomic to antigenomic to genomic strands. The amount of antigenomic strands counts less than a few percent compared with that of the genomic strands. W doublestranded RNA (7), a double-stranded form of the 20S RNA genome, is present in virus-harboring cells and is a byproduct of replication (8). Cells with vegetative growth contain a low amount of 20S RNA virus. Transferring them to nitrogen-starvation conditions, a common procedure to induce sporulation in diploid cells, increases the copy number of viral RNA to almost equivalent to those of rRNAs (9,10). Increase in viral load may contribute to the efficient distribution of the virus to meiotic segregants.
The reading frame for p91 spans almost the entire length of the genome, with only 12 nt at both ends as non-coding regions. 20S RNA has 5-nt inverted repeats at both termini (5Ј-GGGGC . . . GCCCC-OH) (6). It lacks a poly(A) tail at the 3Ј-end and has perhaps no cap structure at the 5Ј-end, thus resembling an intermediate of the mRNA degradation pathway. 20S RNA is not encapsidated into proteinous or membranous structures and it migrates in sucrose gradients almost as naked RNA (11,12). It thus appears that the viral genome would be vulnerable to the exonucleases involved in mRNA degradation. 20S RNA virus, however, resides and replicates stably in the cytoplasm (13). When antibodies against the RNA polymerase became available, it was found that p91 is associated with the viral genome (12). Further studies showed that most (if not all) of the genomic strands exist in the form of ribonucleoprotein complexes with p91 in a 1:1 stoichiometry (13), suggesting that formation of these complexes is a key factor for the stabilization of the viral genome inside of the cell. In this work we investigate the interactions of p91 with the viral RNA in ribonucleoprotein complexes. Our results indicate that p91 interacts with the genomic strand at the 5Ј-and 3Ј-end regions and also weakly at the central part. Mutations at the 5Ј-and 3Ј-sites that impaired formation of complexes in vivo also abolished replication, indicating the importance of formation of ribonucleoprotein complexes in the viral life cycle. We will discuss possible roles of these complexes, especially on the stability of the viral genome during 20S RNA persistent infection.

EXPERIMENTAL PROCEDURES
Immunoprecipitation of RNase-treated 20S RNA/p91 Complexes-Ribonucleoprotein complexes of 20S RNA virus were isolated from strain 37-4C (a kar1-1 leu1 20S RNA, 23S RNA, L-A-o, L-BC-o) (7) after converting the cells to spheroplasts and breaking them through a French pressure cell (5). Then complexes were purified by a differential centrifugation followed by a sucrose gradient, as described in Ref. 14. 100 l of a sucrose gradient fraction (containing ϳ50 g of protein) were diluted 10 times with TBST (10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 0.05% Tween 20) containing 5 mM MgCl 2 , and digested for 10 min at 25°C with different concentrations of RNase A alone or along with RNase V1. 50 g of yeast tRNA (Invitrogen) and 2 l of anti-p91 antiserum (13) were added to the solution. The reaction mixture was kept at 4°C for 30 min. 50 l (wet volume) of protein A-conjugated Sepharose CL-4B (Amersham Biosciences) was added to the mixture, and it was incubated at 4°C for another 30 min. The Sepharose was washed four times with 1 ml of TBST and suspended in 200 l of 5 mM EDTA and 0.5% SDS. The RNA bound to Sepharose was extracted with phenol, phenol/chloroform and precipitated with ethanol. The pellet was dissolved in 100 l of 50% dimethyl sulfoxide, 1 M glyoxal, and 10 mM sodium phosphate, pH 7.0, and incubated at 50°C for 1 h. The solution was diluted with 1 ml of 10ϫ SSC and divided into aliquots. Each aliquot was blotted onto a Nytran N membrane (Schleicher and Schüell) using a Bio-Dot SF apparatus (Bio-Rad). RNA on the membranes was detected by hybridization using T7 or T3 RNA polymerase run-off transcripts as described in Ref. 15 or using deoxyoligo probes (16). Quantification of the RNA recovered was performed with a phosphorimager.
Hydroxyl Radical Footprinting-Hydroxyl radical treatment was done as described in Ref. 17. We used 20S RNA/p91 complexes purified by a differential centrifugation instead of using a sucrose gradient fraction to avoid quenching of radicals by sucrose. The cleavage pattern was detected by reverse transcription (17) using the 5Ј-32 P-labeled oligo primer NR2 (5Ј-GACGGCTCC-AACCGTAG-3Ј) complementary to nt 159 -175 in 20S RNA (6).
Launching Assay-20S RNA virus was generated in vivo from a launching plasmid as described in Ref. 18.
Formation of Ribonucleoprotein Complexes in Vivo-Pulldown assays of ribonucleoprotein complexes formed in vivo were performed as described in Ref. 19. Briefly, transformants of 20S RNA-negative strain 2928-4 (a ura3 trp1 his3, 20S RNA-o, 23S RNA-o, L-A-o) (18) with expression plasmids were grown and harvested in late logarithmic phase. Cells (3 ϫ 10 8 ) were broken with glass beads by vortex mixing. After removing the cell debris and unbroken cells, lysates were incubated with anti-p91 antiserum or its preimmune serum and immunocomplexes were isolated with protein A-conjugated Sepharose CL-4B. The RNA pulled-down was extracted and slot-blotted onto a Nytran N membrane, and 20S RNA transcripts on the membrane were detected by hybridization. Quantification of band intensities was done with a phosphorimager. Throughout FIGURE 1. p91 interacts with 20S RNA at three sites in a ribonucleoprotein complex. A, immunoprecipitation of intact 20S RNA molecules by anti-p91 antiserum. Sucrose gradient-purified 20S RNA/p91 complexes were immunoprecipitated with anti-p91 antiserum (Anti-p91) or the preimmune serum (Preimmune). RNA was extracted from the precipitate, separated in an agarose gel, blotted, and hybridized with a 20S RNA genomic strand-specific probe. As control, 20S RNA directly extracted with phenol from the gradient fraction was analyzed in the same gel (Phenol). The mobility of some Lambda HindIII fragments is shown. B, 20S RNA fragments associated with p91 after RNase A treatment. 20S RNA/p91 complexes were digested with RNase A at the concentrations indicated and then immunoprecipitated with anti-p91 antiserum. RNA in the precipitates was extracted and blotted onto nylon membranes. We made nine sets of blots, and each membrane was hybridized with a 32 P-labeled probe that recognizes a different part of the 20S RNA molecule (1)(2)(3)(4)(5)(6)(7)(8)(9). The percentages of the RNA fragments recovered in the precipitates after treatment with 1 g/ml of RNase A are shown below each panel. C, RNA probes (1-9) used for hybridization. this work we used T3 run-off transcripts from SmaI-digested pALI-18 (18) as a 20S RNA-specific probe unless described otherwise. The probe contains the entire 20S RNA antigenomic strand sequence. RNA was also extracted directly from the lysates (Total RNA), slot-blotted, and hybridized with either the 20S RNA-specific probe or the 32 P-labeled oligonucleotide RE368 (5Ј-CCTCATAAAACTGATACGAGCTTCTGCTA-TCC-3Ј) complementary to nt 963-994 in 25S rRNA. The latter was used to detect 25S rRNA as a loading control.
In Vitro Mutagenesis-Site-directed in vitro mutagenesis was done as described (20). All mutations introduced were confirmed by DNA sequencing.

RESULTS
p91 Interacts with the 20S RNA Viral Genome at Three Sites-20S RNA forms ribonucleoprotein complexes with p91. These complexes, partially purified through a sucrose gradient, can be specifically immunoprecipitated with anti-p91 antiserum and intact 20S RNA molecules can be recovered in the precipitates (Fig. 1A). We wished to determine the p91 binding site(s) on the RNA. For this, complexes were first treated with various amounts of RNase A and then immunoprecipitated with anti-p91 antiserum. RNA fragments pulled down by the antibodies were analyzed by Northern blots using different RNA probes. We made nine sets of blots, and each membrane was hybridized with a 32 P-labeled probe that recognized a different part of the 20S RNA molecule (Fig. 1C). After treatment with the highest concentration of RNase A, the 5Ј-and 3Ј-most fragments (detected by probes 1 and 9, respectively) were found in the precipitates at a level (70 -80%) similar to the undigested control (Fig. 1B). Probes 3, 4, and 6 -8, however, detected almost no RNA fragments. Probe 5 that recognized the central portion (nt 1253-1513) of 20S RNA also gave a lesser but significant signal (35% of the control). A secondary structure analysis with the MFOLD program (21) predicts that the 20S RNA genome has long distance interactions that connect the 5Ј-end, central, and 3Ј-end regions detected by probes 1, 5, and 9, respectively (Fig .  4A). Therefore, these data suggest that p91 interacts only with one or two of these sites and that the other sites are pulled down indirectly through RNA-RNA interactions. Alternatively, p91 may interact with all of three sites.
The 3Ј cis Site Is Identical to or Overlaps with the 3Ј cis Signal for Replication-The 20S RNA genome has 5-nt inverted repeats at both termini (5Ј-GGGGC . . . GCCCC-OH) and the 3Ј-terminal four C nt are attached to a stem loop structure (Fig.  2B). Previously we established an in vivo launching system of 20S RNA virus from a yeast expression vector (18). The vector contained the complete cDNA of the 20S RNA genome with a ribozyme sequence directly fused to its 3Ј-end. Using this system we found that the third and fourth C nt from the 3Ј-end are essential for virus generation. While the terminal and penultimate C nt were dispensable for in vivo generation, the launched virus recovered these terminal nucleotides. Furthermore, the fifth nucleotide from the 3Ј-end, which is a G, is dispensable but the nucleotide at this position needs to be hydrogen-bonded at the bottom of the stem for virus generation (18). The inactivation of the ribozyme by changing GGG 3Ј to the cleavage site to AAA results in the failure of virus generation. The vector (the expression plasmid; Fig. 2A) still can transcribe the 20S RNA genome from the promoter and p91 can be translated from the transcript. This p91 forms complexes with the 20S RNA transcripts in vivo even though they have non-viral extensions at their 3Ј-ends (not shown). Thus these complexes can be pulled down with anti-p91 antiserum and the 20S RNA transcripts in the precipitates can be detected by hybridization. Using this in vivo system we analyzed the effect of mutations at the 3Ј-end on complex formation. As shown in Fig. 2C, we found that, while changing the 3Ј-terminal or penultimate C to other nucleotides did not affect formation of complexes (Fig. 2, lanes 1 and 2), a modification at the third or fourth C impaired the activity by about 90% (Fig. 2, lanes 3 and 4). The replication-negative mutants still retain a basal level of complex forming activity (10%). As will be shown later, the cis site at the central part appears to be responsible for this basal activity. A mutation at the fifth nucleotide (G5A) that abolished replication also affected the formation of complexes (Fig. 2, lane 5). Because this mutation also shortened the stem by one base pair, the result indicates that formation of complexes requires either a G at this position or the nucleotide in this place to be hydrogenbonded for full activity. To examine these possibilities we car-  [1][2][3][4][5]. Lysates were prepared from transformants or from untransformed cells (lane Ϫ) and incubated with anti-p91 antiserum (Anti-p91) or its preimmune serum (Preimmune). RNA was extracted from immunoprecipitates, slot-blotted, and hybridized with a 20S RNA-specific probe. RNA was also extracted directly from the lysates (Total RNA), blotted, and hybridized with a 20S RNA-specific probe or a 25S rRNA-specific oligo probe. The results with the rRNA probe indicate that the extracts contained fairly equal amounts of total RNA. The effects of these mutations on virus generation observed previously in launching experiments (18) are shown on the right of the panel (Launching). The relative amounts of 20S RNA transcripts pulleddown with anti-p91 antiserum are also shown (Recovery).
ried out the experiments shown in Fig. 3. Consistent with the result with the G5A mutation, a G5C change also gave a basal level of complex forming activity (Fig. 3, lane 1). To restore base pairing at the bottom of the stem in the latter mutant, we introduced a compensatory mutation (C35G, numbered from the 3Ј-end) to the other side of the stem (Fig. 3, lane 3). Because this mutation also changed the amino acid sequence of p91 from Thr to Ser at codon 823 we made another control to examine the effects of this amino acid modification. p91 was changed from Thr to Ser at the same codon by A36U (numbered from the 3Ј-end) without modifying the sequence and structure at the 3Ј-stem loop in the transcripts (Fig. 3, lane 4). Both transcripts formed complexes with the modified p91 at a wild-type level. The results thus indicate that p91 tolerates the change from Thr to Ser at codon 823 to form complexes and that the fifth nucleotide G from the 3Ј-end is dispensable for complex formation. The results further suggest that this fifth nucleotide from the 3Ј-end needs to be hydrogen-bonded for full activity. This was confirmed by the failure of transcripts having the C35G mutation alone in forming complexes even though they contained a wild-type G at the fifth position (Fig. 3, lane 2). There is a clear correlation between replication and formation of complexes among the mutants analyzed in Figs. 2 and 3. All the mutants that formed complexes at a wild type level generated 20S RNA virus when expressed from a launching plasmid, whereas the mutants that formed complexes at a basal level failed to generate the virus from the launching vector. These results demonstrate that the 3Ј cis site for formation of ribonucleoprotein complexes is identical to or overlaps with the 3Ј cis signal for replication identified previously.
The 5Ј-Binding Site Is Located in the Second Stem Loop Structure from the 5Ј-End-To locate the 5Ј-binding site more precisely, we took two different approaches. First, in an experiment analogous to the one shown in Fig. 1B, we used five small deoxyoligo probes that covered the region recognized by RNA probe 1 (nt 1-326), to finely map the 5Ј-end site (Fig. 4B). After treatment with RNase A, the 5Ј-most stem loop structure (Stem-loop 1, nt 1-71) was not found in the precipitate, while the second stem loop structure (Stem-loop 2, nt 72-104) or internal regions were recovered efficiently. These results suggest that the 5Ј-binding site is located at the Stem-loop 2 or a more internal region. The probes failed to detect RNA fragments when complexes were exhaustively digested with a mixture of RNases A and V1. In the second approach, we treated native complexes with hydroxyl radicals and determined the regions protected from cleavages by reverse transcription (Fig.   FIGURE 3. The fifth nucleotide G from the 3-end in the 20S RNA genome is dispensable for formation of complexes. A, cells were transformed with the standard expression plasmid (lane C, control) or with expression plasmids modified as described for B (lanes 1-4). 20S RNA transcripts pulled-down from cell lysates with anti-p91 antiserum (Anti-p91) or with the preimmune serum (Preimmune) or 20S RNA transcripts and 25S rRNA in cell lysates (Total RNA) were detected as described in the legend to Fig. 2C. RNA from untransformed cells was also analyzed (lane -). The effects of these mutations on virus generation from launching plasmids observed previously (18)   The last two stem loop structures at the 3Ј-end have been detected in purified 20S RNA by S1 and V1 mapping (6). Areas complementary to probes (a-e) are indicated by bold lines. Nucleotides are numbered from the 5Ј-end. As a reference to data in Fig. 1, the 5Ј-end, internal, and 3Ј-end regions recognized by probes 1, 5, and 9 are indicated by blue, green, and red colors, respectively. The central region is not drawn to a scale. B, interaction of p91 with 20S RNA at the 5Ј-end region. 20S RNA/p91 complexes were treated with RNase A alone or with RNase V1 at the concentrations indicated. After digestion, the complexes were immunoprecipitated with anti-p91 antiserum, and RNA in the precipitates was blotted onto nylon membranes and hybridized with the 32 P-labeled oligo probes indicated. 5A). Non-cleaved RNA control produced a pattern of premature reverse transcription termination (Fig. 5A, lane 3), consistent with features of the two stem loop structures shown in Fig.  5B. When native complexes were treated with hydroxyl radicals, we found two wide areas (denoted as I and II) protected from cleavages (Fig. 5A, lane 2) compared with the pattern generated from phenol-extracted 20S RNA (Fig. 5A, lane 1). The protected nucleotides are located at both sides of the stem structure in Stem-loop 2, consistent with data from the pulleddown assay with deoxyoligo probes (Fig. 4). This result suggests that p91 physically interacts with the RNA at the 5Ј-site and eliminates the possibility that RNA fragments containing this site were pulled down indirectly by anti-p91 antibodies through RNA-RNA interactions.
Stem-loop 2 at the 5Ј-End Contains cis Sites for Replication and for Formation of Complexes-To confirm the above results, we introduced mutations into Stem-loop 2 and examined their effects on formation of complexes and also on virus generation. The long stem of this structure is divided into two parts by an asymmetric internal loop; a 9-bp upper stem (St-2a) and a 3-bp lower stem (St-2b) (Fig. 6A). To avoid changing the amino acid sequence of p91 we first introduced single mutations at the wobble positions in the structure. Although most of them did not affect virus generation and formation of complexes, C84G, C96G (Fig. 6), and G102C (not shown) reduced both virus generation and formation of complexes moderately. C84G and G102C also shorten St-2a and St-2b by one base pair, respectively. Then, to modify the Stem-loop 2 structure more extensively we changed the two Arg codons present in Stemloop 2. When three bases of the Arg codon 97AGG99 were changed to another Arg codon CGC, thus opening St-2a at the lower edge by three base pairs, this mutant failed to generate 20S RNA virus from the launching plasmid and also formed complexes only at a basal level (10%) when transcribed from the expression plasmid (Fig. 6C, lane 1). Neither activity requires specific nucleotides at the Arg codon because single mutations at positions 97 and 99 (A97C and G99A, respectively) showed wild-type phenotypes for both activities (Fig. 6C, lanes 2 and  3). Furthermore, when a second mutation (U78G) was introduced into the 97CGC99 mutant, thus recovering two base pairs of hydrogen bonding, this compensatory mutation not only restored the ability to generate the virus but also brought complex formation back to a wild-type level (Fig. 6C, lane 4). When the other Arg codon 82CGC84 was changed to AGA, thus shortening the St-2a at the upper edge by three base pairs, this mutation resulted in the complete loss of virus generation and also reduced formation of complexes to a basal level (Fig. 6B, lane 3). Even the single mutation C84G, that shortens the stem by one base pair, reduced virus generation and formation of complexes moderately (Fig. 6B, lane 2). On the other hand, a double mutation in the loop (Fig. 6B, lane 1) or a mismatched pair at the upper part of St-2a (Fig. 6B, lane 4) affected none of the activities. When St-2a was disturbed in its middle part by the creation of a symmetric internal loop of four or six nucleotides by combining two single mutations (U78A and C96G, or U81A and C96G), the modified RNA failed to generate the virus and also formed complexes only at a basal level (Fig. 6D, lanes 4 and 5). U78A or U81A alone did not affect either activity (Fig. 6D, lanes  2 and 3) and, as afore mentioned, C96G reduced virus generation and formation of complexes moderately (Fig. 6D, lane 1). These results indicate that Stem-loop 2 contains cis signals for replication and also for formation of complexes and that the secondary structure is important for both activities.
The mutations in Stem-loop 2 described above can be divided into three groups based on the effects on replication and formation of complexes. 1) Those that did not affect virus generation from a launching vector produced complexes in amounts comparable to the wild type control. 2) Those that failed to generate the virus formed complexes at a basal level (about 10% or less compared with the wild-type control) and 3) Those (C84G, C96G, and G102C) that generated virus at a reduced level formed complexes at an intermediate level. After curing the launching plasmids, the viruses generated from the last group of mutants were induced poorly under nitrogen starvation conditions (not shown), most likely because of the instability of ribonucleoprotein complexes. These data suggest that formation of stable complexes is a prerequisite for efficient replication. Furthermore, we observed that the amounts of 20S RNA transcripts in lysates (see Total RNA columns in Figs. 2, 3, 6, and 7) decreased by 30 -60% compared with the wild-type control when mutants failed to form complexes at full activity. Because these amounts represent steady-state levels, the data suggest that turnover of 20S RNA transcripts was attenuated by formation of complexes with p91. Interestingly, the modification of the third and fourth Gs from the 5Ј-end to UU, which abolished virus generation from a launching vector (18), did not affect formation of complexes (Fig. 6C, lane 5). This is consistent with the data from the pull-down experiment shown in Fig.  4. Because this modification also changes the third and fourth C nt from the 3Ј-end in the antigenomic strand, if copied, this result strongly suggests that its failure to generate 20S RNA virus is caused by the alteration of the 3Ј cis signal for replication present in the antigenomic strand (18).
Involvement of the Central cis Site in Complex Formation-As shown above, mutations at the 3Ј or 5Ј cis sites that abolished virus generation from a launching vector still produced ribonucleoprotein complexes at a basal level. To test whether the interactions of p91 with these two cis sites are accountable for all complex-forming activity we constructed a double mutant in these two cis sites (Fig. 7). The mutation at the 3Ј cis site (C2511A) or at the 5Ј cis site (97CGC 99) reduced the formation of complexes to a basal level (10%) (Fig. 7A, lanes 1 and 2). The double mutant still had the same basal level of complexes and there was no further reduction in its activity (Fig. 7A, lane 3). Anti-p91 antiserum failed to pull down 20S RNA transcripts when ribonucleoprotein complexes had been treated with phenol or when a p91 fragment truncated at Arg 106 or Leu 274 was expressed from the vector (not shown). This indicates that the transcripts bearing the double mutation were still specifically associated with p91, thus suggesting the presence of another cis site for complex formation. These results, therefore, strongly suggest that the interaction at the central site is responsible for the basal level of complex formation exhibited by mutants at the 5Ј and/or 3Ј cis site.

DISCUSSION
mRNA degradation in eukaryotes usually begins with the shortening of the poly(A) tail at the 3Ј-end, followed by decapping at the 5Ј-end. Then the decapped RNA is degraded by the 5Ј XRN1/SKI1 exonuclease. Alternatively, deadenylated RNA is digested by a 3Ј exonuclease complex called the exosome. The exosome consists of multiple 3Ј exonucleases, and to digest mRNA in the cytoplasm the exosome interacts with a heterotrimeric complex of Ski2p, Ski3p, and Ski8p through its cytoplasm specific subunit Ski7p (22,23). Because many RNA viruses lack the 5Ј-cap structure and the 3Ј poly(A) tail in their genomes, the proteins involved in mRNA degradation pathways can play an important role in antiviral defense. Indeed, SKI genes were initially identified in yeast by their antiviral activities on the totivirus L-A and its satellite RNA (24). SKI2, SKI3, SKI7, and SKI8 genes also repress the copy number of 20S RNA virus 3 (25). Because the 20S RNA genome is not sequestered into proteinous or membranous structures in the cytoplasm, it suggests that the 20S RNA genome is constantly challenged by the exosome.
In this work we have investigated the interactions of p91 with the 20S RNA genome in ribonucleoprotein complexes. From complexes predigested with RNase A, anti-p91 antiserum pulled down 20S RNA fragments containing the 5Ј-and 3Ј-end sequences and, to a lesser extent, a central region of the molecule as well. This suggests that p91 interacts with the RNA at three sites. The 3Ј-site is identical to or overlaps with the 3Ј cis signal for replication identified previously. As shown in Fig. 2C, the modification of the third or fourth C from the 3Ј-end to A reduced formation of complexes to a basal level while changing the 3Ј-terminal or penultimate C to A did not affect complex formation. The fifth nucleotide G is located at the bottom of a stem adjacent to the 3Ј-end. Modifying it to A impaired both replication and formation of complexes. This nucleotide, however, is dispensable and can be changed to C without affecting either activity, provided that a compensatory mutation is introduced to other side of the stem to re-establish base pairing with the modified nucleotide (Fig. 3). This tight correlation between replication and formation of complexes is similar to the one observed in 23S RNA virus, another narnavirus present in yeast. Its genome contains a 3Ј bipartite cis signal essential for generation of 23S RNA virus in vivo, consisting of the third and fourth C nt from the 3Ј-end and a mismatched pair of purines in the stem adjacent to the 3Ј-end (26). This signal is also required for formation of complexes between the 23S RNA genome and its RNA polymerase, p104 (19).  Hybridization with deoxyoligo probes and hydroxyl radical protection experiments located the 5Ј-site at Stem-loop 2. This stem loop structure also contains a cis signal for replication. Mutations in the structure analyzed in this work can be divided into three groups based on their effects on replication and formation of complexes. First, most of single mutations at the wobble positions did not impair replication or formation of complexes. Second, some of them, such as C84G, affected both activities moderately. Finally, those that caused extensive disturbance on the stem structure abolished replication and reduced complexes to a basal level. Here we again observed a good correlation between replication and formation of complexes. It suggests that formation of stable complexes is a prerequisite for efficient replication. A wide area of the stem structure was protected from hydroxyl radical in the complex (Fig.  5). It suggests that p91 interacts with the structure along the long stem. It is conceivable that single mutations in the second group partially disturbed the 5Ј cis signal. This may also explain why an extensive disturbance in the stem structure was necessary to abolish replication completely. Fine mapping of the 5Ј cis signal, however, awaits the development of a system in which p91 and modified RNA can be expressed separately.
Secondary structure analysis in silica predicts that the 5Ј-end, central, and 3Ј-end regions are brought together into proximity by intramolecular long distance interactions (Fig.  4A). This raises the possibility that anti-p91 antibodies pulled down some fragments indirectly through RNA-RNA interactions. We consider it unlikely based on the following evidence. Protection experiments from hydroxyl radicals suggest that p91 physically interacts with Stem-loop 2 in the 5Ј-region. Furthermore, mutations at the 5Ј and 3Ј cis sites reduced formation of complexes to a basal level. This strongly suggests that p91 interacts with the RNA at these two sites. Finally, a double mutant at these two sites still keeps the same basal level of complexes (Fig.  7). This suggests that the interactions at these two sites are coordinated and that the binding of p91 to the central site is responsible for the basal activity. It may be envisaged that in the complex 20S RNA folds into a highly organized structure by intramolecular interactions (Fig. 4A), and that p91 may help to stabilize the structure by holding the RNA at the three sites. It is not known whether a host protein(s) other than p91 is also involved in formation of complexes. Such protein could mediate the interaction of p91 with 20S RNA.
As shown previously, the 3Ј-terminal and penultimate C nt were dispensable for 20S RNA virus generation, while the third and fourth C nt were essential (18). This is consistent with the idea that p91 interacts with the third and fourth C nt from the 3Ј-end in the complex and protects it from the exosome. In accordance with this, formation of complexes attenuated the turnover of 20S RNA transcripts. In 23S RNA virus we observed a similar but more profound stabilization of 23S RNA transcripts by formation of complexes with p104 (19). Moreover, the turnover of 20S RNA transcripts is not affected by complex formation any longer in SKI2-or SKI7-deleted strains. 3 This suggests that the viral RNA is vulnerable to the exosome in the absence of complex formation. The 3Ј-terminal and penultimate positions may be accessible to the exosome. Thus mutations introduced to these positions could be excised and then repaired to the wild-type sequence by an unknown mechanism. The secondary structures found at the 3Ј-ends in both 20S and 23S RNAs resemble the so-called top half domain of tRNA (18,26). Therefore, the enzymes involved in tRNA 3Ј-end turnover, such as the CCA-adding enzyme, could be responsible for this repair. Because antigenomic strands are intermediates of replication, they also need to be protected in the cytoplasm. These molecules can be immunoprecipitated by anti-p91 antiserum (27). They posses the same 5-nt inverted repeats at both termini and the third and fourth C nt from the 3Ј-end are essential for replication (18). Thus it is likely that the antigenomic strands are also protected at the 3Ј-ends by formation of complexes.
In contrast to the 3Ј-end, binding of p91 at the 5Ј-site occurs not at the very end of the molecule, but to the second stem loop structure from the 5Ј-end. The 20S RNA genome has a cluster of four G nt at the 5Ј-end. Furthermore, these G nt are buried at the bottom of a long stem structure (Fig. 5B). It is known that oligo G tracts and strong secondary structures inhibit the progression of the 5Ј XRN1/SKI1 exonuclease (28,29). Therefore, the 20S RNA genome itself appears to be relatively resistant to the enzyme. Consistently, a ski1⌬ mutation does not affect significantly the copy number of 20S RNA virus while it considerably increases the copy number of L-A totivirus (5-10-fold), whose plus strand lacks a strong secondary structure and has an A-U-rich sequence at its 5Ј-terminus. Furthermore, destabilization of the 5Ј-end stem structure makes 20S RNA virus vulnerable to the XRN1/SKI1 exonuuclease. 4 In this context, it is noteworthy to mention that the initiation codon of p91 is located in the middle of the first long stem (Fig. 5B). If p91 bound to the first stem loop structure, then such tight binding might interfere with translation of p91. We observed a weak protection from hydroxyl radical cleavages in the complex at positions 61 and 62, just opposite to the initiation codon in the stem. This raises the possibility that p91 bound to the second stem structure can, by taking advantage of its proximity, influence translational events from the initiation codon. It is tempting to speculate that the two driving forces; protection of the 5Ј-end and translation of p91, might have shaped the secondary structure at the 5Ј-end in 20S RNA virus.
20S RNA virus has no extracellular transmission pathway. It infects yeast cells persistently without killing them. The formation of ribonucleoprotein complexes with its RNA polymerase represents a prominent feature in the exclusively intracellular form of this virus; this structure is not only essential for replication but perhaps it also reflects the necessity of this virus to dodge or cope with host exonucleases surveillance. Although an infectious virus can protect its RNA genome by forming virions in the extracellular environment, the virus still needs to replicate inside of the cell, and thus has to deal with exonucleases. In poliovirus, a cloverleaf structure at the 5Ј-terminus of the genomic RNA forms a ternary ribonucleoprotein complex with the poliovirus 3CD protein and a cellular protein, poly(rC)-binding protein. It has been proposed that the viral genome is circularized by a protein bridge between the ternary complex and poly(A)-binding protein associated with the poly(A) tail at the 3Ј-end of the genome (30, 31); a circularization model analogous to the one proposed for stabilization of mRNA (32). Influenza virus has a segmented negative strand RNA genome, and each RNA segment is encapsidated into a ribonucleoprotein particle by the nucleoprotein NP. These particles serve as the template for transcription and replication. The polymerase of influenza virus is a heterotrimer and binds to both ends of viral RNA in the particle (33,34). It is conceivable that the association of the polymerase may stabilize the RNA and protect it from exonucleases. Because most RNA viruses have linear RNA genomes, a mechanism(s) to maintain the integrity of both ends is vital for their survival. This could be a potential target to develop a new type of antiviral drugs against RNA virus infection.