The Bipartite 3′-cis-Acting Signal for Replication Is Required for Formation of a Ribonucleoprotein Complex in Vivo between the Viral Genome and Its RNA Polymerase in Yeast 23 S RNA Virus*

23 S RNA narnavirus is a persistent positive strand RNA virus found in Saccharomyces cerevisiae. The viral genome (2.9 kb) encodes only its RNA-dependent RNA polymerase, p104, and forms a ribonucleoprotein complex with p104 in vivo. Previously we succeeded in generating 23 S RNA virus in yeast from an expression vector containing the entire viral cDNA sequence. Using this system, we have recently identified a bipartite 3′ cis-acting signal for replication. The signal consists of a stretch of four cytidines (Cs) at the 3′ end and a mismatched pair of purines in a stem-loop structure that partially overlaps the terminal four Cs. Although the 3′ terminal and penultimate Cs are not essential for virus launching, the generated viruses efficiently recovered these terminal nucleotides. In this work, we expressed RNA transcripts containing the entire 23 S RNA genome but incapable of generating the virus because of the presence of non-viral extra sequences at the 3′ ends. These transcripts could form complexes with p104 in vivo, and a detailed analysis indicated that the mismatched pair of purines as well as the third and fourth Cs from the viral 3′ end was essential for this complex-forming activity. Given that 23 S RNA virus does not have genes for capsid proteins, the binding of p104 to the viral 3′ end, in addition to the efficient 3′ terminal repair, may play a crucial role in virus persistence by protecting and maintaining the correct viral 3′ end in vivo.

A wide variety of viruses can establish persistent infections in the host. For a long-term infection, the viruses must evade host immune responses and limit their cytolytic effects but also need to replicate their genomes to be maintained stably in the cells. It is well known that retroviral DNA is integrated into the host genome after its synthesis by the reverse transcriptase. Viral DNA can also be maintained as an extrachromosomal element (1)(2)(3), and its synthesis is linked to the replication of the host chromosome. Although persistent infection also occurs in RNA viruses, such as the clinically important hepatitis A and C viruses, the basis for their persistence, especially how they replicate and maintain their genomes within the host cells, remains to be elucidated. 20 S and 23 S RNA narnaviruses are persistent positive strand RNA viruses found in the yeast Saccharomyces cerevisiae (4). These viruses were discovered originally as RNA species induced under nitrogen starvation conditions (5,6). As is typical of fungal viruses, they do not kill the host or render phenotypic changes to the host. The viruses share many features. Their RNA genomes are small (2514 and 2891 nt 1 in 20 S and 23 S RNAs, respectively) and each genome encodes a single protein, a 91-kDa protein (p91) by 20 S RNA and a 104 kDa protein (p104) by 23 S RNA (7)(8)(9)(10). Both proteins contain four amino acid motifs well conserved among RNA-dependent RNA polymerases (RdRps) and these motifs are related most closely to those of RdRps of RNA coliphages (11). The doublestranded forms of 20 S and 23 S RNAs are called W and T, respectively (10,12). Because 20 S and 23 S RNA viruses do not encode capsid proteins, their RNA genomes are not encapsidated into viral particles (13)(14)(15). Instead, these RNAs form ribonucleoprotein complexes with their cognate RdRps in a 1:1 stoichiometry and reside in the host cytoplasm (14). 20 S and 23 S RNA genomes lack 3Ј-poly(A) tails and have perhaps no cap structures at the 5Ј ends, thus resembling intermediates of mRNA decay. Therefore, one of the interesting questions concerning these viruses is how they can reside and persist in the host cytoplasm without their genomes being degraded by the exonucleases involved in mRNA degradation pathways.
We succeeded recently in generating 23 S RNA virus in yeast from an expression vector containing the entire 23 S RNA cDNA sequence (15). The hepatitis ␦ virus antigenomic ribozyme was directly fused to the 3Ј terminus of the viral genome. An active ribozyme as well as an active p104 were essential for virus generation. Using this launching system, we have identified a bipartite cis-acting signal for replication in the 3Ј non-coding region of the 23 S RNA genome (16) (Fig. 1C). The signal consists of a cluster of 4 Cs at the 3Ј terminus and a mismatched pair of purines in a stem-loop structure that partially overlaps this cluster of 4 Cs. Any combination of purines at the mismatched pair enabled the RNA to generate 23 S RNA virus; however, eliminating the mismatched pair or substituting the purines by pyrimidines abolished the activity. Although 20 S and 23 S RNAs share the same 5-nt inverted repeats at the 5Ј and 3Ј termini (GGGGC . . . GCCCC-OH), the innermost G at the 3Ј terminus was not essential for 23 S RNA replication. The terminal and penultimate Cs at the 3Ј end were not essential for virus launching; the generated viruses, however, restored these terminal Cs. This indicates that the four Cs at the 3Ј end are essential for virus replication and that 23 S RNA virus has an effective 3Ј terminal repair mechanism(s) in vivo.
In this work, we analyzed the formation of 23 S RNA/p104 ribonucleoprotein complexes in vivo and found that the 3Ј bipartite cis-signal for replication (the mismatched pair of purines as well as the third and fourth Cs at the 3Ј end) is essential for the binding of p104, thus suggesting physical interactions between them. The formation of complexes, along with the efficient 3Ј end repair in vivo, may play an essential role in protecting and maintaining the proper viral 3Ј end for 23 S RNA virus persistence.
Plasmids-All of the 23 S RNA expression plasmids used in this work were derivatives of the 23 S RNA launching plasmids used in the previous study (16). The original launching plasmids were modified either by deleting the antigenomic hepatitis delta virus ribozyme sequence fused to the 3Ј end of the viral sequence, or by substituting the GGG sequence at the ribozyme cleavage site with AAA by site-directed in vitro mutagenesis (19). Because the generation of the viral 3Ј end by ribozyme cleavage was critical to launch 23 S RNA virus, these derivatives were unable to generate 23 S RNA virus in vivo. All of modifications were confirmed by DNA-sequencing.
Immunoprecipitation-Yeast transformants were grown in H-Trp medium and harvested at the logarithmic growing phase. These cells had been grown ϳ30 generations after receiving a 23 S RNA expression plasmid to obtain enough quantity. Cells (3 ϫ 10 8 ) were washed with lysis buffer (20 mM Tris, pH 8.0, 100 mM NaCl), re-suspended in the same buffer (100 l) and broken with glass beads by vortex mixing (15 s, 10 times). After adding 350 l of fresh buffer, the cell suspension was centrifuged at 13,000 rpm for 1 min to remove cell debris and unbroken cells. The lysate was then used 1) to analyze p104 by Western blotting, 2) to extract total 23 S RNA transcripts with phenol for Northern blot analysis, and 3) to immunoprecipitate 23 S RNA transcripts with anti-p104 antibodies. Immunoprecipitation of 23 S RNA transcripts was carried out as follows. To 150 l of the lysate prepared as described above, 1 ml of Tris-buffered saline-Tween 20 (10 mM Tris, pH 8.0, 150 mM NaCl, and 0.05% Tween 20), 1 mM dithiothreitol, 40 units of RNasin (Promega), 2 g of yeast tRNA (Invitrogen) and 2 l of anti-p104 antiserum or its preimmune serum were added and incubated at 4°C for 30 min. 25 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 three times with 1 ml of Tris-buffered saline-Tween 20 and 1 mM dithiothreitol and then suspended in 150 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 12 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 150 l of 10ϫ SSC and blotted onto a Nytran N membrane (Schleicher and Schü ll) using a Bio-Dot SF apparatus (Bio-Rad). RNA on the membrane was detected by Northern hybridization as described previously (20). Treatment of lysates with bentonite was done as follows: to 50 l of the cell lysate described above, 2.5 l of bentonite (3% w/v) was added. After a brief incubation at 4°C, the lysate was centrifuged at 13,000 rpm for 1 min to separate bentonite. p104 in the supernatant was analyzed by Western blotting.
Northern Probes-A 23 S RNA positive strand-specific probe was made from SmaI-digested pALI-38 by run-off transcription with T3 RNA polymerase. The probe contained the entire nucleotide sequence of the 23 S RNA negative strand without extra non-viral sequences at both ends (14) and was used throughout this work unless described otherwise. pRE491, pRE479, and pRE443 contained 5Ј XhoI (1-223), BamHI-XhoI (1257-1600), and 3Ј SpeI (2750 -2891) fragments of the 23 S RNA cDNA sequence, respectively, in the Bluescript-SK ϩ or -KS ϩ vector (Stratagene), and were used to make the 5Ј end, middle, and 3Ј endspecific probes for the 23 S RNA genome. A 0.35 kb EcoRI-SpeI fragment from pI2 vector (17) was subcloned into the KS ϩ vector (pRE636). This plasmid was used to make a probe to detect the vector sequence downstream of the 23 S RNA genome present in the transcripts immunoprecipitated with anti-p104 antiserum.
Antiserum-Anti-p104 antiserum used in this work was described previously (21). It was raised against an N-terminal fragment (amino acids 11 to 265) of p104.

S RNA Expression
Plasmids-23 S RNA virus exists in vivo in the form of a ribonucleoprotein complex between the 23 S RNA genome and its RdRp p104. To analyze cis-acting signals for the formation of these complexes we expressed 23 S RNA transcripts from a vector bearing various mutations at the 3Ј non-coding region in the viral genome and examined their ability to form complexes in vivo with p104 by an immunoprecipitation assay using anti-p104 antibodies. To avoid any complications derived from 23 S RNA virus generation, we used launching plasmids constructed previously (16) but modified them as follows (Fig. 1A). The standard launching plasmid (pRE637) contains the complete 23 S RNA cDNA sequence (2891 base pairs) downstream of the PGK1 promoter. Transcripts from the promoter have the positive polarity of the viral genome. The hepatitis ␦ virus antigenomic ribozyme is directly fused to the 3Ј terminus of the viral genome. The transcription termination site for the FLP gene (23) of the 2 M plasmid is located 0.7 kb downstream of the 23 S RNA genome. Because the ribozyme sequence is flanked by two unique restriction sites (SmaI and EcoRI), we digested the plasmid with these enzymes to eliminate the ribozyme. The larger fragment containing the 23 S RNA sequence was Klenow-treated and ligated to produce plasmid pTF662 (Fig. 1A). We have shown previously that an active ribozyme fused to the viral 3Ј end was critical for virus generation. Consistently, transcripts from pTF662 failed to generate 23 S RNA virus in vivo even after 100 generations, as judged by the absence of T double-stranded RNA, an indicative of negative strand synthesis, as well as the absence of 23 S RNA after curing the plasmid (data not shown). All the 23 S RNA expression plasmids used in this work except for those shown in Fig. 9 were constructed in this way; they are thus incapable of generating 23 S RNA virus and contained the same flanking sequence to the viral 3Ј end.
p104 Can Bind to 23 S RNA with Extra Non-viral Sequence at the 3Ј End-Yeast cells having no endogenous 23 S RNA virus were transformed with pTF662 or the standard launching plasmid (pRE637). Lysates prepared from growing transformants contained p104 when analyzed by Western blots using anti-p104 antiserum ( Fig. 2A). The amount of p104 expressed from the standard launching plasmid was severalfold higher than the one from pTF662. This reflects that a significant proportion of cells (20 -50%) transformed with the launching plasmid had already generated 23 S RNA virus (15). When p104 in the lysates was immunoprecipitated with anti-p104 antiserum, the precipitates from both transformants were found positive with a 23 S RNA-specific probe in Northern blots (Fig. 2B). Control experiments with the preimmune serum or without serum gave no signals. Because pTF662 does not generate 23 S RNA virus, the results suggest that transcripts from this plasmid can bind to p104 even though they contain extra non-viral sequences at the 3Ј ends. To confirm this, we analyzed the transcripts in the immunoprecipitates with a probe specific to the vector sequence downstream of the 23 S RNA genome and also with probes recognizing different parts of the 23 S RNA genome (Fig. 3B). As control, we prepared a lysate from 23 S RNA virus-generated cells from which the launching plasmid (pRE637) had been cured. The three probes specific to different parts of 23 S RNA genome hybridized with RNA extracted from both immunoprecipitates and the relative intensities of the signals between them are quite similar irrespective of whether the probe recognized the 5Ј, middle, or 3Ј portions of the 23 S RNA genome (Fig. 3A). These results suggest that the pTF662 transcripts immunoprecipitated with anti-p104 antiserum contained intact 23 S RNA viral genome. Consistently, the probe specific to the vector sequence flanking the viral 3Ј end recognized the transcripts from pTF662 in the immunoprecipitates but not the viral RNA in the control (Fig.  3A). Therefore, these results clearly indicate that p104 could form complexes with transcripts from pTF662 even though they contained extra non-viral sequences at the 3Ј ends.
The Stem Structure at the Viral 3Ј End Is Essential for the Formation of Complexes with p104 -The 23 S RNA genome contains a 59-nt untranslated region at the 3Ј end. This region can form three stem-loop structures (Fig. 4B). When the nucleotide sequences of each loop (I-III) were modified separately in the vector, none of the modifications significantly affected the formation of complexes between the modified transcripts and p104 (Fig. 4A), although the modification in the loop III sequence also changed the last two amino acids in p104. As shown previously (16), these modifications did not affect replication of 23 S RNA virus. In contrast, the change of the fourth C from the 3Ј end to a U, an essential nucleotide for replication, abolished the formation of complexes. This suggests an involvement of the bipartite cis signal for replication in the formation of 23 S RNA/p104 complexes. The importance of the stem I structure containing the mismatched pair of purines to form complexes is shown in Fig. 5. When the nucleotide sequence of one side of the lower stem was changed to that of the other side of the stem, or vice versa, thus destroying the lower stem structure, the modified transcripts no longer formed complexes with p104 (lanes 4 and 5). However, exchanging them simultaneously, thus re-establishing the lower stem structure but with nucleotide sequences different from the original ones, restored the activity to form complexes with p104 (lane 6). We obtained the same results concerning the upper stem structure (lanes 1-3). These results, therefore, indicate the importance of the lower and upper stem structures surrounding the mismatched pair of purines in the formation of 23 S RNA/p104 complexes.
p104 Expression-Because the formation of ribonucleoprotein complexes requires 23 S RNA transcripts as well as p104 decoded from them, we examined the expression of p104 in transformants analyzed in Figs. 4 and 5. As shown in Fig. 6, A and C, all of the transformants examined expressed p104. There is, however, variation in the amounts of p104 expressed among them. We noticed that cells expressing 23 S RNA transcripts capable of forming complexes contain higher amounts of p104 compared with cells producing transcripts defective in forming complexes. This may be a mere coincidence, and slight modifications at the 3Ј non-coding region might have affected the expression of p104 as observed. On the other hand, because we were detecting the steady state levels of p104, this can be explained in terms of a stabilizing effect on the protein caused by the formation of a ribonucleoprotein complex. Consistent with the latter explanation, we also noticed that the steady state levels of 23 S RNA transcripts were higher when they could form complexes with p104 ( Figs. 4 and 5).
In the course of experiments, we found that p104, depending shown. The ribozyme sequence was eliminated from pRE637 by digesting the plasmid with SmaI and EcoRI and the larger fragment was Klenow-treated and re-ligated, resulting in the 23 S RNA expression plasmid pTF662. Because of the lack of the ribozyme sequence, pTF662 cannot generate 23 S RNA virus in vivo. B, the bipartite 3Ј cis signal for replication consists of a stretch of 4 Cs at the 3Ј end (boxed) and a mismatch pair of purines (circled) at the stem adjacent to the 3Ј end.
on whether it was part of ribonucleoprotein complexes or not, interacted differently with bentonite. When lysates were incubated with bentonite and then centrifuged to remove it, the p104 that was able to form complexes with 23 S RNA transcripts remained in the supernatant, whereas the protein incapable of forming complexes with 23 S RNA transcripts was found to precipitate with bentonite during the centrifugation (Fig. 6, B and D). In both cases, 23 S RNA transcripts remained in the supernatant (not shown). The underlying mechanism of this phenomenon is unclear. However, because all p104 in the transformants analyzed in this work contained the same wildtype amino acid sequence (except for the loop III mutant with two amino acids modified at the C terminus), this has to be related with the status of p104 in the lysates. The bulky structure of 23 S RNA transcripts may prevent the bound protein from gaining access to bentonite, or the p104 bound to 23 S RNA transcripts may fold differently from the unbound protein.
We ran sucrose gradients to confirm that the immunoprecipitation-Northern blot assay employed in this work faithfully reflected the formation of ribonucleoprotein complexes (Fig. 7). As control, we prepared a lysate from plasmid-cured cells carrying 23 S RNA virus generated from the standard launching plasmid (pRE637). The viral RNA and p104 comigrated in the gradient and peaked at fraction 17 (Fig. 7A). 23 S RNA transcripts from pTF662 that had the intact 23 S RNA genome with the 3Ј non-viral sequence ran as fast as or slightly faster than the authentic 23 S RNA viral genome, and p104 also peaked at fraction 17 in the gradient (Fig. 7B). 23 S RNA transcripts with a single mutation at the fourth C from the viral 3Ј end (C4U) from pTF689 migrated in the gradient with a sedimentation profile indistinguishable from that of the wild-type transcripts. However, p104 expressed from this plasmid, although containing the wild-type amino acid sequence, failed to form complexes with the RNA transcripts and remained at the top of the gradient (Fig. 7C). Therefore, these results confirm that the immunoprecipitation-Northern blot assay used in this work properly reflects the status of the formation of ribonucleoprotein complexes in vivo.
The Mismatched Pair of Purines Is Essential for Formation of Complexes-The stem-exchange experiments (Fig. 5) indicate that the stem structure of the stem-loop I is important for the formation of complexes. Because the loop sequence was not involved in this complex formation (Fig. 4), these results suggest that the stem structure is necessary to form a platform to mount the mismatched pair of purines that is essential for the formation of 23 S RNA/p104 complexes. To test this possibility, we modified nucleotides at the mismatched pair and examined their effects on the formation of complexes (Fig. 8). When the 5Ј-A of the mismatched pair was eliminated (lane 1) or replaced with C (lane 3) or U (data not shown), the modified transcripts no longer formed complexes with p104. However, when this A was changed to G, the modified transcripts with the GϽ ϾG mismatch could form complexes with p104 with a slightly diminished activity (lane 2). Likewise, changing the 3Ј-G of the

3Ј-cis Signal for 23 S RNA/p104 Complex Formation
mismatched pair to C (lane 6) or eliminating this nucleotide (lane 4) abolished the complex-forming activity in the modified transcripts. Again, an A could substitute for this G but with a slightly reduced activity (lane 5). These results suggest that the bulged non-base-paired nucleotides at the stem are essential for the formation of complexes and that their bases should be purines. To confirm this, we introduced concerted mutations at the mismatched pair and analyzed their effects on the formation of complexes. Because changing the 5Ј-A at the mismatched pair to C (lane 3) or U resulted in the elimination of the mismatch from the stem by a C-G or U-G pairing, we changed the wild-type mismatched pair to the pair (CϽ ϾA) and found that transcripts bearing these mutations failed to form complexes with p104 (lane 7). This suggests that the 5Ј base of the mismatched pair should also be a purine. Consistently, transcripts bearing a AϽ ϾA (lane 5) or GϽ ϾA (lane 8) mismatched pair formed complexes with p104. Furthermore, transcripts with a UϽ ϾU mismatched pair failed to form complexes (data not shown). These results clearly indicate that the bases of the mismatched pair should be purines and that any combination of purines bestowed the activity of forming complexes with p104 on 23 S RNA transcripts. Therefore, 23 S RNA virus requires the same spectrum of nucleotide bases at the mismatched pair, not only for replication but also for the formation of ribonucleoprotein complexes with p104.
The Third and Fourth Cs from the Viral 3Ј End Are Essential for the Formation of Complexes-The 23 S RNA viral genome has 5-nt terminal inverted repeats (GGGGC . . . GCCCC-OH). As shown previously (16), the fifth G from the 3Ј end could be changed to C, along with the compensatory mutation C34G (numbered from the 3Ј end) at the other side of stem I, in the launching plasmid without losing virus-generating activity, and the generated viruses retained the modified nucleotides. Likewise, transcripts having these G5C and the compensatory C34G mutations retained a complex-forming activity with p104, although less compared with the control (Fig. 8, lane 9). As shown in Fig. 4B, when the fourth C from the 3Ј end was changed to U, the modified transcripts were unable to form complexes with p104. We chose a U to minimize the effects on the stem I structure. We also tried to analyze the last three Cs at the viral 3Ј end. Because these Cs are part of the unique SmaI site used to eliminate the ribozyme sequence from launching plasmids, we modified each of these terminal Cs in the standard launching plasmid by site-directed mutagenesis. To avoid virus generation, we also changed the GGG sequence 3Ј to the cleavage site to AAA. This modification destroys or modifies the substrate-bearing PI helix as well as a G-U wobble at the cleavage site in the ribozyme core structure (23)(24)(25) and, as shown previously, eliminated virus generating activity from the launching plasmid (15). The control plasmids with the complete 23 S RNA virus sequence, constructed either in this way or by deleting the ribozyme sequence, expressed comparable amounts of transcripts and both transcripts had similar activities to form complexes with p104. When the terminal C at the viral 3Ј end was changed to A, the modified transcripts showed a slightly reduced activity to form complexes (Fig. 9,  lane 1). Changing the penultimate C to A significantly reduced the activity (lane 2). Modifying the third C to A (lane 3) or the fourth C to A or U (lanes 4 and 5) completely eliminated the ability to form complexes from the transcripts. Therefore, these nucleotides at the third and fourth positions constitute an essential part of the 3Ј cis signal to form complexes with p104. These results together indicate that the bipartite 3Ј cis signal for replication determined by virus generation from a vector is also involved in the formation of ribonucleoprotein complexes in vivo with the viral RdRp, p104.  4). The lysates were incubated with anti-p104 anti-serum (Anti-p104) or its preimmune serum (Preimmune) and the 23 S RNA transcripts complexed with p104 were isolated and detected as described in the legend to Fig. 2B. RNA extracted from the lysates without immunoprecipitation was also blotted and analyzed in the same membrane (Total RNA). The effects of these mutations on virus launching observed in the previous work (16) are shown on the right. B, the 3Ј terminal region of 23 S RNA contains three stem-loop structures (I-III). The termination codon of p104 is boxed. Nucleotides are numbered from the 3Ј end and mutations introduced are indicated by filled circles. C, the modified nucleotides in the mutants and the corresponding wild-type sequences are shown.
ing or, vertically, to daughter cells (26). Because mating or hyphal fusion is part of the host life cycle, it is, perhaps, more economic for the viruses to rely on the host's ability to find partners to be transmitted rather than to have their own ex-tracellular infectious cycle. Therefore, two characteristics are evident among these viruses. First, they are persistent viruses. Because there is no extracellular phase that allows the virus to escape, they cannot kill the host. They have to replicate with- out hurting the cells and have to evade the intracellular surveillance of the host. Second, because there is no infectious cycle, the viruses do not need elaborate machinery for exit and reentry to a new host. This makes their structures and genomes much simpler than those of infectious counterparts found in other organisms. L-A double-stranded RNA virus and Ty retroelements have no envelopes, and their genomes only contain gag and pol genes (27,28). L-A particles resemble, structurally and functionally, the inner core of reoviruses in higher eukaryotes. 20 S and 23 S RNA narnaviruses even have no coat proteins, and their genomes only encode their RdRps. Their simpler genomes, or a small number of proteins encoded by them might have helped these viruses to establish persistent infections. It is well known that persistent infection, for viruses that are normally lytic, generally requires restriction of viral gene expression or the generation of deletion mutants lacking one or more components of the wild-type genome. Considering the availability of versatile genetics and well-developed molecular biology techniques in yeast; therefore, these viruses are ideal to study not only their replication mechanisms but also intracellular virus-host interactions involved in viral persistence.
Recently, we have established a launching system to generate 23 S RNA virus in vivo from a vector containing the entire viral cDNA. Using this system, we have begun reverse genetics to study the replication mechanism of 23 S RNA virus. We found that the 23 S RNA genome contains a bipartite cis signal for replication in the 3Ј non-coding region. The signal consists of a row of four Cs at the 3Ј end and a mismatched pair of purines in a stem-loop structure adjacent to the 3Ј end. Although the terminal and penultimate Cs at the 3Ј end are not essential for virus generation, the generated viruses recovered these terminal Cs. This suggests that 23 S RNA virus has an efficient 3Ј terminal repair mechanism(s). Because these viruses have no coat proteins, their genomes are not encapsidated into intracellular particles. Instead, 20 S and 23 S RNAs form ribonucleoprotein complexes with their cognate RdRps, p91 and p104, respectively, in a 1:1 stoichiometry. As shown in this work, RNA transcripts containing the entire 23 S RNA genome could form complexes with the polymerase p104, even though they had non-viral extra sequences at the 3Ј ends. Our results indicate that the bipartite cis-signal for replication is also essential for forming these complexes with p104.
The 23 S RNA genome contains three stem-loop structures FIG. 6. p104 expressed from 23 S RNA-expression plasmids. p104 in cell lysates (A and C) was analyzed by Western blotting using anti-p104 antiserum. B and D, the lysates used in A and C were first incubated with bentonite. After centrifugation to eliminate bentonite, protein in the supernatants was analyzed as described above. In A and B, lysates were prepared from cells transcribing 23 S RNA sequences bearing mutations in loops I-III as well as at the fourth position from the 3Ј end as described in Fig. 4. In C and D, the lysates were prepared from cells transcribing (I-III) at the 3Ј non-coding region, and none of these loop sequences is important for the formation of complexes. However, the stem I structure surrounding the mismatched pair of purines was essential for this complex-forming activity. This suggests the importance of the mismatched pair of purines for this activity. Consistently, eliminating one of the mismatched pair of purines or changing it to C or U abolished the activity to form complexes with p104. Any combination of purines at the mismatched pair bestowed the activity to form complexes on the RNA. Therefore, the spectrum of nucleotides at the mismatched pair necessary to form complexes with p104 is identical to the one required for 23 S RNA virus replication. Furthermore, the third and fourth Cs from the viral 3Ј end were also essential to form complexes with p104. The same bipartite 3Ј cis signal identified by a virus-launching assay is also required for the formation of complexes with p104 in vivo, which indicates the importance of complex forming in 23 S RNA virus replication.
There are two implications of this finding relevant to 23 S RNA virus replication. First, the formation of ribonucleoprotein complexes brings the viral RNA and its RdRp in close contact. This may be important because 23 S RNA virus resides and replicates in the host cytoplasm that is filled with a great variety of host RNAs. Therefore, the formation of complexes between them will not only ensure an efficient replication of the viral genome but also increase its fidelity by reducing synthesis of non-viral RNA. Furthermore, the involvement of the bipartite 3Ј cis signal in the formation of complexes indicates that p104 can bind only to replication-competent 23 S RNA genomes having intact viral 3Ј ends. Second, the formation of complexes may protect and stabilize the viral genome. 23 S RNA genome has no poly(A) tail at the 3Ј end and perhaps no 5Ј-cap structure, thus resembling degradation intermediates of host mRNAs. Shortening the poly(A) tail triggers decapping of mRNA and then the decapped mRNA is quickly degraded by the potent XRN1/SKI1 5Ј-3Ј exonuclease as well as by a 3Ј-5Ј exonuclease complex called exosome (29,30). The copy numbers of 20 S and 23 S RNA viruses are increased by host mutations in SKI2, SKI6, and SKI8 genes (31). 2 These gene products are reported to be a component of the exosome or modulators of the exosome (32)(33)(34). It has also been proposed that these gene products suppress the expression of RNAs lacking the poly(A) tails, a common feature seen in yeast viral RNAs (35)(36)(37). We do not know whether p104 interacts directly with the bipartite 3Ј cis signal to form complexes or a host factor(s) might also be involved in this recognition process. At any rate, because the third and fourth Cs from the viral 3Ј end are essential for the complex-forming activity, we expect that such protein/RNA interactions will protect the viral 3Ј end from 2 R. Esteban and T. Fujimura, unpublished results. We also changed the fifth G numbered from the 3Ј end of the 23 S RNA sequence to C with the compensatory mutation (C34G) at the 5Ј side of the stem-loop I (lane 9). Lane C, control with the wild-type mismatched pair. The effects of these mutations in launching plasmids to generate 23 S RNA virus observed in the previous work (16) are shown on the right.  1-4). In lane 5, the fourth C from the 3Ј end was changed to U. Lane C, control with the wild-type 23 S RNA sequence. Modified nucleotides are indicated by dots. To avoid virus launching, the 5Ј terminal 3 nucleotides (GGG) of the ribozyme adjacent to the cleavage site were also changed to AAA. Lysates were prepared from cells transformed with the plasmids and 23 S RNA transcripts immunoprecipitated with anti-p104 antiserum (Anti-p104) or its preimmune serum (Preimmune) and total RNA extracted from the lysates (Total RNA) were analyzed by Northern blot using a 23 S RNA-specific probe. The effects of these mutations to generate 23 S RNA virus with the active wild-type ribozyme sequence observed in previous work (16) are shown on the right. Open arrowheads indicate the boundaries between the 23 S RNA genome and the modified ribozyme. exonuclease cleavage. The steady-state levels of 23 S RNA transcripts unable to form complexes with p104 were consistently lower than those of transcripts able to bind to the protein.
It is interesting that a modification of the 3Ј terminal C to A slightly affected the complex-forming activity, and changing the penultimate C to A substantially reduced the activity. We observed previously that modification or elimination of any of these last two Cs was efficiently repaired to wild-type nucleotides during virus launching. It is tempting, therefore, to speculate that these alterations destabilize the complex so that the modified 3Ј end is now allowed to interact more easily with the repair machinery. Because the stem-loop structure at the 3Ј end of 23 S RNA that contains the bipartite cis signal resemble the top half domain of tRNA, the yeast CCA-adding enzyme (tRNA nucleotidyl transferase) may be involved in this repair process (16).
We do not know whether the bipartite 3Ј cis signal is sufficient for forming complexes with p104 or other cis sites are also required. If this 3Ј signal is important to protect the viral 3Ј end from degradation, then, the 5Ј end of 23 S RNA genome must also be protected from 5Ј-3Ј exonucleases by binding to p104 or by other means. In the case of the closely related 20 S RNA narnavirus, our data indicate that its RdRp, p91, interacts with the 20 S RNA genome at both ends and that the 5Ј site is located within the first 150 bases. 3 The immunoprecipitation assay of complexes assembled in vivo described in this work can be applied to investigate such a hypothetical 5Ј-or internal cis site to bind to p104. Because the 23 S RNA genome has only 6 nt in the 5Ј untranslated region, the wild-type p104 protein and the modified 23 S RNA transcripts to be examined need to be expressed separately from two vectors.
Infectious RNA viruses also replicate inside the host cells. Their RNA genomes must be protected from degradation within the cells. In this context, we could draw some features parallel between 23 S RNA virus and the intracellular form of infectious RNA viruses. It has been proposed that poliovirus RNA forms a ribonucleoprotein complex in which the RNA termini are circularized by a protein-protein bridge consisting of the viral polymerase 3CD and host proteins (38,39). The formation of the complex is not only essential for replication but also implicated in the stability of the viral RNA. Replication of many eukaryotic positive strand RNA viruses, including poliovirus, takes place in intracellular membranous structures (40). In brome mosaic virus, it has been demonstrated that these structures protect the RNA genome from RNase treatment (41). In negative strand RNA viruses such as influenza virus, the templates for transcription and replication are ribonucleoproteins in which the negative strand RNA is complexed with the nucleoprotein NP. The polymerase of influenza virus is a heterotrimer and binds to both ends of the viral RNA covered with many copies of NP (42). The three-dimensional structure of the polymerase bound to both ends of viral RNA in the ribonucleoprotein complex has recently been determined (43). Because the polymerase visualized is an inactive resting form, it is likely that the binding of the polymerase protects the viral genome from exonuclease cleavages. Transcription and replication of double-stranded RNA viruses such as reovirus occurs in viral cores or core-like particles where the templates and polymerases are co-localized. It is known that L-A virions isolated from yeast cells can protect the packaged RNA genome from RNase treatment (20). L-A virions have a T ϭ 1 icosahedral capsid composed of 60 Gag asymmetric dimers, an architecture commonly seen in the inner cores of double-stranded RNA viruses (44,45). The virion has 60 holes of ϳ15 Å diameter that extend through the capsid wall. These holes seem to be large enough to allow ingress of nucleotides for RNA synthesis and egress of synthesized positive strand RNA yet small enough to exclude potentially harmful nucleases and proteases. In summary, these intracellular supramolecular structures, although diverse in architecture and composition among viruses, thus provide a place where the template RNA and polymerase are concentrated to achieve an efficient synthesis of viral RNA. These structures also provide a place where the replicating viral machinery is sequestered for protection from host exonucleases. As infectious viruses require virion structures to protect their genomes physically and chemically during the extracellular stage, RNA viruses in eukaryotes may also have developed these intracellular supramolecular structures to achieve dual tasks: to protect the transcription and replication machinery in the hostile intracellular environment and, at the same time, to fulfill viruses' reliance on the host metabolism.