Launching of the Yeast 20 S RNA Narnavirus by Expressing the Genomic or Antigenomic Viral RNA in Vivo*

20 S RNA virus is a persistent positive strand RNA virus found in Saccharomyces cerevisiae. The viral genome encodes only its RNA polymerase, p91, and resides in the cytoplasm in the form of a ribonucleoprotein complex with p91. We succeeded in generating 20 S RNA virus in vivo by expressing, from a vector, genomic strands fused at the 3′-ends to the hepatitis delta virus antigenomic ribozyme. Using this launching system, we analyzed 3′-cis-signals present in the genomic strand for replication. The viral genome has five-nucleotide inverted repeats at both termini (5′-GGGGC... GCCCC-OH). The fifth G from the 3′-end was dispensable for replication, whereas the third and fourth Cs were essential. The 3′-terminal and penultimate Cs could be eliminated or modified to other nucleotides; however, the generated viruses recovered these terminal Cs. Furthermore, extra nucleotides added at the viral 3′-end were eliminated in the launched viruses. Therefore, 20 S RNA virus has a mechanism(s) to maintain the correct size and sequence of the viral 3′-end. This may contribute to its persistent infection in yeast. We also succeeded in generating 20 S RNA virus similarly from antigenomic strands provided active p91 was supplied from a second vector in trans. Again, a cluster of four Cs at the 3′-end in the antigenomic strand was essential for replication. In this work, we also present the first conclusive evidence that 20 S and 23 S RNA viruses are independent replicons.

Positive strand RNA viruses encode RNA-dependent RNA polymerases in their genomes and utilize them to synthesize viral RNA in conjunction with other viral or host proteins (1). The antigenomic (negative) strand RNA is an intermediate of replication and serves as a template to synthesize progeny positive strands. Because viral replication takes place within the host cell and because there are plenty of cellular RNAs in these cells, the viruses must find efficiently not only the positive strands but also the negative strands during replication. Therefore, both the positive and negative strands bear cis-acting signals for replication, including those necessary to interact with the polymerase machinery. To analyze and characterize these signals, it is essential to develop appropriate in vivo or in vitro systems in which changes in the signals can be tested. 20 S and 23 S RNAs are positive strand RNA viruses found in Saccharomyces cerevisiae and belong to the genus Narnavirus (2). These viruses were initially identified as RNA species induced in cells under nitrogen starvation conditions (3,4). Most laboratory strains carry 20 S RNA, and fewer strains harbor 23 S RNA along with 20 S RNA. They are compatible in the same host cell. Because all known strains that carry 23 S RNA also carry 20 S RNA, it has not been clear whether 23 S RNA virus can replicate without 20 S RNA virus. Typical of fungal viruses, they are not infectious and have no extracellular pathways of transmission. They are transmitted horizontally through mating or vertically from mother to daughter cells. No curing methods to eliminate these viruses from yeast are known so far. Furthermore, 20 S and 23 S RNA viruses do not confer any phenotypic changes on the host. This makes their genetic manipulation difficult.
The 20 S and 23 S RNA genomes are small (2514 and 2891 nucleotides, respectively), and each RNA encodes only a single protein: a 91-kDa protein (p91) and a 104-kDa (p104), respectively (5)(6)(7)(8). Both proteins contain four amino acid motifs well conserved among RNAdependent RNA polymerases (9). When yeast cells are grown at 37°C, the cells accumulate double-stranded RNAs called W and T (10), the double-stranded forms of 20 S and 23 S RNA genomes, respectively (8). They are not replication intermediates, but by-products (11). Because these viruses do not have genes for capsid proteins, their genomes are not encapsidated into intracellular viral particles (12)(13)(14). Instead, the RNAs form ribonucleoprotein complexes with their cognate RNA-dependent RNA polymerases at a 1:1 stoichiometry and reside in the cytoplasm (15). The viral genomes lack poly(A) tails at the 3Ј-ends and have perhaps no 5Ј-cap structures (8), thus resembling degradation intermediates of mRNAs. How these viruses can survive as persistent viruses in the host cytoplasm without their RNA genomes being digested by exonucleases involved in mRNA degradation is therefore interesting.
Recently, we succeeded in generating 23 S RNA virus in vivo from a vector containing the entire cDNA sequence of the viral genome (16). Using this launching system, we began reverse genetics to investigate cis-acting signals for replication in 23 S RNA virus. 20 S and 23 S RNA genomes share five-nucleotide inverted repeats at the 5Ј and 3Ј termini (5Ј-GGGGC . . . GCCCC-OH) (8). The 23 S RNA genome contains a bipartite cis-signal in the 3Ј-region that consists of the cluster of the terminal four Cs and a mismatched pair of purines present in a stem structure adjacent to the 3Ј-end (17). Although the 3Ј-terminal and penultimate Cs are dispensable for launching 23 S RNA virus, the generated viruses recover these Cs. This indicates that the virus has an efficient 3Ј-terminal repair mechanism(s). Subsequently, we found that the bipartite 3Ј-cis-signal for replication is also essential for formation of ribonucleoprotein complexes in vivo with its RNA-dependent RNA polymerase, p104 (18). This indicates the importance of complex formation for 23 S RNA virus and suggests that p104 protects the viral 3Ј-ends from degradation by binding to the 3Ј-cis-signal.
In this work, we describe the generation of 20 S RNA virus in vivo from a vector, a system similar to the one developed previously for 23 S RNA virus (16). In contrast to the previous work, however, we suc-ceeded in producing 20 S RNA virus not only from positive strands but also from negative strands if, in the latter case, active p91 was provided in trans from a second vector. Using these systems, we modified the viral 3Ј-ends and found that the clusters of the 3Ј-terminal four Cs in both the positive and negative strands are 3Ј-cis-signals for replication. We also provide, for the first time, conclusive evidence that 23 S RNA virus does not require 20 S RNA virus for replication; thus, they are independent replicons.
Northern Hybridization-Cells were broken with glass beads (15), and RNA was extracted from cell lysates once with phenol and twice with phenol/chloroform and precipitated with ethanol. RNA was separated on an agarose gel, blotted onto a neutral nylon membrane (Hybond-N, Amersham Biosciences), and hybridized with a 32 P-labeled 20 S RNA or 23 S RNA positive or negative strand-specific probe (20). We analyzed 5-10 independent transformants in each experiment. For the sake of simplicity, only a representative of each experiment is presented in the figures.
Plasmids-The standard 20 S RNA virus-launching plasmid pRE740 contains the entire 20 S RNA cDNA sequence (2514 bp) downstream of the PGK1 promoter in a pI2 derivative (21). The hepatitis delta virus (HDV) 4 antigenomic ribozyme was directly fused to the 3Ј-end of the 20 S RNA cDNA genome (see Fig. 2A). There are 42 nucleotides between the major transcription start site from the PGK1 promoter and the 20 S RNA sequence. Transcripts from this vector have 20 S RNA positive strand polarity. The template plasmid pRE805 is similar to pRE740, but the orientation of the 20 S RNA cDNA is reversed. Therefore, transcripts from the PGK1 promoter have the negative strand polarity of the viral genome. In addition, the TRP1 marker of this plasmid was disrupted at the unique EcoRV site by insertion of a 1.1-kb URA3 fragment. 20 S RNA positive and negative strand-specific probes were made by T7 and T3 runoff transcription from pRE449 predigested with appropriate restriction enzymes. pRE449 contains a 1262-bp BamHI-SmaI fragment (nucleotides 1253-2514) of the 20 S RNA cDNA sequence between the BamHI and SamI sites of the pBluescript KS ϩ vector (Stratagene). 23 S RNA positive strand-specific probes were made from plasmid pALI38 as described (15). In vitro mutagenesis was performed as described (22). All mutations introduced were confirmed by DNA sequencing.
Reverse Transcription (RT)-PCR-To prove that 20 S RNA virus was generated from plasmids, a DNA fragment of 842 bp was amplified by RT-PCR from lysates using internal primers RE171 (5Ј-CGCTTCTGC-GATCGTAGATG-3Ј) and ALI-3 (5Ј-TAAAACTGTATGCAGCAG-3Ј). The lysates were prepared from 20 S RNA virus-generated cells from which the launching plasmids had been cured or from cells containing endogenous 20 S RNA virus. In either case, the lysate was pretreated with DNase I before RT-PCR. The RT-PCR products were then digested with SmaI and analyzed on agarose gels.
3Ј-Rapid Amplification of cDNA Ends (RACE)-Total RNA isolated by launching plasmid-cured cells was poly(A)-tailed using poly(A) polymerase (Invitrogen). The poly(A)-tailed RNA was denatured with hydroxymethylmercuric hydroxide as described (23), and cDNA was synthesized using oligonucleotide RE156 (5Ј-GACTCGAGTCGAG-GATCCTTTTTTTTTTTTTTTTT-3Ј) and SuperScript II RNase H Ϫ reverse transcriptase (Invitrogen). After cDNA synthesis, RNA was digested with RNase A, and the unincorporated primer was eliminated using a Sephadex G-50 mini spin column (Worthington). cDNA containing the 3Ј-ends of 20 S RNA positive strands was PCR-amplified for 30 cycles using Taq polymerase (Promega) and primers RE157 (5Ј-GACTCGAGTCGAGGATCC-3Ј) and PG6 (5Ј-CGAATCGTCGC-CAGTAG-3Ј). For amplification of the negative strand 3Ј-ends, we used oligonucleotide RE157 and oligonucleotide RE233 (5Ј-GCGTCGAAA-GACGACAGC-3Ј). The amplified products from positive strand 3Ј-ends were digested with HindIII and BamHI and ligated into the pBluescript KS ϩ vector. In the case of negative strand 3Ј-ends, PCR products were digested with SalI and BamHI and ligated into the pBluescript KS ϩ vector.

S and 23 S RNA Viruses
Are Independent Replicons-23 S RNA virus can be generated in vivo from an expression plasmid containing the entire viral cDNA sequence. We wished to establish a similar launching system for 20 S RNA virus to carry out reverse genetics. However, most laboratory strains harbor 20 S RNA virus. During the course of 23 S RNA virus-launching experiments, we happened to obtain cells that had lost endogenous 20 S RNA virus. Strain 2928 L-A-o was transformed with the standard 23 S RNA-launching plasmid pRE637 (16). When transformants were analyzed, all of them generated 23 S RNA virus from the plasmid, and one of them (transformant 1) had apparently lost its endogenous 20 S RNA virus (Fig. 1A). We eliminated the launching plasmid from transformant 1 by isolating single colonies on nonselective YPAD agar plates. When the plasmid-cured colonies were analyzed, we found two types of segregants; the majority of colonies (10 of 12) contained only 23 S RNA virus generated from the plasmid, and the rest harbored neither 20 S nor 23 S RNA virus. Both types of segregants retained the genetic markers of the original strain. The segregants carrying 23 S RNA virus maintained the virus stably and did not produce virus-free colonies anymore. This suggests that the loss of endogenous 20 S RNA virus preceded the generation of 23 S RNA virus from the launching plasmid.
We confirmed the curing of 20 S RNA virus in these segregants by three criteria. First, we did not detect 20 S RNA by Northern blotting (Fig. 1B). Even when the gels were overloaded, we found no signal for 20 S RNA. The conditions used should have allowed us to detect 20 S RNA if the virus were present at as low as one copy/cell. Second, we could not amplify 20 S RNA cDNA from the lysates of these cells by RT-PCR (data not shown). Finally, as will be shown below, we could generate 20 S RNA virus tagged with a silent mutation from a launching plasmid using these 20 S RNA-negative segregants as hosts. We confirmed that the virus retained the mutation and was not the result of amplification of any residual endogenous 20 S RNA virus. Generated 20 S RNA virus could be maintained stably and was also induced normally under nitrogen starvation conditions in these cells. Therefore, the 20 S RNA-negative segregants have no genetic defects to harbor 20 S RNA virus. We designated segregants 4 and 5 shown in Fig. 1B as 2928-4 and 2928-5, respectively. These segregants showed the same doubling time as the original strain, and we noticed no phenotypic changes in them. Strain 2928-5 carries only 23 S RNA virus derived from plasmid pRE637, but no 20 S RNA virus. Strain 2928-4 harbors none of the narnaviruses and can be used successfully as a host to generate 23 S RNA virus from the 23 S RNA-launching plasmid (17). Although it has been reported that some industrial yeast strains (Saccharomyces diastaticus) appear to harbor only 23 S RNA-like molecules (24), these results are the first conclusive evidence that 23 S RNA virus does not require 20 S RNA virus for replication. Therefore, 20 S and 23 S RNA viruses are independent replicons.
Launching of 20 S RNA Virus from Expression Plasmids-We constructed a launching plasmid (pRE740) to generate 20 S RNA virus in yeast by inserting the entire 20 S RNA cDNA sequence (2514 bp) downstream of the constitutive PGK1 promoter ( Fig. 2A). Positive strands of 20 S RNA can be transcribed from the promoter. To generate transcripts in vivo with the precise viral 3Ј-end sequence at the 3Ј termini, an 82-nucleotide HDV antigenomic ribozyme cDNA sequence (25) was directly fused to the 3Ј-end of the 20 S RNA sequence. The plasmid has the TRP1 gene as a selectable marker. We transformed strains 2928-4 and 2928-5 with pRE740. Transformants were transferred to 1% potassium acetate to induce 20 S RNA, and the RNA was extracted from the induced cells. By Northern hybridization, we detected single-stranded 20 S RNA positive strands in all of the transformants. The amounts of positive strands among them were similar (Fig. 2B, lanes 1-4). We also found small amounts of W, the double-stranded form of 20 S RNA, in all transformants. Furthermore, we detected single-stranded 20 S RNA negative strands with a negative strand-specific probe (data not shown).   4) were transformed with the standard launching plasmid pRE740. Strain 2928-4 was also transformed with launching plasmids in which the amino acid sequence of p91 was modified (lanes 5 and 6) as indicated in the diagrams. RNA was extracted from transformed cells (lanes 1-6) or untransformed 2928-4 cells (lane C (control)) and analyzed by Northern blotting using a 20 S RNA positive strand-specific probe. dsRNA, double-stranded RNA. C, proof that 20 S RNA virus was generated from the launching plasmid is shown. A unique SmaI site was created by a silent mutation (1476T3 G, numbered from the 5Ј-end of the positive strand) of the viral genome in the launching plasmid. Strains 2928-4 (lanes 1) and 2928-5 (lanes 2) were transformed with the modified plasmid (pRE749). After virus generation, the plasmid was cured from the cells. RNA was prepared from these cells or from cells harboring endogenous 20 S RNA virus (lanes C). Then, an 842-bp cDNA fragment encompassing the SmaI site was generated by RT-PCR. The products with (ϩ) and without (Ϫ) SmaI treatment were separated on an agarose gel and visualized by ethidium bromide staining. HindIII-digested markers and the positions of undigested or SmaItreated PCR products are indicated. The diagram shows endogenous and plasmid-generated 20 S RNAs and the expected sizes of their RT-PCR products. Arrows show the oligonucleotides primers used for amplification.
These results indicate that 20 S RNA negative strands are synthesized from the positive strand transcripts in vivo and suggest the generation of 20 S RNA virus from the plasmid. To confirm this, we took two experimental approaches. First, we cured the plasmid from the transformants by isolating single colonies in nonselective YPAD agar plates. More than 70% of the plasmid-cured colonies retained 20 S RNA virus. Once generated, the virus replicated autonomously and could be maintained stably in the cells for Ͼ100 generations (so far examined) in the absence of the plasmid. The generated virus could be induced under nitrogen starvation conditions, and the viral RNA could be seen directly by ethidium bromide staining on agarose gels. The amount of induced RNA and the positive/negative strand ratios were similar to those found in endogenous 20 S RNA virus. We did not notice any effect of 23 S RNA virus on the efficiency of 20 S RNA virus generation (Fig. 2B, compare lanes 1 and 2 with lanes 3 and 4). Second, we introduced a unique SmaI site into the 20 S RNA cDNA sequence in pRE740 by changing the T at position 1476 to G (numbered from the 20 S RNA 5Ј-end) (Fig. 2C). This marking did not alter the amino acid sequence of the encoded p91 protein.
We transformed strains 2928-4 and 2928-5 with the SmaI-tagged plasmid. Transformants from both strains generated 20 S RNA virus. After curing the plasmid, RNA was extracted from both strains, and an 842-bp 20 S RNA cDNA fragment encompassing the SmaI site was amplified by RT-PCR. We also amplified the 842-bp fragment from endogenous 20 S RNA virus present in the original 2928 L-A-o strain as a control. As shown in Fig. 2C (lower panel), the amplified cDNA fragment from generated 20 S RNA virus in strain 2928-4 (lane 1) or 2928-5 (lane 2) was completely digested with SmaI, whereas the control from endogenous 20 S RNA virus was fully resistant to the enzyme (lane C). Therefore, these results demonstrate that 20 S RNA virus was generated from the launching plasmid and also confirm that the host strains 2928-4 and 2928-5 do not carry any endogenous 20 S RNA virus.
p91 Is Essential-20 S RNA encodes a single protein, p91. p91 has the four amino acid motifs well conserved among RNA-dependent RNA polymerases. When one of the motifs ( 460 GDD 462 ) was changed to 460 EFD 462 in the launching plasmid, the modified vector failed to produce 20 S RNA virus (Fig. 2B, lane 5), indicating that active p91 is essential for replication. The presence of 23 S RNA virus in the host could not rescue the modified plasmid to generate 20 S RNA virus (data not shown). This indicates that p104 encoded by 23 S RNA cannot substitute p91 for replication of 20 S RNA virus and is consistent with the fact that 20 S and 23 S RNA viruses are independent replicons. p91 and p104 share an eight-amino acid sequence (R(V/I)CGDDLI) surrounding the GDD motif with only one mismatch. When the mismatched amino acid (Val 458 ) in p91 was replaced with Ile, thus making the stretch identical to the one found in p104, the plasmid with the modified p91 sequence produced 20 S RNA virus without any deleterious effects (Fig. 2B, lane 6). Previously, we found that 23 S RNA virus can also tolerate the reverse modification (Ile to Val) in the eight-amino acid sequence in p104 (16).
A Cluster of Four Cs at the 3Ј-End of the Positive Strand Is Essential for Replication-The transcription termination site for the FLP gene of the 2 plasmid is located 0.7 kb downstream of the 20 S RNA genome in the vector (26). When the ribozyme GGG sequence 3Ј to the cleavage site was substituted with AAA in the standard launching plasmid pRE740, the modified plasmid failed to generate 20 S RNA virus (data not shown). Therefore, it is important to generate transcripts in vivo with the precise viral 3Ј termini for virus launching. 20 S RNA has fivenucleotide inverted repeats at both termini (5Ј-GGGGC . . . GCCCC-OH). We examined the role of the 3Ј-terminal nucleotides of the 20 S RNA positive strand in replication by modifying each nucleotide with A in the vector. As shown in Fig. 3, changing the 3Ј-terminal or penultimate C did not affect the generation of 20 S RNA virus. On the contrary, a modification at the third or fourth C did not produce the virus. Therefore, the third and fourth Cs from the 3Ј-end are essential for replication. In good agreement, deletions of up to two Cs (but not three) generated 20 S RNA virus (Fig. 3) (data not shown). We were interested in the 3Ј-terminal sequences of 20 S RNA viruses generated from these modified vectors. We cured the launching plasmids from the cells, and the viral RNAs were isolated. The 3Ј-terminal sequences of these RNAs were then amplified by 3Ј-RACE, and 8 -10 independently isolated clones were sequenced. As shown in TABLE ONE, the substitution of the terminal (pRE757) or penultimate (pRE758) C with A at the viral 3Ј-end was corrected to the wild-type C in the generated viruses. Furthermore, viruses launched from 20 S RNA cDNA lacking two Cs at the 3Ј-end (pRE759) recovered these nucleotides. Therefore, these results indicate that 20 S RNA virus, like 23 S RNA virus, has an efficient 3Ј-end repair mechanism(s) in vivo. The major transcription start site in the launching plasmid is located at position Ϫ42, relative to the 5Ј terminus  (5). The plasmid was also changed by the addition of a single nucleotide (6 -8) or three nucleotides (9 and 10) at the boundary between the 3Ј-end of the positive strand and the ribozyme cleavage site (underlined). The expected nucleotide sequences of the transcripts after cleavage by the ribozyme are shown. C indicates the control without modification. C, cells (strain 2928-4) were transformed with the standard launching plasmid (lanes C (control)) or with plasmids modified as described for B (lanes 1-10). RNA was extracted from transformants, and 20 S RNA was detected by Northern blotting as described in the legend to Fig. 1A using a positive strand-specific probe. dsRNA, double-stranded RNA.
of the 20 S RNA genome. These extra nucleotides were not present in the generated viruses as judged from the 3Ј-sequences of the negative strands (TABLE ONE). We also added extra nucleotides between the 3Ј-end of 20 S RNA and the ribozyme sequence and examined their effects on virus launching. As shown in Fig. 3, the addition of a single U (lane 6), C (lane 7), or A (data not shown) did not affect the generation of 20 S RNA virus. However, the addition of a single G reduced virus launching (lane 8). Likewise, the addition of three Gs (lane 10), but not three Cs (lane 9), severely affected 20 S RNA virus generation. Extra Gs at the 3Ј-end reduced the efficiency of virus launching by severalfold as judged by counting virus-positive and virus-negative colonies after curing the plasmids. Once the plasmids were cured, however, the amount of 20 S RNA in the cells was similar to that generated from the wild-type cDNA without extra nucleotides at the 3Ј-end. We isolated viral RNA from plasmid-cured cells and analyzed the 3Ј-ends of the positive strands by 3Ј-RACE. As shown in TABLE ONE, the launched viruses examined eliminated the extra nucleotides present at the 3Ј-ends in the vectors. Therefore, these results clearly indicate that 20 S RNA virus has a mechanism(s) to correctly maintain the size and sequence of the 3Ј-end of the positive strand and is thus capable of repairing limited damage of the 3Ј-end.
The Fifth G from the 3Ј-End in the Positive Strand Is Dispensable for Replication-When the fifth G from the 3Ј-end was changed to C (5G3 C, numbered from the 3Ј-end) in the vector, the modified plasmid failed to generate 20 S RNA virus (Fig. 4, lane 1), thus suggesting that this nucleotide is essential for replication. Alternatively, because the fifth G is part of a structure adjacent to the 3Ј-end, the inability of this plasmid to generate the virus may be a secondary effect due to a perturbation of the stem structure caused by the modification. Because the corresponding nucleotide in the 23 S RNA genome 3Ј-end is dispensable for its replication (17), we were interested in this result and tried to distinguish these two possibilities. To restore the stem structure in the 5G3 C mutant, we introduced a second mutation, 35C3 G, at the other side of the stem (lane 3). Because this compensatory mutation resulted in the change of one amino acid (Thr 823 to Ser) at the C terminus of p91, we made two more constructs as controls. One construct contained the single mutation 36A3 U with the wild-type stem and wild-type . . . GCCCC-OH terminus and, as shown in lane 4, produced 20 S RNA virus. Thus, p91 can tolerate the amino acid change T823S for its activities. The second control contained the single mutation 35C3 G with the wild-type . . . GCCCC-OH terminus. This construct, as shown in lane 2, failed to generate 20 S RNA virus. Therefore, these control experiments suggest that the fifth nucleotide needs to be hydrogenbonded. Consistently, the compensatory mutation 35C3 G restored the ability to generate virus on the 5G3 C mutant (lane 3). This result clearly indicates that, as in the case of 23 S RNA virus, the fifth G in 20 S RNA virus is dispensable for replication. To rule out the possibility that this double mutant generated replicable revertants during launching, we isolated viral RNA after curing the plasmid and determined the 3Ј-terminal sequences by 3Ј-RACE. We sequenced 17 independently isolated clones. All of them retained the compensatory mutation 35C3 G at the 5Ј-side of the stem. Of these 17, 11 clones had the full 3Ј-terminal sequence ( . . . GCCCCC-OH), and the rest possessed the 3Ј termini truncated by one nucleotide ( . . . GCCCC_-OH; four clones) or by two nucleotides ( . . . GCCC__-OH; two clones). Therefore, these results show that no revertants were generated during virus launching. Furthermore, as the majority (11 of 17) retained the full 3Ј-terminal sequence, the sixth G from the 3Ј-end did not substitute for the modified nucleotide at position 5 in replication.
Launching of 20 S RNA Virus from Negative Strands-We also tried to generate 20 S RNA virus from negative strands. Plasmid pRE805 (the template plasmid) has the complete sequence of 20 S RNA downstream of the PGK1 promoter (Fig. 5A). The orientation of the cDNA was reversed so that transcripts from the promoter would have the negative strand polarity of the viral genome. The HDV antigenomic ribozyme was directly fused to the 3Ј-end of the viral negative strand. When virusfree strain 2928-4 was transformed with pRE805, we found no generation of 20 S RNA virus from the plasmid (Fig. 5C). This was expected because p91 cannot be translated from the negative strands transcribed from the PGK1 promoter. Plasmid pRE787 (the polymerase plasmid) is the standard launching plasmid, but with the mutation 4C3 A in the 3Ј-cis-signal in the positive strand. Although wild-type p91 could be translated from the transcripts, this plasmid could not generate 20 S RNA virus because of this mutation (Fig. 3C, lane 4). When yeast cells were transformed with both plasmids, the transformants could generate 20 S RNA virus (Fig. 5B). The efficiency of virus generation was high. After curing the plasmids, Ͼ70% of the plasmid-free colonies generated the virus, an efficiency similar to that of the standard launching plasmid   OCTOBER 7, 2005 • VOLUME 280 • NUMBER 40 FIGURE 5. Generation of 20 S RNA virus from negative strands in a two-vector system. A, the template plasmid (pRE805) has URA3 as a selective marker and contains the entire 20 S RNA cDNA sequence downstream of the PGK1 promoter. The orientation of 20 S RNA is reversed so that the negative strand of the viral genome can be transcribed from the promoter. The viral genome was tagged by a silent mutation (1476T3 G, numbered from the 5Ј-end of the positive strand) to create a unique SmaI site. The HDV antigenomic ribozyme (R) is directly fused to the 3Ј-end of the negative strand. The polymerase plasmid (pRE787) has a TRP1 selective marker and is a derivative of the standard launching plasmid pRE740 with a 4C3 A mutation at the 3Ј-end of the positive strand genome (indicated by the asterisk). Transcripts from this plasmid can be translated into wild-type p91 but could not generate 20 S RNA virus because of the 4C3 A mutation (see Fig. 3C, lane 4). The 20 S RNA genome in the polymerase plasmid has no SmaI site. B, cells were transformed with either the template plasmid (pRE805) or the polymerase plasmid (pRE787) alone or together. As a control, cells were also transformed with the standard launching plasmid pRE740. RNA was extracted from transformants, and 20 S RNA was detected by Northern blotting using a positive strand-specific probe as described in the legend to Fig. 1A. dsRNA, double-stranded RNA. C, proof that the virus was generated from negative strands is shown. RNA was prepared from 20 S RNA-generated cells from which both plasmids had been cured, and an 842-bp cDNA fragment encompassing the SmaI site was obtained by RT-PCR (lanes 2 and 4). As a control, we also amplified the cDNA fragment from RNA extracted from 2928 L-A-o cells carrying its endogenous 20 S RNA virus (lanes 1 and 3). The cDNA fragments, either undigested (lanes 1 and 2) or digested with SmaI (lanes 3 and 4), were separated on an agarose gel. Ethidium bromide staining of the gel is shown.  1-4). RNA was extracted from transformants, and 20 S RNA was detected by Northern blotting as described in the legend to Fig. 1A  pRE740 with positive strand transcripts. We interpreted this result as follows. The primary transcripts from the template plasmid were processed by ribozyme cleavage in cis to produce negative strand templates with the correct viral 3Ј-ends. p91 produced from the polymerase plasmid acted in trans to produce positive strand copies from the negative strand templates. A positive or negative strand viral RNA with the correct 5Ј-and 3Ј-ends was generated from the positive strand copies or negative strand templates by a mechanism(s) yet to be identified. Once generated, it was amplified by p91, thus creating a new virus. To confirm this interpretation, we did three experiments. First, when the ribozyme sequence in the template plasmid was modified (substitution of the GGG sequence 3Ј-end to the cleavage site with AAA), thus destroying its cleavage activity (16), the launching activity was abolished (data not shown). This indicates the importance of creating negative strands with the correct viral 3Ј-ends for virus launching in this two-vector system. Second, we introduced a mutation (EFD) of the GDD motif of p91 into the polymerase plasmid. The modified plasmid failed to generate 20 S RNA virus in this two-vector system (data not shown). This indicates that active p91 from the polymerase plasmid is essential for virus generation. Finally, we confirmed that the virus originated from the template plasmid (pRE805). The 20 S RNA genome in pRE805 was tagged with a unique SmaI site by a silent mutation (1476T3 G, numbered from the 5Ј-end of the positive strand), whereas the viral genome in the polymerase plasmid (pRE787) had no SmaI sites. From RNA isolated from the plasmid-cured cells, we amplified a cDNA fragment of 842 bp encompassing the SmaI site by RT-PCR. We also amplified the corresponding fragment from endogenous 20 S RNA virus as a control. When these fragments were digested with SmaI (Fig. 5C), the cDNA from the launched virus was digested completely, whereas the cDNA from the endogenous virus was totally resistant to the enzyme. Therefore, these results demonstrate that the virus was derived from the negative strands transcribed from the template plasmid (pRE805). Using this two-vector system, we examined the role of the 3Ј-terminal four Cs of the negative strand in viral replication. We introduced a single mutation (C to A) at each position in the row of four Cs in the template plasmid. As shown in Fig. 6, the mutation at the third or fourth C from the 3Ј-end abolished the generation of 20 S RNA virus, whereas changing the 3Ј-terminal or penultimate C to A did not affect the generation of 20 S RNA virus. Consistently, up to two Cs could be eliminated from the 3Ј-end in the vector without any effects on virus generation ( Fig. 6) (data not shown). These results indicate that the third and fourth Cs from the 3Ј-end of the negative strand are essential for 20 S RNA virus replication, whereas the 3Ј-terminal or penultimate C is dispensable for virus launching. We isolated viral RNAs from plasmidcured cells and determined their 3Ј-end sequences by 3Ј-RACE (TABLE  TWO). A mutation at the penultimate C (pLOR-2) or deletion of the 3Ј-terminal nucleotide (pLOR-4) introduced into the vector was corrected to the wild-type C in the generated viruses. Therefore, the cluster of four Cs at the 3Ј-end of the negative strand is again a 3Ј-cis-signal for replication (or part of it), and the virus also has a repair mechanism(s) to correctly maintain this 3Ј-end. Interestingly, we observed that negative strands cloned after 3Ј-RACE possessed an extra G at the 3Ј-ends at high frequencies (50% in some cases). Positive strands cloned from the same RNA preparations had an extra G at the 3Ј-ends but at much lower frequencies. At present, we do not know the biological significance of this observation.

DISCUSSION
In this work, we have described the generation of 20 S RNA virus from a vector by expressing the viral RNA in vivo. The RNA can be either the positive or negative strand of the 20 S RNA genome. In the latter case, active p91 needs to be provided in trans from a second vector. Once 20 S RNA virus is generated, however, the virus does not require these vectors any longer and can be propagated autonomously. Because 20 S RNA virus does not confer any phenotypic changes on the host, it has been impossible to perform comprehensive genetic analysis on this virus. Therefore, these launching systems have allowed us to carry out reverse genetics to determine and analyze cis-acting signals necessary for replication present not only in the positive strand but also in the negative strand of the viral genome.
Host Strains-During the course of experiments launching 23 S RNA virus from a vector, we happened to obtain 20 S RNA-cured cells. Interference in viral replication by defective interfering particles or by other viruses is well known among DNA and RNA viruses. In yeast, L-A and L-BC double-stranded RNA viruses can be eliminated from the cells by transcribing L-A and L-BC positive strand RNAs, respectively, from a vector (27,28). Although 20 S and 23 S RNA viruses are compatible in the same host and do not exclude each other, their copy numbers in logarithmically growing cells are extremely low compared with those under induction conditions (nitrogen starvation). Therefore, overexpression of 23 S RNA or its gene product (p104) from a vector in growing cells may have taken up a hypothetical limiting cellular factor shared by 20 S and 23 S RNA viruses necessary for replication or for stable transmission to daughter cells. We have not yet determined the frequency of exclusion, but it appears to be quite low.  1-4). A single C was also deleted (_) from the 3Ј-end (5). The diagrams show the expected 3Ј-end sequences of the transcripts generated after ribozyme cleavage. C indicates the wild-type control. C, cells containing the polymerase plasmid (lane C) were transformed with the template plasmids (lanes 1-5) modified as described for B. RNA was isolated from transformants, and 20 S RNA was detected by Northern blotting as described in the legend to Fig. 1A using a positive strand-specific probe. dsRNA, double-stranded RNA. OCTOBER 7, 2005 • VOLUME 280 • NUMBER 40 p91 Is Essential-When the GDD sequence in p91 was changed to EFD in the vector, the modified launching plasmid failed to generate 20 S RNA virus. Crystallographic studies have demonstrated that RNA-dependent RNA polymerases of RNA viruses, like DNA-dependent DNA or RNA polymerases or reverse transcriptases, form tertiary structures resembling a right hand (29 -31). The four amino acid motifs well conserved among RNA-dependent RNA polymerases, including the GDD motif, are located within a subdomain called the palm. The first aspartic acid residue of the GDD motif coordinates two Mg 2ϩ ions, which are essential for the catalytic activity in polymerization. Therefore, the failure of the EFD mutant to generate 20 S RNA virus is due to destruction of the catalytic site of RNA polymerization in p91. Short stretches of amino acids flanking the GDD motif form a ␤-hairpin structure with GDD in the hairpin loop. p91 and p104 polymerases share the same eight-amino acid sequence (R(V/I)CGDDLI) surrounding the GDD motif with one mismatch. When the mismatched Val in p91 was changed to Ile, thus making the eight-amino acid stretch identical to the one in p104, p91 tolerated this change, and 20 S RNA virus was generated from the modified vector. Previously, we found that 23 S RNA virus also tolerates the reverse change (Ile to Val) in the same stretch in p104 (16). Because these aliphatic amino acids are favored for ␤-sheet formation, these exchanges apparently do not significantly affect the overall structure of each polymerase. These results indicate the potential usefulness of the launching systems of 20 S and 23 S RNA viruses for structure-function analysis of these polymerases. Furthermore, as p91 and p104 also form ribonucleoprotein complexes with their cognate viral RNAs, these launching systems will be powerful tools for assigning and analyzing domains with distinct functions in these polymerases.

Launching of 20 S RNA Virus from Positive/Negative Strands
Positive Strand 3Ј-cis-Signal-In this work, we developed two launching systems to generate 20 S RNA virus in vivo: one from the genomic strands and the other from the antigenomic strands. The former system is similar to the one established previously to generate 23 S RNA virus (16). Using this launching system, we analyzed 3Ј-cis-signals for replication present in the 20 S RNA positive strand. 20 S and 23 S RNA viruses share the same five-nucleotide inverted repeats at both termini (5Ј-GGGGC . . . GCCCC-OH). Our results indicate that the third and fourth Cs from the 3Ј-end of the positive strand are essential for replication, whereas the terminal and penultimate Cs are dispensable, at least for virus launching. In the latter cases, however, we found that the viruses regained the wild-type four Cs. Thus, 20 S RNA virus, like 23 S RNA virus, has an efficient repair mechanism(s) for the viral 3Ј-ends, and the stretch of four Cs at the 3Ј-end of the 20 S RNA positive strand is essential for replication. In contrast to these terminal four Cs, the fifth G is not essential; however, it needs to be hydrogen-bonded with the nucleotide at the other side of the stem (Fig. 4). There are two probable explanations for this. 20 S and 23 S RNA viruses are closely related but independent replicons. Therefore, their replication machineries have to discriminate each other and recognize their own RNA template for replication. 23 S RNA virus has a bipartite 3Ј-cis-signal that consists of a stretch of four Cs at the 3Ј-end and a mismatched pair of purines in a stem-loop structure adjacent to the 3Ј-end (Fig. 7). In 23 S RNA virus, the fourth C from the 3Ј-end is hydrogen-bonded and located at the lower edge of the stem structure. Therefore, the replication machinery of 20 S RNA virus may recognize a stretch of singlestranded four Cs at the 3Ј-end flanked by a stem structure at the 5Ј-side without a spacer as a cis-signal for replication. Alternatively, 20 S RNA virus may also have a bipartite 3Ј-cis-signal with the second site yet to be located in the stem-loop structure adjacent to the 3Ј-end. Thus, elimination of the hydrogen bonding at the lower edge of the stem may alter  (43). Y, R, and N stand for pyrimidine, purine, and any base, respectively. the bipartite signal by shortening the distance between the two sites or changing their topological configuration. 23 S RNA virus did not tolerate a change of adding or eliminating 2 base pairs at the lower stem that separates both sites. 5 Launching of 20 S RNA Virus from Antigenomic Strands-In the second launching system, we used two vectors to separately express the 20 S RNA antigenomic strands and the polymerase. Either vector alone could not generate the virus. However, when these two plasmids were introduced together into the same host, the cells generated 20 S RNA virus. That the virus was not generated by recombination between these two plasmids or between their transcripts is illustrated by the following evidence. First, transcripts from these two vectors had opposite polarities. Therefore, the 20 S RNA wild-type genome cannot be produced by recombination between them. Second, the high efficiency of virus generation (as efficient as that from a single vector expressing positive strands) suggests that the virus was not generated through rare events such as RNA-RNA recombination. Third, when the essential GDD motif of p91 in the polymerase plasmid was changed to EFD, the modified plasmid failed to assist virus generation. This indicates the importance of an active p91 protein rather than its DNA or RNA sequence in the polymerase plasmid for virus generation. Fourth, when the ribozyme GGG sequence 3Ј to the cleavage site was changed to AAA in the template plasmid, thus destroying its cleavage activity, the modified template plasmid failed to generate 20 S RNA virus. This indicates that the negative strand transcripts need to be processed by the ribozyme to produce correct 3Ј-ends at the RNA termini. Finally, we found that the viruses retained the silent mutation (1476T3 G) tagged in the template plasmid. This indicates that their RNA genome derived from the negative strands transcribed from the template plasmid. This also eliminates the possibility that the 4C3 A mutation in the polymerase plasmid was rescued by recombination between the two plasmids. Because the GDD motif and the silent mutation were separated by only 85 nucleotides in the viral genome, such recombination should also efficiently rescue the EFD mutation in the polymerase plasmid.
Using this two-vector system, we found that the third and fourth Cs from the 3Ј-end of the negative strand are essential for replication. The 3Ј-terminal and penultimate Cs were dispensable for virus launching; however, the generated viruses regained the wild-type Cs at these positions. Therefore, the cluster of four consecutive Cs at the 3Ј-end of the negative strand is again a 3Ј-cis-signal for replication or part of it.
3Ј-End Repair and Viral Persistence-20 S RNA virus can repair mutations introduced at the terminal or penultimate position from the 3Ј-end or deletion of up to two nucleotides from the 3Ј-end in the positive or negative strand. Likewise, the addition of three extra Cs to the 3Ј-end of the positive strand did not affect virus launching, and the virus generated possessed the wild-type termini without these extra nucleotides. Therefore, 20 S RNA virus has a mechanism(s) to maintain not only the correct sequence but also the correct length of the viral 3Ј-ends. Interestingly, the addition of three extra Gs reduced the efficiency of virus generation, although the virus generated did not retain these extra Gs at the 3Ј-end. We observed previously that the addition of a row of eight Gs to the 5Ј-end of the 23 S RNA cDNA genome reduces the generation of 23 S RNA virus from a launching vector (16). However, the addition of eight Cs does not have such an effect. It is known that oligo G tracts inhibit the progression of 5Ј-and 3Ј-exonucleases in yeast (32,33). This is thus consistent with the idea that the 5Ј-and 3Ј-non-viral sequences in the transcripts after ribozyme cleavage were eliminated by host exonucleases. The launching vectors expressing 20 S RNA positive and negative strands contained non-viral sequences of 42 and 41 nucleotides, respectively, between the major transcription start site and the 5Ј-end of the viral genome. Although these extra sequences are almost 10 nucleotides longer than the one present in the 23 S RNAlaunching vector, we found that launching of 20 S RNA virus from these vectors was as efficient as (or better than) the generation of 23 S RNA virus observed previously. On the other hand, the addition of one or three extra Gs to the 3Ј-end greatly reduced the generation of 20 S RNA virus. These results correlate with the fact that the SKI1/XRN1 5Ј-exonuclease is more active than the 3Ј-exonuclease (exosome) in mRNA degradation in yeast (34).
A mutation introduced at the 3Ј-terminal or penultimate position could be mended by the polymerase machinery. RNA polymerases are known to incorporate mismatched nucleotides at high frequency. Alternatively, the machinery could synthesize oligonucleotides in a templateindependent manner and use these oligonucleotides as a primer to initiate complementary strand synthesis (35). Host enzymes might also be involved in the 3Ј-end repair. Previously, we pointed out the similarity between the 3Ј-end stem-loop structures present in the positive and negative strands of 23 S RNA virus and the top half-domain of tRNA (17). The acceptor stem and T stem of tRNA stack on each other coaxially and form a stem-loop structure with a single-stranded 3Ј-tail referred as the top half of tRNA (Fig. 7). The structure consists of a long stem of 12-13 bp with a seven-nucleotide T⌿C loop. The top halfdomain contains the determinants necessary for specific interactions with tRNA-related enzymes, including the CCA-adding enzyme (36,37). The positive strand of 20 S RNA virus also possesses a stem-loop structure with a tail of four single-stranded Cs at the 3Ј-end that resembles the top half-domain of tRNA. This molecule could be a substrate of the CCA-adding enzyme (Fig. 7). A mutation introduced at the 3Ј-end of the positive strand may be excised by an exonuclease. It is known that RNase T, a non-processive 3Ј-exonuclease, is responsible for the end turnover of tRNA in Escherichia coli. The progression of the cleavage by this enzyme is blocked by consecutive Cs (38). Then, the shortened 3Ј-end would be filled to the wild-type sequence by the CCA-adding enzyme. Consistent with this possibility is the evidence that 20 -30% of 20 S RNA positive and negative strands possessed a non-templated A at their 3Ј-ends when W double-stranded RNA was analyzed after pCp labeling (39). The negative strand of 20 S RNA has a similar secondary structure at the 3Ј-end but with a much shorter stem (8 bp). This raises the possibility that the negative strand is not a good substrate for the CCA-adding enzyme. In this work, we observed that the negative strands of 20 S RNA viruses generated from a vector possessed, at a high frequency, a non-templated G at the 3Ј-ends (TABLE TWO). Because endogenous 20 S RNA virus also possesses this G in the negative strands at high frequency, 6 it is not a phenomenon related to virus generation from a vector. In E. coli, it is known that poly(A) polymerase I and polynucleotide phosphorylase can partially compensate for the absence of the CCA-adding enzyme in the repair of the 3Ј-terminal sequence of tRNA (40,41). Therefore, the extra G at the 3Ј-end could be the manifestation that an enzyme other than the CCA-adding enzyme is involved in the repair of the 3Ј-end of the negative strand and that the virus somehow tolerated this G during its replication. Alternatively, because most of the negative strands in the lysates are present in replication complexes engaged in positive strand synthesis (11), this G may be involved in the replication process.
The positive strands of 20 S and 23 S RNA viruses form ribonucleoprotein complexes with their cognate RNA-dependent RNA poly-merases at a 1:1 stoichiometry and reside in the host cytoplasm. These RNAs have no poly(A) tails and perhaps no 5Ј-cap structures, thus resembling degradation intermediates of mRNAs. We expect them to be vulnerable to the exonucleases involved in mRNA turnover. This raises the possibility that the formation of ribonucleoprotein complexes is essential for virus persistence by protecting these RNA genomes in the host cytoplasm. The positive strand of 23 S RNA virus possesses the bipartite 3Ј-cis-acting signal for replication (Fig. 7). Previously, we found that the bipartite 3Ј-signal is also important for the formation of a complex between the 23 S RNA genome and its RNA-dependent RNA polymerase, p104 (18). The mismatched pair of purines and the third and fourth Cs from the 3Ј-end are essential for this activity. A mutation at the penultimate position slightly reduced the activity, whereas a modification at the 3Ј-terminal C did not affect the formation of the complex. These results suggest that p104 directly interacts with the third and fourth Cs from the 3Ј-end and protects the viral RNA from exonucleases. The terminal and penultimate nucleotides at the 3Ј-end may be accessible to the 3Ј-repair machinery. As demonstrated in this work, the stretch of four Cs at the 3Ј-end in the 20 S RNA positive strand is also a cis-acting signal for replication. Based on analogy to 23 S RNA virus, we expect that p91 interacts with the third and fourth Cs and protects the viral 3Ј-end from exonucleases. Because the negative strands are replication intermediates, they also need to be protected in the cytoplasm. The negative strands of 20 S RNA virus also form complexes with p91, as anti-p91 antiserum can immunoprecipitate the negative strands from cell lysates (42). It is likely that the stretch of four Cs at the 3Ј-end in the negative strand is also involved in the formation of complexes with p91.
The 20 S RNA genome contains only 12 nucleotides in the 3Ј-noncoding region (Fig. 3). This makes the comprehensive analysis of the 3Ј-cis-acting signals (for replication or for complex formation) quite difficult without modifying the amino acid sequence of p91. As shown in this work, however, p91 can act in trans to generate 20 S RNA virus from two vectors. Therefore, it will be possible to develop an in vivo system to analyze these cis-acting signals by separately expressing the viral RNA and the polymerase p91 from two vectors. These experiments will provide information essential to understanding the molecular basis of narnavirus replication and its persistence in yeast.